Linux Audio

Check our new training course

Loading...
v6.2
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * Budget Fair Queueing (BFQ) I/O scheduler.
   4 *
   5 * Based on ideas and code from CFQ:
   6 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
   7 *
   8 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
   9 *		      Paolo Valente <paolo.valente@unimore.it>
  10 *
  11 * Copyright (C) 2010 Paolo Valente <paolo.valente@unimore.it>
  12 *                    Arianna Avanzini <avanzini@google.com>
  13 *
  14 * Copyright (C) 2017 Paolo Valente <paolo.valente@linaro.org>
  15 *
  16 * BFQ is a proportional-share I/O scheduler, with some extra
  17 * low-latency capabilities. BFQ also supports full hierarchical
  18 * scheduling through cgroups. Next paragraphs provide an introduction
  19 * on BFQ inner workings. Details on BFQ benefits, usage and
  20 * limitations can be found in Documentation/block/bfq-iosched.rst.
  21 *
  22 * BFQ is a proportional-share storage-I/O scheduling algorithm based
  23 * on the slice-by-slice service scheme of CFQ. But BFQ assigns
  24 * budgets, measured in number of sectors, to processes instead of
  25 * time slices. The device is not granted to the in-service process
  26 * for a given time slice, but until it has exhausted its assigned
  27 * budget. This change from the time to the service domain enables BFQ
  28 * to distribute the device throughput among processes as desired,
  29 * without any distortion due to throughput fluctuations, or to device
  30 * internal queueing. BFQ uses an ad hoc internal scheduler, called
  31 * B-WF2Q+, to schedule processes according to their budgets. More
  32 * precisely, BFQ schedules queues associated with processes. Each
  33 * process/queue is assigned a user-configurable weight, and B-WF2Q+
  34 * guarantees that each queue receives a fraction of the throughput
  35 * proportional to its weight. Thanks to the accurate policy of
  36 * B-WF2Q+, BFQ can afford to assign high budgets to I/O-bound
  37 * processes issuing sequential requests (to boost the throughput),
  38 * and yet guarantee a low latency to interactive and soft real-time
  39 * applications.
  40 *
  41 * In particular, to provide these low-latency guarantees, BFQ
  42 * explicitly privileges the I/O of two classes of time-sensitive
  43 * applications: interactive and soft real-time. In more detail, BFQ
  44 * behaves this way if the low_latency parameter is set (default
  45 * configuration). This feature enables BFQ to provide applications in
  46 * these classes with a very low latency.
  47 *
  48 * To implement this feature, BFQ constantly tries to detect whether
  49 * the I/O requests in a bfq_queue come from an interactive or a soft
  50 * real-time application. For brevity, in these cases, the queue is
  51 * said to be interactive or soft real-time. In both cases, BFQ
  52 * privileges the service of the queue, over that of non-interactive
  53 * and non-soft-real-time queues. This privileging is performed,
  54 * mainly, by raising the weight of the queue. So, for brevity, we
  55 * call just weight-raising periods the time periods during which a
  56 * queue is privileged, because deemed interactive or soft real-time.
  57 *
  58 * The detection of soft real-time queues/applications is described in
  59 * detail in the comments on the function
  60 * bfq_bfqq_softrt_next_start. On the other hand, the detection of an
  61 * interactive queue works as follows: a queue is deemed interactive
  62 * if it is constantly non empty only for a limited time interval,
  63 * after which it does become empty. The queue may be deemed
  64 * interactive again (for a limited time), if it restarts being
  65 * constantly non empty, provided that this happens only after the
  66 * queue has remained empty for a given minimum idle time.
  67 *
  68 * By default, BFQ computes automatically the above maximum time
  69 * interval, i.e., the time interval after which a constantly
  70 * non-empty queue stops being deemed interactive. Since a queue is
  71 * weight-raised while it is deemed interactive, this maximum time
  72 * interval happens to coincide with the (maximum) duration of the
  73 * weight-raising for interactive queues.
  74 *
  75 * Finally, BFQ also features additional heuristics for
  76 * preserving both a low latency and a high throughput on NCQ-capable,
  77 * rotational or flash-based devices, and to get the job done quickly
  78 * for applications consisting in many I/O-bound processes.
  79 *
  80 * NOTE: if the main or only goal, with a given device, is to achieve
  81 * the maximum-possible throughput at all times, then do switch off
  82 * all low-latency heuristics for that device, by setting low_latency
  83 * to 0.
  84 *
  85 * BFQ is described in [1], where also a reference to the initial,
  86 * more theoretical paper on BFQ can be found. The interested reader
  87 * can find in the latter paper full details on the main algorithm, as
  88 * well as formulas of the guarantees and formal proofs of all the
  89 * properties.  With respect to the version of BFQ presented in these
  90 * papers, this implementation adds a few more heuristics, such as the
  91 * ones that guarantee a low latency to interactive and soft real-time
  92 * applications, and a hierarchical extension based on H-WF2Q+.
  93 *
  94 * B-WF2Q+ is based on WF2Q+, which is described in [2], together with
  95 * H-WF2Q+, while the augmented tree used here to implement B-WF2Q+
  96 * with O(log N) complexity derives from the one introduced with EEVDF
  97 * in [3].
  98 *
  99 * [1] P. Valente, A. Avanzini, "Evolution of the BFQ Storage I/O
 100 *     Scheduler", Proceedings of the First Workshop on Mobile System
 101 *     Technologies (MST-2015), May 2015.
 102 *     http://algogroup.unimore.it/people/paolo/disk_sched/mst-2015.pdf
 103 *
 104 * [2] Jon C.R. Bennett and H. Zhang, "Hierarchical Packet Fair Queueing
 105 *     Algorithms", IEEE/ACM Transactions on Networking, 5(5):675-689,
 106 *     Oct 1997.
 107 *
 108 * http://www.cs.cmu.edu/~hzhang/papers/TON-97-Oct.ps.gz
 109 *
 110 * [3] I. Stoica and H. Abdel-Wahab, "Earliest Eligible Virtual Deadline
 111 *     First: A Flexible and Accurate Mechanism for Proportional Share
 112 *     Resource Allocation", technical report.
 113 *
 114 * http://www.cs.berkeley.edu/~istoica/papers/eevdf-tr-95.pdf
 115 */
 116#include <linux/module.h>
 117#include <linux/slab.h>
 118#include <linux/blkdev.h>
 119#include <linux/cgroup.h>
 
 120#include <linux/ktime.h>
 121#include <linux/rbtree.h>
 122#include <linux/ioprio.h>
 123#include <linux/sbitmap.h>
 124#include <linux/delay.h>
 125#include <linux/backing-dev.h>
 126
 127#include <trace/events/block.h>
 128
 129#include "elevator.h"
 130#include "blk.h"
 131#include "blk-mq.h"
 132#include "blk-mq-tag.h"
 133#include "blk-mq-sched.h"
 134#include "bfq-iosched.h"
 135#include "blk-wbt.h"
 136
 137#define BFQ_BFQQ_FNS(name)						\
 138void bfq_mark_bfqq_##name(struct bfq_queue *bfqq)			\
 139{									\
 140	__set_bit(BFQQF_##name, &(bfqq)->flags);			\
 141}									\
 142void bfq_clear_bfqq_##name(struct bfq_queue *bfqq)			\
 143{									\
 144	__clear_bit(BFQQF_##name, &(bfqq)->flags);		\
 145}									\
 146int bfq_bfqq_##name(const struct bfq_queue *bfqq)			\
 147{									\
 148	return test_bit(BFQQF_##name, &(bfqq)->flags);		\
 149}
 150
 151BFQ_BFQQ_FNS(just_created);
 152BFQ_BFQQ_FNS(busy);
 153BFQ_BFQQ_FNS(wait_request);
 154BFQ_BFQQ_FNS(non_blocking_wait_rq);
 155BFQ_BFQQ_FNS(fifo_expire);
 156BFQ_BFQQ_FNS(has_short_ttime);
 157BFQ_BFQQ_FNS(sync);
 158BFQ_BFQQ_FNS(IO_bound);
 159BFQ_BFQQ_FNS(in_large_burst);
 160BFQ_BFQQ_FNS(coop);
 161BFQ_BFQQ_FNS(split_coop);
 162BFQ_BFQQ_FNS(softrt_update);
 
 163#undef BFQ_BFQQ_FNS						\
 164
 165/* Expiration time of async (0) and sync (1) requests, in ns. */
 166static const u64 bfq_fifo_expire[2] = { NSEC_PER_SEC / 4, NSEC_PER_SEC / 8 };
 167
 168/* Maximum backwards seek (magic number lifted from CFQ), in KiB. */
 169static const int bfq_back_max = 16 * 1024;
 170
 171/* Penalty of a backwards seek, in number of sectors. */
 172static const int bfq_back_penalty = 2;
 173
 174/* Idling period duration, in ns. */
 175static u64 bfq_slice_idle = NSEC_PER_SEC / 125;
 176
 177/* Minimum number of assigned budgets for which stats are safe to compute. */
 178static const int bfq_stats_min_budgets = 194;
 179
 180/* Default maximum budget values, in sectors and number of requests. */
 181static const int bfq_default_max_budget = 16 * 1024;
 182
 183/*
 184 * When a sync request is dispatched, the queue that contains that
 185 * request, and all the ancestor entities of that queue, are charged
 186 * with the number of sectors of the request. In contrast, if the
 187 * request is async, then the queue and its ancestor entities are
 188 * charged with the number of sectors of the request, multiplied by
 189 * the factor below. This throttles the bandwidth for async I/O,
 190 * w.r.t. to sync I/O, and it is done to counter the tendency of async
 191 * writes to steal I/O throughput to reads.
 192 *
 193 * The current value of this parameter is the result of a tuning with
 194 * several hardware and software configurations. We tried to find the
 195 * lowest value for which writes do not cause noticeable problems to
 196 * reads. In fact, the lower this parameter, the stabler I/O control,
 197 * in the following respect.  The lower this parameter is, the less
 198 * the bandwidth enjoyed by a group decreases
 199 * - when the group does writes, w.r.t. to when it does reads;
 200 * - when other groups do reads, w.r.t. to when they do writes.
 201 */
 202static const int bfq_async_charge_factor = 3;
 203
 204/* Default timeout values, in jiffies, approximating CFQ defaults. */
 205const int bfq_timeout = HZ / 8;
 206
 207/*
 208 * Time limit for merging (see comments in bfq_setup_cooperator). Set
 209 * to the slowest value that, in our tests, proved to be effective in
 210 * removing false positives, while not causing true positives to miss
 211 * queue merging.
 212 *
 213 * As can be deduced from the low time limit below, queue merging, if
 214 * successful, happens at the very beginning of the I/O of the involved
 215 * cooperating processes, as a consequence of the arrival of the very
 216 * first requests from each cooperator.  After that, there is very
 217 * little chance to find cooperators.
 218 */
 219static const unsigned long bfq_merge_time_limit = HZ/10;
 220
 221static struct kmem_cache *bfq_pool;
 222
 223/* Below this threshold (in ns), we consider thinktime immediate. */
 224#define BFQ_MIN_TT		(2 * NSEC_PER_MSEC)
 225
 226/* hw_tag detection: parallel requests threshold and min samples needed. */
 227#define BFQ_HW_QUEUE_THRESHOLD	3
 228#define BFQ_HW_QUEUE_SAMPLES	32
 229
 230#define BFQQ_SEEK_THR		(sector_t)(8 * 100)
 231#define BFQQ_SECT_THR_NONROT	(sector_t)(2 * 32)
 232#define BFQ_RQ_SEEKY(bfqd, last_pos, rq) \
 233	(get_sdist(last_pos, rq) >			\
 234	 BFQQ_SEEK_THR &&				\
 235	 (!blk_queue_nonrot(bfqd->queue) ||		\
 236	  blk_rq_sectors(rq) < BFQQ_SECT_THR_NONROT))
 237#define BFQQ_CLOSE_THR		(sector_t)(8 * 1024)
 238#define BFQQ_SEEKY(bfqq)	(hweight32(bfqq->seek_history) > 19)
 239/*
 240 * Sync random I/O is likely to be confused with soft real-time I/O,
 241 * because it is characterized by limited throughput and apparently
 242 * isochronous arrival pattern. To avoid false positives, queues
 243 * containing only random (seeky) I/O are prevented from being tagged
 244 * as soft real-time.
 245 */
 246#define BFQQ_TOTALLY_SEEKY(bfqq)	(bfqq->seek_history == -1)
 247
 248/* Min number of samples required to perform peak-rate update */
 249#define BFQ_RATE_MIN_SAMPLES	32
 250/* Min observation time interval required to perform a peak-rate update (ns) */
 251#define BFQ_RATE_MIN_INTERVAL	(300*NSEC_PER_MSEC)
 252/* Target observation time interval for a peak-rate update (ns) */
 253#define BFQ_RATE_REF_INTERVAL	NSEC_PER_SEC
 254
 255/*
 256 * Shift used for peak-rate fixed precision calculations.
 257 * With
 258 * - the current shift: 16 positions
 259 * - the current type used to store rate: u32
 260 * - the current unit of measure for rate: [sectors/usec], or, more precisely,
 261 *   [(sectors/usec) / 2^BFQ_RATE_SHIFT] to take into account the shift,
 262 * the range of rates that can be stored is
 263 * [1 / 2^BFQ_RATE_SHIFT, 2^(32 - BFQ_RATE_SHIFT)] sectors/usec =
 264 * [1 / 2^16, 2^16] sectors/usec = [15e-6, 65536] sectors/usec =
 265 * [15, 65G] sectors/sec
 266 * Which, assuming a sector size of 512B, corresponds to a range of
 267 * [7.5K, 33T] B/sec
 268 */
 269#define BFQ_RATE_SHIFT		16
 270
 271/*
 272 * When configured for computing the duration of the weight-raising
 273 * for interactive queues automatically (see the comments at the
 274 * beginning of this file), BFQ does it using the following formula:
 275 * duration = (ref_rate / r) * ref_wr_duration,
 276 * where r is the peak rate of the device, and ref_rate and
 277 * ref_wr_duration are two reference parameters.  In particular,
 278 * ref_rate is the peak rate of the reference storage device (see
 279 * below), and ref_wr_duration is about the maximum time needed, with
 280 * BFQ and while reading two files in parallel, to load typical large
 281 * applications on the reference device (see the comments on
 282 * max_service_from_wr below, for more details on how ref_wr_duration
 283 * is obtained).  In practice, the slower/faster the device at hand
 284 * is, the more/less it takes to load applications with respect to the
 285 * reference device.  Accordingly, the longer/shorter BFQ grants
 286 * weight raising to interactive applications.
 287 *
 288 * BFQ uses two different reference pairs (ref_rate, ref_wr_duration),
 289 * depending on whether the device is rotational or non-rotational.
 290 *
 291 * In the following definitions, ref_rate[0] and ref_wr_duration[0]
 292 * are the reference values for a rotational device, whereas
 293 * ref_rate[1] and ref_wr_duration[1] are the reference values for a
 294 * non-rotational device. The reference rates are not the actual peak
 295 * rates of the devices used as a reference, but slightly lower
 296 * values. The reason for using slightly lower values is that the
 297 * peak-rate estimator tends to yield slightly lower values than the
 298 * actual peak rate (it can yield the actual peak rate only if there
 299 * is only one process doing I/O, and the process does sequential
 300 * I/O).
 301 *
 302 * The reference peak rates are measured in sectors/usec, left-shifted
 303 * by BFQ_RATE_SHIFT.
 304 */
 305static int ref_rate[2] = {14000, 33000};
 306/*
 307 * To improve readability, a conversion function is used to initialize
 308 * the following array, which entails that the array can be
 309 * initialized only in a function.
 310 */
 311static int ref_wr_duration[2];
 312
 313/*
 314 * BFQ uses the above-detailed, time-based weight-raising mechanism to
 315 * privilege interactive tasks. This mechanism is vulnerable to the
 316 * following false positives: I/O-bound applications that will go on
 317 * doing I/O for much longer than the duration of weight
 318 * raising. These applications have basically no benefit from being
 319 * weight-raised at the beginning of their I/O. On the opposite end,
 320 * while being weight-raised, these applications
 321 * a) unjustly steal throughput to applications that may actually need
 322 * low latency;
 323 * b) make BFQ uselessly perform device idling; device idling results
 324 * in loss of device throughput with most flash-based storage, and may
 325 * increase latencies when used purposelessly.
 326 *
 327 * BFQ tries to reduce these problems, by adopting the following
 328 * countermeasure. To introduce this countermeasure, we need first to
 329 * finish explaining how the duration of weight-raising for
 330 * interactive tasks is computed.
 331 *
 332 * For a bfq_queue deemed as interactive, the duration of weight
 333 * raising is dynamically adjusted, as a function of the estimated
 334 * peak rate of the device, so as to be equal to the time needed to
 335 * execute the 'largest' interactive task we benchmarked so far. By
 336 * largest task, we mean the task for which each involved process has
 337 * to do more I/O than for any of the other tasks we benchmarked. This
 338 * reference interactive task is the start-up of LibreOffice Writer,
 339 * and in this task each process/bfq_queue needs to have at most ~110K
 340 * sectors transferred.
 341 *
 342 * This last piece of information enables BFQ to reduce the actual
 343 * duration of weight-raising for at least one class of I/O-bound
 344 * applications: those doing sequential or quasi-sequential I/O. An
 345 * example is file copy. In fact, once started, the main I/O-bound
 346 * processes of these applications usually consume the above 110K
 347 * sectors in much less time than the processes of an application that
 348 * is starting, because these I/O-bound processes will greedily devote
 349 * almost all their CPU cycles only to their target,
 350 * throughput-friendly I/O operations. This is even more true if BFQ
 351 * happens to be underestimating the device peak rate, and thus
 352 * overestimating the duration of weight raising. But, according to
 353 * our measurements, once transferred 110K sectors, these processes
 354 * have no right to be weight-raised any longer.
 355 *
 356 * Basing on the last consideration, BFQ ends weight-raising for a
 357 * bfq_queue if the latter happens to have received an amount of
 358 * service at least equal to the following constant. The constant is
 359 * set to slightly more than 110K, to have a minimum safety margin.
 360 *
 361 * This early ending of weight-raising reduces the amount of time
 362 * during which interactive false positives cause the two problems
 363 * described at the beginning of these comments.
 364 */
 365static const unsigned long max_service_from_wr = 120000;
 366
 367/*
 368 * Maximum time between the creation of two queues, for stable merge
 369 * to be activated (in ms)
 370 */
 371static const unsigned long bfq_activation_stable_merging = 600;
 372/*
 373 * Minimum time to be waited before evaluating delayed stable merge (in ms)
 374 */
 375static const unsigned long bfq_late_stable_merging = 600;
 376
 377#define RQ_BIC(rq)		((struct bfq_io_cq *)((rq)->elv.priv[0]))
 378#define RQ_BFQQ(rq)		((rq)->elv.priv[1])
 379
 380struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync)
 381{
 382	return bic->bfqq[is_sync];
 383}
 384
 385static void bfq_put_stable_ref(struct bfq_queue *bfqq);
 386
 387void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync)
 388{
 389	struct bfq_queue *old_bfqq = bic->bfqq[is_sync];
 390
 391	/* Clear bic pointer if bfqq is detached from this bic */
 392	if (old_bfqq && old_bfqq->bic == bic)
 393		old_bfqq->bic = NULL;
 394
 395	/*
 396	 * If bfqq != NULL, then a non-stable queue merge between
 397	 * bic->bfqq and bfqq is happening here. This causes troubles
 398	 * in the following case: bic->bfqq has also been scheduled
 399	 * for a possible stable merge with bic->stable_merge_bfqq,
 400	 * and bic->stable_merge_bfqq == bfqq happens to
 401	 * hold. Troubles occur because bfqq may then undergo a split,
 402	 * thereby becoming eligible for a stable merge. Yet, if
 403	 * bic->stable_merge_bfqq points exactly to bfqq, then bfqq
 404	 * would be stably merged with itself. To avoid this anomaly,
 405	 * we cancel the stable merge if
 406	 * bic->stable_merge_bfqq == bfqq.
 407	 */
 408	bic->bfqq[is_sync] = bfqq;
 409
 410	if (bfqq && bic->stable_merge_bfqq == bfqq) {
 411		/*
 412		 * Actually, these same instructions are executed also
 413		 * in bfq_setup_cooperator, in case of abort or actual
 414		 * execution of a stable merge. We could avoid
 415		 * repeating these instructions there too, but if we
 416		 * did so, we would nest even more complexity in this
 417		 * function.
 418		 */
 419		bfq_put_stable_ref(bic->stable_merge_bfqq);
 420
 421		bic->stable_merge_bfqq = NULL;
 422	}
 423}
 424
 425struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic)
 426{
 427	return bic->icq.q->elevator->elevator_data;
 428}
 429
 430/**
 431 * icq_to_bic - convert iocontext queue structure to bfq_io_cq.
 432 * @icq: the iocontext queue.
 433 */
 434static struct bfq_io_cq *icq_to_bic(struct io_cq *icq)
 435{
 436	/* bic->icq is the first member, %NULL will convert to %NULL */
 437	return container_of(icq, struct bfq_io_cq, icq);
 438}
 439
 440/**
 441 * bfq_bic_lookup - search into @ioc a bic associated to @bfqd.
 
 
 442 * @q: the request queue.
 443 */
 444static struct bfq_io_cq *bfq_bic_lookup(struct request_queue *q)
 
 
 445{
 446	struct bfq_io_cq *icq;
 447	unsigned long flags;
 
 448
 449	if (!current->io_context)
 450		return NULL;
 
 451
 452	spin_lock_irqsave(&q->queue_lock, flags);
 453	icq = icq_to_bic(ioc_lookup_icq(q));
 454	spin_unlock_irqrestore(&q->queue_lock, flags);
 455
 456	return icq;
 457}
 458
 459/*
 460 * Scheduler run of queue, if there are requests pending and no one in the
 461 * driver that will restart queueing.
 462 */
 463void bfq_schedule_dispatch(struct bfq_data *bfqd)
 464{
 465	lockdep_assert_held(&bfqd->lock);
 466
 467	if (bfqd->queued != 0) {
 468		bfq_log(bfqd, "schedule dispatch");
 469		blk_mq_run_hw_queues(bfqd->queue, true);
 470	}
 471}
 472
 473#define bfq_class_idle(bfqq)	((bfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
 474
 475#define bfq_sample_valid(samples)	((samples) > 80)
 476
 477/*
 478 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
 479 * We choose the request that is closer to the head right now.  Distance
 480 * behind the head is penalized and only allowed to a certain extent.
 481 */
 482static struct request *bfq_choose_req(struct bfq_data *bfqd,
 483				      struct request *rq1,
 484				      struct request *rq2,
 485				      sector_t last)
 486{
 487	sector_t s1, s2, d1 = 0, d2 = 0;
 488	unsigned long back_max;
 489#define BFQ_RQ1_WRAP	0x01 /* request 1 wraps */
 490#define BFQ_RQ2_WRAP	0x02 /* request 2 wraps */
 491	unsigned int wrap = 0; /* bit mask: requests behind the disk head? */
 492
 493	if (!rq1 || rq1 == rq2)
 494		return rq2;
 495	if (!rq2)
 496		return rq1;
 497
 498	if (rq_is_sync(rq1) && !rq_is_sync(rq2))
 499		return rq1;
 500	else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
 501		return rq2;
 502	if ((rq1->cmd_flags & REQ_META) && !(rq2->cmd_flags & REQ_META))
 503		return rq1;
 504	else if ((rq2->cmd_flags & REQ_META) && !(rq1->cmd_flags & REQ_META))
 505		return rq2;
 506
 507	s1 = blk_rq_pos(rq1);
 508	s2 = blk_rq_pos(rq2);
 509
 510	/*
 511	 * By definition, 1KiB is 2 sectors.
 512	 */
 513	back_max = bfqd->bfq_back_max * 2;
 514
 515	/*
 516	 * Strict one way elevator _except_ in the case where we allow
 517	 * short backward seeks which are biased as twice the cost of a
 518	 * similar forward seek.
 519	 */
 520	if (s1 >= last)
 521		d1 = s1 - last;
 522	else if (s1 + back_max >= last)
 523		d1 = (last - s1) * bfqd->bfq_back_penalty;
 524	else
 525		wrap |= BFQ_RQ1_WRAP;
 526
 527	if (s2 >= last)
 528		d2 = s2 - last;
 529	else if (s2 + back_max >= last)
 530		d2 = (last - s2) * bfqd->bfq_back_penalty;
 531	else
 532		wrap |= BFQ_RQ2_WRAP;
 533
 534	/* Found required data */
 535
 536	/*
 537	 * By doing switch() on the bit mask "wrap" we avoid having to
 538	 * check two variables for all permutations: --> faster!
 539	 */
 540	switch (wrap) {
 541	case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
 542		if (d1 < d2)
 543			return rq1;
 544		else if (d2 < d1)
 545			return rq2;
 546
 547		if (s1 >= s2)
 548			return rq1;
 549		else
 550			return rq2;
 551
 552	case BFQ_RQ2_WRAP:
 553		return rq1;
 554	case BFQ_RQ1_WRAP:
 555		return rq2;
 556	case BFQ_RQ1_WRAP|BFQ_RQ2_WRAP: /* both rqs wrapped */
 557	default:
 558		/*
 559		 * Since both rqs are wrapped,
 560		 * start with the one that's further behind head
 561		 * (--> only *one* back seek required),
 562		 * since back seek takes more time than forward.
 563		 */
 564		if (s1 <= s2)
 565			return rq1;
 566		else
 567			return rq2;
 568	}
 569}
 570
 571#define BFQ_LIMIT_INLINE_DEPTH 16
 572
 573#ifdef CONFIG_BFQ_GROUP_IOSCHED
 574static bool bfqq_request_over_limit(struct bfq_queue *bfqq, int limit)
 575{
 576	struct bfq_data *bfqd = bfqq->bfqd;
 577	struct bfq_entity *entity = &bfqq->entity;
 578	struct bfq_entity *inline_entities[BFQ_LIMIT_INLINE_DEPTH];
 579	struct bfq_entity **entities = inline_entities;
 580	int depth, level, alloc_depth = BFQ_LIMIT_INLINE_DEPTH;
 581	int class_idx = bfqq->ioprio_class - 1;
 582	struct bfq_sched_data *sched_data;
 583	unsigned long wsum;
 584	bool ret = false;
 585
 586	if (!entity->on_st_or_in_serv)
 587		return false;
 588
 589retry:
 590	spin_lock_irq(&bfqd->lock);
 591	/* +1 for bfqq entity, root cgroup not included */
 592	depth = bfqg_to_blkg(bfqq_group(bfqq))->blkcg->css.cgroup->level + 1;
 593	if (depth > alloc_depth) {
 594		spin_unlock_irq(&bfqd->lock);
 595		if (entities != inline_entities)
 596			kfree(entities);
 597		entities = kmalloc_array(depth, sizeof(*entities), GFP_NOIO);
 598		if (!entities)
 599			return false;
 600		alloc_depth = depth;
 601		goto retry;
 602	}
 603
 604	sched_data = entity->sched_data;
 605	/* Gather our ancestors as we need to traverse them in reverse order */
 606	level = 0;
 607	for_each_entity(entity) {
 608		/*
 609		 * If at some level entity is not even active, allow request
 610		 * queueing so that BFQ knows there's work to do and activate
 611		 * entities.
 612		 */
 613		if (!entity->on_st_or_in_serv)
 614			goto out;
 615		/* Uh, more parents than cgroup subsystem thinks? */
 616		if (WARN_ON_ONCE(level >= depth))
 617			break;
 618		entities[level++] = entity;
 619	}
 620	WARN_ON_ONCE(level != depth);
 621	for (level--; level >= 0; level--) {
 622		entity = entities[level];
 623		if (level > 0) {
 624			wsum = bfq_entity_service_tree(entity)->wsum;
 625		} else {
 626			int i;
 627			/*
 628			 * For bfqq itself we take into account service trees
 629			 * of all higher priority classes and multiply their
 630			 * weights so that low prio queue from higher class
 631			 * gets more requests than high prio queue from lower
 632			 * class.
 633			 */
 634			wsum = 0;
 635			for (i = 0; i <= class_idx; i++) {
 636				wsum = wsum * IOPRIO_BE_NR +
 637					sched_data->service_tree[i].wsum;
 638			}
 639		}
 640		limit = DIV_ROUND_CLOSEST(limit * entity->weight, wsum);
 641		if (entity->allocated >= limit) {
 642			bfq_log_bfqq(bfqq->bfqd, bfqq,
 643				"too many requests: allocated %d limit %d level %d",
 644				entity->allocated, limit, level);
 645			ret = true;
 646			break;
 647		}
 648	}
 649out:
 650	spin_unlock_irq(&bfqd->lock);
 651	if (entities != inline_entities)
 652		kfree(entities);
 653	return ret;
 654}
 655#else
 656static bool bfqq_request_over_limit(struct bfq_queue *bfqq, int limit)
 657{
 658	return false;
 659}
 660#endif
 661
 662/*
 663 * Async I/O can easily starve sync I/O (both sync reads and sync
 664 * writes), by consuming all tags. Similarly, storms of sync writes,
 665 * such as those that sync(2) may trigger, can starve sync reads.
 666 * Limit depths of async I/O and sync writes so as to counter both
 667 * problems.
 668 *
 669 * Also if a bfq queue or its parent cgroup consume more tags than would be
 670 * appropriate for their weight, we trim the available tag depth to 1. This
 671 * avoids a situation where one cgroup can starve another cgroup from tags and
 672 * thus block service differentiation among cgroups. Note that because the
 673 * queue / cgroup already has many requests allocated and queued, this does not
 674 * significantly affect service guarantees coming from the BFQ scheduling
 675 * algorithm.
 676 */
 677static void bfq_limit_depth(blk_opf_t opf, struct blk_mq_alloc_data *data)
 678{
 679	struct bfq_data *bfqd = data->q->elevator->elevator_data;
 680	struct bfq_io_cq *bic = bfq_bic_lookup(data->q);
 681	struct bfq_queue *bfqq = bic ? bic_to_bfqq(bic, op_is_sync(opf)) : NULL;
 682	int depth;
 683	unsigned limit = data->q->nr_requests;
 684
 685	/* Sync reads have full depth available */
 686	if (op_is_sync(opf) && !op_is_write(opf)) {
 687		depth = 0;
 688	} else {
 689		depth = bfqd->word_depths[!!bfqd->wr_busy_queues][op_is_sync(opf)];
 690		limit = (limit * depth) >> bfqd->full_depth_shift;
 691	}
 692
 693	/*
 694	 * Does queue (or any parent entity) exceed number of requests that
 695	 * should be available to it? Heavily limit depth so that it cannot
 696	 * consume more available requests and thus starve other entities.
 697	 */
 698	if (bfqq && bfqq_request_over_limit(bfqq, limit))
 699		depth = 1;
 700
 701	bfq_log(bfqd, "[%s] wr_busy %d sync %d depth %u",
 702		__func__, bfqd->wr_busy_queues, op_is_sync(opf), depth);
 703	if (depth)
 704		data->shallow_depth = depth;
 705}
 706
 707static struct bfq_queue *
 708bfq_rq_pos_tree_lookup(struct bfq_data *bfqd, struct rb_root *root,
 709		     sector_t sector, struct rb_node **ret_parent,
 710		     struct rb_node ***rb_link)
 711{
 712	struct rb_node **p, *parent;
 713	struct bfq_queue *bfqq = NULL;
 714
 715	parent = NULL;
 716	p = &root->rb_node;
 717	while (*p) {
 718		struct rb_node **n;
 719
 720		parent = *p;
 721		bfqq = rb_entry(parent, struct bfq_queue, pos_node);
 722
 723		/*
 724		 * Sort strictly based on sector. Smallest to the left,
 725		 * largest to the right.
 726		 */
 727		if (sector > blk_rq_pos(bfqq->next_rq))
 728			n = &(*p)->rb_right;
 729		else if (sector < blk_rq_pos(bfqq->next_rq))
 730			n = &(*p)->rb_left;
 731		else
 732			break;
 733		p = n;
 734		bfqq = NULL;
 735	}
 736
 737	*ret_parent = parent;
 738	if (rb_link)
 739		*rb_link = p;
 740
 741	bfq_log(bfqd, "rq_pos_tree_lookup %llu: returning %d",
 742		(unsigned long long)sector,
 743		bfqq ? bfqq->pid : 0);
 744
 745	return bfqq;
 746}
 747
 748static bool bfq_too_late_for_merging(struct bfq_queue *bfqq)
 749{
 750	return bfqq->service_from_backlogged > 0 &&
 751		time_is_before_jiffies(bfqq->first_IO_time +
 752				       bfq_merge_time_limit);
 753}
 754
 755/*
 756 * The following function is not marked as __cold because it is
 757 * actually cold, but for the same performance goal described in the
 758 * comments on the likely() at the beginning of
 759 * bfq_setup_cooperator(). Unexpectedly, to reach an even lower
 760 * execution time for the case where this function is not invoked, we
 761 * had to add an unlikely() in each involved if().
 762 */
 763void __cold
 764bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq)
 765{
 766	struct rb_node **p, *parent;
 767	struct bfq_queue *__bfqq;
 768
 769	if (bfqq->pos_root) {
 770		rb_erase(&bfqq->pos_node, bfqq->pos_root);
 771		bfqq->pos_root = NULL;
 772	}
 773
 774	/* oom_bfqq does not participate in queue merging */
 775	if (bfqq == &bfqd->oom_bfqq)
 776		return;
 777
 778	/*
 779	 * bfqq cannot be merged any longer (see comments in
 780	 * bfq_setup_cooperator): no point in adding bfqq into the
 781	 * position tree.
 782	 */
 783	if (bfq_too_late_for_merging(bfqq))
 784		return;
 785
 786	if (bfq_class_idle(bfqq))
 787		return;
 788	if (!bfqq->next_rq)
 789		return;
 790
 791	bfqq->pos_root = &bfqq_group(bfqq)->rq_pos_tree;
 792	__bfqq = bfq_rq_pos_tree_lookup(bfqd, bfqq->pos_root,
 793			blk_rq_pos(bfqq->next_rq), &parent, &p);
 794	if (!__bfqq) {
 795		rb_link_node(&bfqq->pos_node, parent, p);
 796		rb_insert_color(&bfqq->pos_node, bfqq->pos_root);
 797	} else
 798		bfqq->pos_root = NULL;
 799}
 800
 801/*
 802 * The following function returns false either if every active queue
 803 * must receive the same share of the throughput (symmetric scenario),
 804 * or, as a special case, if bfqq must receive a share of the
 805 * throughput lower than or equal to the share that every other active
 806 * queue must receive.  If bfqq does sync I/O, then these are the only
 807 * two cases where bfqq happens to be guaranteed its share of the
 808 * throughput even if I/O dispatching is not plugged when bfqq remains
 809 * temporarily empty (for more details, see the comments in the
 810 * function bfq_better_to_idle()). For this reason, the return value
 811 * of this function is used to check whether I/O-dispatch plugging can
 812 * be avoided.
 813 *
 814 * The above first case (symmetric scenario) occurs when:
 815 * 1) all active queues have the same weight,
 816 * 2) all active queues belong to the same I/O-priority class,
 817 * 3) all active groups at the same level in the groups tree have the same
 818 *    weight,
 819 * 4) all active groups at the same level in the groups tree have the same
 820 *    number of children.
 821 *
 822 * Unfortunately, keeping the necessary state for evaluating exactly
 823 * the last two symmetry sub-conditions above would be quite complex
 824 * and time consuming. Therefore this function evaluates, instead,
 825 * only the following stronger three sub-conditions, for which it is
 826 * much easier to maintain the needed state:
 827 * 1) all active queues have the same weight,
 828 * 2) all active queues belong to the same I/O-priority class,
 829 * 3) there is at most one active group.
 830 * In particular, the last condition is always true if hierarchical
 831 * support or the cgroups interface are not enabled, thus no state
 832 * needs to be maintained in this case.
 833 */
 834static bool bfq_asymmetric_scenario(struct bfq_data *bfqd,
 835				   struct bfq_queue *bfqq)
 836{
 837	bool smallest_weight = bfqq &&
 838		bfqq->weight_counter &&
 839		bfqq->weight_counter ==
 840		container_of(
 841			rb_first_cached(&bfqd->queue_weights_tree),
 842			struct bfq_weight_counter,
 843			weights_node);
 844
 845	/*
 846	 * For queue weights to differ, queue_weights_tree must contain
 847	 * at least two nodes.
 848	 */
 849	bool varied_queue_weights = !smallest_weight &&
 850		!RB_EMPTY_ROOT(&bfqd->queue_weights_tree.rb_root) &&
 851		(bfqd->queue_weights_tree.rb_root.rb_node->rb_left ||
 852		 bfqd->queue_weights_tree.rb_root.rb_node->rb_right);
 853
 854	bool multiple_classes_busy =
 855		(bfqd->busy_queues[0] && bfqd->busy_queues[1]) ||
 856		(bfqd->busy_queues[0] && bfqd->busy_queues[2]) ||
 857		(bfqd->busy_queues[1] && bfqd->busy_queues[2]);
 858
 859	return varied_queue_weights || multiple_classes_busy
 860#ifdef CONFIG_BFQ_GROUP_IOSCHED
 861	       || bfqd->num_groups_with_pending_reqs > 1
 862#endif
 863		;
 864}
 865
 866/*
 867 * If the weight-counter tree passed as input contains no counter for
 868 * the weight of the input queue, then add that counter; otherwise just
 869 * increment the existing counter.
 870 *
 871 * Note that weight-counter trees contain few nodes in mostly symmetric
 872 * scenarios. For example, if all queues have the same weight, then the
 873 * weight-counter tree for the queues may contain at most one node.
 874 * This holds even if low_latency is on, because weight-raised queues
 875 * are not inserted in the tree.
 876 * In most scenarios, the rate at which nodes are created/destroyed
 877 * should be low too.
 878 */
 879void bfq_weights_tree_add(struct bfq_queue *bfqq)
 
 880{
 881	struct rb_root_cached *root = &bfqq->bfqd->queue_weights_tree;
 882	struct bfq_entity *entity = &bfqq->entity;
 883	struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL;
 884	bool leftmost = true;
 885
 886	/*
 887	 * Do not insert if the queue is already associated with a
 888	 * counter, which happens if:
 889	 *   1) a request arrival has caused the queue to become both
 890	 *      non-weight-raised, and hence change its weight, and
 891	 *      backlogged; in this respect, each of the two events
 892	 *      causes an invocation of this function,
 893	 *   2) this is the invocation of this function caused by the
 894	 *      second event. This second invocation is actually useless,
 895	 *      and we handle this fact by exiting immediately. More
 896	 *      efficient or clearer solutions might possibly be adopted.
 897	 */
 898	if (bfqq->weight_counter)
 899		return;
 900
 901	while (*new) {
 902		struct bfq_weight_counter *__counter = container_of(*new,
 903						struct bfq_weight_counter,
 904						weights_node);
 905		parent = *new;
 906
 907		if (entity->weight == __counter->weight) {
 908			bfqq->weight_counter = __counter;
 909			goto inc_counter;
 910		}
 911		if (entity->weight < __counter->weight)
 912			new = &((*new)->rb_left);
 913		else {
 914			new = &((*new)->rb_right);
 915			leftmost = false;
 916		}
 917	}
 918
 919	bfqq->weight_counter = kzalloc(sizeof(struct bfq_weight_counter),
 920				       GFP_ATOMIC);
 921
 922	/*
 923	 * In the unlucky event of an allocation failure, we just
 924	 * exit. This will cause the weight of queue to not be
 925	 * considered in bfq_asymmetric_scenario, which, in its turn,
 926	 * causes the scenario to be deemed wrongly symmetric in case
 927	 * bfqq's weight would have been the only weight making the
 928	 * scenario asymmetric.  On the bright side, no unbalance will
 929	 * however occur when bfqq becomes inactive again (the
 930	 * invocation of this function is triggered by an activation
 931	 * of queue).  In fact, bfq_weights_tree_remove does nothing
 932	 * if !bfqq->weight_counter.
 933	 */
 934	if (unlikely(!bfqq->weight_counter))
 935		return;
 936
 937	bfqq->weight_counter->weight = entity->weight;
 938	rb_link_node(&bfqq->weight_counter->weights_node, parent, new);
 939	rb_insert_color_cached(&bfqq->weight_counter->weights_node, root,
 940				leftmost);
 941
 942inc_counter:
 943	bfqq->weight_counter->num_active++;
 944	bfqq->ref++;
 945}
 946
 947/*
 948 * Decrement the weight counter associated with the queue, and, if the
 949 * counter reaches 0, remove the counter from the tree.
 950 * See the comments to the function bfq_weights_tree_add() for considerations
 951 * about overhead.
 952 */
 953void bfq_weights_tree_remove(struct bfq_queue *bfqq)
 
 
 954{
 955	struct rb_root_cached *root;
 956
 957	if (!bfqq->weight_counter)
 958		return;
 959
 960	root = &bfqq->bfqd->queue_weights_tree;
 961	bfqq->weight_counter->num_active--;
 962	if (bfqq->weight_counter->num_active > 0)
 963		goto reset_entity_pointer;
 964
 965	rb_erase_cached(&bfqq->weight_counter->weights_node, root);
 966	kfree(bfqq->weight_counter);
 967
 968reset_entity_pointer:
 969	bfqq->weight_counter = NULL;
 970	bfq_put_queue(bfqq);
 971}
 972
 973/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 974 * Return expired entry, or NULL to just start from scratch in rbtree.
 975 */
 976static struct request *bfq_check_fifo(struct bfq_queue *bfqq,
 977				      struct request *last)
 978{
 979	struct request *rq;
 980
 981	if (bfq_bfqq_fifo_expire(bfqq))
 982		return NULL;
 983
 984	bfq_mark_bfqq_fifo_expire(bfqq);
 985
 986	rq = rq_entry_fifo(bfqq->fifo.next);
 987
 988	if (rq == last || ktime_get_ns() < rq->fifo_time)
 989		return NULL;
 990
 991	bfq_log_bfqq(bfqq->bfqd, bfqq, "check_fifo: returned %p", rq);
 992	return rq;
 993}
 994
 995static struct request *bfq_find_next_rq(struct bfq_data *bfqd,
 996					struct bfq_queue *bfqq,
 997					struct request *last)
 998{
 999	struct rb_node *rbnext = rb_next(&last->rb_node);
1000	struct rb_node *rbprev = rb_prev(&last->rb_node);
1001	struct request *next, *prev = NULL;
1002
1003	/* Follow expired path, else get first next available. */
1004	next = bfq_check_fifo(bfqq, last);
1005	if (next)
1006		return next;
1007
1008	if (rbprev)
1009		prev = rb_entry_rq(rbprev);
1010
1011	if (rbnext)
1012		next = rb_entry_rq(rbnext);
1013	else {
1014		rbnext = rb_first(&bfqq->sort_list);
1015		if (rbnext && rbnext != &last->rb_node)
1016			next = rb_entry_rq(rbnext);
1017	}
1018
1019	return bfq_choose_req(bfqd, next, prev, blk_rq_pos(last));
1020}
1021
1022/* see the definition of bfq_async_charge_factor for details */
1023static unsigned long bfq_serv_to_charge(struct request *rq,
1024					struct bfq_queue *bfqq)
1025{
1026	if (bfq_bfqq_sync(bfqq) || bfqq->wr_coeff > 1 ||
1027	    bfq_asymmetric_scenario(bfqq->bfqd, bfqq))
1028		return blk_rq_sectors(rq);
1029
1030	return blk_rq_sectors(rq) * bfq_async_charge_factor;
1031}
1032
1033/**
1034 * bfq_updated_next_req - update the queue after a new next_rq selection.
1035 * @bfqd: the device data the queue belongs to.
1036 * @bfqq: the queue to update.
1037 *
1038 * If the first request of a queue changes we make sure that the queue
1039 * has enough budget to serve at least its first request (if the
1040 * request has grown).  We do this because if the queue has not enough
1041 * budget for its first request, it has to go through two dispatch
1042 * rounds to actually get it dispatched.
1043 */
1044static void bfq_updated_next_req(struct bfq_data *bfqd,
1045				 struct bfq_queue *bfqq)
1046{
1047	struct bfq_entity *entity = &bfqq->entity;
1048	struct request *next_rq = bfqq->next_rq;
1049	unsigned long new_budget;
1050
1051	if (!next_rq)
1052		return;
1053
1054	if (bfqq == bfqd->in_service_queue)
1055		/*
1056		 * In order not to break guarantees, budgets cannot be
1057		 * changed after an entity has been selected.
1058		 */
1059		return;
1060
1061	new_budget = max_t(unsigned long,
1062			   max_t(unsigned long, bfqq->max_budget,
1063				 bfq_serv_to_charge(next_rq, bfqq)),
1064			   entity->service);
1065	if (entity->budget != new_budget) {
1066		entity->budget = new_budget;
1067		bfq_log_bfqq(bfqd, bfqq, "updated next rq: new budget %lu",
1068					 new_budget);
1069		bfq_requeue_bfqq(bfqd, bfqq, false);
1070	}
1071}
1072
1073static unsigned int bfq_wr_duration(struct bfq_data *bfqd)
1074{
1075	u64 dur;
1076
1077	if (bfqd->bfq_wr_max_time > 0)
1078		return bfqd->bfq_wr_max_time;
1079
1080	dur = bfqd->rate_dur_prod;
1081	do_div(dur, bfqd->peak_rate);
1082
1083	/*
1084	 * Limit duration between 3 and 25 seconds. The upper limit
1085	 * has been conservatively set after the following worst case:
1086	 * on a QEMU/KVM virtual machine
1087	 * - running in a slow PC
1088	 * - with a virtual disk stacked on a slow low-end 5400rpm HDD
1089	 * - serving a heavy I/O workload, such as the sequential reading
1090	 *   of several files
1091	 * mplayer took 23 seconds to start, if constantly weight-raised.
1092	 *
1093	 * As for higher values than that accommodating the above bad
1094	 * scenario, tests show that higher values would often yield
1095	 * the opposite of the desired result, i.e., would worsen
1096	 * responsiveness by allowing non-interactive applications to
1097	 * preserve weight raising for too long.
1098	 *
1099	 * On the other end, lower values than 3 seconds make it
1100	 * difficult for most interactive tasks to complete their jobs
1101	 * before weight-raising finishes.
1102	 */
1103	return clamp_val(dur, msecs_to_jiffies(3000), msecs_to_jiffies(25000));
1104}
1105
1106/* switch back from soft real-time to interactive weight raising */
1107static void switch_back_to_interactive_wr(struct bfq_queue *bfqq,
1108					  struct bfq_data *bfqd)
1109{
1110	bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1111	bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1112	bfqq->last_wr_start_finish = bfqq->wr_start_at_switch_to_srt;
1113}
1114
1115static void
1116bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd,
1117		      struct bfq_io_cq *bic, bool bfq_already_existing)
1118{
1119	unsigned int old_wr_coeff = 1;
1120	bool busy = bfq_already_existing && bfq_bfqq_busy(bfqq);
1121
1122	if (bic->saved_has_short_ttime)
1123		bfq_mark_bfqq_has_short_ttime(bfqq);
1124	else
1125		bfq_clear_bfqq_has_short_ttime(bfqq);
1126
1127	if (bic->saved_IO_bound)
1128		bfq_mark_bfqq_IO_bound(bfqq);
1129	else
1130		bfq_clear_bfqq_IO_bound(bfqq);
1131
1132	bfqq->last_serv_time_ns = bic->saved_last_serv_time_ns;
1133	bfqq->inject_limit = bic->saved_inject_limit;
1134	bfqq->decrease_time_jif = bic->saved_decrease_time_jif;
1135
1136	bfqq->entity.new_weight = bic->saved_weight;
1137	bfqq->ttime = bic->saved_ttime;
1138	bfqq->io_start_time = bic->saved_io_start_time;
1139	bfqq->tot_idle_time = bic->saved_tot_idle_time;
1140	/*
1141	 * Restore weight coefficient only if low_latency is on
1142	 */
1143	if (bfqd->low_latency) {
1144		old_wr_coeff = bfqq->wr_coeff;
1145		bfqq->wr_coeff = bic->saved_wr_coeff;
1146	}
1147	bfqq->service_from_wr = bic->saved_service_from_wr;
1148	bfqq->wr_start_at_switch_to_srt = bic->saved_wr_start_at_switch_to_srt;
1149	bfqq->last_wr_start_finish = bic->saved_last_wr_start_finish;
1150	bfqq->wr_cur_max_time = bic->saved_wr_cur_max_time;
1151
1152	if (bfqq->wr_coeff > 1 && (bfq_bfqq_in_large_burst(bfqq) ||
1153	    time_is_before_jiffies(bfqq->last_wr_start_finish +
1154				   bfqq->wr_cur_max_time))) {
1155		if (bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
1156		    !bfq_bfqq_in_large_burst(bfqq) &&
1157		    time_is_after_eq_jiffies(bfqq->wr_start_at_switch_to_srt +
1158					     bfq_wr_duration(bfqd))) {
1159			switch_back_to_interactive_wr(bfqq, bfqd);
1160		} else {
1161			bfqq->wr_coeff = 1;
1162			bfq_log_bfqq(bfqq->bfqd, bfqq,
1163				     "resume state: switching off wr");
1164		}
1165	}
1166
1167	/* make sure weight will be updated, however we got here */
1168	bfqq->entity.prio_changed = 1;
1169
1170	if (likely(!busy))
1171		return;
1172
1173	if (old_wr_coeff == 1 && bfqq->wr_coeff > 1)
1174		bfqd->wr_busy_queues++;
1175	else if (old_wr_coeff > 1 && bfqq->wr_coeff == 1)
1176		bfqd->wr_busy_queues--;
1177}
1178
1179static int bfqq_process_refs(struct bfq_queue *bfqq)
1180{
1181	return bfqq->ref - bfqq->entity.allocated -
1182		bfqq->entity.on_st_or_in_serv -
1183		(bfqq->weight_counter != NULL) - bfqq->stable_ref;
1184}
1185
1186/* Empty burst list and add just bfqq (see comments on bfq_handle_burst) */
1187static void bfq_reset_burst_list(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1188{
1189	struct bfq_queue *item;
1190	struct hlist_node *n;
1191
1192	hlist_for_each_entry_safe(item, n, &bfqd->burst_list, burst_list_node)
1193		hlist_del_init(&item->burst_list_node);
1194
1195	/*
1196	 * Start the creation of a new burst list only if there is no
1197	 * active queue. See comments on the conditional invocation of
1198	 * bfq_handle_burst().
1199	 */
1200	if (bfq_tot_busy_queues(bfqd) == 0) {
1201		hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list);
1202		bfqd->burst_size = 1;
1203	} else
1204		bfqd->burst_size = 0;
1205
1206	bfqd->burst_parent_entity = bfqq->entity.parent;
1207}
1208
1209/* Add bfqq to the list of queues in current burst (see bfq_handle_burst) */
1210static void bfq_add_to_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1211{
1212	/* Increment burst size to take into account also bfqq */
1213	bfqd->burst_size++;
1214
1215	if (bfqd->burst_size == bfqd->bfq_large_burst_thresh) {
1216		struct bfq_queue *pos, *bfqq_item;
1217		struct hlist_node *n;
1218
1219		/*
1220		 * Enough queues have been activated shortly after each
1221		 * other to consider this burst as large.
1222		 */
1223		bfqd->large_burst = true;
1224
1225		/*
1226		 * We can now mark all queues in the burst list as
1227		 * belonging to a large burst.
1228		 */
1229		hlist_for_each_entry(bfqq_item, &bfqd->burst_list,
1230				     burst_list_node)
1231			bfq_mark_bfqq_in_large_burst(bfqq_item);
1232		bfq_mark_bfqq_in_large_burst(bfqq);
1233
1234		/*
1235		 * From now on, and until the current burst finishes, any
1236		 * new queue being activated shortly after the last queue
1237		 * was inserted in the burst can be immediately marked as
1238		 * belonging to a large burst. So the burst list is not
1239		 * needed any more. Remove it.
1240		 */
1241		hlist_for_each_entry_safe(pos, n, &bfqd->burst_list,
1242					  burst_list_node)
1243			hlist_del_init(&pos->burst_list_node);
1244	} else /*
1245		* Burst not yet large: add bfqq to the burst list. Do
1246		* not increment the ref counter for bfqq, because bfqq
1247		* is removed from the burst list before freeing bfqq
1248		* in put_queue.
1249		*/
1250		hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list);
1251}
1252
1253/*
1254 * If many queues belonging to the same group happen to be created
1255 * shortly after each other, then the processes associated with these
1256 * queues have typically a common goal. In particular, bursts of queue
1257 * creations are usually caused by services or applications that spawn
1258 * many parallel threads/processes. Examples are systemd during boot,
1259 * or git grep. To help these processes get their job done as soon as
1260 * possible, it is usually better to not grant either weight-raising
1261 * or device idling to their queues, unless these queues must be
1262 * protected from the I/O flowing through other active queues.
1263 *
1264 * In this comment we describe, firstly, the reasons why this fact
1265 * holds, and, secondly, the next function, which implements the main
1266 * steps needed to properly mark these queues so that they can then be
1267 * treated in a different way.
1268 *
1269 * The above services or applications benefit mostly from a high
1270 * throughput: the quicker the requests of the activated queues are
1271 * cumulatively served, the sooner the target job of these queues gets
1272 * completed. As a consequence, weight-raising any of these queues,
1273 * which also implies idling the device for it, is almost always
1274 * counterproductive, unless there are other active queues to isolate
1275 * these new queues from. If there no other active queues, then
1276 * weight-raising these new queues just lowers throughput in most
1277 * cases.
1278 *
1279 * On the other hand, a burst of queue creations may be caused also by
1280 * the start of an application that does not consist of a lot of
1281 * parallel I/O-bound threads. In fact, with a complex application,
1282 * several short processes may need to be executed to start-up the
1283 * application. In this respect, to start an application as quickly as
1284 * possible, the best thing to do is in any case to privilege the I/O
1285 * related to the application with respect to all other
1286 * I/O. Therefore, the best strategy to start as quickly as possible
1287 * an application that causes a burst of queue creations is to
1288 * weight-raise all the queues created during the burst. This is the
1289 * exact opposite of the best strategy for the other type of bursts.
1290 *
1291 * In the end, to take the best action for each of the two cases, the
1292 * two types of bursts need to be distinguished. Fortunately, this
1293 * seems relatively easy, by looking at the sizes of the bursts. In
1294 * particular, we found a threshold such that only bursts with a
1295 * larger size than that threshold are apparently caused by
1296 * services or commands such as systemd or git grep. For brevity,
1297 * hereafter we call just 'large' these bursts. BFQ *does not*
1298 * weight-raise queues whose creation occurs in a large burst. In
1299 * addition, for each of these queues BFQ performs or does not perform
1300 * idling depending on which choice boosts the throughput more. The
1301 * exact choice depends on the device and request pattern at
1302 * hand.
1303 *
1304 * Unfortunately, false positives may occur while an interactive task
1305 * is starting (e.g., an application is being started). The
1306 * consequence is that the queues associated with the task do not
1307 * enjoy weight raising as expected. Fortunately these false positives
1308 * are very rare. They typically occur if some service happens to
1309 * start doing I/O exactly when the interactive task starts.
1310 *
1311 * Turning back to the next function, it is invoked only if there are
1312 * no active queues (apart from active queues that would belong to the
1313 * same, possible burst bfqq would belong to), and it implements all
1314 * the steps needed to detect the occurrence of a large burst and to
1315 * properly mark all the queues belonging to it (so that they can then
1316 * be treated in a different way). This goal is achieved by
1317 * maintaining a "burst list" that holds, temporarily, the queues that
1318 * belong to the burst in progress. The list is then used to mark
1319 * these queues as belonging to a large burst if the burst does become
1320 * large. The main steps are the following.
1321 *
1322 * . when the very first queue is created, the queue is inserted into the
1323 *   list (as it could be the first queue in a possible burst)
1324 *
1325 * . if the current burst has not yet become large, and a queue Q that does
1326 *   not yet belong to the burst is activated shortly after the last time
1327 *   at which a new queue entered the burst list, then the function appends
1328 *   Q to the burst list
1329 *
1330 * . if, as a consequence of the previous step, the burst size reaches
1331 *   the large-burst threshold, then
1332 *
1333 *     . all the queues in the burst list are marked as belonging to a
1334 *       large burst
1335 *
1336 *     . the burst list is deleted; in fact, the burst list already served
1337 *       its purpose (keeping temporarily track of the queues in a burst,
1338 *       so as to be able to mark them as belonging to a large burst in the
1339 *       previous sub-step), and now is not needed any more
1340 *
1341 *     . the device enters a large-burst mode
1342 *
1343 * . if a queue Q that does not belong to the burst is created while
1344 *   the device is in large-burst mode and shortly after the last time
1345 *   at which a queue either entered the burst list or was marked as
1346 *   belonging to the current large burst, then Q is immediately marked
1347 *   as belonging to a large burst.
1348 *
1349 * . if a queue Q that does not belong to the burst is created a while
1350 *   later, i.e., not shortly after, than the last time at which a queue
1351 *   either entered the burst list or was marked as belonging to the
1352 *   current large burst, then the current burst is deemed as finished and:
1353 *
1354 *        . the large-burst mode is reset if set
1355 *
1356 *        . the burst list is emptied
1357 *
1358 *        . Q is inserted in the burst list, as Q may be the first queue
1359 *          in a possible new burst (then the burst list contains just Q
1360 *          after this step).
1361 */
1362static void bfq_handle_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1363{
1364	/*
1365	 * If bfqq is already in the burst list or is part of a large
1366	 * burst, or finally has just been split, then there is
1367	 * nothing else to do.
1368	 */
1369	if (!hlist_unhashed(&bfqq->burst_list_node) ||
1370	    bfq_bfqq_in_large_burst(bfqq) ||
1371	    time_is_after_eq_jiffies(bfqq->split_time +
1372				     msecs_to_jiffies(10)))
1373		return;
1374
1375	/*
1376	 * If bfqq's creation happens late enough, or bfqq belongs to
1377	 * a different group than the burst group, then the current
1378	 * burst is finished, and related data structures must be
1379	 * reset.
1380	 *
1381	 * In this respect, consider the special case where bfqq is
1382	 * the very first queue created after BFQ is selected for this
1383	 * device. In this case, last_ins_in_burst and
1384	 * burst_parent_entity are not yet significant when we get
1385	 * here. But it is easy to verify that, whether or not the
1386	 * following condition is true, bfqq will end up being
1387	 * inserted into the burst list. In particular the list will
1388	 * happen to contain only bfqq. And this is exactly what has
1389	 * to happen, as bfqq may be the first queue of the first
1390	 * burst.
1391	 */
1392	if (time_is_before_jiffies(bfqd->last_ins_in_burst +
1393	    bfqd->bfq_burst_interval) ||
1394	    bfqq->entity.parent != bfqd->burst_parent_entity) {
1395		bfqd->large_burst = false;
1396		bfq_reset_burst_list(bfqd, bfqq);
1397		goto end;
1398	}
1399
1400	/*
1401	 * If we get here, then bfqq is being activated shortly after the
1402	 * last queue. So, if the current burst is also large, we can mark
1403	 * bfqq as belonging to this large burst immediately.
1404	 */
1405	if (bfqd->large_burst) {
1406		bfq_mark_bfqq_in_large_burst(bfqq);
1407		goto end;
1408	}
1409
1410	/*
1411	 * If we get here, then a large-burst state has not yet been
1412	 * reached, but bfqq is being activated shortly after the last
1413	 * queue. Then we add bfqq to the burst.
1414	 */
1415	bfq_add_to_burst(bfqd, bfqq);
1416end:
1417	/*
1418	 * At this point, bfqq either has been added to the current
1419	 * burst or has caused the current burst to terminate and a
1420	 * possible new burst to start. In particular, in the second
1421	 * case, bfqq has become the first queue in the possible new
1422	 * burst.  In both cases last_ins_in_burst needs to be moved
1423	 * forward.
1424	 */
1425	bfqd->last_ins_in_burst = jiffies;
1426}
1427
1428static int bfq_bfqq_budget_left(struct bfq_queue *bfqq)
1429{
1430	struct bfq_entity *entity = &bfqq->entity;
1431
1432	return entity->budget - entity->service;
1433}
1434
1435/*
1436 * If enough samples have been computed, return the current max budget
1437 * stored in bfqd, which is dynamically updated according to the
1438 * estimated disk peak rate; otherwise return the default max budget
1439 */
1440static int bfq_max_budget(struct bfq_data *bfqd)
1441{
1442	if (bfqd->budgets_assigned < bfq_stats_min_budgets)
1443		return bfq_default_max_budget;
1444	else
1445		return bfqd->bfq_max_budget;
1446}
1447
1448/*
1449 * Return min budget, which is a fraction of the current or default
1450 * max budget (trying with 1/32)
1451 */
1452static int bfq_min_budget(struct bfq_data *bfqd)
1453{
1454	if (bfqd->budgets_assigned < bfq_stats_min_budgets)
1455		return bfq_default_max_budget / 32;
1456	else
1457		return bfqd->bfq_max_budget / 32;
1458}
1459
1460/*
1461 * The next function, invoked after the input queue bfqq switches from
1462 * idle to busy, updates the budget of bfqq. The function also tells
1463 * whether the in-service queue should be expired, by returning
1464 * true. The purpose of expiring the in-service queue is to give bfqq
1465 * the chance to possibly preempt the in-service queue, and the reason
1466 * for preempting the in-service queue is to achieve one of the two
1467 * goals below.
1468 *
1469 * 1. Guarantee to bfqq its reserved bandwidth even if bfqq has
1470 * expired because it has remained idle. In particular, bfqq may have
1471 * expired for one of the following two reasons:
1472 *
1473 * - BFQQE_NO_MORE_REQUESTS bfqq did not enjoy any device idling
1474 *   and did not make it to issue a new request before its last
1475 *   request was served;
1476 *
1477 * - BFQQE_TOO_IDLE bfqq did enjoy device idling, but did not issue
1478 *   a new request before the expiration of the idling-time.
1479 *
1480 * Even if bfqq has expired for one of the above reasons, the process
1481 * associated with the queue may be however issuing requests greedily,
1482 * and thus be sensitive to the bandwidth it receives (bfqq may have
1483 * remained idle for other reasons: CPU high load, bfqq not enjoying
1484 * idling, I/O throttling somewhere in the path from the process to
1485 * the I/O scheduler, ...). But if, after every expiration for one of
1486 * the above two reasons, bfqq has to wait for the service of at least
1487 * one full budget of another queue before being served again, then
1488 * bfqq is likely to get a much lower bandwidth or resource time than
1489 * its reserved ones. To address this issue, two countermeasures need
1490 * to be taken.
1491 *
1492 * First, the budget and the timestamps of bfqq need to be updated in
1493 * a special way on bfqq reactivation: they need to be updated as if
1494 * bfqq did not remain idle and did not expire. In fact, if they are
1495 * computed as if bfqq expired and remained idle until reactivation,
1496 * then the process associated with bfqq is treated as if, instead of
1497 * being greedy, it stopped issuing requests when bfqq remained idle,
1498 * and restarts issuing requests only on this reactivation. In other
1499 * words, the scheduler does not help the process recover the "service
1500 * hole" between bfqq expiration and reactivation. As a consequence,
1501 * the process receives a lower bandwidth than its reserved one. In
1502 * contrast, to recover this hole, the budget must be updated as if
1503 * bfqq was not expired at all before this reactivation, i.e., it must
1504 * be set to the value of the remaining budget when bfqq was
1505 * expired. Along the same line, timestamps need to be assigned the
1506 * value they had the last time bfqq was selected for service, i.e.,
1507 * before last expiration. Thus timestamps need to be back-shifted
1508 * with respect to their normal computation (see [1] for more details
1509 * on this tricky aspect).
1510 *
1511 * Secondly, to allow the process to recover the hole, the in-service
1512 * queue must be expired too, to give bfqq the chance to preempt it
1513 * immediately. In fact, if bfqq has to wait for a full budget of the
1514 * in-service queue to be completed, then it may become impossible to
1515 * let the process recover the hole, even if the back-shifted
1516 * timestamps of bfqq are lower than those of the in-service queue. If
1517 * this happens for most or all of the holes, then the process may not
1518 * receive its reserved bandwidth. In this respect, it is worth noting
1519 * that, being the service of outstanding requests unpreemptible, a
1520 * little fraction of the holes may however be unrecoverable, thereby
1521 * causing a little loss of bandwidth.
1522 *
1523 * The last important point is detecting whether bfqq does need this
1524 * bandwidth recovery. In this respect, the next function deems the
1525 * process associated with bfqq greedy, and thus allows it to recover
1526 * the hole, if: 1) the process is waiting for the arrival of a new
1527 * request (which implies that bfqq expired for one of the above two
1528 * reasons), and 2) such a request has arrived soon. The first
1529 * condition is controlled through the flag non_blocking_wait_rq,
1530 * while the second through the flag arrived_in_time. If both
1531 * conditions hold, then the function computes the budget in the
1532 * above-described special way, and signals that the in-service queue
1533 * should be expired. Timestamp back-shifting is done later in
1534 * __bfq_activate_entity.
1535 *
1536 * 2. Reduce latency. Even if timestamps are not backshifted to let
1537 * the process associated with bfqq recover a service hole, bfqq may
1538 * however happen to have, after being (re)activated, a lower finish
1539 * timestamp than the in-service queue.	 That is, the next budget of
1540 * bfqq may have to be completed before the one of the in-service
1541 * queue. If this is the case, then preempting the in-service queue
1542 * allows this goal to be achieved, apart from the unpreemptible,
1543 * outstanding requests mentioned above.
1544 *
1545 * Unfortunately, regardless of which of the above two goals one wants
1546 * to achieve, service trees need first to be updated to know whether
1547 * the in-service queue must be preempted. To have service trees
1548 * correctly updated, the in-service queue must be expired and
1549 * rescheduled, and bfqq must be scheduled too. This is one of the
1550 * most costly operations (in future versions, the scheduling
1551 * mechanism may be re-designed in such a way to make it possible to
1552 * know whether preemption is needed without needing to update service
1553 * trees). In addition, queue preemptions almost always cause random
1554 * I/O, which may in turn cause loss of throughput. Finally, there may
1555 * even be no in-service queue when the next function is invoked (so,
1556 * no queue to compare timestamps with). Because of these facts, the
1557 * next function adopts the following simple scheme to avoid costly
1558 * operations, too frequent preemptions and too many dependencies on
1559 * the state of the scheduler: it requests the expiration of the
1560 * in-service queue (unconditionally) only for queues that need to
1561 * recover a hole. Then it delegates to other parts of the code the
1562 * responsibility of handling the above case 2.
1563 */
1564static bool bfq_bfqq_update_budg_for_activation(struct bfq_data *bfqd,
1565						struct bfq_queue *bfqq,
1566						bool arrived_in_time)
1567{
1568	struct bfq_entity *entity = &bfqq->entity;
1569
1570	/*
1571	 * In the next compound condition, we check also whether there
1572	 * is some budget left, because otherwise there is no point in
1573	 * trying to go on serving bfqq with this same budget: bfqq
1574	 * would be expired immediately after being selected for
1575	 * service. This would only cause useless overhead.
1576	 */
1577	if (bfq_bfqq_non_blocking_wait_rq(bfqq) && arrived_in_time &&
1578	    bfq_bfqq_budget_left(bfqq) > 0) {
1579		/*
1580		 * We do not clear the flag non_blocking_wait_rq here, as
1581		 * the latter is used in bfq_activate_bfqq to signal
1582		 * that timestamps need to be back-shifted (and is
1583		 * cleared right after).
1584		 */
1585
1586		/*
1587		 * In next assignment we rely on that either
1588		 * entity->service or entity->budget are not updated
1589		 * on expiration if bfqq is empty (see
1590		 * __bfq_bfqq_recalc_budget). Thus both quantities
1591		 * remain unchanged after such an expiration, and the
1592		 * following statement therefore assigns to
1593		 * entity->budget the remaining budget on such an
1594		 * expiration.
1595		 */
1596		entity->budget = min_t(unsigned long,
1597				       bfq_bfqq_budget_left(bfqq),
1598				       bfqq->max_budget);
1599
1600		/*
1601		 * At this point, we have used entity->service to get
1602		 * the budget left (needed for updating
1603		 * entity->budget). Thus we finally can, and have to,
1604		 * reset entity->service. The latter must be reset
1605		 * because bfqq would otherwise be charged again for
1606		 * the service it has received during its previous
1607		 * service slot(s).
1608		 */
1609		entity->service = 0;
1610
1611		return true;
1612	}
1613
1614	/*
1615	 * We can finally complete expiration, by setting service to 0.
1616	 */
1617	entity->service = 0;
1618	entity->budget = max_t(unsigned long, bfqq->max_budget,
1619			       bfq_serv_to_charge(bfqq->next_rq, bfqq));
1620	bfq_clear_bfqq_non_blocking_wait_rq(bfqq);
1621	return false;
1622}
1623
1624/*
1625 * Return the farthest past time instant according to jiffies
1626 * macros.
1627 */
1628static unsigned long bfq_smallest_from_now(void)
1629{
1630	return jiffies - MAX_JIFFY_OFFSET;
1631}
1632
1633static void bfq_update_bfqq_wr_on_rq_arrival(struct bfq_data *bfqd,
1634					     struct bfq_queue *bfqq,
1635					     unsigned int old_wr_coeff,
1636					     bool wr_or_deserves_wr,
1637					     bool interactive,
1638					     bool in_burst,
1639					     bool soft_rt)
1640{
1641	if (old_wr_coeff == 1 && wr_or_deserves_wr) {
1642		/* start a weight-raising period */
1643		if (interactive) {
1644			bfqq->service_from_wr = 0;
1645			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1646			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1647		} else {
1648			/*
1649			 * No interactive weight raising in progress
1650			 * here: assign minus infinity to
1651			 * wr_start_at_switch_to_srt, to make sure
1652			 * that, at the end of the soft-real-time
1653			 * weight raising periods that is starting
1654			 * now, no interactive weight-raising period
1655			 * may be wrongly considered as still in
1656			 * progress (and thus actually started by
1657			 * mistake).
1658			 */
1659			bfqq->wr_start_at_switch_to_srt =
1660				bfq_smallest_from_now();
1661			bfqq->wr_coeff = bfqd->bfq_wr_coeff *
1662				BFQ_SOFTRT_WEIGHT_FACTOR;
1663			bfqq->wr_cur_max_time =
1664				bfqd->bfq_wr_rt_max_time;
1665		}
1666
1667		/*
1668		 * If needed, further reduce budget to make sure it is
1669		 * close to bfqq's backlog, so as to reduce the
1670		 * scheduling-error component due to a too large
1671		 * budget. Do not care about throughput consequences,
1672		 * but only about latency. Finally, do not assign a
1673		 * too small budget either, to avoid increasing
1674		 * latency by causing too frequent expirations.
1675		 */
1676		bfqq->entity.budget = min_t(unsigned long,
1677					    bfqq->entity.budget,
1678					    2 * bfq_min_budget(bfqd));
1679	} else if (old_wr_coeff > 1) {
1680		if (interactive) { /* update wr coeff and duration */
1681			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1682			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1683		} else if (in_burst)
1684			bfqq->wr_coeff = 1;
1685		else if (soft_rt) {
1686			/*
1687			 * The application is now or still meeting the
1688			 * requirements for being deemed soft rt.  We
1689			 * can then correctly and safely (re)charge
1690			 * the weight-raising duration for the
1691			 * application with the weight-raising
1692			 * duration for soft rt applications.
1693			 *
1694			 * In particular, doing this recharge now, i.e.,
1695			 * before the weight-raising period for the
1696			 * application finishes, reduces the probability
1697			 * of the following negative scenario:
1698			 * 1) the weight of a soft rt application is
1699			 *    raised at startup (as for any newly
1700			 *    created application),
1701			 * 2) since the application is not interactive,
1702			 *    at a certain time weight-raising is
1703			 *    stopped for the application,
1704			 * 3) at that time the application happens to
1705			 *    still have pending requests, and hence
1706			 *    is destined to not have a chance to be
1707			 *    deemed soft rt before these requests are
1708			 *    completed (see the comments to the
1709			 *    function bfq_bfqq_softrt_next_start()
1710			 *    for details on soft rt detection),
1711			 * 4) these pending requests experience a high
1712			 *    latency because the application is not
1713			 *    weight-raised while they are pending.
1714			 */
1715			if (bfqq->wr_cur_max_time !=
1716				bfqd->bfq_wr_rt_max_time) {
1717				bfqq->wr_start_at_switch_to_srt =
1718					bfqq->last_wr_start_finish;
1719
1720				bfqq->wr_cur_max_time =
1721					bfqd->bfq_wr_rt_max_time;
1722				bfqq->wr_coeff = bfqd->bfq_wr_coeff *
1723					BFQ_SOFTRT_WEIGHT_FACTOR;
1724			}
1725			bfqq->last_wr_start_finish = jiffies;
1726		}
1727	}
1728}
1729
1730static bool bfq_bfqq_idle_for_long_time(struct bfq_data *bfqd,
1731					struct bfq_queue *bfqq)
1732{
1733	return bfqq->dispatched == 0 &&
1734		time_is_before_jiffies(
1735			bfqq->budget_timeout +
1736			bfqd->bfq_wr_min_idle_time);
1737}
1738
1739
1740/*
1741 * Return true if bfqq is in a higher priority class, or has a higher
1742 * weight than the in-service queue.
1743 */
1744static bool bfq_bfqq_higher_class_or_weight(struct bfq_queue *bfqq,
1745					    struct bfq_queue *in_serv_bfqq)
1746{
1747	int bfqq_weight, in_serv_weight;
1748
1749	if (bfqq->ioprio_class < in_serv_bfqq->ioprio_class)
1750		return true;
1751
1752	if (in_serv_bfqq->entity.parent == bfqq->entity.parent) {
1753		bfqq_weight = bfqq->entity.weight;
1754		in_serv_weight = in_serv_bfqq->entity.weight;
1755	} else {
1756		if (bfqq->entity.parent)
1757			bfqq_weight = bfqq->entity.parent->weight;
1758		else
1759			bfqq_weight = bfqq->entity.weight;
1760		if (in_serv_bfqq->entity.parent)
1761			in_serv_weight = in_serv_bfqq->entity.parent->weight;
1762		else
1763			in_serv_weight = in_serv_bfqq->entity.weight;
1764	}
1765
1766	return bfqq_weight > in_serv_weight;
1767}
1768
1769static bool bfq_better_to_idle(struct bfq_queue *bfqq);
1770
1771static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd,
1772					     struct bfq_queue *bfqq,
1773					     int old_wr_coeff,
1774					     struct request *rq,
1775					     bool *interactive)
1776{
1777	bool soft_rt, in_burst,	wr_or_deserves_wr,
1778		bfqq_wants_to_preempt,
1779		idle_for_long_time = bfq_bfqq_idle_for_long_time(bfqd, bfqq),
1780		/*
1781		 * See the comments on
1782		 * bfq_bfqq_update_budg_for_activation for
1783		 * details on the usage of the next variable.
1784		 */
1785		arrived_in_time =  ktime_get_ns() <=
1786			bfqq->ttime.last_end_request +
1787			bfqd->bfq_slice_idle * 3;
1788
1789
1790	/*
1791	 * bfqq deserves to be weight-raised if:
1792	 * - it is sync,
1793	 * - it does not belong to a large burst,
1794	 * - it has been idle for enough time or is soft real-time,
1795	 * - is linked to a bfq_io_cq (it is not shared in any sense),
1796	 * - has a default weight (otherwise we assume the user wanted
1797	 *   to control its weight explicitly)
1798	 */
1799	in_burst = bfq_bfqq_in_large_burst(bfqq);
1800	soft_rt = bfqd->bfq_wr_max_softrt_rate > 0 &&
1801		!BFQQ_TOTALLY_SEEKY(bfqq) &&
1802		!in_burst &&
1803		time_is_before_jiffies(bfqq->soft_rt_next_start) &&
1804		bfqq->dispatched == 0 &&
1805		bfqq->entity.new_weight == 40;
1806	*interactive = !in_burst && idle_for_long_time &&
1807		bfqq->entity.new_weight == 40;
1808	/*
1809	 * Merged bfq_queues are kept out of weight-raising
1810	 * (low-latency) mechanisms. The reason is that these queues
1811	 * are usually created for non-interactive and
1812	 * non-soft-real-time tasks. Yet this is not the case for
1813	 * stably-merged queues. These queues are merged just because
1814	 * they are created shortly after each other. So they may
1815	 * easily serve the I/O of an interactive or soft-real time
1816	 * application, if the application happens to spawn multiple
1817	 * processes. So let also stably-merged queued enjoy weight
1818	 * raising.
1819	 */
1820	wr_or_deserves_wr = bfqd->low_latency &&
1821		(bfqq->wr_coeff > 1 ||
1822		 (bfq_bfqq_sync(bfqq) &&
1823		  (bfqq->bic || RQ_BIC(rq)->stably_merged) &&
1824		   (*interactive || soft_rt)));
1825
1826	/*
1827	 * Using the last flag, update budget and check whether bfqq
1828	 * may want to preempt the in-service queue.
1829	 */
1830	bfqq_wants_to_preempt =
1831		bfq_bfqq_update_budg_for_activation(bfqd, bfqq,
1832						    arrived_in_time);
1833
1834	/*
1835	 * If bfqq happened to be activated in a burst, but has been
1836	 * idle for much more than an interactive queue, then we
1837	 * assume that, in the overall I/O initiated in the burst, the
1838	 * I/O associated with bfqq is finished. So bfqq does not need
1839	 * to be treated as a queue belonging to a burst
1840	 * anymore. Accordingly, we reset bfqq's in_large_burst flag
1841	 * if set, and remove bfqq from the burst list if it's
1842	 * there. We do not decrement burst_size, because the fact
1843	 * that bfqq does not need to belong to the burst list any
1844	 * more does not invalidate the fact that bfqq was created in
1845	 * a burst.
1846	 */
1847	if (likely(!bfq_bfqq_just_created(bfqq)) &&
1848	    idle_for_long_time &&
1849	    time_is_before_jiffies(
1850		    bfqq->budget_timeout +
1851		    msecs_to_jiffies(10000))) {
1852		hlist_del_init(&bfqq->burst_list_node);
1853		bfq_clear_bfqq_in_large_burst(bfqq);
1854	}
1855
1856	bfq_clear_bfqq_just_created(bfqq);
1857
 
 
 
 
 
 
 
 
 
 
 
1858	if (bfqd->low_latency) {
1859		if (unlikely(time_is_after_jiffies(bfqq->split_time)))
1860			/* wraparound */
1861			bfqq->split_time =
1862				jiffies - bfqd->bfq_wr_min_idle_time - 1;
1863
1864		if (time_is_before_jiffies(bfqq->split_time +
1865					   bfqd->bfq_wr_min_idle_time)) {
1866			bfq_update_bfqq_wr_on_rq_arrival(bfqd, bfqq,
1867							 old_wr_coeff,
1868							 wr_or_deserves_wr,
1869							 *interactive,
1870							 in_burst,
1871							 soft_rt);
1872
1873			if (old_wr_coeff != bfqq->wr_coeff)
1874				bfqq->entity.prio_changed = 1;
1875		}
1876	}
1877
1878	bfqq->last_idle_bklogged = jiffies;
1879	bfqq->service_from_backlogged = 0;
1880	bfq_clear_bfqq_softrt_update(bfqq);
1881
1882	bfq_add_bfqq_busy(bfqq);
1883
1884	/*
1885	 * Expire in-service queue if preemption may be needed for
1886	 * guarantees or throughput. As for guarantees, we care
1887	 * explicitly about two cases. The first is that bfqq has to
1888	 * recover a service hole, as explained in the comments on
1889	 * bfq_bfqq_update_budg_for_activation(), i.e., that
1890	 * bfqq_wants_to_preempt is true. However, if bfqq does not
1891	 * carry time-critical I/O, then bfqq's bandwidth is less
1892	 * important than that of queues that carry time-critical I/O.
1893	 * So, as a further constraint, we consider this case only if
1894	 * bfqq is at least as weight-raised, i.e., at least as time
1895	 * critical, as the in-service queue.
1896	 *
1897	 * The second case is that bfqq is in a higher priority class,
1898	 * or has a higher weight than the in-service queue. If this
1899	 * condition does not hold, we don't care because, even if
1900	 * bfqq does not start to be served immediately, the resulting
1901	 * delay for bfqq's I/O is however lower or much lower than
1902	 * the ideal completion time to be guaranteed to bfqq's I/O.
1903	 *
1904	 * In both cases, preemption is needed only if, according to
1905	 * the timestamps of both bfqq and of the in-service queue,
1906	 * bfqq actually is the next queue to serve. So, to reduce
1907	 * useless preemptions, the return value of
1908	 * next_queue_may_preempt() is considered in the next compound
1909	 * condition too. Yet next_queue_may_preempt() just checks a
1910	 * simple, necessary condition for bfqq to be the next queue
1911	 * to serve. In fact, to evaluate a sufficient condition, the
1912	 * timestamps of the in-service queue would need to be
1913	 * updated, and this operation is quite costly (see the
1914	 * comments on bfq_bfqq_update_budg_for_activation()).
1915	 *
1916	 * As for throughput, we ask bfq_better_to_idle() whether we
1917	 * still need to plug I/O dispatching. If bfq_better_to_idle()
1918	 * says no, then plugging is not needed any longer, either to
1919	 * boost throughput or to perserve service guarantees. Then
1920	 * the best option is to stop plugging I/O, as not doing so
1921	 * would certainly lower throughput. We may end up in this
1922	 * case if: (1) upon a dispatch attempt, we detected that it
1923	 * was better to plug I/O dispatch, and to wait for a new
1924	 * request to arrive for the currently in-service queue, but
1925	 * (2) this switch of bfqq to busy changes the scenario.
1926	 */
1927	if (bfqd->in_service_queue &&
1928	    ((bfqq_wants_to_preempt &&
1929	      bfqq->wr_coeff >= bfqd->in_service_queue->wr_coeff) ||
1930	     bfq_bfqq_higher_class_or_weight(bfqq, bfqd->in_service_queue) ||
1931	     !bfq_better_to_idle(bfqd->in_service_queue)) &&
1932	    next_queue_may_preempt(bfqd))
1933		bfq_bfqq_expire(bfqd, bfqd->in_service_queue,
1934				false, BFQQE_PREEMPTED);
1935}
1936
1937static void bfq_reset_inject_limit(struct bfq_data *bfqd,
1938				   struct bfq_queue *bfqq)
1939{
1940	/* invalidate baseline total service time */
1941	bfqq->last_serv_time_ns = 0;
1942
1943	/*
1944	 * Reset pointer in case we are waiting for
1945	 * some request completion.
1946	 */
1947	bfqd->waited_rq = NULL;
1948
1949	/*
1950	 * If bfqq has a short think time, then start by setting the
1951	 * inject limit to 0 prudentially, because the service time of
1952	 * an injected I/O request may be higher than the think time
1953	 * of bfqq, and therefore, if one request was injected when
1954	 * bfqq remains empty, this injected request might delay the
1955	 * service of the next I/O request for bfqq significantly. In
1956	 * case bfqq can actually tolerate some injection, then the
1957	 * adaptive update will however raise the limit soon. This
1958	 * lucky circumstance holds exactly because bfqq has a short
1959	 * think time, and thus, after remaining empty, is likely to
1960	 * get new I/O enqueued---and then completed---before being
1961	 * expired. This is the very pattern that gives the
1962	 * limit-update algorithm the chance to measure the effect of
1963	 * injection on request service times, and then to update the
1964	 * limit accordingly.
1965	 *
1966	 * However, in the following special case, the inject limit is
1967	 * left to 1 even if the think time is short: bfqq's I/O is
1968	 * synchronized with that of some other queue, i.e., bfqq may
1969	 * receive new I/O only after the I/O of the other queue is
1970	 * completed. Keeping the inject limit to 1 allows the
1971	 * blocking I/O to be served while bfqq is in service. And
1972	 * this is very convenient both for bfqq and for overall
1973	 * throughput, as explained in detail in the comments in
1974	 * bfq_update_has_short_ttime().
1975	 *
1976	 * On the opposite end, if bfqq has a long think time, then
1977	 * start directly by 1, because:
1978	 * a) on the bright side, keeping at most one request in
1979	 * service in the drive is unlikely to cause any harm to the
1980	 * latency of bfqq's requests, as the service time of a single
1981	 * request is likely to be lower than the think time of bfqq;
1982	 * b) on the downside, after becoming empty, bfqq is likely to
1983	 * expire before getting its next request. With this request
1984	 * arrival pattern, it is very hard to sample total service
1985	 * times and update the inject limit accordingly (see comments
1986	 * on bfq_update_inject_limit()). So the limit is likely to be
1987	 * never, or at least seldom, updated.  As a consequence, by
1988	 * setting the limit to 1, we avoid that no injection ever
1989	 * occurs with bfqq. On the downside, this proactive step
1990	 * further reduces chances to actually compute the baseline
1991	 * total service time. Thus it reduces chances to execute the
1992	 * limit-update algorithm and possibly raise the limit to more
1993	 * than 1.
1994	 */
1995	if (bfq_bfqq_has_short_ttime(bfqq))
1996		bfqq->inject_limit = 0;
1997	else
1998		bfqq->inject_limit = 1;
1999
2000	bfqq->decrease_time_jif = jiffies;
2001}
2002
2003static void bfq_update_io_intensity(struct bfq_queue *bfqq, u64 now_ns)
2004{
2005	u64 tot_io_time = now_ns - bfqq->io_start_time;
2006
2007	if (RB_EMPTY_ROOT(&bfqq->sort_list) && bfqq->dispatched == 0)
2008		bfqq->tot_idle_time +=
2009			now_ns - bfqq->ttime.last_end_request;
2010
2011	if (unlikely(bfq_bfqq_just_created(bfqq)))
2012		return;
2013
2014	/*
2015	 * Must be busy for at least about 80% of the time to be
2016	 * considered I/O bound.
2017	 */
2018	if (bfqq->tot_idle_time * 5 > tot_io_time)
2019		bfq_clear_bfqq_IO_bound(bfqq);
2020	else
2021		bfq_mark_bfqq_IO_bound(bfqq);
2022
2023	/*
2024	 * Keep an observation window of at most 200 ms in the past
2025	 * from now.
2026	 */
2027	if (tot_io_time > 200 * NSEC_PER_MSEC) {
2028		bfqq->io_start_time = now_ns - (tot_io_time>>1);
2029		bfqq->tot_idle_time >>= 1;
2030	}
2031}
2032
2033/*
2034 * Detect whether bfqq's I/O seems synchronized with that of some
2035 * other queue, i.e., whether bfqq, after remaining empty, happens to
2036 * receive new I/O only right after some I/O request of the other
2037 * queue has been completed. We call waker queue the other queue, and
2038 * we assume, for simplicity, that bfqq may have at most one waker
2039 * queue.
2040 *
2041 * A remarkable throughput boost can be reached by unconditionally
2042 * injecting the I/O of the waker queue, every time a new
2043 * bfq_dispatch_request happens to be invoked while I/O is being
2044 * plugged for bfqq.  In addition to boosting throughput, this
2045 * unblocks bfqq's I/O, thereby improving bandwidth and latency for
2046 * bfqq. Note that these same results may be achieved with the general
2047 * injection mechanism, but less effectively. For details on this
2048 * aspect, see the comments on the choice of the queue for injection
2049 * in bfq_select_queue().
2050 *
2051 * Turning back to the detection of a waker queue, a queue Q is deemed as a
2052 * waker queue for bfqq if, for three consecutive times, bfqq happens to become
2053 * non empty right after a request of Q has been completed within given
2054 * timeout. In this respect, even if bfqq is empty, we do not check for a waker
2055 * if it still has some in-flight I/O. In fact, in this case bfqq is actually
2056 * still being served by the drive, and may receive new I/O on the completion
2057 * of some of the in-flight requests. In particular, on the first time, Q is
2058 * tentatively set as a candidate waker queue, while on the third consecutive
2059 * time that Q is detected, the field waker_bfqq is set to Q, to confirm that Q
2060 * is a waker queue for bfqq. These detection steps are performed only if bfqq
2061 * has a long think time, so as to make it more likely that bfqq's I/O is
2062 * actually being blocked by a synchronization. This last filter, plus the
2063 * above three-times requirement and time limit for detection, make false
2064 * positives less likely.
2065 *
2066 * NOTE
2067 *
2068 * The sooner a waker queue is detected, the sooner throughput can be
2069 * boosted by injecting I/O from the waker queue. Fortunately,
2070 * detection is likely to be actually fast, for the following
2071 * reasons. While blocked by synchronization, bfqq has a long think
2072 * time. This implies that bfqq's inject limit is at least equal to 1
2073 * (see the comments in bfq_update_inject_limit()). So, thanks to
2074 * injection, the waker queue is likely to be served during the very
2075 * first I/O-plugging time interval for bfqq. This triggers the first
2076 * step of the detection mechanism. Thanks again to injection, the
2077 * candidate waker queue is then likely to be confirmed no later than
2078 * during the next I/O-plugging interval for bfqq.
2079 *
2080 * ISSUE
2081 *
2082 * On queue merging all waker information is lost.
2083 */
2084static void bfq_check_waker(struct bfq_data *bfqd, struct bfq_queue *bfqq,
2085			    u64 now_ns)
2086{
2087	char waker_name[MAX_BFQQ_NAME_LENGTH];
2088
2089	if (!bfqd->last_completed_rq_bfqq ||
2090	    bfqd->last_completed_rq_bfqq == bfqq ||
2091	    bfq_bfqq_has_short_ttime(bfqq) ||
2092	    now_ns - bfqd->last_completion >= 4 * NSEC_PER_MSEC ||
2093	    bfqd->last_completed_rq_bfqq == &bfqd->oom_bfqq ||
2094	    bfqq == &bfqd->oom_bfqq)
2095		return;
2096
2097	/*
2098	 * We reset waker detection logic also if too much time has passed
2099 	 * since the first detection. If wakeups are rare, pointless idling
2100	 * doesn't hurt throughput that much. The condition below makes sure
2101	 * we do not uselessly idle blocking waker in more than 1/64 cases. 
2102	 */
2103	if (bfqd->last_completed_rq_bfqq !=
2104	    bfqq->tentative_waker_bfqq ||
2105	    now_ns > bfqq->waker_detection_started +
2106					128 * (u64)bfqd->bfq_slice_idle) {
2107		/*
2108		 * First synchronization detected with a
2109		 * candidate waker queue, or with a different
2110		 * candidate waker queue from the current one.
2111		 */
2112		bfqq->tentative_waker_bfqq =
2113			bfqd->last_completed_rq_bfqq;
2114		bfqq->num_waker_detections = 1;
2115		bfqq->waker_detection_started = now_ns;
2116		bfq_bfqq_name(bfqq->tentative_waker_bfqq, waker_name,
2117			      MAX_BFQQ_NAME_LENGTH);
2118		bfq_log_bfqq(bfqd, bfqq, "set tentative waker %s", waker_name);
2119	} else /* Same tentative waker queue detected again */
2120		bfqq->num_waker_detections++;
2121
2122	if (bfqq->num_waker_detections == 3) {
2123		bfqq->waker_bfqq = bfqd->last_completed_rq_bfqq;
2124		bfqq->tentative_waker_bfqq = NULL;
2125		bfq_bfqq_name(bfqq->waker_bfqq, waker_name,
2126			      MAX_BFQQ_NAME_LENGTH);
2127		bfq_log_bfqq(bfqd, bfqq, "set waker %s", waker_name);
2128
2129		/*
2130		 * If the waker queue disappears, then
2131		 * bfqq->waker_bfqq must be reset. To
2132		 * this goal, we maintain in each
2133		 * waker queue a list, woken_list, of
2134		 * all the queues that reference the
2135		 * waker queue through their
2136		 * waker_bfqq pointer. When the waker
2137		 * queue exits, the waker_bfqq pointer
2138		 * of all the queues in the woken_list
2139		 * is reset.
2140		 *
2141		 * In addition, if bfqq is already in
2142		 * the woken_list of a waker queue,
2143		 * then, before being inserted into
2144		 * the woken_list of a new waker
2145		 * queue, bfqq must be removed from
2146		 * the woken_list of the old waker
2147		 * queue.
2148		 */
2149		if (!hlist_unhashed(&bfqq->woken_list_node))
2150			hlist_del_init(&bfqq->woken_list_node);
2151		hlist_add_head(&bfqq->woken_list_node,
2152			       &bfqd->last_completed_rq_bfqq->woken_list);
2153	}
2154}
2155
2156static void bfq_add_request(struct request *rq)
2157{
2158	struct bfq_queue *bfqq = RQ_BFQQ(rq);
2159	struct bfq_data *bfqd = bfqq->bfqd;
2160	struct request *next_rq, *prev;
2161	unsigned int old_wr_coeff = bfqq->wr_coeff;
2162	bool interactive = false;
2163	u64 now_ns = ktime_get_ns();
2164
2165	bfq_log_bfqq(bfqd, bfqq, "add_request %d", rq_is_sync(rq));
2166	bfqq->queued[rq_is_sync(rq)]++;
2167	/*
2168	 * Updating of 'bfqd->queued' is protected by 'bfqd->lock', however, it
2169	 * may be read without holding the lock in bfq_has_work().
2170	 */
2171	WRITE_ONCE(bfqd->queued, bfqd->queued + 1);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2172
2173	if (bfq_bfqq_sync(bfqq) && RQ_BIC(rq)->requests <= 1) {
2174		bfq_check_waker(bfqd, bfqq, now_ns);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2175
2176		/*
2177		 * Periodically reset inject limit, to make sure that
2178		 * the latter eventually drops in case workload
2179		 * changes, see step (3) in the comments on
2180		 * bfq_update_inject_limit().
2181		 */
2182		if (time_is_before_eq_jiffies(bfqq->decrease_time_jif +
2183					     msecs_to_jiffies(1000)))
2184			bfq_reset_inject_limit(bfqd, bfqq);
2185
2186		/*
2187		 * The following conditions must hold to setup a new
2188		 * sampling of total service time, and then a new
2189		 * update of the inject limit:
2190		 * - bfqq is in service, because the total service
2191		 *   time is evaluated only for the I/O requests of
2192		 *   the queues in service;
2193		 * - this is the right occasion to compute or to
2194		 *   lower the baseline total service time, because
2195		 *   there are actually no requests in the drive,
2196		 *   or
2197		 *   the baseline total service time is available, and
2198		 *   this is the right occasion to compute the other
2199		 *   quantity needed to update the inject limit, i.e.,
2200		 *   the total service time caused by the amount of
2201		 *   injection allowed by the current value of the
2202		 *   limit. It is the right occasion because injection
2203		 *   has actually been performed during the service
2204		 *   hole, and there are still in-flight requests,
2205		 *   which are very likely to be exactly the injected
2206		 *   requests, or part of them;
2207		 * - the minimum interval for sampling the total
2208		 *   service time and updating the inject limit has
2209		 *   elapsed.
2210		 */
2211		if (bfqq == bfqd->in_service_queue &&
2212		    (bfqd->rq_in_driver == 0 ||
2213		     (bfqq->last_serv_time_ns > 0 &&
2214		      bfqd->rqs_injected && bfqd->rq_in_driver > 0)) &&
2215		    time_is_before_eq_jiffies(bfqq->decrease_time_jif +
2216					      msecs_to_jiffies(10))) {
2217			bfqd->last_empty_occupied_ns = ktime_get_ns();
2218			/*
2219			 * Start the state machine for measuring the
2220			 * total service time of rq: setting
2221			 * wait_dispatch will cause bfqd->waited_rq to
2222			 * be set when rq will be dispatched.
2223			 */
2224			bfqd->wait_dispatch = true;
2225			/*
2226			 * If there is no I/O in service in the drive,
2227			 * then possible injection occurred before the
2228			 * arrival of rq will not affect the total
2229			 * service time of rq. So the injection limit
2230			 * must not be updated as a function of such
2231			 * total service time, unless new injection
2232			 * occurs before rq is completed. To have the
2233			 * injection limit updated only in the latter
2234			 * case, reset rqs_injected here (rqs_injected
2235			 * will be set in case injection is performed
2236			 * on bfqq before rq is completed).
2237			 */
2238			if (bfqd->rq_in_driver == 0)
2239				bfqd->rqs_injected = false;
2240		}
2241	}
2242
2243	if (bfq_bfqq_sync(bfqq))
2244		bfq_update_io_intensity(bfqq, now_ns);
2245
2246	elv_rb_add(&bfqq->sort_list, rq);
2247
2248	/*
2249	 * Check if this request is a better next-serve candidate.
2250	 */
2251	prev = bfqq->next_rq;
2252	next_rq = bfq_choose_req(bfqd, bfqq->next_rq, rq, bfqd->last_position);
2253	bfqq->next_rq = next_rq;
2254
2255	/*
2256	 * Adjust priority tree position, if next_rq changes.
2257	 * See comments on bfq_pos_tree_add_move() for the unlikely().
2258	 */
2259	if (unlikely(!bfqd->nonrot_with_queueing && prev != bfqq->next_rq))
2260		bfq_pos_tree_add_move(bfqd, bfqq);
2261
2262	if (!bfq_bfqq_busy(bfqq)) /* switching to busy ... */
2263		bfq_bfqq_handle_idle_busy_switch(bfqd, bfqq, old_wr_coeff,
2264						 rq, &interactive);
2265	else {
2266		if (bfqd->low_latency && old_wr_coeff == 1 && !rq_is_sync(rq) &&
2267		    time_is_before_jiffies(
2268				bfqq->last_wr_start_finish +
2269				bfqd->bfq_wr_min_inter_arr_async)) {
2270			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
2271			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
2272
2273			bfqd->wr_busy_queues++;
2274			bfqq->entity.prio_changed = 1;
2275		}
2276		if (prev != bfqq->next_rq)
2277			bfq_updated_next_req(bfqd, bfqq);
2278	}
2279
2280	/*
2281	 * Assign jiffies to last_wr_start_finish in the following
2282	 * cases:
2283	 *
2284	 * . if bfqq is not going to be weight-raised, because, for
2285	 *   non weight-raised queues, last_wr_start_finish stores the
2286	 *   arrival time of the last request; as of now, this piece
2287	 *   of information is used only for deciding whether to
2288	 *   weight-raise async queues
2289	 *
2290	 * . if bfqq is not weight-raised, because, if bfqq is now
2291	 *   switching to weight-raised, then last_wr_start_finish
2292	 *   stores the time when weight-raising starts
2293	 *
2294	 * . if bfqq is interactive, because, regardless of whether
2295	 *   bfqq is currently weight-raised, the weight-raising
2296	 *   period must start or restart (this case is considered
2297	 *   separately because it is not detected by the above
2298	 *   conditions, if bfqq is already weight-raised)
2299	 *
2300	 * last_wr_start_finish has to be updated also if bfqq is soft
2301	 * real-time, because the weight-raising period is constantly
2302	 * restarted on idle-to-busy transitions for these queues, but
2303	 * this is already done in bfq_bfqq_handle_idle_busy_switch if
2304	 * needed.
2305	 */
2306	if (bfqd->low_latency &&
2307		(old_wr_coeff == 1 || bfqq->wr_coeff == 1 || interactive))
2308		bfqq->last_wr_start_finish = jiffies;
2309}
2310
2311static struct request *bfq_find_rq_fmerge(struct bfq_data *bfqd,
2312					  struct bio *bio,
2313					  struct request_queue *q)
2314{
2315	struct bfq_queue *bfqq = bfqd->bio_bfqq;
2316
2317
2318	if (bfqq)
2319		return elv_rb_find(&bfqq->sort_list, bio_end_sector(bio));
2320
2321	return NULL;
2322}
2323
2324static sector_t get_sdist(sector_t last_pos, struct request *rq)
2325{
2326	if (last_pos)
2327		return abs(blk_rq_pos(rq) - last_pos);
2328
2329	return 0;
2330}
2331
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2332static void bfq_remove_request(struct request_queue *q,
2333			       struct request *rq)
2334{
2335	struct bfq_queue *bfqq = RQ_BFQQ(rq);
2336	struct bfq_data *bfqd = bfqq->bfqd;
2337	const int sync = rq_is_sync(rq);
2338
2339	if (bfqq->next_rq == rq) {
2340		bfqq->next_rq = bfq_find_next_rq(bfqd, bfqq, rq);
2341		bfq_updated_next_req(bfqd, bfqq);
2342	}
2343
2344	if (rq->queuelist.prev != &rq->queuelist)
2345		list_del_init(&rq->queuelist);
2346	bfqq->queued[sync]--;
2347	/*
2348	 * Updating of 'bfqd->queued' is protected by 'bfqd->lock', however, it
2349	 * may be read without holding the lock in bfq_has_work().
2350	 */
2351	WRITE_ONCE(bfqd->queued, bfqd->queued - 1);
2352	elv_rb_del(&bfqq->sort_list, rq);
2353
2354	elv_rqhash_del(q, rq);
2355	if (q->last_merge == rq)
2356		q->last_merge = NULL;
2357
2358	if (RB_EMPTY_ROOT(&bfqq->sort_list)) {
2359		bfqq->next_rq = NULL;
2360
2361		if (bfq_bfqq_busy(bfqq) && bfqq != bfqd->in_service_queue) {
2362			bfq_del_bfqq_busy(bfqq, false);
2363			/*
2364			 * bfqq emptied. In normal operation, when
2365			 * bfqq is empty, bfqq->entity.service and
2366			 * bfqq->entity.budget must contain,
2367			 * respectively, the service received and the
2368			 * budget used last time bfqq emptied. These
2369			 * facts do not hold in this case, as at least
2370			 * this last removal occurred while bfqq is
2371			 * not in service. To avoid inconsistencies,
2372			 * reset both bfqq->entity.service and
2373			 * bfqq->entity.budget, if bfqq has still a
2374			 * process that may issue I/O requests to it.
2375			 */
2376			bfqq->entity.budget = bfqq->entity.service = 0;
2377		}
2378
2379		/*
2380		 * Remove queue from request-position tree as it is empty.
2381		 */
2382		if (bfqq->pos_root) {
2383			rb_erase(&bfqq->pos_node, bfqq->pos_root);
2384			bfqq->pos_root = NULL;
2385		}
2386	} else {
2387		/* see comments on bfq_pos_tree_add_move() for the unlikely() */
2388		if (unlikely(!bfqd->nonrot_with_queueing))
2389			bfq_pos_tree_add_move(bfqd, bfqq);
2390	}
2391
2392	if (rq->cmd_flags & REQ_META)
2393		bfqq->meta_pending--;
2394
2395}
2396
2397static bool bfq_bio_merge(struct request_queue *q, struct bio *bio,
2398		unsigned int nr_segs)
2399{
 
2400	struct bfq_data *bfqd = q->elevator->elevator_data;
2401	struct request *free = NULL;
2402	/*
2403	 * bfq_bic_lookup grabs the queue_lock: invoke it now and
2404	 * store its return value for later use, to avoid nesting
2405	 * queue_lock inside the bfqd->lock. We assume that the bic
2406	 * returned by bfq_bic_lookup does not go away before
2407	 * bfqd->lock is taken.
2408	 */
2409	struct bfq_io_cq *bic = bfq_bic_lookup(q);
2410	bool ret;
2411
2412	spin_lock_irq(&bfqd->lock);
2413
2414	if (bic) {
2415		/*
2416		 * Make sure cgroup info is uptodate for current process before
2417		 * considering the merge.
2418		 */
2419		bfq_bic_update_cgroup(bic, bio);
2420
2421		bfqd->bio_bfqq = bic_to_bfqq(bic, op_is_sync(bio->bi_opf));
2422	} else {
2423		bfqd->bio_bfqq = NULL;
2424	}
2425	bfqd->bio_bic = bic;
2426
2427	ret = blk_mq_sched_try_merge(q, bio, nr_segs, &free);
2428
2429	spin_unlock_irq(&bfqd->lock);
2430	if (free)
2431		blk_mq_free_request(free);
 
2432
2433	return ret;
2434}
2435
2436static int bfq_request_merge(struct request_queue *q, struct request **req,
2437			     struct bio *bio)
2438{
2439	struct bfq_data *bfqd = q->elevator->elevator_data;
2440	struct request *__rq;
2441
2442	__rq = bfq_find_rq_fmerge(bfqd, bio, q);
2443	if (__rq && elv_bio_merge_ok(__rq, bio)) {
2444		*req = __rq;
2445
2446		if (blk_discard_mergable(__rq))
2447			return ELEVATOR_DISCARD_MERGE;
2448		return ELEVATOR_FRONT_MERGE;
2449	}
2450
2451	return ELEVATOR_NO_MERGE;
2452}
2453
 
 
2454static void bfq_request_merged(struct request_queue *q, struct request *req,
2455			       enum elv_merge type)
2456{
2457	if (type == ELEVATOR_FRONT_MERGE &&
2458	    rb_prev(&req->rb_node) &&
2459	    blk_rq_pos(req) <
2460	    blk_rq_pos(container_of(rb_prev(&req->rb_node),
2461				    struct request, rb_node))) {
2462		struct bfq_queue *bfqq = RQ_BFQQ(req);
2463		struct bfq_data *bfqd;
2464		struct request *prev, *next_rq;
2465
2466		if (!bfqq)
2467			return;
2468
2469		bfqd = bfqq->bfqd;
2470
2471		/* Reposition request in its sort_list */
2472		elv_rb_del(&bfqq->sort_list, req);
2473		elv_rb_add(&bfqq->sort_list, req);
2474
2475		/* Choose next request to be served for bfqq */
2476		prev = bfqq->next_rq;
2477		next_rq = bfq_choose_req(bfqd, bfqq->next_rq, req,
2478					 bfqd->last_position);
2479		bfqq->next_rq = next_rq;
2480		/*
2481		 * If next_rq changes, update both the queue's budget to
2482		 * fit the new request and the queue's position in its
2483		 * rq_pos_tree.
2484		 */
2485		if (prev != bfqq->next_rq) {
2486			bfq_updated_next_req(bfqd, bfqq);
2487			/*
2488			 * See comments on bfq_pos_tree_add_move() for
2489			 * the unlikely().
2490			 */
2491			if (unlikely(!bfqd->nonrot_with_queueing))
2492				bfq_pos_tree_add_move(bfqd, bfqq);
2493		}
2494	}
2495}
2496
2497/*
2498 * This function is called to notify the scheduler that the requests
2499 * rq and 'next' have been merged, with 'next' going away.  BFQ
2500 * exploits this hook to address the following issue: if 'next' has a
2501 * fifo_time lower that rq, then the fifo_time of rq must be set to
2502 * the value of 'next', to not forget the greater age of 'next'.
2503 *
2504 * NOTE: in this function we assume that rq is in a bfq_queue, basing
2505 * on that rq is picked from the hash table q->elevator->hash, which,
2506 * in its turn, is filled only with I/O requests present in
2507 * bfq_queues, while BFQ is in use for the request queue q. In fact,
2508 * the function that fills this hash table (elv_rqhash_add) is called
2509 * only by bfq_insert_request.
2510 */
2511static void bfq_requests_merged(struct request_queue *q, struct request *rq,
2512				struct request *next)
2513{
2514	struct bfq_queue *bfqq = RQ_BFQQ(rq),
2515		*next_bfqq = RQ_BFQQ(next);
2516
2517	if (!bfqq)
2518		goto remove;
2519
2520	/*
2521	 * If next and rq belong to the same bfq_queue and next is older
2522	 * than rq, then reposition rq in the fifo (by substituting next
2523	 * with rq). Otherwise, if next and rq belong to different
2524	 * bfq_queues, never reposition rq: in fact, we would have to
2525	 * reposition it with respect to next's position in its own fifo,
2526	 * which would most certainly be too expensive with respect to
2527	 * the benefits.
2528	 */
2529	if (bfqq == next_bfqq &&
2530	    !list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2531	    next->fifo_time < rq->fifo_time) {
2532		list_del_init(&rq->queuelist);
2533		list_replace_init(&next->queuelist, &rq->queuelist);
2534		rq->fifo_time = next->fifo_time;
2535	}
2536
2537	if (bfqq->next_rq == next)
2538		bfqq->next_rq = rq;
2539
2540	bfqg_stats_update_io_merged(bfqq_group(bfqq), next->cmd_flags);
2541remove:
2542	/* Merged request may be in the IO scheduler. Remove it. */
2543	if (!RB_EMPTY_NODE(&next->rb_node)) {
2544		bfq_remove_request(next->q, next);
2545		if (next_bfqq)
2546			bfqg_stats_update_io_remove(bfqq_group(next_bfqq),
2547						    next->cmd_flags);
2548	}
2549}
2550
2551/* Must be called with bfqq != NULL */
2552static void bfq_bfqq_end_wr(struct bfq_queue *bfqq)
2553{
2554	/*
2555	 * If bfqq has been enjoying interactive weight-raising, then
2556	 * reset soft_rt_next_start. We do it for the following
2557	 * reason. bfqq may have been conveying the I/O needed to load
2558	 * a soft real-time application. Such an application actually
2559	 * exhibits a soft real-time I/O pattern after it finishes
2560	 * loading, and finally starts doing its job. But, if bfqq has
2561	 * been receiving a lot of bandwidth so far (likely to happen
2562	 * on a fast device), then soft_rt_next_start now contains a
2563	 * high value that. So, without this reset, bfqq would be
2564	 * prevented from being possibly considered as soft_rt for a
2565	 * very long time.
2566	 */
2567
2568	if (bfqq->wr_cur_max_time !=
2569	    bfqq->bfqd->bfq_wr_rt_max_time)
2570		bfqq->soft_rt_next_start = jiffies;
2571
2572	if (bfq_bfqq_busy(bfqq))
2573		bfqq->bfqd->wr_busy_queues--;
2574	bfqq->wr_coeff = 1;
2575	bfqq->wr_cur_max_time = 0;
2576	bfqq->last_wr_start_finish = jiffies;
2577	/*
2578	 * Trigger a weight change on the next invocation of
2579	 * __bfq_entity_update_weight_prio.
2580	 */
2581	bfqq->entity.prio_changed = 1;
2582}
2583
2584void bfq_end_wr_async_queues(struct bfq_data *bfqd,
2585			     struct bfq_group *bfqg)
2586{
2587	int i, j;
2588
2589	for (i = 0; i < 2; i++)
2590		for (j = 0; j < IOPRIO_NR_LEVELS; j++)
2591			if (bfqg->async_bfqq[i][j])
2592				bfq_bfqq_end_wr(bfqg->async_bfqq[i][j]);
2593	if (bfqg->async_idle_bfqq)
2594		bfq_bfqq_end_wr(bfqg->async_idle_bfqq);
2595}
2596
2597static void bfq_end_wr(struct bfq_data *bfqd)
2598{
2599	struct bfq_queue *bfqq;
2600
2601	spin_lock_irq(&bfqd->lock);
2602
2603	list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list)
2604		bfq_bfqq_end_wr(bfqq);
2605	list_for_each_entry(bfqq, &bfqd->idle_list, bfqq_list)
2606		bfq_bfqq_end_wr(bfqq);
2607	bfq_end_wr_async(bfqd);
2608
2609	spin_unlock_irq(&bfqd->lock);
2610}
2611
2612static sector_t bfq_io_struct_pos(void *io_struct, bool request)
2613{
2614	if (request)
2615		return blk_rq_pos(io_struct);
2616	else
2617		return ((struct bio *)io_struct)->bi_iter.bi_sector;
2618}
2619
2620static int bfq_rq_close_to_sector(void *io_struct, bool request,
2621				  sector_t sector)
2622{
2623	return abs(bfq_io_struct_pos(io_struct, request) - sector) <=
2624	       BFQQ_CLOSE_THR;
2625}
2626
2627static struct bfq_queue *bfqq_find_close(struct bfq_data *bfqd,
2628					 struct bfq_queue *bfqq,
2629					 sector_t sector)
2630{
2631	struct rb_root *root = &bfqq_group(bfqq)->rq_pos_tree;
2632	struct rb_node *parent, *node;
2633	struct bfq_queue *__bfqq;
2634
2635	if (RB_EMPTY_ROOT(root))
2636		return NULL;
2637
2638	/*
2639	 * First, if we find a request starting at the end of the last
2640	 * request, choose it.
2641	 */
2642	__bfqq = bfq_rq_pos_tree_lookup(bfqd, root, sector, &parent, NULL);
2643	if (__bfqq)
2644		return __bfqq;
2645
2646	/*
2647	 * If the exact sector wasn't found, the parent of the NULL leaf
2648	 * will contain the closest sector (rq_pos_tree sorted by
2649	 * next_request position).
2650	 */
2651	__bfqq = rb_entry(parent, struct bfq_queue, pos_node);
2652	if (bfq_rq_close_to_sector(__bfqq->next_rq, true, sector))
2653		return __bfqq;
2654
2655	if (blk_rq_pos(__bfqq->next_rq) < sector)
2656		node = rb_next(&__bfqq->pos_node);
2657	else
2658		node = rb_prev(&__bfqq->pos_node);
2659	if (!node)
2660		return NULL;
2661
2662	__bfqq = rb_entry(node, struct bfq_queue, pos_node);
2663	if (bfq_rq_close_to_sector(__bfqq->next_rq, true, sector))
2664		return __bfqq;
2665
2666	return NULL;
2667}
2668
2669static struct bfq_queue *bfq_find_close_cooperator(struct bfq_data *bfqd,
2670						   struct bfq_queue *cur_bfqq,
2671						   sector_t sector)
2672{
2673	struct bfq_queue *bfqq;
2674
2675	/*
2676	 * We shall notice if some of the queues are cooperating,
2677	 * e.g., working closely on the same area of the device. In
2678	 * that case, we can group them together and: 1) don't waste
2679	 * time idling, and 2) serve the union of their requests in
2680	 * the best possible order for throughput.
2681	 */
2682	bfqq = bfqq_find_close(bfqd, cur_bfqq, sector);
2683	if (!bfqq || bfqq == cur_bfqq)
2684		return NULL;
2685
2686	return bfqq;
2687}
2688
2689static struct bfq_queue *
2690bfq_setup_merge(struct bfq_queue *bfqq, struct bfq_queue *new_bfqq)
2691{
2692	int process_refs, new_process_refs;
2693	struct bfq_queue *__bfqq;
2694
2695	/*
2696	 * If there are no process references on the new_bfqq, then it is
2697	 * unsafe to follow the ->new_bfqq chain as other bfqq's in the chain
2698	 * may have dropped their last reference (not just their last process
2699	 * reference).
2700	 */
2701	if (!bfqq_process_refs(new_bfqq))
2702		return NULL;
2703
2704	/* Avoid a circular list and skip interim queue merges. */
2705	while ((__bfqq = new_bfqq->new_bfqq)) {
2706		if (__bfqq == bfqq)
2707			return NULL;
2708		new_bfqq = __bfqq;
2709	}
2710
2711	process_refs = bfqq_process_refs(bfqq);
2712	new_process_refs = bfqq_process_refs(new_bfqq);
2713	/*
2714	 * If the process for the bfqq has gone away, there is no
2715	 * sense in merging the queues.
2716	 */
2717	if (process_refs == 0 || new_process_refs == 0)
2718		return NULL;
2719
2720	/*
2721	 * Make sure merged queues belong to the same parent. Parents could
2722	 * have changed since the time we decided the two queues are suitable
2723	 * for merging.
2724	 */
2725	if (new_bfqq->entity.parent != bfqq->entity.parent)
2726		return NULL;
2727
2728	bfq_log_bfqq(bfqq->bfqd, bfqq, "scheduling merge with queue %d",
2729		new_bfqq->pid);
2730
2731	/*
2732	 * Merging is just a redirection: the requests of the process
2733	 * owning one of the two queues are redirected to the other queue.
2734	 * The latter queue, in its turn, is set as shared if this is the
2735	 * first time that the requests of some process are redirected to
2736	 * it.
2737	 *
2738	 * We redirect bfqq to new_bfqq and not the opposite, because
2739	 * we are in the context of the process owning bfqq, thus we
2740	 * have the io_cq of this process. So we can immediately
2741	 * configure this io_cq to redirect the requests of the
2742	 * process to new_bfqq. In contrast, the io_cq of new_bfqq is
2743	 * not available any more (new_bfqq->bic == NULL).
2744	 *
2745	 * Anyway, even in case new_bfqq coincides with the in-service
2746	 * queue, redirecting requests the in-service queue is the
2747	 * best option, as we feed the in-service queue with new
2748	 * requests close to the last request served and, by doing so,
2749	 * are likely to increase the throughput.
2750	 */
2751	bfqq->new_bfqq = new_bfqq;
2752	/*
2753	 * The above assignment schedules the following redirections:
2754	 * each time some I/O for bfqq arrives, the process that
2755	 * generated that I/O is disassociated from bfqq and
2756	 * associated with new_bfqq. Here we increases new_bfqq->ref
2757	 * in advance, adding the number of processes that are
2758	 * expected to be associated with new_bfqq as they happen to
2759	 * issue I/O.
2760	 */
2761	new_bfqq->ref += process_refs;
2762	return new_bfqq;
2763}
2764
2765static bool bfq_may_be_close_cooperator(struct bfq_queue *bfqq,
2766					struct bfq_queue *new_bfqq)
2767{
2768	if (bfq_too_late_for_merging(new_bfqq))
2769		return false;
2770
2771	if (bfq_class_idle(bfqq) || bfq_class_idle(new_bfqq) ||
2772	    (bfqq->ioprio_class != new_bfqq->ioprio_class))
2773		return false;
2774
2775	/*
2776	 * If either of the queues has already been detected as seeky,
2777	 * then merging it with the other queue is unlikely to lead to
2778	 * sequential I/O.
2779	 */
2780	if (BFQQ_SEEKY(bfqq) || BFQQ_SEEKY(new_bfqq))
2781		return false;
2782
2783	/*
2784	 * Interleaved I/O is known to be done by (some) applications
2785	 * only for reads, so it does not make sense to merge async
2786	 * queues.
2787	 */
2788	if (!bfq_bfqq_sync(bfqq) || !bfq_bfqq_sync(new_bfqq))
2789		return false;
2790
2791	return true;
2792}
2793
2794static bool idling_boosts_thr_without_issues(struct bfq_data *bfqd,
2795					     struct bfq_queue *bfqq);
2796
2797/*
2798 * Attempt to schedule a merge of bfqq with the currently in-service
2799 * queue or with a close queue among the scheduled queues.  Return
2800 * NULL if no merge was scheduled, a pointer to the shared bfq_queue
2801 * structure otherwise.
2802 *
2803 * The OOM queue is not allowed to participate to cooperation: in fact, since
2804 * the requests temporarily redirected to the OOM queue could be redirected
2805 * again to dedicated queues at any time, the state needed to correctly
2806 * handle merging with the OOM queue would be quite complex and expensive
2807 * to maintain. Besides, in such a critical condition as an out of memory,
2808 * the benefits of queue merging may be little relevant, or even negligible.
2809 *
2810 * WARNING: queue merging may impair fairness among non-weight raised
2811 * queues, for at least two reasons: 1) the original weight of a
2812 * merged queue may change during the merged state, 2) even being the
2813 * weight the same, a merged queue may be bloated with many more
2814 * requests than the ones produced by its originally-associated
2815 * process.
2816 */
2817static struct bfq_queue *
2818bfq_setup_cooperator(struct bfq_data *bfqd, struct bfq_queue *bfqq,
2819		     void *io_struct, bool request, struct bfq_io_cq *bic)
2820{
2821	struct bfq_queue *in_service_bfqq, *new_bfqq;
2822
2823	/* if a merge has already been setup, then proceed with that first */
2824	if (bfqq->new_bfqq)
2825		return bfqq->new_bfqq;
2826
2827	/*
2828	 * Check delayed stable merge for rotational or non-queueing
2829	 * devs. For this branch to be executed, bfqq must not be
2830	 * currently merged with some other queue (i.e., bfqq->bic
2831	 * must be non null). If we considered also merged queues,
2832	 * then we should also check whether bfqq has already been
2833	 * merged with bic->stable_merge_bfqq. But this would be
2834	 * costly and complicated.
2835	 */
2836	if (unlikely(!bfqd->nonrot_with_queueing)) {
2837		/*
2838		 * Make sure also that bfqq is sync, because
2839		 * bic->stable_merge_bfqq may point to some queue (for
2840		 * stable merging) also if bic is associated with a
2841		 * sync queue, but this bfqq is async
2842		 */
2843		if (bfq_bfqq_sync(bfqq) && bic->stable_merge_bfqq &&
2844		    !bfq_bfqq_just_created(bfqq) &&
2845		    time_is_before_jiffies(bfqq->split_time +
2846					  msecs_to_jiffies(bfq_late_stable_merging)) &&
2847		    time_is_before_jiffies(bfqq->creation_time +
2848					   msecs_to_jiffies(bfq_late_stable_merging))) {
2849			struct bfq_queue *stable_merge_bfqq =
2850				bic->stable_merge_bfqq;
2851			int proc_ref = min(bfqq_process_refs(bfqq),
2852					   bfqq_process_refs(stable_merge_bfqq));
2853
2854			/* deschedule stable merge, because done or aborted here */
2855			bfq_put_stable_ref(stable_merge_bfqq);
2856
2857			bic->stable_merge_bfqq = NULL;
2858
2859			if (!idling_boosts_thr_without_issues(bfqd, bfqq) &&
2860			    proc_ref > 0) {
2861				/* next function will take at least one ref */
2862				struct bfq_queue *new_bfqq =
2863					bfq_setup_merge(bfqq, stable_merge_bfqq);
2864
2865				if (new_bfqq) {
2866					bic->stably_merged = true;
2867					if (new_bfqq->bic)
2868						new_bfqq->bic->stably_merged =
2869									true;
2870				}
2871				return new_bfqq;
2872			} else
2873				return NULL;
2874		}
2875	}
2876
2877	/*
2878	 * Do not perform queue merging if the device is non
2879	 * rotational and performs internal queueing. In fact, such a
2880	 * device reaches a high speed through internal parallelism
2881	 * and pipelining. This means that, to reach a high
2882	 * throughput, it must have many requests enqueued at the same
2883	 * time. But, in this configuration, the internal scheduling
2884	 * algorithm of the device does exactly the job of queue
2885	 * merging: it reorders requests so as to obtain as much as
2886	 * possible a sequential I/O pattern. As a consequence, with
2887	 * the workload generated by processes doing interleaved I/O,
2888	 * the throughput reached by the device is likely to be the
2889	 * same, with and without queue merging.
2890	 *
2891	 * Disabling merging also provides a remarkable benefit in
2892	 * terms of throughput. Merging tends to make many workloads
2893	 * artificially more uneven, because of shared queues
2894	 * remaining non empty for incomparably more time than
2895	 * non-merged queues. This may accentuate workload
2896	 * asymmetries. For example, if one of the queues in a set of
2897	 * merged queues has a higher weight than a normal queue, then
2898	 * the shared queue may inherit such a high weight and, by
2899	 * staying almost always active, may force BFQ to perform I/O
2900	 * plugging most of the time. This evidently makes it harder
2901	 * for BFQ to let the device reach a high throughput.
2902	 *
2903	 * Finally, the likely() macro below is not used because one
2904	 * of the two branches is more likely than the other, but to
2905	 * have the code path after the following if() executed as
2906	 * fast as possible for the case of a non rotational device
2907	 * with queueing. We want it because this is the fastest kind
2908	 * of device. On the opposite end, the likely() may lengthen
2909	 * the execution time of BFQ for the case of slower devices
2910	 * (rotational or at least without queueing). But in this case
2911	 * the execution time of BFQ matters very little, if not at
2912	 * all.
2913	 */
2914	if (likely(bfqd->nonrot_with_queueing))
2915		return NULL;
2916
2917	/*
2918	 * Prevent bfqq from being merged if it has been created too
2919	 * long ago. The idea is that true cooperating processes, and
2920	 * thus their associated bfq_queues, are supposed to be
2921	 * created shortly after each other. This is the case, e.g.,
2922	 * for KVM/QEMU and dump I/O threads. Basing on this
2923	 * assumption, the following filtering greatly reduces the
2924	 * probability that two non-cooperating processes, which just
2925	 * happen to do close I/O for some short time interval, have
2926	 * their queues merged by mistake.
2927	 */
2928	if (bfq_too_late_for_merging(bfqq))
2929		return NULL;
2930
 
 
 
2931	if (!io_struct || unlikely(bfqq == &bfqd->oom_bfqq))
2932		return NULL;
2933
2934	/* If there is only one backlogged queue, don't search. */
2935	if (bfq_tot_busy_queues(bfqd) == 1)
2936		return NULL;
2937
2938	in_service_bfqq = bfqd->in_service_queue;
2939
2940	if (in_service_bfqq && in_service_bfqq != bfqq &&
2941	    likely(in_service_bfqq != &bfqd->oom_bfqq) &&
2942	    bfq_rq_close_to_sector(io_struct, request,
2943				   bfqd->in_serv_last_pos) &&
2944	    bfqq->entity.parent == in_service_bfqq->entity.parent &&
2945	    bfq_may_be_close_cooperator(bfqq, in_service_bfqq)) {
2946		new_bfqq = bfq_setup_merge(bfqq, in_service_bfqq);
2947		if (new_bfqq)
2948			return new_bfqq;
2949	}
2950	/*
2951	 * Check whether there is a cooperator among currently scheduled
2952	 * queues. The only thing we need is that the bio/request is not
2953	 * NULL, as we need it to establish whether a cooperator exists.
2954	 */
2955	new_bfqq = bfq_find_close_cooperator(bfqd, bfqq,
2956			bfq_io_struct_pos(io_struct, request));
2957
2958	if (new_bfqq && likely(new_bfqq != &bfqd->oom_bfqq) &&
2959	    bfq_may_be_close_cooperator(bfqq, new_bfqq))
2960		return bfq_setup_merge(bfqq, new_bfqq);
2961
2962	return NULL;
2963}
2964
2965static void bfq_bfqq_save_state(struct bfq_queue *bfqq)
2966{
2967	struct bfq_io_cq *bic = bfqq->bic;
2968
2969	/*
2970	 * If !bfqq->bic, the queue is already shared or its requests
2971	 * have already been redirected to a shared queue; both idle window
2972	 * and weight raising state have already been saved. Do nothing.
2973	 */
2974	if (!bic)
2975		return;
2976
2977	bic->saved_last_serv_time_ns = bfqq->last_serv_time_ns;
2978	bic->saved_inject_limit = bfqq->inject_limit;
2979	bic->saved_decrease_time_jif = bfqq->decrease_time_jif;
2980
2981	bic->saved_weight = bfqq->entity.orig_weight;
2982	bic->saved_ttime = bfqq->ttime;
2983	bic->saved_has_short_ttime = bfq_bfqq_has_short_ttime(bfqq);
2984	bic->saved_IO_bound = bfq_bfqq_IO_bound(bfqq);
2985	bic->saved_io_start_time = bfqq->io_start_time;
2986	bic->saved_tot_idle_time = bfqq->tot_idle_time;
2987	bic->saved_in_large_burst = bfq_bfqq_in_large_burst(bfqq);
2988	bic->was_in_burst_list = !hlist_unhashed(&bfqq->burst_list_node);
2989	if (unlikely(bfq_bfqq_just_created(bfqq) &&
2990		     !bfq_bfqq_in_large_burst(bfqq) &&
2991		     bfqq->bfqd->low_latency)) {
2992		/*
2993		 * bfqq being merged right after being created: bfqq
2994		 * would have deserved interactive weight raising, but
2995		 * did not make it to be set in a weight-raised state,
2996		 * because of this early merge.	Store directly the
2997		 * weight-raising state that would have been assigned
2998		 * to bfqq, so that to avoid that bfqq unjustly fails
2999		 * to enjoy weight raising if split soon.
3000		 */
3001		bic->saved_wr_coeff = bfqq->bfqd->bfq_wr_coeff;
3002		bic->saved_wr_start_at_switch_to_srt = bfq_smallest_from_now();
3003		bic->saved_wr_cur_max_time = bfq_wr_duration(bfqq->bfqd);
3004		bic->saved_last_wr_start_finish = jiffies;
3005	} else {
3006		bic->saved_wr_coeff = bfqq->wr_coeff;
3007		bic->saved_wr_start_at_switch_to_srt =
3008			bfqq->wr_start_at_switch_to_srt;
3009		bic->saved_service_from_wr = bfqq->service_from_wr;
3010		bic->saved_last_wr_start_finish = bfqq->last_wr_start_finish;
3011		bic->saved_wr_cur_max_time = bfqq->wr_cur_max_time;
3012	}
3013}
3014
3015
3016static void
3017bfq_reassign_last_bfqq(struct bfq_queue *cur_bfqq, struct bfq_queue *new_bfqq)
3018{
3019	if (cur_bfqq->entity.parent &&
3020	    cur_bfqq->entity.parent->last_bfqq_created == cur_bfqq)
3021		cur_bfqq->entity.parent->last_bfqq_created = new_bfqq;
3022	else if (cur_bfqq->bfqd && cur_bfqq->bfqd->last_bfqq_created == cur_bfqq)
3023		cur_bfqq->bfqd->last_bfqq_created = new_bfqq;
3024}
3025
3026void bfq_release_process_ref(struct bfq_data *bfqd, struct bfq_queue *bfqq)
3027{
3028	/*
3029	 * To prevent bfqq's service guarantees from being violated,
3030	 * bfqq may be left busy, i.e., queued for service, even if
3031	 * empty (see comments in __bfq_bfqq_expire() for
3032	 * details). But, if no process will send requests to bfqq any
3033	 * longer, then there is no point in keeping bfqq queued for
3034	 * service. In addition, keeping bfqq queued for service, but
3035	 * with no process ref any longer, may have caused bfqq to be
3036	 * freed when dequeued from service. But this is assumed to
3037	 * never happen.
3038	 */
3039	if (bfq_bfqq_busy(bfqq) && RB_EMPTY_ROOT(&bfqq->sort_list) &&
3040	    bfqq != bfqd->in_service_queue)
3041		bfq_del_bfqq_busy(bfqq, false);
3042
3043	bfq_reassign_last_bfqq(bfqq, NULL);
3044
3045	bfq_put_queue(bfqq);
3046}
3047
3048static void
3049bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic,
3050		struct bfq_queue *bfqq, struct bfq_queue *new_bfqq)
3051{
3052	bfq_log_bfqq(bfqd, bfqq, "merging with queue %lu",
3053		(unsigned long)new_bfqq->pid);
3054	/* Save weight raising and idle window of the merged queues */
3055	bfq_bfqq_save_state(bfqq);
3056	bfq_bfqq_save_state(new_bfqq);
3057	if (bfq_bfqq_IO_bound(bfqq))
3058		bfq_mark_bfqq_IO_bound(new_bfqq);
3059	bfq_clear_bfqq_IO_bound(bfqq);
3060
3061	/*
3062	 * The processes associated with bfqq are cooperators of the
3063	 * processes associated with new_bfqq. So, if bfqq has a
3064	 * waker, then assume that all these processes will be happy
3065	 * to let bfqq's waker freely inject I/O when they have no
3066	 * I/O.
3067	 */
3068	if (bfqq->waker_bfqq && !new_bfqq->waker_bfqq &&
3069	    bfqq->waker_bfqq != new_bfqq) {
3070		new_bfqq->waker_bfqq = bfqq->waker_bfqq;
3071		new_bfqq->tentative_waker_bfqq = NULL;
3072
3073		/*
3074		 * If the waker queue disappears, then
3075		 * new_bfqq->waker_bfqq must be reset. So insert
3076		 * new_bfqq into the woken_list of the waker. See
3077		 * bfq_check_waker for details.
3078		 */
3079		hlist_add_head(&new_bfqq->woken_list_node,
3080			       &new_bfqq->waker_bfqq->woken_list);
3081
3082	}
3083
3084	/*
3085	 * If bfqq is weight-raised, then let new_bfqq inherit
3086	 * weight-raising. To reduce false positives, neglect the case
3087	 * where bfqq has just been created, but has not yet made it
3088	 * to be weight-raised (which may happen because EQM may merge
3089	 * bfqq even before bfq_add_request is executed for the first
3090	 * time for bfqq). Handling this case would however be very
3091	 * easy, thanks to the flag just_created.
3092	 */
3093	if (new_bfqq->wr_coeff == 1 && bfqq->wr_coeff > 1) {
3094		new_bfqq->wr_coeff = bfqq->wr_coeff;
3095		new_bfqq->wr_cur_max_time = bfqq->wr_cur_max_time;
3096		new_bfqq->last_wr_start_finish = bfqq->last_wr_start_finish;
3097		new_bfqq->wr_start_at_switch_to_srt =
3098			bfqq->wr_start_at_switch_to_srt;
3099		if (bfq_bfqq_busy(new_bfqq))
3100			bfqd->wr_busy_queues++;
3101		new_bfqq->entity.prio_changed = 1;
3102	}
3103
3104	if (bfqq->wr_coeff > 1) { /* bfqq has given its wr to new_bfqq */
3105		bfqq->wr_coeff = 1;
3106		bfqq->entity.prio_changed = 1;
3107		if (bfq_bfqq_busy(bfqq))
3108			bfqd->wr_busy_queues--;
3109	}
3110
3111	bfq_log_bfqq(bfqd, new_bfqq, "merge_bfqqs: wr_busy %d",
3112		     bfqd->wr_busy_queues);
3113
3114	/*
3115	 * Merge queues (that is, let bic redirect its requests to new_bfqq)
3116	 */
3117	bic_set_bfqq(bic, new_bfqq, true);
3118	bfq_mark_bfqq_coop(new_bfqq);
3119	/*
3120	 * new_bfqq now belongs to at least two bics (it is a shared queue):
3121	 * set new_bfqq->bic to NULL. bfqq either:
3122	 * - does not belong to any bic any more, and hence bfqq->bic must
3123	 *   be set to NULL, or
3124	 * - is a queue whose owning bics have already been redirected to a
3125	 *   different queue, hence the queue is destined to not belong to
3126	 *   any bic soon and bfqq->bic is already NULL (therefore the next
3127	 *   assignment causes no harm).
3128	 */
3129	new_bfqq->bic = NULL;
3130	/*
3131	 * If the queue is shared, the pid is the pid of one of the associated
3132	 * processes. Which pid depends on the exact sequence of merge events
3133	 * the queue underwent. So printing such a pid is useless and confusing
3134	 * because it reports a random pid between those of the associated
3135	 * processes.
3136	 * We mark such a queue with a pid -1, and then print SHARED instead of
3137	 * a pid in logging messages.
3138	 */
3139	new_bfqq->pid = -1;
3140	bfqq->bic = NULL;
3141
3142	bfq_reassign_last_bfqq(bfqq, new_bfqq);
3143
3144	bfq_release_process_ref(bfqd, bfqq);
3145}
3146
3147static bool bfq_allow_bio_merge(struct request_queue *q, struct request *rq,
3148				struct bio *bio)
3149{
3150	struct bfq_data *bfqd = q->elevator->elevator_data;
3151	bool is_sync = op_is_sync(bio->bi_opf);
3152	struct bfq_queue *bfqq = bfqd->bio_bfqq, *new_bfqq;
3153
3154	/*
3155	 * Disallow merge of a sync bio into an async request.
3156	 */
3157	if (is_sync && !rq_is_sync(rq))
3158		return false;
3159
3160	/*
3161	 * Lookup the bfqq that this bio will be queued with. Allow
3162	 * merge only if rq is queued there.
3163	 */
3164	if (!bfqq)
3165		return false;
3166
3167	/*
3168	 * We take advantage of this function to perform an early merge
3169	 * of the queues of possible cooperating processes.
3170	 */
3171	new_bfqq = bfq_setup_cooperator(bfqd, bfqq, bio, false, bfqd->bio_bic);
3172	if (new_bfqq) {
3173		/*
3174		 * bic still points to bfqq, then it has not yet been
3175		 * redirected to some other bfq_queue, and a queue
3176		 * merge between bfqq and new_bfqq can be safely
3177		 * fulfilled, i.e., bic can be redirected to new_bfqq
3178		 * and bfqq can be put.
3179		 */
3180		bfq_merge_bfqqs(bfqd, bfqd->bio_bic, bfqq,
3181				new_bfqq);
3182		/*
3183		 * If we get here, bio will be queued into new_queue,
3184		 * so use new_bfqq to decide whether bio and rq can be
3185		 * merged.
3186		 */
3187		bfqq = new_bfqq;
3188
3189		/*
3190		 * Change also bqfd->bio_bfqq, as
3191		 * bfqd->bio_bic now points to new_bfqq, and
3192		 * this function may be invoked again (and then may
3193		 * use again bqfd->bio_bfqq).
3194		 */
3195		bfqd->bio_bfqq = bfqq;
3196	}
3197
3198	return bfqq == RQ_BFQQ(rq);
3199}
3200
3201/*
3202 * Set the maximum time for the in-service queue to consume its
3203 * budget. This prevents seeky processes from lowering the throughput.
3204 * In practice, a time-slice service scheme is used with seeky
3205 * processes.
3206 */
3207static void bfq_set_budget_timeout(struct bfq_data *bfqd,
3208				   struct bfq_queue *bfqq)
3209{
3210	unsigned int timeout_coeff;
3211
3212	if (bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time)
3213		timeout_coeff = 1;
3214	else
3215		timeout_coeff = bfqq->entity.weight / bfqq->entity.orig_weight;
3216
3217	bfqd->last_budget_start = ktime_get();
3218
3219	bfqq->budget_timeout = jiffies +
3220		bfqd->bfq_timeout * timeout_coeff;
3221}
3222
3223static void __bfq_set_in_service_queue(struct bfq_data *bfqd,
3224				       struct bfq_queue *bfqq)
3225{
3226	if (bfqq) {
3227		bfq_clear_bfqq_fifo_expire(bfqq);
3228
3229		bfqd->budgets_assigned = (bfqd->budgets_assigned * 7 + 256) / 8;
3230
3231		if (time_is_before_jiffies(bfqq->last_wr_start_finish) &&
3232		    bfqq->wr_coeff > 1 &&
3233		    bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
3234		    time_is_before_jiffies(bfqq->budget_timeout)) {
3235			/*
3236			 * For soft real-time queues, move the start
3237			 * of the weight-raising period forward by the
3238			 * time the queue has not received any
3239			 * service. Otherwise, a relatively long
3240			 * service delay is likely to cause the
3241			 * weight-raising period of the queue to end,
3242			 * because of the short duration of the
3243			 * weight-raising period of a soft real-time
3244			 * queue.  It is worth noting that this move
3245			 * is not so dangerous for the other queues,
3246			 * because soft real-time queues are not
3247			 * greedy.
3248			 *
3249			 * To not add a further variable, we use the
3250			 * overloaded field budget_timeout to
3251			 * determine for how long the queue has not
3252			 * received service, i.e., how much time has
3253			 * elapsed since the queue expired. However,
3254			 * this is a little imprecise, because
3255			 * budget_timeout is set to jiffies if bfqq
3256			 * not only expires, but also remains with no
3257			 * request.
3258			 */
3259			if (time_after(bfqq->budget_timeout,
3260				       bfqq->last_wr_start_finish))
3261				bfqq->last_wr_start_finish +=
3262					jiffies - bfqq->budget_timeout;
3263			else
3264				bfqq->last_wr_start_finish = jiffies;
3265		}
3266
3267		bfq_set_budget_timeout(bfqd, bfqq);
3268		bfq_log_bfqq(bfqd, bfqq,
3269			     "set_in_service_queue, cur-budget = %d",
3270			     bfqq->entity.budget);
3271	}
3272
3273	bfqd->in_service_queue = bfqq;
3274	bfqd->in_serv_last_pos = 0;
3275}
3276
3277/*
3278 * Get and set a new queue for service.
3279 */
3280static struct bfq_queue *bfq_set_in_service_queue(struct bfq_data *bfqd)
3281{
3282	struct bfq_queue *bfqq = bfq_get_next_queue(bfqd);
3283
3284	__bfq_set_in_service_queue(bfqd, bfqq);
3285	return bfqq;
3286}
3287
3288static void bfq_arm_slice_timer(struct bfq_data *bfqd)
3289{
3290	struct bfq_queue *bfqq = bfqd->in_service_queue;
3291	u32 sl;
3292
3293	bfq_mark_bfqq_wait_request(bfqq);
3294
3295	/*
3296	 * We don't want to idle for seeks, but we do want to allow
3297	 * fair distribution of slice time for a process doing back-to-back
3298	 * seeks. So allow a little bit of time for him to submit a new rq.
3299	 */
3300	sl = bfqd->bfq_slice_idle;
3301	/*
3302	 * Unless the queue is being weight-raised or the scenario is
3303	 * asymmetric, grant only minimum idle time if the queue
3304	 * is seeky. A long idling is preserved for a weight-raised
3305	 * queue, or, more in general, in an asymmetric scenario,
3306	 * because a long idling is needed for guaranteeing to a queue
3307	 * its reserved share of the throughput (in particular, it is
3308	 * needed if the queue has a higher weight than some other
3309	 * queue).
3310	 */
3311	if (BFQQ_SEEKY(bfqq) && bfqq->wr_coeff == 1 &&
3312	    !bfq_asymmetric_scenario(bfqd, bfqq))
3313		sl = min_t(u64, sl, BFQ_MIN_TT);
3314	else if (bfqq->wr_coeff > 1)
3315		sl = max_t(u32, sl, 20ULL * NSEC_PER_MSEC);
3316
3317	bfqd->last_idling_start = ktime_get();
3318	bfqd->last_idling_start_jiffies = jiffies;
3319
3320	hrtimer_start(&bfqd->idle_slice_timer, ns_to_ktime(sl),
3321		      HRTIMER_MODE_REL);
3322	bfqg_stats_set_start_idle_time(bfqq_group(bfqq));
3323}
3324
3325/*
3326 * In autotuning mode, max_budget is dynamically recomputed as the
3327 * amount of sectors transferred in timeout at the estimated peak
3328 * rate. This enables BFQ to utilize a full timeslice with a full
3329 * budget, even if the in-service queue is served at peak rate. And
3330 * this maximises throughput with sequential workloads.
3331 */
3332static unsigned long bfq_calc_max_budget(struct bfq_data *bfqd)
3333{
3334	return (u64)bfqd->peak_rate * USEC_PER_MSEC *
3335		jiffies_to_msecs(bfqd->bfq_timeout)>>BFQ_RATE_SHIFT;
3336}
3337
3338/*
3339 * Update parameters related to throughput and responsiveness, as a
3340 * function of the estimated peak rate. See comments on
3341 * bfq_calc_max_budget(), and on the ref_wr_duration array.
3342 */
3343static void update_thr_responsiveness_params(struct bfq_data *bfqd)
3344{
3345	if (bfqd->bfq_user_max_budget == 0) {
3346		bfqd->bfq_max_budget =
3347			bfq_calc_max_budget(bfqd);
3348		bfq_log(bfqd, "new max_budget = %d", bfqd->bfq_max_budget);
3349	}
3350}
3351
3352static void bfq_reset_rate_computation(struct bfq_data *bfqd,
3353				       struct request *rq)
3354{
3355	if (rq != NULL) { /* new rq dispatch now, reset accordingly */
3356		bfqd->last_dispatch = bfqd->first_dispatch = ktime_get_ns();
3357		bfqd->peak_rate_samples = 1;
3358		bfqd->sequential_samples = 0;
3359		bfqd->tot_sectors_dispatched = bfqd->last_rq_max_size =
3360			blk_rq_sectors(rq);
3361	} else /* no new rq dispatched, just reset the number of samples */
3362		bfqd->peak_rate_samples = 0; /* full re-init on next disp. */
3363
3364	bfq_log(bfqd,
3365		"reset_rate_computation at end, sample %u/%u tot_sects %llu",
3366		bfqd->peak_rate_samples, bfqd->sequential_samples,
3367		bfqd->tot_sectors_dispatched);
3368}
3369
3370static void bfq_update_rate_reset(struct bfq_data *bfqd, struct request *rq)
3371{
3372	u32 rate, weight, divisor;
3373
3374	/*
3375	 * For the convergence property to hold (see comments on
3376	 * bfq_update_peak_rate()) and for the assessment to be
3377	 * reliable, a minimum number of samples must be present, and
3378	 * a minimum amount of time must have elapsed. If not so, do
3379	 * not compute new rate. Just reset parameters, to get ready
3380	 * for a new evaluation attempt.
3381	 */
3382	if (bfqd->peak_rate_samples < BFQ_RATE_MIN_SAMPLES ||
3383	    bfqd->delta_from_first < BFQ_RATE_MIN_INTERVAL)
3384		goto reset_computation;
3385
3386	/*
3387	 * If a new request completion has occurred after last
3388	 * dispatch, then, to approximate the rate at which requests
3389	 * have been served by the device, it is more precise to
3390	 * extend the observation interval to the last completion.
3391	 */
3392	bfqd->delta_from_first =
3393		max_t(u64, bfqd->delta_from_first,
3394		      bfqd->last_completion - bfqd->first_dispatch);
3395
3396	/*
3397	 * Rate computed in sects/usec, and not sects/nsec, for
3398	 * precision issues.
3399	 */
3400	rate = div64_ul(bfqd->tot_sectors_dispatched<<BFQ_RATE_SHIFT,
3401			div_u64(bfqd->delta_from_first, NSEC_PER_USEC));
3402
3403	/*
3404	 * Peak rate not updated if:
3405	 * - the percentage of sequential dispatches is below 3/4 of the
3406	 *   total, and rate is below the current estimated peak rate
3407	 * - rate is unreasonably high (> 20M sectors/sec)
3408	 */
3409	if ((bfqd->sequential_samples < (3 * bfqd->peak_rate_samples)>>2 &&
3410	     rate <= bfqd->peak_rate) ||
3411		rate > 20<<BFQ_RATE_SHIFT)
3412		goto reset_computation;
3413
3414	/*
3415	 * We have to update the peak rate, at last! To this purpose,
3416	 * we use a low-pass filter. We compute the smoothing constant
3417	 * of the filter as a function of the 'weight' of the new
3418	 * measured rate.
3419	 *
3420	 * As can be seen in next formulas, we define this weight as a
3421	 * quantity proportional to how sequential the workload is,
3422	 * and to how long the observation time interval is.
3423	 *
3424	 * The weight runs from 0 to 8. The maximum value of the
3425	 * weight, 8, yields the minimum value for the smoothing
3426	 * constant. At this minimum value for the smoothing constant,
3427	 * the measured rate contributes for half of the next value of
3428	 * the estimated peak rate.
3429	 *
3430	 * So, the first step is to compute the weight as a function
3431	 * of how sequential the workload is. Note that the weight
3432	 * cannot reach 9, because bfqd->sequential_samples cannot
3433	 * become equal to bfqd->peak_rate_samples, which, in its
3434	 * turn, holds true because bfqd->sequential_samples is not
3435	 * incremented for the first sample.
3436	 */
3437	weight = (9 * bfqd->sequential_samples) / bfqd->peak_rate_samples;
3438
3439	/*
3440	 * Second step: further refine the weight as a function of the
3441	 * duration of the observation interval.
3442	 */
3443	weight = min_t(u32, 8,
3444		       div_u64(weight * bfqd->delta_from_first,
3445			       BFQ_RATE_REF_INTERVAL));
3446
3447	/*
3448	 * Divisor ranging from 10, for minimum weight, to 2, for
3449	 * maximum weight.
3450	 */
3451	divisor = 10 - weight;
3452
3453	/*
3454	 * Finally, update peak rate:
3455	 *
3456	 * peak_rate = peak_rate * (divisor-1) / divisor  +  rate / divisor
3457	 */
3458	bfqd->peak_rate *= divisor-1;
3459	bfqd->peak_rate /= divisor;
3460	rate /= divisor; /* smoothing constant alpha = 1/divisor */
3461
3462	bfqd->peak_rate += rate;
3463
3464	/*
3465	 * For a very slow device, bfqd->peak_rate can reach 0 (see
3466	 * the minimum representable values reported in the comments
3467	 * on BFQ_RATE_SHIFT). Push to 1 if this happens, to avoid
3468	 * divisions by zero where bfqd->peak_rate is used as a
3469	 * divisor.
3470	 */
3471	bfqd->peak_rate = max_t(u32, 1, bfqd->peak_rate);
3472
3473	update_thr_responsiveness_params(bfqd);
3474
3475reset_computation:
3476	bfq_reset_rate_computation(bfqd, rq);
3477}
3478
3479/*
3480 * Update the read/write peak rate (the main quantity used for
3481 * auto-tuning, see update_thr_responsiveness_params()).
3482 *
3483 * It is not trivial to estimate the peak rate (correctly): because of
3484 * the presence of sw and hw queues between the scheduler and the
3485 * device components that finally serve I/O requests, it is hard to
3486 * say exactly when a given dispatched request is served inside the
3487 * device, and for how long. As a consequence, it is hard to know
3488 * precisely at what rate a given set of requests is actually served
3489 * by the device.
3490 *
3491 * On the opposite end, the dispatch time of any request is trivially
3492 * available, and, from this piece of information, the "dispatch rate"
3493 * of requests can be immediately computed. So, the idea in the next
3494 * function is to use what is known, namely request dispatch times
3495 * (plus, when useful, request completion times), to estimate what is
3496 * unknown, namely in-device request service rate.
3497 *
3498 * The main issue is that, because of the above facts, the rate at
3499 * which a certain set of requests is dispatched over a certain time
3500 * interval can vary greatly with respect to the rate at which the
3501 * same requests are then served. But, since the size of any
3502 * intermediate queue is limited, and the service scheme is lossless
3503 * (no request is silently dropped), the following obvious convergence
3504 * property holds: the number of requests dispatched MUST become
3505 * closer and closer to the number of requests completed as the
3506 * observation interval grows. This is the key property used in
3507 * the next function to estimate the peak service rate as a function
3508 * of the observed dispatch rate. The function assumes to be invoked
3509 * on every request dispatch.
3510 */
3511static void bfq_update_peak_rate(struct bfq_data *bfqd, struct request *rq)
3512{
3513	u64 now_ns = ktime_get_ns();
3514
3515	if (bfqd->peak_rate_samples == 0) { /* first dispatch */
3516		bfq_log(bfqd, "update_peak_rate: goto reset, samples %d",
3517			bfqd->peak_rate_samples);
3518		bfq_reset_rate_computation(bfqd, rq);
3519		goto update_last_values; /* will add one sample */
3520	}
3521
3522	/*
3523	 * Device idle for very long: the observation interval lasting
3524	 * up to this dispatch cannot be a valid observation interval
3525	 * for computing a new peak rate (similarly to the late-
3526	 * completion event in bfq_completed_request()). Go to
3527	 * update_rate_and_reset to have the following three steps
3528	 * taken:
3529	 * - close the observation interval at the last (previous)
3530	 *   request dispatch or completion
3531	 * - compute rate, if possible, for that observation interval
3532	 * - start a new observation interval with this dispatch
3533	 */
3534	if (now_ns - bfqd->last_dispatch > 100*NSEC_PER_MSEC &&
3535	    bfqd->rq_in_driver == 0)
3536		goto update_rate_and_reset;
3537
3538	/* Update sampling information */
3539	bfqd->peak_rate_samples++;
3540
3541	if ((bfqd->rq_in_driver > 0 ||
3542		now_ns - bfqd->last_completion < BFQ_MIN_TT)
3543	    && !BFQ_RQ_SEEKY(bfqd, bfqd->last_position, rq))
3544		bfqd->sequential_samples++;
3545
3546	bfqd->tot_sectors_dispatched += blk_rq_sectors(rq);
3547
3548	/* Reset max observed rq size every 32 dispatches */
3549	if (likely(bfqd->peak_rate_samples % 32))
3550		bfqd->last_rq_max_size =
3551			max_t(u32, blk_rq_sectors(rq), bfqd->last_rq_max_size);
3552	else
3553		bfqd->last_rq_max_size = blk_rq_sectors(rq);
3554
3555	bfqd->delta_from_first = now_ns - bfqd->first_dispatch;
3556
3557	/* Target observation interval not yet reached, go on sampling */
3558	if (bfqd->delta_from_first < BFQ_RATE_REF_INTERVAL)
3559		goto update_last_values;
3560
3561update_rate_and_reset:
3562	bfq_update_rate_reset(bfqd, rq);
3563update_last_values:
3564	bfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
3565	if (RQ_BFQQ(rq) == bfqd->in_service_queue)
3566		bfqd->in_serv_last_pos = bfqd->last_position;
3567	bfqd->last_dispatch = now_ns;
3568}
3569
3570/*
3571 * Remove request from internal lists.
3572 */
3573static void bfq_dispatch_remove(struct request_queue *q, struct request *rq)
3574{
3575	struct bfq_queue *bfqq = RQ_BFQQ(rq);
3576
3577	/*
3578	 * For consistency, the next instruction should have been
3579	 * executed after removing the request from the queue and
3580	 * dispatching it.  We execute instead this instruction before
3581	 * bfq_remove_request() (and hence introduce a temporary
3582	 * inconsistency), for efficiency.  In fact, should this
3583	 * dispatch occur for a non in-service bfqq, this anticipated
3584	 * increment prevents two counters related to bfqq->dispatched
3585	 * from risking to be, first, uselessly decremented, and then
3586	 * incremented again when the (new) value of bfqq->dispatched
3587	 * happens to be taken into account.
3588	 */
3589	bfqq->dispatched++;
3590	bfq_update_peak_rate(q->elevator->elevator_data, rq);
3591
3592	bfq_remove_request(q, rq);
3593}
3594
3595/*
3596 * There is a case where idling does not have to be performed for
3597 * throughput concerns, but to preserve the throughput share of
3598 * the process associated with bfqq.
3599 *
3600 * To introduce this case, we can note that allowing the drive
3601 * to enqueue more than one request at a time, and hence
3602 * delegating de facto final scheduling decisions to the
3603 * drive's internal scheduler, entails loss of control on the
3604 * actual request service order. In particular, the critical
3605 * situation is when requests from different processes happen
3606 * to be present, at the same time, in the internal queue(s)
3607 * of the drive. In such a situation, the drive, by deciding
3608 * the service order of the internally-queued requests, does
3609 * determine also the actual throughput distribution among
3610 * these processes. But the drive typically has no notion or
3611 * concern about per-process throughput distribution, and
3612 * makes its decisions only on a per-request basis. Therefore,
3613 * the service distribution enforced by the drive's internal
3614 * scheduler is likely to coincide with the desired throughput
3615 * distribution only in a completely symmetric, or favorably
3616 * skewed scenario where:
3617 * (i-a) each of these processes must get the same throughput as
3618 *	 the others,
3619 * (i-b) in case (i-a) does not hold, it holds that the process
3620 *       associated with bfqq must receive a lower or equal
3621 *	 throughput than any of the other processes;
3622 * (ii)  the I/O of each process has the same properties, in
3623 *       terms of locality (sequential or random), direction
3624 *       (reads or writes), request sizes, greediness
3625 *       (from I/O-bound to sporadic), and so on;
3626
3627 * In fact, in such a scenario, the drive tends to treat the requests
3628 * of each process in about the same way as the requests of the
3629 * others, and thus to provide each of these processes with about the
3630 * same throughput.  This is exactly the desired throughput
3631 * distribution if (i-a) holds, or, if (i-b) holds instead, this is an
3632 * even more convenient distribution for (the process associated with)
3633 * bfqq.
3634 *
3635 * In contrast, in any asymmetric or unfavorable scenario, device
3636 * idling (I/O-dispatch plugging) is certainly needed to guarantee
3637 * that bfqq receives its assigned fraction of the device throughput
3638 * (see [1] for details).
3639 *
3640 * The problem is that idling may significantly reduce throughput with
3641 * certain combinations of types of I/O and devices. An important
3642 * example is sync random I/O on flash storage with command
3643 * queueing. So, unless bfqq falls in cases where idling also boosts
3644 * throughput, it is important to check conditions (i-a), i(-b) and
3645 * (ii) accurately, so as to avoid idling when not strictly needed for
3646 * service guarantees.
3647 *
3648 * Unfortunately, it is extremely difficult to thoroughly check
3649 * condition (ii). And, in case there are active groups, it becomes
3650 * very difficult to check conditions (i-a) and (i-b) too.  In fact,
3651 * if there are active groups, then, for conditions (i-a) or (i-b) to
3652 * become false 'indirectly', it is enough that an active group
3653 * contains more active processes or sub-groups than some other active
3654 * group. More precisely, for conditions (i-a) or (i-b) to become
3655 * false because of such a group, it is not even necessary that the
3656 * group is (still) active: it is sufficient that, even if the group
3657 * has become inactive, some of its descendant processes still have
3658 * some request already dispatched but still waiting for
3659 * completion. In fact, requests have still to be guaranteed their
3660 * share of the throughput even after being dispatched. In this
3661 * respect, it is easy to show that, if a group frequently becomes
3662 * inactive while still having in-flight requests, and if, when this
3663 * happens, the group is not considered in the calculation of whether
3664 * the scenario is asymmetric, then the group may fail to be
3665 * guaranteed its fair share of the throughput (basically because
3666 * idling may not be performed for the descendant processes of the
3667 * group, but it had to be).  We address this issue with the following
3668 * bi-modal behavior, implemented in the function
3669 * bfq_asymmetric_scenario().
3670 *
3671 * If there are groups with requests waiting for completion
3672 * (as commented above, some of these groups may even be
3673 * already inactive), then the scenario is tagged as
3674 * asymmetric, conservatively, without checking any of the
3675 * conditions (i-a), (i-b) or (ii). So the device is idled for bfqq.
3676 * This behavior matches also the fact that groups are created
3677 * exactly if controlling I/O is a primary concern (to
3678 * preserve bandwidth and latency guarantees).
3679 *
3680 * On the opposite end, if there are no groups with requests waiting
3681 * for completion, then only conditions (i-a) and (i-b) are actually
3682 * controlled, i.e., provided that conditions (i-a) or (i-b) holds,
3683 * idling is not performed, regardless of whether condition (ii)
3684 * holds.  In other words, only if conditions (i-a) and (i-b) do not
3685 * hold, then idling is allowed, and the device tends to be prevented
3686 * from queueing many requests, possibly of several processes. Since
3687 * there are no groups with requests waiting for completion, then, to
3688 * control conditions (i-a) and (i-b) it is enough to check just
3689 * whether all the queues with requests waiting for completion also
3690 * have the same weight.
3691 *
3692 * Not checking condition (ii) evidently exposes bfqq to the
3693 * risk of getting less throughput than its fair share.
3694 * However, for queues with the same weight, a further
3695 * mechanism, preemption, mitigates or even eliminates this
3696 * problem. And it does so without consequences on overall
3697 * throughput. This mechanism and its benefits are explained
3698 * in the next three paragraphs.
3699 *
3700 * Even if a queue, say Q, is expired when it remains idle, Q
3701 * can still preempt the new in-service queue if the next
3702 * request of Q arrives soon (see the comments on
3703 * bfq_bfqq_update_budg_for_activation). If all queues and
3704 * groups have the same weight, this form of preemption,
3705 * combined with the hole-recovery heuristic described in the
3706 * comments on function bfq_bfqq_update_budg_for_activation,
3707 * are enough to preserve a correct bandwidth distribution in
3708 * the mid term, even without idling. In fact, even if not
3709 * idling allows the internal queues of the device to contain
3710 * many requests, and thus to reorder requests, we can rather
3711 * safely assume that the internal scheduler still preserves a
3712 * minimum of mid-term fairness.
3713 *
3714 * More precisely, this preemption-based, idleless approach
3715 * provides fairness in terms of IOPS, and not sectors per
3716 * second. This can be seen with a simple example. Suppose
3717 * that there are two queues with the same weight, but that
3718 * the first queue receives requests of 8 sectors, while the
3719 * second queue receives requests of 1024 sectors. In
3720 * addition, suppose that each of the two queues contains at
3721 * most one request at a time, which implies that each queue
3722 * always remains idle after it is served. Finally, after
3723 * remaining idle, each queue receives very quickly a new
3724 * request. It follows that the two queues are served
3725 * alternatively, preempting each other if needed. This
3726 * implies that, although both queues have the same weight,
3727 * the queue with large requests receives a service that is
3728 * 1024/8 times as high as the service received by the other
3729 * queue.
3730 *
3731 * The motivation for using preemption instead of idling (for
3732 * queues with the same weight) is that, by not idling,
3733 * service guarantees are preserved (completely or at least in
3734 * part) without minimally sacrificing throughput. And, if
3735 * there is no active group, then the primary expectation for
3736 * this device is probably a high throughput.
3737 *
3738 * We are now left only with explaining the two sub-conditions in the
3739 * additional compound condition that is checked below for deciding
3740 * whether the scenario is asymmetric. To explain the first
3741 * sub-condition, we need to add that the function
3742 * bfq_asymmetric_scenario checks the weights of only
3743 * non-weight-raised queues, for efficiency reasons (see comments on
3744 * bfq_weights_tree_add()). Then the fact that bfqq is weight-raised
3745 * is checked explicitly here. More precisely, the compound condition
3746 * below takes into account also the fact that, even if bfqq is being
3747 * weight-raised, the scenario is still symmetric if all queues with
3748 * requests waiting for completion happen to be
3749 * weight-raised. Actually, we should be even more precise here, and
3750 * differentiate between interactive weight raising and soft real-time
3751 * weight raising.
3752 *
3753 * The second sub-condition checked in the compound condition is
3754 * whether there is a fair amount of already in-flight I/O not
3755 * belonging to bfqq. If so, I/O dispatching is to be plugged, for the
3756 * following reason. The drive may decide to serve in-flight
3757 * non-bfqq's I/O requests before bfqq's ones, thereby delaying the
3758 * arrival of new I/O requests for bfqq (recall that bfqq is sync). If
3759 * I/O-dispatching is not plugged, then, while bfqq remains empty, a
3760 * basically uncontrolled amount of I/O from other queues may be
3761 * dispatched too, possibly causing the service of bfqq's I/O to be
3762 * delayed even longer in the drive. This problem gets more and more
3763 * serious as the speed and the queue depth of the drive grow,
3764 * because, as these two quantities grow, the probability to find no
3765 * queue busy but many requests in flight grows too. By contrast,
3766 * plugging I/O dispatching minimizes the delay induced by already
3767 * in-flight I/O, and enables bfqq to recover the bandwidth it may
3768 * lose because of this delay.
3769 *
3770 * As a side note, it is worth considering that the above
3771 * device-idling countermeasures may however fail in the following
3772 * unlucky scenario: if I/O-dispatch plugging is (correctly) disabled
3773 * in a time period during which all symmetry sub-conditions hold, and
3774 * therefore the device is allowed to enqueue many requests, but at
3775 * some later point in time some sub-condition stops to hold, then it
3776 * may become impossible to make requests be served in the desired
3777 * order until all the requests already queued in the device have been
3778 * served. The last sub-condition commented above somewhat mitigates
3779 * this problem for weight-raised queues.
3780 *
3781 * However, as an additional mitigation for this problem, we preserve
3782 * plugging for a special symmetric case that may suddenly turn into
3783 * asymmetric: the case where only bfqq is busy. In this case, not
3784 * expiring bfqq does not cause any harm to any other queues in terms
3785 * of service guarantees. In contrast, it avoids the following unlucky
3786 * sequence of events: (1) bfqq is expired, (2) a new queue with a
3787 * lower weight than bfqq becomes busy (or more queues), (3) the new
3788 * queue is served until a new request arrives for bfqq, (4) when bfqq
3789 * is finally served, there are so many requests of the new queue in
3790 * the drive that the pending requests for bfqq take a lot of time to
3791 * be served. In particular, event (2) may case even already
3792 * dispatched requests of bfqq to be delayed, inside the drive. So, to
3793 * avoid this series of events, the scenario is preventively declared
3794 * as asymmetric also if bfqq is the only busy queues
3795 */
3796static bool idling_needed_for_service_guarantees(struct bfq_data *bfqd,
3797						 struct bfq_queue *bfqq)
3798{
3799	int tot_busy_queues = bfq_tot_busy_queues(bfqd);
3800
3801	/* No point in idling for bfqq if it won't get requests any longer */
3802	if (unlikely(!bfqq_process_refs(bfqq)))
3803		return false;
3804
3805	return (bfqq->wr_coeff > 1 &&
3806		(bfqd->wr_busy_queues <
3807		 tot_busy_queues ||
3808		 bfqd->rq_in_driver >=
3809		 bfqq->dispatched + 4)) ||
3810		bfq_asymmetric_scenario(bfqd, bfqq) ||
3811		tot_busy_queues == 1;
3812}
3813
3814static bool __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq,
3815			      enum bfqq_expiration reason)
3816{
3817	/*
3818	 * If this bfqq is shared between multiple processes, check
3819	 * to make sure that those processes are still issuing I/Os
3820	 * within the mean seek distance. If not, it may be time to
3821	 * break the queues apart again.
3822	 */
3823	if (bfq_bfqq_coop(bfqq) && BFQQ_SEEKY(bfqq))
3824		bfq_mark_bfqq_split_coop(bfqq);
3825
3826	/*
3827	 * Consider queues with a higher finish virtual time than
3828	 * bfqq. If idling_needed_for_service_guarantees(bfqq) returns
3829	 * true, then bfqq's bandwidth would be violated if an
3830	 * uncontrolled amount of I/O from these queues were
3831	 * dispatched while bfqq is waiting for its new I/O to
3832	 * arrive. This is exactly what may happen if this is a forced
3833	 * expiration caused by a preemption attempt, and if bfqq is
3834	 * not re-scheduled. To prevent this from happening, re-queue
3835	 * bfqq if it needs I/O-dispatch plugging, even if it is
3836	 * empty. By doing so, bfqq is granted to be served before the
3837	 * above queues (provided that bfqq is of course eligible).
3838	 */
3839	if (RB_EMPTY_ROOT(&bfqq->sort_list) &&
3840	    !(reason == BFQQE_PREEMPTED &&
3841	      idling_needed_for_service_guarantees(bfqd, bfqq))) {
3842		if (bfqq->dispatched == 0)
3843			/*
3844			 * Overloading budget_timeout field to store
3845			 * the time at which the queue remains with no
3846			 * backlog and no outstanding request; used by
3847			 * the weight-raising mechanism.
3848			 */
3849			bfqq->budget_timeout = jiffies;
3850
3851		bfq_del_bfqq_busy(bfqq, true);
3852	} else {
3853		bfq_requeue_bfqq(bfqd, bfqq, true);
3854		/*
3855		 * Resort priority tree of potential close cooperators.
3856		 * See comments on bfq_pos_tree_add_move() for the unlikely().
3857		 */
3858		if (unlikely(!bfqd->nonrot_with_queueing &&
3859			     !RB_EMPTY_ROOT(&bfqq->sort_list)))
3860			bfq_pos_tree_add_move(bfqd, bfqq);
3861	}
3862
3863	/*
3864	 * All in-service entities must have been properly deactivated
3865	 * or requeued before executing the next function, which
3866	 * resets all in-service entities as no more in service. This
3867	 * may cause bfqq to be freed. If this happens, the next
3868	 * function returns true.
3869	 */
3870	return __bfq_bfqd_reset_in_service(bfqd);
3871}
3872
3873/**
3874 * __bfq_bfqq_recalc_budget - try to adapt the budget to the @bfqq behavior.
3875 * @bfqd: device data.
3876 * @bfqq: queue to update.
3877 * @reason: reason for expiration.
3878 *
3879 * Handle the feedback on @bfqq budget at queue expiration.
3880 * See the body for detailed comments.
3881 */
3882static void __bfq_bfqq_recalc_budget(struct bfq_data *bfqd,
3883				     struct bfq_queue *bfqq,
3884				     enum bfqq_expiration reason)
3885{
3886	struct request *next_rq;
3887	int budget, min_budget;
3888
3889	min_budget = bfq_min_budget(bfqd);
3890
3891	if (bfqq->wr_coeff == 1)
3892		budget = bfqq->max_budget;
3893	else /*
3894	      * Use a constant, low budget for weight-raised queues,
3895	      * to help achieve a low latency. Keep it slightly higher
3896	      * than the minimum possible budget, to cause a little
3897	      * bit fewer expirations.
3898	      */
3899		budget = 2 * min_budget;
3900
3901	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: last budg %d, budg left %d",
3902		bfqq->entity.budget, bfq_bfqq_budget_left(bfqq));
3903	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: last max_budg %d, min budg %d",
3904		budget, bfq_min_budget(bfqd));
3905	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: sync %d, seeky %d",
3906		bfq_bfqq_sync(bfqq), BFQQ_SEEKY(bfqd->in_service_queue));
3907
3908	if (bfq_bfqq_sync(bfqq) && bfqq->wr_coeff == 1) {
3909		switch (reason) {
3910		/*
3911		 * Caveat: in all the following cases we trade latency
3912		 * for throughput.
3913		 */
3914		case BFQQE_TOO_IDLE:
3915			/*
3916			 * This is the only case where we may reduce
3917			 * the budget: if there is no request of the
3918			 * process still waiting for completion, then
3919			 * we assume (tentatively) that the timer has
3920			 * expired because the batch of requests of
3921			 * the process could have been served with a
3922			 * smaller budget.  Hence, betting that
3923			 * process will behave in the same way when it
3924			 * becomes backlogged again, we reduce its
3925			 * next budget.  As long as we guess right,
3926			 * this budget cut reduces the latency
3927			 * experienced by the process.
3928			 *
3929			 * However, if there are still outstanding
3930			 * requests, then the process may have not yet
3931			 * issued its next request just because it is
3932			 * still waiting for the completion of some of
3933			 * the still outstanding ones.  So in this
3934			 * subcase we do not reduce its budget, on the
3935			 * contrary we increase it to possibly boost
3936			 * the throughput, as discussed in the
3937			 * comments to the BUDGET_TIMEOUT case.
3938			 */
3939			if (bfqq->dispatched > 0) /* still outstanding reqs */
3940				budget = min(budget * 2, bfqd->bfq_max_budget);
3941			else {
3942				if (budget > 5 * min_budget)
3943					budget -= 4 * min_budget;
3944				else
3945					budget = min_budget;
3946			}
3947			break;
3948		case BFQQE_BUDGET_TIMEOUT:
3949			/*
3950			 * We double the budget here because it gives
3951			 * the chance to boost the throughput if this
3952			 * is not a seeky process (and has bumped into
3953			 * this timeout because of, e.g., ZBR).
3954			 */
3955			budget = min(budget * 2, bfqd->bfq_max_budget);
3956			break;
3957		case BFQQE_BUDGET_EXHAUSTED:
3958			/*
3959			 * The process still has backlog, and did not
3960			 * let either the budget timeout or the disk
3961			 * idling timeout expire. Hence it is not
3962			 * seeky, has a short thinktime and may be
3963			 * happy with a higher budget too. So
3964			 * definitely increase the budget of this good
3965			 * candidate to boost the disk throughput.
3966			 */
3967			budget = min(budget * 4, bfqd->bfq_max_budget);
3968			break;
3969		case BFQQE_NO_MORE_REQUESTS:
3970			/*
3971			 * For queues that expire for this reason, it
3972			 * is particularly important to keep the
3973			 * budget close to the actual service they
3974			 * need. Doing so reduces the timestamp
3975			 * misalignment problem described in the
3976			 * comments in the body of
3977			 * __bfq_activate_entity. In fact, suppose
3978			 * that a queue systematically expires for
3979			 * BFQQE_NO_MORE_REQUESTS and presents a
3980			 * new request in time to enjoy timestamp
3981			 * back-shifting. The larger the budget of the
3982			 * queue is with respect to the service the
3983			 * queue actually requests in each service
3984			 * slot, the more times the queue can be
3985			 * reactivated with the same virtual finish
3986			 * time. It follows that, even if this finish
3987			 * time is pushed to the system virtual time
3988			 * to reduce the consequent timestamp
3989			 * misalignment, the queue unjustly enjoys for
3990			 * many re-activations a lower finish time
3991			 * than all newly activated queues.
3992			 *
3993			 * The service needed by bfqq is measured
3994			 * quite precisely by bfqq->entity.service.
3995			 * Since bfqq does not enjoy device idling,
3996			 * bfqq->entity.service is equal to the number
3997			 * of sectors that the process associated with
3998			 * bfqq requested to read/write before waiting
3999			 * for request completions, or blocking for
4000			 * other reasons.
4001			 */
4002			budget = max_t(int, bfqq->entity.service, min_budget);
4003			break;
4004		default:
4005			return;
4006		}
4007	} else if (!bfq_bfqq_sync(bfqq)) {
4008		/*
4009		 * Async queues get always the maximum possible
4010		 * budget, as for them we do not care about latency
4011		 * (in addition, their ability to dispatch is limited
4012		 * by the charging factor).
4013		 */
4014		budget = bfqd->bfq_max_budget;
4015	}
4016
4017	bfqq->max_budget = budget;
4018
4019	if (bfqd->budgets_assigned >= bfq_stats_min_budgets &&
4020	    !bfqd->bfq_user_max_budget)
4021		bfqq->max_budget = min(bfqq->max_budget, bfqd->bfq_max_budget);
4022
4023	/*
4024	 * If there is still backlog, then assign a new budget, making
4025	 * sure that it is large enough for the next request.  Since
4026	 * the finish time of bfqq must be kept in sync with the
4027	 * budget, be sure to call __bfq_bfqq_expire() *after* this
4028	 * update.
4029	 *
4030	 * If there is no backlog, then no need to update the budget;
4031	 * it will be updated on the arrival of a new request.
4032	 */
4033	next_rq = bfqq->next_rq;
4034	if (next_rq)
4035		bfqq->entity.budget = max_t(unsigned long, bfqq->max_budget,
4036					    bfq_serv_to_charge(next_rq, bfqq));
4037
4038	bfq_log_bfqq(bfqd, bfqq, "head sect: %u, new budget %d",
4039			next_rq ? blk_rq_sectors(next_rq) : 0,
4040			bfqq->entity.budget);
4041}
4042
4043/*
4044 * Return true if the process associated with bfqq is "slow". The slow
4045 * flag is used, in addition to the budget timeout, to reduce the
4046 * amount of service provided to seeky processes, and thus reduce
4047 * their chances to lower the throughput. More details in the comments
4048 * on the function bfq_bfqq_expire().
4049 *
4050 * An important observation is in order: as discussed in the comments
4051 * on the function bfq_update_peak_rate(), with devices with internal
4052 * queues, it is hard if ever possible to know when and for how long
4053 * an I/O request is processed by the device (apart from the trivial
4054 * I/O pattern where a new request is dispatched only after the
4055 * previous one has been completed). This makes it hard to evaluate
4056 * the real rate at which the I/O requests of each bfq_queue are
4057 * served.  In fact, for an I/O scheduler like BFQ, serving a
4058 * bfq_queue means just dispatching its requests during its service
4059 * slot (i.e., until the budget of the queue is exhausted, or the
4060 * queue remains idle, or, finally, a timeout fires). But, during the
4061 * service slot of a bfq_queue, around 100 ms at most, the device may
4062 * be even still processing requests of bfq_queues served in previous
4063 * service slots. On the opposite end, the requests of the in-service
4064 * bfq_queue may be completed after the service slot of the queue
4065 * finishes.
4066 *
4067 * Anyway, unless more sophisticated solutions are used
4068 * (where possible), the sum of the sizes of the requests dispatched
4069 * during the service slot of a bfq_queue is probably the only
4070 * approximation available for the service received by the bfq_queue
4071 * during its service slot. And this sum is the quantity used in this
4072 * function to evaluate the I/O speed of a process.
4073 */
4074static bool bfq_bfqq_is_slow(struct bfq_data *bfqd, struct bfq_queue *bfqq,
4075				 bool compensate, enum bfqq_expiration reason,
4076				 unsigned long *delta_ms)
4077{
4078	ktime_t delta_ktime;
4079	u32 delta_usecs;
4080	bool slow = BFQQ_SEEKY(bfqq); /* if delta too short, use seekyness */
4081
4082	if (!bfq_bfqq_sync(bfqq))
4083		return false;
4084
4085	if (compensate)
4086		delta_ktime = bfqd->last_idling_start;
4087	else
4088		delta_ktime = ktime_get();
4089	delta_ktime = ktime_sub(delta_ktime, bfqd->last_budget_start);
4090	delta_usecs = ktime_to_us(delta_ktime);
4091
4092	/* don't use too short time intervals */
4093	if (delta_usecs < 1000) {
4094		if (blk_queue_nonrot(bfqd->queue))
4095			 /*
4096			  * give same worst-case guarantees as idling
4097			  * for seeky
4098			  */
4099			*delta_ms = BFQ_MIN_TT / NSEC_PER_MSEC;
4100		else /* charge at least one seek */
4101			*delta_ms = bfq_slice_idle / NSEC_PER_MSEC;
4102
4103		return slow;
4104	}
4105
4106	*delta_ms = delta_usecs / USEC_PER_MSEC;
4107
4108	/*
4109	 * Use only long (> 20ms) intervals to filter out excessive
4110	 * spikes in service rate estimation.
4111	 */
4112	if (delta_usecs > 20000) {
4113		/*
4114		 * Caveat for rotational devices: processes doing I/O
4115		 * in the slower disk zones tend to be slow(er) even
4116		 * if not seeky. In this respect, the estimated peak
4117		 * rate is likely to be an average over the disk
4118		 * surface. Accordingly, to not be too harsh with
4119		 * unlucky processes, a process is deemed slow only if
4120		 * its rate has been lower than half of the estimated
4121		 * peak rate.
4122		 */
4123		slow = bfqq->entity.service < bfqd->bfq_max_budget / 2;
4124	}
4125
4126	bfq_log_bfqq(bfqd, bfqq, "bfq_bfqq_is_slow: slow %d", slow);
4127
4128	return slow;
4129}
4130
4131/*
4132 * To be deemed as soft real-time, an application must meet two
4133 * requirements. First, the application must not require an average
4134 * bandwidth higher than the approximate bandwidth required to playback or
4135 * record a compressed high-definition video.
4136 * The next function is invoked on the completion of the last request of a
4137 * batch, to compute the next-start time instant, soft_rt_next_start, such
4138 * that, if the next request of the application does not arrive before
4139 * soft_rt_next_start, then the above requirement on the bandwidth is met.
4140 *
4141 * The second requirement is that the request pattern of the application is
4142 * isochronous, i.e., that, after issuing a request or a batch of requests,
4143 * the application stops issuing new requests until all its pending requests
4144 * have been completed. After that, the application may issue a new batch,
4145 * and so on.
4146 * For this reason the next function is invoked to compute
4147 * soft_rt_next_start only for applications that meet this requirement,
4148 * whereas soft_rt_next_start is set to infinity for applications that do
4149 * not.
4150 *
4151 * Unfortunately, even a greedy (i.e., I/O-bound) application may
4152 * happen to meet, occasionally or systematically, both the above
4153 * bandwidth and isochrony requirements. This may happen at least in
4154 * the following circumstances. First, if the CPU load is high. The
4155 * application may stop issuing requests while the CPUs are busy
4156 * serving other processes, then restart, then stop again for a while,
4157 * and so on. The other circumstances are related to the storage
4158 * device: the storage device is highly loaded or reaches a low-enough
4159 * throughput with the I/O of the application (e.g., because the I/O
4160 * is random and/or the device is slow). In all these cases, the
4161 * I/O of the application may be simply slowed down enough to meet
4162 * the bandwidth and isochrony requirements. To reduce the probability
4163 * that greedy applications are deemed as soft real-time in these
4164 * corner cases, a further rule is used in the computation of
4165 * soft_rt_next_start: the return value of this function is forced to
4166 * be higher than the maximum between the following two quantities.
4167 *
4168 * (a) Current time plus: (1) the maximum time for which the arrival
4169 *     of a request is waited for when a sync queue becomes idle,
4170 *     namely bfqd->bfq_slice_idle, and (2) a few extra jiffies. We
4171 *     postpone for a moment the reason for adding a few extra
4172 *     jiffies; we get back to it after next item (b).  Lower-bounding
4173 *     the return value of this function with the current time plus
4174 *     bfqd->bfq_slice_idle tends to filter out greedy applications,
4175 *     because the latter issue their next request as soon as possible
4176 *     after the last one has been completed. In contrast, a soft
4177 *     real-time application spends some time processing data, after a
4178 *     batch of its requests has been completed.
4179 *
4180 * (b) Current value of bfqq->soft_rt_next_start. As pointed out
4181 *     above, greedy applications may happen to meet both the
4182 *     bandwidth and isochrony requirements under heavy CPU or
4183 *     storage-device load. In more detail, in these scenarios, these
4184 *     applications happen, only for limited time periods, to do I/O
4185 *     slowly enough to meet all the requirements described so far,
4186 *     including the filtering in above item (a). These slow-speed
4187 *     time intervals are usually interspersed between other time
4188 *     intervals during which these applications do I/O at a very high
4189 *     speed. Fortunately, exactly because of the high speed of the
4190 *     I/O in the high-speed intervals, the values returned by this
4191 *     function happen to be so high, near the end of any such
4192 *     high-speed interval, to be likely to fall *after* the end of
4193 *     the low-speed time interval that follows. These high values are
4194 *     stored in bfqq->soft_rt_next_start after each invocation of
4195 *     this function. As a consequence, if the last value of
4196 *     bfqq->soft_rt_next_start is constantly used to lower-bound the
4197 *     next value that this function may return, then, from the very
4198 *     beginning of a low-speed interval, bfqq->soft_rt_next_start is
4199 *     likely to be constantly kept so high that any I/O request
4200 *     issued during the low-speed interval is considered as arriving
4201 *     to soon for the application to be deemed as soft
4202 *     real-time. Then, in the high-speed interval that follows, the
4203 *     application will not be deemed as soft real-time, just because
4204 *     it will do I/O at a high speed. And so on.
4205 *
4206 * Getting back to the filtering in item (a), in the following two
4207 * cases this filtering might be easily passed by a greedy
4208 * application, if the reference quantity was just
4209 * bfqd->bfq_slice_idle:
4210 * 1) HZ is so low that the duration of a jiffy is comparable to or
4211 *    higher than bfqd->bfq_slice_idle. This happens, e.g., on slow
4212 *    devices with HZ=100. The time granularity may be so coarse
4213 *    that the approximation, in jiffies, of bfqd->bfq_slice_idle
4214 *    is rather lower than the exact value.
4215 * 2) jiffies, instead of increasing at a constant rate, may stop increasing
4216 *    for a while, then suddenly 'jump' by several units to recover the lost
4217 *    increments. This seems to happen, e.g., inside virtual machines.
4218 * To address this issue, in the filtering in (a) we do not use as a
4219 * reference time interval just bfqd->bfq_slice_idle, but
4220 * bfqd->bfq_slice_idle plus a few jiffies. In particular, we add the
4221 * minimum number of jiffies for which the filter seems to be quite
4222 * precise also in embedded systems and KVM/QEMU virtual machines.
4223 */
4224static unsigned long bfq_bfqq_softrt_next_start(struct bfq_data *bfqd,
4225						struct bfq_queue *bfqq)
4226{
4227	return max3(bfqq->soft_rt_next_start,
4228		    bfqq->last_idle_bklogged +
4229		    HZ * bfqq->service_from_backlogged /
4230		    bfqd->bfq_wr_max_softrt_rate,
4231		    jiffies + nsecs_to_jiffies(bfqq->bfqd->bfq_slice_idle) + 4);
4232}
4233
4234/**
4235 * bfq_bfqq_expire - expire a queue.
4236 * @bfqd: device owning the queue.
4237 * @bfqq: the queue to expire.
4238 * @compensate: if true, compensate for the time spent idling.
4239 * @reason: the reason causing the expiration.
4240 *
4241 * If the process associated with bfqq does slow I/O (e.g., because it
4242 * issues random requests), we charge bfqq with the time it has been
4243 * in service instead of the service it has received (see
4244 * bfq_bfqq_charge_time for details on how this goal is achieved). As
4245 * a consequence, bfqq will typically get higher timestamps upon
4246 * reactivation, and hence it will be rescheduled as if it had
4247 * received more service than what it has actually received. In the
4248 * end, bfqq receives less service in proportion to how slowly its
4249 * associated process consumes its budgets (and hence how seriously it
4250 * tends to lower the throughput). In addition, this time-charging
4251 * strategy guarantees time fairness among slow processes. In
4252 * contrast, if the process associated with bfqq is not slow, we
4253 * charge bfqq exactly with the service it has received.
4254 *
4255 * Charging time to the first type of queues and the exact service to
4256 * the other has the effect of using the WF2Q+ policy to schedule the
4257 * former on a timeslice basis, without violating service domain
4258 * guarantees among the latter.
4259 */
4260void bfq_bfqq_expire(struct bfq_data *bfqd,
4261		     struct bfq_queue *bfqq,
4262		     bool compensate,
4263		     enum bfqq_expiration reason)
4264{
4265	bool slow;
4266	unsigned long delta = 0;
4267	struct bfq_entity *entity = &bfqq->entity;
4268
4269	/*
4270	 * Check whether the process is slow (see bfq_bfqq_is_slow).
4271	 */
4272	slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, reason, &delta);
4273
4274	/*
4275	 * As above explained, charge slow (typically seeky) and
4276	 * timed-out queues with the time and not the service
4277	 * received, to favor sequential workloads.
4278	 *
4279	 * Processes doing I/O in the slower disk zones will tend to
4280	 * be slow(er) even if not seeky. Therefore, since the
4281	 * estimated peak rate is actually an average over the disk
4282	 * surface, these processes may timeout just for bad luck. To
4283	 * avoid punishing them, do not charge time to processes that
4284	 * succeeded in consuming at least 2/3 of their budget. This
4285	 * allows BFQ to preserve enough elasticity to still perform
4286	 * bandwidth, and not time, distribution with little unlucky
4287	 * or quasi-sequential processes.
4288	 */
4289	if (bfqq->wr_coeff == 1 &&
4290	    (slow ||
4291	     (reason == BFQQE_BUDGET_TIMEOUT &&
4292	      bfq_bfqq_budget_left(bfqq) >=  entity->budget / 3)))
4293		bfq_bfqq_charge_time(bfqd, bfqq, delta);
4294
 
 
 
 
4295	if (bfqd->low_latency && bfqq->wr_coeff == 1)
4296		bfqq->last_wr_start_finish = jiffies;
4297
4298	if (bfqd->low_latency && bfqd->bfq_wr_max_softrt_rate > 0 &&
4299	    RB_EMPTY_ROOT(&bfqq->sort_list)) {
4300		/*
4301		 * If we get here, and there are no outstanding
4302		 * requests, then the request pattern is isochronous
4303		 * (see the comments on the function
4304		 * bfq_bfqq_softrt_next_start()). Therefore we can
4305		 * compute soft_rt_next_start.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4306		 *
4307		 * If, instead, the queue still has outstanding
4308		 * requests, then we have to wait for the completion
4309		 * of all the outstanding requests to discover whether
4310		 * the request pattern is actually isochronous.
4311		 */
4312		if (bfqq->dispatched == 0)
 
4313			bfqq->soft_rt_next_start =
4314				bfq_bfqq_softrt_next_start(bfqd, bfqq);
4315		else if (bfqq->dispatched > 0) {
4316			/*
4317			 * Schedule an update of soft_rt_next_start to when
4318			 * the task may be discovered to be isochronous.
4319			 */
4320			bfq_mark_bfqq_softrt_update(bfqq);
4321		}
4322	}
4323
4324	bfq_log_bfqq(bfqd, bfqq,
4325		"expire (%d, slow %d, num_disp %d, short_ttime %d)", reason,
4326		slow, bfqq->dispatched, bfq_bfqq_has_short_ttime(bfqq));
4327
4328	/*
4329	 * bfqq expired, so no total service time needs to be computed
4330	 * any longer: reset state machine for measuring total service
4331	 * times.
4332	 */
4333	bfqd->rqs_injected = bfqd->wait_dispatch = false;
4334	bfqd->waited_rq = NULL;
4335
4336	/*
4337	 * Increase, decrease or leave budget unchanged according to
4338	 * reason.
4339	 */
4340	__bfq_bfqq_recalc_budget(bfqd, bfqq, reason);
4341	if (__bfq_bfqq_expire(bfqd, bfqq, reason))
4342		/* bfqq is gone, no more actions on it */
4343		return;
4344
4345	/* mark bfqq as waiting a request only if a bic still points to it */
4346	if (!bfq_bfqq_busy(bfqq) &&
4347	    reason != BFQQE_BUDGET_TIMEOUT &&
4348	    reason != BFQQE_BUDGET_EXHAUSTED) {
4349		bfq_mark_bfqq_non_blocking_wait_rq(bfqq);
4350		/*
4351		 * Not setting service to 0, because, if the next rq
4352		 * arrives in time, the queue will go on receiving
4353		 * service with this same budget (as if it never expired)
4354		 */
4355	} else
4356		entity->service = 0;
4357
4358	/*
4359	 * Reset the received-service counter for every parent entity.
4360	 * Differently from what happens with bfqq->entity.service,
4361	 * the resetting of this counter never needs to be postponed
4362	 * for parent entities. In fact, in case bfqq may have a
4363	 * chance to go on being served using the last, partially
4364	 * consumed budget, bfqq->entity.service needs to be kept,
4365	 * because if bfqq then actually goes on being served using
4366	 * the same budget, the last value of bfqq->entity.service is
4367	 * needed to properly decrement bfqq->entity.budget by the
4368	 * portion already consumed. In contrast, it is not necessary
4369	 * to keep entity->service for parent entities too, because
4370	 * the bubble up of the new value of bfqq->entity.budget will
4371	 * make sure that the budgets of parent entities are correct,
4372	 * even in case bfqq and thus parent entities go on receiving
4373	 * service with the same budget.
4374	 */
4375	entity = entity->parent;
4376	for_each_entity(entity)
4377		entity->service = 0;
4378}
4379
4380/*
4381 * Budget timeout is not implemented through a dedicated timer, but
4382 * just checked on request arrivals and completions, as well as on
4383 * idle timer expirations.
4384 */
4385static bool bfq_bfqq_budget_timeout(struct bfq_queue *bfqq)
4386{
4387	return time_is_before_eq_jiffies(bfqq->budget_timeout);
4388}
4389
4390/*
4391 * If we expire a queue that is actively waiting (i.e., with the
4392 * device idled) for the arrival of a new request, then we may incur
4393 * the timestamp misalignment problem described in the body of the
4394 * function __bfq_activate_entity. Hence we return true only if this
4395 * condition does not hold, or if the queue is slow enough to deserve
4396 * only to be kicked off for preserving a high throughput.
4397 */
4398static bool bfq_may_expire_for_budg_timeout(struct bfq_queue *bfqq)
4399{
4400	bfq_log_bfqq(bfqq->bfqd, bfqq,
4401		"may_budget_timeout: wait_request %d left %d timeout %d",
4402		bfq_bfqq_wait_request(bfqq),
4403			bfq_bfqq_budget_left(bfqq) >=  bfqq->entity.budget / 3,
4404		bfq_bfqq_budget_timeout(bfqq));
4405
4406	return (!bfq_bfqq_wait_request(bfqq) ||
4407		bfq_bfqq_budget_left(bfqq) >=  bfqq->entity.budget / 3)
4408		&&
4409		bfq_bfqq_budget_timeout(bfqq);
4410}
4411
4412static bool idling_boosts_thr_without_issues(struct bfq_data *bfqd,
4413					     struct bfq_queue *bfqq)
4414{
4415	bool rot_without_queueing =
4416		!blk_queue_nonrot(bfqd->queue) && !bfqd->hw_tag,
4417		bfqq_sequential_and_IO_bound,
4418		idling_boosts_thr;
4419
4420	/* No point in idling for bfqq if it won't get requests any longer */
4421	if (unlikely(!bfqq_process_refs(bfqq)))
4422		return false;
4423
4424	bfqq_sequential_and_IO_bound = !BFQQ_SEEKY(bfqq) &&
4425		bfq_bfqq_IO_bound(bfqq) && bfq_bfqq_has_short_ttime(bfqq);
4426
4427	/*
4428	 * The next variable takes into account the cases where idling
4429	 * boosts the throughput.
4430	 *
4431	 * The value of the variable is computed considering, first, that
4432	 * idling is virtually always beneficial for the throughput if:
4433	 * (a) the device is not NCQ-capable and rotational, or
4434	 * (b) regardless of the presence of NCQ, the device is rotational and
4435	 *     the request pattern for bfqq is I/O-bound and sequential, or
4436	 * (c) regardless of whether it is rotational, the device is
4437	 *     not NCQ-capable and the request pattern for bfqq is
4438	 *     I/O-bound and sequential.
4439	 *
4440	 * Secondly, and in contrast to the above item (b), idling an
4441	 * NCQ-capable flash-based device would not boost the
4442	 * throughput even with sequential I/O; rather it would lower
4443	 * the throughput in proportion to how fast the device
4444	 * is. Accordingly, the next variable is true if any of the
4445	 * above conditions (a), (b) or (c) is true, and, in
4446	 * particular, happens to be false if bfqd is an NCQ-capable
4447	 * flash-based device.
4448	 */
4449	idling_boosts_thr = rot_without_queueing ||
4450		((!blk_queue_nonrot(bfqd->queue) || !bfqd->hw_tag) &&
4451		 bfqq_sequential_and_IO_bound);
4452
4453	/*
4454	 * The return value of this function is equal to that of
4455	 * idling_boosts_thr, unless a special case holds. In this
4456	 * special case, described below, idling may cause problems to
4457	 * weight-raised queues.
4458	 *
4459	 * When the request pool is saturated (e.g., in the presence
4460	 * of write hogs), if the processes associated with
4461	 * non-weight-raised queues ask for requests at a lower rate,
4462	 * then processes associated with weight-raised queues have a
4463	 * higher probability to get a request from the pool
4464	 * immediately (or at least soon) when they need one. Thus
4465	 * they have a higher probability to actually get a fraction
4466	 * of the device throughput proportional to their high
4467	 * weight. This is especially true with NCQ-capable drives,
4468	 * which enqueue several requests in advance, and further
4469	 * reorder internally-queued requests.
4470	 *
4471	 * For this reason, we force to false the return value if
4472	 * there are weight-raised busy queues. In this case, and if
4473	 * bfqq is not weight-raised, this guarantees that the device
4474	 * is not idled for bfqq (if, instead, bfqq is weight-raised,
4475	 * then idling will be guaranteed by another variable, see
4476	 * below). Combined with the timestamping rules of BFQ (see
4477	 * [1] for details), this behavior causes bfqq, and hence any
4478	 * sync non-weight-raised queue, to get a lower number of
4479	 * requests served, and thus to ask for a lower number of
4480	 * requests from the request pool, before the busy
4481	 * weight-raised queues get served again. This often mitigates
4482	 * starvation problems in the presence of heavy write
4483	 * workloads and NCQ, thereby guaranteeing a higher
4484	 * application and system responsiveness in these hostile
4485	 * scenarios.
4486	 */
4487	return idling_boosts_thr &&
4488		bfqd->wr_busy_queues == 0;
4489}
4490
4491/*
4492 * For a queue that becomes empty, device idling is allowed only if
4493 * this function returns true for that queue. As a consequence, since
4494 * device idling plays a critical role for both throughput boosting
4495 * and service guarantees, the return value of this function plays a
4496 * critical role as well.
4497 *
4498 * In a nutshell, this function returns true only if idling is
4499 * beneficial for throughput or, even if detrimental for throughput,
4500 * idling is however necessary to preserve service guarantees (low
4501 * latency, desired throughput distribution, ...). In particular, on
4502 * NCQ-capable devices, this function tries to return false, so as to
4503 * help keep the drives' internal queues full, whenever this helps the
4504 * device boost the throughput without causing any service-guarantee
4505 * issue.
4506 *
4507 * Most of the issues taken into account to get the return value of
4508 * this function are not trivial. We discuss these issues in the two
4509 * functions providing the main pieces of information needed by this
4510 * function.
4511 */
4512static bool bfq_better_to_idle(struct bfq_queue *bfqq)
4513{
4514	struct bfq_data *bfqd = bfqq->bfqd;
4515	bool idling_boosts_thr_with_no_issue, idling_needed_for_service_guar;
4516
4517	/* No point in idling for bfqq if it won't get requests any longer */
4518	if (unlikely(!bfqq_process_refs(bfqq)))
4519		return false;
4520
4521	if (unlikely(bfqd->strict_guarantees))
4522		return true;
4523
4524	/*
4525	 * Idling is performed only if slice_idle > 0. In addition, we
4526	 * do not idle if
4527	 * (a) bfqq is async
4528	 * (b) bfqq is in the idle io prio class: in this case we do
4529	 * not idle because we want to minimize the bandwidth that
4530	 * queues in this class can steal to higher-priority queues
4531	 */
4532	if (bfqd->bfq_slice_idle == 0 || !bfq_bfqq_sync(bfqq) ||
4533	   bfq_class_idle(bfqq))
4534		return false;
4535
4536	idling_boosts_thr_with_no_issue =
4537		idling_boosts_thr_without_issues(bfqd, bfqq);
4538
4539	idling_needed_for_service_guar =
4540		idling_needed_for_service_guarantees(bfqd, bfqq);
4541
4542	/*
4543	 * We have now the two components we need to compute the
4544	 * return value of the function, which is true only if idling
4545	 * either boosts the throughput (without issues), or is
4546	 * necessary to preserve service guarantees.
4547	 */
4548	return idling_boosts_thr_with_no_issue ||
4549		idling_needed_for_service_guar;
4550}
4551
4552/*
4553 * If the in-service queue is empty but the function bfq_better_to_idle
4554 * returns true, then:
4555 * 1) the queue must remain in service and cannot be expired, and
4556 * 2) the device must be idled to wait for the possible arrival of a new
4557 *    request for the queue.
4558 * See the comments on the function bfq_better_to_idle for the reasons
4559 * why performing device idling is the best choice to boost the throughput
4560 * and preserve service guarantees when bfq_better_to_idle itself
4561 * returns true.
4562 */
4563static bool bfq_bfqq_must_idle(struct bfq_queue *bfqq)
4564{
4565	return RB_EMPTY_ROOT(&bfqq->sort_list) && bfq_better_to_idle(bfqq);
4566}
4567
4568/*
4569 * This function chooses the queue from which to pick the next extra
4570 * I/O request to inject, if it finds a compatible queue. See the
4571 * comments on bfq_update_inject_limit() for details on the injection
4572 * mechanism, and for the definitions of the quantities mentioned
4573 * below.
4574 */
4575static struct bfq_queue *
4576bfq_choose_bfqq_for_injection(struct bfq_data *bfqd)
4577{
4578	struct bfq_queue *bfqq, *in_serv_bfqq = bfqd->in_service_queue;
4579	unsigned int limit = in_serv_bfqq->inject_limit;
4580	/*
4581	 * If
4582	 * - bfqq is not weight-raised and therefore does not carry
4583	 *   time-critical I/O,
4584	 * or
4585	 * - regardless of whether bfqq is weight-raised, bfqq has
4586	 *   however a long think time, during which it can absorb the
4587	 *   effect of an appropriate number of extra I/O requests
4588	 *   from other queues (see bfq_update_inject_limit for
4589	 *   details on the computation of this number);
4590	 * then injection can be performed without restrictions.
4591	 */
4592	bool in_serv_always_inject = in_serv_bfqq->wr_coeff == 1 ||
4593		!bfq_bfqq_has_short_ttime(in_serv_bfqq);
4594
4595	/*
4596	 * If
4597	 * - the baseline total service time could not be sampled yet,
4598	 *   so the inject limit happens to be still 0, and
4599	 * - a lot of time has elapsed since the plugging of I/O
4600	 *   dispatching started, so drive speed is being wasted
4601	 *   significantly;
4602	 * then temporarily raise inject limit to one request.
4603	 */
4604	if (limit == 0 && in_serv_bfqq->last_serv_time_ns == 0 &&
4605	    bfq_bfqq_wait_request(in_serv_bfqq) &&
4606	    time_is_before_eq_jiffies(bfqd->last_idling_start_jiffies +
4607				      bfqd->bfq_slice_idle)
4608		)
4609		limit = 1;
4610
4611	if (bfqd->rq_in_driver >= limit)
4612		return NULL;
4613
4614	/*
4615	 * Linear search of the source queue for injection; but, with
4616	 * a high probability, very few steps are needed to find a
4617	 * candidate queue, i.e., a queue with enough budget left for
4618	 * its next request. In fact:
4619	 * - BFQ dynamically updates the budget of every queue so as
4620	 *   to accommodate the expected backlog of the queue;
4621	 * - if a queue gets all its requests dispatched as injected
4622	 *   service, then the queue is removed from the active list
4623	 *   (and re-added only if it gets new requests, but then it
4624	 *   is assigned again enough budget for its new backlog).
4625	 */
4626	list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list)
4627		if (!RB_EMPTY_ROOT(&bfqq->sort_list) &&
4628		    (in_serv_always_inject || bfqq->wr_coeff > 1) &&
4629		    bfq_serv_to_charge(bfqq->next_rq, bfqq) <=
4630		    bfq_bfqq_budget_left(bfqq)) {
4631			/*
4632			 * Allow for only one large in-flight request
4633			 * on non-rotational devices, for the
4634			 * following reason. On non-rotationl drives,
4635			 * large requests take much longer than
4636			 * smaller requests to be served. In addition,
4637			 * the drive prefers to serve large requests
4638			 * w.r.t. to small ones, if it can choose. So,
4639			 * having more than one large requests queued
4640			 * in the drive may easily make the next first
4641			 * request of the in-service queue wait for so
4642			 * long to break bfqq's service guarantees. On
4643			 * the bright side, large requests let the
4644			 * drive reach a very high throughput, even if
4645			 * there is only one in-flight large request
4646			 * at a time.
4647			 */
4648			if (blk_queue_nonrot(bfqd->queue) &&
4649			    blk_rq_sectors(bfqq->next_rq) >=
4650			    BFQQ_SECT_THR_NONROT)
4651				limit = min_t(unsigned int, 1, limit);
4652			else
4653				limit = in_serv_bfqq->inject_limit;
4654
4655			if (bfqd->rq_in_driver < limit) {
4656				bfqd->rqs_injected = true;
4657				return bfqq;
4658			}
4659		}
4660
4661	return NULL;
4662}
4663
4664/*
4665 * Select a queue for service.  If we have a current queue in service,
4666 * check whether to continue servicing it, or retrieve and set a new one.
4667 */
4668static struct bfq_queue *bfq_select_queue(struct bfq_data *bfqd)
4669{
4670	struct bfq_queue *bfqq;
4671	struct request *next_rq;
4672	enum bfqq_expiration reason = BFQQE_BUDGET_TIMEOUT;
4673
4674	bfqq = bfqd->in_service_queue;
4675	if (!bfqq)
4676		goto new_queue;
4677
4678	bfq_log_bfqq(bfqd, bfqq, "select_queue: already in-service queue");
4679
4680	/*
4681	 * Do not expire bfqq for budget timeout if bfqq may be about
4682	 * to enjoy device idling. The reason why, in this case, we
4683	 * prevent bfqq from expiring is the same as in the comments
4684	 * on the case where bfq_bfqq_must_idle() returns true, in
4685	 * bfq_completed_request().
4686	 */
4687	if (bfq_may_expire_for_budg_timeout(bfqq) &&
4688	    !bfq_bfqq_must_idle(bfqq))
4689		goto expire;
4690
4691check_queue:
4692	/*
4693	 * This loop is rarely executed more than once. Even when it
4694	 * happens, it is much more convenient to re-execute this loop
4695	 * than to return NULL and trigger a new dispatch to get a
4696	 * request served.
4697	 */
4698	next_rq = bfqq->next_rq;
4699	/*
4700	 * If bfqq has requests queued and it has enough budget left to
4701	 * serve them, keep the queue, otherwise expire it.
4702	 */
4703	if (next_rq) {
4704		if (bfq_serv_to_charge(next_rq, bfqq) >
4705			bfq_bfqq_budget_left(bfqq)) {
4706			/*
4707			 * Expire the queue for budget exhaustion,
4708			 * which makes sure that the next budget is
4709			 * enough to serve the next request, even if
4710			 * it comes from the fifo expired path.
4711			 */
4712			reason = BFQQE_BUDGET_EXHAUSTED;
4713			goto expire;
4714		} else {
4715			/*
4716			 * The idle timer may be pending because we may
4717			 * not disable disk idling even when a new request
4718			 * arrives.
4719			 */
4720			if (bfq_bfqq_wait_request(bfqq)) {
4721				/*
4722				 * If we get here: 1) at least a new request
4723				 * has arrived but we have not disabled the
4724				 * timer because the request was too small,
4725				 * 2) then the block layer has unplugged
4726				 * the device, causing the dispatch to be
4727				 * invoked.
4728				 *
4729				 * Since the device is unplugged, now the
4730				 * requests are probably large enough to
4731				 * provide a reasonable throughput.
4732				 * So we disable idling.
4733				 */
4734				bfq_clear_bfqq_wait_request(bfqq);
4735				hrtimer_try_to_cancel(&bfqd->idle_slice_timer);
4736			}
4737			goto keep_queue;
4738		}
4739	}
4740
4741	/*
4742	 * No requests pending. However, if the in-service queue is idling
4743	 * for a new request, or has requests waiting for a completion and
4744	 * may idle after their completion, then keep it anyway.
4745	 *
4746	 * Yet, inject service from other queues if it boosts
4747	 * throughput and is possible.
4748	 */
4749	if (bfq_bfqq_wait_request(bfqq) ||
4750	    (bfqq->dispatched != 0 && bfq_better_to_idle(bfqq))) {
4751		struct bfq_queue *async_bfqq =
4752			bfqq->bic && bfqq->bic->bfqq[0] &&
4753			bfq_bfqq_busy(bfqq->bic->bfqq[0]) &&
4754			bfqq->bic->bfqq[0]->next_rq ?
4755			bfqq->bic->bfqq[0] : NULL;
4756		struct bfq_queue *blocked_bfqq =
4757			!hlist_empty(&bfqq->woken_list) ?
4758			container_of(bfqq->woken_list.first,
4759				     struct bfq_queue,
4760				     woken_list_node)
4761			: NULL;
4762
4763		/*
4764		 * The next four mutually-exclusive ifs decide
4765		 * whether to try injection, and choose the queue to
4766		 * pick an I/O request from.
4767		 *
4768		 * The first if checks whether the process associated
4769		 * with bfqq has also async I/O pending. If so, it
4770		 * injects such I/O unconditionally. Injecting async
4771		 * I/O from the same process can cause no harm to the
4772		 * process. On the contrary, it can only increase
4773		 * bandwidth and reduce latency for the process.
4774		 *
4775		 * The second if checks whether there happens to be a
4776		 * non-empty waker queue for bfqq, i.e., a queue whose
4777		 * I/O needs to be completed for bfqq to receive new
4778		 * I/O. This happens, e.g., if bfqq is associated with
4779		 * a process that does some sync. A sync generates
4780		 * extra blocking I/O, which must be completed before
4781		 * the process associated with bfqq can go on with its
4782		 * I/O. If the I/O of the waker queue is not served,
4783		 * then bfqq remains empty, and no I/O is dispatched,
4784		 * until the idle timeout fires for bfqq. This is
4785		 * likely to result in lower bandwidth and higher
4786		 * latencies for bfqq, and in a severe loss of total
4787		 * throughput. The best action to take is therefore to
4788		 * serve the waker queue as soon as possible. So do it
4789		 * (without relying on the third alternative below for
4790		 * eventually serving waker_bfqq's I/O; see the last
4791		 * paragraph for further details). This systematic
4792		 * injection of I/O from the waker queue does not
4793		 * cause any delay to bfqq's I/O. On the contrary,
4794		 * next bfqq's I/O is brought forward dramatically,
4795		 * for it is not blocked for milliseconds.
4796		 *
4797		 * The third if checks whether there is a queue woken
4798		 * by bfqq, and currently with pending I/O. Such a
4799		 * woken queue does not steal bandwidth from bfqq,
4800		 * because it remains soon without I/O if bfqq is not
4801		 * served. So there is virtually no risk of loss of
4802		 * bandwidth for bfqq if this woken queue has I/O
4803		 * dispatched while bfqq is waiting for new I/O.
4804		 *
4805		 * The fourth if checks whether bfqq is a queue for
4806		 * which it is better to avoid injection. It is so if
4807		 * bfqq delivers more throughput when served without
4808		 * any further I/O from other queues in the middle, or
4809		 * if the service times of bfqq's I/O requests both
4810		 * count more than overall throughput, and may be
4811		 * easily increased by injection (this happens if bfqq
4812		 * has a short think time). If none of these
4813		 * conditions holds, then a candidate queue for
4814		 * injection is looked for through
4815		 * bfq_choose_bfqq_for_injection(). Note that the
4816		 * latter may return NULL (for example if the inject
4817		 * limit for bfqq is currently 0).
4818		 *
4819		 * NOTE: motivation for the second alternative
4820		 *
4821		 * Thanks to the way the inject limit is updated in
4822		 * bfq_update_has_short_ttime(), it is rather likely
4823		 * that, if I/O is being plugged for bfqq and the
4824		 * waker queue has pending I/O requests that are
4825		 * blocking bfqq's I/O, then the fourth alternative
4826		 * above lets the waker queue get served before the
4827		 * I/O-plugging timeout fires. So one may deem the
4828		 * second alternative superfluous. It is not, because
4829		 * the fourth alternative may be way less effective in
4830		 * case of a synchronization. For two main
4831		 * reasons. First, throughput may be low because the
4832		 * inject limit may be too low to guarantee the same
4833		 * amount of injected I/O, from the waker queue or
4834		 * other queues, that the second alternative
4835		 * guarantees (the second alternative unconditionally
4836		 * injects a pending I/O request of the waker queue
4837		 * for each bfq_dispatch_request()). Second, with the
4838		 * fourth alternative, the duration of the plugging,
4839		 * i.e., the time before bfqq finally receives new I/O,
4840		 * may not be minimized, because the waker queue may
4841		 * happen to be served only after other queues.
4842		 */
4843		if (async_bfqq &&
4844		    icq_to_bic(async_bfqq->next_rq->elv.icq) == bfqq->bic &&
4845		    bfq_serv_to_charge(async_bfqq->next_rq, async_bfqq) <=
4846		    bfq_bfqq_budget_left(async_bfqq))
4847			bfqq = bfqq->bic->bfqq[0];
4848		else if (bfqq->waker_bfqq &&
4849			   bfq_bfqq_busy(bfqq->waker_bfqq) &&
4850			   bfqq->waker_bfqq->next_rq &&
4851			   bfq_serv_to_charge(bfqq->waker_bfqq->next_rq,
4852					      bfqq->waker_bfqq) <=
4853			   bfq_bfqq_budget_left(bfqq->waker_bfqq)
4854			)
4855			bfqq = bfqq->waker_bfqq;
4856		else if (blocked_bfqq &&
4857			   bfq_bfqq_busy(blocked_bfqq) &&
4858			   blocked_bfqq->next_rq &&
4859			   bfq_serv_to_charge(blocked_bfqq->next_rq,
4860					      blocked_bfqq) <=
4861			   bfq_bfqq_budget_left(blocked_bfqq)
4862			)
4863			bfqq = blocked_bfqq;
4864		else if (!idling_boosts_thr_without_issues(bfqd, bfqq) &&
4865			 (bfqq->wr_coeff == 1 || bfqd->wr_busy_queues > 1 ||
4866			  !bfq_bfqq_has_short_ttime(bfqq)))
4867			bfqq = bfq_choose_bfqq_for_injection(bfqd);
4868		else
4869			bfqq = NULL;
4870
4871		goto keep_queue;
4872	}
4873
4874	reason = BFQQE_NO_MORE_REQUESTS;
4875expire:
4876	bfq_bfqq_expire(bfqd, bfqq, false, reason);
4877new_queue:
4878	bfqq = bfq_set_in_service_queue(bfqd);
4879	if (bfqq) {
4880		bfq_log_bfqq(bfqd, bfqq, "select_queue: checking new queue");
4881		goto check_queue;
4882	}
4883keep_queue:
4884	if (bfqq)
4885		bfq_log_bfqq(bfqd, bfqq, "select_queue: returned this queue");
4886	else
4887		bfq_log(bfqd, "select_queue: no queue returned");
4888
4889	return bfqq;
4890}
4891
4892static void bfq_update_wr_data(struct bfq_data *bfqd, struct bfq_queue *bfqq)
4893{
4894	struct bfq_entity *entity = &bfqq->entity;
4895
4896	if (bfqq->wr_coeff > 1) { /* queue is being weight-raised */
4897		bfq_log_bfqq(bfqd, bfqq,
4898			"raising period dur %u/%u msec, old coeff %u, w %d(%d)",
4899			jiffies_to_msecs(jiffies - bfqq->last_wr_start_finish),
4900			jiffies_to_msecs(bfqq->wr_cur_max_time),
4901			bfqq->wr_coeff,
4902			bfqq->entity.weight, bfqq->entity.orig_weight);
4903
4904		if (entity->prio_changed)
4905			bfq_log_bfqq(bfqd, bfqq, "WARN: pending prio change");
4906
4907		/*
4908		 * If the queue was activated in a burst, or too much
4909		 * time has elapsed from the beginning of this
4910		 * weight-raising period, then end weight raising.
4911		 */
4912		if (bfq_bfqq_in_large_burst(bfqq))
4913			bfq_bfqq_end_wr(bfqq);
4914		else if (time_is_before_jiffies(bfqq->last_wr_start_finish +
4915						bfqq->wr_cur_max_time)) {
4916			if (bfqq->wr_cur_max_time != bfqd->bfq_wr_rt_max_time ||
4917			time_is_before_jiffies(bfqq->wr_start_at_switch_to_srt +
4918					       bfq_wr_duration(bfqd))) {
4919				/*
4920				 * Either in interactive weight
4921				 * raising, or in soft_rt weight
4922				 * raising with the
4923				 * interactive-weight-raising period
4924				 * elapsed (so no switch back to
4925				 * interactive weight raising).
4926				 */
4927				bfq_bfqq_end_wr(bfqq);
4928			} else { /*
4929				  * soft_rt finishing while still in
4930				  * interactive period, switch back to
4931				  * interactive weight raising
4932				  */
4933				switch_back_to_interactive_wr(bfqq, bfqd);
4934				bfqq->entity.prio_changed = 1;
4935			}
4936		}
4937		if (bfqq->wr_coeff > 1 &&
4938		    bfqq->wr_cur_max_time != bfqd->bfq_wr_rt_max_time &&
4939		    bfqq->service_from_wr > max_service_from_wr) {
4940			/* see comments on max_service_from_wr */
4941			bfq_bfqq_end_wr(bfqq);
4942		}
4943	}
4944	/*
4945	 * To improve latency (for this or other queues), immediately
4946	 * update weight both if it must be raised and if it must be
4947	 * lowered. Since, entity may be on some active tree here, and
4948	 * might have a pending change of its ioprio class, invoke
4949	 * next function with the last parameter unset (see the
4950	 * comments on the function).
4951	 */
4952	if ((entity->weight > entity->orig_weight) != (bfqq->wr_coeff > 1))
4953		__bfq_entity_update_weight_prio(bfq_entity_service_tree(entity),
4954						entity, false);
4955}
4956
4957/*
4958 * Dispatch next request from bfqq.
4959 */
4960static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
4961						 struct bfq_queue *bfqq)
4962{
4963	struct request *rq = bfqq->next_rq;
4964	unsigned long service_to_charge;
4965
4966	service_to_charge = bfq_serv_to_charge(rq, bfqq);
4967
4968	bfq_bfqq_served(bfqq, service_to_charge);
4969
4970	if (bfqq == bfqd->in_service_queue && bfqd->wait_dispatch) {
4971		bfqd->wait_dispatch = false;
4972		bfqd->waited_rq = rq;
4973	}
4974
4975	bfq_dispatch_remove(bfqd->queue, rq);
4976
4977	if (bfqq != bfqd->in_service_queue)
4978		goto return_rq;
4979
4980	/*
4981	 * If weight raising has to terminate for bfqq, then next
4982	 * function causes an immediate update of bfqq's weight,
4983	 * without waiting for next activation. As a consequence, on
4984	 * expiration, bfqq will be timestamped as if has never been
4985	 * weight-raised during this service slot, even if it has
4986	 * received part or even most of the service as a
4987	 * weight-raised queue. This inflates bfqq's timestamps, which
4988	 * is beneficial, as bfqq is then more willing to leave the
4989	 * device immediately to possible other weight-raised queues.
4990	 */
4991	bfq_update_wr_data(bfqd, bfqq);
4992
4993	/*
4994	 * Expire bfqq, pretending that its budget expired, if bfqq
4995	 * belongs to CLASS_IDLE and other queues are waiting for
4996	 * service.
4997	 */
4998	if (!(bfq_tot_busy_queues(bfqd) > 1 && bfq_class_idle(bfqq)))
4999		goto return_rq;
5000
5001	bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED);
5002
5003return_rq:
5004	return rq;
5005}
5006
5007static bool bfq_has_work(struct blk_mq_hw_ctx *hctx)
5008{
5009	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
5010
5011	/*
5012	 * Avoiding lock: a race on bfqd->queued should cause at
5013	 * most a call to dispatch for nothing
5014	 */
5015	return !list_empty_careful(&bfqd->dispatch) ||
5016		READ_ONCE(bfqd->queued);
5017}
5018
5019static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx)
5020{
5021	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
5022	struct request *rq = NULL;
5023	struct bfq_queue *bfqq = NULL;
5024
5025	if (!list_empty(&bfqd->dispatch)) {
5026		rq = list_first_entry(&bfqd->dispatch, struct request,
5027				      queuelist);
5028		list_del_init(&rq->queuelist);
5029
5030		bfqq = RQ_BFQQ(rq);
5031
5032		if (bfqq) {
5033			/*
5034			 * Increment counters here, because this
5035			 * dispatch does not follow the standard
5036			 * dispatch flow (where counters are
5037			 * incremented)
5038			 */
5039			bfqq->dispatched++;
5040
5041			goto inc_in_driver_start_rq;
5042		}
5043
5044		/*
5045		 * We exploit the bfq_finish_requeue_request hook to
5046		 * decrement rq_in_driver, but
5047		 * bfq_finish_requeue_request will not be invoked on
5048		 * this request. So, to avoid unbalance, just start
5049		 * this request, without incrementing rq_in_driver. As
5050		 * a negative consequence, rq_in_driver is deceptively
5051		 * lower than it should be while this request is in
5052		 * service. This may cause bfq_schedule_dispatch to be
5053		 * invoked uselessly.
5054		 *
5055		 * As for implementing an exact solution, the
5056		 * bfq_finish_requeue_request hook, if defined, is
5057		 * probably invoked also on this request. So, by
5058		 * exploiting this hook, we could 1) increment
5059		 * rq_in_driver here, and 2) decrement it in
5060		 * bfq_finish_requeue_request. Such a solution would
5061		 * let the value of the counter be always accurate,
5062		 * but it would entail using an extra interface
5063		 * function. This cost seems higher than the benefit,
5064		 * being the frequency of non-elevator-private
5065		 * requests very low.
5066		 */
5067		goto start_rq;
5068	}
5069
5070	bfq_log(bfqd, "dispatch requests: %d busy queues",
5071		bfq_tot_busy_queues(bfqd));
5072
5073	if (bfq_tot_busy_queues(bfqd) == 0)
5074		goto exit;
5075
5076	/*
5077	 * Force device to serve one request at a time if
5078	 * strict_guarantees is true. Forcing this service scheme is
5079	 * currently the ONLY way to guarantee that the request
5080	 * service order enforced by the scheduler is respected by a
5081	 * queueing device. Otherwise the device is free even to make
5082	 * some unlucky request wait for as long as the device
5083	 * wishes.
5084	 *
5085	 * Of course, serving one request at a time may cause loss of
5086	 * throughput.
5087	 */
5088	if (bfqd->strict_guarantees && bfqd->rq_in_driver > 0)
5089		goto exit;
5090
5091	bfqq = bfq_select_queue(bfqd);
5092	if (!bfqq)
5093		goto exit;
5094
5095	rq = bfq_dispatch_rq_from_bfqq(bfqd, bfqq);
5096
5097	if (rq) {
5098inc_in_driver_start_rq:
5099		bfqd->rq_in_driver++;
5100start_rq:
5101		rq->rq_flags |= RQF_STARTED;
5102	}
5103exit:
5104	return rq;
5105}
5106
5107#ifdef CONFIG_BFQ_CGROUP_DEBUG
5108static void bfq_update_dispatch_stats(struct request_queue *q,
5109				      struct request *rq,
5110				      struct bfq_queue *in_serv_queue,
5111				      bool idle_timer_disabled)
5112{
5113	struct bfq_queue *bfqq = rq ? RQ_BFQQ(rq) : NULL;
5114
5115	if (!idle_timer_disabled && !bfqq)
5116		return;
5117
5118	/*
5119	 * rq and bfqq are guaranteed to exist until this function
5120	 * ends, for the following reasons. First, rq can be
5121	 * dispatched to the device, and then can be completed and
5122	 * freed, only after this function ends. Second, rq cannot be
5123	 * merged (and thus freed because of a merge) any longer,
5124	 * because it has already started. Thus rq cannot be freed
5125	 * before this function ends, and, since rq has a reference to
5126	 * bfqq, the same guarantee holds for bfqq too.
5127	 *
5128	 * In addition, the following queue lock guarantees that
5129	 * bfqq_group(bfqq) exists as well.
5130	 */
5131	spin_lock_irq(&q->queue_lock);
5132	if (idle_timer_disabled)
5133		/*
5134		 * Since the idle timer has been disabled,
5135		 * in_serv_queue contained some request when
5136		 * __bfq_dispatch_request was invoked above, which
5137		 * implies that rq was picked exactly from
5138		 * in_serv_queue. Thus in_serv_queue == bfqq, and is
5139		 * therefore guaranteed to exist because of the above
5140		 * arguments.
5141		 */
5142		bfqg_stats_update_idle_time(bfqq_group(in_serv_queue));
5143	if (bfqq) {
5144		struct bfq_group *bfqg = bfqq_group(bfqq);
5145
5146		bfqg_stats_update_avg_queue_size(bfqg);
5147		bfqg_stats_set_start_empty_time(bfqg);
5148		bfqg_stats_update_io_remove(bfqg, rq->cmd_flags);
5149	}
5150	spin_unlock_irq(&q->queue_lock);
5151}
5152#else
5153static inline void bfq_update_dispatch_stats(struct request_queue *q,
5154					     struct request *rq,
5155					     struct bfq_queue *in_serv_queue,
5156					     bool idle_timer_disabled) {}
5157#endif /* CONFIG_BFQ_CGROUP_DEBUG */
5158
5159static struct request *bfq_dispatch_request(struct blk_mq_hw_ctx *hctx)
5160{
5161	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
5162	struct request *rq;
5163	struct bfq_queue *in_serv_queue;
5164	bool waiting_rq, idle_timer_disabled = false;
5165
5166	spin_lock_irq(&bfqd->lock);
5167
5168	in_serv_queue = bfqd->in_service_queue;
5169	waiting_rq = in_serv_queue && bfq_bfqq_wait_request(in_serv_queue);
5170
5171	rq = __bfq_dispatch_request(hctx);
5172	if (in_serv_queue == bfqd->in_service_queue) {
5173		idle_timer_disabled =
5174			waiting_rq && !bfq_bfqq_wait_request(in_serv_queue);
5175	}
5176
5177	spin_unlock_irq(&bfqd->lock);
5178	bfq_update_dispatch_stats(hctx->queue, rq,
5179			idle_timer_disabled ? in_serv_queue : NULL,
5180				idle_timer_disabled);
5181
5182	return rq;
5183}
5184
5185/*
5186 * Task holds one reference to the queue, dropped when task exits.  Each rq
5187 * in-flight on this queue also holds a reference, dropped when rq is freed.
5188 *
5189 * Scheduler lock must be held here. Recall not to use bfqq after calling
5190 * this function on it.
5191 */
5192void bfq_put_queue(struct bfq_queue *bfqq)
5193{
5194	struct bfq_queue *item;
5195	struct hlist_node *n;
5196	struct bfq_group *bfqg = bfqq_group(bfqq);
5197
5198	bfq_log_bfqq(bfqq->bfqd, bfqq, "put_queue: %p %d", bfqq, bfqq->ref);
 
 
5199
5200	bfqq->ref--;
5201	if (bfqq->ref)
5202		return;
5203
5204	if (!hlist_unhashed(&bfqq->burst_list_node)) {
5205		hlist_del_init(&bfqq->burst_list_node);
5206		/*
5207		 * Decrement also burst size after the removal, if the
5208		 * process associated with bfqq is exiting, and thus
5209		 * does not contribute to the burst any longer. This
5210		 * decrement helps filter out false positives of large
5211		 * bursts, when some short-lived process (often due to
5212		 * the execution of commands by some service) happens
5213		 * to start and exit while a complex application is
5214		 * starting, and thus spawning several processes that
5215		 * do I/O (and that *must not* be treated as a large
5216		 * burst, see comments on bfq_handle_burst).
5217		 *
5218		 * In particular, the decrement is performed only if:
5219		 * 1) bfqq is not a merged queue, because, if it is,
5220		 * then this free of bfqq is not triggered by the exit
5221		 * of the process bfqq is associated with, but exactly
5222		 * by the fact that bfqq has just been merged.
5223		 * 2) burst_size is greater than 0, to handle
5224		 * unbalanced decrements. Unbalanced decrements may
5225		 * happen in te following case: bfqq is inserted into
5226		 * the current burst list--without incrementing
5227		 * bust_size--because of a split, but the current
5228		 * burst list is not the burst list bfqq belonged to
5229		 * (see comments on the case of a split in
5230		 * bfq_set_request).
5231		 */
5232		if (bfqq->bic && bfqq->bfqd->burst_size > 0)
5233			bfqq->bfqd->burst_size--;
5234	}
5235
5236	/*
5237	 * bfqq does not exist any longer, so it cannot be woken by
5238	 * any other queue, and cannot wake any other queue. Then bfqq
5239	 * must be removed from the woken list of its possible waker
5240	 * queue, and all queues in the woken list of bfqq must stop
5241	 * having a waker queue. Strictly speaking, these updates
5242	 * should be performed when bfqq remains with no I/O source
5243	 * attached to it, which happens before bfqq gets freed. In
5244	 * particular, this happens when the last process associated
5245	 * with bfqq exits or gets associated with a different
5246	 * queue. However, both events lead to bfqq being freed soon,
5247	 * and dangling references would come out only after bfqq gets
5248	 * freed. So these updates are done here, as a simple and safe
5249	 * way to handle all cases.
5250	 */
5251	/* remove bfqq from woken list */
5252	if (!hlist_unhashed(&bfqq->woken_list_node))
5253		hlist_del_init(&bfqq->woken_list_node);
5254
5255	/* reset waker for all queues in woken list */
5256	hlist_for_each_entry_safe(item, n, &bfqq->woken_list,
5257				  woken_list_node) {
5258		item->waker_bfqq = NULL;
 
5259		hlist_del_init(&item->woken_list_node);
5260	}
5261
5262	if (bfqq->bfqd->last_completed_rq_bfqq == bfqq)
5263		bfqq->bfqd->last_completed_rq_bfqq = NULL;
5264
5265	kmem_cache_free(bfq_pool, bfqq);
5266	bfqg_and_blkg_put(bfqg);
5267}
5268
5269static void bfq_put_stable_ref(struct bfq_queue *bfqq)
5270{
5271	bfqq->stable_ref--;
5272	bfq_put_queue(bfqq);
5273}
5274
5275void bfq_put_cooperator(struct bfq_queue *bfqq)
5276{
5277	struct bfq_queue *__bfqq, *next;
5278
5279	/*
5280	 * If this queue was scheduled to merge with another queue, be
5281	 * sure to drop the reference taken on that queue (and others in
5282	 * the merge chain). See bfq_setup_merge and bfq_merge_bfqqs.
5283	 */
5284	__bfqq = bfqq->new_bfqq;
5285	while (__bfqq) {
5286		if (__bfqq == bfqq)
5287			break;
5288		next = __bfqq->new_bfqq;
5289		bfq_put_queue(__bfqq);
5290		__bfqq = next;
5291	}
5292}
5293
5294static void bfq_exit_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq)
5295{
5296	if (bfqq == bfqd->in_service_queue) {
5297		__bfq_bfqq_expire(bfqd, bfqq, BFQQE_BUDGET_TIMEOUT);
5298		bfq_schedule_dispatch(bfqd);
5299	}
5300
5301	bfq_log_bfqq(bfqd, bfqq, "exit_bfqq: %p, %d", bfqq, bfqq->ref);
5302
5303	bfq_put_cooperator(bfqq);
5304
5305	bfq_release_process_ref(bfqd, bfqq);
5306}
5307
5308static void bfq_exit_icq_bfqq(struct bfq_io_cq *bic, bool is_sync)
5309{
5310	struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync);
5311	struct bfq_data *bfqd;
5312
5313	if (bfqq)
5314		bfqd = bfqq->bfqd; /* NULL if scheduler already exited */
5315
5316	if (bfqq && bfqd) {
5317		unsigned long flags;
5318
5319		spin_lock_irqsave(&bfqd->lock, flags);
5320		bic_set_bfqq(bic, NULL, is_sync);
5321		bfq_exit_bfqq(bfqd, bfqq);
 
5322		spin_unlock_irqrestore(&bfqd->lock, flags);
5323	}
5324}
5325
5326static void bfq_exit_icq(struct io_cq *icq)
5327{
5328	struct bfq_io_cq *bic = icq_to_bic(icq);
5329
5330	if (bic->stable_merge_bfqq) {
5331		struct bfq_data *bfqd = bic->stable_merge_bfqq->bfqd;
5332
5333		/*
5334		 * bfqd is NULL if scheduler already exited, and in
5335		 * that case this is the last time bfqq is accessed.
5336		 */
5337		if (bfqd) {
5338			unsigned long flags;
5339
5340			spin_lock_irqsave(&bfqd->lock, flags);
5341			bfq_put_stable_ref(bic->stable_merge_bfqq);
5342			spin_unlock_irqrestore(&bfqd->lock, flags);
5343		} else {
5344			bfq_put_stable_ref(bic->stable_merge_bfqq);
5345		}
5346	}
5347
5348	bfq_exit_icq_bfqq(bic, true);
5349	bfq_exit_icq_bfqq(bic, false);
5350}
5351
5352/*
5353 * Update the entity prio values; note that the new values will not
5354 * be used until the next (re)activation.
5355 */
5356static void
5357bfq_set_next_ioprio_data(struct bfq_queue *bfqq, struct bfq_io_cq *bic)
5358{
5359	struct task_struct *tsk = current;
5360	int ioprio_class;
5361	struct bfq_data *bfqd = bfqq->bfqd;
5362
5363	if (!bfqd)
5364		return;
5365
5366	ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio);
5367	switch (ioprio_class) {
5368	default:
5369		pr_err("bdi %s: bfq: bad prio class %d\n",
5370			bdi_dev_name(bfqq->bfqd->queue->disk->bdi),
5371			ioprio_class);
5372		fallthrough;
5373	case IOPRIO_CLASS_NONE:
5374		/*
5375		 * No prio set, inherit CPU scheduling settings.
5376		 */
5377		bfqq->new_ioprio = task_nice_ioprio(tsk);
5378		bfqq->new_ioprio_class = task_nice_ioclass(tsk);
5379		break;
5380	case IOPRIO_CLASS_RT:
5381		bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
5382		bfqq->new_ioprio_class = IOPRIO_CLASS_RT;
5383		break;
5384	case IOPRIO_CLASS_BE:
5385		bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
5386		bfqq->new_ioprio_class = IOPRIO_CLASS_BE;
5387		break;
5388	case IOPRIO_CLASS_IDLE:
5389		bfqq->new_ioprio_class = IOPRIO_CLASS_IDLE;
5390		bfqq->new_ioprio = 7;
5391		break;
5392	}
5393
5394	if (bfqq->new_ioprio >= IOPRIO_NR_LEVELS) {
5395		pr_crit("bfq_set_next_ioprio_data: new_ioprio %d\n",
5396			bfqq->new_ioprio);
5397		bfqq->new_ioprio = IOPRIO_NR_LEVELS - 1;
5398	}
5399
5400	bfqq->entity.new_weight = bfq_ioprio_to_weight(bfqq->new_ioprio);
5401	bfq_log_bfqq(bfqd, bfqq, "new_ioprio %d new_weight %d",
5402		     bfqq->new_ioprio, bfqq->entity.new_weight);
5403	bfqq->entity.prio_changed = 1;
5404}
5405
5406static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd,
5407				       struct bio *bio, bool is_sync,
5408				       struct bfq_io_cq *bic,
5409				       bool respawn);
5410
5411static void bfq_check_ioprio_change(struct bfq_io_cq *bic, struct bio *bio)
5412{
5413	struct bfq_data *bfqd = bic_to_bfqd(bic);
5414	struct bfq_queue *bfqq;
5415	int ioprio = bic->icq.ioc->ioprio;
5416
5417	/*
5418	 * This condition may trigger on a newly created bic, be sure to
5419	 * drop the lock before returning.
5420	 */
5421	if (unlikely(!bfqd) || likely(bic->ioprio == ioprio))
5422		return;
5423
5424	bic->ioprio = ioprio;
5425
5426	bfqq = bic_to_bfqq(bic, false);
5427	if (bfqq) {
5428		struct bfq_queue *old_bfqq = bfqq;
5429
5430		bfqq = bfq_get_queue(bfqd, bio, false, bic, true);
5431		bic_set_bfqq(bic, bfqq, false);
5432		bfq_release_process_ref(bfqd, old_bfqq);
5433	}
5434
5435	bfqq = bic_to_bfqq(bic, true);
5436	if (bfqq)
5437		bfq_set_next_ioprio_data(bfqq, bic);
5438}
5439
5440static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5441			  struct bfq_io_cq *bic, pid_t pid, int is_sync)
5442{
5443	u64 now_ns = ktime_get_ns();
5444
5445	RB_CLEAR_NODE(&bfqq->entity.rb_node);
5446	INIT_LIST_HEAD(&bfqq->fifo);
5447	INIT_HLIST_NODE(&bfqq->burst_list_node);
5448	INIT_HLIST_NODE(&bfqq->woken_list_node);
5449	INIT_HLIST_HEAD(&bfqq->woken_list);
5450
5451	bfqq->ref = 0;
5452	bfqq->bfqd = bfqd;
5453
5454	if (bic)
5455		bfq_set_next_ioprio_data(bfqq, bic);
5456
5457	if (is_sync) {
5458		/*
5459		 * No need to mark as has_short_ttime if in
5460		 * idle_class, because no device idling is performed
5461		 * for queues in idle class
5462		 */
5463		if (!bfq_class_idle(bfqq))
5464			/* tentatively mark as has_short_ttime */
5465			bfq_mark_bfqq_has_short_ttime(bfqq);
5466		bfq_mark_bfqq_sync(bfqq);
5467		bfq_mark_bfqq_just_created(bfqq);
5468	} else
5469		bfq_clear_bfqq_sync(bfqq);
5470
5471	/* set end request to minus infinity from now */
5472	bfqq->ttime.last_end_request = now_ns + 1;
5473
5474	bfqq->creation_time = jiffies;
5475
5476	bfqq->io_start_time = now_ns;
5477
5478	bfq_mark_bfqq_IO_bound(bfqq);
5479
5480	bfqq->pid = pid;
5481
5482	/* Tentative initial value to trade off between thr and lat */
5483	bfqq->max_budget = (2 * bfq_max_budget(bfqd)) / 3;
5484	bfqq->budget_timeout = bfq_smallest_from_now();
5485
5486	bfqq->wr_coeff = 1;
5487	bfqq->last_wr_start_finish = jiffies;
5488	bfqq->wr_start_at_switch_to_srt = bfq_smallest_from_now();
5489	bfqq->split_time = bfq_smallest_from_now();
5490
5491	/*
5492	 * To not forget the possibly high bandwidth consumed by a
5493	 * process/queue in the recent past,
5494	 * bfq_bfqq_softrt_next_start() returns a value at least equal
5495	 * to the current value of bfqq->soft_rt_next_start (see
5496	 * comments on bfq_bfqq_softrt_next_start).  Set
5497	 * soft_rt_next_start to now, to mean that bfqq has consumed
5498	 * no bandwidth so far.
5499	 */
5500	bfqq->soft_rt_next_start = jiffies;
5501
5502	/* first request is almost certainly seeky */
5503	bfqq->seek_history = 1;
5504}
5505
5506static struct bfq_queue **bfq_async_queue_prio(struct bfq_data *bfqd,
5507					       struct bfq_group *bfqg,
5508					       int ioprio_class, int ioprio)
5509{
5510	switch (ioprio_class) {
5511	case IOPRIO_CLASS_RT:
5512		return &bfqg->async_bfqq[0][ioprio];
5513	case IOPRIO_CLASS_NONE:
5514		ioprio = IOPRIO_BE_NORM;
5515		fallthrough;
5516	case IOPRIO_CLASS_BE:
5517		return &bfqg->async_bfqq[1][ioprio];
5518	case IOPRIO_CLASS_IDLE:
5519		return &bfqg->async_idle_bfqq;
5520	default:
5521		return NULL;
5522	}
5523}
5524
5525static struct bfq_queue *
5526bfq_do_early_stable_merge(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5527			  struct bfq_io_cq *bic,
5528			  struct bfq_queue *last_bfqq_created)
5529{
5530	struct bfq_queue *new_bfqq =
5531		bfq_setup_merge(bfqq, last_bfqq_created);
5532
5533	if (!new_bfqq)
5534		return bfqq;
5535
5536	if (new_bfqq->bic)
5537		new_bfqq->bic->stably_merged = true;
5538	bic->stably_merged = true;
5539
5540	/*
5541	 * Reusing merge functions. This implies that
5542	 * bfqq->bic must be set too, for
5543	 * bfq_merge_bfqqs to correctly save bfqq's
5544	 * state before killing it.
5545	 */
5546	bfqq->bic = bic;
5547	bfq_merge_bfqqs(bfqd, bic, bfqq, new_bfqq);
5548
5549	return new_bfqq;
5550}
5551
5552/*
5553 * Many throughput-sensitive workloads are made of several parallel
5554 * I/O flows, with all flows generated by the same application, or
5555 * more generically by the same task (e.g., system boot). The most
5556 * counterproductive action with these workloads is plugging I/O
5557 * dispatch when one of the bfq_queues associated with these flows
5558 * remains temporarily empty.
5559 *
5560 * To avoid this plugging, BFQ has been using a burst-handling
5561 * mechanism for years now. This mechanism has proven effective for
5562 * throughput, and not detrimental for service guarantees. The
5563 * following function pushes this mechanism a little bit further,
5564 * basing on the following two facts.
5565 *
5566 * First, all the I/O flows of a the same application or task
5567 * contribute to the execution/completion of that common application
5568 * or task. So the performance figures that matter are total
5569 * throughput of the flows and task-wide I/O latency.  In particular,
5570 * these flows do not need to be protected from each other, in terms
5571 * of individual bandwidth or latency.
5572 *
5573 * Second, the above fact holds regardless of the number of flows.
5574 *
5575 * Putting these two facts together, this commits merges stably the
5576 * bfq_queues associated with these I/O flows, i.e., with the
5577 * processes that generate these IO/ flows, regardless of how many the
5578 * involved processes are.
5579 *
5580 * To decide whether a set of bfq_queues is actually associated with
5581 * the I/O flows of a common application or task, and to merge these
5582 * queues stably, this function operates as follows: given a bfq_queue,
5583 * say Q2, currently being created, and the last bfq_queue, say Q1,
5584 * created before Q2, Q2 is merged stably with Q1 if
5585 * - very little time has elapsed since when Q1 was created
5586 * - Q2 has the same ioprio as Q1
5587 * - Q2 belongs to the same group as Q1
5588 *
5589 * Merging bfq_queues also reduces scheduling overhead. A fio test
5590 * with ten random readers on /dev/nullb shows a throughput boost of
5591 * 40%, with a quadcore. Since BFQ's execution time amounts to ~50% of
5592 * the total per-request processing time, the above throughput boost
5593 * implies that BFQ's overhead is reduced by more than 50%.
5594 *
5595 * This new mechanism most certainly obsoletes the current
5596 * burst-handling heuristics. We keep those heuristics for the moment.
5597 */
5598static struct bfq_queue *bfq_do_or_sched_stable_merge(struct bfq_data *bfqd,
5599						      struct bfq_queue *bfqq,
5600						      struct bfq_io_cq *bic)
5601{
5602	struct bfq_queue **source_bfqq = bfqq->entity.parent ?
5603		&bfqq->entity.parent->last_bfqq_created :
5604		&bfqd->last_bfqq_created;
5605
5606	struct bfq_queue *last_bfqq_created = *source_bfqq;
5607
5608	/*
5609	 * If last_bfqq_created has not been set yet, then init it. If
5610	 * it has been set already, but too long ago, then move it
5611	 * forward to bfqq. Finally, move also if bfqq belongs to a
5612	 * different group than last_bfqq_created, or if bfqq has a
5613	 * different ioprio or ioprio_class. If none of these
5614	 * conditions holds true, then try an early stable merge or
5615	 * schedule a delayed stable merge.
5616	 *
5617	 * A delayed merge is scheduled (instead of performing an
5618	 * early merge), in case bfqq might soon prove to be more
5619	 * throughput-beneficial if not merged. Currently this is
5620	 * possible only if bfqd is rotational with no queueing. For
5621	 * such a drive, not merging bfqq is better for throughput if
5622	 * bfqq happens to contain sequential I/O. So, we wait a
5623	 * little bit for enough I/O to flow through bfqq. After that,
5624	 * if such an I/O is sequential, then the merge is
5625	 * canceled. Otherwise the merge is finally performed.
5626	 */
5627	if (!last_bfqq_created ||
5628	    time_before(last_bfqq_created->creation_time +
5629			msecs_to_jiffies(bfq_activation_stable_merging),
5630			bfqq->creation_time) ||
5631		bfqq->entity.parent != last_bfqq_created->entity.parent ||
5632		bfqq->ioprio != last_bfqq_created->ioprio ||
5633		bfqq->ioprio_class != last_bfqq_created->ioprio_class)
5634		*source_bfqq = bfqq;
5635	else if (time_after_eq(last_bfqq_created->creation_time +
5636				 bfqd->bfq_burst_interval,
5637				 bfqq->creation_time)) {
5638		if (likely(bfqd->nonrot_with_queueing))
5639			/*
5640			 * With this type of drive, leaving
5641			 * bfqq alone may provide no
5642			 * throughput benefits compared with
5643			 * merging bfqq. So merge bfqq now.
5644			 */
5645			bfqq = bfq_do_early_stable_merge(bfqd, bfqq,
5646							 bic,
5647							 last_bfqq_created);
5648		else { /* schedule tentative stable merge */
5649			/*
5650			 * get reference on last_bfqq_created,
5651			 * to prevent it from being freed,
5652			 * until we decide whether to merge
5653			 */
5654			last_bfqq_created->ref++;
5655			/*
5656			 * need to keep track of stable refs, to
5657			 * compute process refs correctly
5658			 */
5659			last_bfqq_created->stable_ref++;
5660			/*
5661			 * Record the bfqq to merge to.
5662			 */
5663			bic->stable_merge_bfqq = last_bfqq_created;
5664		}
5665	}
5666
5667	return bfqq;
5668}
5669
5670
5671static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd,
5672				       struct bio *bio, bool is_sync,
5673				       struct bfq_io_cq *bic,
5674				       bool respawn)
5675{
5676	const int ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
5677	const int ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio);
5678	struct bfq_queue **async_bfqq = NULL;
5679	struct bfq_queue *bfqq;
5680	struct bfq_group *bfqg;
5681
5682	bfqg = bfq_bio_bfqg(bfqd, bio);
 
 
 
 
 
 
 
5683	if (!is_sync) {
5684		async_bfqq = bfq_async_queue_prio(bfqd, bfqg, ioprio_class,
5685						  ioprio);
5686		bfqq = *async_bfqq;
5687		if (bfqq)
5688			goto out;
5689	}
5690
5691	bfqq = kmem_cache_alloc_node(bfq_pool,
5692				     GFP_NOWAIT | __GFP_ZERO | __GFP_NOWARN,
5693				     bfqd->queue->node);
5694
5695	if (bfqq) {
5696		bfq_init_bfqq(bfqd, bfqq, bic, current->pid,
5697			      is_sync);
5698		bfq_init_entity(&bfqq->entity, bfqg);
5699		bfq_log_bfqq(bfqd, bfqq, "allocated");
5700	} else {
5701		bfqq = &bfqd->oom_bfqq;
5702		bfq_log_bfqq(bfqd, bfqq, "using oom bfqq");
5703		goto out;
5704	}
5705
5706	/*
5707	 * Pin the queue now that it's allocated, scheduler exit will
5708	 * prune it.
5709	 */
5710	if (async_bfqq) {
5711		bfqq->ref++; /*
5712			      * Extra group reference, w.r.t. sync
5713			      * queue. This extra reference is removed
5714			      * only if bfqq->bfqg disappears, to
5715			      * guarantee that this queue is not freed
5716			      * until its group goes away.
5717			      */
5718		bfq_log_bfqq(bfqd, bfqq, "get_queue, bfqq not in async: %p, %d",
5719			     bfqq, bfqq->ref);
5720		*async_bfqq = bfqq;
5721	}
5722
5723out:
5724	bfqq->ref++; /* get a process reference to this queue */
5725
5726	if (bfqq != &bfqd->oom_bfqq && is_sync && !respawn)
5727		bfqq = bfq_do_or_sched_stable_merge(bfqd, bfqq, bic);
5728	return bfqq;
5729}
5730
5731static void bfq_update_io_thinktime(struct bfq_data *bfqd,
5732				    struct bfq_queue *bfqq)
5733{
5734	struct bfq_ttime *ttime = &bfqq->ttime;
5735	u64 elapsed;
5736
5737	/*
5738	 * We are really interested in how long it takes for the queue to
5739	 * become busy when there is no outstanding IO for this queue. So
5740	 * ignore cases when the bfq queue has already IO queued.
5741	 */
5742	if (bfqq->dispatched || bfq_bfqq_busy(bfqq))
5743		return;
5744	elapsed = ktime_get_ns() - bfqq->ttime.last_end_request;
5745	elapsed = min_t(u64, elapsed, 2ULL * bfqd->bfq_slice_idle);
5746
5747	ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8;
5748	ttime->ttime_total = div_u64(7*ttime->ttime_total + 256*elapsed,  8);
5749	ttime->ttime_mean = div64_ul(ttime->ttime_total + 128,
5750				     ttime->ttime_samples);
5751}
5752
5753static void
5754bfq_update_io_seektime(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5755		       struct request *rq)
5756{
5757	bfqq->seek_history <<= 1;
5758	bfqq->seek_history |= BFQ_RQ_SEEKY(bfqd, bfqq->last_request_pos, rq);
5759
5760	if (bfqq->wr_coeff > 1 &&
5761	    bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
5762	    BFQQ_TOTALLY_SEEKY(bfqq)) {
5763		if (time_is_before_jiffies(bfqq->wr_start_at_switch_to_srt +
5764					   bfq_wr_duration(bfqd))) {
5765			/*
5766			 * In soft_rt weight raising with the
5767			 * interactive-weight-raising period
5768			 * elapsed (so no switch back to
5769			 * interactive weight raising).
5770			 */
5771			bfq_bfqq_end_wr(bfqq);
5772		} else { /*
5773			  * stopping soft_rt weight raising
5774			  * while still in interactive period,
5775			  * switch back to interactive weight
5776			  * raising
5777			  */
5778			switch_back_to_interactive_wr(bfqq, bfqd);
5779			bfqq->entity.prio_changed = 1;
5780		}
5781	}
5782}
5783
5784static void bfq_update_has_short_ttime(struct bfq_data *bfqd,
5785				       struct bfq_queue *bfqq,
5786				       struct bfq_io_cq *bic)
5787{
5788	bool has_short_ttime = true, state_changed;
5789
5790	/*
5791	 * No need to update has_short_ttime if bfqq is async or in
5792	 * idle io prio class, or if bfq_slice_idle is zero, because
5793	 * no device idling is performed for bfqq in this case.
5794	 */
5795	if (!bfq_bfqq_sync(bfqq) || bfq_class_idle(bfqq) ||
5796	    bfqd->bfq_slice_idle == 0)
5797		return;
5798
5799	/* Idle window just restored, statistics are meaningless. */
5800	if (time_is_after_eq_jiffies(bfqq->split_time +
5801				     bfqd->bfq_wr_min_idle_time))
5802		return;
5803
5804	/* Think time is infinite if no process is linked to
5805	 * bfqq. Otherwise check average think time to decide whether
5806	 * to mark as has_short_ttime. To this goal, compare average
5807	 * think time with half the I/O-plugging timeout.
5808	 */
5809	if (atomic_read(&bic->icq.ioc->active_ref) == 0 ||
5810	    (bfq_sample_valid(bfqq->ttime.ttime_samples) &&
5811	     bfqq->ttime.ttime_mean > bfqd->bfq_slice_idle>>1))
5812		has_short_ttime = false;
5813
5814	state_changed = has_short_ttime != bfq_bfqq_has_short_ttime(bfqq);
5815
5816	if (has_short_ttime)
5817		bfq_mark_bfqq_has_short_ttime(bfqq);
5818	else
5819		bfq_clear_bfqq_has_short_ttime(bfqq);
5820
5821	/*
5822	 * Until the base value for the total service time gets
5823	 * finally computed for bfqq, the inject limit does depend on
5824	 * the think-time state (short|long). In particular, the limit
5825	 * is 0 or 1 if the think time is deemed, respectively, as
5826	 * short or long (details in the comments in
5827	 * bfq_update_inject_limit()). Accordingly, the next
5828	 * instructions reset the inject limit if the think-time state
5829	 * has changed and the above base value is still to be
5830	 * computed.
5831	 *
5832	 * However, the reset is performed only if more than 100 ms
5833	 * have elapsed since the last update of the inject limit, or
5834	 * (inclusive) if the change is from short to long think
5835	 * time. The reason for this waiting is as follows.
5836	 *
5837	 * bfqq may have a long think time because of a
5838	 * synchronization with some other queue, i.e., because the
5839	 * I/O of some other queue may need to be completed for bfqq
5840	 * to receive new I/O. Details in the comments on the choice
5841	 * of the queue for injection in bfq_select_queue().
5842	 *
5843	 * As stressed in those comments, if such a synchronization is
5844	 * actually in place, then, without injection on bfqq, the
5845	 * blocking I/O cannot happen to served while bfqq is in
5846	 * service. As a consequence, if bfqq is granted
5847	 * I/O-dispatch-plugging, then bfqq remains empty, and no I/O
5848	 * is dispatched, until the idle timeout fires. This is likely
5849	 * to result in lower bandwidth and higher latencies for bfqq,
5850	 * and in a severe loss of total throughput.
5851	 *
5852	 * On the opposite end, a non-zero inject limit may allow the
5853	 * I/O that blocks bfqq to be executed soon, and therefore
5854	 * bfqq to receive new I/O soon.
5855	 *
5856	 * But, if the blocking gets actually eliminated, then the
5857	 * next think-time sample for bfqq may be very low. This in
5858	 * turn may cause bfqq's think time to be deemed
5859	 * short. Without the 100 ms barrier, this new state change
5860	 * would cause the body of the next if to be executed
5861	 * immediately. But this would set to 0 the inject
5862	 * limit. Without injection, the blocking I/O would cause the
5863	 * think time of bfqq to become long again, and therefore the
5864	 * inject limit to be raised again, and so on. The only effect
5865	 * of such a steady oscillation between the two think-time
5866	 * states would be to prevent effective injection on bfqq.
5867	 *
5868	 * In contrast, if the inject limit is not reset during such a
5869	 * long time interval as 100 ms, then the number of short
5870	 * think time samples can grow significantly before the reset
5871	 * is performed. As a consequence, the think time state can
5872	 * become stable before the reset. Therefore there will be no
5873	 * state change when the 100 ms elapse, and no reset of the
5874	 * inject limit. The inject limit remains steadily equal to 1
5875	 * both during and after the 100 ms. So injection can be
5876	 * performed at all times, and throughput gets boosted.
5877	 *
5878	 * An inject limit equal to 1 is however in conflict, in
5879	 * general, with the fact that the think time of bfqq is
5880	 * short, because injection may be likely to delay bfqq's I/O
5881	 * (as explained in the comments in
5882	 * bfq_update_inject_limit()). But this does not happen in
5883	 * this special case, because bfqq's low think time is due to
5884	 * an effective handling of a synchronization, through
5885	 * injection. In this special case, bfqq's I/O does not get
5886	 * delayed by injection; on the contrary, bfqq's I/O is
5887	 * brought forward, because it is not blocked for
5888	 * milliseconds.
5889	 *
5890	 * In addition, serving the blocking I/O much sooner, and much
5891	 * more frequently than once per I/O-plugging timeout, makes
5892	 * it much quicker to detect a waker queue (the concept of
5893	 * waker queue is defined in the comments in
5894	 * bfq_add_request()). This makes it possible to start sooner
5895	 * to boost throughput more effectively, by injecting the I/O
5896	 * of the waker queue unconditionally on every
5897	 * bfq_dispatch_request().
5898	 *
5899	 * One last, important benefit of not resetting the inject
5900	 * limit before 100 ms is that, during this time interval, the
5901	 * base value for the total service time is likely to get
5902	 * finally computed for bfqq, freeing the inject limit from
5903	 * its relation with the think time.
5904	 */
5905	if (state_changed && bfqq->last_serv_time_ns == 0 &&
5906	    (time_is_before_eq_jiffies(bfqq->decrease_time_jif +
5907				      msecs_to_jiffies(100)) ||
5908	     !has_short_ttime))
5909		bfq_reset_inject_limit(bfqd, bfqq);
5910}
5911
5912/*
5913 * Called when a new fs request (rq) is added to bfqq.  Check if there's
5914 * something we should do about it.
5915 */
5916static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5917			    struct request *rq)
5918{
5919	if (rq->cmd_flags & REQ_META)
5920		bfqq->meta_pending++;
5921
5922	bfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
5923
5924	if (bfqq == bfqd->in_service_queue && bfq_bfqq_wait_request(bfqq)) {
5925		bool small_req = bfqq->queued[rq_is_sync(rq)] == 1 &&
5926				 blk_rq_sectors(rq) < 32;
5927		bool budget_timeout = bfq_bfqq_budget_timeout(bfqq);
5928
5929		/*
5930		 * There is just this request queued: if
5931		 * - the request is small, and
5932		 * - we are idling to boost throughput, and
5933		 * - the queue is not to be expired,
5934		 * then just exit.
5935		 *
5936		 * In this way, if the device is being idled to wait
5937		 * for a new request from the in-service queue, we
5938		 * avoid unplugging the device and committing the
5939		 * device to serve just a small request. In contrast
5940		 * we wait for the block layer to decide when to
5941		 * unplug the device: hopefully, new requests will be
5942		 * merged to this one quickly, then the device will be
5943		 * unplugged and larger requests will be dispatched.
5944		 */
5945		if (small_req && idling_boosts_thr_without_issues(bfqd, bfqq) &&
5946		    !budget_timeout)
5947			return;
5948
5949		/*
5950		 * A large enough request arrived, or idling is being
5951		 * performed to preserve service guarantees, or
5952		 * finally the queue is to be expired: in all these
5953		 * cases disk idling is to be stopped, so clear
5954		 * wait_request flag and reset timer.
5955		 */
5956		bfq_clear_bfqq_wait_request(bfqq);
5957		hrtimer_try_to_cancel(&bfqd->idle_slice_timer);
5958
5959		/*
5960		 * The queue is not empty, because a new request just
5961		 * arrived. Hence we can safely expire the queue, in
5962		 * case of budget timeout, without risking that the
5963		 * timestamps of the queue are not updated correctly.
5964		 * See [1] for more details.
5965		 */
5966		if (budget_timeout)
5967			bfq_bfqq_expire(bfqd, bfqq, false,
5968					BFQQE_BUDGET_TIMEOUT);
5969	}
5970}
5971
5972static void bfqq_request_allocated(struct bfq_queue *bfqq)
5973{
5974	struct bfq_entity *entity = &bfqq->entity;
5975
5976	for_each_entity(entity)
5977		entity->allocated++;
5978}
5979
5980static void bfqq_request_freed(struct bfq_queue *bfqq)
5981{
5982	struct bfq_entity *entity = &bfqq->entity;
5983
5984	for_each_entity(entity)
5985		entity->allocated--;
5986}
5987
5988/* returns true if it causes the idle timer to be disabled */
5989static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
5990{
5991	struct bfq_queue *bfqq = RQ_BFQQ(rq),
5992		*new_bfqq = bfq_setup_cooperator(bfqd, bfqq, rq, true,
5993						 RQ_BIC(rq));
5994	bool waiting, idle_timer_disabled = false;
5995
5996	if (new_bfqq) {
5997		/*
5998		 * Release the request's reference to the old bfqq
5999		 * and make sure one is taken to the shared queue.
6000		 */
6001		bfqq_request_allocated(new_bfqq);
6002		bfqq_request_freed(bfqq);
6003		new_bfqq->ref++;
6004		/*
6005		 * If the bic associated with the process
6006		 * issuing this request still points to bfqq
6007		 * (and thus has not been already redirected
6008		 * to new_bfqq or even some other bfq_queue),
6009		 * then complete the merge and redirect it to
6010		 * new_bfqq.
6011		 */
6012		if (bic_to_bfqq(RQ_BIC(rq), 1) == bfqq)
6013			bfq_merge_bfqqs(bfqd, RQ_BIC(rq),
6014					bfqq, new_bfqq);
6015
6016		bfq_clear_bfqq_just_created(bfqq);
6017		/*
6018		 * rq is about to be enqueued into new_bfqq,
6019		 * release rq reference on bfqq
6020		 */
6021		bfq_put_queue(bfqq);
6022		rq->elv.priv[1] = new_bfqq;
6023		bfqq = new_bfqq;
6024	}
6025
6026	bfq_update_io_thinktime(bfqd, bfqq);
6027	bfq_update_has_short_ttime(bfqd, bfqq, RQ_BIC(rq));
6028	bfq_update_io_seektime(bfqd, bfqq, rq);
6029
6030	waiting = bfqq && bfq_bfqq_wait_request(bfqq);
6031	bfq_add_request(rq);
6032	idle_timer_disabled = waiting && !bfq_bfqq_wait_request(bfqq);
6033
6034	rq->fifo_time = ktime_get_ns() + bfqd->bfq_fifo_expire[rq_is_sync(rq)];
6035	list_add_tail(&rq->queuelist, &bfqq->fifo);
6036
6037	bfq_rq_enqueued(bfqd, bfqq, rq);
6038
6039	return idle_timer_disabled;
6040}
6041
6042#ifdef CONFIG_BFQ_CGROUP_DEBUG
6043static void bfq_update_insert_stats(struct request_queue *q,
6044				    struct bfq_queue *bfqq,
6045				    bool idle_timer_disabled,
6046				    blk_opf_t cmd_flags)
6047{
6048	if (!bfqq)
6049		return;
6050
6051	/*
6052	 * bfqq still exists, because it can disappear only after
6053	 * either it is merged with another queue, or the process it
6054	 * is associated with exits. But both actions must be taken by
6055	 * the same process currently executing this flow of
6056	 * instructions.
6057	 *
6058	 * In addition, the following queue lock guarantees that
6059	 * bfqq_group(bfqq) exists as well.
6060	 */
6061	spin_lock_irq(&q->queue_lock);
6062	bfqg_stats_update_io_add(bfqq_group(bfqq), bfqq, cmd_flags);
6063	if (idle_timer_disabled)
6064		bfqg_stats_update_idle_time(bfqq_group(bfqq));
6065	spin_unlock_irq(&q->queue_lock);
6066}
6067#else
6068static inline void bfq_update_insert_stats(struct request_queue *q,
6069					   struct bfq_queue *bfqq,
6070					   bool idle_timer_disabled,
6071					   blk_opf_t cmd_flags) {}
6072#endif /* CONFIG_BFQ_CGROUP_DEBUG */
6073
6074static struct bfq_queue *bfq_init_rq(struct request *rq);
6075
6076static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
6077			       bool at_head)
6078{
6079	struct request_queue *q = hctx->queue;
6080	struct bfq_data *bfqd = q->elevator->elevator_data;
6081	struct bfq_queue *bfqq;
6082	bool idle_timer_disabled = false;
6083	blk_opf_t cmd_flags;
6084	LIST_HEAD(free);
6085
6086#ifdef CONFIG_BFQ_GROUP_IOSCHED
6087	if (!cgroup_subsys_on_dfl(io_cgrp_subsys) && rq->bio)
6088		bfqg_stats_update_legacy_io(q, rq);
6089#endif
6090	spin_lock_irq(&bfqd->lock);
6091	bfqq = bfq_init_rq(rq);
6092	if (blk_mq_sched_try_insert_merge(q, rq, &free)) {
6093		spin_unlock_irq(&bfqd->lock);
6094		blk_mq_free_requests(&free);
6095		return;
6096	}
6097
6098	trace_block_rq_insert(rq);
6099
6100	if (!bfqq || at_head) {
 
 
 
 
6101		if (at_head)
6102			list_add(&rq->queuelist, &bfqd->dispatch);
6103		else
6104			list_add_tail(&rq->queuelist, &bfqd->dispatch);
6105	} else {
6106		idle_timer_disabled = __bfq_insert_request(bfqd, rq);
6107		/*
6108		 * Update bfqq, because, if a queue merge has occurred
6109		 * in __bfq_insert_request, then rq has been
6110		 * redirected into a new queue.
6111		 */
6112		bfqq = RQ_BFQQ(rq);
6113
6114		if (rq_mergeable(rq)) {
6115			elv_rqhash_add(q, rq);
6116			if (!q->last_merge)
6117				q->last_merge = rq;
6118		}
6119	}
6120
6121	/*
6122	 * Cache cmd_flags before releasing scheduler lock, because rq
6123	 * may disappear afterwards (for example, because of a request
6124	 * merge).
6125	 */
6126	cmd_flags = rq->cmd_flags;
 
6127	spin_unlock_irq(&bfqd->lock);
6128
6129	bfq_update_insert_stats(q, bfqq, idle_timer_disabled,
6130				cmd_flags);
6131}
6132
6133static void bfq_insert_requests(struct blk_mq_hw_ctx *hctx,
6134				struct list_head *list, bool at_head)
6135{
6136	while (!list_empty(list)) {
6137		struct request *rq;
6138
6139		rq = list_first_entry(list, struct request, queuelist);
6140		list_del_init(&rq->queuelist);
6141		bfq_insert_request(hctx, rq, at_head);
6142	}
6143}
6144
6145static void bfq_update_hw_tag(struct bfq_data *bfqd)
6146{
6147	struct bfq_queue *bfqq = bfqd->in_service_queue;
6148
6149	bfqd->max_rq_in_driver = max_t(int, bfqd->max_rq_in_driver,
6150				       bfqd->rq_in_driver);
6151
6152	if (bfqd->hw_tag == 1)
6153		return;
6154
6155	/*
6156	 * This sample is valid if the number of outstanding requests
6157	 * is large enough to allow a queueing behavior.  Note that the
6158	 * sum is not exact, as it's not taking into account deactivated
6159	 * requests.
6160	 */
6161	if (bfqd->rq_in_driver + bfqd->queued <= BFQ_HW_QUEUE_THRESHOLD)
6162		return;
6163
6164	/*
6165	 * If active queue hasn't enough requests and can idle, bfq might not
6166	 * dispatch sufficient requests to hardware. Don't zero hw_tag in this
6167	 * case
6168	 */
6169	if (bfqq && bfq_bfqq_has_short_ttime(bfqq) &&
6170	    bfqq->dispatched + bfqq->queued[0] + bfqq->queued[1] <
6171	    BFQ_HW_QUEUE_THRESHOLD &&
6172	    bfqd->rq_in_driver < BFQ_HW_QUEUE_THRESHOLD)
6173		return;
6174
6175	if (bfqd->hw_tag_samples++ < BFQ_HW_QUEUE_SAMPLES)
6176		return;
6177
6178	bfqd->hw_tag = bfqd->max_rq_in_driver > BFQ_HW_QUEUE_THRESHOLD;
6179	bfqd->max_rq_in_driver = 0;
6180	bfqd->hw_tag_samples = 0;
6181
6182	bfqd->nonrot_with_queueing =
6183		blk_queue_nonrot(bfqd->queue) && bfqd->hw_tag;
6184}
6185
6186static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd)
6187{
6188	u64 now_ns;
6189	u32 delta_us;
6190
6191	bfq_update_hw_tag(bfqd);
6192
6193	bfqd->rq_in_driver--;
6194	bfqq->dispatched--;
6195
6196	if (!bfqq->dispatched && !bfq_bfqq_busy(bfqq)) {
6197		/*
6198		 * Set budget_timeout (which we overload to store the
6199		 * time at which the queue remains with no backlog and
6200		 * no outstanding request; used by the weight-raising
6201		 * mechanism).
6202		 */
6203		bfqq->budget_timeout = jiffies;
6204
6205		bfq_del_bfqq_in_groups_with_pending_reqs(bfqq);
6206		bfq_weights_tree_remove(bfqq);
6207	}
6208
6209	now_ns = ktime_get_ns();
6210
6211	bfqq->ttime.last_end_request = now_ns;
6212
6213	/*
6214	 * Using us instead of ns, to get a reasonable precision in
6215	 * computing rate in next check.
6216	 */
6217	delta_us = div_u64(now_ns - bfqd->last_completion, NSEC_PER_USEC);
6218
6219	/*
6220	 * If the request took rather long to complete, and, according
6221	 * to the maximum request size recorded, this completion latency
6222	 * implies that the request was certainly served at a very low
6223	 * rate (less than 1M sectors/sec), then the whole observation
6224	 * interval that lasts up to this time instant cannot be a
6225	 * valid time interval for computing a new peak rate.  Invoke
6226	 * bfq_update_rate_reset to have the following three steps
6227	 * taken:
6228	 * - close the observation interval at the last (previous)
6229	 *   request dispatch or completion
6230	 * - compute rate, if possible, for that observation interval
6231	 * - reset to zero samples, which will trigger a proper
6232	 *   re-initialization of the observation interval on next
6233	 *   dispatch
6234	 */
6235	if (delta_us > BFQ_MIN_TT/NSEC_PER_USEC &&
6236	   (bfqd->last_rq_max_size<<BFQ_RATE_SHIFT)/delta_us <
6237			1UL<<(BFQ_RATE_SHIFT - 10))
6238		bfq_update_rate_reset(bfqd, NULL);
6239	bfqd->last_completion = now_ns;
6240	/*
6241	 * Shared queues are likely to receive I/O at a high
6242	 * rate. This may deceptively let them be considered as wakers
6243	 * of other queues. But a false waker will unjustly steal
6244	 * bandwidth to its supposedly woken queue. So considering
6245	 * also shared queues in the waking mechanism may cause more
6246	 * control troubles than throughput benefits. Then reset
6247	 * last_completed_rq_bfqq if bfqq is a shared queue.
6248	 */
6249	if (!bfq_bfqq_coop(bfqq))
6250		bfqd->last_completed_rq_bfqq = bfqq;
6251	else
6252		bfqd->last_completed_rq_bfqq = NULL;
6253
6254	/*
6255	 * If we are waiting to discover whether the request pattern
6256	 * of the task associated with the queue is actually
6257	 * isochronous, and both requisites for this condition to hold
6258	 * are now satisfied, then compute soft_rt_next_start (see the
6259	 * comments on the function bfq_bfqq_softrt_next_start()). We
6260	 * do not compute soft_rt_next_start if bfqq is in interactive
6261	 * weight raising (see the comments in bfq_bfqq_expire() for
6262	 * an explanation). We schedule this delayed update when bfqq
6263	 * expires, if it still has in-flight requests.
6264	 */
6265	if (bfq_bfqq_softrt_update(bfqq) && bfqq->dispatched == 0 &&
6266	    RB_EMPTY_ROOT(&bfqq->sort_list) &&
6267	    bfqq->wr_coeff != bfqd->bfq_wr_coeff)
6268		bfqq->soft_rt_next_start =
6269			bfq_bfqq_softrt_next_start(bfqd, bfqq);
6270
6271	/*
6272	 * If this is the in-service queue, check if it needs to be expired,
6273	 * or if we want to idle in case it has no pending requests.
6274	 */
6275	if (bfqd->in_service_queue == bfqq) {
6276		if (bfq_bfqq_must_idle(bfqq)) {
6277			if (bfqq->dispatched == 0)
6278				bfq_arm_slice_timer(bfqd);
6279			/*
6280			 * If we get here, we do not expire bfqq, even
6281			 * if bfqq was in budget timeout or had no
6282			 * more requests (as controlled in the next
6283			 * conditional instructions). The reason for
6284			 * not expiring bfqq is as follows.
6285			 *
6286			 * Here bfqq->dispatched > 0 holds, but
6287			 * bfq_bfqq_must_idle() returned true. This
6288			 * implies that, even if no request arrives
6289			 * for bfqq before bfqq->dispatched reaches 0,
6290			 * bfqq will, however, not be expired on the
6291			 * completion event that causes bfqq->dispatch
6292			 * to reach zero. In contrast, on this event,
6293			 * bfqq will start enjoying device idling
6294			 * (I/O-dispatch plugging).
6295			 *
6296			 * But, if we expired bfqq here, bfqq would
6297			 * not have the chance to enjoy device idling
6298			 * when bfqq->dispatched finally reaches
6299			 * zero. This would expose bfqq to violation
6300			 * of its reserved service guarantees.
6301			 */
6302			return;
6303		} else if (bfq_may_expire_for_budg_timeout(bfqq))
6304			bfq_bfqq_expire(bfqd, bfqq, false,
6305					BFQQE_BUDGET_TIMEOUT);
6306		else if (RB_EMPTY_ROOT(&bfqq->sort_list) &&
6307			 (bfqq->dispatched == 0 ||
6308			  !bfq_better_to_idle(bfqq)))
6309			bfq_bfqq_expire(bfqd, bfqq, false,
6310					BFQQE_NO_MORE_REQUESTS);
6311	}
6312
6313	if (!bfqd->rq_in_driver)
6314		bfq_schedule_dispatch(bfqd);
6315}
6316
 
 
 
 
 
 
 
6317/*
6318 * The processes associated with bfqq may happen to generate their
6319 * cumulative I/O at a lower rate than the rate at which the device
6320 * could serve the same I/O. This is rather probable, e.g., if only
6321 * one process is associated with bfqq and the device is an SSD. It
6322 * results in bfqq becoming often empty while in service. In this
6323 * respect, if BFQ is allowed to switch to another queue when bfqq
6324 * remains empty, then the device goes on being fed with I/O requests,
6325 * and the throughput is not affected. In contrast, if BFQ is not
6326 * allowed to switch to another queue---because bfqq is sync and
6327 * I/O-dispatch needs to be plugged while bfqq is temporarily
6328 * empty---then, during the service of bfqq, there will be frequent
6329 * "service holes", i.e., time intervals during which bfqq gets empty
6330 * and the device can only consume the I/O already queued in its
6331 * hardware queues. During service holes, the device may even get to
6332 * remaining idle. In the end, during the service of bfqq, the device
6333 * is driven at a lower speed than the one it can reach with the kind
6334 * of I/O flowing through bfqq.
6335 *
6336 * To counter this loss of throughput, BFQ implements a "request
6337 * injection mechanism", which tries to fill the above service holes
6338 * with I/O requests taken from other queues. The hard part in this
6339 * mechanism is finding the right amount of I/O to inject, so as to
6340 * both boost throughput and not break bfqq's bandwidth and latency
6341 * guarantees. In this respect, the mechanism maintains a per-queue
6342 * inject limit, computed as below. While bfqq is empty, the injection
6343 * mechanism dispatches extra I/O requests only until the total number
6344 * of I/O requests in flight---i.e., already dispatched but not yet
6345 * completed---remains lower than this limit.
6346 *
6347 * A first definition comes in handy to introduce the algorithm by
6348 * which the inject limit is computed.  We define as first request for
6349 * bfqq, an I/O request for bfqq that arrives while bfqq is in
6350 * service, and causes bfqq to switch from empty to non-empty. The
6351 * algorithm updates the limit as a function of the effect of
6352 * injection on the service times of only the first requests of
6353 * bfqq. The reason for this restriction is that these are the
6354 * requests whose service time is affected most, because they are the
6355 * first to arrive after injection possibly occurred.
6356 *
6357 * To evaluate the effect of injection, the algorithm measures the
6358 * "total service time" of first requests. We define as total service
6359 * time of an I/O request, the time that elapses since when the
6360 * request is enqueued into bfqq, to when it is completed. This
6361 * quantity allows the whole effect of injection to be measured. It is
6362 * easy to see why. Suppose that some requests of other queues are
6363 * actually injected while bfqq is empty, and that a new request R
6364 * then arrives for bfqq. If the device does start to serve all or
6365 * part of the injected requests during the service hole, then,
6366 * because of this extra service, it may delay the next invocation of
6367 * the dispatch hook of BFQ. Then, even after R gets eventually
6368 * dispatched, the device may delay the actual service of R if it is
6369 * still busy serving the extra requests, or if it decides to serve,
6370 * before R, some extra request still present in its queues. As a
6371 * conclusion, the cumulative extra delay caused by injection can be
6372 * easily evaluated by just comparing the total service time of first
6373 * requests with and without injection.
6374 *
6375 * The limit-update algorithm works as follows. On the arrival of a
6376 * first request of bfqq, the algorithm measures the total time of the
6377 * request only if one of the three cases below holds, and, for each
6378 * case, it updates the limit as described below:
6379 *
6380 * (1) If there is no in-flight request. This gives a baseline for the
6381 *     total service time of the requests of bfqq. If the baseline has
6382 *     not been computed yet, then, after computing it, the limit is
6383 *     set to 1, to start boosting throughput, and to prepare the
6384 *     ground for the next case. If the baseline has already been
6385 *     computed, then it is updated, in case it results to be lower
6386 *     than the previous value.
6387 *
6388 * (2) If the limit is higher than 0 and there are in-flight
6389 *     requests. By comparing the total service time in this case with
6390 *     the above baseline, it is possible to know at which extent the
6391 *     current value of the limit is inflating the total service
6392 *     time. If the inflation is below a certain threshold, then bfqq
6393 *     is assumed to be suffering from no perceivable loss of its
6394 *     service guarantees, and the limit is even tentatively
6395 *     increased. If the inflation is above the threshold, then the
6396 *     limit is decreased. Due to the lack of any hysteresis, this
6397 *     logic makes the limit oscillate even in steady workload
6398 *     conditions. Yet we opted for it, because it is fast in reaching
6399 *     the best value for the limit, as a function of the current I/O
6400 *     workload. To reduce oscillations, this step is disabled for a
6401 *     short time interval after the limit happens to be decreased.
6402 *
6403 * (3) Periodically, after resetting the limit, to make sure that the
6404 *     limit eventually drops in case the workload changes. This is
6405 *     needed because, after the limit has gone safely up for a
6406 *     certain workload, it is impossible to guess whether the
6407 *     baseline total service time may have changed, without measuring
6408 *     it again without injection. A more effective version of this
6409 *     step might be to just sample the baseline, by interrupting
6410 *     injection only once, and then to reset/lower the limit only if
6411 *     the total service time with the current limit does happen to be
6412 *     too large.
6413 *
6414 * More details on each step are provided in the comments on the
6415 * pieces of code that implement these steps: the branch handling the
6416 * transition from empty to non empty in bfq_add_request(), the branch
6417 * handling injection in bfq_select_queue(), and the function
6418 * bfq_choose_bfqq_for_injection(). These comments also explain some
6419 * exceptions, made by the injection mechanism in some special cases.
6420 */
6421static void bfq_update_inject_limit(struct bfq_data *bfqd,
6422				    struct bfq_queue *bfqq)
6423{
6424	u64 tot_time_ns = ktime_get_ns() - bfqd->last_empty_occupied_ns;
6425	unsigned int old_limit = bfqq->inject_limit;
6426
6427	if (bfqq->last_serv_time_ns > 0 && bfqd->rqs_injected) {
6428		u64 threshold = (bfqq->last_serv_time_ns * 3)>>1;
6429
6430		if (tot_time_ns >= threshold && old_limit > 0) {
6431			bfqq->inject_limit--;
6432			bfqq->decrease_time_jif = jiffies;
6433		} else if (tot_time_ns < threshold &&
6434			   old_limit <= bfqd->max_rq_in_driver)
6435			bfqq->inject_limit++;
6436	}
6437
6438	/*
6439	 * Either we still have to compute the base value for the
6440	 * total service time, and there seem to be the right
6441	 * conditions to do it, or we can lower the last base value
6442	 * computed.
6443	 *
6444	 * NOTE: (bfqd->rq_in_driver == 1) means that there is no I/O
6445	 * request in flight, because this function is in the code
6446	 * path that handles the completion of a request of bfqq, and,
6447	 * in particular, this function is executed before
6448	 * bfqd->rq_in_driver is decremented in such a code path.
6449	 */
6450	if ((bfqq->last_serv_time_ns == 0 && bfqd->rq_in_driver == 1) ||
6451	    tot_time_ns < bfqq->last_serv_time_ns) {
6452		if (bfqq->last_serv_time_ns == 0) {
6453			/*
6454			 * Now we certainly have a base value: make sure we
6455			 * start trying injection.
6456			 */
6457			bfqq->inject_limit = max_t(unsigned int, 1, old_limit);
6458		}
6459		bfqq->last_serv_time_ns = tot_time_ns;
6460	} else if (!bfqd->rqs_injected && bfqd->rq_in_driver == 1)
6461		/*
6462		 * No I/O injected and no request still in service in
6463		 * the drive: these are the exact conditions for
6464		 * computing the base value of the total service time
6465		 * for bfqq. So let's update this value, because it is
6466		 * rather variable. For example, it varies if the size
6467		 * or the spatial locality of the I/O requests in bfqq
6468		 * change.
6469		 */
6470		bfqq->last_serv_time_ns = tot_time_ns;
6471
6472
6473	/* update complete, not waiting for any request completion any longer */
6474	bfqd->waited_rq = NULL;
6475	bfqd->rqs_injected = false;
6476}
6477
6478/*
6479 * Handle either a requeue or a finish for rq. The things to do are
6480 * the same in both cases: all references to rq are to be dropped. In
6481 * particular, rq is considered completed from the point of view of
6482 * the scheduler.
6483 */
6484static void bfq_finish_requeue_request(struct request *rq)
6485{
6486	struct bfq_queue *bfqq = RQ_BFQQ(rq);
6487	struct bfq_data *bfqd;
6488	unsigned long flags;
6489
6490	/*
6491	 * rq either is not associated with any icq, or is an already
6492	 * requeued request that has not (yet) been re-inserted into
6493	 * a bfq_queue.
6494	 */
6495	if (!rq->elv.icq || !bfqq)
6496		return;
6497
6498	bfqd = bfqq->bfqd;
6499
6500	if (rq->rq_flags & RQF_STARTED)
6501		bfqg_stats_update_completion(bfqq_group(bfqq),
6502					     rq->start_time_ns,
6503					     rq->io_start_time_ns,
6504					     rq->cmd_flags);
6505
6506	spin_lock_irqsave(&bfqd->lock, flags);
6507	if (likely(rq->rq_flags & RQF_STARTED)) {
 
 
 
 
6508		if (rq == bfqd->waited_rq)
6509			bfq_update_inject_limit(bfqd, bfqq);
6510
6511		bfq_completed_request(bfqq, bfqd);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6512	}
6513	bfqq_request_freed(bfqq);
6514	bfq_put_queue(bfqq);
6515	RQ_BIC(rq)->requests--;
6516	spin_unlock_irqrestore(&bfqd->lock, flags);
6517
6518	/*
6519	 * Reset private fields. In case of a requeue, this allows
6520	 * this function to correctly do nothing if it is spuriously
6521	 * invoked again on this same request (see the check at the
6522	 * beginning of the function). Probably, a better general
6523	 * design would be to prevent blk-mq from invoking the requeue
6524	 * or finish hooks of an elevator, for a request that is not
6525	 * referred by that elevator.
6526	 *
6527	 * Resetting the following fields would break the
6528	 * request-insertion logic if rq is re-inserted into a bfq
6529	 * internal queue, without a re-preparation. Here we assume
6530	 * that re-insertions of requeued requests, without
6531	 * re-preparation, can happen only for pass_through or at_head
6532	 * requests (which are not re-inserted into bfq internal
6533	 * queues).
6534	 */
6535	rq->elv.priv[0] = NULL;
6536	rq->elv.priv[1] = NULL;
6537}
6538
6539static void bfq_finish_request(struct request *rq)
6540{
6541	bfq_finish_requeue_request(rq);
6542
6543	if (rq->elv.icq) {
6544		put_io_context(rq->elv.icq->ioc);
6545		rq->elv.icq = NULL;
6546	}
6547}
6548
6549/*
6550 * Removes the association between the current task and bfqq, assuming
6551 * that bic points to the bfq iocontext of the task.
6552 * Returns NULL if a new bfqq should be allocated, or the old bfqq if this
6553 * was the last process referring to that bfqq.
6554 */
6555static struct bfq_queue *
6556bfq_split_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq)
6557{
6558	bfq_log_bfqq(bfqq->bfqd, bfqq, "splitting queue");
6559
6560	if (bfqq_process_refs(bfqq) == 1) {
6561		bfqq->pid = current->pid;
6562		bfq_clear_bfqq_coop(bfqq);
6563		bfq_clear_bfqq_split_coop(bfqq);
6564		return bfqq;
6565	}
6566
6567	bic_set_bfqq(bic, NULL, true);
6568
6569	bfq_put_cooperator(bfqq);
6570
6571	bfq_release_process_ref(bfqq->bfqd, bfqq);
6572	return NULL;
6573}
6574
6575static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd,
6576						   struct bfq_io_cq *bic,
6577						   struct bio *bio,
6578						   bool split, bool is_sync,
6579						   bool *new_queue)
6580{
6581	struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync);
6582
6583	if (likely(bfqq && bfqq != &bfqd->oom_bfqq))
6584		return bfqq;
6585
6586	if (new_queue)
6587		*new_queue = true;
6588
6589	if (bfqq)
6590		bfq_put_queue(bfqq);
6591	bfqq = bfq_get_queue(bfqd, bio, is_sync, bic, split);
6592
6593	bic_set_bfqq(bic, bfqq, is_sync);
6594	if (split && is_sync) {
6595		if ((bic->was_in_burst_list && bfqd->large_burst) ||
6596		    bic->saved_in_large_burst)
6597			bfq_mark_bfqq_in_large_burst(bfqq);
6598		else {
6599			bfq_clear_bfqq_in_large_burst(bfqq);
6600			if (bic->was_in_burst_list)
6601				/*
6602				 * If bfqq was in the current
6603				 * burst list before being
6604				 * merged, then we have to add
6605				 * it back. And we do not need
6606				 * to increase burst_size, as
6607				 * we did not decrement
6608				 * burst_size when we removed
6609				 * bfqq from the burst list as
6610				 * a consequence of a merge
6611				 * (see comments in
6612				 * bfq_put_queue). In this
6613				 * respect, it would be rather
6614				 * costly to know whether the
6615				 * current burst list is still
6616				 * the same burst list from
6617				 * which bfqq was removed on
6618				 * the merge. To avoid this
6619				 * cost, if bfqq was in a
6620				 * burst list, then we add
6621				 * bfqq to the current burst
6622				 * list without any further
6623				 * check. This can cause
6624				 * inappropriate insertions,
6625				 * but rarely enough to not
6626				 * harm the detection of large
6627				 * bursts significantly.
6628				 */
6629				hlist_add_head(&bfqq->burst_list_node,
6630					       &bfqd->burst_list);
6631		}
6632		bfqq->split_time = jiffies;
6633	}
6634
6635	return bfqq;
6636}
6637
6638/*
6639 * Only reset private fields. The actual request preparation will be
6640 * performed by bfq_init_rq, when rq is either inserted or merged. See
6641 * comments on bfq_init_rq for the reason behind this delayed
6642 * preparation.
6643 */
6644static void bfq_prepare_request(struct request *rq)
6645{
6646	rq->elv.icq = ioc_find_get_icq(rq->q);
6647
6648	/*
6649	 * Regardless of whether we have an icq attached, we have to
6650	 * clear the scheduler pointers, as they might point to
6651	 * previously allocated bic/bfqq structs.
6652	 */
6653	rq->elv.priv[0] = rq->elv.priv[1] = NULL;
6654}
6655
6656/*
6657 * If needed, init rq, allocate bfq data structures associated with
6658 * rq, and increment reference counters in the destination bfq_queue
6659 * for rq. Return the destination bfq_queue for rq, or NULL is rq is
6660 * not associated with any bfq_queue.
6661 *
6662 * This function is invoked by the functions that perform rq insertion
6663 * or merging. One may have expected the above preparation operations
6664 * to be performed in bfq_prepare_request, and not delayed to when rq
6665 * is inserted or merged. The rationale behind this delayed
6666 * preparation is that, after the prepare_request hook is invoked for
6667 * rq, rq may still be transformed into a request with no icq, i.e., a
6668 * request not associated with any queue. No bfq hook is invoked to
6669 * signal this transformation. As a consequence, should these
6670 * preparation operations be performed when the prepare_request hook
6671 * is invoked, and should rq be transformed one moment later, bfq
6672 * would end up in an inconsistent state, because it would have
6673 * incremented some queue counters for an rq destined to
6674 * transformation, without any chance to correctly lower these
6675 * counters back. In contrast, no transformation can still happen for
6676 * rq after rq has been inserted or merged. So, it is safe to execute
6677 * these preparation operations when rq is finally inserted or merged.
6678 */
6679static struct bfq_queue *bfq_init_rq(struct request *rq)
6680{
6681	struct request_queue *q = rq->q;
6682	struct bio *bio = rq->bio;
6683	struct bfq_data *bfqd = q->elevator->elevator_data;
6684	struct bfq_io_cq *bic;
6685	const int is_sync = rq_is_sync(rq);
6686	struct bfq_queue *bfqq;
6687	bool new_queue = false;
6688	bool bfqq_already_existing = false, split = false;
6689
6690	if (unlikely(!rq->elv.icq))
6691		return NULL;
6692
6693	/*
6694	 * Assuming that elv.priv[1] is set only if everything is set
6695	 * for this rq. This holds true, because this function is
6696	 * invoked only for insertion or merging, and, after such
6697	 * events, a request cannot be manipulated any longer before
6698	 * being removed from bfq.
6699	 */
6700	if (rq->elv.priv[1])
6701		return rq->elv.priv[1];
6702
6703	bic = icq_to_bic(rq->elv.icq);
6704
6705	bfq_check_ioprio_change(bic, bio);
6706
6707	bfq_bic_update_cgroup(bic, bio);
6708
6709	bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio, false, is_sync,
6710					 &new_queue);
6711
6712	if (likely(!new_queue)) {
6713		/* If the queue was seeky for too long, break it apart. */
6714		if (bfq_bfqq_coop(bfqq) && bfq_bfqq_split_coop(bfqq) &&
6715			!bic->stably_merged) {
6716			struct bfq_queue *old_bfqq = bfqq;
6717
6718			/* Update bic before losing reference to bfqq */
6719			if (bfq_bfqq_in_large_burst(bfqq))
6720				bic->saved_in_large_burst = true;
6721
6722			bfqq = bfq_split_bfqq(bic, bfqq);
6723			split = true;
6724
6725			if (!bfqq) {
6726				bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio,
6727								 true, is_sync,
6728								 NULL);
6729				if (unlikely(bfqq == &bfqd->oom_bfqq))
6730					bfqq_already_existing = true;
6731			} else
6732				bfqq_already_existing = true;
6733
6734			if (!bfqq_already_existing) {
6735				bfqq->waker_bfqq = old_bfqq->waker_bfqq;
6736				bfqq->tentative_waker_bfqq = NULL;
6737
6738				/*
6739				 * If the waker queue disappears, then
6740				 * new_bfqq->waker_bfqq must be
6741				 * reset. So insert new_bfqq into the
6742				 * woken_list of the waker. See
6743				 * bfq_check_waker for details.
6744				 */
6745				if (bfqq->waker_bfqq)
6746					hlist_add_head(&bfqq->woken_list_node,
6747						       &bfqq->waker_bfqq->woken_list);
6748			}
6749		}
6750	}
6751
6752	bfqq_request_allocated(bfqq);
6753	bfqq->ref++;
6754	bic->requests++;
6755	bfq_log_bfqq(bfqd, bfqq, "get_request %p: bfqq %p, %d",
6756		     rq, bfqq, bfqq->ref);
6757
6758	rq->elv.priv[0] = bic;
6759	rq->elv.priv[1] = bfqq;
6760
6761	/*
6762	 * If a bfq_queue has only one process reference, it is owned
6763	 * by only this bic: we can then set bfqq->bic = bic. in
6764	 * addition, if the queue has also just been split, we have to
6765	 * resume its state.
6766	 */
6767	if (likely(bfqq != &bfqd->oom_bfqq) && bfqq_process_refs(bfqq) == 1) {
6768		bfqq->bic = bic;
6769		if (split) {
6770			/*
6771			 * The queue has just been split from a shared
6772			 * queue: restore the idle window and the
6773			 * possible weight raising period.
6774			 */
6775			bfq_bfqq_resume_state(bfqq, bfqd, bic,
6776					      bfqq_already_existing);
6777		}
6778	}
6779
6780	/*
6781	 * Consider bfqq as possibly belonging to a burst of newly
6782	 * created queues only if:
6783	 * 1) A burst is actually happening (bfqd->burst_size > 0)
6784	 * or
6785	 * 2) There is no other active queue. In fact, if, in
6786	 *    contrast, there are active queues not belonging to the
6787	 *    possible burst bfqq may belong to, then there is no gain
6788	 *    in considering bfqq as belonging to a burst, and
6789	 *    therefore in not weight-raising bfqq. See comments on
6790	 *    bfq_handle_burst().
6791	 *
6792	 * This filtering also helps eliminating false positives,
6793	 * occurring when bfqq does not belong to an actual large
6794	 * burst, but some background task (e.g., a service) happens
6795	 * to trigger the creation of new queues very close to when
6796	 * bfqq and its possible companion queues are created. See
6797	 * comments on bfq_handle_burst() for further details also on
6798	 * this issue.
6799	 */
6800	if (unlikely(bfq_bfqq_just_created(bfqq) &&
6801		     (bfqd->burst_size > 0 ||
6802		      bfq_tot_busy_queues(bfqd) == 0)))
6803		bfq_handle_burst(bfqd, bfqq);
6804
6805	return bfqq;
6806}
6807
6808static void
6809bfq_idle_slice_timer_body(struct bfq_data *bfqd, struct bfq_queue *bfqq)
6810{
6811	enum bfqq_expiration reason;
6812	unsigned long flags;
6813
6814	spin_lock_irqsave(&bfqd->lock, flags);
6815
6816	/*
6817	 * Considering that bfqq may be in race, we should firstly check
6818	 * whether bfqq is in service before doing something on it. If
6819	 * the bfqq in race is not in service, it has already been expired
6820	 * through __bfq_bfqq_expire func and its wait_request flags has
6821	 * been cleared in __bfq_bfqd_reset_in_service func.
6822	 */
6823	if (bfqq != bfqd->in_service_queue) {
6824		spin_unlock_irqrestore(&bfqd->lock, flags);
6825		return;
6826	}
6827
6828	bfq_clear_bfqq_wait_request(bfqq);
6829
6830	if (bfq_bfqq_budget_timeout(bfqq))
6831		/*
6832		 * Also here the queue can be safely expired
6833		 * for budget timeout without wasting
6834		 * guarantees
6835		 */
6836		reason = BFQQE_BUDGET_TIMEOUT;
6837	else if (bfqq->queued[0] == 0 && bfqq->queued[1] == 0)
6838		/*
6839		 * The queue may not be empty upon timer expiration,
6840		 * because we may not disable the timer when the
6841		 * first request of the in-service queue arrives
6842		 * during disk idling.
6843		 */
6844		reason = BFQQE_TOO_IDLE;
6845	else
6846		goto schedule_dispatch;
6847
6848	bfq_bfqq_expire(bfqd, bfqq, true, reason);
6849
6850schedule_dispatch:
6851	bfq_schedule_dispatch(bfqd);
6852	spin_unlock_irqrestore(&bfqd->lock, flags);
 
6853}
6854
6855/*
6856 * Handler of the expiration of the timer running if the in-service queue
6857 * is idling inside its time slice.
6858 */
6859static enum hrtimer_restart bfq_idle_slice_timer(struct hrtimer *timer)
6860{
6861	struct bfq_data *bfqd = container_of(timer, struct bfq_data,
6862					     idle_slice_timer);
6863	struct bfq_queue *bfqq = bfqd->in_service_queue;
6864
6865	/*
6866	 * Theoretical race here: the in-service queue can be NULL or
6867	 * different from the queue that was idling if a new request
6868	 * arrives for the current queue and there is a full dispatch
6869	 * cycle that changes the in-service queue.  This can hardly
6870	 * happen, but in the worst case we just expire a queue too
6871	 * early.
6872	 */
6873	if (bfqq)
6874		bfq_idle_slice_timer_body(bfqd, bfqq);
6875
6876	return HRTIMER_NORESTART;
6877}
6878
6879static void __bfq_put_async_bfqq(struct bfq_data *bfqd,
6880				 struct bfq_queue **bfqq_ptr)
6881{
6882	struct bfq_queue *bfqq = *bfqq_ptr;
6883
6884	bfq_log(bfqd, "put_async_bfqq: %p", bfqq);
6885	if (bfqq) {
6886		bfq_bfqq_move(bfqd, bfqq, bfqd->root_group);
6887
6888		bfq_log_bfqq(bfqd, bfqq, "put_async_bfqq: putting %p, %d",
6889			     bfqq, bfqq->ref);
6890		bfq_put_queue(bfqq);
6891		*bfqq_ptr = NULL;
6892	}
6893}
6894
6895/*
6896 * Release all the bfqg references to its async queues.  If we are
6897 * deallocating the group these queues may still contain requests, so
6898 * we reparent them to the root cgroup (i.e., the only one that will
6899 * exist for sure until all the requests on a device are gone).
6900 */
6901void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg)
6902{
6903	int i, j;
6904
6905	for (i = 0; i < 2; i++)
6906		for (j = 0; j < IOPRIO_NR_LEVELS; j++)
6907			__bfq_put_async_bfqq(bfqd, &bfqg->async_bfqq[i][j]);
6908
6909	__bfq_put_async_bfqq(bfqd, &bfqg->async_idle_bfqq);
6910}
6911
6912/*
6913 * See the comments on bfq_limit_depth for the purpose of
6914 * the depths set in the function. Return minimum shallow depth we'll use.
6915 */
6916static void bfq_update_depths(struct bfq_data *bfqd, struct sbitmap_queue *bt)
 
6917{
6918	unsigned int depth = 1U << bt->sb.shift;
6919
6920	bfqd->full_depth_shift = bt->sb.shift;
6921	/*
6922	 * In-word depths if no bfq_queue is being weight-raised:
6923	 * leaving 25% of tags only for sync reads.
6924	 *
6925	 * In next formulas, right-shift the value
6926	 * (1U<<bt->sb.shift), instead of computing directly
6927	 * (1U<<(bt->sb.shift - something)), to be robust against
6928	 * any possible value of bt->sb.shift, without having to
6929	 * limit 'something'.
6930	 */
6931	/* no more than 50% of tags for async I/O */
6932	bfqd->word_depths[0][0] = max(depth >> 1, 1U);
6933	/*
6934	 * no more than 75% of tags for sync writes (25% extra tags
6935	 * w.r.t. async I/O, to prevent async I/O from starving sync
6936	 * writes)
6937	 */
6938	bfqd->word_depths[0][1] = max((depth * 3) >> 2, 1U);
6939
6940	/*
6941	 * In-word depths in case some bfq_queue is being weight-
6942	 * raised: leaving ~63% of tags for sync reads. This is the
6943	 * highest percentage for which, in our tests, application
6944	 * start-up times didn't suffer from any regression due to tag
6945	 * shortage.
6946	 */
6947	/* no more than ~18% of tags for async I/O */
6948	bfqd->word_depths[1][0] = max((depth * 3) >> 4, 1U);
6949	/* no more than ~37% of tags for sync writes (~20% extra tags) */
6950	bfqd->word_depths[1][1] = max((depth * 6) >> 4, 1U);
 
 
 
 
 
 
6951}
6952
6953static void bfq_depth_updated(struct blk_mq_hw_ctx *hctx)
6954{
6955	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
6956	struct blk_mq_tags *tags = hctx->sched_tags;
 
6957
6958	bfq_update_depths(bfqd, &tags->bitmap_tags);
6959	sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, 1);
6960}
6961
6962static int bfq_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int index)
6963{
6964	bfq_depth_updated(hctx);
6965	return 0;
6966}
6967
6968static void bfq_exit_queue(struct elevator_queue *e)
6969{
6970	struct bfq_data *bfqd = e->elevator_data;
6971	struct bfq_queue *bfqq, *n;
6972
6973	hrtimer_cancel(&bfqd->idle_slice_timer);
6974
6975	spin_lock_irq(&bfqd->lock);
6976	list_for_each_entry_safe(bfqq, n, &bfqd->idle_list, bfqq_list)
6977		bfq_deactivate_bfqq(bfqd, bfqq, false, false);
6978	spin_unlock_irq(&bfqd->lock);
6979
6980	hrtimer_cancel(&bfqd->idle_slice_timer);
6981
6982	/* release oom-queue reference to root group */
6983	bfqg_and_blkg_put(bfqd->root_group);
6984
6985#ifdef CONFIG_BFQ_GROUP_IOSCHED
6986	blkcg_deactivate_policy(bfqd->queue, &blkcg_policy_bfq);
6987#else
6988	spin_lock_irq(&bfqd->lock);
6989	bfq_put_async_queues(bfqd, bfqd->root_group);
6990	kfree(bfqd->root_group);
6991	spin_unlock_irq(&bfqd->lock);
6992#endif
6993
6994	blk_stat_disable_accounting(bfqd->queue);
6995	clear_bit(ELEVATOR_FLAG_DISABLE_WBT, &e->flags);
6996	wbt_enable_default(bfqd->queue);
6997
6998	kfree(bfqd);
6999}
7000
7001static void bfq_init_root_group(struct bfq_group *root_group,
7002				struct bfq_data *bfqd)
7003{
7004	int i;
7005
7006#ifdef CONFIG_BFQ_GROUP_IOSCHED
7007	root_group->entity.parent = NULL;
7008	root_group->my_entity = NULL;
7009	root_group->bfqd = bfqd;
7010#endif
7011	root_group->rq_pos_tree = RB_ROOT;
7012	for (i = 0; i < BFQ_IOPRIO_CLASSES; i++)
7013		root_group->sched_data.service_tree[i] = BFQ_SERVICE_TREE_INIT;
7014	root_group->sched_data.bfq_class_idle_last_service = jiffies;
7015}
7016
7017static int bfq_init_queue(struct request_queue *q, struct elevator_type *e)
7018{
7019	struct bfq_data *bfqd;
7020	struct elevator_queue *eq;
7021
7022	eq = elevator_alloc(q, e);
7023	if (!eq)
7024		return -ENOMEM;
7025
7026	bfqd = kzalloc_node(sizeof(*bfqd), GFP_KERNEL, q->node);
7027	if (!bfqd) {
7028		kobject_put(&eq->kobj);
7029		return -ENOMEM;
7030	}
7031	eq->elevator_data = bfqd;
7032
7033	spin_lock_irq(&q->queue_lock);
7034	q->elevator = eq;
7035	spin_unlock_irq(&q->queue_lock);
7036
7037	/*
7038	 * Our fallback bfqq if bfq_find_alloc_queue() runs into OOM issues.
7039	 * Grab a permanent reference to it, so that the normal code flow
7040	 * will not attempt to free it.
7041	 */
7042	bfq_init_bfqq(bfqd, &bfqd->oom_bfqq, NULL, 1, 0);
7043	bfqd->oom_bfqq.ref++;
7044	bfqd->oom_bfqq.new_ioprio = BFQ_DEFAULT_QUEUE_IOPRIO;
7045	bfqd->oom_bfqq.new_ioprio_class = IOPRIO_CLASS_BE;
7046	bfqd->oom_bfqq.entity.new_weight =
7047		bfq_ioprio_to_weight(bfqd->oom_bfqq.new_ioprio);
7048
7049	/* oom_bfqq does not participate to bursts */
7050	bfq_clear_bfqq_just_created(&bfqd->oom_bfqq);
7051
7052	/*
7053	 * Trigger weight initialization, according to ioprio, at the
7054	 * oom_bfqq's first activation. The oom_bfqq's ioprio and ioprio
7055	 * class won't be changed any more.
7056	 */
7057	bfqd->oom_bfqq.entity.prio_changed = 1;
7058
7059	bfqd->queue = q;
7060
7061	INIT_LIST_HEAD(&bfqd->dispatch);
7062
7063	hrtimer_init(&bfqd->idle_slice_timer, CLOCK_MONOTONIC,
7064		     HRTIMER_MODE_REL);
7065	bfqd->idle_slice_timer.function = bfq_idle_slice_timer;
7066
7067	bfqd->queue_weights_tree = RB_ROOT_CACHED;
7068#ifdef CONFIG_BFQ_GROUP_IOSCHED
7069	bfqd->num_groups_with_pending_reqs = 0;
7070#endif
7071
7072	INIT_LIST_HEAD(&bfqd->active_list);
7073	INIT_LIST_HEAD(&bfqd->idle_list);
7074	INIT_HLIST_HEAD(&bfqd->burst_list);
7075
7076	bfqd->hw_tag = -1;
7077	bfqd->nonrot_with_queueing = blk_queue_nonrot(bfqd->queue);
7078
7079	bfqd->bfq_max_budget = bfq_default_max_budget;
7080
7081	bfqd->bfq_fifo_expire[0] = bfq_fifo_expire[0];
7082	bfqd->bfq_fifo_expire[1] = bfq_fifo_expire[1];
7083	bfqd->bfq_back_max = bfq_back_max;
7084	bfqd->bfq_back_penalty = bfq_back_penalty;
7085	bfqd->bfq_slice_idle = bfq_slice_idle;
7086	bfqd->bfq_timeout = bfq_timeout;
7087
 
 
7088	bfqd->bfq_large_burst_thresh = 8;
7089	bfqd->bfq_burst_interval = msecs_to_jiffies(180);
7090
7091	bfqd->low_latency = true;
7092
7093	/*
7094	 * Trade-off between responsiveness and fairness.
7095	 */
7096	bfqd->bfq_wr_coeff = 30;
7097	bfqd->bfq_wr_rt_max_time = msecs_to_jiffies(300);
7098	bfqd->bfq_wr_max_time = 0;
7099	bfqd->bfq_wr_min_idle_time = msecs_to_jiffies(2000);
7100	bfqd->bfq_wr_min_inter_arr_async = msecs_to_jiffies(500);
7101	bfqd->bfq_wr_max_softrt_rate = 7000; /*
7102					      * Approximate rate required
7103					      * to playback or record a
7104					      * high-definition compressed
7105					      * video.
7106					      */
7107	bfqd->wr_busy_queues = 0;
7108
7109	/*
7110	 * Begin by assuming, optimistically, that the device peak
7111	 * rate is equal to 2/3 of the highest reference rate.
7112	 */
7113	bfqd->rate_dur_prod = ref_rate[blk_queue_nonrot(bfqd->queue)] *
7114		ref_wr_duration[blk_queue_nonrot(bfqd->queue)];
7115	bfqd->peak_rate = ref_rate[blk_queue_nonrot(bfqd->queue)] * 2 / 3;
7116
7117	spin_lock_init(&bfqd->lock);
7118
7119	/*
7120	 * The invocation of the next bfq_create_group_hierarchy
7121	 * function is the head of a chain of function calls
7122	 * (bfq_create_group_hierarchy->blkcg_activate_policy->
7123	 * blk_mq_freeze_queue) that may lead to the invocation of the
7124	 * has_work hook function. For this reason,
7125	 * bfq_create_group_hierarchy is invoked only after all
7126	 * scheduler data has been initialized, apart from the fields
7127	 * that can be initialized only after invoking
7128	 * bfq_create_group_hierarchy. This, in particular, enables
7129	 * has_work to correctly return false. Of course, to avoid
7130	 * other inconsistencies, the blk-mq stack must then refrain
7131	 * from invoking further scheduler hooks before this init
7132	 * function is finished.
7133	 */
7134	bfqd->root_group = bfq_create_group_hierarchy(bfqd, q->node);
7135	if (!bfqd->root_group)
7136		goto out_free;
7137	bfq_init_root_group(bfqd->root_group, bfqd);
7138	bfq_init_entity(&bfqd->oom_bfqq.entity, bfqd->root_group);
7139
7140	/* We dispatch from request queue wide instead of hw queue */
7141	blk_queue_flag_set(QUEUE_FLAG_SQ_SCHED, q);
7142
7143	set_bit(ELEVATOR_FLAG_DISABLE_WBT, &eq->flags);
7144	wbt_disable_default(q);
7145	blk_stat_enable_accounting(q);
7146
7147	return 0;
7148
7149out_free:
7150	kfree(bfqd);
7151	kobject_put(&eq->kobj);
7152	return -ENOMEM;
7153}
7154
7155static void bfq_slab_kill(void)
7156{
7157	kmem_cache_destroy(bfq_pool);
7158}
7159
7160static int __init bfq_slab_setup(void)
7161{
7162	bfq_pool = KMEM_CACHE(bfq_queue, 0);
7163	if (!bfq_pool)
7164		return -ENOMEM;
7165	return 0;
7166}
7167
7168static ssize_t bfq_var_show(unsigned int var, char *page)
7169{
7170	return sprintf(page, "%u\n", var);
7171}
7172
7173static int bfq_var_store(unsigned long *var, const char *page)
7174{
7175	unsigned long new_val;
7176	int ret = kstrtoul(page, 10, &new_val);
7177
7178	if (ret)
7179		return ret;
7180	*var = new_val;
7181	return 0;
7182}
7183
7184#define SHOW_FUNCTION(__FUNC, __VAR, __CONV)				\
7185static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
7186{									\
7187	struct bfq_data *bfqd = e->elevator_data;			\
7188	u64 __data = __VAR;						\
7189	if (__CONV == 1)						\
7190		__data = jiffies_to_msecs(__data);			\
7191	else if (__CONV == 2)						\
7192		__data = div_u64(__data, NSEC_PER_MSEC);		\
7193	return bfq_var_show(__data, (page));				\
7194}
7195SHOW_FUNCTION(bfq_fifo_expire_sync_show, bfqd->bfq_fifo_expire[1], 2);
7196SHOW_FUNCTION(bfq_fifo_expire_async_show, bfqd->bfq_fifo_expire[0], 2);
7197SHOW_FUNCTION(bfq_back_seek_max_show, bfqd->bfq_back_max, 0);
7198SHOW_FUNCTION(bfq_back_seek_penalty_show, bfqd->bfq_back_penalty, 0);
7199SHOW_FUNCTION(bfq_slice_idle_show, bfqd->bfq_slice_idle, 2);
7200SHOW_FUNCTION(bfq_max_budget_show, bfqd->bfq_user_max_budget, 0);
7201SHOW_FUNCTION(bfq_timeout_sync_show, bfqd->bfq_timeout, 1);
7202SHOW_FUNCTION(bfq_strict_guarantees_show, bfqd->strict_guarantees, 0);
7203SHOW_FUNCTION(bfq_low_latency_show, bfqd->low_latency, 0);
7204#undef SHOW_FUNCTION
7205
7206#define USEC_SHOW_FUNCTION(__FUNC, __VAR)				\
7207static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
7208{									\
7209	struct bfq_data *bfqd = e->elevator_data;			\
7210	u64 __data = __VAR;						\
7211	__data = div_u64(__data, NSEC_PER_USEC);			\
7212	return bfq_var_show(__data, (page));				\
7213}
7214USEC_SHOW_FUNCTION(bfq_slice_idle_us_show, bfqd->bfq_slice_idle);
7215#undef USEC_SHOW_FUNCTION
7216
7217#define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)			\
7218static ssize_t								\
7219__FUNC(struct elevator_queue *e, const char *page, size_t count)	\
7220{									\
7221	struct bfq_data *bfqd = e->elevator_data;			\
7222	unsigned long __data, __min = (MIN), __max = (MAX);		\
7223	int ret;							\
7224									\
7225	ret = bfq_var_store(&__data, (page));				\
7226	if (ret)							\
7227		return ret;						\
7228	if (__data < __min)						\
7229		__data = __min;						\
7230	else if (__data > __max)					\
7231		__data = __max;						\
7232	if (__CONV == 1)						\
7233		*(__PTR) = msecs_to_jiffies(__data);			\
7234	else if (__CONV == 2)						\
7235		*(__PTR) = (u64)__data * NSEC_PER_MSEC;			\
7236	else								\
7237		*(__PTR) = __data;					\
7238	return count;							\
7239}
7240STORE_FUNCTION(bfq_fifo_expire_sync_store, &bfqd->bfq_fifo_expire[1], 1,
7241		INT_MAX, 2);
7242STORE_FUNCTION(bfq_fifo_expire_async_store, &bfqd->bfq_fifo_expire[0], 1,
7243		INT_MAX, 2);
7244STORE_FUNCTION(bfq_back_seek_max_store, &bfqd->bfq_back_max, 0, INT_MAX, 0);
7245STORE_FUNCTION(bfq_back_seek_penalty_store, &bfqd->bfq_back_penalty, 1,
7246		INT_MAX, 0);
7247STORE_FUNCTION(bfq_slice_idle_store, &bfqd->bfq_slice_idle, 0, INT_MAX, 2);
7248#undef STORE_FUNCTION
7249
7250#define USEC_STORE_FUNCTION(__FUNC, __PTR, MIN, MAX)			\
7251static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count)\
7252{									\
7253	struct bfq_data *bfqd = e->elevator_data;			\
7254	unsigned long __data, __min = (MIN), __max = (MAX);		\
7255	int ret;							\
7256									\
7257	ret = bfq_var_store(&__data, (page));				\
7258	if (ret)							\
7259		return ret;						\
7260	if (__data < __min)						\
7261		__data = __min;						\
7262	else if (__data > __max)					\
7263		__data = __max;						\
7264	*(__PTR) = (u64)__data * NSEC_PER_USEC;				\
7265	return count;							\
7266}
7267USEC_STORE_FUNCTION(bfq_slice_idle_us_store, &bfqd->bfq_slice_idle, 0,
7268		    UINT_MAX);
7269#undef USEC_STORE_FUNCTION
7270
7271static ssize_t bfq_max_budget_store(struct elevator_queue *e,
7272				    const char *page, size_t count)
7273{
7274	struct bfq_data *bfqd = e->elevator_data;
7275	unsigned long __data;
7276	int ret;
7277
7278	ret = bfq_var_store(&__data, (page));
7279	if (ret)
7280		return ret;
7281
7282	if (__data == 0)
7283		bfqd->bfq_max_budget = bfq_calc_max_budget(bfqd);
7284	else {
7285		if (__data > INT_MAX)
7286			__data = INT_MAX;
7287		bfqd->bfq_max_budget = __data;
7288	}
7289
7290	bfqd->bfq_user_max_budget = __data;
7291
7292	return count;
7293}
7294
7295/*
7296 * Leaving this name to preserve name compatibility with cfq
7297 * parameters, but this timeout is used for both sync and async.
7298 */
7299static ssize_t bfq_timeout_sync_store(struct elevator_queue *e,
7300				      const char *page, size_t count)
7301{
7302	struct bfq_data *bfqd = e->elevator_data;
7303	unsigned long __data;
7304	int ret;
7305
7306	ret = bfq_var_store(&__data, (page));
7307	if (ret)
7308		return ret;
7309
7310	if (__data < 1)
7311		__data = 1;
7312	else if (__data > INT_MAX)
7313		__data = INT_MAX;
7314
7315	bfqd->bfq_timeout = msecs_to_jiffies(__data);
7316	if (bfqd->bfq_user_max_budget == 0)
7317		bfqd->bfq_max_budget = bfq_calc_max_budget(bfqd);
7318
7319	return count;
7320}
7321
7322static ssize_t bfq_strict_guarantees_store(struct elevator_queue *e,
7323				     const char *page, size_t count)
7324{
7325	struct bfq_data *bfqd = e->elevator_data;
7326	unsigned long __data;
7327	int ret;
7328
7329	ret = bfq_var_store(&__data, (page));
7330	if (ret)
7331		return ret;
7332
7333	if (__data > 1)
7334		__data = 1;
7335	if (!bfqd->strict_guarantees && __data == 1
7336	    && bfqd->bfq_slice_idle < 8 * NSEC_PER_MSEC)
7337		bfqd->bfq_slice_idle = 8 * NSEC_PER_MSEC;
7338
7339	bfqd->strict_guarantees = __data;
7340
7341	return count;
7342}
7343
7344static ssize_t bfq_low_latency_store(struct elevator_queue *e,
7345				     const char *page, size_t count)
7346{
7347	struct bfq_data *bfqd = e->elevator_data;
7348	unsigned long __data;
7349	int ret;
7350
7351	ret = bfq_var_store(&__data, (page));
7352	if (ret)
7353		return ret;
7354
7355	if (__data > 1)
7356		__data = 1;
7357	if (__data == 0 && bfqd->low_latency != 0)
7358		bfq_end_wr(bfqd);
7359	bfqd->low_latency = __data;
7360
7361	return count;
7362}
7363
7364#define BFQ_ATTR(name) \
7365	__ATTR(name, 0644, bfq_##name##_show, bfq_##name##_store)
7366
7367static struct elv_fs_entry bfq_attrs[] = {
7368	BFQ_ATTR(fifo_expire_sync),
7369	BFQ_ATTR(fifo_expire_async),
7370	BFQ_ATTR(back_seek_max),
7371	BFQ_ATTR(back_seek_penalty),
7372	BFQ_ATTR(slice_idle),
7373	BFQ_ATTR(slice_idle_us),
7374	BFQ_ATTR(max_budget),
7375	BFQ_ATTR(timeout_sync),
7376	BFQ_ATTR(strict_guarantees),
7377	BFQ_ATTR(low_latency),
7378	__ATTR_NULL
7379};
7380
7381static struct elevator_type iosched_bfq_mq = {
7382	.ops = {
7383		.limit_depth		= bfq_limit_depth,
7384		.prepare_request	= bfq_prepare_request,
7385		.requeue_request        = bfq_finish_requeue_request,
7386		.finish_request		= bfq_finish_request,
7387		.exit_icq		= bfq_exit_icq,
7388		.insert_requests	= bfq_insert_requests,
7389		.dispatch_request	= bfq_dispatch_request,
7390		.next_request		= elv_rb_latter_request,
7391		.former_request		= elv_rb_former_request,
7392		.allow_merge		= bfq_allow_bio_merge,
7393		.bio_merge		= bfq_bio_merge,
7394		.request_merge		= bfq_request_merge,
7395		.requests_merged	= bfq_requests_merged,
7396		.request_merged		= bfq_request_merged,
7397		.has_work		= bfq_has_work,
7398		.depth_updated		= bfq_depth_updated,
7399		.init_hctx		= bfq_init_hctx,
7400		.init_sched		= bfq_init_queue,
7401		.exit_sched		= bfq_exit_queue,
7402	},
7403
7404	.icq_size =		sizeof(struct bfq_io_cq),
7405	.icq_align =		__alignof__(struct bfq_io_cq),
7406	.elevator_attrs =	bfq_attrs,
7407	.elevator_name =	"bfq",
7408	.elevator_owner =	THIS_MODULE,
7409};
7410MODULE_ALIAS("bfq-iosched");
7411
7412static int __init bfq_init(void)
7413{
7414	int ret;
7415
7416#ifdef CONFIG_BFQ_GROUP_IOSCHED
7417	ret = blkcg_policy_register(&blkcg_policy_bfq);
7418	if (ret)
7419		return ret;
7420#endif
7421
7422	ret = -ENOMEM;
7423	if (bfq_slab_setup())
7424		goto err_pol_unreg;
7425
7426	/*
7427	 * Times to load large popular applications for the typical
7428	 * systems installed on the reference devices (see the
7429	 * comments before the definition of the next
7430	 * array). Actually, we use slightly lower values, as the
7431	 * estimated peak rate tends to be smaller than the actual
7432	 * peak rate.  The reason for this last fact is that estimates
7433	 * are computed over much shorter time intervals than the long
7434	 * intervals typically used for benchmarking. Why? First, to
7435	 * adapt more quickly to variations. Second, because an I/O
7436	 * scheduler cannot rely on a peak-rate-evaluation workload to
7437	 * be run for a long time.
7438	 */
7439	ref_wr_duration[0] = msecs_to_jiffies(7000); /* actually 8 sec */
7440	ref_wr_duration[1] = msecs_to_jiffies(2500); /* actually 3 sec */
7441
7442	ret = elv_register(&iosched_bfq_mq);
7443	if (ret)
7444		goto slab_kill;
7445
7446	return 0;
7447
7448slab_kill:
7449	bfq_slab_kill();
7450err_pol_unreg:
7451#ifdef CONFIG_BFQ_GROUP_IOSCHED
7452	blkcg_policy_unregister(&blkcg_policy_bfq);
7453#endif
7454	return ret;
7455}
7456
7457static void __exit bfq_exit(void)
7458{
7459	elv_unregister(&iosched_bfq_mq);
7460#ifdef CONFIG_BFQ_GROUP_IOSCHED
7461	blkcg_policy_unregister(&blkcg_policy_bfq);
7462#endif
7463	bfq_slab_kill();
7464}
7465
7466module_init(bfq_init);
7467module_exit(bfq_exit);
7468
7469MODULE_AUTHOR("Paolo Valente");
7470MODULE_LICENSE("GPL");
7471MODULE_DESCRIPTION("MQ Budget Fair Queueing I/O Scheduler");
v5.9
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 * Budget Fair Queueing (BFQ) I/O scheduler.
   4 *
   5 * Based on ideas and code from CFQ:
   6 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
   7 *
   8 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
   9 *		      Paolo Valente <paolo.valente@unimore.it>
  10 *
  11 * Copyright (C) 2010 Paolo Valente <paolo.valente@unimore.it>
  12 *                    Arianna Avanzini <avanzini@google.com>
  13 *
  14 * Copyright (C) 2017 Paolo Valente <paolo.valente@linaro.org>
  15 *
  16 * BFQ is a proportional-share I/O scheduler, with some extra
  17 * low-latency capabilities. BFQ also supports full hierarchical
  18 * scheduling through cgroups. Next paragraphs provide an introduction
  19 * on BFQ inner workings. Details on BFQ benefits, usage and
  20 * limitations can be found in Documentation/block/bfq-iosched.rst.
  21 *
  22 * BFQ is a proportional-share storage-I/O scheduling algorithm based
  23 * on the slice-by-slice service scheme of CFQ. But BFQ assigns
  24 * budgets, measured in number of sectors, to processes instead of
  25 * time slices. The device is not granted to the in-service process
  26 * for a given time slice, but until it has exhausted its assigned
  27 * budget. This change from the time to the service domain enables BFQ
  28 * to distribute the device throughput among processes as desired,
  29 * without any distortion due to throughput fluctuations, or to device
  30 * internal queueing. BFQ uses an ad hoc internal scheduler, called
  31 * B-WF2Q+, to schedule processes according to their budgets. More
  32 * precisely, BFQ schedules queues associated with processes. Each
  33 * process/queue is assigned a user-configurable weight, and B-WF2Q+
  34 * guarantees that each queue receives a fraction of the throughput
  35 * proportional to its weight. Thanks to the accurate policy of
  36 * B-WF2Q+, BFQ can afford to assign high budgets to I/O-bound
  37 * processes issuing sequential requests (to boost the throughput),
  38 * and yet guarantee a low latency to interactive and soft real-time
  39 * applications.
  40 *
  41 * In particular, to provide these low-latency guarantees, BFQ
  42 * explicitly privileges the I/O of two classes of time-sensitive
  43 * applications: interactive and soft real-time. In more detail, BFQ
  44 * behaves this way if the low_latency parameter is set (default
  45 * configuration). This feature enables BFQ to provide applications in
  46 * these classes with a very low latency.
  47 *
  48 * To implement this feature, BFQ constantly tries to detect whether
  49 * the I/O requests in a bfq_queue come from an interactive or a soft
  50 * real-time application. For brevity, in these cases, the queue is
  51 * said to be interactive or soft real-time. In both cases, BFQ
  52 * privileges the service of the queue, over that of non-interactive
  53 * and non-soft-real-time queues. This privileging is performed,
  54 * mainly, by raising the weight of the queue. So, for brevity, we
  55 * call just weight-raising periods the time periods during which a
  56 * queue is privileged, because deemed interactive or soft real-time.
  57 *
  58 * The detection of soft real-time queues/applications is described in
  59 * detail in the comments on the function
  60 * bfq_bfqq_softrt_next_start. On the other hand, the detection of an
  61 * interactive queue works as follows: a queue is deemed interactive
  62 * if it is constantly non empty only for a limited time interval,
  63 * after which it does become empty. The queue may be deemed
  64 * interactive again (for a limited time), if it restarts being
  65 * constantly non empty, provided that this happens only after the
  66 * queue has remained empty for a given minimum idle time.
  67 *
  68 * By default, BFQ computes automatically the above maximum time
  69 * interval, i.e., the time interval after which a constantly
  70 * non-empty queue stops being deemed interactive. Since a queue is
  71 * weight-raised while it is deemed interactive, this maximum time
  72 * interval happens to coincide with the (maximum) duration of the
  73 * weight-raising for interactive queues.
  74 *
  75 * Finally, BFQ also features additional heuristics for
  76 * preserving both a low latency and a high throughput on NCQ-capable,
  77 * rotational or flash-based devices, and to get the job done quickly
  78 * for applications consisting in many I/O-bound processes.
  79 *
  80 * NOTE: if the main or only goal, with a given device, is to achieve
  81 * the maximum-possible throughput at all times, then do switch off
  82 * all low-latency heuristics for that device, by setting low_latency
  83 * to 0.
  84 *
  85 * BFQ is described in [1], where also a reference to the initial,
  86 * more theoretical paper on BFQ can be found. The interested reader
  87 * can find in the latter paper full details on the main algorithm, as
  88 * well as formulas of the guarantees and formal proofs of all the
  89 * properties.  With respect to the version of BFQ presented in these
  90 * papers, this implementation adds a few more heuristics, such as the
  91 * ones that guarantee a low latency to interactive and soft real-time
  92 * applications, and a hierarchical extension based on H-WF2Q+.
  93 *
  94 * B-WF2Q+ is based on WF2Q+, which is described in [2], together with
  95 * H-WF2Q+, while the augmented tree used here to implement B-WF2Q+
  96 * with O(log N) complexity derives from the one introduced with EEVDF
  97 * in [3].
  98 *
  99 * [1] P. Valente, A. Avanzini, "Evolution of the BFQ Storage I/O
 100 *     Scheduler", Proceedings of the First Workshop on Mobile System
 101 *     Technologies (MST-2015), May 2015.
 102 *     http://algogroup.unimore.it/people/paolo/disk_sched/mst-2015.pdf
 103 *
 104 * [2] Jon C.R. Bennett and H. Zhang, "Hierarchical Packet Fair Queueing
 105 *     Algorithms", IEEE/ACM Transactions on Networking, 5(5):675-689,
 106 *     Oct 1997.
 107 *
 108 * http://www.cs.cmu.edu/~hzhang/papers/TON-97-Oct.ps.gz
 109 *
 110 * [3] I. Stoica and H. Abdel-Wahab, "Earliest Eligible Virtual Deadline
 111 *     First: A Flexible and Accurate Mechanism for Proportional Share
 112 *     Resource Allocation", technical report.
 113 *
 114 * http://www.cs.berkeley.edu/~istoica/papers/eevdf-tr-95.pdf
 115 */
 116#include <linux/module.h>
 117#include <linux/slab.h>
 118#include <linux/blkdev.h>
 119#include <linux/cgroup.h>
 120#include <linux/elevator.h>
 121#include <linux/ktime.h>
 122#include <linux/rbtree.h>
 123#include <linux/ioprio.h>
 124#include <linux/sbitmap.h>
 125#include <linux/delay.h>
 126#include <linux/backing-dev.h>
 127
 
 
 
 128#include "blk.h"
 129#include "blk-mq.h"
 130#include "blk-mq-tag.h"
 131#include "blk-mq-sched.h"
 132#include "bfq-iosched.h"
 133#include "blk-wbt.h"
 134
 135#define BFQ_BFQQ_FNS(name)						\
 136void bfq_mark_bfqq_##name(struct bfq_queue *bfqq)			\
 137{									\
 138	__set_bit(BFQQF_##name, &(bfqq)->flags);			\
 139}									\
 140void bfq_clear_bfqq_##name(struct bfq_queue *bfqq)			\
 141{									\
 142	__clear_bit(BFQQF_##name, &(bfqq)->flags);		\
 143}									\
 144int bfq_bfqq_##name(const struct bfq_queue *bfqq)			\
 145{									\
 146	return test_bit(BFQQF_##name, &(bfqq)->flags);		\
 147}
 148
 149BFQ_BFQQ_FNS(just_created);
 150BFQ_BFQQ_FNS(busy);
 151BFQ_BFQQ_FNS(wait_request);
 152BFQ_BFQQ_FNS(non_blocking_wait_rq);
 153BFQ_BFQQ_FNS(fifo_expire);
 154BFQ_BFQQ_FNS(has_short_ttime);
 155BFQ_BFQQ_FNS(sync);
 156BFQ_BFQQ_FNS(IO_bound);
 157BFQ_BFQQ_FNS(in_large_burst);
 158BFQ_BFQQ_FNS(coop);
 159BFQ_BFQQ_FNS(split_coop);
 160BFQ_BFQQ_FNS(softrt_update);
 161BFQ_BFQQ_FNS(has_waker);
 162#undef BFQ_BFQQ_FNS						\
 163
 164/* Expiration time of sync (0) and async (1) requests, in ns. */
 165static const u64 bfq_fifo_expire[2] = { NSEC_PER_SEC / 4, NSEC_PER_SEC / 8 };
 166
 167/* Maximum backwards seek (magic number lifted from CFQ), in KiB. */
 168static const int bfq_back_max = 16 * 1024;
 169
 170/* Penalty of a backwards seek, in number of sectors. */
 171static const int bfq_back_penalty = 2;
 172
 173/* Idling period duration, in ns. */
 174static u64 bfq_slice_idle = NSEC_PER_SEC / 125;
 175
 176/* Minimum number of assigned budgets for which stats are safe to compute. */
 177static const int bfq_stats_min_budgets = 194;
 178
 179/* Default maximum budget values, in sectors and number of requests. */
 180static const int bfq_default_max_budget = 16 * 1024;
 181
 182/*
 183 * When a sync request is dispatched, the queue that contains that
 184 * request, and all the ancestor entities of that queue, are charged
 185 * with the number of sectors of the request. In contrast, if the
 186 * request is async, then the queue and its ancestor entities are
 187 * charged with the number of sectors of the request, multiplied by
 188 * the factor below. This throttles the bandwidth for async I/O,
 189 * w.r.t. to sync I/O, and it is done to counter the tendency of async
 190 * writes to steal I/O throughput to reads.
 191 *
 192 * The current value of this parameter is the result of a tuning with
 193 * several hardware and software configurations. We tried to find the
 194 * lowest value for which writes do not cause noticeable problems to
 195 * reads. In fact, the lower this parameter, the stabler I/O control,
 196 * in the following respect.  The lower this parameter is, the less
 197 * the bandwidth enjoyed by a group decreases
 198 * - when the group does writes, w.r.t. to when it does reads;
 199 * - when other groups do reads, w.r.t. to when they do writes.
 200 */
 201static const int bfq_async_charge_factor = 3;
 202
 203/* Default timeout values, in jiffies, approximating CFQ defaults. */
 204const int bfq_timeout = HZ / 8;
 205
 206/*
 207 * Time limit for merging (see comments in bfq_setup_cooperator). Set
 208 * to the slowest value that, in our tests, proved to be effective in
 209 * removing false positives, while not causing true positives to miss
 210 * queue merging.
 211 *
 212 * As can be deduced from the low time limit below, queue merging, if
 213 * successful, happens at the very beginning of the I/O of the involved
 214 * cooperating processes, as a consequence of the arrival of the very
 215 * first requests from each cooperator.  After that, there is very
 216 * little chance to find cooperators.
 217 */
 218static const unsigned long bfq_merge_time_limit = HZ/10;
 219
 220static struct kmem_cache *bfq_pool;
 221
 222/* Below this threshold (in ns), we consider thinktime immediate. */
 223#define BFQ_MIN_TT		(2 * NSEC_PER_MSEC)
 224
 225/* hw_tag detection: parallel requests threshold and min samples needed. */
 226#define BFQ_HW_QUEUE_THRESHOLD	3
 227#define BFQ_HW_QUEUE_SAMPLES	32
 228
 229#define BFQQ_SEEK_THR		(sector_t)(8 * 100)
 230#define BFQQ_SECT_THR_NONROT	(sector_t)(2 * 32)
 231#define BFQ_RQ_SEEKY(bfqd, last_pos, rq) \
 232	(get_sdist(last_pos, rq) >			\
 233	 BFQQ_SEEK_THR &&				\
 234	 (!blk_queue_nonrot(bfqd->queue) ||		\
 235	  blk_rq_sectors(rq) < BFQQ_SECT_THR_NONROT))
 236#define BFQQ_CLOSE_THR		(sector_t)(8 * 1024)
 237#define BFQQ_SEEKY(bfqq)	(hweight32(bfqq->seek_history) > 19)
 238/*
 239 * Sync random I/O is likely to be confused with soft real-time I/O,
 240 * because it is characterized by limited throughput and apparently
 241 * isochronous arrival pattern. To avoid false positives, queues
 242 * containing only random (seeky) I/O are prevented from being tagged
 243 * as soft real-time.
 244 */
 245#define BFQQ_TOTALLY_SEEKY(bfqq)	(bfqq->seek_history == -1)
 246
 247/* Min number of samples required to perform peak-rate update */
 248#define BFQ_RATE_MIN_SAMPLES	32
 249/* Min observation time interval required to perform a peak-rate update (ns) */
 250#define BFQ_RATE_MIN_INTERVAL	(300*NSEC_PER_MSEC)
 251/* Target observation time interval for a peak-rate update (ns) */
 252#define BFQ_RATE_REF_INTERVAL	NSEC_PER_SEC
 253
 254/*
 255 * Shift used for peak-rate fixed precision calculations.
 256 * With
 257 * - the current shift: 16 positions
 258 * - the current type used to store rate: u32
 259 * - the current unit of measure for rate: [sectors/usec], or, more precisely,
 260 *   [(sectors/usec) / 2^BFQ_RATE_SHIFT] to take into account the shift,
 261 * the range of rates that can be stored is
 262 * [1 / 2^BFQ_RATE_SHIFT, 2^(32 - BFQ_RATE_SHIFT)] sectors/usec =
 263 * [1 / 2^16, 2^16] sectors/usec = [15e-6, 65536] sectors/usec =
 264 * [15, 65G] sectors/sec
 265 * Which, assuming a sector size of 512B, corresponds to a range of
 266 * [7.5K, 33T] B/sec
 267 */
 268#define BFQ_RATE_SHIFT		16
 269
 270/*
 271 * When configured for computing the duration of the weight-raising
 272 * for interactive queues automatically (see the comments at the
 273 * beginning of this file), BFQ does it using the following formula:
 274 * duration = (ref_rate / r) * ref_wr_duration,
 275 * where r is the peak rate of the device, and ref_rate and
 276 * ref_wr_duration are two reference parameters.  In particular,
 277 * ref_rate is the peak rate of the reference storage device (see
 278 * below), and ref_wr_duration is about the maximum time needed, with
 279 * BFQ and while reading two files in parallel, to load typical large
 280 * applications on the reference device (see the comments on
 281 * max_service_from_wr below, for more details on how ref_wr_duration
 282 * is obtained).  In practice, the slower/faster the device at hand
 283 * is, the more/less it takes to load applications with respect to the
 284 * reference device.  Accordingly, the longer/shorter BFQ grants
 285 * weight raising to interactive applications.
 286 *
 287 * BFQ uses two different reference pairs (ref_rate, ref_wr_duration),
 288 * depending on whether the device is rotational or non-rotational.
 289 *
 290 * In the following definitions, ref_rate[0] and ref_wr_duration[0]
 291 * are the reference values for a rotational device, whereas
 292 * ref_rate[1] and ref_wr_duration[1] are the reference values for a
 293 * non-rotational device. The reference rates are not the actual peak
 294 * rates of the devices used as a reference, but slightly lower
 295 * values. The reason for using slightly lower values is that the
 296 * peak-rate estimator tends to yield slightly lower values than the
 297 * actual peak rate (it can yield the actual peak rate only if there
 298 * is only one process doing I/O, and the process does sequential
 299 * I/O).
 300 *
 301 * The reference peak rates are measured in sectors/usec, left-shifted
 302 * by BFQ_RATE_SHIFT.
 303 */
 304static int ref_rate[2] = {14000, 33000};
 305/*
 306 * To improve readability, a conversion function is used to initialize
 307 * the following array, which entails that the array can be
 308 * initialized only in a function.
 309 */
 310static int ref_wr_duration[2];
 311
 312/*
 313 * BFQ uses the above-detailed, time-based weight-raising mechanism to
 314 * privilege interactive tasks. This mechanism is vulnerable to the
 315 * following false positives: I/O-bound applications that will go on
 316 * doing I/O for much longer than the duration of weight
 317 * raising. These applications have basically no benefit from being
 318 * weight-raised at the beginning of their I/O. On the opposite end,
 319 * while being weight-raised, these applications
 320 * a) unjustly steal throughput to applications that may actually need
 321 * low latency;
 322 * b) make BFQ uselessly perform device idling; device idling results
 323 * in loss of device throughput with most flash-based storage, and may
 324 * increase latencies when used purposelessly.
 325 *
 326 * BFQ tries to reduce these problems, by adopting the following
 327 * countermeasure. To introduce this countermeasure, we need first to
 328 * finish explaining how the duration of weight-raising for
 329 * interactive tasks is computed.
 330 *
 331 * For a bfq_queue deemed as interactive, the duration of weight
 332 * raising is dynamically adjusted, as a function of the estimated
 333 * peak rate of the device, so as to be equal to the time needed to
 334 * execute the 'largest' interactive task we benchmarked so far. By
 335 * largest task, we mean the task for which each involved process has
 336 * to do more I/O than for any of the other tasks we benchmarked. This
 337 * reference interactive task is the start-up of LibreOffice Writer,
 338 * and in this task each process/bfq_queue needs to have at most ~110K
 339 * sectors transferred.
 340 *
 341 * This last piece of information enables BFQ to reduce the actual
 342 * duration of weight-raising for at least one class of I/O-bound
 343 * applications: those doing sequential or quasi-sequential I/O. An
 344 * example is file copy. In fact, once started, the main I/O-bound
 345 * processes of these applications usually consume the above 110K
 346 * sectors in much less time than the processes of an application that
 347 * is starting, because these I/O-bound processes will greedily devote
 348 * almost all their CPU cycles only to their target,
 349 * throughput-friendly I/O operations. This is even more true if BFQ
 350 * happens to be underestimating the device peak rate, and thus
 351 * overestimating the duration of weight raising. But, according to
 352 * our measurements, once transferred 110K sectors, these processes
 353 * have no right to be weight-raised any longer.
 354 *
 355 * Basing on the last consideration, BFQ ends weight-raising for a
 356 * bfq_queue if the latter happens to have received an amount of
 357 * service at least equal to the following constant. The constant is
 358 * set to slightly more than 110K, to have a minimum safety margin.
 359 *
 360 * This early ending of weight-raising reduces the amount of time
 361 * during which interactive false positives cause the two problems
 362 * described at the beginning of these comments.
 363 */
 364static const unsigned long max_service_from_wr = 120000;
 365
 366#define RQ_BIC(rq)		icq_to_bic((rq)->elv.priv[0])
 
 
 
 
 
 
 
 
 
 
 367#define RQ_BFQQ(rq)		((rq)->elv.priv[1])
 368
 369struct bfq_queue *bic_to_bfqq(struct bfq_io_cq *bic, bool is_sync)
 370{
 371	return bic->bfqq[is_sync];
 372}
 373
 
 
 374void bic_set_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq, bool is_sync)
 375{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 376	bic->bfqq[is_sync] = bfqq;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 377}
 378
 379struct bfq_data *bic_to_bfqd(struct bfq_io_cq *bic)
 380{
 381	return bic->icq.q->elevator->elevator_data;
 382}
 383
 384/**
 385 * icq_to_bic - convert iocontext queue structure to bfq_io_cq.
 386 * @icq: the iocontext queue.
 387 */
 388static struct bfq_io_cq *icq_to_bic(struct io_cq *icq)
 389{
 390	/* bic->icq is the first member, %NULL will convert to %NULL */
 391	return container_of(icq, struct bfq_io_cq, icq);
 392}
 393
 394/**
 395 * bfq_bic_lookup - search into @ioc a bic associated to @bfqd.
 396 * @bfqd: the lookup key.
 397 * @ioc: the io_context of the process doing I/O.
 398 * @q: the request queue.
 399 */
 400static struct bfq_io_cq *bfq_bic_lookup(struct bfq_data *bfqd,
 401					struct io_context *ioc,
 402					struct request_queue *q)
 403{
 404	if (ioc) {
 405		unsigned long flags;
 406		struct bfq_io_cq *icq;
 407
 408		spin_lock_irqsave(&q->queue_lock, flags);
 409		icq = icq_to_bic(ioc_lookup_icq(ioc, q));
 410		spin_unlock_irqrestore(&q->queue_lock, flags);
 411
 412		return icq;
 413	}
 
 414
 415	return NULL;
 416}
 417
 418/*
 419 * Scheduler run of queue, if there are requests pending and no one in the
 420 * driver that will restart queueing.
 421 */
 422void bfq_schedule_dispatch(struct bfq_data *bfqd)
 423{
 
 
 424	if (bfqd->queued != 0) {
 425		bfq_log(bfqd, "schedule dispatch");
 426		blk_mq_run_hw_queues(bfqd->queue, true);
 427	}
 428}
 429
 430#define bfq_class_idle(bfqq)	((bfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
 431
 432#define bfq_sample_valid(samples)	((samples) > 80)
 433
 434/*
 435 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
 436 * We choose the request that is closer to the head right now.  Distance
 437 * behind the head is penalized and only allowed to a certain extent.
 438 */
 439static struct request *bfq_choose_req(struct bfq_data *bfqd,
 440				      struct request *rq1,
 441				      struct request *rq2,
 442				      sector_t last)
 443{
 444	sector_t s1, s2, d1 = 0, d2 = 0;
 445	unsigned long back_max;
 446#define BFQ_RQ1_WRAP	0x01 /* request 1 wraps */
 447#define BFQ_RQ2_WRAP	0x02 /* request 2 wraps */
 448	unsigned int wrap = 0; /* bit mask: requests behind the disk head? */
 449
 450	if (!rq1 || rq1 == rq2)
 451		return rq2;
 452	if (!rq2)
 453		return rq1;
 454
 455	if (rq_is_sync(rq1) && !rq_is_sync(rq2))
 456		return rq1;
 457	else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
 458		return rq2;
 459	if ((rq1->cmd_flags & REQ_META) && !(rq2->cmd_flags & REQ_META))
 460		return rq1;
 461	else if ((rq2->cmd_flags & REQ_META) && !(rq1->cmd_flags & REQ_META))
 462		return rq2;
 463
 464	s1 = blk_rq_pos(rq1);
 465	s2 = blk_rq_pos(rq2);
 466
 467	/*
 468	 * By definition, 1KiB is 2 sectors.
 469	 */
 470	back_max = bfqd->bfq_back_max * 2;
 471
 472	/*
 473	 * Strict one way elevator _except_ in the case where we allow
 474	 * short backward seeks which are biased as twice the cost of a
 475	 * similar forward seek.
 476	 */
 477	if (s1 >= last)
 478		d1 = s1 - last;
 479	else if (s1 + back_max >= last)
 480		d1 = (last - s1) * bfqd->bfq_back_penalty;
 481	else
 482		wrap |= BFQ_RQ1_WRAP;
 483
 484	if (s2 >= last)
 485		d2 = s2 - last;
 486	else if (s2 + back_max >= last)
 487		d2 = (last - s2) * bfqd->bfq_back_penalty;
 488	else
 489		wrap |= BFQ_RQ2_WRAP;
 490
 491	/* Found required data */
 492
 493	/*
 494	 * By doing switch() on the bit mask "wrap" we avoid having to
 495	 * check two variables for all permutations: --> faster!
 496	 */
 497	switch (wrap) {
 498	case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
 499		if (d1 < d2)
 500			return rq1;
 501		else if (d2 < d1)
 502			return rq2;
 503
 504		if (s1 >= s2)
 505			return rq1;
 506		else
 507			return rq2;
 508
 509	case BFQ_RQ2_WRAP:
 510		return rq1;
 511	case BFQ_RQ1_WRAP:
 512		return rq2;
 513	case BFQ_RQ1_WRAP|BFQ_RQ2_WRAP: /* both rqs wrapped */
 514	default:
 515		/*
 516		 * Since both rqs are wrapped,
 517		 * start with the one that's further behind head
 518		 * (--> only *one* back seek required),
 519		 * since back seek takes more time than forward.
 520		 */
 521		if (s1 <= s2)
 522			return rq1;
 523		else
 524			return rq2;
 525	}
 526}
 527
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 528/*
 529 * Async I/O can easily starve sync I/O (both sync reads and sync
 530 * writes), by consuming all tags. Similarly, storms of sync writes,
 531 * such as those that sync(2) may trigger, can starve sync reads.
 532 * Limit depths of async I/O and sync writes so as to counter both
 533 * problems.
 
 
 
 
 
 
 
 
 534 */
 535static void bfq_limit_depth(unsigned int op, struct blk_mq_alloc_data *data)
 536{
 537	struct bfq_data *bfqd = data->q->elevator->elevator_data;
 
 
 
 
 
 
 
 
 
 
 
 
 538
 539	if (op_is_sync(op) && !op_is_write(op))
 540		return;
 541
 542	data->shallow_depth =
 543		bfqd->word_depths[!!bfqd->wr_busy_queues][op_is_sync(op)];
 
 
 544
 545	bfq_log(bfqd, "[%s] wr_busy %d sync %d depth %u",
 546			__func__, bfqd->wr_busy_queues, op_is_sync(op),
 547			data->shallow_depth);
 
 548}
 549
 550static struct bfq_queue *
 551bfq_rq_pos_tree_lookup(struct bfq_data *bfqd, struct rb_root *root,
 552		     sector_t sector, struct rb_node **ret_parent,
 553		     struct rb_node ***rb_link)
 554{
 555	struct rb_node **p, *parent;
 556	struct bfq_queue *bfqq = NULL;
 557
 558	parent = NULL;
 559	p = &root->rb_node;
 560	while (*p) {
 561		struct rb_node **n;
 562
 563		parent = *p;
 564		bfqq = rb_entry(parent, struct bfq_queue, pos_node);
 565
 566		/*
 567		 * Sort strictly based on sector. Smallest to the left,
 568		 * largest to the right.
 569		 */
 570		if (sector > blk_rq_pos(bfqq->next_rq))
 571			n = &(*p)->rb_right;
 572		else if (sector < blk_rq_pos(bfqq->next_rq))
 573			n = &(*p)->rb_left;
 574		else
 575			break;
 576		p = n;
 577		bfqq = NULL;
 578	}
 579
 580	*ret_parent = parent;
 581	if (rb_link)
 582		*rb_link = p;
 583
 584	bfq_log(bfqd, "rq_pos_tree_lookup %llu: returning %d",
 585		(unsigned long long)sector,
 586		bfqq ? bfqq->pid : 0);
 587
 588	return bfqq;
 589}
 590
 591static bool bfq_too_late_for_merging(struct bfq_queue *bfqq)
 592{
 593	return bfqq->service_from_backlogged > 0 &&
 594		time_is_before_jiffies(bfqq->first_IO_time +
 595				       bfq_merge_time_limit);
 596}
 597
 598/*
 599 * The following function is not marked as __cold because it is
 600 * actually cold, but for the same performance goal described in the
 601 * comments on the likely() at the beginning of
 602 * bfq_setup_cooperator(). Unexpectedly, to reach an even lower
 603 * execution time for the case where this function is not invoked, we
 604 * had to add an unlikely() in each involved if().
 605 */
 606void __cold
 607bfq_pos_tree_add_move(struct bfq_data *bfqd, struct bfq_queue *bfqq)
 608{
 609	struct rb_node **p, *parent;
 610	struct bfq_queue *__bfqq;
 611
 612	if (bfqq->pos_root) {
 613		rb_erase(&bfqq->pos_node, bfqq->pos_root);
 614		bfqq->pos_root = NULL;
 615	}
 616
 617	/* oom_bfqq does not participate in queue merging */
 618	if (bfqq == &bfqd->oom_bfqq)
 619		return;
 620
 621	/*
 622	 * bfqq cannot be merged any longer (see comments in
 623	 * bfq_setup_cooperator): no point in adding bfqq into the
 624	 * position tree.
 625	 */
 626	if (bfq_too_late_for_merging(bfqq))
 627		return;
 628
 629	if (bfq_class_idle(bfqq))
 630		return;
 631	if (!bfqq->next_rq)
 632		return;
 633
 634	bfqq->pos_root = &bfq_bfqq_to_bfqg(bfqq)->rq_pos_tree;
 635	__bfqq = bfq_rq_pos_tree_lookup(bfqd, bfqq->pos_root,
 636			blk_rq_pos(bfqq->next_rq), &parent, &p);
 637	if (!__bfqq) {
 638		rb_link_node(&bfqq->pos_node, parent, p);
 639		rb_insert_color(&bfqq->pos_node, bfqq->pos_root);
 640	} else
 641		bfqq->pos_root = NULL;
 642}
 643
 644/*
 645 * The following function returns false either if every active queue
 646 * must receive the same share of the throughput (symmetric scenario),
 647 * or, as a special case, if bfqq must receive a share of the
 648 * throughput lower than or equal to the share that every other active
 649 * queue must receive.  If bfqq does sync I/O, then these are the only
 650 * two cases where bfqq happens to be guaranteed its share of the
 651 * throughput even if I/O dispatching is not plugged when bfqq remains
 652 * temporarily empty (for more details, see the comments in the
 653 * function bfq_better_to_idle()). For this reason, the return value
 654 * of this function is used to check whether I/O-dispatch plugging can
 655 * be avoided.
 656 *
 657 * The above first case (symmetric scenario) occurs when:
 658 * 1) all active queues have the same weight,
 659 * 2) all active queues belong to the same I/O-priority class,
 660 * 3) all active groups at the same level in the groups tree have the same
 661 *    weight,
 662 * 4) all active groups at the same level in the groups tree have the same
 663 *    number of children.
 664 *
 665 * Unfortunately, keeping the necessary state for evaluating exactly
 666 * the last two symmetry sub-conditions above would be quite complex
 667 * and time consuming. Therefore this function evaluates, instead,
 668 * only the following stronger three sub-conditions, for which it is
 669 * much easier to maintain the needed state:
 670 * 1) all active queues have the same weight,
 671 * 2) all active queues belong to the same I/O-priority class,
 672 * 3) there are no active groups.
 673 * In particular, the last condition is always true if hierarchical
 674 * support or the cgroups interface are not enabled, thus no state
 675 * needs to be maintained in this case.
 676 */
 677static bool bfq_asymmetric_scenario(struct bfq_data *bfqd,
 678				   struct bfq_queue *bfqq)
 679{
 680	bool smallest_weight = bfqq &&
 681		bfqq->weight_counter &&
 682		bfqq->weight_counter ==
 683		container_of(
 684			rb_first_cached(&bfqd->queue_weights_tree),
 685			struct bfq_weight_counter,
 686			weights_node);
 687
 688	/*
 689	 * For queue weights to differ, queue_weights_tree must contain
 690	 * at least two nodes.
 691	 */
 692	bool varied_queue_weights = !smallest_weight &&
 693		!RB_EMPTY_ROOT(&bfqd->queue_weights_tree.rb_root) &&
 694		(bfqd->queue_weights_tree.rb_root.rb_node->rb_left ||
 695		 bfqd->queue_weights_tree.rb_root.rb_node->rb_right);
 696
 697	bool multiple_classes_busy =
 698		(bfqd->busy_queues[0] && bfqd->busy_queues[1]) ||
 699		(bfqd->busy_queues[0] && bfqd->busy_queues[2]) ||
 700		(bfqd->busy_queues[1] && bfqd->busy_queues[2]);
 701
 702	return varied_queue_weights || multiple_classes_busy
 703#ifdef CONFIG_BFQ_GROUP_IOSCHED
 704	       || bfqd->num_groups_with_pending_reqs > 0
 705#endif
 706		;
 707}
 708
 709/*
 710 * If the weight-counter tree passed as input contains no counter for
 711 * the weight of the input queue, then add that counter; otherwise just
 712 * increment the existing counter.
 713 *
 714 * Note that weight-counter trees contain few nodes in mostly symmetric
 715 * scenarios. For example, if all queues have the same weight, then the
 716 * weight-counter tree for the queues may contain at most one node.
 717 * This holds even if low_latency is on, because weight-raised queues
 718 * are not inserted in the tree.
 719 * In most scenarios, the rate at which nodes are created/destroyed
 720 * should be low too.
 721 */
 722void bfq_weights_tree_add(struct bfq_data *bfqd, struct bfq_queue *bfqq,
 723			  struct rb_root_cached *root)
 724{
 
 725	struct bfq_entity *entity = &bfqq->entity;
 726	struct rb_node **new = &(root->rb_root.rb_node), *parent = NULL;
 727	bool leftmost = true;
 728
 729	/*
 730	 * Do not insert if the queue is already associated with a
 731	 * counter, which happens if:
 732	 *   1) a request arrival has caused the queue to become both
 733	 *      non-weight-raised, and hence change its weight, and
 734	 *      backlogged; in this respect, each of the two events
 735	 *      causes an invocation of this function,
 736	 *   2) this is the invocation of this function caused by the
 737	 *      second event. This second invocation is actually useless,
 738	 *      and we handle this fact by exiting immediately. More
 739	 *      efficient or clearer solutions might possibly be adopted.
 740	 */
 741	if (bfqq->weight_counter)
 742		return;
 743
 744	while (*new) {
 745		struct bfq_weight_counter *__counter = container_of(*new,
 746						struct bfq_weight_counter,
 747						weights_node);
 748		parent = *new;
 749
 750		if (entity->weight == __counter->weight) {
 751			bfqq->weight_counter = __counter;
 752			goto inc_counter;
 753		}
 754		if (entity->weight < __counter->weight)
 755			new = &((*new)->rb_left);
 756		else {
 757			new = &((*new)->rb_right);
 758			leftmost = false;
 759		}
 760	}
 761
 762	bfqq->weight_counter = kzalloc(sizeof(struct bfq_weight_counter),
 763				       GFP_ATOMIC);
 764
 765	/*
 766	 * In the unlucky event of an allocation failure, we just
 767	 * exit. This will cause the weight of queue to not be
 768	 * considered in bfq_asymmetric_scenario, which, in its turn,
 769	 * causes the scenario to be deemed wrongly symmetric in case
 770	 * bfqq's weight would have been the only weight making the
 771	 * scenario asymmetric.  On the bright side, no unbalance will
 772	 * however occur when bfqq becomes inactive again (the
 773	 * invocation of this function is triggered by an activation
 774	 * of queue).  In fact, bfq_weights_tree_remove does nothing
 775	 * if !bfqq->weight_counter.
 776	 */
 777	if (unlikely(!bfqq->weight_counter))
 778		return;
 779
 780	bfqq->weight_counter->weight = entity->weight;
 781	rb_link_node(&bfqq->weight_counter->weights_node, parent, new);
 782	rb_insert_color_cached(&bfqq->weight_counter->weights_node, root,
 783				leftmost);
 784
 785inc_counter:
 786	bfqq->weight_counter->num_active++;
 787	bfqq->ref++;
 788}
 789
 790/*
 791 * Decrement the weight counter associated with the queue, and, if the
 792 * counter reaches 0, remove the counter from the tree.
 793 * See the comments to the function bfq_weights_tree_add() for considerations
 794 * about overhead.
 795 */
 796void __bfq_weights_tree_remove(struct bfq_data *bfqd,
 797			       struct bfq_queue *bfqq,
 798			       struct rb_root_cached *root)
 799{
 
 
 800	if (!bfqq->weight_counter)
 801		return;
 802
 
 803	bfqq->weight_counter->num_active--;
 804	if (bfqq->weight_counter->num_active > 0)
 805		goto reset_entity_pointer;
 806
 807	rb_erase_cached(&bfqq->weight_counter->weights_node, root);
 808	kfree(bfqq->weight_counter);
 809
 810reset_entity_pointer:
 811	bfqq->weight_counter = NULL;
 812	bfq_put_queue(bfqq);
 813}
 814
 815/*
 816 * Invoke __bfq_weights_tree_remove on bfqq and decrement the number
 817 * of active groups for each queue's inactive parent entity.
 818 */
 819void bfq_weights_tree_remove(struct bfq_data *bfqd,
 820			     struct bfq_queue *bfqq)
 821{
 822	struct bfq_entity *entity = bfqq->entity.parent;
 823
 824	for_each_entity(entity) {
 825		struct bfq_sched_data *sd = entity->my_sched_data;
 826
 827		if (sd->next_in_service || sd->in_service_entity) {
 828			/*
 829			 * entity is still active, because either
 830			 * next_in_service or in_service_entity is not
 831			 * NULL (see the comments on the definition of
 832			 * next_in_service for details on why
 833			 * in_service_entity must be checked too).
 834			 *
 835			 * As a consequence, its parent entities are
 836			 * active as well, and thus this loop must
 837			 * stop here.
 838			 */
 839			break;
 840		}
 841
 842		/*
 843		 * The decrement of num_groups_with_pending_reqs is
 844		 * not performed immediately upon the deactivation of
 845		 * entity, but it is delayed to when it also happens
 846		 * that the first leaf descendant bfqq of entity gets
 847		 * all its pending requests completed. The following
 848		 * instructions perform this delayed decrement, if
 849		 * needed. See the comments on
 850		 * num_groups_with_pending_reqs for details.
 851		 */
 852		if (entity->in_groups_with_pending_reqs) {
 853			entity->in_groups_with_pending_reqs = false;
 854			bfqd->num_groups_with_pending_reqs--;
 855		}
 856	}
 857
 858	/*
 859	 * Next function is invoked last, because it causes bfqq to be
 860	 * freed if the following holds: bfqq is not in service and
 861	 * has no dispatched request. DO NOT use bfqq after the next
 862	 * function invocation.
 863	 */
 864	__bfq_weights_tree_remove(bfqd, bfqq,
 865				  &bfqd->queue_weights_tree);
 866}
 867
 868/*
 869 * Return expired entry, or NULL to just start from scratch in rbtree.
 870 */
 871static struct request *bfq_check_fifo(struct bfq_queue *bfqq,
 872				      struct request *last)
 873{
 874	struct request *rq;
 875
 876	if (bfq_bfqq_fifo_expire(bfqq))
 877		return NULL;
 878
 879	bfq_mark_bfqq_fifo_expire(bfqq);
 880
 881	rq = rq_entry_fifo(bfqq->fifo.next);
 882
 883	if (rq == last || ktime_get_ns() < rq->fifo_time)
 884		return NULL;
 885
 886	bfq_log_bfqq(bfqq->bfqd, bfqq, "check_fifo: returned %p", rq);
 887	return rq;
 888}
 889
 890static struct request *bfq_find_next_rq(struct bfq_data *bfqd,
 891					struct bfq_queue *bfqq,
 892					struct request *last)
 893{
 894	struct rb_node *rbnext = rb_next(&last->rb_node);
 895	struct rb_node *rbprev = rb_prev(&last->rb_node);
 896	struct request *next, *prev = NULL;
 897
 898	/* Follow expired path, else get first next available. */
 899	next = bfq_check_fifo(bfqq, last);
 900	if (next)
 901		return next;
 902
 903	if (rbprev)
 904		prev = rb_entry_rq(rbprev);
 905
 906	if (rbnext)
 907		next = rb_entry_rq(rbnext);
 908	else {
 909		rbnext = rb_first(&bfqq->sort_list);
 910		if (rbnext && rbnext != &last->rb_node)
 911			next = rb_entry_rq(rbnext);
 912	}
 913
 914	return bfq_choose_req(bfqd, next, prev, blk_rq_pos(last));
 915}
 916
 917/* see the definition of bfq_async_charge_factor for details */
 918static unsigned long bfq_serv_to_charge(struct request *rq,
 919					struct bfq_queue *bfqq)
 920{
 921	if (bfq_bfqq_sync(bfqq) || bfqq->wr_coeff > 1 ||
 922	    bfq_asymmetric_scenario(bfqq->bfqd, bfqq))
 923		return blk_rq_sectors(rq);
 924
 925	return blk_rq_sectors(rq) * bfq_async_charge_factor;
 926}
 927
 928/**
 929 * bfq_updated_next_req - update the queue after a new next_rq selection.
 930 * @bfqd: the device data the queue belongs to.
 931 * @bfqq: the queue to update.
 932 *
 933 * If the first request of a queue changes we make sure that the queue
 934 * has enough budget to serve at least its first request (if the
 935 * request has grown).  We do this because if the queue has not enough
 936 * budget for its first request, it has to go through two dispatch
 937 * rounds to actually get it dispatched.
 938 */
 939static void bfq_updated_next_req(struct bfq_data *bfqd,
 940				 struct bfq_queue *bfqq)
 941{
 942	struct bfq_entity *entity = &bfqq->entity;
 943	struct request *next_rq = bfqq->next_rq;
 944	unsigned long new_budget;
 945
 946	if (!next_rq)
 947		return;
 948
 949	if (bfqq == bfqd->in_service_queue)
 950		/*
 951		 * In order not to break guarantees, budgets cannot be
 952		 * changed after an entity has been selected.
 953		 */
 954		return;
 955
 956	new_budget = max_t(unsigned long,
 957			   max_t(unsigned long, bfqq->max_budget,
 958				 bfq_serv_to_charge(next_rq, bfqq)),
 959			   entity->service);
 960	if (entity->budget != new_budget) {
 961		entity->budget = new_budget;
 962		bfq_log_bfqq(bfqd, bfqq, "updated next rq: new budget %lu",
 963					 new_budget);
 964		bfq_requeue_bfqq(bfqd, bfqq, false);
 965	}
 966}
 967
 968static unsigned int bfq_wr_duration(struct bfq_data *bfqd)
 969{
 970	u64 dur;
 971
 972	if (bfqd->bfq_wr_max_time > 0)
 973		return bfqd->bfq_wr_max_time;
 974
 975	dur = bfqd->rate_dur_prod;
 976	do_div(dur, bfqd->peak_rate);
 977
 978	/*
 979	 * Limit duration between 3 and 25 seconds. The upper limit
 980	 * has been conservatively set after the following worst case:
 981	 * on a QEMU/KVM virtual machine
 982	 * - running in a slow PC
 983	 * - with a virtual disk stacked on a slow low-end 5400rpm HDD
 984	 * - serving a heavy I/O workload, such as the sequential reading
 985	 *   of several files
 986	 * mplayer took 23 seconds to start, if constantly weight-raised.
 987	 *
 988	 * As for higher values than that accommodating the above bad
 989	 * scenario, tests show that higher values would often yield
 990	 * the opposite of the desired result, i.e., would worsen
 991	 * responsiveness by allowing non-interactive applications to
 992	 * preserve weight raising for too long.
 993	 *
 994	 * On the other end, lower values than 3 seconds make it
 995	 * difficult for most interactive tasks to complete their jobs
 996	 * before weight-raising finishes.
 997	 */
 998	return clamp_val(dur, msecs_to_jiffies(3000), msecs_to_jiffies(25000));
 999}
1000
1001/* switch back from soft real-time to interactive weight raising */
1002static void switch_back_to_interactive_wr(struct bfq_queue *bfqq,
1003					  struct bfq_data *bfqd)
1004{
1005	bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1006	bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1007	bfqq->last_wr_start_finish = bfqq->wr_start_at_switch_to_srt;
1008}
1009
1010static void
1011bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd,
1012		      struct bfq_io_cq *bic, bool bfq_already_existing)
1013{
1014	unsigned int old_wr_coeff = bfqq->wr_coeff;
1015	bool busy = bfq_already_existing && bfq_bfqq_busy(bfqq);
1016
1017	if (bic->saved_has_short_ttime)
1018		bfq_mark_bfqq_has_short_ttime(bfqq);
1019	else
1020		bfq_clear_bfqq_has_short_ttime(bfqq);
1021
1022	if (bic->saved_IO_bound)
1023		bfq_mark_bfqq_IO_bound(bfqq);
1024	else
1025		bfq_clear_bfqq_IO_bound(bfqq);
1026
 
 
 
 
1027	bfqq->entity.new_weight = bic->saved_weight;
1028	bfqq->ttime = bic->saved_ttime;
1029	bfqq->wr_coeff = bic->saved_wr_coeff;
 
 
 
 
 
 
 
 
 
1030	bfqq->wr_start_at_switch_to_srt = bic->saved_wr_start_at_switch_to_srt;
1031	bfqq->last_wr_start_finish = bic->saved_last_wr_start_finish;
1032	bfqq->wr_cur_max_time = bic->saved_wr_cur_max_time;
1033
1034	if (bfqq->wr_coeff > 1 && (bfq_bfqq_in_large_burst(bfqq) ||
1035	    time_is_before_jiffies(bfqq->last_wr_start_finish +
1036				   bfqq->wr_cur_max_time))) {
1037		if (bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
1038		    !bfq_bfqq_in_large_burst(bfqq) &&
1039		    time_is_after_eq_jiffies(bfqq->wr_start_at_switch_to_srt +
1040					     bfq_wr_duration(bfqd))) {
1041			switch_back_to_interactive_wr(bfqq, bfqd);
1042		} else {
1043			bfqq->wr_coeff = 1;
1044			bfq_log_bfqq(bfqq->bfqd, bfqq,
1045				     "resume state: switching off wr");
1046		}
1047	}
1048
1049	/* make sure weight will be updated, however we got here */
1050	bfqq->entity.prio_changed = 1;
1051
1052	if (likely(!busy))
1053		return;
1054
1055	if (old_wr_coeff == 1 && bfqq->wr_coeff > 1)
1056		bfqd->wr_busy_queues++;
1057	else if (old_wr_coeff > 1 && bfqq->wr_coeff == 1)
1058		bfqd->wr_busy_queues--;
1059}
1060
1061static int bfqq_process_refs(struct bfq_queue *bfqq)
1062{
1063	return bfqq->ref - bfqq->allocated - bfqq->entity.on_st_or_in_serv -
1064		(bfqq->weight_counter != NULL);
 
1065}
1066
1067/* Empty burst list and add just bfqq (see comments on bfq_handle_burst) */
1068static void bfq_reset_burst_list(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1069{
1070	struct bfq_queue *item;
1071	struct hlist_node *n;
1072
1073	hlist_for_each_entry_safe(item, n, &bfqd->burst_list, burst_list_node)
1074		hlist_del_init(&item->burst_list_node);
1075
1076	/*
1077	 * Start the creation of a new burst list only if there is no
1078	 * active queue. See comments on the conditional invocation of
1079	 * bfq_handle_burst().
1080	 */
1081	if (bfq_tot_busy_queues(bfqd) == 0) {
1082		hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list);
1083		bfqd->burst_size = 1;
1084	} else
1085		bfqd->burst_size = 0;
1086
1087	bfqd->burst_parent_entity = bfqq->entity.parent;
1088}
1089
1090/* Add bfqq to the list of queues in current burst (see bfq_handle_burst) */
1091static void bfq_add_to_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1092{
1093	/* Increment burst size to take into account also bfqq */
1094	bfqd->burst_size++;
1095
1096	if (bfqd->burst_size == bfqd->bfq_large_burst_thresh) {
1097		struct bfq_queue *pos, *bfqq_item;
1098		struct hlist_node *n;
1099
1100		/*
1101		 * Enough queues have been activated shortly after each
1102		 * other to consider this burst as large.
1103		 */
1104		bfqd->large_burst = true;
1105
1106		/*
1107		 * We can now mark all queues in the burst list as
1108		 * belonging to a large burst.
1109		 */
1110		hlist_for_each_entry(bfqq_item, &bfqd->burst_list,
1111				     burst_list_node)
1112			bfq_mark_bfqq_in_large_burst(bfqq_item);
1113		bfq_mark_bfqq_in_large_burst(bfqq);
1114
1115		/*
1116		 * From now on, and until the current burst finishes, any
1117		 * new queue being activated shortly after the last queue
1118		 * was inserted in the burst can be immediately marked as
1119		 * belonging to a large burst. So the burst list is not
1120		 * needed any more. Remove it.
1121		 */
1122		hlist_for_each_entry_safe(pos, n, &bfqd->burst_list,
1123					  burst_list_node)
1124			hlist_del_init(&pos->burst_list_node);
1125	} else /*
1126		* Burst not yet large: add bfqq to the burst list. Do
1127		* not increment the ref counter for bfqq, because bfqq
1128		* is removed from the burst list before freeing bfqq
1129		* in put_queue.
1130		*/
1131		hlist_add_head(&bfqq->burst_list_node, &bfqd->burst_list);
1132}
1133
1134/*
1135 * If many queues belonging to the same group happen to be created
1136 * shortly after each other, then the processes associated with these
1137 * queues have typically a common goal. In particular, bursts of queue
1138 * creations are usually caused by services or applications that spawn
1139 * many parallel threads/processes. Examples are systemd during boot,
1140 * or git grep. To help these processes get their job done as soon as
1141 * possible, it is usually better to not grant either weight-raising
1142 * or device idling to their queues, unless these queues must be
1143 * protected from the I/O flowing through other active queues.
1144 *
1145 * In this comment we describe, firstly, the reasons why this fact
1146 * holds, and, secondly, the next function, which implements the main
1147 * steps needed to properly mark these queues so that they can then be
1148 * treated in a different way.
1149 *
1150 * The above services or applications benefit mostly from a high
1151 * throughput: the quicker the requests of the activated queues are
1152 * cumulatively served, the sooner the target job of these queues gets
1153 * completed. As a consequence, weight-raising any of these queues,
1154 * which also implies idling the device for it, is almost always
1155 * counterproductive, unless there are other active queues to isolate
1156 * these new queues from. If there no other active queues, then
1157 * weight-raising these new queues just lowers throughput in most
1158 * cases.
1159 *
1160 * On the other hand, a burst of queue creations may be caused also by
1161 * the start of an application that does not consist of a lot of
1162 * parallel I/O-bound threads. In fact, with a complex application,
1163 * several short processes may need to be executed to start-up the
1164 * application. In this respect, to start an application as quickly as
1165 * possible, the best thing to do is in any case to privilege the I/O
1166 * related to the application with respect to all other
1167 * I/O. Therefore, the best strategy to start as quickly as possible
1168 * an application that causes a burst of queue creations is to
1169 * weight-raise all the queues created during the burst. This is the
1170 * exact opposite of the best strategy for the other type of bursts.
1171 *
1172 * In the end, to take the best action for each of the two cases, the
1173 * two types of bursts need to be distinguished. Fortunately, this
1174 * seems relatively easy, by looking at the sizes of the bursts. In
1175 * particular, we found a threshold such that only bursts with a
1176 * larger size than that threshold are apparently caused by
1177 * services or commands such as systemd or git grep. For brevity,
1178 * hereafter we call just 'large' these bursts. BFQ *does not*
1179 * weight-raise queues whose creation occurs in a large burst. In
1180 * addition, for each of these queues BFQ performs or does not perform
1181 * idling depending on which choice boosts the throughput more. The
1182 * exact choice depends on the device and request pattern at
1183 * hand.
1184 *
1185 * Unfortunately, false positives may occur while an interactive task
1186 * is starting (e.g., an application is being started). The
1187 * consequence is that the queues associated with the task do not
1188 * enjoy weight raising as expected. Fortunately these false positives
1189 * are very rare. They typically occur if some service happens to
1190 * start doing I/O exactly when the interactive task starts.
1191 *
1192 * Turning back to the next function, it is invoked only if there are
1193 * no active queues (apart from active queues that would belong to the
1194 * same, possible burst bfqq would belong to), and it implements all
1195 * the steps needed to detect the occurrence of a large burst and to
1196 * properly mark all the queues belonging to it (so that they can then
1197 * be treated in a different way). This goal is achieved by
1198 * maintaining a "burst list" that holds, temporarily, the queues that
1199 * belong to the burst in progress. The list is then used to mark
1200 * these queues as belonging to a large burst if the burst does become
1201 * large. The main steps are the following.
1202 *
1203 * . when the very first queue is created, the queue is inserted into the
1204 *   list (as it could be the first queue in a possible burst)
1205 *
1206 * . if the current burst has not yet become large, and a queue Q that does
1207 *   not yet belong to the burst is activated shortly after the last time
1208 *   at which a new queue entered the burst list, then the function appends
1209 *   Q to the burst list
1210 *
1211 * . if, as a consequence of the previous step, the burst size reaches
1212 *   the large-burst threshold, then
1213 *
1214 *     . all the queues in the burst list are marked as belonging to a
1215 *       large burst
1216 *
1217 *     . the burst list is deleted; in fact, the burst list already served
1218 *       its purpose (keeping temporarily track of the queues in a burst,
1219 *       so as to be able to mark them as belonging to a large burst in the
1220 *       previous sub-step), and now is not needed any more
1221 *
1222 *     . the device enters a large-burst mode
1223 *
1224 * . if a queue Q that does not belong to the burst is created while
1225 *   the device is in large-burst mode and shortly after the last time
1226 *   at which a queue either entered the burst list or was marked as
1227 *   belonging to the current large burst, then Q is immediately marked
1228 *   as belonging to a large burst.
1229 *
1230 * . if a queue Q that does not belong to the burst is created a while
1231 *   later, i.e., not shortly after, than the last time at which a queue
1232 *   either entered the burst list or was marked as belonging to the
1233 *   current large burst, then the current burst is deemed as finished and:
1234 *
1235 *        . the large-burst mode is reset if set
1236 *
1237 *        . the burst list is emptied
1238 *
1239 *        . Q is inserted in the burst list, as Q may be the first queue
1240 *          in a possible new burst (then the burst list contains just Q
1241 *          after this step).
1242 */
1243static void bfq_handle_burst(struct bfq_data *bfqd, struct bfq_queue *bfqq)
1244{
1245	/*
1246	 * If bfqq is already in the burst list or is part of a large
1247	 * burst, or finally has just been split, then there is
1248	 * nothing else to do.
1249	 */
1250	if (!hlist_unhashed(&bfqq->burst_list_node) ||
1251	    bfq_bfqq_in_large_burst(bfqq) ||
1252	    time_is_after_eq_jiffies(bfqq->split_time +
1253				     msecs_to_jiffies(10)))
1254		return;
1255
1256	/*
1257	 * If bfqq's creation happens late enough, or bfqq belongs to
1258	 * a different group than the burst group, then the current
1259	 * burst is finished, and related data structures must be
1260	 * reset.
1261	 *
1262	 * In this respect, consider the special case where bfqq is
1263	 * the very first queue created after BFQ is selected for this
1264	 * device. In this case, last_ins_in_burst and
1265	 * burst_parent_entity are not yet significant when we get
1266	 * here. But it is easy to verify that, whether or not the
1267	 * following condition is true, bfqq will end up being
1268	 * inserted into the burst list. In particular the list will
1269	 * happen to contain only bfqq. And this is exactly what has
1270	 * to happen, as bfqq may be the first queue of the first
1271	 * burst.
1272	 */
1273	if (time_is_before_jiffies(bfqd->last_ins_in_burst +
1274	    bfqd->bfq_burst_interval) ||
1275	    bfqq->entity.parent != bfqd->burst_parent_entity) {
1276		bfqd->large_burst = false;
1277		bfq_reset_burst_list(bfqd, bfqq);
1278		goto end;
1279	}
1280
1281	/*
1282	 * If we get here, then bfqq is being activated shortly after the
1283	 * last queue. So, if the current burst is also large, we can mark
1284	 * bfqq as belonging to this large burst immediately.
1285	 */
1286	if (bfqd->large_burst) {
1287		bfq_mark_bfqq_in_large_burst(bfqq);
1288		goto end;
1289	}
1290
1291	/*
1292	 * If we get here, then a large-burst state has not yet been
1293	 * reached, but bfqq is being activated shortly after the last
1294	 * queue. Then we add bfqq to the burst.
1295	 */
1296	bfq_add_to_burst(bfqd, bfqq);
1297end:
1298	/*
1299	 * At this point, bfqq either has been added to the current
1300	 * burst or has caused the current burst to terminate and a
1301	 * possible new burst to start. In particular, in the second
1302	 * case, bfqq has become the first queue in the possible new
1303	 * burst.  In both cases last_ins_in_burst needs to be moved
1304	 * forward.
1305	 */
1306	bfqd->last_ins_in_burst = jiffies;
1307}
1308
1309static int bfq_bfqq_budget_left(struct bfq_queue *bfqq)
1310{
1311	struct bfq_entity *entity = &bfqq->entity;
1312
1313	return entity->budget - entity->service;
1314}
1315
1316/*
1317 * If enough samples have been computed, return the current max budget
1318 * stored in bfqd, which is dynamically updated according to the
1319 * estimated disk peak rate; otherwise return the default max budget
1320 */
1321static int bfq_max_budget(struct bfq_data *bfqd)
1322{
1323	if (bfqd->budgets_assigned < bfq_stats_min_budgets)
1324		return bfq_default_max_budget;
1325	else
1326		return bfqd->bfq_max_budget;
1327}
1328
1329/*
1330 * Return min budget, which is a fraction of the current or default
1331 * max budget (trying with 1/32)
1332 */
1333static int bfq_min_budget(struct bfq_data *bfqd)
1334{
1335	if (bfqd->budgets_assigned < bfq_stats_min_budgets)
1336		return bfq_default_max_budget / 32;
1337	else
1338		return bfqd->bfq_max_budget / 32;
1339}
1340
1341/*
1342 * The next function, invoked after the input queue bfqq switches from
1343 * idle to busy, updates the budget of bfqq. The function also tells
1344 * whether the in-service queue should be expired, by returning
1345 * true. The purpose of expiring the in-service queue is to give bfqq
1346 * the chance to possibly preempt the in-service queue, and the reason
1347 * for preempting the in-service queue is to achieve one of the two
1348 * goals below.
1349 *
1350 * 1. Guarantee to bfqq its reserved bandwidth even if bfqq has
1351 * expired because it has remained idle. In particular, bfqq may have
1352 * expired for one of the following two reasons:
1353 *
1354 * - BFQQE_NO_MORE_REQUESTS bfqq did not enjoy any device idling
1355 *   and did not make it to issue a new request before its last
1356 *   request was served;
1357 *
1358 * - BFQQE_TOO_IDLE bfqq did enjoy device idling, but did not issue
1359 *   a new request before the expiration of the idling-time.
1360 *
1361 * Even if bfqq has expired for one of the above reasons, the process
1362 * associated with the queue may be however issuing requests greedily,
1363 * and thus be sensitive to the bandwidth it receives (bfqq may have
1364 * remained idle for other reasons: CPU high load, bfqq not enjoying
1365 * idling, I/O throttling somewhere in the path from the process to
1366 * the I/O scheduler, ...). But if, after every expiration for one of
1367 * the above two reasons, bfqq has to wait for the service of at least
1368 * one full budget of another queue before being served again, then
1369 * bfqq is likely to get a much lower bandwidth or resource time than
1370 * its reserved ones. To address this issue, two countermeasures need
1371 * to be taken.
1372 *
1373 * First, the budget and the timestamps of bfqq need to be updated in
1374 * a special way on bfqq reactivation: they need to be updated as if
1375 * bfqq did not remain idle and did not expire. In fact, if they are
1376 * computed as if bfqq expired and remained idle until reactivation,
1377 * then the process associated with bfqq is treated as if, instead of
1378 * being greedy, it stopped issuing requests when bfqq remained idle,
1379 * and restarts issuing requests only on this reactivation. In other
1380 * words, the scheduler does not help the process recover the "service
1381 * hole" between bfqq expiration and reactivation. As a consequence,
1382 * the process receives a lower bandwidth than its reserved one. In
1383 * contrast, to recover this hole, the budget must be updated as if
1384 * bfqq was not expired at all before this reactivation, i.e., it must
1385 * be set to the value of the remaining budget when bfqq was
1386 * expired. Along the same line, timestamps need to be assigned the
1387 * value they had the last time bfqq was selected for service, i.e.,
1388 * before last expiration. Thus timestamps need to be back-shifted
1389 * with respect to their normal computation (see [1] for more details
1390 * on this tricky aspect).
1391 *
1392 * Secondly, to allow the process to recover the hole, the in-service
1393 * queue must be expired too, to give bfqq the chance to preempt it
1394 * immediately. In fact, if bfqq has to wait for a full budget of the
1395 * in-service queue to be completed, then it may become impossible to
1396 * let the process recover the hole, even if the back-shifted
1397 * timestamps of bfqq are lower than those of the in-service queue. If
1398 * this happens for most or all of the holes, then the process may not
1399 * receive its reserved bandwidth. In this respect, it is worth noting
1400 * that, being the service of outstanding requests unpreemptible, a
1401 * little fraction of the holes may however be unrecoverable, thereby
1402 * causing a little loss of bandwidth.
1403 *
1404 * The last important point is detecting whether bfqq does need this
1405 * bandwidth recovery. In this respect, the next function deems the
1406 * process associated with bfqq greedy, and thus allows it to recover
1407 * the hole, if: 1) the process is waiting for the arrival of a new
1408 * request (which implies that bfqq expired for one of the above two
1409 * reasons), and 2) such a request has arrived soon. The first
1410 * condition is controlled through the flag non_blocking_wait_rq,
1411 * while the second through the flag arrived_in_time. If both
1412 * conditions hold, then the function computes the budget in the
1413 * above-described special way, and signals that the in-service queue
1414 * should be expired. Timestamp back-shifting is done later in
1415 * __bfq_activate_entity.
1416 *
1417 * 2. Reduce latency. Even if timestamps are not backshifted to let
1418 * the process associated with bfqq recover a service hole, bfqq may
1419 * however happen to have, after being (re)activated, a lower finish
1420 * timestamp than the in-service queue.	 That is, the next budget of
1421 * bfqq may have to be completed before the one of the in-service
1422 * queue. If this is the case, then preempting the in-service queue
1423 * allows this goal to be achieved, apart from the unpreemptible,
1424 * outstanding requests mentioned above.
1425 *
1426 * Unfortunately, regardless of which of the above two goals one wants
1427 * to achieve, service trees need first to be updated to know whether
1428 * the in-service queue must be preempted. To have service trees
1429 * correctly updated, the in-service queue must be expired and
1430 * rescheduled, and bfqq must be scheduled too. This is one of the
1431 * most costly operations (in future versions, the scheduling
1432 * mechanism may be re-designed in such a way to make it possible to
1433 * know whether preemption is needed without needing to update service
1434 * trees). In addition, queue preemptions almost always cause random
1435 * I/O, which may in turn cause loss of throughput. Finally, there may
1436 * even be no in-service queue when the next function is invoked (so,
1437 * no queue to compare timestamps with). Because of these facts, the
1438 * next function adopts the following simple scheme to avoid costly
1439 * operations, too frequent preemptions and too many dependencies on
1440 * the state of the scheduler: it requests the expiration of the
1441 * in-service queue (unconditionally) only for queues that need to
1442 * recover a hole. Then it delegates to other parts of the code the
1443 * responsibility of handling the above case 2.
1444 */
1445static bool bfq_bfqq_update_budg_for_activation(struct bfq_data *bfqd,
1446						struct bfq_queue *bfqq,
1447						bool arrived_in_time)
1448{
1449	struct bfq_entity *entity = &bfqq->entity;
1450
1451	/*
1452	 * In the next compound condition, we check also whether there
1453	 * is some budget left, because otherwise there is no point in
1454	 * trying to go on serving bfqq with this same budget: bfqq
1455	 * would be expired immediately after being selected for
1456	 * service. This would only cause useless overhead.
1457	 */
1458	if (bfq_bfqq_non_blocking_wait_rq(bfqq) && arrived_in_time &&
1459	    bfq_bfqq_budget_left(bfqq) > 0) {
1460		/*
1461		 * We do not clear the flag non_blocking_wait_rq here, as
1462		 * the latter is used in bfq_activate_bfqq to signal
1463		 * that timestamps need to be back-shifted (and is
1464		 * cleared right after).
1465		 */
1466
1467		/*
1468		 * In next assignment we rely on that either
1469		 * entity->service or entity->budget are not updated
1470		 * on expiration if bfqq is empty (see
1471		 * __bfq_bfqq_recalc_budget). Thus both quantities
1472		 * remain unchanged after such an expiration, and the
1473		 * following statement therefore assigns to
1474		 * entity->budget the remaining budget on such an
1475		 * expiration.
1476		 */
1477		entity->budget = min_t(unsigned long,
1478				       bfq_bfqq_budget_left(bfqq),
1479				       bfqq->max_budget);
1480
1481		/*
1482		 * At this point, we have used entity->service to get
1483		 * the budget left (needed for updating
1484		 * entity->budget). Thus we finally can, and have to,
1485		 * reset entity->service. The latter must be reset
1486		 * because bfqq would otherwise be charged again for
1487		 * the service it has received during its previous
1488		 * service slot(s).
1489		 */
1490		entity->service = 0;
1491
1492		return true;
1493	}
1494
1495	/*
1496	 * We can finally complete expiration, by setting service to 0.
1497	 */
1498	entity->service = 0;
1499	entity->budget = max_t(unsigned long, bfqq->max_budget,
1500			       bfq_serv_to_charge(bfqq->next_rq, bfqq));
1501	bfq_clear_bfqq_non_blocking_wait_rq(bfqq);
1502	return false;
1503}
1504
1505/*
1506 * Return the farthest past time instant according to jiffies
1507 * macros.
1508 */
1509static unsigned long bfq_smallest_from_now(void)
1510{
1511	return jiffies - MAX_JIFFY_OFFSET;
1512}
1513
1514static void bfq_update_bfqq_wr_on_rq_arrival(struct bfq_data *bfqd,
1515					     struct bfq_queue *bfqq,
1516					     unsigned int old_wr_coeff,
1517					     bool wr_or_deserves_wr,
1518					     bool interactive,
1519					     bool in_burst,
1520					     bool soft_rt)
1521{
1522	if (old_wr_coeff == 1 && wr_or_deserves_wr) {
1523		/* start a weight-raising period */
1524		if (interactive) {
1525			bfqq->service_from_wr = 0;
1526			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1527			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1528		} else {
1529			/*
1530			 * No interactive weight raising in progress
1531			 * here: assign minus infinity to
1532			 * wr_start_at_switch_to_srt, to make sure
1533			 * that, at the end of the soft-real-time
1534			 * weight raising periods that is starting
1535			 * now, no interactive weight-raising period
1536			 * may be wrongly considered as still in
1537			 * progress (and thus actually started by
1538			 * mistake).
1539			 */
1540			bfqq->wr_start_at_switch_to_srt =
1541				bfq_smallest_from_now();
1542			bfqq->wr_coeff = bfqd->bfq_wr_coeff *
1543				BFQ_SOFTRT_WEIGHT_FACTOR;
1544			bfqq->wr_cur_max_time =
1545				bfqd->bfq_wr_rt_max_time;
1546		}
1547
1548		/*
1549		 * If needed, further reduce budget to make sure it is
1550		 * close to bfqq's backlog, so as to reduce the
1551		 * scheduling-error component due to a too large
1552		 * budget. Do not care about throughput consequences,
1553		 * but only about latency. Finally, do not assign a
1554		 * too small budget either, to avoid increasing
1555		 * latency by causing too frequent expirations.
1556		 */
1557		bfqq->entity.budget = min_t(unsigned long,
1558					    bfqq->entity.budget,
1559					    2 * bfq_min_budget(bfqd));
1560	} else if (old_wr_coeff > 1) {
1561		if (interactive) { /* update wr coeff and duration */
1562			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
1563			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
1564		} else if (in_burst)
1565			bfqq->wr_coeff = 1;
1566		else if (soft_rt) {
1567			/*
1568			 * The application is now or still meeting the
1569			 * requirements for being deemed soft rt.  We
1570			 * can then correctly and safely (re)charge
1571			 * the weight-raising duration for the
1572			 * application with the weight-raising
1573			 * duration for soft rt applications.
1574			 *
1575			 * In particular, doing this recharge now, i.e.,
1576			 * before the weight-raising period for the
1577			 * application finishes, reduces the probability
1578			 * of the following negative scenario:
1579			 * 1) the weight of a soft rt application is
1580			 *    raised at startup (as for any newly
1581			 *    created application),
1582			 * 2) since the application is not interactive,
1583			 *    at a certain time weight-raising is
1584			 *    stopped for the application,
1585			 * 3) at that time the application happens to
1586			 *    still have pending requests, and hence
1587			 *    is destined to not have a chance to be
1588			 *    deemed soft rt before these requests are
1589			 *    completed (see the comments to the
1590			 *    function bfq_bfqq_softrt_next_start()
1591			 *    for details on soft rt detection),
1592			 * 4) these pending requests experience a high
1593			 *    latency because the application is not
1594			 *    weight-raised while they are pending.
1595			 */
1596			if (bfqq->wr_cur_max_time !=
1597				bfqd->bfq_wr_rt_max_time) {
1598				bfqq->wr_start_at_switch_to_srt =
1599					bfqq->last_wr_start_finish;
1600
1601				bfqq->wr_cur_max_time =
1602					bfqd->bfq_wr_rt_max_time;
1603				bfqq->wr_coeff = bfqd->bfq_wr_coeff *
1604					BFQ_SOFTRT_WEIGHT_FACTOR;
1605			}
1606			bfqq->last_wr_start_finish = jiffies;
1607		}
1608	}
1609}
1610
1611static bool bfq_bfqq_idle_for_long_time(struct bfq_data *bfqd,
1612					struct bfq_queue *bfqq)
1613{
1614	return bfqq->dispatched == 0 &&
1615		time_is_before_jiffies(
1616			bfqq->budget_timeout +
1617			bfqd->bfq_wr_min_idle_time);
1618}
1619
1620
1621/*
1622 * Return true if bfqq is in a higher priority class, or has a higher
1623 * weight than the in-service queue.
1624 */
1625static bool bfq_bfqq_higher_class_or_weight(struct bfq_queue *bfqq,
1626					    struct bfq_queue *in_serv_bfqq)
1627{
1628	int bfqq_weight, in_serv_weight;
1629
1630	if (bfqq->ioprio_class < in_serv_bfqq->ioprio_class)
1631		return true;
1632
1633	if (in_serv_bfqq->entity.parent == bfqq->entity.parent) {
1634		bfqq_weight = bfqq->entity.weight;
1635		in_serv_weight = in_serv_bfqq->entity.weight;
1636	} else {
1637		if (bfqq->entity.parent)
1638			bfqq_weight = bfqq->entity.parent->weight;
1639		else
1640			bfqq_weight = bfqq->entity.weight;
1641		if (in_serv_bfqq->entity.parent)
1642			in_serv_weight = in_serv_bfqq->entity.parent->weight;
1643		else
1644			in_serv_weight = in_serv_bfqq->entity.weight;
1645	}
1646
1647	return bfqq_weight > in_serv_weight;
1648}
1649
 
 
1650static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd,
1651					     struct bfq_queue *bfqq,
1652					     int old_wr_coeff,
1653					     struct request *rq,
1654					     bool *interactive)
1655{
1656	bool soft_rt, in_burst,	wr_or_deserves_wr,
1657		bfqq_wants_to_preempt,
1658		idle_for_long_time = bfq_bfqq_idle_for_long_time(bfqd, bfqq),
1659		/*
1660		 * See the comments on
1661		 * bfq_bfqq_update_budg_for_activation for
1662		 * details on the usage of the next variable.
1663		 */
1664		arrived_in_time =  ktime_get_ns() <=
1665			bfqq->ttime.last_end_request +
1666			bfqd->bfq_slice_idle * 3;
1667
1668
1669	/*
1670	 * bfqq deserves to be weight-raised if:
1671	 * - it is sync,
1672	 * - it does not belong to a large burst,
1673	 * - it has been idle for enough time or is soft real-time,
1674	 * - is linked to a bfq_io_cq (it is not shared in any sense).
 
 
1675	 */
1676	in_burst = bfq_bfqq_in_large_burst(bfqq);
1677	soft_rt = bfqd->bfq_wr_max_softrt_rate > 0 &&
1678		!BFQQ_TOTALLY_SEEKY(bfqq) &&
1679		!in_burst &&
1680		time_is_before_jiffies(bfqq->soft_rt_next_start) &&
1681		bfqq->dispatched == 0;
1682	*interactive = !in_burst && idle_for_long_time;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1683	wr_or_deserves_wr = bfqd->low_latency &&
1684		(bfqq->wr_coeff > 1 ||
1685		 (bfq_bfqq_sync(bfqq) &&
1686		  bfqq->bic && (*interactive || soft_rt)));
 
1687
1688	/*
1689	 * Using the last flag, update budget and check whether bfqq
1690	 * may want to preempt the in-service queue.
1691	 */
1692	bfqq_wants_to_preempt =
1693		bfq_bfqq_update_budg_for_activation(bfqd, bfqq,
1694						    arrived_in_time);
1695
1696	/*
1697	 * If bfqq happened to be activated in a burst, but has been
1698	 * idle for much more than an interactive queue, then we
1699	 * assume that, in the overall I/O initiated in the burst, the
1700	 * I/O associated with bfqq is finished. So bfqq does not need
1701	 * to be treated as a queue belonging to a burst
1702	 * anymore. Accordingly, we reset bfqq's in_large_burst flag
1703	 * if set, and remove bfqq from the burst list if it's
1704	 * there. We do not decrement burst_size, because the fact
1705	 * that bfqq does not need to belong to the burst list any
1706	 * more does not invalidate the fact that bfqq was created in
1707	 * a burst.
1708	 */
1709	if (likely(!bfq_bfqq_just_created(bfqq)) &&
1710	    idle_for_long_time &&
1711	    time_is_before_jiffies(
1712		    bfqq->budget_timeout +
1713		    msecs_to_jiffies(10000))) {
1714		hlist_del_init(&bfqq->burst_list_node);
1715		bfq_clear_bfqq_in_large_burst(bfqq);
1716	}
1717
1718	bfq_clear_bfqq_just_created(bfqq);
1719
1720
1721	if (!bfq_bfqq_IO_bound(bfqq)) {
1722		if (arrived_in_time) {
1723			bfqq->requests_within_timer++;
1724			if (bfqq->requests_within_timer >=
1725			    bfqd->bfq_requests_within_timer)
1726				bfq_mark_bfqq_IO_bound(bfqq);
1727		} else
1728			bfqq->requests_within_timer = 0;
1729	}
1730
1731	if (bfqd->low_latency) {
1732		if (unlikely(time_is_after_jiffies(bfqq->split_time)))
1733			/* wraparound */
1734			bfqq->split_time =
1735				jiffies - bfqd->bfq_wr_min_idle_time - 1;
1736
1737		if (time_is_before_jiffies(bfqq->split_time +
1738					   bfqd->bfq_wr_min_idle_time)) {
1739			bfq_update_bfqq_wr_on_rq_arrival(bfqd, bfqq,
1740							 old_wr_coeff,
1741							 wr_or_deserves_wr,
1742							 *interactive,
1743							 in_burst,
1744							 soft_rt);
1745
1746			if (old_wr_coeff != bfqq->wr_coeff)
1747				bfqq->entity.prio_changed = 1;
1748		}
1749	}
1750
1751	bfqq->last_idle_bklogged = jiffies;
1752	bfqq->service_from_backlogged = 0;
1753	bfq_clear_bfqq_softrt_update(bfqq);
1754
1755	bfq_add_bfqq_busy(bfqd, bfqq);
1756
1757	/*
1758	 * Expire in-service queue only if preemption may be needed
1759	 * for guarantees. In particular, we care only about two
1760	 * cases. The first is that bfqq has to recover a service
1761	 * hole, as explained in the comments on
1762	 * bfq_bfqq_update_budg_for_activation(), i.e., that
1763	 * bfqq_wants_to_preempt is true. However, if bfqq does not
1764	 * carry time-critical I/O, then bfqq's bandwidth is less
1765	 * important than that of queues that carry time-critical I/O.
1766	 * So, as a further constraint, we consider this case only if
1767	 * bfqq is at least as weight-raised, i.e., at least as time
1768	 * critical, as the in-service queue.
1769	 *
1770	 * The second case is that bfqq is in a higher priority class,
1771	 * or has a higher weight than the in-service queue. If this
1772	 * condition does not hold, we don't care because, even if
1773	 * bfqq does not start to be served immediately, the resulting
1774	 * delay for bfqq's I/O is however lower or much lower than
1775	 * the ideal completion time to be guaranteed to bfqq's I/O.
1776	 *
1777	 * In both cases, preemption is needed only if, according to
1778	 * the timestamps of both bfqq and of the in-service queue,
1779	 * bfqq actually is the next queue to serve. So, to reduce
1780	 * useless preemptions, the return value of
1781	 * next_queue_may_preempt() is considered in the next compound
1782	 * condition too. Yet next_queue_may_preempt() just checks a
1783	 * simple, necessary condition for bfqq to be the next queue
1784	 * to serve. In fact, to evaluate a sufficient condition, the
1785	 * timestamps of the in-service queue would need to be
1786	 * updated, and this operation is quite costly (see the
1787	 * comments on bfq_bfqq_update_budg_for_activation()).
 
 
 
 
 
 
 
 
 
 
 
1788	 */
1789	if (bfqd->in_service_queue &&
1790	    ((bfqq_wants_to_preempt &&
1791	      bfqq->wr_coeff >= bfqd->in_service_queue->wr_coeff) ||
1792	     bfq_bfqq_higher_class_or_weight(bfqq, bfqd->in_service_queue)) &&
 
1793	    next_queue_may_preempt(bfqd))
1794		bfq_bfqq_expire(bfqd, bfqd->in_service_queue,
1795				false, BFQQE_PREEMPTED);
1796}
1797
1798static void bfq_reset_inject_limit(struct bfq_data *bfqd,
1799				   struct bfq_queue *bfqq)
1800{
1801	/* invalidate baseline total service time */
1802	bfqq->last_serv_time_ns = 0;
1803
1804	/*
1805	 * Reset pointer in case we are waiting for
1806	 * some request completion.
1807	 */
1808	bfqd->waited_rq = NULL;
1809
1810	/*
1811	 * If bfqq has a short think time, then start by setting the
1812	 * inject limit to 0 prudentially, because the service time of
1813	 * an injected I/O request may be higher than the think time
1814	 * of bfqq, and therefore, if one request was injected when
1815	 * bfqq remains empty, this injected request might delay the
1816	 * service of the next I/O request for bfqq significantly. In
1817	 * case bfqq can actually tolerate some injection, then the
1818	 * adaptive update will however raise the limit soon. This
1819	 * lucky circumstance holds exactly because bfqq has a short
1820	 * think time, and thus, after remaining empty, is likely to
1821	 * get new I/O enqueued---and then completed---before being
1822	 * expired. This is the very pattern that gives the
1823	 * limit-update algorithm the chance to measure the effect of
1824	 * injection on request service times, and then to update the
1825	 * limit accordingly.
1826	 *
1827	 * However, in the following special case, the inject limit is
1828	 * left to 1 even if the think time is short: bfqq's I/O is
1829	 * synchronized with that of some other queue, i.e., bfqq may
1830	 * receive new I/O only after the I/O of the other queue is
1831	 * completed. Keeping the inject limit to 1 allows the
1832	 * blocking I/O to be served while bfqq is in service. And
1833	 * this is very convenient both for bfqq and for overall
1834	 * throughput, as explained in detail in the comments in
1835	 * bfq_update_has_short_ttime().
1836	 *
1837	 * On the opposite end, if bfqq has a long think time, then
1838	 * start directly by 1, because:
1839	 * a) on the bright side, keeping at most one request in
1840	 * service in the drive is unlikely to cause any harm to the
1841	 * latency of bfqq's requests, as the service time of a single
1842	 * request is likely to be lower than the think time of bfqq;
1843	 * b) on the downside, after becoming empty, bfqq is likely to
1844	 * expire before getting its next request. With this request
1845	 * arrival pattern, it is very hard to sample total service
1846	 * times and update the inject limit accordingly (see comments
1847	 * on bfq_update_inject_limit()). So the limit is likely to be
1848	 * never, or at least seldom, updated.  As a consequence, by
1849	 * setting the limit to 1, we avoid that no injection ever
1850	 * occurs with bfqq. On the downside, this proactive step
1851	 * further reduces chances to actually compute the baseline
1852	 * total service time. Thus it reduces chances to execute the
1853	 * limit-update algorithm and possibly raise the limit to more
1854	 * than 1.
1855	 */
1856	if (bfq_bfqq_has_short_ttime(bfqq))
1857		bfqq->inject_limit = 0;
1858	else
1859		bfqq->inject_limit = 1;
1860
1861	bfqq->decrease_time_jif = jiffies;
1862}
1863
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1864static void bfq_add_request(struct request *rq)
1865{
1866	struct bfq_queue *bfqq = RQ_BFQQ(rq);
1867	struct bfq_data *bfqd = bfqq->bfqd;
1868	struct request *next_rq, *prev;
1869	unsigned int old_wr_coeff = bfqq->wr_coeff;
1870	bool interactive = false;
 
1871
1872	bfq_log_bfqq(bfqd, bfqq, "add_request %d", rq_is_sync(rq));
1873	bfqq->queued[rq_is_sync(rq)]++;
1874	bfqd->queued++;
1875
1876	if (RB_EMPTY_ROOT(&bfqq->sort_list) && bfq_bfqq_sync(bfqq)) {
1877		/*
1878		 * Detect whether bfqq's I/O seems synchronized with
1879		 * that of some other queue, i.e., whether bfqq, after
1880		 * remaining empty, happens to receive new I/O only
1881		 * right after some I/O request of the other queue has
1882		 * been completed. We call waker queue the other
1883		 * queue, and we assume, for simplicity, that bfqq may
1884		 * have at most one waker queue.
1885		 *
1886		 * A remarkable throughput boost can be reached by
1887		 * unconditionally injecting the I/O of the waker
1888		 * queue, every time a new bfq_dispatch_request
1889		 * happens to be invoked while I/O is being plugged
1890		 * for bfqq.  In addition to boosting throughput, this
1891		 * unblocks bfqq's I/O, thereby improving bandwidth
1892		 * and latency for bfqq. Note that these same results
1893		 * may be achieved with the general injection
1894		 * mechanism, but less effectively. For details on
1895		 * this aspect, see the comments on the choice of the
1896		 * queue for injection in bfq_select_queue().
1897		 *
1898		 * Turning back to the detection of a waker queue, a
1899		 * queue Q is deemed as a waker queue for bfqq if, for
1900		 * two consecutive times, bfqq happens to become non
1901		 * empty right after a request of Q has been
1902		 * completed. In particular, on the first time, Q is
1903		 * tentatively set as a candidate waker queue, while
1904		 * on the second time, the flag
1905		 * bfq_bfqq_has_waker(bfqq) is set to confirm that Q
1906		 * is a waker queue for bfqq. These detection steps
1907		 * are performed only if bfqq has a long think time,
1908		 * so as to make it more likely that bfqq's I/O is
1909		 * actually being blocked by a synchronization. This
1910		 * last filter, plus the above two-times requirement,
1911		 * make false positives less likely.
1912		 *
1913		 * NOTE
1914		 *
1915		 * The sooner a waker queue is detected, the sooner
1916		 * throughput can be boosted by injecting I/O from the
1917		 * waker queue. Fortunately, detection is likely to be
1918		 * actually fast, for the following reasons. While
1919		 * blocked by synchronization, bfqq has a long think
1920		 * time. This implies that bfqq's inject limit is at
1921		 * least equal to 1 (see the comments in
1922		 * bfq_update_inject_limit()). So, thanks to
1923		 * injection, the waker queue is likely to be served
1924		 * during the very first I/O-plugging time interval
1925		 * for bfqq. This triggers the first step of the
1926		 * detection mechanism. Thanks again to injection, the
1927		 * candidate waker queue is then likely to be
1928		 * confirmed no later than during the next
1929		 * I/O-plugging interval for bfqq.
1930		 */
1931		if (bfqd->last_completed_rq_bfqq &&
1932		    !bfq_bfqq_has_short_ttime(bfqq) &&
1933		    ktime_get_ns() - bfqd->last_completion <
1934		    200 * NSEC_PER_USEC) {
1935			if (bfqd->last_completed_rq_bfqq != bfqq &&
1936			    bfqd->last_completed_rq_bfqq !=
1937			    bfqq->waker_bfqq) {
1938				/*
1939				 * First synchronization detected with
1940				 * a candidate waker queue, or with a
1941				 * different candidate waker queue
1942				 * from the current one.
1943				 */
1944				bfqq->waker_bfqq = bfqd->last_completed_rq_bfqq;
1945
1946				/*
1947				 * If the waker queue disappears, then
1948				 * bfqq->waker_bfqq must be reset. To
1949				 * this goal, we maintain in each
1950				 * waker queue a list, woken_list, of
1951				 * all the queues that reference the
1952				 * waker queue through their
1953				 * waker_bfqq pointer. When the waker
1954				 * queue exits, the waker_bfqq pointer
1955				 * of all the queues in the woken_list
1956				 * is reset.
1957				 *
1958				 * In addition, if bfqq is already in
1959				 * the woken_list of a waker queue,
1960				 * then, before being inserted into
1961				 * the woken_list of a new waker
1962				 * queue, bfqq must be removed from
1963				 * the woken_list of the old waker
1964				 * queue.
1965				 */
1966				if (!hlist_unhashed(&bfqq->woken_list_node))
1967					hlist_del_init(&bfqq->woken_list_node);
1968				hlist_add_head(&bfqq->woken_list_node,
1969				    &bfqd->last_completed_rq_bfqq->woken_list);
1970
1971				bfq_clear_bfqq_has_waker(bfqq);
1972			} else if (bfqd->last_completed_rq_bfqq ==
1973				   bfqq->waker_bfqq &&
1974				   !bfq_bfqq_has_waker(bfqq)) {
1975				/*
1976				 * synchronization with waker_bfqq
1977				 * seen for the second time
1978				 */
1979				bfq_mark_bfqq_has_waker(bfqq);
1980			}
1981		}
1982
1983		/*
1984		 * Periodically reset inject limit, to make sure that
1985		 * the latter eventually drops in case workload
1986		 * changes, see step (3) in the comments on
1987		 * bfq_update_inject_limit().
1988		 */
1989		if (time_is_before_eq_jiffies(bfqq->decrease_time_jif +
1990					     msecs_to_jiffies(1000)))
1991			bfq_reset_inject_limit(bfqd, bfqq);
1992
1993		/*
1994		 * The following conditions must hold to setup a new
1995		 * sampling of total service time, and then a new
1996		 * update of the inject limit:
1997		 * - bfqq is in service, because the total service
1998		 *   time is evaluated only for the I/O requests of
1999		 *   the queues in service;
2000		 * - this is the right occasion to compute or to
2001		 *   lower the baseline total service time, because
2002		 *   there are actually no requests in the drive,
2003		 *   or
2004		 *   the baseline total service time is available, and
2005		 *   this is the right occasion to compute the other
2006		 *   quantity needed to update the inject limit, i.e.,
2007		 *   the total service time caused by the amount of
2008		 *   injection allowed by the current value of the
2009		 *   limit. It is the right occasion because injection
2010		 *   has actually been performed during the service
2011		 *   hole, and there are still in-flight requests,
2012		 *   which are very likely to be exactly the injected
2013		 *   requests, or part of them;
2014		 * - the minimum interval for sampling the total
2015		 *   service time and updating the inject limit has
2016		 *   elapsed.
2017		 */
2018		if (bfqq == bfqd->in_service_queue &&
2019		    (bfqd->rq_in_driver == 0 ||
2020		     (bfqq->last_serv_time_ns > 0 &&
2021		      bfqd->rqs_injected && bfqd->rq_in_driver > 0)) &&
2022		    time_is_before_eq_jiffies(bfqq->decrease_time_jif +
2023					      msecs_to_jiffies(10))) {
2024			bfqd->last_empty_occupied_ns = ktime_get_ns();
2025			/*
2026			 * Start the state machine for measuring the
2027			 * total service time of rq: setting
2028			 * wait_dispatch will cause bfqd->waited_rq to
2029			 * be set when rq will be dispatched.
2030			 */
2031			bfqd->wait_dispatch = true;
2032			/*
2033			 * If there is no I/O in service in the drive,
2034			 * then possible injection occurred before the
2035			 * arrival of rq will not affect the total
2036			 * service time of rq. So the injection limit
2037			 * must not be updated as a function of such
2038			 * total service time, unless new injection
2039			 * occurs before rq is completed. To have the
2040			 * injection limit updated only in the latter
2041			 * case, reset rqs_injected here (rqs_injected
2042			 * will be set in case injection is performed
2043			 * on bfqq before rq is completed).
2044			 */
2045			if (bfqd->rq_in_driver == 0)
2046				bfqd->rqs_injected = false;
2047		}
2048	}
2049
 
 
 
2050	elv_rb_add(&bfqq->sort_list, rq);
2051
2052	/*
2053	 * Check if this request is a better next-serve candidate.
2054	 */
2055	prev = bfqq->next_rq;
2056	next_rq = bfq_choose_req(bfqd, bfqq->next_rq, rq, bfqd->last_position);
2057	bfqq->next_rq = next_rq;
2058
2059	/*
2060	 * Adjust priority tree position, if next_rq changes.
2061	 * See comments on bfq_pos_tree_add_move() for the unlikely().
2062	 */
2063	if (unlikely(!bfqd->nonrot_with_queueing && prev != bfqq->next_rq))
2064		bfq_pos_tree_add_move(bfqd, bfqq);
2065
2066	if (!bfq_bfqq_busy(bfqq)) /* switching to busy ... */
2067		bfq_bfqq_handle_idle_busy_switch(bfqd, bfqq, old_wr_coeff,
2068						 rq, &interactive);
2069	else {
2070		if (bfqd->low_latency && old_wr_coeff == 1 && !rq_is_sync(rq) &&
2071		    time_is_before_jiffies(
2072				bfqq->last_wr_start_finish +
2073				bfqd->bfq_wr_min_inter_arr_async)) {
2074			bfqq->wr_coeff = bfqd->bfq_wr_coeff;
2075			bfqq->wr_cur_max_time = bfq_wr_duration(bfqd);
2076
2077			bfqd->wr_busy_queues++;
2078			bfqq->entity.prio_changed = 1;
2079		}
2080		if (prev != bfqq->next_rq)
2081			bfq_updated_next_req(bfqd, bfqq);
2082	}
2083
2084	/*
2085	 * Assign jiffies to last_wr_start_finish in the following
2086	 * cases:
2087	 *
2088	 * . if bfqq is not going to be weight-raised, because, for
2089	 *   non weight-raised queues, last_wr_start_finish stores the
2090	 *   arrival time of the last request; as of now, this piece
2091	 *   of information is used only for deciding whether to
2092	 *   weight-raise async queues
2093	 *
2094	 * . if bfqq is not weight-raised, because, if bfqq is now
2095	 *   switching to weight-raised, then last_wr_start_finish
2096	 *   stores the time when weight-raising starts
2097	 *
2098	 * . if bfqq is interactive, because, regardless of whether
2099	 *   bfqq is currently weight-raised, the weight-raising
2100	 *   period must start or restart (this case is considered
2101	 *   separately because it is not detected by the above
2102	 *   conditions, if bfqq is already weight-raised)
2103	 *
2104	 * last_wr_start_finish has to be updated also if bfqq is soft
2105	 * real-time, because the weight-raising period is constantly
2106	 * restarted on idle-to-busy transitions for these queues, but
2107	 * this is already done in bfq_bfqq_handle_idle_busy_switch if
2108	 * needed.
2109	 */
2110	if (bfqd->low_latency &&
2111		(old_wr_coeff == 1 || bfqq->wr_coeff == 1 || interactive))
2112		bfqq->last_wr_start_finish = jiffies;
2113}
2114
2115static struct request *bfq_find_rq_fmerge(struct bfq_data *bfqd,
2116					  struct bio *bio,
2117					  struct request_queue *q)
2118{
2119	struct bfq_queue *bfqq = bfqd->bio_bfqq;
2120
2121
2122	if (bfqq)
2123		return elv_rb_find(&bfqq->sort_list, bio_end_sector(bio));
2124
2125	return NULL;
2126}
2127
2128static sector_t get_sdist(sector_t last_pos, struct request *rq)
2129{
2130	if (last_pos)
2131		return abs(blk_rq_pos(rq) - last_pos);
2132
2133	return 0;
2134}
2135
2136#if 0 /* Still not clear if we can do without next two functions */
2137static void bfq_activate_request(struct request_queue *q, struct request *rq)
2138{
2139	struct bfq_data *bfqd = q->elevator->elevator_data;
2140
2141	bfqd->rq_in_driver++;
2142}
2143
2144static void bfq_deactivate_request(struct request_queue *q, struct request *rq)
2145{
2146	struct bfq_data *bfqd = q->elevator->elevator_data;
2147
2148	bfqd->rq_in_driver--;
2149}
2150#endif
2151
2152static void bfq_remove_request(struct request_queue *q,
2153			       struct request *rq)
2154{
2155	struct bfq_queue *bfqq = RQ_BFQQ(rq);
2156	struct bfq_data *bfqd = bfqq->bfqd;
2157	const int sync = rq_is_sync(rq);
2158
2159	if (bfqq->next_rq == rq) {
2160		bfqq->next_rq = bfq_find_next_rq(bfqd, bfqq, rq);
2161		bfq_updated_next_req(bfqd, bfqq);
2162	}
2163
2164	if (rq->queuelist.prev != &rq->queuelist)
2165		list_del_init(&rq->queuelist);
2166	bfqq->queued[sync]--;
2167	bfqd->queued--;
 
 
 
 
2168	elv_rb_del(&bfqq->sort_list, rq);
2169
2170	elv_rqhash_del(q, rq);
2171	if (q->last_merge == rq)
2172		q->last_merge = NULL;
2173
2174	if (RB_EMPTY_ROOT(&bfqq->sort_list)) {
2175		bfqq->next_rq = NULL;
2176
2177		if (bfq_bfqq_busy(bfqq) && bfqq != bfqd->in_service_queue) {
2178			bfq_del_bfqq_busy(bfqd, bfqq, false);
2179			/*
2180			 * bfqq emptied. In normal operation, when
2181			 * bfqq is empty, bfqq->entity.service and
2182			 * bfqq->entity.budget must contain,
2183			 * respectively, the service received and the
2184			 * budget used last time bfqq emptied. These
2185			 * facts do not hold in this case, as at least
2186			 * this last removal occurred while bfqq is
2187			 * not in service. To avoid inconsistencies,
2188			 * reset both bfqq->entity.service and
2189			 * bfqq->entity.budget, if bfqq has still a
2190			 * process that may issue I/O requests to it.
2191			 */
2192			bfqq->entity.budget = bfqq->entity.service = 0;
2193		}
2194
2195		/*
2196		 * Remove queue from request-position tree as it is empty.
2197		 */
2198		if (bfqq->pos_root) {
2199			rb_erase(&bfqq->pos_node, bfqq->pos_root);
2200			bfqq->pos_root = NULL;
2201		}
2202	} else {
2203		/* see comments on bfq_pos_tree_add_move() for the unlikely() */
2204		if (unlikely(!bfqd->nonrot_with_queueing))
2205			bfq_pos_tree_add_move(bfqd, bfqq);
2206	}
2207
2208	if (rq->cmd_flags & REQ_META)
2209		bfqq->meta_pending--;
2210
2211}
2212
2213static bool bfq_bio_merge(struct blk_mq_hw_ctx *hctx, struct bio *bio,
2214		unsigned int nr_segs)
2215{
2216	struct request_queue *q = hctx->queue;
2217	struct bfq_data *bfqd = q->elevator->elevator_data;
2218	struct request *free = NULL;
2219	/*
2220	 * bfq_bic_lookup grabs the queue_lock: invoke it now and
2221	 * store its return value for later use, to avoid nesting
2222	 * queue_lock inside the bfqd->lock. We assume that the bic
2223	 * returned by bfq_bic_lookup does not go away before
2224	 * bfqd->lock is taken.
2225	 */
2226	struct bfq_io_cq *bic = bfq_bic_lookup(bfqd, current->io_context, q);
2227	bool ret;
2228
2229	spin_lock_irq(&bfqd->lock);
2230
2231	if (bic)
 
 
 
 
 
 
2232		bfqd->bio_bfqq = bic_to_bfqq(bic, op_is_sync(bio->bi_opf));
2233	else
2234		bfqd->bio_bfqq = NULL;
 
2235	bfqd->bio_bic = bic;
2236
2237	ret = blk_mq_sched_try_merge(q, bio, nr_segs, &free);
2238
 
2239	if (free)
2240		blk_mq_free_request(free);
2241	spin_unlock_irq(&bfqd->lock);
2242
2243	return ret;
2244}
2245
2246static int bfq_request_merge(struct request_queue *q, struct request **req,
2247			     struct bio *bio)
2248{
2249	struct bfq_data *bfqd = q->elevator->elevator_data;
2250	struct request *__rq;
2251
2252	__rq = bfq_find_rq_fmerge(bfqd, bio, q);
2253	if (__rq && elv_bio_merge_ok(__rq, bio)) {
2254		*req = __rq;
 
 
 
2255		return ELEVATOR_FRONT_MERGE;
2256	}
2257
2258	return ELEVATOR_NO_MERGE;
2259}
2260
2261static struct bfq_queue *bfq_init_rq(struct request *rq);
2262
2263static void bfq_request_merged(struct request_queue *q, struct request *req,
2264			       enum elv_merge type)
2265{
2266	if (type == ELEVATOR_FRONT_MERGE &&
2267	    rb_prev(&req->rb_node) &&
2268	    blk_rq_pos(req) <
2269	    blk_rq_pos(container_of(rb_prev(&req->rb_node),
2270				    struct request, rb_node))) {
2271		struct bfq_queue *bfqq = bfq_init_rq(req);
2272		struct bfq_data *bfqd;
2273		struct request *prev, *next_rq;
2274
2275		if (!bfqq)
2276			return;
2277
2278		bfqd = bfqq->bfqd;
2279
2280		/* Reposition request in its sort_list */
2281		elv_rb_del(&bfqq->sort_list, req);
2282		elv_rb_add(&bfqq->sort_list, req);
2283
2284		/* Choose next request to be served for bfqq */
2285		prev = bfqq->next_rq;
2286		next_rq = bfq_choose_req(bfqd, bfqq->next_rq, req,
2287					 bfqd->last_position);
2288		bfqq->next_rq = next_rq;
2289		/*
2290		 * If next_rq changes, update both the queue's budget to
2291		 * fit the new request and the queue's position in its
2292		 * rq_pos_tree.
2293		 */
2294		if (prev != bfqq->next_rq) {
2295			bfq_updated_next_req(bfqd, bfqq);
2296			/*
2297			 * See comments on bfq_pos_tree_add_move() for
2298			 * the unlikely().
2299			 */
2300			if (unlikely(!bfqd->nonrot_with_queueing))
2301				bfq_pos_tree_add_move(bfqd, bfqq);
2302		}
2303	}
2304}
2305
2306/*
2307 * This function is called to notify the scheduler that the requests
2308 * rq and 'next' have been merged, with 'next' going away.  BFQ
2309 * exploits this hook to address the following issue: if 'next' has a
2310 * fifo_time lower that rq, then the fifo_time of rq must be set to
2311 * the value of 'next', to not forget the greater age of 'next'.
2312 *
2313 * NOTE: in this function we assume that rq is in a bfq_queue, basing
2314 * on that rq is picked from the hash table q->elevator->hash, which,
2315 * in its turn, is filled only with I/O requests present in
2316 * bfq_queues, while BFQ is in use for the request queue q. In fact,
2317 * the function that fills this hash table (elv_rqhash_add) is called
2318 * only by bfq_insert_request.
2319 */
2320static void bfq_requests_merged(struct request_queue *q, struct request *rq,
2321				struct request *next)
2322{
2323	struct bfq_queue *bfqq = bfq_init_rq(rq),
2324		*next_bfqq = bfq_init_rq(next);
2325
2326	if (!bfqq)
2327		return;
2328
2329	/*
2330	 * If next and rq belong to the same bfq_queue and next is older
2331	 * than rq, then reposition rq in the fifo (by substituting next
2332	 * with rq). Otherwise, if next and rq belong to different
2333	 * bfq_queues, never reposition rq: in fact, we would have to
2334	 * reposition it with respect to next's position in its own fifo,
2335	 * which would most certainly be too expensive with respect to
2336	 * the benefits.
2337	 */
2338	if (bfqq == next_bfqq &&
2339	    !list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
2340	    next->fifo_time < rq->fifo_time) {
2341		list_del_init(&rq->queuelist);
2342		list_replace_init(&next->queuelist, &rq->queuelist);
2343		rq->fifo_time = next->fifo_time;
2344	}
2345
2346	if (bfqq->next_rq == next)
2347		bfqq->next_rq = rq;
2348
2349	bfqg_stats_update_io_merged(bfqq_group(bfqq), next->cmd_flags);
 
 
 
 
 
 
 
 
2350}
2351
2352/* Must be called with bfqq != NULL */
2353static void bfq_bfqq_end_wr(struct bfq_queue *bfqq)
2354{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2355	if (bfq_bfqq_busy(bfqq))
2356		bfqq->bfqd->wr_busy_queues--;
2357	bfqq->wr_coeff = 1;
2358	bfqq->wr_cur_max_time = 0;
2359	bfqq->last_wr_start_finish = jiffies;
2360	/*
2361	 * Trigger a weight change on the next invocation of
2362	 * __bfq_entity_update_weight_prio.
2363	 */
2364	bfqq->entity.prio_changed = 1;
2365}
2366
2367void bfq_end_wr_async_queues(struct bfq_data *bfqd,
2368			     struct bfq_group *bfqg)
2369{
2370	int i, j;
2371
2372	for (i = 0; i < 2; i++)
2373		for (j = 0; j < IOPRIO_BE_NR; j++)
2374			if (bfqg->async_bfqq[i][j])
2375				bfq_bfqq_end_wr(bfqg->async_bfqq[i][j]);
2376	if (bfqg->async_idle_bfqq)
2377		bfq_bfqq_end_wr(bfqg->async_idle_bfqq);
2378}
2379
2380static void bfq_end_wr(struct bfq_data *bfqd)
2381{
2382	struct bfq_queue *bfqq;
2383
2384	spin_lock_irq(&bfqd->lock);
2385
2386	list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list)
2387		bfq_bfqq_end_wr(bfqq);
2388	list_for_each_entry(bfqq, &bfqd->idle_list, bfqq_list)
2389		bfq_bfqq_end_wr(bfqq);
2390	bfq_end_wr_async(bfqd);
2391
2392	spin_unlock_irq(&bfqd->lock);
2393}
2394
2395static sector_t bfq_io_struct_pos(void *io_struct, bool request)
2396{
2397	if (request)
2398		return blk_rq_pos(io_struct);
2399	else
2400		return ((struct bio *)io_struct)->bi_iter.bi_sector;
2401}
2402
2403static int bfq_rq_close_to_sector(void *io_struct, bool request,
2404				  sector_t sector)
2405{
2406	return abs(bfq_io_struct_pos(io_struct, request) - sector) <=
2407	       BFQQ_CLOSE_THR;
2408}
2409
2410static struct bfq_queue *bfqq_find_close(struct bfq_data *bfqd,
2411					 struct bfq_queue *bfqq,
2412					 sector_t sector)
2413{
2414	struct rb_root *root = &bfq_bfqq_to_bfqg(bfqq)->rq_pos_tree;
2415	struct rb_node *parent, *node;
2416	struct bfq_queue *__bfqq;
2417
2418	if (RB_EMPTY_ROOT(root))
2419		return NULL;
2420
2421	/*
2422	 * First, if we find a request starting at the end of the last
2423	 * request, choose it.
2424	 */
2425	__bfqq = bfq_rq_pos_tree_lookup(bfqd, root, sector, &parent, NULL);
2426	if (__bfqq)
2427		return __bfqq;
2428
2429	/*
2430	 * If the exact sector wasn't found, the parent of the NULL leaf
2431	 * will contain the closest sector (rq_pos_tree sorted by
2432	 * next_request position).
2433	 */
2434	__bfqq = rb_entry(parent, struct bfq_queue, pos_node);
2435	if (bfq_rq_close_to_sector(__bfqq->next_rq, true, sector))
2436		return __bfqq;
2437
2438	if (blk_rq_pos(__bfqq->next_rq) < sector)
2439		node = rb_next(&__bfqq->pos_node);
2440	else
2441		node = rb_prev(&__bfqq->pos_node);
2442	if (!node)
2443		return NULL;
2444
2445	__bfqq = rb_entry(node, struct bfq_queue, pos_node);
2446	if (bfq_rq_close_to_sector(__bfqq->next_rq, true, sector))
2447		return __bfqq;
2448
2449	return NULL;
2450}
2451
2452static struct bfq_queue *bfq_find_close_cooperator(struct bfq_data *bfqd,
2453						   struct bfq_queue *cur_bfqq,
2454						   sector_t sector)
2455{
2456	struct bfq_queue *bfqq;
2457
2458	/*
2459	 * We shall notice if some of the queues are cooperating,
2460	 * e.g., working closely on the same area of the device. In
2461	 * that case, we can group them together and: 1) don't waste
2462	 * time idling, and 2) serve the union of their requests in
2463	 * the best possible order for throughput.
2464	 */
2465	bfqq = bfqq_find_close(bfqd, cur_bfqq, sector);
2466	if (!bfqq || bfqq == cur_bfqq)
2467		return NULL;
2468
2469	return bfqq;
2470}
2471
2472static struct bfq_queue *
2473bfq_setup_merge(struct bfq_queue *bfqq, struct bfq_queue *new_bfqq)
2474{
2475	int process_refs, new_process_refs;
2476	struct bfq_queue *__bfqq;
2477
2478	/*
2479	 * If there are no process references on the new_bfqq, then it is
2480	 * unsafe to follow the ->new_bfqq chain as other bfqq's in the chain
2481	 * may have dropped their last reference (not just their last process
2482	 * reference).
2483	 */
2484	if (!bfqq_process_refs(new_bfqq))
2485		return NULL;
2486
2487	/* Avoid a circular list and skip interim queue merges. */
2488	while ((__bfqq = new_bfqq->new_bfqq)) {
2489		if (__bfqq == bfqq)
2490			return NULL;
2491		new_bfqq = __bfqq;
2492	}
2493
2494	process_refs = bfqq_process_refs(bfqq);
2495	new_process_refs = bfqq_process_refs(new_bfqq);
2496	/*
2497	 * If the process for the bfqq has gone away, there is no
2498	 * sense in merging the queues.
2499	 */
2500	if (process_refs == 0 || new_process_refs == 0)
2501		return NULL;
2502
 
 
 
 
 
 
 
 
2503	bfq_log_bfqq(bfqq->bfqd, bfqq, "scheduling merge with queue %d",
2504		new_bfqq->pid);
2505
2506	/*
2507	 * Merging is just a redirection: the requests of the process
2508	 * owning one of the two queues are redirected to the other queue.
2509	 * The latter queue, in its turn, is set as shared if this is the
2510	 * first time that the requests of some process are redirected to
2511	 * it.
2512	 *
2513	 * We redirect bfqq to new_bfqq and not the opposite, because
2514	 * we are in the context of the process owning bfqq, thus we
2515	 * have the io_cq of this process. So we can immediately
2516	 * configure this io_cq to redirect the requests of the
2517	 * process to new_bfqq. In contrast, the io_cq of new_bfqq is
2518	 * not available any more (new_bfqq->bic == NULL).
2519	 *
2520	 * Anyway, even in case new_bfqq coincides with the in-service
2521	 * queue, redirecting requests the in-service queue is the
2522	 * best option, as we feed the in-service queue with new
2523	 * requests close to the last request served and, by doing so,
2524	 * are likely to increase the throughput.
2525	 */
2526	bfqq->new_bfqq = new_bfqq;
 
 
 
 
 
 
 
 
 
2527	new_bfqq->ref += process_refs;
2528	return new_bfqq;
2529}
2530
2531static bool bfq_may_be_close_cooperator(struct bfq_queue *bfqq,
2532					struct bfq_queue *new_bfqq)
2533{
2534	if (bfq_too_late_for_merging(new_bfqq))
2535		return false;
2536
2537	if (bfq_class_idle(bfqq) || bfq_class_idle(new_bfqq) ||
2538	    (bfqq->ioprio_class != new_bfqq->ioprio_class))
2539		return false;
2540
2541	/*
2542	 * If either of the queues has already been detected as seeky,
2543	 * then merging it with the other queue is unlikely to lead to
2544	 * sequential I/O.
2545	 */
2546	if (BFQQ_SEEKY(bfqq) || BFQQ_SEEKY(new_bfqq))
2547		return false;
2548
2549	/*
2550	 * Interleaved I/O is known to be done by (some) applications
2551	 * only for reads, so it does not make sense to merge async
2552	 * queues.
2553	 */
2554	if (!bfq_bfqq_sync(bfqq) || !bfq_bfqq_sync(new_bfqq))
2555		return false;
2556
2557	return true;
2558}
2559
 
 
 
2560/*
2561 * Attempt to schedule a merge of bfqq with the currently in-service
2562 * queue or with a close queue among the scheduled queues.  Return
2563 * NULL if no merge was scheduled, a pointer to the shared bfq_queue
2564 * structure otherwise.
2565 *
2566 * The OOM queue is not allowed to participate to cooperation: in fact, since
2567 * the requests temporarily redirected to the OOM queue could be redirected
2568 * again to dedicated queues at any time, the state needed to correctly
2569 * handle merging with the OOM queue would be quite complex and expensive
2570 * to maintain. Besides, in such a critical condition as an out of memory,
2571 * the benefits of queue merging may be little relevant, or even negligible.
2572 *
2573 * WARNING: queue merging may impair fairness among non-weight raised
2574 * queues, for at least two reasons: 1) the original weight of a
2575 * merged queue may change during the merged state, 2) even being the
2576 * weight the same, a merged queue may be bloated with many more
2577 * requests than the ones produced by its originally-associated
2578 * process.
2579 */
2580static struct bfq_queue *
2581bfq_setup_cooperator(struct bfq_data *bfqd, struct bfq_queue *bfqq,
2582		     void *io_struct, bool request)
2583{
2584	struct bfq_queue *in_service_bfqq, *new_bfqq;
2585
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2586	/*
2587	 * Do not perform queue merging if the device is non
2588	 * rotational and performs internal queueing. In fact, such a
2589	 * device reaches a high speed through internal parallelism
2590	 * and pipelining. This means that, to reach a high
2591	 * throughput, it must have many requests enqueued at the same
2592	 * time. But, in this configuration, the internal scheduling
2593	 * algorithm of the device does exactly the job of queue
2594	 * merging: it reorders requests so as to obtain as much as
2595	 * possible a sequential I/O pattern. As a consequence, with
2596	 * the workload generated by processes doing interleaved I/O,
2597	 * the throughput reached by the device is likely to be the
2598	 * same, with and without queue merging.
2599	 *
2600	 * Disabling merging also provides a remarkable benefit in
2601	 * terms of throughput. Merging tends to make many workloads
2602	 * artificially more uneven, because of shared queues
2603	 * remaining non empty for incomparably more time than
2604	 * non-merged queues. This may accentuate workload
2605	 * asymmetries. For example, if one of the queues in a set of
2606	 * merged queues has a higher weight than a normal queue, then
2607	 * the shared queue may inherit such a high weight and, by
2608	 * staying almost always active, may force BFQ to perform I/O
2609	 * plugging most of the time. This evidently makes it harder
2610	 * for BFQ to let the device reach a high throughput.
2611	 *
2612	 * Finally, the likely() macro below is not used because one
2613	 * of the two branches is more likely than the other, but to
2614	 * have the code path after the following if() executed as
2615	 * fast as possible for the case of a non rotational device
2616	 * with queueing. We want it because this is the fastest kind
2617	 * of device. On the opposite end, the likely() may lengthen
2618	 * the execution time of BFQ for the case of slower devices
2619	 * (rotational or at least without queueing). But in this case
2620	 * the execution time of BFQ matters very little, if not at
2621	 * all.
2622	 */
2623	if (likely(bfqd->nonrot_with_queueing))
2624		return NULL;
2625
2626	/*
2627	 * Prevent bfqq from being merged if it has been created too
2628	 * long ago. The idea is that true cooperating processes, and
2629	 * thus their associated bfq_queues, are supposed to be
2630	 * created shortly after each other. This is the case, e.g.,
2631	 * for KVM/QEMU and dump I/O threads. Basing on this
2632	 * assumption, the following filtering greatly reduces the
2633	 * probability that two non-cooperating processes, which just
2634	 * happen to do close I/O for some short time interval, have
2635	 * their queues merged by mistake.
2636	 */
2637	if (bfq_too_late_for_merging(bfqq))
2638		return NULL;
2639
2640	if (bfqq->new_bfqq)
2641		return bfqq->new_bfqq;
2642
2643	if (!io_struct || unlikely(bfqq == &bfqd->oom_bfqq))
2644		return NULL;
2645
2646	/* If there is only one backlogged queue, don't search. */
2647	if (bfq_tot_busy_queues(bfqd) == 1)
2648		return NULL;
2649
2650	in_service_bfqq = bfqd->in_service_queue;
2651
2652	if (in_service_bfqq && in_service_bfqq != bfqq &&
2653	    likely(in_service_bfqq != &bfqd->oom_bfqq) &&
2654	    bfq_rq_close_to_sector(io_struct, request,
2655				   bfqd->in_serv_last_pos) &&
2656	    bfqq->entity.parent == in_service_bfqq->entity.parent &&
2657	    bfq_may_be_close_cooperator(bfqq, in_service_bfqq)) {
2658		new_bfqq = bfq_setup_merge(bfqq, in_service_bfqq);
2659		if (new_bfqq)
2660			return new_bfqq;
2661	}
2662	/*
2663	 * Check whether there is a cooperator among currently scheduled
2664	 * queues. The only thing we need is that the bio/request is not
2665	 * NULL, as we need it to establish whether a cooperator exists.
2666	 */
2667	new_bfqq = bfq_find_close_cooperator(bfqd, bfqq,
2668			bfq_io_struct_pos(io_struct, request));
2669
2670	if (new_bfqq && likely(new_bfqq != &bfqd->oom_bfqq) &&
2671	    bfq_may_be_close_cooperator(bfqq, new_bfqq))
2672		return bfq_setup_merge(bfqq, new_bfqq);
2673
2674	return NULL;
2675}
2676
2677static void bfq_bfqq_save_state(struct bfq_queue *bfqq)
2678{
2679	struct bfq_io_cq *bic = bfqq->bic;
2680
2681	/*
2682	 * If !bfqq->bic, the queue is already shared or its requests
2683	 * have already been redirected to a shared queue; both idle window
2684	 * and weight raising state have already been saved. Do nothing.
2685	 */
2686	if (!bic)
2687		return;
2688
 
 
 
 
2689	bic->saved_weight = bfqq->entity.orig_weight;
2690	bic->saved_ttime = bfqq->ttime;
2691	bic->saved_has_short_ttime = bfq_bfqq_has_short_ttime(bfqq);
2692	bic->saved_IO_bound = bfq_bfqq_IO_bound(bfqq);
 
 
2693	bic->saved_in_large_burst = bfq_bfqq_in_large_burst(bfqq);
2694	bic->was_in_burst_list = !hlist_unhashed(&bfqq->burst_list_node);
2695	if (unlikely(bfq_bfqq_just_created(bfqq) &&
2696		     !bfq_bfqq_in_large_burst(bfqq) &&
2697		     bfqq->bfqd->low_latency)) {
2698		/*
2699		 * bfqq being merged right after being created: bfqq
2700		 * would have deserved interactive weight raising, but
2701		 * did not make it to be set in a weight-raised state,
2702		 * because of this early merge.	Store directly the
2703		 * weight-raising state that would have been assigned
2704		 * to bfqq, so that to avoid that bfqq unjustly fails
2705		 * to enjoy weight raising if split soon.
2706		 */
2707		bic->saved_wr_coeff = bfqq->bfqd->bfq_wr_coeff;
2708		bic->saved_wr_start_at_switch_to_srt = bfq_smallest_from_now();
2709		bic->saved_wr_cur_max_time = bfq_wr_duration(bfqq->bfqd);
2710		bic->saved_last_wr_start_finish = jiffies;
2711	} else {
2712		bic->saved_wr_coeff = bfqq->wr_coeff;
2713		bic->saved_wr_start_at_switch_to_srt =
2714			bfqq->wr_start_at_switch_to_srt;
 
2715		bic->saved_last_wr_start_finish = bfqq->last_wr_start_finish;
2716		bic->saved_wr_cur_max_time = bfqq->wr_cur_max_time;
2717	}
2718}
2719
 
 
 
 
 
 
 
 
 
 
 
2720void bfq_release_process_ref(struct bfq_data *bfqd, struct bfq_queue *bfqq)
2721{
2722	/*
2723	 * To prevent bfqq's service guarantees from being violated,
2724	 * bfqq may be left busy, i.e., queued for service, even if
2725	 * empty (see comments in __bfq_bfqq_expire() for
2726	 * details). But, if no process will send requests to bfqq any
2727	 * longer, then there is no point in keeping bfqq queued for
2728	 * service. In addition, keeping bfqq queued for service, but
2729	 * with no process ref any longer, may have caused bfqq to be
2730	 * freed when dequeued from service. But this is assumed to
2731	 * never happen.
2732	 */
2733	if (bfq_bfqq_busy(bfqq) && RB_EMPTY_ROOT(&bfqq->sort_list) &&
2734	    bfqq != bfqd->in_service_queue)
2735		bfq_del_bfqq_busy(bfqd, bfqq, false);
 
 
2736
2737	bfq_put_queue(bfqq);
2738}
2739
2740static void
2741bfq_merge_bfqqs(struct bfq_data *bfqd, struct bfq_io_cq *bic,
2742		struct bfq_queue *bfqq, struct bfq_queue *new_bfqq)
2743{
2744	bfq_log_bfqq(bfqd, bfqq, "merging with queue %lu",
2745		(unsigned long)new_bfqq->pid);
2746	/* Save weight raising and idle window of the merged queues */
2747	bfq_bfqq_save_state(bfqq);
2748	bfq_bfqq_save_state(new_bfqq);
2749	if (bfq_bfqq_IO_bound(bfqq))
2750		bfq_mark_bfqq_IO_bound(new_bfqq);
2751	bfq_clear_bfqq_IO_bound(bfqq);
2752
2753	/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2754	 * If bfqq is weight-raised, then let new_bfqq inherit
2755	 * weight-raising. To reduce false positives, neglect the case
2756	 * where bfqq has just been created, but has not yet made it
2757	 * to be weight-raised (which may happen because EQM may merge
2758	 * bfqq even before bfq_add_request is executed for the first
2759	 * time for bfqq). Handling this case would however be very
2760	 * easy, thanks to the flag just_created.
2761	 */
2762	if (new_bfqq->wr_coeff == 1 && bfqq->wr_coeff > 1) {
2763		new_bfqq->wr_coeff = bfqq->wr_coeff;
2764		new_bfqq->wr_cur_max_time = bfqq->wr_cur_max_time;
2765		new_bfqq->last_wr_start_finish = bfqq->last_wr_start_finish;
2766		new_bfqq->wr_start_at_switch_to_srt =
2767			bfqq->wr_start_at_switch_to_srt;
2768		if (bfq_bfqq_busy(new_bfqq))
2769			bfqd->wr_busy_queues++;
2770		new_bfqq->entity.prio_changed = 1;
2771	}
2772
2773	if (bfqq->wr_coeff > 1) { /* bfqq has given its wr to new_bfqq */
2774		bfqq->wr_coeff = 1;
2775		bfqq->entity.prio_changed = 1;
2776		if (bfq_bfqq_busy(bfqq))
2777			bfqd->wr_busy_queues--;
2778	}
2779
2780	bfq_log_bfqq(bfqd, new_bfqq, "merge_bfqqs: wr_busy %d",
2781		     bfqd->wr_busy_queues);
2782
2783	/*
2784	 * Merge queues (that is, let bic redirect its requests to new_bfqq)
2785	 */
2786	bic_set_bfqq(bic, new_bfqq, 1);
2787	bfq_mark_bfqq_coop(new_bfqq);
2788	/*
2789	 * new_bfqq now belongs to at least two bics (it is a shared queue):
2790	 * set new_bfqq->bic to NULL. bfqq either:
2791	 * - does not belong to any bic any more, and hence bfqq->bic must
2792	 *   be set to NULL, or
2793	 * - is a queue whose owning bics have already been redirected to a
2794	 *   different queue, hence the queue is destined to not belong to
2795	 *   any bic soon and bfqq->bic is already NULL (therefore the next
2796	 *   assignment causes no harm).
2797	 */
2798	new_bfqq->bic = NULL;
2799	/*
2800	 * If the queue is shared, the pid is the pid of one of the associated
2801	 * processes. Which pid depends on the exact sequence of merge events
2802	 * the queue underwent. So printing such a pid is useless and confusing
2803	 * because it reports a random pid between those of the associated
2804	 * processes.
2805	 * We mark such a queue with a pid -1, and then print SHARED instead of
2806	 * a pid in logging messages.
2807	 */
2808	new_bfqq->pid = -1;
2809	bfqq->bic = NULL;
 
 
 
2810	bfq_release_process_ref(bfqd, bfqq);
2811}
2812
2813static bool bfq_allow_bio_merge(struct request_queue *q, struct request *rq,
2814				struct bio *bio)
2815{
2816	struct bfq_data *bfqd = q->elevator->elevator_data;
2817	bool is_sync = op_is_sync(bio->bi_opf);
2818	struct bfq_queue *bfqq = bfqd->bio_bfqq, *new_bfqq;
2819
2820	/*
2821	 * Disallow merge of a sync bio into an async request.
2822	 */
2823	if (is_sync && !rq_is_sync(rq))
2824		return false;
2825
2826	/*
2827	 * Lookup the bfqq that this bio will be queued with. Allow
2828	 * merge only if rq is queued there.
2829	 */
2830	if (!bfqq)
2831		return false;
2832
2833	/*
2834	 * We take advantage of this function to perform an early merge
2835	 * of the queues of possible cooperating processes.
2836	 */
2837	new_bfqq = bfq_setup_cooperator(bfqd, bfqq, bio, false);
2838	if (new_bfqq) {
2839		/*
2840		 * bic still points to bfqq, then it has not yet been
2841		 * redirected to some other bfq_queue, and a queue
2842		 * merge between bfqq and new_bfqq can be safely
2843		 * fulfilled, i.e., bic can be redirected to new_bfqq
2844		 * and bfqq can be put.
2845		 */
2846		bfq_merge_bfqqs(bfqd, bfqd->bio_bic, bfqq,
2847				new_bfqq);
2848		/*
2849		 * If we get here, bio will be queued into new_queue,
2850		 * so use new_bfqq to decide whether bio and rq can be
2851		 * merged.
2852		 */
2853		bfqq = new_bfqq;
2854
2855		/*
2856		 * Change also bqfd->bio_bfqq, as
2857		 * bfqd->bio_bic now points to new_bfqq, and
2858		 * this function may be invoked again (and then may
2859		 * use again bqfd->bio_bfqq).
2860		 */
2861		bfqd->bio_bfqq = bfqq;
2862	}
2863
2864	return bfqq == RQ_BFQQ(rq);
2865}
2866
2867/*
2868 * Set the maximum time for the in-service queue to consume its
2869 * budget. This prevents seeky processes from lowering the throughput.
2870 * In practice, a time-slice service scheme is used with seeky
2871 * processes.
2872 */
2873static void bfq_set_budget_timeout(struct bfq_data *bfqd,
2874				   struct bfq_queue *bfqq)
2875{
2876	unsigned int timeout_coeff;
2877
2878	if (bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time)
2879		timeout_coeff = 1;
2880	else
2881		timeout_coeff = bfqq->entity.weight / bfqq->entity.orig_weight;
2882
2883	bfqd->last_budget_start = ktime_get();
2884
2885	bfqq->budget_timeout = jiffies +
2886		bfqd->bfq_timeout * timeout_coeff;
2887}
2888
2889static void __bfq_set_in_service_queue(struct bfq_data *bfqd,
2890				       struct bfq_queue *bfqq)
2891{
2892	if (bfqq) {
2893		bfq_clear_bfqq_fifo_expire(bfqq);
2894
2895		bfqd->budgets_assigned = (bfqd->budgets_assigned * 7 + 256) / 8;
2896
2897		if (time_is_before_jiffies(bfqq->last_wr_start_finish) &&
2898		    bfqq->wr_coeff > 1 &&
2899		    bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
2900		    time_is_before_jiffies(bfqq->budget_timeout)) {
2901			/*
2902			 * For soft real-time queues, move the start
2903			 * of the weight-raising period forward by the
2904			 * time the queue has not received any
2905			 * service. Otherwise, a relatively long
2906			 * service delay is likely to cause the
2907			 * weight-raising period of the queue to end,
2908			 * because of the short duration of the
2909			 * weight-raising period of a soft real-time
2910			 * queue.  It is worth noting that this move
2911			 * is not so dangerous for the other queues,
2912			 * because soft real-time queues are not
2913			 * greedy.
2914			 *
2915			 * To not add a further variable, we use the
2916			 * overloaded field budget_timeout to
2917			 * determine for how long the queue has not
2918			 * received service, i.e., how much time has
2919			 * elapsed since the queue expired. However,
2920			 * this is a little imprecise, because
2921			 * budget_timeout is set to jiffies if bfqq
2922			 * not only expires, but also remains with no
2923			 * request.
2924			 */
2925			if (time_after(bfqq->budget_timeout,
2926				       bfqq->last_wr_start_finish))
2927				bfqq->last_wr_start_finish +=
2928					jiffies - bfqq->budget_timeout;
2929			else
2930				bfqq->last_wr_start_finish = jiffies;
2931		}
2932
2933		bfq_set_budget_timeout(bfqd, bfqq);
2934		bfq_log_bfqq(bfqd, bfqq,
2935			     "set_in_service_queue, cur-budget = %d",
2936			     bfqq->entity.budget);
2937	}
2938
2939	bfqd->in_service_queue = bfqq;
 
2940}
2941
2942/*
2943 * Get and set a new queue for service.
2944 */
2945static struct bfq_queue *bfq_set_in_service_queue(struct bfq_data *bfqd)
2946{
2947	struct bfq_queue *bfqq = bfq_get_next_queue(bfqd);
2948
2949	__bfq_set_in_service_queue(bfqd, bfqq);
2950	return bfqq;
2951}
2952
2953static void bfq_arm_slice_timer(struct bfq_data *bfqd)
2954{
2955	struct bfq_queue *bfqq = bfqd->in_service_queue;
2956	u32 sl;
2957
2958	bfq_mark_bfqq_wait_request(bfqq);
2959
2960	/*
2961	 * We don't want to idle for seeks, but we do want to allow
2962	 * fair distribution of slice time for a process doing back-to-back
2963	 * seeks. So allow a little bit of time for him to submit a new rq.
2964	 */
2965	sl = bfqd->bfq_slice_idle;
2966	/*
2967	 * Unless the queue is being weight-raised or the scenario is
2968	 * asymmetric, grant only minimum idle time if the queue
2969	 * is seeky. A long idling is preserved for a weight-raised
2970	 * queue, or, more in general, in an asymmetric scenario,
2971	 * because a long idling is needed for guaranteeing to a queue
2972	 * its reserved share of the throughput (in particular, it is
2973	 * needed if the queue has a higher weight than some other
2974	 * queue).
2975	 */
2976	if (BFQQ_SEEKY(bfqq) && bfqq->wr_coeff == 1 &&
2977	    !bfq_asymmetric_scenario(bfqd, bfqq))
2978		sl = min_t(u64, sl, BFQ_MIN_TT);
2979	else if (bfqq->wr_coeff > 1)
2980		sl = max_t(u32, sl, 20ULL * NSEC_PER_MSEC);
2981
2982	bfqd->last_idling_start = ktime_get();
2983	bfqd->last_idling_start_jiffies = jiffies;
2984
2985	hrtimer_start(&bfqd->idle_slice_timer, ns_to_ktime(sl),
2986		      HRTIMER_MODE_REL);
2987	bfqg_stats_set_start_idle_time(bfqq_group(bfqq));
2988}
2989
2990/*
2991 * In autotuning mode, max_budget is dynamically recomputed as the
2992 * amount of sectors transferred in timeout at the estimated peak
2993 * rate. This enables BFQ to utilize a full timeslice with a full
2994 * budget, even if the in-service queue is served at peak rate. And
2995 * this maximises throughput with sequential workloads.
2996 */
2997static unsigned long bfq_calc_max_budget(struct bfq_data *bfqd)
2998{
2999	return (u64)bfqd->peak_rate * USEC_PER_MSEC *
3000		jiffies_to_msecs(bfqd->bfq_timeout)>>BFQ_RATE_SHIFT;
3001}
3002
3003/*
3004 * Update parameters related to throughput and responsiveness, as a
3005 * function of the estimated peak rate. See comments on
3006 * bfq_calc_max_budget(), and on the ref_wr_duration array.
3007 */
3008static void update_thr_responsiveness_params(struct bfq_data *bfqd)
3009{
3010	if (bfqd->bfq_user_max_budget == 0) {
3011		bfqd->bfq_max_budget =
3012			bfq_calc_max_budget(bfqd);
3013		bfq_log(bfqd, "new max_budget = %d", bfqd->bfq_max_budget);
3014	}
3015}
3016
3017static void bfq_reset_rate_computation(struct bfq_data *bfqd,
3018				       struct request *rq)
3019{
3020	if (rq != NULL) { /* new rq dispatch now, reset accordingly */
3021		bfqd->last_dispatch = bfqd->first_dispatch = ktime_get_ns();
3022		bfqd->peak_rate_samples = 1;
3023		bfqd->sequential_samples = 0;
3024		bfqd->tot_sectors_dispatched = bfqd->last_rq_max_size =
3025			blk_rq_sectors(rq);
3026	} else /* no new rq dispatched, just reset the number of samples */
3027		bfqd->peak_rate_samples = 0; /* full re-init on next disp. */
3028
3029	bfq_log(bfqd,
3030		"reset_rate_computation at end, sample %u/%u tot_sects %llu",
3031		bfqd->peak_rate_samples, bfqd->sequential_samples,
3032		bfqd->tot_sectors_dispatched);
3033}
3034
3035static void bfq_update_rate_reset(struct bfq_data *bfqd, struct request *rq)
3036{
3037	u32 rate, weight, divisor;
3038
3039	/*
3040	 * For the convergence property to hold (see comments on
3041	 * bfq_update_peak_rate()) and for the assessment to be
3042	 * reliable, a minimum number of samples must be present, and
3043	 * a minimum amount of time must have elapsed. If not so, do
3044	 * not compute new rate. Just reset parameters, to get ready
3045	 * for a new evaluation attempt.
3046	 */
3047	if (bfqd->peak_rate_samples < BFQ_RATE_MIN_SAMPLES ||
3048	    bfqd->delta_from_first < BFQ_RATE_MIN_INTERVAL)
3049		goto reset_computation;
3050
3051	/*
3052	 * If a new request completion has occurred after last
3053	 * dispatch, then, to approximate the rate at which requests
3054	 * have been served by the device, it is more precise to
3055	 * extend the observation interval to the last completion.
3056	 */
3057	bfqd->delta_from_first =
3058		max_t(u64, bfqd->delta_from_first,
3059		      bfqd->last_completion - bfqd->first_dispatch);
3060
3061	/*
3062	 * Rate computed in sects/usec, and not sects/nsec, for
3063	 * precision issues.
3064	 */
3065	rate = div64_ul(bfqd->tot_sectors_dispatched<<BFQ_RATE_SHIFT,
3066			div_u64(bfqd->delta_from_first, NSEC_PER_USEC));
3067
3068	/*
3069	 * Peak rate not updated if:
3070	 * - the percentage of sequential dispatches is below 3/4 of the
3071	 *   total, and rate is below the current estimated peak rate
3072	 * - rate is unreasonably high (> 20M sectors/sec)
3073	 */
3074	if ((bfqd->sequential_samples < (3 * bfqd->peak_rate_samples)>>2 &&
3075	     rate <= bfqd->peak_rate) ||
3076		rate > 20<<BFQ_RATE_SHIFT)
3077		goto reset_computation;
3078
3079	/*
3080	 * We have to update the peak rate, at last! To this purpose,
3081	 * we use a low-pass filter. We compute the smoothing constant
3082	 * of the filter as a function of the 'weight' of the new
3083	 * measured rate.
3084	 *
3085	 * As can be seen in next formulas, we define this weight as a
3086	 * quantity proportional to how sequential the workload is,
3087	 * and to how long the observation time interval is.
3088	 *
3089	 * The weight runs from 0 to 8. The maximum value of the
3090	 * weight, 8, yields the minimum value for the smoothing
3091	 * constant. At this minimum value for the smoothing constant,
3092	 * the measured rate contributes for half of the next value of
3093	 * the estimated peak rate.
3094	 *
3095	 * So, the first step is to compute the weight as a function
3096	 * of how sequential the workload is. Note that the weight
3097	 * cannot reach 9, because bfqd->sequential_samples cannot
3098	 * become equal to bfqd->peak_rate_samples, which, in its
3099	 * turn, holds true because bfqd->sequential_samples is not
3100	 * incremented for the first sample.
3101	 */
3102	weight = (9 * bfqd->sequential_samples) / bfqd->peak_rate_samples;
3103
3104	/*
3105	 * Second step: further refine the weight as a function of the
3106	 * duration of the observation interval.
3107	 */
3108	weight = min_t(u32, 8,
3109		       div_u64(weight * bfqd->delta_from_first,
3110			       BFQ_RATE_REF_INTERVAL));
3111
3112	/*
3113	 * Divisor ranging from 10, for minimum weight, to 2, for
3114	 * maximum weight.
3115	 */
3116	divisor = 10 - weight;
3117
3118	/*
3119	 * Finally, update peak rate:
3120	 *
3121	 * peak_rate = peak_rate * (divisor-1) / divisor  +  rate / divisor
3122	 */
3123	bfqd->peak_rate *= divisor-1;
3124	bfqd->peak_rate /= divisor;
3125	rate /= divisor; /* smoothing constant alpha = 1/divisor */
3126
3127	bfqd->peak_rate += rate;
3128
3129	/*
3130	 * For a very slow device, bfqd->peak_rate can reach 0 (see
3131	 * the minimum representable values reported in the comments
3132	 * on BFQ_RATE_SHIFT). Push to 1 if this happens, to avoid
3133	 * divisions by zero where bfqd->peak_rate is used as a
3134	 * divisor.
3135	 */
3136	bfqd->peak_rate = max_t(u32, 1, bfqd->peak_rate);
3137
3138	update_thr_responsiveness_params(bfqd);
3139
3140reset_computation:
3141	bfq_reset_rate_computation(bfqd, rq);
3142}
3143
3144/*
3145 * Update the read/write peak rate (the main quantity used for
3146 * auto-tuning, see update_thr_responsiveness_params()).
3147 *
3148 * It is not trivial to estimate the peak rate (correctly): because of
3149 * the presence of sw and hw queues between the scheduler and the
3150 * device components that finally serve I/O requests, it is hard to
3151 * say exactly when a given dispatched request is served inside the
3152 * device, and for how long. As a consequence, it is hard to know
3153 * precisely at what rate a given set of requests is actually served
3154 * by the device.
3155 *
3156 * On the opposite end, the dispatch time of any request is trivially
3157 * available, and, from this piece of information, the "dispatch rate"
3158 * of requests can be immediately computed. So, the idea in the next
3159 * function is to use what is known, namely request dispatch times
3160 * (plus, when useful, request completion times), to estimate what is
3161 * unknown, namely in-device request service rate.
3162 *
3163 * The main issue is that, because of the above facts, the rate at
3164 * which a certain set of requests is dispatched over a certain time
3165 * interval can vary greatly with respect to the rate at which the
3166 * same requests are then served. But, since the size of any
3167 * intermediate queue is limited, and the service scheme is lossless
3168 * (no request is silently dropped), the following obvious convergence
3169 * property holds: the number of requests dispatched MUST become
3170 * closer and closer to the number of requests completed as the
3171 * observation interval grows. This is the key property used in
3172 * the next function to estimate the peak service rate as a function
3173 * of the observed dispatch rate. The function assumes to be invoked
3174 * on every request dispatch.
3175 */
3176static void bfq_update_peak_rate(struct bfq_data *bfqd, struct request *rq)
3177{
3178	u64 now_ns = ktime_get_ns();
3179
3180	if (bfqd->peak_rate_samples == 0) { /* first dispatch */
3181		bfq_log(bfqd, "update_peak_rate: goto reset, samples %d",
3182			bfqd->peak_rate_samples);
3183		bfq_reset_rate_computation(bfqd, rq);
3184		goto update_last_values; /* will add one sample */
3185	}
3186
3187	/*
3188	 * Device idle for very long: the observation interval lasting
3189	 * up to this dispatch cannot be a valid observation interval
3190	 * for computing a new peak rate (similarly to the late-
3191	 * completion event in bfq_completed_request()). Go to
3192	 * update_rate_and_reset to have the following three steps
3193	 * taken:
3194	 * - close the observation interval at the last (previous)
3195	 *   request dispatch or completion
3196	 * - compute rate, if possible, for that observation interval
3197	 * - start a new observation interval with this dispatch
3198	 */
3199	if (now_ns - bfqd->last_dispatch > 100*NSEC_PER_MSEC &&
3200	    bfqd->rq_in_driver == 0)
3201		goto update_rate_and_reset;
3202
3203	/* Update sampling information */
3204	bfqd->peak_rate_samples++;
3205
3206	if ((bfqd->rq_in_driver > 0 ||
3207		now_ns - bfqd->last_completion < BFQ_MIN_TT)
3208	    && !BFQ_RQ_SEEKY(bfqd, bfqd->last_position, rq))
3209		bfqd->sequential_samples++;
3210
3211	bfqd->tot_sectors_dispatched += blk_rq_sectors(rq);
3212
3213	/* Reset max observed rq size every 32 dispatches */
3214	if (likely(bfqd->peak_rate_samples % 32))
3215		bfqd->last_rq_max_size =
3216			max_t(u32, blk_rq_sectors(rq), bfqd->last_rq_max_size);
3217	else
3218		bfqd->last_rq_max_size = blk_rq_sectors(rq);
3219
3220	bfqd->delta_from_first = now_ns - bfqd->first_dispatch;
3221
3222	/* Target observation interval not yet reached, go on sampling */
3223	if (bfqd->delta_from_first < BFQ_RATE_REF_INTERVAL)
3224		goto update_last_values;
3225
3226update_rate_and_reset:
3227	bfq_update_rate_reset(bfqd, rq);
3228update_last_values:
3229	bfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
3230	if (RQ_BFQQ(rq) == bfqd->in_service_queue)
3231		bfqd->in_serv_last_pos = bfqd->last_position;
3232	bfqd->last_dispatch = now_ns;
3233}
3234
3235/*
3236 * Remove request from internal lists.
3237 */
3238static void bfq_dispatch_remove(struct request_queue *q, struct request *rq)
3239{
3240	struct bfq_queue *bfqq = RQ_BFQQ(rq);
3241
3242	/*
3243	 * For consistency, the next instruction should have been
3244	 * executed after removing the request from the queue and
3245	 * dispatching it.  We execute instead this instruction before
3246	 * bfq_remove_request() (and hence introduce a temporary
3247	 * inconsistency), for efficiency.  In fact, should this
3248	 * dispatch occur for a non in-service bfqq, this anticipated
3249	 * increment prevents two counters related to bfqq->dispatched
3250	 * from risking to be, first, uselessly decremented, and then
3251	 * incremented again when the (new) value of bfqq->dispatched
3252	 * happens to be taken into account.
3253	 */
3254	bfqq->dispatched++;
3255	bfq_update_peak_rate(q->elevator->elevator_data, rq);
3256
3257	bfq_remove_request(q, rq);
3258}
3259
3260/*
3261 * There is a case where idling does not have to be performed for
3262 * throughput concerns, but to preserve the throughput share of
3263 * the process associated with bfqq.
3264 *
3265 * To introduce this case, we can note that allowing the drive
3266 * to enqueue more than one request at a time, and hence
3267 * delegating de facto final scheduling decisions to the
3268 * drive's internal scheduler, entails loss of control on the
3269 * actual request service order. In particular, the critical
3270 * situation is when requests from different processes happen
3271 * to be present, at the same time, in the internal queue(s)
3272 * of the drive. In such a situation, the drive, by deciding
3273 * the service order of the internally-queued requests, does
3274 * determine also the actual throughput distribution among
3275 * these processes. But the drive typically has no notion or
3276 * concern about per-process throughput distribution, and
3277 * makes its decisions only on a per-request basis. Therefore,
3278 * the service distribution enforced by the drive's internal
3279 * scheduler is likely to coincide with the desired throughput
3280 * distribution only in a completely symmetric, or favorably
3281 * skewed scenario where:
3282 * (i-a) each of these processes must get the same throughput as
3283 *	 the others,
3284 * (i-b) in case (i-a) does not hold, it holds that the process
3285 *       associated with bfqq must receive a lower or equal
3286 *	 throughput than any of the other processes;
3287 * (ii)  the I/O of each process has the same properties, in
3288 *       terms of locality (sequential or random), direction
3289 *       (reads or writes), request sizes, greediness
3290 *       (from I/O-bound to sporadic), and so on;
3291
3292 * In fact, in such a scenario, the drive tends to treat the requests
3293 * of each process in about the same way as the requests of the
3294 * others, and thus to provide each of these processes with about the
3295 * same throughput.  This is exactly the desired throughput
3296 * distribution if (i-a) holds, or, if (i-b) holds instead, this is an
3297 * even more convenient distribution for (the process associated with)
3298 * bfqq.
3299 *
3300 * In contrast, in any asymmetric or unfavorable scenario, device
3301 * idling (I/O-dispatch plugging) is certainly needed to guarantee
3302 * that bfqq receives its assigned fraction of the device throughput
3303 * (see [1] for details).
3304 *
3305 * The problem is that idling may significantly reduce throughput with
3306 * certain combinations of types of I/O and devices. An important
3307 * example is sync random I/O on flash storage with command
3308 * queueing. So, unless bfqq falls in cases where idling also boosts
3309 * throughput, it is important to check conditions (i-a), i(-b) and
3310 * (ii) accurately, so as to avoid idling when not strictly needed for
3311 * service guarantees.
3312 *
3313 * Unfortunately, it is extremely difficult to thoroughly check
3314 * condition (ii). And, in case there are active groups, it becomes
3315 * very difficult to check conditions (i-a) and (i-b) too.  In fact,
3316 * if there are active groups, then, for conditions (i-a) or (i-b) to
3317 * become false 'indirectly', it is enough that an active group
3318 * contains more active processes or sub-groups than some other active
3319 * group. More precisely, for conditions (i-a) or (i-b) to become
3320 * false because of such a group, it is not even necessary that the
3321 * group is (still) active: it is sufficient that, even if the group
3322 * has become inactive, some of its descendant processes still have
3323 * some request already dispatched but still waiting for
3324 * completion. In fact, requests have still to be guaranteed their
3325 * share of the throughput even after being dispatched. In this
3326 * respect, it is easy to show that, if a group frequently becomes
3327 * inactive while still having in-flight requests, and if, when this
3328 * happens, the group is not considered in the calculation of whether
3329 * the scenario is asymmetric, then the group may fail to be
3330 * guaranteed its fair share of the throughput (basically because
3331 * idling may not be performed for the descendant processes of the
3332 * group, but it had to be).  We address this issue with the following
3333 * bi-modal behavior, implemented in the function
3334 * bfq_asymmetric_scenario().
3335 *
3336 * If there are groups with requests waiting for completion
3337 * (as commented above, some of these groups may even be
3338 * already inactive), then the scenario is tagged as
3339 * asymmetric, conservatively, without checking any of the
3340 * conditions (i-a), (i-b) or (ii). So the device is idled for bfqq.
3341 * This behavior matches also the fact that groups are created
3342 * exactly if controlling I/O is a primary concern (to
3343 * preserve bandwidth and latency guarantees).
3344 *
3345 * On the opposite end, if there are no groups with requests waiting
3346 * for completion, then only conditions (i-a) and (i-b) are actually
3347 * controlled, i.e., provided that conditions (i-a) or (i-b) holds,
3348 * idling is not performed, regardless of whether condition (ii)
3349 * holds.  In other words, only if conditions (i-a) and (i-b) do not
3350 * hold, then idling is allowed, and the device tends to be prevented
3351 * from queueing many requests, possibly of several processes. Since
3352 * there are no groups with requests waiting for completion, then, to
3353 * control conditions (i-a) and (i-b) it is enough to check just
3354 * whether all the queues with requests waiting for completion also
3355 * have the same weight.
3356 *
3357 * Not checking condition (ii) evidently exposes bfqq to the
3358 * risk of getting less throughput than its fair share.
3359 * However, for queues with the same weight, a further
3360 * mechanism, preemption, mitigates or even eliminates this
3361 * problem. And it does so without consequences on overall
3362 * throughput. This mechanism and its benefits are explained
3363 * in the next three paragraphs.
3364 *
3365 * Even if a queue, say Q, is expired when it remains idle, Q
3366 * can still preempt the new in-service queue if the next
3367 * request of Q arrives soon (see the comments on
3368 * bfq_bfqq_update_budg_for_activation). If all queues and
3369 * groups have the same weight, this form of preemption,
3370 * combined with the hole-recovery heuristic described in the
3371 * comments on function bfq_bfqq_update_budg_for_activation,
3372 * are enough to preserve a correct bandwidth distribution in
3373 * the mid term, even without idling. In fact, even if not
3374 * idling allows the internal queues of the device to contain
3375 * many requests, and thus to reorder requests, we can rather
3376 * safely assume that the internal scheduler still preserves a
3377 * minimum of mid-term fairness.
3378 *
3379 * More precisely, this preemption-based, idleless approach
3380 * provides fairness in terms of IOPS, and not sectors per
3381 * second. This can be seen with a simple example. Suppose
3382 * that there are two queues with the same weight, but that
3383 * the first queue receives requests of 8 sectors, while the
3384 * second queue receives requests of 1024 sectors. In
3385 * addition, suppose that each of the two queues contains at
3386 * most one request at a time, which implies that each queue
3387 * always remains idle after it is served. Finally, after
3388 * remaining idle, each queue receives very quickly a new
3389 * request. It follows that the two queues are served
3390 * alternatively, preempting each other if needed. This
3391 * implies that, although both queues have the same weight,
3392 * the queue with large requests receives a service that is
3393 * 1024/8 times as high as the service received by the other
3394 * queue.
3395 *
3396 * The motivation for using preemption instead of idling (for
3397 * queues with the same weight) is that, by not idling,
3398 * service guarantees are preserved (completely or at least in
3399 * part) without minimally sacrificing throughput. And, if
3400 * there is no active group, then the primary expectation for
3401 * this device is probably a high throughput.
3402 *
3403 * We are now left only with explaining the two sub-conditions in the
3404 * additional compound condition that is checked below for deciding
3405 * whether the scenario is asymmetric. To explain the first
3406 * sub-condition, we need to add that the function
3407 * bfq_asymmetric_scenario checks the weights of only
3408 * non-weight-raised queues, for efficiency reasons (see comments on
3409 * bfq_weights_tree_add()). Then the fact that bfqq is weight-raised
3410 * is checked explicitly here. More precisely, the compound condition
3411 * below takes into account also the fact that, even if bfqq is being
3412 * weight-raised, the scenario is still symmetric if all queues with
3413 * requests waiting for completion happen to be
3414 * weight-raised. Actually, we should be even more precise here, and
3415 * differentiate between interactive weight raising and soft real-time
3416 * weight raising.
3417 *
3418 * The second sub-condition checked in the compound condition is
3419 * whether there is a fair amount of already in-flight I/O not
3420 * belonging to bfqq. If so, I/O dispatching is to be plugged, for the
3421 * following reason. The drive may decide to serve in-flight
3422 * non-bfqq's I/O requests before bfqq's ones, thereby delaying the
3423 * arrival of new I/O requests for bfqq (recall that bfqq is sync). If
3424 * I/O-dispatching is not plugged, then, while bfqq remains empty, a
3425 * basically uncontrolled amount of I/O from other queues may be
3426 * dispatched too, possibly causing the service of bfqq's I/O to be
3427 * delayed even longer in the drive. This problem gets more and more
3428 * serious as the speed and the queue depth of the drive grow,
3429 * because, as these two quantities grow, the probability to find no
3430 * queue busy but many requests in flight grows too. By contrast,
3431 * plugging I/O dispatching minimizes the delay induced by already
3432 * in-flight I/O, and enables bfqq to recover the bandwidth it may
3433 * lose because of this delay.
3434 *
3435 * As a side note, it is worth considering that the above
3436 * device-idling countermeasures may however fail in the following
3437 * unlucky scenario: if I/O-dispatch plugging is (correctly) disabled
3438 * in a time period during which all symmetry sub-conditions hold, and
3439 * therefore the device is allowed to enqueue many requests, but at
3440 * some later point in time some sub-condition stops to hold, then it
3441 * may become impossible to make requests be served in the desired
3442 * order until all the requests already queued in the device have been
3443 * served. The last sub-condition commented above somewhat mitigates
3444 * this problem for weight-raised queues.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3445 */
3446static bool idling_needed_for_service_guarantees(struct bfq_data *bfqd,
3447						 struct bfq_queue *bfqq)
3448{
 
 
3449	/* No point in idling for bfqq if it won't get requests any longer */
3450	if (unlikely(!bfqq_process_refs(bfqq)))
3451		return false;
3452
3453	return (bfqq->wr_coeff > 1 &&
3454		(bfqd->wr_busy_queues <
3455		 bfq_tot_busy_queues(bfqd) ||
3456		 bfqd->rq_in_driver >=
3457		 bfqq->dispatched + 4)) ||
3458		bfq_asymmetric_scenario(bfqd, bfqq);
 
3459}
3460
3461static bool __bfq_bfqq_expire(struct bfq_data *bfqd, struct bfq_queue *bfqq,
3462			      enum bfqq_expiration reason)
3463{
3464	/*
3465	 * If this bfqq is shared between multiple processes, check
3466	 * to make sure that those processes are still issuing I/Os
3467	 * within the mean seek distance. If not, it may be time to
3468	 * break the queues apart again.
3469	 */
3470	if (bfq_bfqq_coop(bfqq) && BFQQ_SEEKY(bfqq))
3471		bfq_mark_bfqq_split_coop(bfqq);
3472
3473	/*
3474	 * Consider queues with a higher finish virtual time than
3475	 * bfqq. If idling_needed_for_service_guarantees(bfqq) returns
3476	 * true, then bfqq's bandwidth would be violated if an
3477	 * uncontrolled amount of I/O from these queues were
3478	 * dispatched while bfqq is waiting for its new I/O to
3479	 * arrive. This is exactly what may happen if this is a forced
3480	 * expiration caused by a preemption attempt, and if bfqq is
3481	 * not re-scheduled. To prevent this from happening, re-queue
3482	 * bfqq if it needs I/O-dispatch plugging, even if it is
3483	 * empty. By doing so, bfqq is granted to be served before the
3484	 * above queues (provided that bfqq is of course eligible).
3485	 */
3486	if (RB_EMPTY_ROOT(&bfqq->sort_list) &&
3487	    !(reason == BFQQE_PREEMPTED &&
3488	      idling_needed_for_service_guarantees(bfqd, bfqq))) {
3489		if (bfqq->dispatched == 0)
3490			/*
3491			 * Overloading budget_timeout field to store
3492			 * the time at which the queue remains with no
3493			 * backlog and no outstanding request; used by
3494			 * the weight-raising mechanism.
3495			 */
3496			bfqq->budget_timeout = jiffies;
3497
3498		bfq_del_bfqq_busy(bfqd, bfqq, true);
3499	} else {
3500		bfq_requeue_bfqq(bfqd, bfqq, true);
3501		/*
3502		 * Resort priority tree of potential close cooperators.
3503		 * See comments on bfq_pos_tree_add_move() for the unlikely().
3504		 */
3505		if (unlikely(!bfqd->nonrot_with_queueing &&
3506			     !RB_EMPTY_ROOT(&bfqq->sort_list)))
3507			bfq_pos_tree_add_move(bfqd, bfqq);
3508	}
3509
3510	/*
3511	 * All in-service entities must have been properly deactivated
3512	 * or requeued before executing the next function, which
3513	 * resets all in-service entities as no more in service. This
3514	 * may cause bfqq to be freed. If this happens, the next
3515	 * function returns true.
3516	 */
3517	return __bfq_bfqd_reset_in_service(bfqd);
3518}
3519
3520/**
3521 * __bfq_bfqq_recalc_budget - try to adapt the budget to the @bfqq behavior.
3522 * @bfqd: device data.
3523 * @bfqq: queue to update.
3524 * @reason: reason for expiration.
3525 *
3526 * Handle the feedback on @bfqq budget at queue expiration.
3527 * See the body for detailed comments.
3528 */
3529static void __bfq_bfqq_recalc_budget(struct bfq_data *bfqd,
3530				     struct bfq_queue *bfqq,
3531				     enum bfqq_expiration reason)
3532{
3533	struct request *next_rq;
3534	int budget, min_budget;
3535
3536	min_budget = bfq_min_budget(bfqd);
3537
3538	if (bfqq->wr_coeff == 1)
3539		budget = bfqq->max_budget;
3540	else /*
3541	      * Use a constant, low budget for weight-raised queues,
3542	      * to help achieve a low latency. Keep it slightly higher
3543	      * than the minimum possible budget, to cause a little
3544	      * bit fewer expirations.
3545	      */
3546		budget = 2 * min_budget;
3547
3548	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: last budg %d, budg left %d",
3549		bfqq->entity.budget, bfq_bfqq_budget_left(bfqq));
3550	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: last max_budg %d, min budg %d",
3551		budget, bfq_min_budget(bfqd));
3552	bfq_log_bfqq(bfqd, bfqq, "recalc_budg: sync %d, seeky %d",
3553		bfq_bfqq_sync(bfqq), BFQQ_SEEKY(bfqd->in_service_queue));
3554
3555	if (bfq_bfqq_sync(bfqq) && bfqq->wr_coeff == 1) {
3556		switch (reason) {
3557		/*
3558		 * Caveat: in all the following cases we trade latency
3559		 * for throughput.
3560		 */
3561		case BFQQE_TOO_IDLE:
3562			/*
3563			 * This is the only case where we may reduce
3564			 * the budget: if there is no request of the
3565			 * process still waiting for completion, then
3566			 * we assume (tentatively) that the timer has
3567			 * expired because the batch of requests of
3568			 * the process could have been served with a
3569			 * smaller budget.  Hence, betting that
3570			 * process will behave in the same way when it
3571			 * becomes backlogged again, we reduce its
3572			 * next budget.  As long as we guess right,
3573			 * this budget cut reduces the latency
3574			 * experienced by the process.
3575			 *
3576			 * However, if there are still outstanding
3577			 * requests, then the process may have not yet
3578			 * issued its next request just because it is
3579			 * still waiting for the completion of some of
3580			 * the still outstanding ones.  So in this
3581			 * subcase we do not reduce its budget, on the
3582			 * contrary we increase it to possibly boost
3583			 * the throughput, as discussed in the
3584			 * comments to the BUDGET_TIMEOUT case.
3585			 */
3586			if (bfqq->dispatched > 0) /* still outstanding reqs */
3587				budget = min(budget * 2, bfqd->bfq_max_budget);
3588			else {
3589				if (budget > 5 * min_budget)
3590					budget -= 4 * min_budget;
3591				else
3592					budget = min_budget;
3593			}
3594			break;
3595		case BFQQE_BUDGET_TIMEOUT:
3596			/*
3597			 * We double the budget here because it gives
3598			 * the chance to boost the throughput if this
3599			 * is not a seeky process (and has bumped into
3600			 * this timeout because of, e.g., ZBR).
3601			 */
3602			budget = min(budget * 2, bfqd->bfq_max_budget);
3603			break;
3604		case BFQQE_BUDGET_EXHAUSTED:
3605			/*
3606			 * The process still has backlog, and did not
3607			 * let either the budget timeout or the disk
3608			 * idling timeout expire. Hence it is not
3609			 * seeky, has a short thinktime and may be
3610			 * happy with a higher budget too. So
3611			 * definitely increase the budget of this good
3612			 * candidate to boost the disk throughput.
3613			 */
3614			budget = min(budget * 4, bfqd->bfq_max_budget);
3615			break;
3616		case BFQQE_NO_MORE_REQUESTS:
3617			/*
3618			 * For queues that expire for this reason, it
3619			 * is particularly important to keep the
3620			 * budget close to the actual service they
3621			 * need. Doing so reduces the timestamp
3622			 * misalignment problem described in the
3623			 * comments in the body of
3624			 * __bfq_activate_entity. In fact, suppose
3625			 * that a queue systematically expires for
3626			 * BFQQE_NO_MORE_REQUESTS and presents a
3627			 * new request in time to enjoy timestamp
3628			 * back-shifting. The larger the budget of the
3629			 * queue is with respect to the service the
3630			 * queue actually requests in each service
3631			 * slot, the more times the queue can be
3632			 * reactivated with the same virtual finish
3633			 * time. It follows that, even if this finish
3634			 * time is pushed to the system virtual time
3635			 * to reduce the consequent timestamp
3636			 * misalignment, the queue unjustly enjoys for
3637			 * many re-activations a lower finish time
3638			 * than all newly activated queues.
3639			 *
3640			 * The service needed by bfqq is measured
3641			 * quite precisely by bfqq->entity.service.
3642			 * Since bfqq does not enjoy device idling,
3643			 * bfqq->entity.service is equal to the number
3644			 * of sectors that the process associated with
3645			 * bfqq requested to read/write before waiting
3646			 * for request completions, or blocking for
3647			 * other reasons.
3648			 */
3649			budget = max_t(int, bfqq->entity.service, min_budget);
3650			break;
3651		default:
3652			return;
3653		}
3654	} else if (!bfq_bfqq_sync(bfqq)) {
3655		/*
3656		 * Async queues get always the maximum possible
3657		 * budget, as for them we do not care about latency
3658		 * (in addition, their ability to dispatch is limited
3659		 * by the charging factor).
3660		 */
3661		budget = bfqd->bfq_max_budget;
3662	}
3663
3664	bfqq->max_budget = budget;
3665
3666	if (bfqd->budgets_assigned >= bfq_stats_min_budgets &&
3667	    !bfqd->bfq_user_max_budget)
3668		bfqq->max_budget = min(bfqq->max_budget, bfqd->bfq_max_budget);
3669
3670	/*
3671	 * If there is still backlog, then assign a new budget, making
3672	 * sure that it is large enough for the next request.  Since
3673	 * the finish time of bfqq must be kept in sync with the
3674	 * budget, be sure to call __bfq_bfqq_expire() *after* this
3675	 * update.
3676	 *
3677	 * If there is no backlog, then no need to update the budget;
3678	 * it will be updated on the arrival of a new request.
3679	 */
3680	next_rq = bfqq->next_rq;
3681	if (next_rq)
3682		bfqq->entity.budget = max_t(unsigned long, bfqq->max_budget,
3683					    bfq_serv_to_charge(next_rq, bfqq));
3684
3685	bfq_log_bfqq(bfqd, bfqq, "head sect: %u, new budget %d",
3686			next_rq ? blk_rq_sectors(next_rq) : 0,
3687			bfqq->entity.budget);
3688}
3689
3690/*
3691 * Return true if the process associated with bfqq is "slow". The slow
3692 * flag is used, in addition to the budget timeout, to reduce the
3693 * amount of service provided to seeky processes, and thus reduce
3694 * their chances to lower the throughput. More details in the comments
3695 * on the function bfq_bfqq_expire().
3696 *
3697 * An important observation is in order: as discussed in the comments
3698 * on the function bfq_update_peak_rate(), with devices with internal
3699 * queues, it is hard if ever possible to know when and for how long
3700 * an I/O request is processed by the device (apart from the trivial
3701 * I/O pattern where a new request is dispatched only after the
3702 * previous one has been completed). This makes it hard to evaluate
3703 * the real rate at which the I/O requests of each bfq_queue are
3704 * served.  In fact, for an I/O scheduler like BFQ, serving a
3705 * bfq_queue means just dispatching its requests during its service
3706 * slot (i.e., until the budget of the queue is exhausted, or the
3707 * queue remains idle, or, finally, a timeout fires). But, during the
3708 * service slot of a bfq_queue, around 100 ms at most, the device may
3709 * be even still processing requests of bfq_queues served in previous
3710 * service slots. On the opposite end, the requests of the in-service
3711 * bfq_queue may be completed after the service slot of the queue
3712 * finishes.
3713 *
3714 * Anyway, unless more sophisticated solutions are used
3715 * (where possible), the sum of the sizes of the requests dispatched
3716 * during the service slot of a bfq_queue is probably the only
3717 * approximation available for the service received by the bfq_queue
3718 * during its service slot. And this sum is the quantity used in this
3719 * function to evaluate the I/O speed of a process.
3720 */
3721static bool bfq_bfqq_is_slow(struct bfq_data *bfqd, struct bfq_queue *bfqq,
3722				 bool compensate, enum bfqq_expiration reason,
3723				 unsigned long *delta_ms)
3724{
3725	ktime_t delta_ktime;
3726	u32 delta_usecs;
3727	bool slow = BFQQ_SEEKY(bfqq); /* if delta too short, use seekyness */
3728
3729	if (!bfq_bfqq_sync(bfqq))
3730		return false;
3731
3732	if (compensate)
3733		delta_ktime = bfqd->last_idling_start;
3734	else
3735		delta_ktime = ktime_get();
3736	delta_ktime = ktime_sub(delta_ktime, bfqd->last_budget_start);
3737	delta_usecs = ktime_to_us(delta_ktime);
3738
3739	/* don't use too short time intervals */
3740	if (delta_usecs < 1000) {
3741		if (blk_queue_nonrot(bfqd->queue))
3742			 /*
3743			  * give same worst-case guarantees as idling
3744			  * for seeky
3745			  */
3746			*delta_ms = BFQ_MIN_TT / NSEC_PER_MSEC;
3747		else /* charge at least one seek */
3748			*delta_ms = bfq_slice_idle / NSEC_PER_MSEC;
3749
3750		return slow;
3751	}
3752
3753	*delta_ms = delta_usecs / USEC_PER_MSEC;
3754
3755	/*
3756	 * Use only long (> 20ms) intervals to filter out excessive
3757	 * spikes in service rate estimation.
3758	 */
3759	if (delta_usecs > 20000) {
3760		/*
3761		 * Caveat for rotational devices: processes doing I/O
3762		 * in the slower disk zones tend to be slow(er) even
3763		 * if not seeky. In this respect, the estimated peak
3764		 * rate is likely to be an average over the disk
3765		 * surface. Accordingly, to not be too harsh with
3766		 * unlucky processes, a process is deemed slow only if
3767		 * its rate has been lower than half of the estimated
3768		 * peak rate.
3769		 */
3770		slow = bfqq->entity.service < bfqd->bfq_max_budget / 2;
3771	}
3772
3773	bfq_log_bfqq(bfqd, bfqq, "bfq_bfqq_is_slow: slow %d", slow);
3774
3775	return slow;
3776}
3777
3778/*
3779 * To be deemed as soft real-time, an application must meet two
3780 * requirements. First, the application must not require an average
3781 * bandwidth higher than the approximate bandwidth required to playback or
3782 * record a compressed high-definition video.
3783 * The next function is invoked on the completion of the last request of a
3784 * batch, to compute the next-start time instant, soft_rt_next_start, such
3785 * that, if the next request of the application does not arrive before
3786 * soft_rt_next_start, then the above requirement on the bandwidth is met.
3787 *
3788 * The second requirement is that the request pattern of the application is
3789 * isochronous, i.e., that, after issuing a request or a batch of requests,
3790 * the application stops issuing new requests until all its pending requests
3791 * have been completed. After that, the application may issue a new batch,
3792 * and so on.
3793 * For this reason the next function is invoked to compute
3794 * soft_rt_next_start only for applications that meet this requirement,
3795 * whereas soft_rt_next_start is set to infinity for applications that do
3796 * not.
3797 *
3798 * Unfortunately, even a greedy (i.e., I/O-bound) application may
3799 * happen to meet, occasionally or systematically, both the above
3800 * bandwidth and isochrony requirements. This may happen at least in
3801 * the following circumstances. First, if the CPU load is high. The
3802 * application may stop issuing requests while the CPUs are busy
3803 * serving other processes, then restart, then stop again for a while,
3804 * and so on. The other circumstances are related to the storage
3805 * device: the storage device is highly loaded or reaches a low-enough
3806 * throughput with the I/O of the application (e.g., because the I/O
3807 * is random and/or the device is slow). In all these cases, the
3808 * I/O of the application may be simply slowed down enough to meet
3809 * the bandwidth and isochrony requirements. To reduce the probability
3810 * that greedy applications are deemed as soft real-time in these
3811 * corner cases, a further rule is used in the computation of
3812 * soft_rt_next_start: the return value of this function is forced to
3813 * be higher than the maximum between the following two quantities.
3814 *
3815 * (a) Current time plus: (1) the maximum time for which the arrival
3816 *     of a request is waited for when a sync queue becomes idle,
3817 *     namely bfqd->bfq_slice_idle, and (2) a few extra jiffies. We
3818 *     postpone for a moment the reason for adding a few extra
3819 *     jiffies; we get back to it after next item (b).  Lower-bounding
3820 *     the return value of this function with the current time plus
3821 *     bfqd->bfq_slice_idle tends to filter out greedy applications,
3822 *     because the latter issue their next request as soon as possible
3823 *     after the last one has been completed. In contrast, a soft
3824 *     real-time application spends some time processing data, after a
3825 *     batch of its requests has been completed.
3826 *
3827 * (b) Current value of bfqq->soft_rt_next_start. As pointed out
3828 *     above, greedy applications may happen to meet both the
3829 *     bandwidth and isochrony requirements under heavy CPU or
3830 *     storage-device load. In more detail, in these scenarios, these
3831 *     applications happen, only for limited time periods, to do I/O
3832 *     slowly enough to meet all the requirements described so far,
3833 *     including the filtering in above item (a). These slow-speed
3834 *     time intervals are usually interspersed between other time
3835 *     intervals during which these applications do I/O at a very high
3836 *     speed. Fortunately, exactly because of the high speed of the
3837 *     I/O in the high-speed intervals, the values returned by this
3838 *     function happen to be so high, near the end of any such
3839 *     high-speed interval, to be likely to fall *after* the end of
3840 *     the low-speed time interval that follows. These high values are
3841 *     stored in bfqq->soft_rt_next_start after each invocation of
3842 *     this function. As a consequence, if the last value of
3843 *     bfqq->soft_rt_next_start is constantly used to lower-bound the
3844 *     next value that this function may return, then, from the very
3845 *     beginning of a low-speed interval, bfqq->soft_rt_next_start is
3846 *     likely to be constantly kept so high that any I/O request
3847 *     issued during the low-speed interval is considered as arriving
3848 *     to soon for the application to be deemed as soft
3849 *     real-time. Then, in the high-speed interval that follows, the
3850 *     application will not be deemed as soft real-time, just because
3851 *     it will do I/O at a high speed. And so on.
3852 *
3853 * Getting back to the filtering in item (a), in the following two
3854 * cases this filtering might be easily passed by a greedy
3855 * application, if the reference quantity was just
3856 * bfqd->bfq_slice_idle:
3857 * 1) HZ is so low that the duration of a jiffy is comparable to or
3858 *    higher than bfqd->bfq_slice_idle. This happens, e.g., on slow
3859 *    devices with HZ=100. The time granularity may be so coarse
3860 *    that the approximation, in jiffies, of bfqd->bfq_slice_idle
3861 *    is rather lower than the exact value.
3862 * 2) jiffies, instead of increasing at a constant rate, may stop increasing
3863 *    for a while, then suddenly 'jump' by several units to recover the lost
3864 *    increments. This seems to happen, e.g., inside virtual machines.
3865 * To address this issue, in the filtering in (a) we do not use as a
3866 * reference time interval just bfqd->bfq_slice_idle, but
3867 * bfqd->bfq_slice_idle plus a few jiffies. In particular, we add the
3868 * minimum number of jiffies for which the filter seems to be quite
3869 * precise also in embedded systems and KVM/QEMU virtual machines.
3870 */
3871static unsigned long bfq_bfqq_softrt_next_start(struct bfq_data *bfqd,
3872						struct bfq_queue *bfqq)
3873{
3874	return max3(bfqq->soft_rt_next_start,
3875		    bfqq->last_idle_bklogged +
3876		    HZ * bfqq->service_from_backlogged /
3877		    bfqd->bfq_wr_max_softrt_rate,
3878		    jiffies + nsecs_to_jiffies(bfqq->bfqd->bfq_slice_idle) + 4);
3879}
3880
3881/**
3882 * bfq_bfqq_expire - expire a queue.
3883 * @bfqd: device owning the queue.
3884 * @bfqq: the queue to expire.
3885 * @compensate: if true, compensate for the time spent idling.
3886 * @reason: the reason causing the expiration.
3887 *
3888 * If the process associated with bfqq does slow I/O (e.g., because it
3889 * issues random requests), we charge bfqq with the time it has been
3890 * in service instead of the service it has received (see
3891 * bfq_bfqq_charge_time for details on how this goal is achieved). As
3892 * a consequence, bfqq will typically get higher timestamps upon
3893 * reactivation, and hence it will be rescheduled as if it had
3894 * received more service than what it has actually received. In the
3895 * end, bfqq receives less service in proportion to how slowly its
3896 * associated process consumes its budgets (and hence how seriously it
3897 * tends to lower the throughput). In addition, this time-charging
3898 * strategy guarantees time fairness among slow processes. In
3899 * contrast, if the process associated with bfqq is not slow, we
3900 * charge bfqq exactly with the service it has received.
3901 *
3902 * Charging time to the first type of queues and the exact service to
3903 * the other has the effect of using the WF2Q+ policy to schedule the
3904 * former on a timeslice basis, without violating service domain
3905 * guarantees among the latter.
3906 */
3907void bfq_bfqq_expire(struct bfq_data *bfqd,
3908		     struct bfq_queue *bfqq,
3909		     bool compensate,
3910		     enum bfqq_expiration reason)
3911{
3912	bool slow;
3913	unsigned long delta = 0;
3914	struct bfq_entity *entity = &bfqq->entity;
3915
3916	/*
3917	 * Check whether the process is slow (see bfq_bfqq_is_slow).
3918	 */
3919	slow = bfq_bfqq_is_slow(bfqd, bfqq, compensate, reason, &delta);
3920
3921	/*
3922	 * As above explained, charge slow (typically seeky) and
3923	 * timed-out queues with the time and not the service
3924	 * received, to favor sequential workloads.
3925	 *
3926	 * Processes doing I/O in the slower disk zones will tend to
3927	 * be slow(er) even if not seeky. Therefore, since the
3928	 * estimated peak rate is actually an average over the disk
3929	 * surface, these processes may timeout just for bad luck. To
3930	 * avoid punishing them, do not charge time to processes that
3931	 * succeeded in consuming at least 2/3 of their budget. This
3932	 * allows BFQ to preserve enough elasticity to still perform
3933	 * bandwidth, and not time, distribution with little unlucky
3934	 * or quasi-sequential processes.
3935	 */
3936	if (bfqq->wr_coeff == 1 &&
3937	    (slow ||
3938	     (reason == BFQQE_BUDGET_TIMEOUT &&
3939	      bfq_bfqq_budget_left(bfqq) >=  entity->budget / 3)))
3940		bfq_bfqq_charge_time(bfqd, bfqq, delta);
3941
3942	if (reason == BFQQE_TOO_IDLE &&
3943	    entity->service <= 2 * entity->budget / 10)
3944		bfq_clear_bfqq_IO_bound(bfqq);
3945
3946	if (bfqd->low_latency && bfqq->wr_coeff == 1)
3947		bfqq->last_wr_start_finish = jiffies;
3948
3949	if (bfqd->low_latency && bfqd->bfq_wr_max_softrt_rate > 0 &&
3950	    RB_EMPTY_ROOT(&bfqq->sort_list)) {
3951		/*
3952		 * If we get here, and there are no outstanding
3953		 * requests, then the request pattern is isochronous
3954		 * (see the comments on the function
3955		 * bfq_bfqq_softrt_next_start()). Thus we can compute
3956		 * soft_rt_next_start. And we do it, unless bfqq is in
3957		 * interactive weight raising. We do not do it in the
3958		 * latter subcase, for the following reason. bfqq may
3959		 * be conveying the I/O needed to load a soft
3960		 * real-time application. Such an application will
3961		 * actually exhibit a soft real-time I/O pattern after
3962		 * it finally starts doing its job. But, if
3963		 * soft_rt_next_start is computed here for an
3964		 * interactive bfqq, and bfqq had received a lot of
3965		 * service before remaining with no outstanding
3966		 * request (likely to happen on a fast device), then
3967		 * soft_rt_next_start would be assigned such a high
3968		 * value that, for a very long time, bfqq would be
3969		 * prevented from being possibly considered as soft
3970		 * real time.
3971		 *
3972		 * If, instead, the queue still has outstanding
3973		 * requests, then we have to wait for the completion
3974		 * of all the outstanding requests to discover whether
3975		 * the request pattern is actually isochronous.
3976		 */
3977		if (bfqq->dispatched == 0 &&
3978		    bfqq->wr_coeff != bfqd->bfq_wr_coeff)
3979			bfqq->soft_rt_next_start =
3980				bfq_bfqq_softrt_next_start(bfqd, bfqq);
3981		else if (bfqq->dispatched > 0) {
3982			/*
3983			 * Schedule an update of soft_rt_next_start to when
3984			 * the task may be discovered to be isochronous.
3985			 */
3986			bfq_mark_bfqq_softrt_update(bfqq);
3987		}
3988	}
3989
3990	bfq_log_bfqq(bfqd, bfqq,
3991		"expire (%d, slow %d, num_disp %d, short_ttime %d)", reason,
3992		slow, bfqq->dispatched, bfq_bfqq_has_short_ttime(bfqq));
3993
3994	/*
3995	 * bfqq expired, so no total service time needs to be computed
3996	 * any longer: reset state machine for measuring total service
3997	 * times.
3998	 */
3999	bfqd->rqs_injected = bfqd->wait_dispatch = false;
4000	bfqd->waited_rq = NULL;
4001
4002	/*
4003	 * Increase, decrease or leave budget unchanged according to
4004	 * reason.
4005	 */
4006	__bfq_bfqq_recalc_budget(bfqd, bfqq, reason);
4007	if (__bfq_bfqq_expire(bfqd, bfqq, reason))
4008		/* bfqq is gone, no more actions on it */
4009		return;
4010
4011	/* mark bfqq as waiting a request only if a bic still points to it */
4012	if (!bfq_bfqq_busy(bfqq) &&
4013	    reason != BFQQE_BUDGET_TIMEOUT &&
4014	    reason != BFQQE_BUDGET_EXHAUSTED) {
4015		bfq_mark_bfqq_non_blocking_wait_rq(bfqq);
4016		/*
4017		 * Not setting service to 0, because, if the next rq
4018		 * arrives in time, the queue will go on receiving
4019		 * service with this same budget (as if it never expired)
4020		 */
4021	} else
4022		entity->service = 0;
4023
4024	/*
4025	 * Reset the received-service counter for every parent entity.
4026	 * Differently from what happens with bfqq->entity.service,
4027	 * the resetting of this counter never needs to be postponed
4028	 * for parent entities. In fact, in case bfqq may have a
4029	 * chance to go on being served using the last, partially
4030	 * consumed budget, bfqq->entity.service needs to be kept,
4031	 * because if bfqq then actually goes on being served using
4032	 * the same budget, the last value of bfqq->entity.service is
4033	 * needed to properly decrement bfqq->entity.budget by the
4034	 * portion already consumed. In contrast, it is not necessary
4035	 * to keep entity->service for parent entities too, because
4036	 * the bubble up of the new value of bfqq->entity.budget will
4037	 * make sure that the budgets of parent entities are correct,
4038	 * even in case bfqq and thus parent entities go on receiving
4039	 * service with the same budget.
4040	 */
4041	entity = entity->parent;
4042	for_each_entity(entity)
4043		entity->service = 0;
4044}
4045
4046/*
4047 * Budget timeout is not implemented through a dedicated timer, but
4048 * just checked on request arrivals and completions, as well as on
4049 * idle timer expirations.
4050 */
4051static bool bfq_bfqq_budget_timeout(struct bfq_queue *bfqq)
4052{
4053	return time_is_before_eq_jiffies(bfqq->budget_timeout);
4054}
4055
4056/*
4057 * If we expire a queue that is actively waiting (i.e., with the
4058 * device idled) for the arrival of a new request, then we may incur
4059 * the timestamp misalignment problem described in the body of the
4060 * function __bfq_activate_entity. Hence we return true only if this
4061 * condition does not hold, or if the queue is slow enough to deserve
4062 * only to be kicked off for preserving a high throughput.
4063 */
4064static bool bfq_may_expire_for_budg_timeout(struct bfq_queue *bfqq)
4065{
4066	bfq_log_bfqq(bfqq->bfqd, bfqq,
4067		"may_budget_timeout: wait_request %d left %d timeout %d",
4068		bfq_bfqq_wait_request(bfqq),
4069			bfq_bfqq_budget_left(bfqq) >=  bfqq->entity.budget / 3,
4070		bfq_bfqq_budget_timeout(bfqq));
4071
4072	return (!bfq_bfqq_wait_request(bfqq) ||
4073		bfq_bfqq_budget_left(bfqq) >=  bfqq->entity.budget / 3)
4074		&&
4075		bfq_bfqq_budget_timeout(bfqq);
4076}
4077
4078static bool idling_boosts_thr_without_issues(struct bfq_data *bfqd,
4079					     struct bfq_queue *bfqq)
4080{
4081	bool rot_without_queueing =
4082		!blk_queue_nonrot(bfqd->queue) && !bfqd->hw_tag,
4083		bfqq_sequential_and_IO_bound,
4084		idling_boosts_thr;
4085
4086	/* No point in idling for bfqq if it won't get requests any longer */
4087	if (unlikely(!bfqq_process_refs(bfqq)))
4088		return false;
4089
4090	bfqq_sequential_and_IO_bound = !BFQQ_SEEKY(bfqq) &&
4091		bfq_bfqq_IO_bound(bfqq) && bfq_bfqq_has_short_ttime(bfqq);
4092
4093	/*
4094	 * The next variable takes into account the cases where idling
4095	 * boosts the throughput.
4096	 *
4097	 * The value of the variable is computed considering, first, that
4098	 * idling is virtually always beneficial for the throughput if:
4099	 * (a) the device is not NCQ-capable and rotational, or
4100	 * (b) regardless of the presence of NCQ, the device is rotational and
4101	 *     the request pattern for bfqq is I/O-bound and sequential, or
4102	 * (c) regardless of whether it is rotational, the device is
4103	 *     not NCQ-capable and the request pattern for bfqq is
4104	 *     I/O-bound and sequential.
4105	 *
4106	 * Secondly, and in contrast to the above item (b), idling an
4107	 * NCQ-capable flash-based device would not boost the
4108	 * throughput even with sequential I/O; rather it would lower
4109	 * the throughput in proportion to how fast the device
4110	 * is. Accordingly, the next variable is true if any of the
4111	 * above conditions (a), (b) or (c) is true, and, in
4112	 * particular, happens to be false if bfqd is an NCQ-capable
4113	 * flash-based device.
4114	 */
4115	idling_boosts_thr = rot_without_queueing ||
4116		((!blk_queue_nonrot(bfqd->queue) || !bfqd->hw_tag) &&
4117		 bfqq_sequential_and_IO_bound);
4118
4119	/*
4120	 * The return value of this function is equal to that of
4121	 * idling_boosts_thr, unless a special case holds. In this
4122	 * special case, described below, idling may cause problems to
4123	 * weight-raised queues.
4124	 *
4125	 * When the request pool is saturated (e.g., in the presence
4126	 * of write hogs), if the processes associated with
4127	 * non-weight-raised queues ask for requests at a lower rate,
4128	 * then processes associated with weight-raised queues have a
4129	 * higher probability to get a request from the pool
4130	 * immediately (or at least soon) when they need one. Thus
4131	 * they have a higher probability to actually get a fraction
4132	 * of the device throughput proportional to their high
4133	 * weight. This is especially true with NCQ-capable drives,
4134	 * which enqueue several requests in advance, and further
4135	 * reorder internally-queued requests.
4136	 *
4137	 * For this reason, we force to false the return value if
4138	 * there are weight-raised busy queues. In this case, and if
4139	 * bfqq is not weight-raised, this guarantees that the device
4140	 * is not idled for bfqq (if, instead, bfqq is weight-raised,
4141	 * then idling will be guaranteed by another variable, see
4142	 * below). Combined with the timestamping rules of BFQ (see
4143	 * [1] for details), this behavior causes bfqq, and hence any
4144	 * sync non-weight-raised queue, to get a lower number of
4145	 * requests served, and thus to ask for a lower number of
4146	 * requests from the request pool, before the busy
4147	 * weight-raised queues get served again. This often mitigates
4148	 * starvation problems in the presence of heavy write
4149	 * workloads and NCQ, thereby guaranteeing a higher
4150	 * application and system responsiveness in these hostile
4151	 * scenarios.
4152	 */
4153	return idling_boosts_thr &&
4154		bfqd->wr_busy_queues == 0;
4155}
4156
4157/*
4158 * For a queue that becomes empty, device idling is allowed only if
4159 * this function returns true for that queue. As a consequence, since
4160 * device idling plays a critical role for both throughput boosting
4161 * and service guarantees, the return value of this function plays a
4162 * critical role as well.
4163 *
4164 * In a nutshell, this function returns true only if idling is
4165 * beneficial for throughput or, even if detrimental for throughput,
4166 * idling is however necessary to preserve service guarantees (low
4167 * latency, desired throughput distribution, ...). In particular, on
4168 * NCQ-capable devices, this function tries to return false, so as to
4169 * help keep the drives' internal queues full, whenever this helps the
4170 * device boost the throughput without causing any service-guarantee
4171 * issue.
4172 *
4173 * Most of the issues taken into account to get the return value of
4174 * this function are not trivial. We discuss these issues in the two
4175 * functions providing the main pieces of information needed by this
4176 * function.
4177 */
4178static bool bfq_better_to_idle(struct bfq_queue *bfqq)
4179{
4180	struct bfq_data *bfqd = bfqq->bfqd;
4181	bool idling_boosts_thr_with_no_issue, idling_needed_for_service_guar;
4182
4183	/* No point in idling for bfqq if it won't get requests any longer */
4184	if (unlikely(!bfqq_process_refs(bfqq)))
4185		return false;
4186
4187	if (unlikely(bfqd->strict_guarantees))
4188		return true;
4189
4190	/*
4191	 * Idling is performed only if slice_idle > 0. In addition, we
4192	 * do not idle if
4193	 * (a) bfqq is async
4194	 * (b) bfqq is in the idle io prio class: in this case we do
4195	 * not idle because we want to minimize the bandwidth that
4196	 * queues in this class can steal to higher-priority queues
4197	 */
4198	if (bfqd->bfq_slice_idle == 0 || !bfq_bfqq_sync(bfqq) ||
4199	   bfq_class_idle(bfqq))
4200		return false;
4201
4202	idling_boosts_thr_with_no_issue =
4203		idling_boosts_thr_without_issues(bfqd, bfqq);
4204
4205	idling_needed_for_service_guar =
4206		idling_needed_for_service_guarantees(bfqd, bfqq);
4207
4208	/*
4209	 * We have now the two components we need to compute the
4210	 * return value of the function, which is true only if idling
4211	 * either boosts the throughput (without issues), or is
4212	 * necessary to preserve service guarantees.
4213	 */
4214	return idling_boosts_thr_with_no_issue ||
4215		idling_needed_for_service_guar;
4216}
4217
4218/*
4219 * If the in-service queue is empty but the function bfq_better_to_idle
4220 * returns true, then:
4221 * 1) the queue must remain in service and cannot be expired, and
4222 * 2) the device must be idled to wait for the possible arrival of a new
4223 *    request for the queue.
4224 * See the comments on the function bfq_better_to_idle for the reasons
4225 * why performing device idling is the best choice to boost the throughput
4226 * and preserve service guarantees when bfq_better_to_idle itself
4227 * returns true.
4228 */
4229static bool bfq_bfqq_must_idle(struct bfq_queue *bfqq)
4230{
4231	return RB_EMPTY_ROOT(&bfqq->sort_list) && bfq_better_to_idle(bfqq);
4232}
4233
4234/*
4235 * This function chooses the queue from which to pick the next extra
4236 * I/O request to inject, if it finds a compatible queue. See the
4237 * comments on bfq_update_inject_limit() for details on the injection
4238 * mechanism, and for the definitions of the quantities mentioned
4239 * below.
4240 */
4241static struct bfq_queue *
4242bfq_choose_bfqq_for_injection(struct bfq_data *bfqd)
4243{
4244	struct bfq_queue *bfqq, *in_serv_bfqq = bfqd->in_service_queue;
4245	unsigned int limit = in_serv_bfqq->inject_limit;
4246	/*
4247	 * If
4248	 * - bfqq is not weight-raised and therefore does not carry
4249	 *   time-critical I/O,
4250	 * or
4251	 * - regardless of whether bfqq is weight-raised, bfqq has
4252	 *   however a long think time, during which it can absorb the
4253	 *   effect of an appropriate number of extra I/O requests
4254	 *   from other queues (see bfq_update_inject_limit for
4255	 *   details on the computation of this number);
4256	 * then injection can be performed without restrictions.
4257	 */
4258	bool in_serv_always_inject = in_serv_bfqq->wr_coeff == 1 ||
4259		!bfq_bfqq_has_short_ttime(in_serv_bfqq);
4260
4261	/*
4262	 * If
4263	 * - the baseline total service time could not be sampled yet,
4264	 *   so the inject limit happens to be still 0, and
4265	 * - a lot of time has elapsed since the plugging of I/O
4266	 *   dispatching started, so drive speed is being wasted
4267	 *   significantly;
4268	 * then temporarily raise inject limit to one request.
4269	 */
4270	if (limit == 0 && in_serv_bfqq->last_serv_time_ns == 0 &&
4271	    bfq_bfqq_wait_request(in_serv_bfqq) &&
4272	    time_is_before_eq_jiffies(bfqd->last_idling_start_jiffies +
4273				      bfqd->bfq_slice_idle)
4274		)
4275		limit = 1;
4276
4277	if (bfqd->rq_in_driver >= limit)
4278		return NULL;
4279
4280	/*
4281	 * Linear search of the source queue for injection; but, with
4282	 * a high probability, very few steps are needed to find a
4283	 * candidate queue, i.e., a queue with enough budget left for
4284	 * its next request. In fact:
4285	 * - BFQ dynamically updates the budget of every queue so as
4286	 *   to accommodate the expected backlog of the queue;
4287	 * - if a queue gets all its requests dispatched as injected
4288	 *   service, then the queue is removed from the active list
4289	 *   (and re-added only if it gets new requests, but then it
4290	 *   is assigned again enough budget for its new backlog).
4291	 */
4292	list_for_each_entry(bfqq, &bfqd->active_list, bfqq_list)
4293		if (!RB_EMPTY_ROOT(&bfqq->sort_list) &&
4294		    (in_serv_always_inject || bfqq->wr_coeff > 1) &&
4295		    bfq_serv_to_charge(bfqq->next_rq, bfqq) <=
4296		    bfq_bfqq_budget_left(bfqq)) {
4297			/*
4298			 * Allow for only one large in-flight request
4299			 * on non-rotational devices, for the
4300			 * following reason. On non-rotationl drives,
4301			 * large requests take much longer than
4302			 * smaller requests to be served. In addition,
4303			 * the drive prefers to serve large requests
4304			 * w.r.t. to small ones, if it can choose. So,
4305			 * having more than one large requests queued
4306			 * in the drive may easily make the next first
4307			 * request of the in-service queue wait for so
4308			 * long to break bfqq's service guarantees. On
4309			 * the bright side, large requests let the
4310			 * drive reach a very high throughput, even if
4311			 * there is only one in-flight large request
4312			 * at a time.
4313			 */
4314			if (blk_queue_nonrot(bfqd->queue) &&
4315			    blk_rq_sectors(bfqq->next_rq) >=
4316			    BFQQ_SECT_THR_NONROT)
4317				limit = min_t(unsigned int, 1, limit);
4318			else
4319				limit = in_serv_bfqq->inject_limit;
4320
4321			if (bfqd->rq_in_driver < limit) {
4322				bfqd->rqs_injected = true;
4323				return bfqq;
4324			}
4325		}
4326
4327	return NULL;
4328}
4329
4330/*
4331 * Select a queue for service.  If we have a current queue in service,
4332 * check whether to continue servicing it, or retrieve and set a new one.
4333 */
4334static struct bfq_queue *bfq_select_queue(struct bfq_data *bfqd)
4335{
4336	struct bfq_queue *bfqq;
4337	struct request *next_rq;
4338	enum bfqq_expiration reason = BFQQE_BUDGET_TIMEOUT;
4339
4340	bfqq = bfqd->in_service_queue;
4341	if (!bfqq)
4342		goto new_queue;
4343
4344	bfq_log_bfqq(bfqd, bfqq, "select_queue: already in-service queue");
4345
4346	/*
4347	 * Do not expire bfqq for budget timeout if bfqq may be about
4348	 * to enjoy device idling. The reason why, in this case, we
4349	 * prevent bfqq from expiring is the same as in the comments
4350	 * on the case where bfq_bfqq_must_idle() returns true, in
4351	 * bfq_completed_request().
4352	 */
4353	if (bfq_may_expire_for_budg_timeout(bfqq) &&
4354	    !bfq_bfqq_must_idle(bfqq))
4355		goto expire;
4356
4357check_queue:
4358	/*
4359	 * This loop is rarely executed more than once. Even when it
4360	 * happens, it is much more convenient to re-execute this loop
4361	 * than to return NULL and trigger a new dispatch to get a
4362	 * request served.
4363	 */
4364	next_rq = bfqq->next_rq;
4365	/*
4366	 * If bfqq has requests queued and it has enough budget left to
4367	 * serve them, keep the queue, otherwise expire it.
4368	 */
4369	if (next_rq) {
4370		if (bfq_serv_to_charge(next_rq, bfqq) >
4371			bfq_bfqq_budget_left(bfqq)) {
4372			/*
4373			 * Expire the queue for budget exhaustion,
4374			 * which makes sure that the next budget is
4375			 * enough to serve the next request, even if
4376			 * it comes from the fifo expired path.
4377			 */
4378			reason = BFQQE_BUDGET_EXHAUSTED;
4379			goto expire;
4380		} else {
4381			/*
4382			 * The idle timer may be pending because we may
4383			 * not disable disk idling even when a new request
4384			 * arrives.
4385			 */
4386			if (bfq_bfqq_wait_request(bfqq)) {
4387				/*
4388				 * If we get here: 1) at least a new request
4389				 * has arrived but we have not disabled the
4390				 * timer because the request was too small,
4391				 * 2) then the block layer has unplugged
4392				 * the device, causing the dispatch to be
4393				 * invoked.
4394				 *
4395				 * Since the device is unplugged, now the
4396				 * requests are probably large enough to
4397				 * provide a reasonable throughput.
4398				 * So we disable idling.
4399				 */
4400				bfq_clear_bfqq_wait_request(bfqq);
4401				hrtimer_try_to_cancel(&bfqd->idle_slice_timer);
4402			}
4403			goto keep_queue;
4404		}
4405	}
4406
4407	/*
4408	 * No requests pending. However, if the in-service queue is idling
4409	 * for a new request, or has requests waiting for a completion and
4410	 * may idle after their completion, then keep it anyway.
4411	 *
4412	 * Yet, inject service from other queues if it boosts
4413	 * throughput and is possible.
4414	 */
4415	if (bfq_bfqq_wait_request(bfqq) ||
4416	    (bfqq->dispatched != 0 && bfq_better_to_idle(bfqq))) {
4417		struct bfq_queue *async_bfqq =
4418			bfqq->bic && bfqq->bic->bfqq[0] &&
4419			bfq_bfqq_busy(bfqq->bic->bfqq[0]) &&
4420			bfqq->bic->bfqq[0]->next_rq ?
4421			bfqq->bic->bfqq[0] : NULL;
 
 
 
 
 
 
4422
4423		/*
4424		 * The next three mutually-exclusive ifs decide
4425		 * whether to try injection, and choose the queue to
4426		 * pick an I/O request from.
4427		 *
4428		 * The first if checks whether the process associated
4429		 * with bfqq has also async I/O pending. If so, it
4430		 * injects such I/O unconditionally. Injecting async
4431		 * I/O from the same process can cause no harm to the
4432		 * process. On the contrary, it can only increase
4433		 * bandwidth and reduce latency for the process.
4434		 *
4435		 * The second if checks whether there happens to be a
4436		 * non-empty waker queue for bfqq, i.e., a queue whose
4437		 * I/O needs to be completed for bfqq to receive new
4438		 * I/O. This happens, e.g., if bfqq is associated with
4439		 * a process that does some sync. A sync generates
4440		 * extra blocking I/O, which must be completed before
4441		 * the process associated with bfqq can go on with its
4442		 * I/O. If the I/O of the waker queue is not served,
4443		 * then bfqq remains empty, and no I/O is dispatched,
4444		 * until the idle timeout fires for bfqq. This is
4445		 * likely to result in lower bandwidth and higher
4446		 * latencies for bfqq, and in a severe loss of total
4447		 * throughput. The best action to take is therefore to
4448		 * serve the waker queue as soon as possible. So do it
4449		 * (without relying on the third alternative below for
4450		 * eventually serving waker_bfqq's I/O; see the last
4451		 * paragraph for further details). This systematic
4452		 * injection of I/O from the waker queue does not
4453		 * cause any delay to bfqq's I/O. On the contrary,
4454		 * next bfqq's I/O is brought forward dramatically,
4455		 * for it is not blocked for milliseconds.
4456		 *
4457		 * The third if checks whether bfqq is a queue for
 
 
 
 
 
 
 
 
4458		 * which it is better to avoid injection. It is so if
4459		 * bfqq delivers more throughput when served without
4460		 * any further I/O from other queues in the middle, or
4461		 * if the service times of bfqq's I/O requests both
4462		 * count more than overall throughput, and may be
4463		 * easily increased by injection (this happens if bfqq
4464		 * has a short think time). If none of these
4465		 * conditions holds, then a candidate queue for
4466		 * injection is looked for through
4467		 * bfq_choose_bfqq_for_injection(). Note that the
4468		 * latter may return NULL (for example if the inject
4469		 * limit for bfqq is currently 0).
4470		 *
4471		 * NOTE: motivation for the second alternative
4472		 *
4473		 * Thanks to the way the inject limit is updated in
4474		 * bfq_update_has_short_ttime(), it is rather likely
4475		 * that, if I/O is being plugged for bfqq and the
4476		 * waker queue has pending I/O requests that are
4477		 * blocking bfqq's I/O, then the third alternative
4478		 * above lets the waker queue get served before the
4479		 * I/O-plugging timeout fires. So one may deem the
4480		 * second alternative superfluous. It is not, because
4481		 * the third alternative may be way less effective in
4482		 * case of a synchronization. For two main
4483		 * reasons. First, throughput may be low because the
4484		 * inject limit may be too low to guarantee the same
4485		 * amount of injected I/O, from the waker queue or
4486		 * other queues, that the second alternative
4487		 * guarantees (the second alternative unconditionally
4488		 * injects a pending I/O request of the waker queue
4489		 * for each bfq_dispatch_request()). Second, with the
4490		 * third alternative, the duration of the plugging,
4491		 * i.e., the time before bfqq finally receives new I/O,
4492		 * may not be minimized, because the waker queue may
4493		 * happen to be served only after other queues.
4494		 */
4495		if (async_bfqq &&
4496		    icq_to_bic(async_bfqq->next_rq->elv.icq) == bfqq->bic &&
4497		    bfq_serv_to_charge(async_bfqq->next_rq, async_bfqq) <=
4498		    bfq_bfqq_budget_left(async_bfqq))
4499			bfqq = bfqq->bic->bfqq[0];
4500		else if (bfq_bfqq_has_waker(bfqq) &&
4501			   bfq_bfqq_busy(bfqq->waker_bfqq) &&
4502			   bfqq->next_rq &&
4503			   bfq_serv_to_charge(bfqq->waker_bfqq->next_rq,
4504					      bfqq->waker_bfqq) <=
4505			   bfq_bfqq_budget_left(bfqq->waker_bfqq)
4506			)
4507			bfqq = bfqq->waker_bfqq;
 
 
 
 
 
 
 
 
4508		else if (!idling_boosts_thr_without_issues(bfqd, bfqq) &&
4509			 (bfqq->wr_coeff == 1 || bfqd->wr_busy_queues > 1 ||
4510			  !bfq_bfqq_has_short_ttime(bfqq)))
4511			bfqq = bfq_choose_bfqq_for_injection(bfqd);
4512		else
4513			bfqq = NULL;
4514
4515		goto keep_queue;
4516	}
4517
4518	reason = BFQQE_NO_MORE_REQUESTS;
4519expire:
4520	bfq_bfqq_expire(bfqd, bfqq, false, reason);
4521new_queue:
4522	bfqq = bfq_set_in_service_queue(bfqd);
4523	if (bfqq) {
4524		bfq_log_bfqq(bfqd, bfqq, "select_queue: checking new queue");
4525		goto check_queue;
4526	}
4527keep_queue:
4528	if (bfqq)
4529		bfq_log_bfqq(bfqd, bfqq, "select_queue: returned this queue");
4530	else
4531		bfq_log(bfqd, "select_queue: no queue returned");
4532
4533	return bfqq;
4534}
4535
4536static void bfq_update_wr_data(struct bfq_data *bfqd, struct bfq_queue *bfqq)
4537{
4538	struct bfq_entity *entity = &bfqq->entity;
4539
4540	if (bfqq->wr_coeff > 1) { /* queue is being weight-raised */
4541		bfq_log_bfqq(bfqd, bfqq,
4542			"raising period dur %u/%u msec, old coeff %u, w %d(%d)",
4543			jiffies_to_msecs(jiffies - bfqq->last_wr_start_finish),
4544			jiffies_to_msecs(bfqq->wr_cur_max_time),
4545			bfqq->wr_coeff,
4546			bfqq->entity.weight, bfqq->entity.orig_weight);
4547
4548		if (entity->prio_changed)
4549			bfq_log_bfqq(bfqd, bfqq, "WARN: pending prio change");
4550
4551		/*
4552		 * If the queue was activated in a burst, or too much
4553		 * time has elapsed from the beginning of this
4554		 * weight-raising period, then end weight raising.
4555		 */
4556		if (bfq_bfqq_in_large_burst(bfqq))
4557			bfq_bfqq_end_wr(bfqq);
4558		else if (time_is_before_jiffies(bfqq->last_wr_start_finish +
4559						bfqq->wr_cur_max_time)) {
4560			if (bfqq->wr_cur_max_time != bfqd->bfq_wr_rt_max_time ||
4561			time_is_before_jiffies(bfqq->wr_start_at_switch_to_srt +
4562					       bfq_wr_duration(bfqd)))
 
 
 
 
 
 
 
 
4563				bfq_bfqq_end_wr(bfqq);
4564			else {
 
 
 
 
4565				switch_back_to_interactive_wr(bfqq, bfqd);
4566				bfqq->entity.prio_changed = 1;
4567			}
4568		}
4569		if (bfqq->wr_coeff > 1 &&
4570		    bfqq->wr_cur_max_time != bfqd->bfq_wr_rt_max_time &&
4571		    bfqq->service_from_wr > max_service_from_wr) {
4572			/* see comments on max_service_from_wr */
4573			bfq_bfqq_end_wr(bfqq);
4574		}
4575	}
4576	/*
4577	 * To improve latency (for this or other queues), immediately
4578	 * update weight both if it must be raised and if it must be
4579	 * lowered. Since, entity may be on some active tree here, and
4580	 * might have a pending change of its ioprio class, invoke
4581	 * next function with the last parameter unset (see the
4582	 * comments on the function).
4583	 */
4584	if ((entity->weight > entity->orig_weight) != (bfqq->wr_coeff > 1))
4585		__bfq_entity_update_weight_prio(bfq_entity_service_tree(entity),
4586						entity, false);
4587}
4588
4589/*
4590 * Dispatch next request from bfqq.
4591 */
4592static struct request *bfq_dispatch_rq_from_bfqq(struct bfq_data *bfqd,
4593						 struct bfq_queue *bfqq)
4594{
4595	struct request *rq = bfqq->next_rq;
4596	unsigned long service_to_charge;
4597
4598	service_to_charge = bfq_serv_to_charge(rq, bfqq);
4599
4600	bfq_bfqq_served(bfqq, service_to_charge);
4601
4602	if (bfqq == bfqd->in_service_queue && bfqd->wait_dispatch) {
4603		bfqd->wait_dispatch = false;
4604		bfqd->waited_rq = rq;
4605	}
4606
4607	bfq_dispatch_remove(bfqd->queue, rq);
4608
4609	if (bfqq != bfqd->in_service_queue)
4610		goto return_rq;
4611
4612	/*
4613	 * If weight raising has to terminate for bfqq, then next
4614	 * function causes an immediate update of bfqq's weight,
4615	 * without waiting for next activation. As a consequence, on
4616	 * expiration, bfqq will be timestamped as if has never been
4617	 * weight-raised during this service slot, even if it has
4618	 * received part or even most of the service as a
4619	 * weight-raised queue. This inflates bfqq's timestamps, which
4620	 * is beneficial, as bfqq is then more willing to leave the
4621	 * device immediately to possible other weight-raised queues.
4622	 */
4623	bfq_update_wr_data(bfqd, bfqq);
4624
4625	/*
4626	 * Expire bfqq, pretending that its budget expired, if bfqq
4627	 * belongs to CLASS_IDLE and other queues are waiting for
4628	 * service.
4629	 */
4630	if (!(bfq_tot_busy_queues(bfqd) > 1 && bfq_class_idle(bfqq)))
4631		goto return_rq;
4632
4633	bfq_bfqq_expire(bfqd, bfqq, false, BFQQE_BUDGET_EXHAUSTED);
4634
4635return_rq:
4636	return rq;
4637}
4638
4639static bool bfq_has_work(struct blk_mq_hw_ctx *hctx)
4640{
4641	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
4642
4643	/*
4644	 * Avoiding lock: a race on bfqd->busy_queues should cause at
4645	 * most a call to dispatch for nothing
4646	 */
4647	return !list_empty_careful(&bfqd->dispatch) ||
4648		bfq_tot_busy_queues(bfqd) > 0;
4649}
4650
4651static struct request *__bfq_dispatch_request(struct blk_mq_hw_ctx *hctx)
4652{
4653	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
4654	struct request *rq = NULL;
4655	struct bfq_queue *bfqq = NULL;
4656
4657	if (!list_empty(&bfqd->dispatch)) {
4658		rq = list_first_entry(&bfqd->dispatch, struct request,
4659				      queuelist);
4660		list_del_init(&rq->queuelist);
4661
4662		bfqq = RQ_BFQQ(rq);
4663
4664		if (bfqq) {
4665			/*
4666			 * Increment counters here, because this
4667			 * dispatch does not follow the standard
4668			 * dispatch flow (where counters are
4669			 * incremented)
4670			 */
4671			bfqq->dispatched++;
4672
4673			goto inc_in_driver_start_rq;
4674		}
4675
4676		/*
4677		 * We exploit the bfq_finish_requeue_request hook to
4678		 * decrement rq_in_driver, but
4679		 * bfq_finish_requeue_request will not be invoked on
4680		 * this request. So, to avoid unbalance, just start
4681		 * this request, without incrementing rq_in_driver. As
4682		 * a negative consequence, rq_in_driver is deceptively
4683		 * lower than it should be while this request is in
4684		 * service. This may cause bfq_schedule_dispatch to be
4685		 * invoked uselessly.
4686		 *
4687		 * As for implementing an exact solution, the
4688		 * bfq_finish_requeue_request hook, if defined, is
4689		 * probably invoked also on this request. So, by
4690		 * exploiting this hook, we could 1) increment
4691		 * rq_in_driver here, and 2) decrement it in
4692		 * bfq_finish_requeue_request. Such a solution would
4693		 * let the value of the counter be always accurate,
4694		 * but it would entail using an extra interface
4695		 * function. This cost seems higher than the benefit,
4696		 * being the frequency of non-elevator-private
4697		 * requests very low.
4698		 */
4699		goto start_rq;
4700	}
4701
4702	bfq_log(bfqd, "dispatch requests: %d busy queues",
4703		bfq_tot_busy_queues(bfqd));
4704
4705	if (bfq_tot_busy_queues(bfqd) == 0)
4706		goto exit;
4707
4708	/*
4709	 * Force device to serve one request at a time if
4710	 * strict_guarantees is true. Forcing this service scheme is
4711	 * currently the ONLY way to guarantee that the request
4712	 * service order enforced by the scheduler is respected by a
4713	 * queueing device. Otherwise the device is free even to make
4714	 * some unlucky request wait for as long as the device
4715	 * wishes.
4716	 *
4717	 * Of course, serving one request at a time may cause loss of
4718	 * throughput.
4719	 */
4720	if (bfqd->strict_guarantees && bfqd->rq_in_driver > 0)
4721		goto exit;
4722
4723	bfqq = bfq_select_queue(bfqd);
4724	if (!bfqq)
4725		goto exit;
4726
4727	rq = bfq_dispatch_rq_from_bfqq(bfqd, bfqq);
4728
4729	if (rq) {
4730inc_in_driver_start_rq:
4731		bfqd->rq_in_driver++;
4732start_rq:
4733		rq->rq_flags |= RQF_STARTED;
4734	}
4735exit:
4736	return rq;
4737}
4738
4739#ifdef CONFIG_BFQ_CGROUP_DEBUG
4740static void bfq_update_dispatch_stats(struct request_queue *q,
4741				      struct request *rq,
4742				      struct bfq_queue *in_serv_queue,
4743				      bool idle_timer_disabled)
4744{
4745	struct bfq_queue *bfqq = rq ? RQ_BFQQ(rq) : NULL;
4746
4747	if (!idle_timer_disabled && !bfqq)
4748		return;
4749
4750	/*
4751	 * rq and bfqq are guaranteed to exist until this function
4752	 * ends, for the following reasons. First, rq can be
4753	 * dispatched to the device, and then can be completed and
4754	 * freed, only after this function ends. Second, rq cannot be
4755	 * merged (and thus freed because of a merge) any longer,
4756	 * because it has already started. Thus rq cannot be freed
4757	 * before this function ends, and, since rq has a reference to
4758	 * bfqq, the same guarantee holds for bfqq too.
4759	 *
4760	 * In addition, the following queue lock guarantees that
4761	 * bfqq_group(bfqq) exists as well.
4762	 */
4763	spin_lock_irq(&q->queue_lock);
4764	if (idle_timer_disabled)
4765		/*
4766		 * Since the idle timer has been disabled,
4767		 * in_serv_queue contained some request when
4768		 * __bfq_dispatch_request was invoked above, which
4769		 * implies that rq was picked exactly from
4770		 * in_serv_queue. Thus in_serv_queue == bfqq, and is
4771		 * therefore guaranteed to exist because of the above
4772		 * arguments.
4773		 */
4774		bfqg_stats_update_idle_time(bfqq_group(in_serv_queue));
4775	if (bfqq) {
4776		struct bfq_group *bfqg = bfqq_group(bfqq);
4777
4778		bfqg_stats_update_avg_queue_size(bfqg);
4779		bfqg_stats_set_start_empty_time(bfqg);
4780		bfqg_stats_update_io_remove(bfqg, rq->cmd_flags);
4781	}
4782	spin_unlock_irq(&q->queue_lock);
4783}
4784#else
4785static inline void bfq_update_dispatch_stats(struct request_queue *q,
4786					     struct request *rq,
4787					     struct bfq_queue *in_serv_queue,
4788					     bool idle_timer_disabled) {}
4789#endif /* CONFIG_BFQ_CGROUP_DEBUG */
4790
4791static struct request *bfq_dispatch_request(struct blk_mq_hw_ctx *hctx)
4792{
4793	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
4794	struct request *rq;
4795	struct bfq_queue *in_serv_queue;
4796	bool waiting_rq, idle_timer_disabled;
4797
4798	spin_lock_irq(&bfqd->lock);
4799
4800	in_serv_queue = bfqd->in_service_queue;
4801	waiting_rq = in_serv_queue && bfq_bfqq_wait_request(in_serv_queue);
4802
4803	rq = __bfq_dispatch_request(hctx);
4804
4805	idle_timer_disabled =
4806		waiting_rq && !bfq_bfqq_wait_request(in_serv_queue);
 
4807
4808	spin_unlock_irq(&bfqd->lock);
4809
4810	bfq_update_dispatch_stats(hctx->queue, rq, in_serv_queue,
4811				  idle_timer_disabled);
4812
4813	return rq;
4814}
4815
4816/*
4817 * Task holds one reference to the queue, dropped when task exits.  Each rq
4818 * in-flight on this queue also holds a reference, dropped when rq is freed.
4819 *
4820 * Scheduler lock must be held here. Recall not to use bfqq after calling
4821 * this function on it.
4822 */
4823void bfq_put_queue(struct bfq_queue *bfqq)
4824{
4825	struct bfq_queue *item;
4826	struct hlist_node *n;
4827	struct bfq_group *bfqg = bfqq_group(bfqq);
4828
4829	if (bfqq->bfqd)
4830		bfq_log_bfqq(bfqq->bfqd, bfqq, "put_queue: %p %d",
4831			     bfqq, bfqq->ref);
4832
4833	bfqq->ref--;
4834	if (bfqq->ref)
4835		return;
4836
4837	if (!hlist_unhashed(&bfqq->burst_list_node)) {
4838		hlist_del_init(&bfqq->burst_list_node);
4839		/*
4840		 * Decrement also burst size after the removal, if the
4841		 * process associated with bfqq is exiting, and thus
4842		 * does not contribute to the burst any longer. This
4843		 * decrement helps filter out false positives of large
4844		 * bursts, when some short-lived process (often due to
4845		 * the execution of commands by some service) happens
4846		 * to start and exit while a complex application is
4847		 * starting, and thus spawning several processes that
4848		 * do I/O (and that *must not* be treated as a large
4849		 * burst, see comments on bfq_handle_burst).
4850		 *
4851		 * In particular, the decrement is performed only if:
4852		 * 1) bfqq is not a merged queue, because, if it is,
4853		 * then this free of bfqq is not triggered by the exit
4854		 * of the process bfqq is associated with, but exactly
4855		 * by the fact that bfqq has just been merged.
4856		 * 2) burst_size is greater than 0, to handle
4857		 * unbalanced decrements. Unbalanced decrements may
4858		 * happen in te following case: bfqq is inserted into
4859		 * the current burst list--without incrementing
4860		 * bust_size--because of a split, but the current
4861		 * burst list is not the burst list bfqq belonged to
4862		 * (see comments on the case of a split in
4863		 * bfq_set_request).
4864		 */
4865		if (bfqq->bic && bfqq->bfqd->burst_size > 0)
4866			bfqq->bfqd->burst_size--;
4867	}
4868
4869	/*
4870	 * bfqq does not exist any longer, so it cannot be woken by
4871	 * any other queue, and cannot wake any other queue. Then bfqq
4872	 * must be removed from the woken list of its possible waker
4873	 * queue, and all queues in the woken list of bfqq must stop
4874	 * having a waker queue. Strictly speaking, these updates
4875	 * should be performed when bfqq remains with no I/O source
4876	 * attached to it, which happens before bfqq gets freed. In
4877	 * particular, this happens when the last process associated
4878	 * with bfqq exits or gets associated with a different
4879	 * queue. However, both events lead to bfqq being freed soon,
4880	 * and dangling references would come out only after bfqq gets
4881	 * freed. So these updates are done here, as a simple and safe
4882	 * way to handle all cases.
4883	 */
4884	/* remove bfqq from woken list */
4885	if (!hlist_unhashed(&bfqq->woken_list_node))
4886		hlist_del_init(&bfqq->woken_list_node);
4887
4888	/* reset waker for all queues in woken list */
4889	hlist_for_each_entry_safe(item, n, &bfqq->woken_list,
4890				  woken_list_node) {
4891		item->waker_bfqq = NULL;
4892		bfq_clear_bfqq_has_waker(item);
4893		hlist_del_init(&item->woken_list_node);
4894	}
4895
4896	if (bfqq->bfqd && bfqq->bfqd->last_completed_rq_bfqq == bfqq)
4897		bfqq->bfqd->last_completed_rq_bfqq = NULL;
4898
4899	kmem_cache_free(bfq_pool, bfqq);
4900	bfqg_and_blkg_put(bfqg);
4901}
4902
4903static void bfq_put_cooperator(struct bfq_queue *bfqq)
 
 
 
 
 
 
4904{
4905	struct bfq_queue *__bfqq, *next;
4906
4907	/*
4908	 * If this queue was scheduled to merge with another queue, be
4909	 * sure to drop the reference taken on that queue (and others in
4910	 * the merge chain). See bfq_setup_merge and bfq_merge_bfqqs.
4911	 */
4912	__bfqq = bfqq->new_bfqq;
4913	while (__bfqq) {
4914		if (__bfqq == bfqq)
4915			break;
4916		next = __bfqq->new_bfqq;
4917		bfq_put_queue(__bfqq);
4918		__bfqq = next;
4919	}
4920}
4921
4922static void bfq_exit_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq)
4923{
4924	if (bfqq == bfqd->in_service_queue) {
4925		__bfq_bfqq_expire(bfqd, bfqq, BFQQE_BUDGET_TIMEOUT);
4926		bfq_schedule_dispatch(bfqd);
4927	}
4928
4929	bfq_log_bfqq(bfqd, bfqq, "exit_bfqq: %p, %d", bfqq, bfqq->ref);
4930
4931	bfq_put_cooperator(bfqq);
4932
4933	bfq_release_process_ref(bfqd, bfqq);
4934}
4935
4936static void bfq_exit_icq_bfqq(struct bfq_io_cq *bic, bool is_sync)
4937{
4938	struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync);
4939	struct bfq_data *bfqd;
4940
4941	if (bfqq)
4942		bfqd = bfqq->bfqd; /* NULL if scheduler already exited */
4943
4944	if (bfqq && bfqd) {
4945		unsigned long flags;
4946
4947		spin_lock_irqsave(&bfqd->lock, flags);
4948		bfqq->bic = NULL;
4949		bfq_exit_bfqq(bfqd, bfqq);
4950		bic_set_bfqq(bic, NULL, is_sync);
4951		spin_unlock_irqrestore(&bfqd->lock, flags);
4952	}
4953}
4954
4955static void bfq_exit_icq(struct io_cq *icq)
4956{
4957	struct bfq_io_cq *bic = icq_to_bic(icq);
4958
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4959	bfq_exit_icq_bfqq(bic, true);
4960	bfq_exit_icq_bfqq(bic, false);
4961}
4962
4963/*
4964 * Update the entity prio values; note that the new values will not
4965 * be used until the next (re)activation.
4966 */
4967static void
4968bfq_set_next_ioprio_data(struct bfq_queue *bfqq, struct bfq_io_cq *bic)
4969{
4970	struct task_struct *tsk = current;
4971	int ioprio_class;
4972	struct bfq_data *bfqd = bfqq->bfqd;
4973
4974	if (!bfqd)
4975		return;
4976
4977	ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio);
4978	switch (ioprio_class) {
4979	default:
4980		pr_err("bdi %s: bfq: bad prio class %d\n",
4981				bdi_dev_name(bfqq->bfqd->queue->backing_dev_info),
4982				ioprio_class);
4983		fallthrough;
4984	case IOPRIO_CLASS_NONE:
4985		/*
4986		 * No prio set, inherit CPU scheduling settings.
4987		 */
4988		bfqq->new_ioprio = task_nice_ioprio(tsk);
4989		bfqq->new_ioprio_class = task_nice_ioclass(tsk);
4990		break;
4991	case IOPRIO_CLASS_RT:
4992		bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
4993		bfqq->new_ioprio_class = IOPRIO_CLASS_RT;
4994		break;
4995	case IOPRIO_CLASS_BE:
4996		bfqq->new_ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
4997		bfqq->new_ioprio_class = IOPRIO_CLASS_BE;
4998		break;
4999	case IOPRIO_CLASS_IDLE:
5000		bfqq->new_ioprio_class = IOPRIO_CLASS_IDLE;
5001		bfqq->new_ioprio = 7;
5002		break;
5003	}
5004
5005	if (bfqq->new_ioprio >= IOPRIO_BE_NR) {
5006		pr_crit("bfq_set_next_ioprio_data: new_ioprio %d\n",
5007			bfqq->new_ioprio);
5008		bfqq->new_ioprio = IOPRIO_BE_NR;
5009	}
5010
5011	bfqq->entity.new_weight = bfq_ioprio_to_weight(bfqq->new_ioprio);
 
 
5012	bfqq->entity.prio_changed = 1;
5013}
5014
5015static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd,
5016				       struct bio *bio, bool is_sync,
5017				       struct bfq_io_cq *bic);
 
5018
5019static void bfq_check_ioprio_change(struct bfq_io_cq *bic, struct bio *bio)
5020{
5021	struct bfq_data *bfqd = bic_to_bfqd(bic);
5022	struct bfq_queue *bfqq;
5023	int ioprio = bic->icq.ioc->ioprio;
5024
5025	/*
5026	 * This condition may trigger on a newly created bic, be sure to
5027	 * drop the lock before returning.
5028	 */
5029	if (unlikely(!bfqd) || likely(bic->ioprio == ioprio))
5030		return;
5031
5032	bic->ioprio = ioprio;
5033
5034	bfqq = bic_to_bfqq(bic, false);
5035	if (bfqq) {
5036		bfq_release_process_ref(bfqd, bfqq);
5037		bfqq = bfq_get_queue(bfqd, bio, BLK_RW_ASYNC, bic);
 
5038		bic_set_bfqq(bic, bfqq, false);
 
5039	}
5040
5041	bfqq = bic_to_bfqq(bic, true);
5042	if (bfqq)
5043		bfq_set_next_ioprio_data(bfqq, bic);
5044}
5045
5046static void bfq_init_bfqq(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5047			  struct bfq_io_cq *bic, pid_t pid, int is_sync)
5048{
 
 
5049	RB_CLEAR_NODE(&bfqq->entity.rb_node);
5050	INIT_LIST_HEAD(&bfqq->fifo);
5051	INIT_HLIST_NODE(&bfqq->burst_list_node);
5052	INIT_HLIST_NODE(&bfqq->woken_list_node);
5053	INIT_HLIST_HEAD(&bfqq->woken_list);
5054
5055	bfqq->ref = 0;
5056	bfqq->bfqd = bfqd;
5057
5058	if (bic)
5059		bfq_set_next_ioprio_data(bfqq, bic);
5060
5061	if (is_sync) {
5062		/*
5063		 * No need to mark as has_short_ttime if in
5064		 * idle_class, because no device idling is performed
5065		 * for queues in idle class
5066		 */
5067		if (!bfq_class_idle(bfqq))
5068			/* tentatively mark as has_short_ttime */
5069			bfq_mark_bfqq_has_short_ttime(bfqq);
5070		bfq_mark_bfqq_sync(bfqq);
5071		bfq_mark_bfqq_just_created(bfqq);
5072	} else
5073		bfq_clear_bfqq_sync(bfqq);
5074
5075	/* set end request to minus infinity from now */
5076	bfqq->ttime.last_end_request = ktime_get_ns() + 1;
 
 
 
 
5077
5078	bfq_mark_bfqq_IO_bound(bfqq);
5079
5080	bfqq->pid = pid;
5081
5082	/* Tentative initial value to trade off between thr and lat */
5083	bfqq->max_budget = (2 * bfq_max_budget(bfqd)) / 3;
5084	bfqq->budget_timeout = bfq_smallest_from_now();
5085
5086	bfqq->wr_coeff = 1;
5087	bfqq->last_wr_start_finish = jiffies;
5088	bfqq->wr_start_at_switch_to_srt = bfq_smallest_from_now();
5089	bfqq->split_time = bfq_smallest_from_now();
5090
5091	/*
5092	 * To not forget the possibly high bandwidth consumed by a
5093	 * process/queue in the recent past,
5094	 * bfq_bfqq_softrt_next_start() returns a value at least equal
5095	 * to the current value of bfqq->soft_rt_next_start (see
5096	 * comments on bfq_bfqq_softrt_next_start).  Set
5097	 * soft_rt_next_start to now, to mean that bfqq has consumed
5098	 * no bandwidth so far.
5099	 */
5100	bfqq->soft_rt_next_start = jiffies;
5101
5102	/* first request is almost certainly seeky */
5103	bfqq->seek_history = 1;
5104}
5105
5106static struct bfq_queue **bfq_async_queue_prio(struct bfq_data *bfqd,
5107					       struct bfq_group *bfqg,
5108					       int ioprio_class, int ioprio)
5109{
5110	switch (ioprio_class) {
5111	case IOPRIO_CLASS_RT:
5112		return &bfqg->async_bfqq[0][ioprio];
5113	case IOPRIO_CLASS_NONE:
5114		ioprio = IOPRIO_NORM;
5115		fallthrough;
5116	case IOPRIO_CLASS_BE:
5117		return &bfqg->async_bfqq[1][ioprio];
5118	case IOPRIO_CLASS_IDLE:
5119		return &bfqg->async_idle_bfqq;
5120	default:
5121		return NULL;
5122	}
5123}
5124
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5125static struct bfq_queue *bfq_get_queue(struct bfq_data *bfqd,
5126				       struct bio *bio, bool is_sync,
5127				       struct bfq_io_cq *bic)
 
5128{
5129	const int ioprio = IOPRIO_PRIO_DATA(bic->ioprio);
5130	const int ioprio_class = IOPRIO_PRIO_CLASS(bic->ioprio);
5131	struct bfq_queue **async_bfqq = NULL;
5132	struct bfq_queue *bfqq;
5133	struct bfq_group *bfqg;
5134
5135	rcu_read_lock();
5136
5137	bfqg = bfq_find_set_group(bfqd, __bio_blkcg(bio));
5138	if (!bfqg) {
5139		bfqq = &bfqd->oom_bfqq;
5140		goto out;
5141	}
5142
5143	if (!is_sync) {
5144		async_bfqq = bfq_async_queue_prio(bfqd, bfqg, ioprio_class,
5145						  ioprio);
5146		bfqq = *async_bfqq;
5147		if (bfqq)
5148			goto out;
5149	}
5150
5151	bfqq = kmem_cache_alloc_node(bfq_pool,
5152				     GFP_NOWAIT | __GFP_ZERO | __GFP_NOWARN,
5153				     bfqd->queue->node);
5154
5155	if (bfqq) {
5156		bfq_init_bfqq(bfqd, bfqq, bic, current->pid,
5157			      is_sync);
5158		bfq_init_entity(&bfqq->entity, bfqg);
5159		bfq_log_bfqq(bfqd, bfqq, "allocated");
5160	} else {
5161		bfqq = &bfqd->oom_bfqq;
5162		bfq_log_bfqq(bfqd, bfqq, "using oom bfqq");
5163		goto out;
5164	}
5165
5166	/*
5167	 * Pin the queue now that it's allocated, scheduler exit will
5168	 * prune it.
5169	 */
5170	if (async_bfqq) {
5171		bfqq->ref++; /*
5172			      * Extra group reference, w.r.t. sync
5173			      * queue. This extra reference is removed
5174			      * only if bfqq->bfqg disappears, to
5175			      * guarantee that this queue is not freed
5176			      * until its group goes away.
5177			      */
5178		bfq_log_bfqq(bfqd, bfqq, "get_queue, bfqq not in async: %p, %d",
5179			     bfqq, bfqq->ref);
5180		*async_bfqq = bfqq;
5181	}
5182
5183out:
5184	bfqq->ref++; /* get a process reference to this queue */
5185	bfq_log_bfqq(bfqd, bfqq, "get_queue, at end: %p, %d", bfqq, bfqq->ref);
5186	rcu_read_unlock();
 
5187	return bfqq;
5188}
5189
5190static void bfq_update_io_thinktime(struct bfq_data *bfqd,
5191				    struct bfq_queue *bfqq)
5192{
5193	struct bfq_ttime *ttime = &bfqq->ttime;
5194	u64 elapsed = ktime_get_ns() - bfqq->ttime.last_end_request;
5195
 
 
 
 
 
 
 
 
5196	elapsed = min_t(u64, elapsed, 2ULL * bfqd->bfq_slice_idle);
5197
5198	ttime->ttime_samples = (7*bfqq->ttime.ttime_samples + 256) / 8;
5199	ttime->ttime_total = div_u64(7*ttime->ttime_total + 256*elapsed,  8);
5200	ttime->ttime_mean = div64_ul(ttime->ttime_total + 128,
5201				     ttime->ttime_samples);
5202}
5203
5204static void
5205bfq_update_io_seektime(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5206		       struct request *rq)
5207{
5208	bfqq->seek_history <<= 1;
5209	bfqq->seek_history |= BFQ_RQ_SEEKY(bfqd, bfqq->last_request_pos, rq);
5210
5211	if (bfqq->wr_coeff > 1 &&
5212	    bfqq->wr_cur_max_time == bfqd->bfq_wr_rt_max_time &&
5213	    BFQQ_TOTALLY_SEEKY(bfqq))
5214		bfq_bfqq_end_wr(bfqq);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5215}
5216
5217static void bfq_update_has_short_ttime(struct bfq_data *bfqd,
5218				       struct bfq_queue *bfqq,
5219				       struct bfq_io_cq *bic)
5220{
5221	bool has_short_ttime = true, state_changed;
5222
5223	/*
5224	 * No need to update has_short_ttime if bfqq is async or in
5225	 * idle io prio class, or if bfq_slice_idle is zero, because
5226	 * no device idling is performed for bfqq in this case.
5227	 */
5228	if (!bfq_bfqq_sync(bfqq) || bfq_class_idle(bfqq) ||
5229	    bfqd->bfq_slice_idle == 0)
5230		return;
5231
5232	/* Idle window just restored, statistics are meaningless. */
5233	if (time_is_after_eq_jiffies(bfqq->split_time +
5234				     bfqd->bfq_wr_min_idle_time))
5235		return;
5236
5237	/* Think time is infinite if no process is linked to
5238	 * bfqq. Otherwise check average think time to
5239	 * decide whether to mark as has_short_ttime
 
5240	 */
5241	if (atomic_read(&bic->icq.ioc->active_ref) == 0 ||
5242	    (bfq_sample_valid(bfqq->ttime.ttime_samples) &&
5243	     bfqq->ttime.ttime_mean > bfqd->bfq_slice_idle))
5244		has_short_ttime = false;
5245
5246	state_changed = has_short_ttime != bfq_bfqq_has_short_ttime(bfqq);
5247
5248	if (has_short_ttime)
5249		bfq_mark_bfqq_has_short_ttime(bfqq);
5250	else
5251		bfq_clear_bfqq_has_short_ttime(bfqq);
5252
5253	/*
5254	 * Until the base value for the total service time gets
5255	 * finally computed for bfqq, the inject limit does depend on
5256	 * the think-time state (short|long). In particular, the limit
5257	 * is 0 or 1 if the think time is deemed, respectively, as
5258	 * short or long (details in the comments in
5259	 * bfq_update_inject_limit()). Accordingly, the next
5260	 * instructions reset the inject limit if the think-time state
5261	 * has changed and the above base value is still to be
5262	 * computed.
5263	 *
5264	 * However, the reset is performed only if more than 100 ms
5265	 * have elapsed since the last update of the inject limit, or
5266	 * (inclusive) if the change is from short to long think
5267	 * time. The reason for this waiting is as follows.
5268	 *
5269	 * bfqq may have a long think time because of a
5270	 * synchronization with some other queue, i.e., because the
5271	 * I/O of some other queue may need to be completed for bfqq
5272	 * to receive new I/O. Details in the comments on the choice
5273	 * of the queue for injection in bfq_select_queue().
5274	 *
5275	 * As stressed in those comments, if such a synchronization is
5276	 * actually in place, then, without injection on bfqq, the
5277	 * blocking I/O cannot happen to served while bfqq is in
5278	 * service. As a consequence, if bfqq is granted
5279	 * I/O-dispatch-plugging, then bfqq remains empty, and no I/O
5280	 * is dispatched, until the idle timeout fires. This is likely
5281	 * to result in lower bandwidth and higher latencies for bfqq,
5282	 * and in a severe loss of total throughput.
5283	 *
5284	 * On the opposite end, a non-zero inject limit may allow the
5285	 * I/O that blocks bfqq to be executed soon, and therefore
5286	 * bfqq to receive new I/O soon.
5287	 *
5288	 * But, if the blocking gets actually eliminated, then the
5289	 * next think-time sample for bfqq may be very low. This in
5290	 * turn may cause bfqq's think time to be deemed
5291	 * short. Without the 100 ms barrier, this new state change
5292	 * would cause the body of the next if to be executed
5293	 * immediately. But this would set to 0 the inject
5294	 * limit. Without injection, the blocking I/O would cause the
5295	 * think time of bfqq to become long again, and therefore the
5296	 * inject limit to be raised again, and so on. The only effect
5297	 * of such a steady oscillation between the two think-time
5298	 * states would be to prevent effective injection on bfqq.
5299	 *
5300	 * In contrast, if the inject limit is not reset during such a
5301	 * long time interval as 100 ms, then the number of short
5302	 * think time samples can grow significantly before the reset
5303	 * is performed. As a consequence, the think time state can
5304	 * become stable before the reset. Therefore there will be no
5305	 * state change when the 100 ms elapse, and no reset of the
5306	 * inject limit. The inject limit remains steadily equal to 1
5307	 * both during and after the 100 ms. So injection can be
5308	 * performed at all times, and throughput gets boosted.
5309	 *
5310	 * An inject limit equal to 1 is however in conflict, in
5311	 * general, with the fact that the think time of bfqq is
5312	 * short, because injection may be likely to delay bfqq's I/O
5313	 * (as explained in the comments in
5314	 * bfq_update_inject_limit()). But this does not happen in
5315	 * this special case, because bfqq's low think time is due to
5316	 * an effective handling of a synchronization, through
5317	 * injection. In this special case, bfqq's I/O does not get
5318	 * delayed by injection; on the contrary, bfqq's I/O is
5319	 * brought forward, because it is not blocked for
5320	 * milliseconds.
5321	 *
5322	 * In addition, serving the blocking I/O much sooner, and much
5323	 * more frequently than once per I/O-plugging timeout, makes
5324	 * it much quicker to detect a waker queue (the concept of
5325	 * waker queue is defined in the comments in
5326	 * bfq_add_request()). This makes it possible to start sooner
5327	 * to boost throughput more effectively, by injecting the I/O
5328	 * of the waker queue unconditionally on every
5329	 * bfq_dispatch_request().
5330	 *
5331	 * One last, important benefit of not resetting the inject
5332	 * limit before 100 ms is that, during this time interval, the
5333	 * base value for the total service time is likely to get
5334	 * finally computed for bfqq, freeing the inject limit from
5335	 * its relation with the think time.
5336	 */
5337	if (state_changed && bfqq->last_serv_time_ns == 0 &&
5338	    (time_is_before_eq_jiffies(bfqq->decrease_time_jif +
5339				      msecs_to_jiffies(100)) ||
5340	     !has_short_ttime))
5341		bfq_reset_inject_limit(bfqd, bfqq);
5342}
5343
5344/*
5345 * Called when a new fs request (rq) is added to bfqq.  Check if there's
5346 * something we should do about it.
5347 */
5348static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq,
5349			    struct request *rq)
5350{
5351	if (rq->cmd_flags & REQ_META)
5352		bfqq->meta_pending++;
5353
5354	bfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
5355
5356	if (bfqq == bfqd->in_service_queue && bfq_bfqq_wait_request(bfqq)) {
5357		bool small_req = bfqq->queued[rq_is_sync(rq)] == 1 &&
5358				 blk_rq_sectors(rq) < 32;
5359		bool budget_timeout = bfq_bfqq_budget_timeout(bfqq);
5360
5361		/*
5362		 * There is just this request queued: if
5363		 * - the request is small, and
5364		 * - we are idling to boost throughput, and
5365		 * - the queue is not to be expired,
5366		 * then just exit.
5367		 *
5368		 * In this way, if the device is being idled to wait
5369		 * for a new request from the in-service queue, we
5370		 * avoid unplugging the device and committing the
5371		 * device to serve just a small request. In contrast
5372		 * we wait for the block layer to decide when to
5373		 * unplug the device: hopefully, new requests will be
5374		 * merged to this one quickly, then the device will be
5375		 * unplugged and larger requests will be dispatched.
5376		 */
5377		if (small_req && idling_boosts_thr_without_issues(bfqd, bfqq) &&
5378		    !budget_timeout)
5379			return;
5380
5381		/*
5382		 * A large enough request arrived, or idling is being
5383		 * performed to preserve service guarantees, or
5384		 * finally the queue is to be expired: in all these
5385		 * cases disk idling is to be stopped, so clear
5386		 * wait_request flag and reset timer.
5387		 */
5388		bfq_clear_bfqq_wait_request(bfqq);
5389		hrtimer_try_to_cancel(&bfqd->idle_slice_timer);
5390
5391		/*
5392		 * The queue is not empty, because a new request just
5393		 * arrived. Hence we can safely expire the queue, in
5394		 * case of budget timeout, without risking that the
5395		 * timestamps of the queue are not updated correctly.
5396		 * See [1] for more details.
5397		 */
5398		if (budget_timeout)
5399			bfq_bfqq_expire(bfqd, bfqq, false,
5400					BFQQE_BUDGET_TIMEOUT);
5401	}
5402}
5403
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5404/* returns true if it causes the idle timer to be disabled */
5405static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
5406{
5407	struct bfq_queue *bfqq = RQ_BFQQ(rq),
5408		*new_bfqq = bfq_setup_cooperator(bfqd, bfqq, rq, true);
 
5409	bool waiting, idle_timer_disabled = false;
5410
5411	if (new_bfqq) {
5412		/*
5413		 * Release the request's reference to the old bfqq
5414		 * and make sure one is taken to the shared queue.
5415		 */
5416		new_bfqq->allocated++;
5417		bfqq->allocated--;
5418		new_bfqq->ref++;
5419		/*
5420		 * If the bic associated with the process
5421		 * issuing this request still points to bfqq
5422		 * (and thus has not been already redirected
5423		 * to new_bfqq or even some other bfq_queue),
5424		 * then complete the merge and redirect it to
5425		 * new_bfqq.
5426		 */
5427		if (bic_to_bfqq(RQ_BIC(rq), 1) == bfqq)
5428			bfq_merge_bfqqs(bfqd, RQ_BIC(rq),
5429					bfqq, new_bfqq);
5430
5431		bfq_clear_bfqq_just_created(bfqq);
5432		/*
5433		 * rq is about to be enqueued into new_bfqq,
5434		 * release rq reference on bfqq
5435		 */
5436		bfq_put_queue(bfqq);
5437		rq->elv.priv[1] = new_bfqq;
5438		bfqq = new_bfqq;
5439	}
5440
5441	bfq_update_io_thinktime(bfqd, bfqq);
5442	bfq_update_has_short_ttime(bfqd, bfqq, RQ_BIC(rq));
5443	bfq_update_io_seektime(bfqd, bfqq, rq);
5444
5445	waiting = bfqq && bfq_bfqq_wait_request(bfqq);
5446	bfq_add_request(rq);
5447	idle_timer_disabled = waiting && !bfq_bfqq_wait_request(bfqq);
5448
5449	rq->fifo_time = ktime_get_ns() + bfqd->bfq_fifo_expire[rq_is_sync(rq)];
5450	list_add_tail(&rq->queuelist, &bfqq->fifo);
5451
5452	bfq_rq_enqueued(bfqd, bfqq, rq);
5453
5454	return idle_timer_disabled;
5455}
5456
5457#ifdef CONFIG_BFQ_CGROUP_DEBUG
5458static void bfq_update_insert_stats(struct request_queue *q,
5459				    struct bfq_queue *bfqq,
5460				    bool idle_timer_disabled,
5461				    unsigned int cmd_flags)
5462{
5463	if (!bfqq)
5464		return;
5465
5466	/*
5467	 * bfqq still exists, because it can disappear only after
5468	 * either it is merged with another queue, or the process it
5469	 * is associated with exits. But both actions must be taken by
5470	 * the same process currently executing this flow of
5471	 * instructions.
5472	 *
5473	 * In addition, the following queue lock guarantees that
5474	 * bfqq_group(bfqq) exists as well.
5475	 */
5476	spin_lock_irq(&q->queue_lock);
5477	bfqg_stats_update_io_add(bfqq_group(bfqq), bfqq, cmd_flags);
5478	if (idle_timer_disabled)
5479		bfqg_stats_update_idle_time(bfqq_group(bfqq));
5480	spin_unlock_irq(&q->queue_lock);
5481}
5482#else
5483static inline void bfq_update_insert_stats(struct request_queue *q,
5484					   struct bfq_queue *bfqq,
5485					   bool idle_timer_disabled,
5486					   unsigned int cmd_flags) {}
5487#endif /* CONFIG_BFQ_CGROUP_DEBUG */
5488
 
 
5489static void bfq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
5490			       bool at_head)
5491{
5492	struct request_queue *q = hctx->queue;
5493	struct bfq_data *bfqd = q->elevator->elevator_data;
5494	struct bfq_queue *bfqq;
5495	bool idle_timer_disabled = false;
5496	unsigned int cmd_flags;
 
5497
5498#ifdef CONFIG_BFQ_GROUP_IOSCHED
5499	if (!cgroup_subsys_on_dfl(io_cgrp_subsys) && rq->bio)
5500		bfqg_stats_update_legacy_io(q, rq);
5501#endif
5502	spin_lock_irq(&bfqd->lock);
5503	if (blk_mq_sched_try_insert_merge(q, rq)) {
 
5504		spin_unlock_irq(&bfqd->lock);
 
5505		return;
5506	}
5507
5508	spin_unlock_irq(&bfqd->lock);
5509
5510	blk_mq_sched_request_inserted(rq);
5511
5512	spin_lock_irq(&bfqd->lock);
5513	bfqq = bfq_init_rq(rq);
5514	if (!bfqq || at_head || blk_rq_is_passthrough(rq)) {
5515		if (at_head)
5516			list_add(&rq->queuelist, &bfqd->dispatch);
5517		else
5518			list_add_tail(&rq->queuelist, &bfqd->dispatch);
5519	} else {
5520		idle_timer_disabled = __bfq_insert_request(bfqd, rq);
5521		/*
5522		 * Update bfqq, because, if a queue merge has occurred
5523		 * in __bfq_insert_request, then rq has been
5524		 * redirected into a new queue.
5525		 */
5526		bfqq = RQ_BFQQ(rq);
5527
5528		if (rq_mergeable(rq)) {
5529			elv_rqhash_add(q, rq);
5530			if (!q->last_merge)
5531				q->last_merge = rq;
5532		}
5533	}
5534
5535	/*
5536	 * Cache cmd_flags before releasing scheduler lock, because rq
5537	 * may disappear afterwards (for example, because of a request
5538	 * merge).
5539	 */
5540	cmd_flags = rq->cmd_flags;
5541
5542	spin_unlock_irq(&bfqd->lock);
5543
5544	bfq_update_insert_stats(q, bfqq, idle_timer_disabled,
5545				cmd_flags);
5546}
5547
5548static void bfq_insert_requests(struct blk_mq_hw_ctx *hctx,
5549				struct list_head *list, bool at_head)
5550{
5551	while (!list_empty(list)) {
5552		struct request *rq;
5553
5554		rq = list_first_entry(list, struct request, queuelist);
5555		list_del_init(&rq->queuelist);
5556		bfq_insert_request(hctx, rq, at_head);
5557	}
5558}
5559
5560static void bfq_update_hw_tag(struct bfq_data *bfqd)
5561{
5562	struct bfq_queue *bfqq = bfqd->in_service_queue;
5563
5564	bfqd->max_rq_in_driver = max_t(int, bfqd->max_rq_in_driver,
5565				       bfqd->rq_in_driver);
5566
5567	if (bfqd->hw_tag == 1)
5568		return;
5569
5570	/*
5571	 * This sample is valid if the number of outstanding requests
5572	 * is large enough to allow a queueing behavior.  Note that the
5573	 * sum is not exact, as it's not taking into account deactivated
5574	 * requests.
5575	 */
5576	if (bfqd->rq_in_driver + bfqd->queued <= BFQ_HW_QUEUE_THRESHOLD)
5577		return;
5578
5579	/*
5580	 * If active queue hasn't enough requests and can idle, bfq might not
5581	 * dispatch sufficient requests to hardware. Don't zero hw_tag in this
5582	 * case
5583	 */
5584	if (bfqq && bfq_bfqq_has_short_ttime(bfqq) &&
5585	    bfqq->dispatched + bfqq->queued[0] + bfqq->queued[1] <
5586	    BFQ_HW_QUEUE_THRESHOLD &&
5587	    bfqd->rq_in_driver < BFQ_HW_QUEUE_THRESHOLD)
5588		return;
5589
5590	if (bfqd->hw_tag_samples++ < BFQ_HW_QUEUE_SAMPLES)
5591		return;
5592
5593	bfqd->hw_tag = bfqd->max_rq_in_driver > BFQ_HW_QUEUE_THRESHOLD;
5594	bfqd->max_rq_in_driver = 0;
5595	bfqd->hw_tag_samples = 0;
5596
5597	bfqd->nonrot_with_queueing =
5598		blk_queue_nonrot(bfqd->queue) && bfqd->hw_tag;
5599}
5600
5601static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd)
5602{
5603	u64 now_ns;
5604	u32 delta_us;
5605
5606	bfq_update_hw_tag(bfqd);
5607
5608	bfqd->rq_in_driver--;
5609	bfqq->dispatched--;
5610
5611	if (!bfqq->dispatched && !bfq_bfqq_busy(bfqq)) {
5612		/*
5613		 * Set budget_timeout (which we overload to store the
5614		 * time at which the queue remains with no backlog and
5615		 * no outstanding request; used by the weight-raising
5616		 * mechanism).
5617		 */
5618		bfqq->budget_timeout = jiffies;
5619
5620		bfq_weights_tree_remove(bfqd, bfqq);
 
5621	}
5622
5623	now_ns = ktime_get_ns();
5624
5625	bfqq->ttime.last_end_request = now_ns;
5626
5627	/*
5628	 * Using us instead of ns, to get a reasonable precision in
5629	 * computing rate in next check.
5630	 */
5631	delta_us = div_u64(now_ns - bfqd->last_completion, NSEC_PER_USEC);
5632
5633	/*
5634	 * If the request took rather long to complete, and, according
5635	 * to the maximum request size recorded, this completion latency
5636	 * implies that the request was certainly served at a very low
5637	 * rate (less than 1M sectors/sec), then the whole observation
5638	 * interval that lasts up to this time instant cannot be a
5639	 * valid time interval for computing a new peak rate.  Invoke
5640	 * bfq_update_rate_reset to have the following three steps
5641	 * taken:
5642	 * - close the observation interval at the last (previous)
5643	 *   request dispatch or completion
5644	 * - compute rate, if possible, for that observation interval
5645	 * - reset to zero samples, which will trigger a proper
5646	 *   re-initialization of the observation interval on next
5647	 *   dispatch
5648	 */
5649	if (delta_us > BFQ_MIN_TT/NSEC_PER_USEC &&
5650	   (bfqd->last_rq_max_size<<BFQ_RATE_SHIFT)/delta_us <
5651			1UL<<(BFQ_RATE_SHIFT - 10))
5652		bfq_update_rate_reset(bfqd, NULL);
5653	bfqd->last_completion = now_ns;
5654	bfqd->last_completed_rq_bfqq = bfqq;
 
 
 
 
 
 
 
 
 
 
 
 
5655
5656	/*
5657	 * If we are waiting to discover whether the request pattern
5658	 * of the task associated with the queue is actually
5659	 * isochronous, and both requisites for this condition to hold
5660	 * are now satisfied, then compute soft_rt_next_start (see the
5661	 * comments on the function bfq_bfqq_softrt_next_start()). We
5662	 * do not compute soft_rt_next_start if bfqq is in interactive
5663	 * weight raising (see the comments in bfq_bfqq_expire() for
5664	 * an explanation). We schedule this delayed update when bfqq
5665	 * expires, if it still has in-flight requests.
5666	 */
5667	if (bfq_bfqq_softrt_update(bfqq) && bfqq->dispatched == 0 &&
5668	    RB_EMPTY_ROOT(&bfqq->sort_list) &&
5669	    bfqq->wr_coeff != bfqd->bfq_wr_coeff)
5670		bfqq->soft_rt_next_start =
5671			bfq_bfqq_softrt_next_start(bfqd, bfqq);
5672
5673	/*
5674	 * If this is the in-service queue, check if it needs to be expired,
5675	 * or if we want to idle in case it has no pending requests.
5676	 */
5677	if (bfqd->in_service_queue == bfqq) {
5678		if (bfq_bfqq_must_idle(bfqq)) {
5679			if (bfqq->dispatched == 0)
5680				bfq_arm_slice_timer(bfqd);
5681			/*
5682			 * If we get here, we do not expire bfqq, even
5683			 * if bfqq was in budget timeout or had no
5684			 * more requests (as controlled in the next
5685			 * conditional instructions). The reason for
5686			 * not expiring bfqq is as follows.
5687			 *
5688			 * Here bfqq->dispatched > 0 holds, but
5689			 * bfq_bfqq_must_idle() returned true. This
5690			 * implies that, even if no request arrives
5691			 * for bfqq before bfqq->dispatched reaches 0,
5692			 * bfqq will, however, not be expired on the
5693			 * completion event that causes bfqq->dispatch
5694			 * to reach zero. In contrast, on this event,
5695			 * bfqq will start enjoying device idling
5696			 * (I/O-dispatch plugging).
5697			 *
5698			 * But, if we expired bfqq here, bfqq would
5699			 * not have the chance to enjoy device idling
5700			 * when bfqq->dispatched finally reaches
5701			 * zero. This would expose bfqq to violation
5702			 * of its reserved service guarantees.
5703			 */
5704			return;
5705		} else if (bfq_may_expire_for_budg_timeout(bfqq))
5706			bfq_bfqq_expire(bfqd, bfqq, false,
5707					BFQQE_BUDGET_TIMEOUT);
5708		else if (RB_EMPTY_ROOT(&bfqq->sort_list) &&
5709			 (bfqq->dispatched == 0 ||
5710			  !bfq_better_to_idle(bfqq)))
5711			bfq_bfqq_expire(bfqd, bfqq, false,
5712					BFQQE_NO_MORE_REQUESTS);
5713	}
5714
5715	if (!bfqd->rq_in_driver)
5716		bfq_schedule_dispatch(bfqd);
5717}
5718
5719static void bfq_finish_requeue_request_body(struct bfq_queue *bfqq)
5720{
5721	bfqq->allocated--;
5722
5723	bfq_put_queue(bfqq);
5724}
5725
5726/*
5727 * The processes associated with bfqq may happen to generate their
5728 * cumulative I/O at a lower rate than the rate at which the device
5729 * could serve the same I/O. This is rather probable, e.g., if only
5730 * one process is associated with bfqq and the device is an SSD. It
5731 * results in bfqq becoming often empty while in service. In this
5732 * respect, if BFQ is allowed to switch to another queue when bfqq
5733 * remains empty, then the device goes on being fed with I/O requests,
5734 * and the throughput is not affected. In contrast, if BFQ is not
5735 * allowed to switch to another queue---because bfqq is sync and
5736 * I/O-dispatch needs to be plugged while bfqq is temporarily
5737 * empty---then, during the service of bfqq, there will be frequent
5738 * "service holes", i.e., time intervals during which bfqq gets empty
5739 * and the device can only consume the I/O already queued in its
5740 * hardware queues. During service holes, the device may even get to
5741 * remaining idle. In the end, during the service of bfqq, the device
5742 * is driven at a lower speed than the one it can reach with the kind
5743 * of I/O flowing through bfqq.
5744 *
5745 * To counter this loss of throughput, BFQ implements a "request
5746 * injection mechanism", which tries to fill the above service holes
5747 * with I/O requests taken from other queues. The hard part in this
5748 * mechanism is finding the right amount of I/O to inject, so as to
5749 * both boost throughput and not break bfqq's bandwidth and latency
5750 * guarantees. In this respect, the mechanism maintains a per-queue
5751 * inject limit, computed as below. While bfqq is empty, the injection
5752 * mechanism dispatches extra I/O requests only until the total number
5753 * of I/O requests in flight---i.e., already dispatched but not yet
5754 * completed---remains lower than this limit.
5755 *
5756 * A first definition comes in handy to introduce the algorithm by
5757 * which the inject limit is computed.  We define as first request for
5758 * bfqq, an I/O request for bfqq that arrives while bfqq is in
5759 * service, and causes bfqq to switch from empty to non-empty. The
5760 * algorithm updates the limit as a function of the effect of
5761 * injection on the service times of only the first requests of
5762 * bfqq. The reason for this restriction is that these are the
5763 * requests whose service time is affected most, because they are the
5764 * first to arrive after injection possibly occurred.
5765 *
5766 * To evaluate the effect of injection, the algorithm measures the
5767 * "total service time" of first requests. We define as total service
5768 * time of an I/O request, the time that elapses since when the
5769 * request is enqueued into bfqq, to when it is completed. This
5770 * quantity allows the whole effect of injection to be measured. It is
5771 * easy to see why. Suppose that some requests of other queues are
5772 * actually injected while bfqq is empty, and that a new request R
5773 * then arrives for bfqq. If the device does start to serve all or
5774 * part of the injected requests during the service hole, then,
5775 * because of this extra service, it may delay the next invocation of
5776 * the dispatch hook of BFQ. Then, even after R gets eventually
5777 * dispatched, the device may delay the actual service of R if it is
5778 * still busy serving the extra requests, or if it decides to serve,
5779 * before R, some extra request still present in its queues. As a
5780 * conclusion, the cumulative extra delay caused by injection can be
5781 * easily evaluated by just comparing the total service time of first
5782 * requests with and without injection.
5783 *
5784 * The limit-update algorithm works as follows. On the arrival of a
5785 * first request of bfqq, the algorithm measures the total time of the
5786 * request only if one of the three cases below holds, and, for each
5787 * case, it updates the limit as described below:
5788 *
5789 * (1) If there is no in-flight request. This gives a baseline for the
5790 *     total service time of the requests of bfqq. If the baseline has
5791 *     not been computed yet, then, after computing it, the limit is
5792 *     set to 1, to start boosting throughput, and to prepare the
5793 *     ground for the next case. If the baseline has already been
5794 *     computed, then it is updated, in case it results to be lower
5795 *     than the previous value.
5796 *
5797 * (2) If the limit is higher than 0 and there are in-flight
5798 *     requests. By comparing the total service time in this case with
5799 *     the above baseline, it is possible to know at which extent the
5800 *     current value of the limit is inflating the total service
5801 *     time. If the inflation is below a certain threshold, then bfqq
5802 *     is assumed to be suffering from no perceivable loss of its
5803 *     service guarantees, and the limit is even tentatively
5804 *     increased. If the inflation is above the threshold, then the
5805 *     limit is decreased. Due to the lack of any hysteresis, this
5806 *     logic makes the limit oscillate even in steady workload
5807 *     conditions. Yet we opted for it, because it is fast in reaching
5808 *     the best value for the limit, as a function of the current I/O
5809 *     workload. To reduce oscillations, this step is disabled for a
5810 *     short time interval after the limit happens to be decreased.
5811 *
5812 * (3) Periodically, after resetting the limit, to make sure that the
5813 *     limit eventually drops in case the workload changes. This is
5814 *     needed because, after the limit has gone safely up for a
5815 *     certain workload, it is impossible to guess whether the
5816 *     baseline total service time may have changed, without measuring
5817 *     it again without injection. A more effective version of this
5818 *     step might be to just sample the baseline, by interrupting
5819 *     injection only once, and then to reset/lower the limit only if
5820 *     the total service time with the current limit does happen to be
5821 *     too large.
5822 *
5823 * More details on each step are provided in the comments on the
5824 * pieces of code that implement these steps: the branch handling the
5825 * transition from empty to non empty in bfq_add_request(), the branch
5826 * handling injection in bfq_select_queue(), and the function
5827 * bfq_choose_bfqq_for_injection(). These comments also explain some
5828 * exceptions, made by the injection mechanism in some special cases.
5829 */
5830static void bfq_update_inject_limit(struct bfq_data *bfqd,
5831				    struct bfq_queue *bfqq)
5832{
5833	u64 tot_time_ns = ktime_get_ns() - bfqd->last_empty_occupied_ns;
5834	unsigned int old_limit = bfqq->inject_limit;
5835
5836	if (bfqq->last_serv_time_ns > 0 && bfqd->rqs_injected) {
5837		u64 threshold = (bfqq->last_serv_time_ns * 3)>>1;
5838
5839		if (tot_time_ns >= threshold && old_limit > 0) {
5840			bfqq->inject_limit--;
5841			bfqq->decrease_time_jif = jiffies;
5842		} else if (tot_time_ns < threshold &&
5843			   old_limit <= bfqd->max_rq_in_driver)
5844			bfqq->inject_limit++;
5845	}
5846
5847	/*
5848	 * Either we still have to compute the base value for the
5849	 * total service time, and there seem to be the right
5850	 * conditions to do it, or we can lower the last base value
5851	 * computed.
5852	 *
5853	 * NOTE: (bfqd->rq_in_driver == 1) means that there is no I/O
5854	 * request in flight, because this function is in the code
5855	 * path that handles the completion of a request of bfqq, and,
5856	 * in particular, this function is executed before
5857	 * bfqd->rq_in_driver is decremented in such a code path.
5858	 */
5859	if ((bfqq->last_serv_time_ns == 0 && bfqd->rq_in_driver == 1) ||
5860	    tot_time_ns < bfqq->last_serv_time_ns) {
5861		if (bfqq->last_serv_time_ns == 0) {
5862			/*
5863			 * Now we certainly have a base value: make sure we
5864			 * start trying injection.
5865			 */
5866			bfqq->inject_limit = max_t(unsigned int, 1, old_limit);
5867		}
5868		bfqq->last_serv_time_ns = tot_time_ns;
5869	} else if (!bfqd->rqs_injected && bfqd->rq_in_driver == 1)
5870		/*
5871		 * No I/O injected and no request still in service in
5872		 * the drive: these are the exact conditions for
5873		 * computing the base value of the total service time
5874		 * for bfqq. So let's update this value, because it is
5875		 * rather variable. For example, it varies if the size
5876		 * or the spatial locality of the I/O requests in bfqq
5877		 * change.
5878		 */
5879		bfqq->last_serv_time_ns = tot_time_ns;
5880
5881
5882	/* update complete, not waiting for any request completion any longer */
5883	bfqd->waited_rq = NULL;
5884	bfqd->rqs_injected = false;
5885}
5886
5887/*
5888 * Handle either a requeue or a finish for rq. The things to do are
5889 * the same in both cases: all references to rq are to be dropped. In
5890 * particular, rq is considered completed from the point of view of
5891 * the scheduler.
5892 */
5893static void bfq_finish_requeue_request(struct request *rq)
5894{
5895	struct bfq_queue *bfqq = RQ_BFQQ(rq);
5896	struct bfq_data *bfqd;
 
5897
5898	/*
5899	 * rq either is not associated with any icq, or is an already
5900	 * requeued request that has not (yet) been re-inserted into
5901	 * a bfq_queue.
5902	 */
5903	if (!rq->elv.icq || !bfqq)
5904		return;
5905
5906	bfqd = bfqq->bfqd;
5907
5908	if (rq->rq_flags & RQF_STARTED)
5909		bfqg_stats_update_completion(bfqq_group(bfqq),
5910					     rq->start_time_ns,
5911					     rq->io_start_time_ns,
5912					     rq->cmd_flags);
5913
 
5914	if (likely(rq->rq_flags & RQF_STARTED)) {
5915		unsigned long flags;
5916
5917		spin_lock_irqsave(&bfqd->lock, flags);
5918
5919		if (rq == bfqd->waited_rq)
5920			bfq_update_inject_limit(bfqd, bfqq);
5921
5922		bfq_completed_request(bfqq, bfqd);
5923		bfq_finish_requeue_request_body(bfqq);
5924
5925		spin_unlock_irqrestore(&bfqd->lock, flags);
5926	} else {
5927		/*
5928		 * Request rq may be still/already in the scheduler,
5929		 * in which case we need to remove it (this should
5930		 * never happen in case of requeue). And we cannot
5931		 * defer such a check and removal, to avoid
5932		 * inconsistencies in the time interval from the end
5933		 * of this function to the start of the deferred work.
5934		 * This situation seems to occur only in process
5935		 * context, as a consequence of a merge. In the
5936		 * current version of the code, this implies that the
5937		 * lock is held.
5938		 */
5939
5940		if (!RB_EMPTY_NODE(&rq->rb_node)) {
5941			bfq_remove_request(rq->q, rq);
5942			bfqg_stats_update_io_remove(bfqq_group(bfqq),
5943						    rq->cmd_flags);
5944		}
5945		bfq_finish_requeue_request_body(bfqq);
5946	}
 
 
 
 
5947
5948	/*
5949	 * Reset private fields. In case of a requeue, this allows
5950	 * this function to correctly do nothing if it is spuriously
5951	 * invoked again on this same request (see the check at the
5952	 * beginning of the function). Probably, a better general
5953	 * design would be to prevent blk-mq from invoking the requeue
5954	 * or finish hooks of an elevator, for a request that is not
5955	 * referred by that elevator.
5956	 *
5957	 * Resetting the following fields would break the
5958	 * request-insertion logic if rq is re-inserted into a bfq
5959	 * internal queue, without a re-preparation. Here we assume
5960	 * that re-insertions of requeued requests, without
5961	 * re-preparation, can happen only for pass_through or at_head
5962	 * requests (which are not re-inserted into bfq internal
5963	 * queues).
5964	 */
5965	rq->elv.priv[0] = NULL;
5966	rq->elv.priv[1] = NULL;
5967}
5968
 
 
 
 
 
 
 
 
 
 
5969/*
5970 * Removes the association between the current task and bfqq, assuming
5971 * that bic points to the bfq iocontext of the task.
5972 * Returns NULL if a new bfqq should be allocated, or the old bfqq if this
5973 * was the last process referring to that bfqq.
5974 */
5975static struct bfq_queue *
5976bfq_split_bfqq(struct bfq_io_cq *bic, struct bfq_queue *bfqq)
5977{
5978	bfq_log_bfqq(bfqq->bfqd, bfqq, "splitting queue");
5979
5980	if (bfqq_process_refs(bfqq) == 1) {
5981		bfqq->pid = current->pid;
5982		bfq_clear_bfqq_coop(bfqq);
5983		bfq_clear_bfqq_split_coop(bfqq);
5984		return bfqq;
5985	}
5986
5987	bic_set_bfqq(bic, NULL, 1);
5988
5989	bfq_put_cooperator(bfqq);
5990
5991	bfq_release_process_ref(bfqq->bfqd, bfqq);
5992	return NULL;
5993}
5994
5995static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd,
5996						   struct bfq_io_cq *bic,
5997						   struct bio *bio,
5998						   bool split, bool is_sync,
5999						   bool *new_queue)
6000{
6001	struct bfq_queue *bfqq = bic_to_bfqq(bic, is_sync);
6002
6003	if (likely(bfqq && bfqq != &bfqd->oom_bfqq))
6004		return bfqq;
6005
6006	if (new_queue)
6007		*new_queue = true;
6008
6009	if (bfqq)
6010		bfq_put_queue(bfqq);
6011	bfqq = bfq_get_queue(bfqd, bio, is_sync, bic);
6012
6013	bic_set_bfqq(bic, bfqq, is_sync);
6014	if (split && is_sync) {
6015		if ((bic->was_in_burst_list && bfqd->large_burst) ||
6016		    bic->saved_in_large_burst)
6017			bfq_mark_bfqq_in_large_burst(bfqq);
6018		else {
6019			bfq_clear_bfqq_in_large_burst(bfqq);
6020			if (bic->was_in_burst_list)
6021				/*
6022				 * If bfqq was in the current
6023				 * burst list before being
6024				 * merged, then we have to add
6025				 * it back. And we do not need
6026				 * to increase burst_size, as
6027				 * we did not decrement
6028				 * burst_size when we removed
6029				 * bfqq from the burst list as
6030				 * a consequence of a merge
6031				 * (see comments in
6032				 * bfq_put_queue). In this
6033				 * respect, it would be rather
6034				 * costly to know whether the
6035				 * current burst list is still
6036				 * the same burst list from
6037				 * which bfqq was removed on
6038				 * the merge. To avoid this
6039				 * cost, if bfqq was in a
6040				 * burst list, then we add
6041				 * bfqq to the current burst
6042				 * list without any further
6043				 * check. This can cause
6044				 * inappropriate insertions,
6045				 * but rarely enough to not
6046				 * harm the detection of large
6047				 * bursts significantly.
6048				 */
6049				hlist_add_head(&bfqq->burst_list_node,
6050					       &bfqd->burst_list);
6051		}
6052		bfqq->split_time = jiffies;
6053	}
6054
6055	return bfqq;
6056}
6057
6058/*
6059 * Only reset private fields. The actual request preparation will be
6060 * performed by bfq_init_rq, when rq is either inserted or merged. See
6061 * comments on bfq_init_rq for the reason behind this delayed
6062 * preparation.
6063 */
6064static void bfq_prepare_request(struct request *rq)
6065{
 
 
6066	/*
6067	 * Regardless of whether we have an icq attached, we have to
6068	 * clear the scheduler pointers, as they might point to
6069	 * previously allocated bic/bfqq structs.
6070	 */
6071	rq->elv.priv[0] = rq->elv.priv[1] = NULL;
6072}
6073
6074/*
6075 * If needed, init rq, allocate bfq data structures associated with
6076 * rq, and increment reference counters in the destination bfq_queue
6077 * for rq. Return the destination bfq_queue for rq, or NULL is rq is
6078 * not associated with any bfq_queue.
6079 *
6080 * This function is invoked by the functions that perform rq insertion
6081 * or merging. One may have expected the above preparation operations
6082 * to be performed in bfq_prepare_request, and not delayed to when rq
6083 * is inserted or merged. The rationale behind this delayed
6084 * preparation is that, after the prepare_request hook is invoked for
6085 * rq, rq may still be transformed into a request with no icq, i.e., a
6086 * request not associated with any queue. No bfq hook is invoked to
6087 * signal this transformation. As a consequence, should these
6088 * preparation operations be performed when the prepare_request hook
6089 * is invoked, and should rq be transformed one moment later, bfq
6090 * would end up in an inconsistent state, because it would have
6091 * incremented some queue counters for an rq destined to
6092 * transformation, without any chance to correctly lower these
6093 * counters back. In contrast, no transformation can still happen for
6094 * rq after rq has been inserted or merged. So, it is safe to execute
6095 * these preparation operations when rq is finally inserted or merged.
6096 */
6097static struct bfq_queue *bfq_init_rq(struct request *rq)
6098{
6099	struct request_queue *q = rq->q;
6100	struct bio *bio = rq->bio;
6101	struct bfq_data *bfqd = q->elevator->elevator_data;
6102	struct bfq_io_cq *bic;
6103	const int is_sync = rq_is_sync(rq);
6104	struct bfq_queue *bfqq;
6105	bool new_queue = false;
6106	bool bfqq_already_existing = false, split = false;
6107
6108	if (unlikely(!rq->elv.icq))
6109		return NULL;
6110
6111	/*
6112	 * Assuming that elv.priv[1] is set only if everything is set
6113	 * for this rq. This holds true, because this function is
6114	 * invoked only for insertion or merging, and, after such
6115	 * events, a request cannot be manipulated any longer before
6116	 * being removed from bfq.
6117	 */
6118	if (rq->elv.priv[1])
6119		return rq->elv.priv[1];
6120
6121	bic = icq_to_bic(rq->elv.icq);
6122
6123	bfq_check_ioprio_change(bic, bio);
6124
6125	bfq_bic_update_cgroup(bic, bio);
6126
6127	bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio, false, is_sync,
6128					 &new_queue);
6129
6130	if (likely(!new_queue)) {
6131		/* If the queue was seeky for too long, break it apart. */
6132		if (bfq_bfqq_coop(bfqq) && bfq_bfqq_split_coop(bfqq)) {
6133			bfq_log_bfqq(bfqd, bfqq, "breaking apart bfqq");
 
6134
6135			/* Update bic before losing reference to bfqq */
6136			if (bfq_bfqq_in_large_burst(bfqq))
6137				bic->saved_in_large_burst = true;
6138
6139			bfqq = bfq_split_bfqq(bic, bfqq);
6140			split = true;
6141
6142			if (!bfqq)
6143				bfqq = bfq_get_bfqq_handle_split(bfqd, bic, bio,
6144								 true, is_sync,
6145								 NULL);
6146			else
 
 
6147				bfqq_already_existing = true;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6148		}
6149	}
6150
6151	bfqq->allocated++;
6152	bfqq->ref++;
 
6153	bfq_log_bfqq(bfqd, bfqq, "get_request %p: bfqq %p, %d",
6154		     rq, bfqq, bfqq->ref);
6155
6156	rq->elv.priv[0] = bic;
6157	rq->elv.priv[1] = bfqq;
6158
6159	/*
6160	 * If a bfq_queue has only one process reference, it is owned
6161	 * by only this bic: we can then set bfqq->bic = bic. in
6162	 * addition, if the queue has also just been split, we have to
6163	 * resume its state.
6164	 */
6165	if (likely(bfqq != &bfqd->oom_bfqq) && bfqq_process_refs(bfqq) == 1) {
6166		bfqq->bic = bic;
6167		if (split) {
6168			/*
6169			 * The queue has just been split from a shared
6170			 * queue: restore the idle window and the
6171			 * possible weight raising period.
6172			 */
6173			bfq_bfqq_resume_state(bfqq, bfqd, bic,
6174					      bfqq_already_existing);
6175		}
6176	}
6177
6178	/*
6179	 * Consider bfqq as possibly belonging to a burst of newly
6180	 * created queues only if:
6181	 * 1) A burst is actually happening (bfqd->burst_size > 0)
6182	 * or
6183	 * 2) There is no other active queue. In fact, if, in
6184	 *    contrast, there are active queues not belonging to the
6185	 *    possible burst bfqq may belong to, then there is no gain
6186	 *    in considering bfqq as belonging to a burst, and
6187	 *    therefore in not weight-raising bfqq. See comments on
6188	 *    bfq_handle_burst().
6189	 *
6190	 * This filtering also helps eliminating false positives,
6191	 * occurring when bfqq does not belong to an actual large
6192	 * burst, but some background task (e.g., a service) happens
6193	 * to trigger the creation of new queues very close to when
6194	 * bfqq and its possible companion queues are created. See
6195	 * comments on bfq_handle_burst() for further details also on
6196	 * this issue.
6197	 */
6198	if (unlikely(bfq_bfqq_just_created(bfqq) &&
6199		     (bfqd->burst_size > 0 ||
6200		      bfq_tot_busy_queues(bfqd) == 0)))
6201		bfq_handle_burst(bfqd, bfqq);
6202
6203	return bfqq;
6204}
6205
6206static void
6207bfq_idle_slice_timer_body(struct bfq_data *bfqd, struct bfq_queue *bfqq)
6208{
6209	enum bfqq_expiration reason;
6210	unsigned long flags;
6211
6212	spin_lock_irqsave(&bfqd->lock, flags);
6213
6214	/*
6215	 * Considering that bfqq may be in race, we should firstly check
6216	 * whether bfqq is in service before doing something on it. If
6217	 * the bfqq in race is not in service, it has already been expired
6218	 * through __bfq_bfqq_expire func and its wait_request flags has
6219	 * been cleared in __bfq_bfqd_reset_in_service func.
6220	 */
6221	if (bfqq != bfqd->in_service_queue) {
6222		spin_unlock_irqrestore(&bfqd->lock, flags);
6223		return;
6224	}
6225
6226	bfq_clear_bfqq_wait_request(bfqq);
6227
6228	if (bfq_bfqq_budget_timeout(bfqq))
6229		/*
6230		 * Also here the queue can be safely expired
6231		 * for budget timeout without wasting
6232		 * guarantees
6233		 */
6234		reason = BFQQE_BUDGET_TIMEOUT;
6235	else if (bfqq->queued[0] == 0 && bfqq->queued[1] == 0)
6236		/*
6237		 * The queue may not be empty upon timer expiration,
6238		 * because we may not disable the timer when the
6239		 * first request of the in-service queue arrives
6240		 * during disk idling.
6241		 */
6242		reason = BFQQE_TOO_IDLE;
6243	else
6244		goto schedule_dispatch;
6245
6246	bfq_bfqq_expire(bfqd, bfqq, true, reason);
6247
6248schedule_dispatch:
 
6249	spin_unlock_irqrestore(&bfqd->lock, flags);
6250	bfq_schedule_dispatch(bfqd);
6251}
6252
6253/*
6254 * Handler of the expiration of the timer running if the in-service queue
6255 * is idling inside its time slice.
6256 */
6257static enum hrtimer_restart bfq_idle_slice_timer(struct hrtimer *timer)
6258{
6259	struct bfq_data *bfqd = container_of(timer, struct bfq_data,
6260					     idle_slice_timer);
6261	struct bfq_queue *bfqq = bfqd->in_service_queue;
6262
6263	/*
6264	 * Theoretical race here: the in-service queue can be NULL or
6265	 * different from the queue that was idling if a new request
6266	 * arrives for the current queue and there is a full dispatch
6267	 * cycle that changes the in-service queue.  This can hardly
6268	 * happen, but in the worst case we just expire a queue too
6269	 * early.
6270	 */
6271	if (bfqq)
6272		bfq_idle_slice_timer_body(bfqd, bfqq);
6273
6274	return HRTIMER_NORESTART;
6275}
6276
6277static void __bfq_put_async_bfqq(struct bfq_data *bfqd,
6278				 struct bfq_queue **bfqq_ptr)
6279{
6280	struct bfq_queue *bfqq = *bfqq_ptr;
6281
6282	bfq_log(bfqd, "put_async_bfqq: %p", bfqq);
6283	if (bfqq) {
6284		bfq_bfqq_move(bfqd, bfqq, bfqd->root_group);
6285
6286		bfq_log_bfqq(bfqd, bfqq, "put_async_bfqq: putting %p, %d",
6287			     bfqq, bfqq->ref);
6288		bfq_put_queue(bfqq);
6289		*bfqq_ptr = NULL;
6290	}
6291}
6292
6293/*
6294 * Release all the bfqg references to its async queues.  If we are
6295 * deallocating the group these queues may still contain requests, so
6296 * we reparent them to the root cgroup (i.e., the only one that will
6297 * exist for sure until all the requests on a device are gone).
6298 */
6299void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg)
6300{
6301	int i, j;
6302
6303	for (i = 0; i < 2; i++)
6304		for (j = 0; j < IOPRIO_BE_NR; j++)
6305			__bfq_put_async_bfqq(bfqd, &bfqg->async_bfqq[i][j]);
6306
6307	__bfq_put_async_bfqq(bfqd, &bfqg->async_idle_bfqq);
6308}
6309
6310/*
6311 * See the comments on bfq_limit_depth for the purpose of
6312 * the depths set in the function. Return minimum shallow depth we'll use.
6313 */
6314static unsigned int bfq_update_depths(struct bfq_data *bfqd,
6315				      struct sbitmap_queue *bt)
6316{
6317	unsigned int i, j, min_shallow = UINT_MAX;
6318
 
6319	/*
6320	 * In-word depths if no bfq_queue is being weight-raised:
6321	 * leaving 25% of tags only for sync reads.
6322	 *
6323	 * In next formulas, right-shift the value
6324	 * (1U<<bt->sb.shift), instead of computing directly
6325	 * (1U<<(bt->sb.shift - something)), to be robust against
6326	 * any possible value of bt->sb.shift, without having to
6327	 * limit 'something'.
6328	 */
6329	/* no more than 50% of tags for async I/O */
6330	bfqd->word_depths[0][0] = max((1U << bt->sb.shift) >> 1, 1U);
6331	/*
6332	 * no more than 75% of tags for sync writes (25% extra tags
6333	 * w.r.t. async I/O, to prevent async I/O from starving sync
6334	 * writes)
6335	 */
6336	bfqd->word_depths[0][1] = max(((1U << bt->sb.shift) * 3) >> 2, 1U);
6337
6338	/*
6339	 * In-word depths in case some bfq_queue is being weight-
6340	 * raised: leaving ~63% of tags for sync reads. This is the
6341	 * highest percentage for which, in our tests, application
6342	 * start-up times didn't suffer from any regression due to tag
6343	 * shortage.
6344	 */
6345	/* no more than ~18% of tags for async I/O */
6346	bfqd->word_depths[1][0] = max(((1U << bt->sb.shift) * 3) >> 4, 1U);
6347	/* no more than ~37% of tags for sync writes (~20% extra tags) */
6348	bfqd->word_depths[1][1] = max(((1U << bt->sb.shift) * 6) >> 4, 1U);
6349
6350	for (i = 0; i < 2; i++)
6351		for (j = 0; j < 2; j++)
6352			min_shallow = min(min_shallow, bfqd->word_depths[i][j]);
6353
6354	return min_shallow;
6355}
6356
6357static void bfq_depth_updated(struct blk_mq_hw_ctx *hctx)
6358{
6359	struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
6360	struct blk_mq_tags *tags = hctx->sched_tags;
6361	unsigned int min_shallow;
6362
6363	min_shallow = bfq_update_depths(bfqd, &tags->bitmap_tags);
6364	sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, min_shallow);
6365}
6366
6367static int bfq_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int index)
6368{
6369	bfq_depth_updated(hctx);
6370	return 0;
6371}
6372
6373static void bfq_exit_queue(struct elevator_queue *e)
6374{
6375	struct bfq_data *bfqd = e->elevator_data;
6376	struct bfq_queue *bfqq, *n;
6377
6378	hrtimer_cancel(&bfqd->idle_slice_timer);
6379
6380	spin_lock_irq(&bfqd->lock);
6381	list_for_each_entry_safe(bfqq, n, &bfqd->idle_list, bfqq_list)
6382		bfq_deactivate_bfqq(bfqd, bfqq, false, false);
6383	spin_unlock_irq(&bfqd->lock);
6384
6385	hrtimer_cancel(&bfqd->idle_slice_timer);
6386
6387	/* release oom-queue reference to root group */
6388	bfqg_and_blkg_put(bfqd->root_group);
6389
6390#ifdef CONFIG_BFQ_GROUP_IOSCHED
6391	blkcg_deactivate_policy(bfqd->queue, &blkcg_policy_bfq);
6392#else
6393	spin_lock_irq(&bfqd->lock);
6394	bfq_put_async_queues(bfqd, bfqd->root_group);
6395	kfree(bfqd->root_group);
6396	spin_unlock_irq(&bfqd->lock);
6397#endif
6398
 
 
 
 
6399	kfree(bfqd);
6400}
6401
6402static void bfq_init_root_group(struct bfq_group *root_group,
6403				struct bfq_data *bfqd)
6404{
6405	int i;
6406
6407#ifdef CONFIG_BFQ_GROUP_IOSCHED
6408	root_group->entity.parent = NULL;
6409	root_group->my_entity = NULL;
6410	root_group->bfqd = bfqd;
6411#endif
6412	root_group->rq_pos_tree = RB_ROOT;
6413	for (i = 0; i < BFQ_IOPRIO_CLASSES; i++)
6414		root_group->sched_data.service_tree[i] = BFQ_SERVICE_TREE_INIT;
6415	root_group->sched_data.bfq_class_idle_last_service = jiffies;
6416}
6417
6418static int bfq_init_queue(struct request_queue *q, struct elevator_type *e)
6419{
6420	struct bfq_data *bfqd;
6421	struct elevator_queue *eq;
6422
6423	eq = elevator_alloc(q, e);
6424	if (!eq)
6425		return -ENOMEM;
6426
6427	bfqd = kzalloc_node(sizeof(*bfqd), GFP_KERNEL, q->node);
6428	if (!bfqd) {
6429		kobject_put(&eq->kobj);
6430		return -ENOMEM;
6431	}
6432	eq->elevator_data = bfqd;
6433
6434	spin_lock_irq(&q->queue_lock);
6435	q->elevator = eq;
6436	spin_unlock_irq(&q->queue_lock);
6437
6438	/*
6439	 * Our fallback bfqq if bfq_find_alloc_queue() runs into OOM issues.
6440	 * Grab a permanent reference to it, so that the normal code flow
6441	 * will not attempt to free it.
6442	 */
6443	bfq_init_bfqq(bfqd, &bfqd->oom_bfqq, NULL, 1, 0);
6444	bfqd->oom_bfqq.ref++;
6445	bfqd->oom_bfqq.new_ioprio = BFQ_DEFAULT_QUEUE_IOPRIO;
6446	bfqd->oom_bfqq.new_ioprio_class = IOPRIO_CLASS_BE;
6447	bfqd->oom_bfqq.entity.new_weight =
6448		bfq_ioprio_to_weight(bfqd->oom_bfqq.new_ioprio);
6449
6450	/* oom_bfqq does not participate to bursts */
6451	bfq_clear_bfqq_just_created(&bfqd->oom_bfqq);
6452
6453	/*
6454	 * Trigger weight initialization, according to ioprio, at the
6455	 * oom_bfqq's first activation. The oom_bfqq's ioprio and ioprio
6456	 * class won't be changed any more.
6457	 */
6458	bfqd->oom_bfqq.entity.prio_changed = 1;
6459
6460	bfqd->queue = q;
6461
6462	INIT_LIST_HEAD(&bfqd->dispatch);
6463
6464	hrtimer_init(&bfqd->idle_slice_timer, CLOCK_MONOTONIC,
6465		     HRTIMER_MODE_REL);
6466	bfqd->idle_slice_timer.function = bfq_idle_slice_timer;
6467
6468	bfqd->queue_weights_tree = RB_ROOT_CACHED;
 
6469	bfqd->num_groups_with_pending_reqs = 0;
 
6470
6471	INIT_LIST_HEAD(&bfqd->active_list);
6472	INIT_LIST_HEAD(&bfqd->idle_list);
6473	INIT_HLIST_HEAD(&bfqd->burst_list);
6474
6475	bfqd->hw_tag = -1;
6476	bfqd->nonrot_with_queueing = blk_queue_nonrot(bfqd->queue);
6477
6478	bfqd->bfq_max_budget = bfq_default_max_budget;
6479
6480	bfqd->bfq_fifo_expire[0] = bfq_fifo_expire[0];
6481	bfqd->bfq_fifo_expire[1] = bfq_fifo_expire[1];
6482	bfqd->bfq_back_max = bfq_back_max;
6483	bfqd->bfq_back_penalty = bfq_back_penalty;
6484	bfqd->bfq_slice_idle = bfq_slice_idle;
6485	bfqd->bfq_timeout = bfq_timeout;
6486
6487	bfqd->bfq_requests_within_timer = 120;
6488
6489	bfqd->bfq_large_burst_thresh = 8;
6490	bfqd->bfq_burst_interval = msecs_to_jiffies(180);
6491
6492	bfqd->low_latency = true;
6493
6494	/*
6495	 * Trade-off between responsiveness and fairness.
6496	 */
6497	bfqd->bfq_wr_coeff = 30;
6498	bfqd->bfq_wr_rt_max_time = msecs_to_jiffies(300);
6499	bfqd->bfq_wr_max_time = 0;
6500	bfqd->bfq_wr_min_idle_time = msecs_to_jiffies(2000);
6501	bfqd->bfq_wr_min_inter_arr_async = msecs_to_jiffies(500);
6502	bfqd->bfq_wr_max_softrt_rate = 7000; /*
6503					      * Approximate rate required
6504					      * to playback or record a
6505					      * high-definition compressed
6506					      * video.
6507					      */
6508	bfqd->wr_busy_queues = 0;
6509
6510	/*
6511	 * Begin by assuming, optimistically, that the device peak
6512	 * rate is equal to 2/3 of the highest reference rate.
6513	 */
6514	bfqd->rate_dur_prod = ref_rate[blk_queue_nonrot(bfqd->queue)] *
6515		ref_wr_duration[blk_queue_nonrot(bfqd->queue)];
6516	bfqd->peak_rate = ref_rate[blk_queue_nonrot(bfqd->queue)] * 2 / 3;
6517
6518	spin_lock_init(&bfqd->lock);
6519
6520	/*
6521	 * The invocation of the next bfq_create_group_hierarchy
6522	 * function is the head of a chain of function calls
6523	 * (bfq_create_group_hierarchy->blkcg_activate_policy->
6524	 * blk_mq_freeze_queue) that may lead to the invocation of the
6525	 * has_work hook function. For this reason,
6526	 * bfq_create_group_hierarchy is invoked only after all
6527	 * scheduler data has been initialized, apart from the fields
6528	 * that can be initialized only after invoking
6529	 * bfq_create_group_hierarchy. This, in particular, enables
6530	 * has_work to correctly return false. Of course, to avoid
6531	 * other inconsistencies, the blk-mq stack must then refrain
6532	 * from invoking further scheduler hooks before this init
6533	 * function is finished.
6534	 */
6535	bfqd->root_group = bfq_create_group_hierarchy(bfqd, q->node);
6536	if (!bfqd->root_group)
6537		goto out_free;
6538	bfq_init_root_group(bfqd->root_group, bfqd);
6539	bfq_init_entity(&bfqd->oom_bfqq.entity, bfqd->root_group);
6540
 
 
 
 
6541	wbt_disable_default(q);
 
 
6542	return 0;
6543
6544out_free:
6545	kfree(bfqd);
6546	kobject_put(&eq->kobj);
6547	return -ENOMEM;
6548}
6549
6550static void bfq_slab_kill(void)
6551{
6552	kmem_cache_destroy(bfq_pool);
6553}
6554
6555static int __init bfq_slab_setup(void)
6556{
6557	bfq_pool = KMEM_CACHE(bfq_queue, 0);
6558	if (!bfq_pool)
6559		return -ENOMEM;
6560	return 0;
6561}
6562
6563static ssize_t bfq_var_show(unsigned int var, char *page)
6564{
6565	return sprintf(page, "%u\n", var);
6566}
6567
6568static int bfq_var_store(unsigned long *var, const char *page)
6569{
6570	unsigned long new_val;
6571	int ret = kstrtoul(page, 10, &new_val);
6572
6573	if (ret)
6574		return ret;
6575	*var = new_val;
6576	return 0;
6577}
6578
6579#define SHOW_FUNCTION(__FUNC, __VAR, __CONV)				\
6580static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
6581{									\
6582	struct bfq_data *bfqd = e->elevator_data;			\
6583	u64 __data = __VAR;						\
6584	if (__CONV == 1)						\
6585		__data = jiffies_to_msecs(__data);			\
6586	else if (__CONV == 2)						\
6587		__data = div_u64(__data, NSEC_PER_MSEC);		\
6588	return bfq_var_show(__data, (page));				\
6589}
6590SHOW_FUNCTION(bfq_fifo_expire_sync_show, bfqd->bfq_fifo_expire[1], 2);
6591SHOW_FUNCTION(bfq_fifo_expire_async_show, bfqd->bfq_fifo_expire[0], 2);
6592SHOW_FUNCTION(bfq_back_seek_max_show, bfqd->bfq_back_max, 0);
6593SHOW_FUNCTION(bfq_back_seek_penalty_show, bfqd->bfq_back_penalty, 0);
6594SHOW_FUNCTION(bfq_slice_idle_show, bfqd->bfq_slice_idle, 2);
6595SHOW_FUNCTION(bfq_max_budget_show, bfqd->bfq_user_max_budget, 0);
6596SHOW_FUNCTION(bfq_timeout_sync_show, bfqd->bfq_timeout, 1);
6597SHOW_FUNCTION(bfq_strict_guarantees_show, bfqd->strict_guarantees, 0);
6598SHOW_FUNCTION(bfq_low_latency_show, bfqd->low_latency, 0);
6599#undef SHOW_FUNCTION
6600
6601#define USEC_SHOW_FUNCTION(__FUNC, __VAR)				\
6602static ssize_t __FUNC(struct elevator_queue *e, char *page)		\
6603{									\
6604	struct bfq_data *bfqd = e->elevator_data;			\
6605	u64 __data = __VAR;						\
6606	__data = div_u64(__data, NSEC_PER_USEC);			\
6607	return bfq_var_show(__data, (page));				\
6608}
6609USEC_SHOW_FUNCTION(bfq_slice_idle_us_show, bfqd->bfq_slice_idle);
6610#undef USEC_SHOW_FUNCTION
6611
6612#define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)			\
6613static ssize_t								\
6614__FUNC(struct elevator_queue *e, const char *page, size_t count)	\
6615{									\
6616	struct bfq_data *bfqd = e->elevator_data;			\
6617	unsigned long __data, __min = (MIN), __max = (MAX);		\
6618	int ret;							\
6619									\
6620	ret = bfq_var_store(&__data, (page));				\
6621	if (ret)							\
6622		return ret;						\
6623	if (__data < __min)						\
6624		__data = __min;						\
6625	else if (__data > __max)					\
6626		__data = __max;						\
6627	if (__CONV == 1)						\
6628		*(__PTR) = msecs_to_jiffies(__data);			\
6629	else if (__CONV == 2)						\
6630		*(__PTR) = (u64)__data * NSEC_PER_MSEC;			\
6631	else								\
6632		*(__PTR) = __data;					\
6633	return count;							\
6634}
6635STORE_FUNCTION(bfq_fifo_expire_sync_store, &bfqd->bfq_fifo_expire[1], 1,
6636		INT_MAX, 2);
6637STORE_FUNCTION(bfq_fifo_expire_async_store, &bfqd->bfq_fifo_expire[0], 1,
6638		INT_MAX, 2);
6639STORE_FUNCTION(bfq_back_seek_max_store, &bfqd->bfq_back_max, 0, INT_MAX, 0);
6640STORE_FUNCTION(bfq_back_seek_penalty_store, &bfqd->bfq_back_penalty, 1,
6641		INT_MAX, 0);
6642STORE_FUNCTION(bfq_slice_idle_store, &bfqd->bfq_slice_idle, 0, INT_MAX, 2);
6643#undef STORE_FUNCTION
6644
6645#define USEC_STORE_FUNCTION(__FUNC, __PTR, MIN, MAX)			\
6646static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count)\
6647{									\
6648	struct bfq_data *bfqd = e->elevator_data;			\
6649	unsigned long __data, __min = (MIN), __max = (MAX);		\
6650	int ret;							\
6651									\
6652	ret = bfq_var_store(&__data, (page));				\
6653	if (ret)							\
6654		return ret;						\
6655	if (__data < __min)						\
6656		__data = __min;						\
6657	else if (__data > __max)					\
6658		__data = __max;						\
6659	*(__PTR) = (u64)__data * NSEC_PER_USEC;				\
6660	return count;							\
6661}
6662USEC_STORE_FUNCTION(bfq_slice_idle_us_store, &bfqd->bfq_slice_idle, 0,
6663		    UINT_MAX);
6664#undef USEC_STORE_FUNCTION
6665
6666static ssize_t bfq_max_budget_store(struct elevator_queue *e,
6667				    const char *page, size_t count)
6668{
6669	struct bfq_data *bfqd = e->elevator_data;
6670	unsigned long __data;
6671	int ret;
6672
6673	ret = bfq_var_store(&__data, (page));
6674	if (ret)
6675		return ret;
6676
6677	if (__data == 0)
6678		bfqd->bfq_max_budget = bfq_calc_max_budget(bfqd);
6679	else {
6680		if (__data > INT_MAX)
6681			__data = INT_MAX;
6682		bfqd->bfq_max_budget = __data;
6683	}
6684
6685	bfqd->bfq_user_max_budget = __data;
6686
6687	return count;
6688}
6689
6690/*
6691 * Leaving this name to preserve name compatibility with cfq
6692 * parameters, but this timeout is used for both sync and async.
6693 */
6694static ssize_t bfq_timeout_sync_store(struct elevator_queue *e,
6695				      const char *page, size_t count)
6696{
6697	struct bfq_data *bfqd = e->elevator_data;
6698	unsigned long __data;
6699	int ret;
6700
6701	ret = bfq_var_store(&__data, (page));
6702	if (ret)
6703		return ret;
6704
6705	if (__data < 1)
6706		__data = 1;
6707	else if (__data > INT_MAX)
6708		__data = INT_MAX;
6709
6710	bfqd->bfq_timeout = msecs_to_jiffies(__data);
6711	if (bfqd->bfq_user_max_budget == 0)
6712		bfqd->bfq_max_budget = bfq_calc_max_budget(bfqd);
6713
6714	return count;
6715}
6716
6717static ssize_t bfq_strict_guarantees_store(struct elevator_queue *e,
6718				     const char *page, size_t count)
6719{
6720	struct bfq_data *bfqd = e->elevator_data;
6721	unsigned long __data;
6722	int ret;
6723
6724	ret = bfq_var_store(&__data, (page));
6725	if (ret)
6726		return ret;
6727
6728	if (__data > 1)
6729		__data = 1;
6730	if (!bfqd->strict_guarantees && __data == 1
6731	    && bfqd->bfq_slice_idle < 8 * NSEC_PER_MSEC)
6732		bfqd->bfq_slice_idle = 8 * NSEC_PER_MSEC;
6733
6734	bfqd->strict_guarantees = __data;
6735
6736	return count;
6737}
6738
6739static ssize_t bfq_low_latency_store(struct elevator_queue *e,
6740				     const char *page, size_t count)
6741{
6742	struct bfq_data *bfqd = e->elevator_data;
6743	unsigned long __data;
6744	int ret;
6745
6746	ret = bfq_var_store(&__data, (page));
6747	if (ret)
6748		return ret;
6749
6750	if (__data > 1)
6751		__data = 1;
6752	if (__data == 0 && bfqd->low_latency != 0)
6753		bfq_end_wr(bfqd);
6754	bfqd->low_latency = __data;
6755
6756	return count;
6757}
6758
6759#define BFQ_ATTR(name) \
6760	__ATTR(name, 0644, bfq_##name##_show, bfq_##name##_store)
6761
6762static struct elv_fs_entry bfq_attrs[] = {
6763	BFQ_ATTR(fifo_expire_sync),
6764	BFQ_ATTR(fifo_expire_async),
6765	BFQ_ATTR(back_seek_max),
6766	BFQ_ATTR(back_seek_penalty),
6767	BFQ_ATTR(slice_idle),
6768	BFQ_ATTR(slice_idle_us),
6769	BFQ_ATTR(max_budget),
6770	BFQ_ATTR(timeout_sync),
6771	BFQ_ATTR(strict_guarantees),
6772	BFQ_ATTR(low_latency),
6773	__ATTR_NULL
6774};
6775
6776static struct elevator_type iosched_bfq_mq = {
6777	.ops = {
6778		.limit_depth		= bfq_limit_depth,
6779		.prepare_request	= bfq_prepare_request,
6780		.requeue_request        = bfq_finish_requeue_request,
6781		.finish_request		= bfq_finish_requeue_request,
6782		.exit_icq		= bfq_exit_icq,
6783		.insert_requests	= bfq_insert_requests,
6784		.dispatch_request	= bfq_dispatch_request,
6785		.next_request		= elv_rb_latter_request,
6786		.former_request		= elv_rb_former_request,
6787		.allow_merge		= bfq_allow_bio_merge,
6788		.bio_merge		= bfq_bio_merge,
6789		.request_merge		= bfq_request_merge,
6790		.requests_merged	= bfq_requests_merged,
6791		.request_merged		= bfq_request_merged,
6792		.has_work		= bfq_has_work,
6793		.depth_updated		= bfq_depth_updated,
6794		.init_hctx		= bfq_init_hctx,
6795		.init_sched		= bfq_init_queue,
6796		.exit_sched		= bfq_exit_queue,
6797	},
6798
6799	.icq_size =		sizeof(struct bfq_io_cq),
6800	.icq_align =		__alignof__(struct bfq_io_cq),
6801	.elevator_attrs =	bfq_attrs,
6802	.elevator_name =	"bfq",
6803	.elevator_owner =	THIS_MODULE,
6804};
6805MODULE_ALIAS("bfq-iosched");
6806
6807static int __init bfq_init(void)
6808{
6809	int ret;
6810
6811#ifdef CONFIG_BFQ_GROUP_IOSCHED
6812	ret = blkcg_policy_register(&blkcg_policy_bfq);
6813	if (ret)
6814		return ret;
6815#endif
6816
6817	ret = -ENOMEM;
6818	if (bfq_slab_setup())
6819		goto err_pol_unreg;
6820
6821	/*
6822	 * Times to load large popular applications for the typical
6823	 * systems installed on the reference devices (see the
6824	 * comments before the definition of the next
6825	 * array). Actually, we use slightly lower values, as the
6826	 * estimated peak rate tends to be smaller than the actual
6827	 * peak rate.  The reason for this last fact is that estimates
6828	 * are computed over much shorter time intervals than the long
6829	 * intervals typically used for benchmarking. Why? First, to
6830	 * adapt more quickly to variations. Second, because an I/O
6831	 * scheduler cannot rely on a peak-rate-evaluation workload to
6832	 * be run for a long time.
6833	 */
6834	ref_wr_duration[0] = msecs_to_jiffies(7000); /* actually 8 sec */
6835	ref_wr_duration[1] = msecs_to_jiffies(2500); /* actually 3 sec */
6836
6837	ret = elv_register(&iosched_bfq_mq);
6838	if (ret)
6839		goto slab_kill;
6840
6841	return 0;
6842
6843slab_kill:
6844	bfq_slab_kill();
6845err_pol_unreg:
6846#ifdef CONFIG_BFQ_GROUP_IOSCHED
6847	blkcg_policy_unregister(&blkcg_policy_bfq);
6848#endif
6849	return ret;
6850}
6851
6852static void __exit bfq_exit(void)
6853{
6854	elv_unregister(&iosched_bfq_mq);
6855#ifdef CONFIG_BFQ_GROUP_IOSCHED
6856	blkcg_policy_unregister(&blkcg_policy_bfq);
6857#endif
6858	bfq_slab_kill();
6859}
6860
6861module_init(bfq_init);
6862module_exit(bfq_exit);
6863
6864MODULE_AUTHOR("Paolo Valente");
6865MODULE_LICENSE("GPL");
6866MODULE_DESCRIPTION("MQ Budget Fair Queueing I/O Scheduler");