Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * linux/net/sunrpc/xprt.c
4 *
5 * This is a generic RPC call interface supporting congestion avoidance,
6 * and asynchronous calls.
7 *
8 * The interface works like this:
9 *
10 * - When a process places a call, it allocates a request slot if
11 * one is available. Otherwise, it sleeps on the backlog queue
12 * (xprt_reserve).
13 * - Next, the caller puts together the RPC message, stuffs it into
14 * the request struct, and calls xprt_transmit().
15 * - xprt_transmit sends the message and installs the caller on the
16 * transport's wait list. At the same time, if a reply is expected,
17 * it installs a timer that is run after the packet's timeout has
18 * expired.
19 * - When a packet arrives, the data_ready handler walks the list of
20 * pending requests for that transport. If a matching XID is found, the
21 * caller is woken up, and the timer removed.
22 * - When no reply arrives within the timeout interval, the timer is
23 * fired by the kernel and runs xprt_timer(). It either adjusts the
24 * timeout values (minor timeout) or wakes up the caller with a status
25 * of -ETIMEDOUT.
26 * - When the caller receives a notification from RPC that a reply arrived,
27 * it should release the RPC slot, and process the reply.
28 * If the call timed out, it may choose to retry the operation by
29 * adjusting the initial timeout value, and simply calling rpc_call
30 * again.
31 *
32 * Support for async RPC is done through a set of RPC-specific scheduling
33 * primitives that `transparently' work for processes as well as async
34 * tasks that rely on callbacks.
35 *
36 * Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
37 *
38 * Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
39 */
40
41#include <linux/module.h>
42
43#include <linux/types.h>
44#include <linux/interrupt.h>
45#include <linux/workqueue.h>
46#include <linux/net.h>
47#include <linux/ktime.h>
48
49#include <linux/sunrpc/clnt.h>
50#include <linux/sunrpc/metrics.h>
51#include <linux/sunrpc/bc_xprt.h>
52#include <linux/rcupdate.h>
53#include <linux/sched/mm.h>
54
55#include <trace/events/sunrpc.h>
56
57#include "sunrpc.h"
58
59/*
60 * Local variables
61 */
62
63#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
64# define RPCDBG_FACILITY RPCDBG_XPRT
65#endif
66
67/*
68 * Local functions
69 */
70static void xprt_init(struct rpc_xprt *xprt, struct net *net);
71static __be32 xprt_alloc_xid(struct rpc_xprt *xprt);
72static void xprt_destroy(struct rpc_xprt *xprt);
73
74static DEFINE_SPINLOCK(xprt_list_lock);
75static LIST_HEAD(xprt_list);
76
77static unsigned long xprt_request_timeout(const struct rpc_rqst *req)
78{
79 unsigned long timeout = jiffies + req->rq_timeout;
80
81 if (time_before(timeout, req->rq_majortimeo))
82 return timeout;
83 return req->rq_majortimeo;
84}
85
86/**
87 * xprt_register_transport - register a transport implementation
88 * @transport: transport to register
89 *
90 * If a transport implementation is loaded as a kernel module, it can
91 * call this interface to make itself known to the RPC client.
92 *
93 * Returns:
94 * 0: transport successfully registered
95 * -EEXIST: transport already registered
96 * -EINVAL: transport module being unloaded
97 */
98int xprt_register_transport(struct xprt_class *transport)
99{
100 struct xprt_class *t;
101 int result;
102
103 result = -EEXIST;
104 spin_lock(&xprt_list_lock);
105 list_for_each_entry(t, &xprt_list, list) {
106 /* don't register the same transport class twice */
107 if (t->ident == transport->ident)
108 goto out;
109 }
110
111 list_add_tail(&transport->list, &xprt_list);
112 printk(KERN_INFO "RPC: Registered %s transport module.\n",
113 transport->name);
114 result = 0;
115
116out:
117 spin_unlock(&xprt_list_lock);
118 return result;
119}
120EXPORT_SYMBOL_GPL(xprt_register_transport);
121
122/**
123 * xprt_unregister_transport - unregister a transport implementation
124 * @transport: transport to unregister
125 *
126 * Returns:
127 * 0: transport successfully unregistered
128 * -ENOENT: transport never registered
129 */
130int xprt_unregister_transport(struct xprt_class *transport)
131{
132 struct xprt_class *t;
133 int result;
134
135 result = 0;
136 spin_lock(&xprt_list_lock);
137 list_for_each_entry(t, &xprt_list, list) {
138 if (t == transport) {
139 printk(KERN_INFO
140 "RPC: Unregistered %s transport module.\n",
141 transport->name);
142 list_del_init(&transport->list);
143 goto out;
144 }
145 }
146 result = -ENOENT;
147
148out:
149 spin_unlock(&xprt_list_lock);
150 return result;
151}
152EXPORT_SYMBOL_GPL(xprt_unregister_transport);
153
154/**
155 * xprt_load_transport - load a transport implementation
156 * @transport_name: transport to load
157 *
158 * Returns:
159 * 0: transport successfully loaded
160 * -ENOENT: transport module not available
161 */
162int xprt_load_transport(const char *transport_name)
163{
164 struct xprt_class *t;
165 int result;
166
167 result = 0;
168 spin_lock(&xprt_list_lock);
169 list_for_each_entry(t, &xprt_list, list) {
170 if (strcmp(t->name, transport_name) == 0) {
171 spin_unlock(&xprt_list_lock);
172 goto out;
173 }
174 }
175 spin_unlock(&xprt_list_lock);
176 result = request_module("xprt%s", transport_name);
177out:
178 return result;
179}
180EXPORT_SYMBOL_GPL(xprt_load_transport);
181
182static void xprt_clear_locked(struct rpc_xprt *xprt)
183{
184 xprt->snd_task = NULL;
185 if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state)) {
186 smp_mb__before_atomic();
187 clear_bit(XPRT_LOCKED, &xprt->state);
188 smp_mb__after_atomic();
189 } else
190 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
191}
192
193/**
194 * xprt_reserve_xprt - serialize write access to transports
195 * @task: task that is requesting access to the transport
196 * @xprt: pointer to the target transport
197 *
198 * This prevents mixing the payload of separate requests, and prevents
199 * transport connects from colliding with writes. No congestion control
200 * is provided.
201 */
202int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
203{
204 struct rpc_rqst *req = task->tk_rqstp;
205
206 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
207 if (task == xprt->snd_task)
208 return 1;
209 goto out_sleep;
210 }
211 if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
212 goto out_unlock;
213 xprt->snd_task = task;
214
215 return 1;
216
217out_unlock:
218 xprt_clear_locked(xprt);
219out_sleep:
220 dprintk("RPC: %5u failed to lock transport %p\n",
221 task->tk_pid, xprt);
222 task->tk_status = -EAGAIN;
223 if (RPC_IS_SOFT(task))
224 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
225 xprt_request_timeout(req));
226 else
227 rpc_sleep_on(&xprt->sending, task, NULL);
228 return 0;
229}
230EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
231
232static bool
233xprt_need_congestion_window_wait(struct rpc_xprt *xprt)
234{
235 return test_bit(XPRT_CWND_WAIT, &xprt->state);
236}
237
238static void
239xprt_set_congestion_window_wait(struct rpc_xprt *xprt)
240{
241 if (!list_empty(&xprt->xmit_queue)) {
242 /* Peek at head of queue to see if it can make progress */
243 if (list_first_entry(&xprt->xmit_queue, struct rpc_rqst,
244 rq_xmit)->rq_cong)
245 return;
246 }
247 set_bit(XPRT_CWND_WAIT, &xprt->state);
248}
249
250static void
251xprt_test_and_clear_congestion_window_wait(struct rpc_xprt *xprt)
252{
253 if (!RPCXPRT_CONGESTED(xprt))
254 clear_bit(XPRT_CWND_WAIT, &xprt->state);
255}
256
257/*
258 * xprt_reserve_xprt_cong - serialize write access to transports
259 * @task: task that is requesting access to the transport
260 *
261 * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
262 * integrated into the decision of whether a request is allowed to be
263 * woken up and given access to the transport.
264 * Note that the lock is only granted if we know there are free slots.
265 */
266int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
267{
268 struct rpc_rqst *req = task->tk_rqstp;
269
270 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
271 if (task == xprt->snd_task)
272 return 1;
273 goto out_sleep;
274 }
275 if (req == NULL) {
276 xprt->snd_task = task;
277 return 1;
278 }
279 if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
280 goto out_unlock;
281 if (!xprt_need_congestion_window_wait(xprt)) {
282 xprt->snd_task = task;
283 return 1;
284 }
285out_unlock:
286 xprt_clear_locked(xprt);
287out_sleep:
288 dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
289 task->tk_status = -EAGAIN;
290 if (RPC_IS_SOFT(task))
291 rpc_sleep_on_timeout(&xprt->sending, task, NULL,
292 xprt_request_timeout(req));
293 else
294 rpc_sleep_on(&xprt->sending, task, NULL);
295 return 0;
296}
297EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
298
299static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
300{
301 int retval;
302
303 if (test_bit(XPRT_LOCKED, &xprt->state) && xprt->snd_task == task)
304 return 1;
305 spin_lock(&xprt->transport_lock);
306 retval = xprt->ops->reserve_xprt(xprt, task);
307 spin_unlock(&xprt->transport_lock);
308 return retval;
309}
310
311static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
312{
313 struct rpc_xprt *xprt = data;
314
315 xprt->snd_task = task;
316 return true;
317}
318
319static void __xprt_lock_write_next(struct rpc_xprt *xprt)
320{
321 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
322 return;
323 if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
324 goto out_unlock;
325 if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
326 __xprt_lock_write_func, xprt))
327 return;
328out_unlock:
329 xprt_clear_locked(xprt);
330}
331
332static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
333{
334 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
335 return;
336 if (test_bit(XPRT_WRITE_SPACE, &xprt->state))
337 goto out_unlock;
338 if (xprt_need_congestion_window_wait(xprt))
339 goto out_unlock;
340 if (rpc_wake_up_first_on_wq(xprtiod_workqueue, &xprt->sending,
341 __xprt_lock_write_func, xprt))
342 return;
343out_unlock:
344 xprt_clear_locked(xprt);
345}
346
347/**
348 * xprt_release_xprt - allow other requests to use a transport
349 * @xprt: transport with other tasks potentially waiting
350 * @task: task that is releasing access to the transport
351 *
352 * Note that "task" can be NULL. No congestion control is provided.
353 */
354void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
355{
356 if (xprt->snd_task == task) {
357 xprt_clear_locked(xprt);
358 __xprt_lock_write_next(xprt);
359 }
360}
361EXPORT_SYMBOL_GPL(xprt_release_xprt);
362
363/**
364 * xprt_release_xprt_cong - allow other requests to use a transport
365 * @xprt: transport with other tasks potentially waiting
366 * @task: task that is releasing access to the transport
367 *
368 * Note that "task" can be NULL. Another task is awoken to use the
369 * transport if the transport's congestion window allows it.
370 */
371void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
372{
373 if (xprt->snd_task == task) {
374 xprt_clear_locked(xprt);
375 __xprt_lock_write_next_cong(xprt);
376 }
377}
378EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
379
380static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
381{
382 if (xprt->snd_task != task)
383 return;
384 spin_lock(&xprt->transport_lock);
385 xprt->ops->release_xprt(xprt, task);
386 spin_unlock(&xprt->transport_lock);
387}
388
389/*
390 * Van Jacobson congestion avoidance. Check if the congestion window
391 * overflowed. Put the task to sleep if this is the case.
392 */
393static int
394__xprt_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
395{
396 if (req->rq_cong)
397 return 1;
398 dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
399 req->rq_task->tk_pid, xprt->cong, xprt->cwnd);
400 if (RPCXPRT_CONGESTED(xprt)) {
401 xprt_set_congestion_window_wait(xprt);
402 return 0;
403 }
404 req->rq_cong = 1;
405 xprt->cong += RPC_CWNDSCALE;
406 return 1;
407}
408
409/*
410 * Adjust the congestion window, and wake up the next task
411 * that has been sleeping due to congestion
412 */
413static void
414__xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
415{
416 if (!req->rq_cong)
417 return;
418 req->rq_cong = 0;
419 xprt->cong -= RPC_CWNDSCALE;
420 xprt_test_and_clear_congestion_window_wait(xprt);
421 __xprt_lock_write_next_cong(xprt);
422}
423
424/**
425 * xprt_request_get_cong - Request congestion control credits
426 * @xprt: pointer to transport
427 * @req: pointer to RPC request
428 *
429 * Useful for transports that require congestion control.
430 */
431bool
432xprt_request_get_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
433{
434 bool ret = false;
435
436 if (req->rq_cong)
437 return true;
438 spin_lock(&xprt->transport_lock);
439 ret = __xprt_get_cong(xprt, req) != 0;
440 spin_unlock(&xprt->transport_lock);
441 return ret;
442}
443EXPORT_SYMBOL_GPL(xprt_request_get_cong);
444
445/**
446 * xprt_release_rqst_cong - housekeeping when request is complete
447 * @task: RPC request that recently completed
448 *
449 * Useful for transports that require congestion control.
450 */
451void xprt_release_rqst_cong(struct rpc_task *task)
452{
453 struct rpc_rqst *req = task->tk_rqstp;
454
455 __xprt_put_cong(req->rq_xprt, req);
456}
457EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
458
459static void xprt_clear_congestion_window_wait_locked(struct rpc_xprt *xprt)
460{
461 if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state))
462 __xprt_lock_write_next_cong(xprt);
463}
464
465/*
466 * Clear the congestion window wait flag and wake up the next
467 * entry on xprt->sending
468 */
469static void
470xprt_clear_congestion_window_wait(struct rpc_xprt *xprt)
471{
472 if (test_and_clear_bit(XPRT_CWND_WAIT, &xprt->state)) {
473 spin_lock(&xprt->transport_lock);
474 __xprt_lock_write_next_cong(xprt);
475 spin_unlock(&xprt->transport_lock);
476 }
477}
478
479/**
480 * xprt_adjust_cwnd - adjust transport congestion window
481 * @xprt: pointer to xprt
482 * @task: recently completed RPC request used to adjust window
483 * @result: result code of completed RPC request
484 *
485 * The transport code maintains an estimate on the maximum number of out-
486 * standing RPC requests, using a smoothed version of the congestion
487 * avoidance implemented in 44BSD. This is basically the Van Jacobson
488 * congestion algorithm: If a retransmit occurs, the congestion window is
489 * halved; otherwise, it is incremented by 1/cwnd when
490 *
491 * - a reply is received and
492 * - a full number of requests are outstanding and
493 * - the congestion window hasn't been updated recently.
494 */
495void xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result)
496{
497 struct rpc_rqst *req = task->tk_rqstp;
498 unsigned long cwnd = xprt->cwnd;
499
500 if (result >= 0 && cwnd <= xprt->cong) {
501 /* The (cwnd >> 1) term makes sure
502 * the result gets rounded properly. */
503 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
504 if (cwnd > RPC_MAXCWND(xprt))
505 cwnd = RPC_MAXCWND(xprt);
506 __xprt_lock_write_next_cong(xprt);
507 } else if (result == -ETIMEDOUT) {
508 cwnd >>= 1;
509 if (cwnd < RPC_CWNDSCALE)
510 cwnd = RPC_CWNDSCALE;
511 }
512 dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
513 xprt->cong, xprt->cwnd, cwnd);
514 xprt->cwnd = cwnd;
515 __xprt_put_cong(xprt, req);
516}
517EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
518
519/**
520 * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
521 * @xprt: transport with waiting tasks
522 * @status: result code to plant in each task before waking it
523 *
524 */
525void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
526{
527 if (status < 0)
528 rpc_wake_up_status(&xprt->pending, status);
529 else
530 rpc_wake_up(&xprt->pending);
531}
532EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
533
534/**
535 * xprt_wait_for_buffer_space - wait for transport output buffer to clear
536 * @xprt: transport
537 *
538 * Note that we only set the timer for the case of RPC_IS_SOFT(), since
539 * we don't in general want to force a socket disconnection due to
540 * an incomplete RPC call transmission.
541 */
542void xprt_wait_for_buffer_space(struct rpc_xprt *xprt)
543{
544 set_bit(XPRT_WRITE_SPACE, &xprt->state);
545}
546EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
547
548static bool
549xprt_clear_write_space_locked(struct rpc_xprt *xprt)
550{
551 if (test_and_clear_bit(XPRT_WRITE_SPACE, &xprt->state)) {
552 __xprt_lock_write_next(xprt);
553 dprintk("RPC: write space: waking waiting task on "
554 "xprt %p\n", xprt);
555 return true;
556 }
557 return false;
558}
559
560/**
561 * xprt_write_space - wake the task waiting for transport output buffer space
562 * @xprt: transport with waiting tasks
563 *
564 * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
565 */
566bool xprt_write_space(struct rpc_xprt *xprt)
567{
568 bool ret;
569
570 if (!test_bit(XPRT_WRITE_SPACE, &xprt->state))
571 return false;
572 spin_lock(&xprt->transport_lock);
573 ret = xprt_clear_write_space_locked(xprt);
574 spin_unlock(&xprt->transport_lock);
575 return ret;
576}
577EXPORT_SYMBOL_GPL(xprt_write_space);
578
579static unsigned long xprt_abs_ktime_to_jiffies(ktime_t abstime)
580{
581 s64 delta = ktime_to_ns(ktime_get() - abstime);
582 return likely(delta >= 0) ?
583 jiffies - nsecs_to_jiffies(delta) :
584 jiffies + nsecs_to_jiffies(-delta);
585}
586
587static unsigned long xprt_calc_majortimeo(struct rpc_rqst *req)
588{
589 const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
590 unsigned long majortimeo = req->rq_timeout;
591
592 if (to->to_exponential)
593 majortimeo <<= to->to_retries;
594 else
595 majortimeo += to->to_increment * to->to_retries;
596 if (majortimeo > to->to_maxval || majortimeo == 0)
597 majortimeo = to->to_maxval;
598 return majortimeo;
599}
600
601static void xprt_reset_majortimeo(struct rpc_rqst *req)
602{
603 req->rq_majortimeo += xprt_calc_majortimeo(req);
604}
605
606static void xprt_init_majortimeo(struct rpc_task *task, struct rpc_rqst *req)
607{
608 unsigned long time_init;
609 struct rpc_xprt *xprt = req->rq_xprt;
610
611 if (likely(xprt && xprt_connected(xprt)))
612 time_init = jiffies;
613 else
614 time_init = xprt_abs_ktime_to_jiffies(task->tk_start);
615 req->rq_timeout = task->tk_client->cl_timeout->to_initval;
616 req->rq_majortimeo = time_init + xprt_calc_majortimeo(req);
617}
618
619/**
620 * xprt_adjust_timeout - adjust timeout values for next retransmit
621 * @req: RPC request containing parameters to use for the adjustment
622 *
623 */
624int xprt_adjust_timeout(struct rpc_rqst *req)
625{
626 struct rpc_xprt *xprt = req->rq_xprt;
627 const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
628 int status = 0;
629
630 if (time_before(jiffies, req->rq_majortimeo)) {
631 if (to->to_exponential)
632 req->rq_timeout <<= 1;
633 else
634 req->rq_timeout += to->to_increment;
635 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
636 req->rq_timeout = to->to_maxval;
637 req->rq_retries++;
638 } else {
639 req->rq_timeout = to->to_initval;
640 req->rq_retries = 0;
641 xprt_reset_majortimeo(req);
642 /* Reset the RTT counters == "slow start" */
643 spin_lock(&xprt->transport_lock);
644 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
645 spin_unlock(&xprt->transport_lock);
646 status = -ETIMEDOUT;
647 }
648
649 if (req->rq_timeout == 0) {
650 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
651 req->rq_timeout = 5 * HZ;
652 }
653 return status;
654}
655
656static void xprt_autoclose(struct work_struct *work)
657{
658 struct rpc_xprt *xprt =
659 container_of(work, struct rpc_xprt, task_cleanup);
660 unsigned int pflags = memalloc_nofs_save();
661
662 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
663 xprt->ops->close(xprt);
664 xprt_release_write(xprt, NULL);
665 wake_up_bit(&xprt->state, XPRT_LOCKED);
666 memalloc_nofs_restore(pflags);
667}
668
669/**
670 * xprt_disconnect_done - mark a transport as disconnected
671 * @xprt: transport to flag for disconnect
672 *
673 */
674void xprt_disconnect_done(struct rpc_xprt *xprt)
675{
676 dprintk("RPC: disconnected transport %p\n", xprt);
677 spin_lock(&xprt->transport_lock);
678 xprt_clear_connected(xprt);
679 xprt_clear_write_space_locked(xprt);
680 xprt_clear_congestion_window_wait_locked(xprt);
681 xprt_wake_pending_tasks(xprt, -ENOTCONN);
682 spin_unlock(&xprt->transport_lock);
683}
684EXPORT_SYMBOL_GPL(xprt_disconnect_done);
685
686/**
687 * xprt_force_disconnect - force a transport to disconnect
688 * @xprt: transport to disconnect
689 *
690 */
691void xprt_force_disconnect(struct rpc_xprt *xprt)
692{
693 /* Don't race with the test_bit() in xprt_clear_locked() */
694 spin_lock(&xprt->transport_lock);
695 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
696 /* Try to schedule an autoclose RPC call */
697 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
698 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
699 else if (xprt->snd_task)
700 rpc_wake_up_queued_task_set_status(&xprt->pending,
701 xprt->snd_task, -ENOTCONN);
702 spin_unlock(&xprt->transport_lock);
703}
704EXPORT_SYMBOL_GPL(xprt_force_disconnect);
705
706static unsigned int
707xprt_connect_cookie(struct rpc_xprt *xprt)
708{
709 return READ_ONCE(xprt->connect_cookie);
710}
711
712static bool
713xprt_request_retransmit_after_disconnect(struct rpc_task *task)
714{
715 struct rpc_rqst *req = task->tk_rqstp;
716 struct rpc_xprt *xprt = req->rq_xprt;
717
718 return req->rq_connect_cookie != xprt_connect_cookie(xprt) ||
719 !xprt_connected(xprt);
720}
721
722/**
723 * xprt_conditional_disconnect - force a transport to disconnect
724 * @xprt: transport to disconnect
725 * @cookie: 'connection cookie'
726 *
727 * This attempts to break the connection if and only if 'cookie' matches
728 * the current transport 'connection cookie'. It ensures that we don't
729 * try to break the connection more than once when we need to retransmit
730 * a batch of RPC requests.
731 *
732 */
733void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
734{
735 /* Don't race with the test_bit() in xprt_clear_locked() */
736 spin_lock(&xprt->transport_lock);
737 if (cookie != xprt->connect_cookie)
738 goto out;
739 if (test_bit(XPRT_CLOSING, &xprt->state))
740 goto out;
741 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
742 /* Try to schedule an autoclose RPC call */
743 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
744 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
745 xprt_wake_pending_tasks(xprt, -EAGAIN);
746out:
747 spin_unlock(&xprt->transport_lock);
748}
749
750static bool
751xprt_has_timer(const struct rpc_xprt *xprt)
752{
753 return xprt->idle_timeout != 0;
754}
755
756static void
757xprt_schedule_autodisconnect(struct rpc_xprt *xprt)
758 __must_hold(&xprt->transport_lock)
759{
760 xprt->last_used = jiffies;
761 if (RB_EMPTY_ROOT(&xprt->recv_queue) && xprt_has_timer(xprt))
762 mod_timer(&xprt->timer, xprt->last_used + xprt->idle_timeout);
763}
764
765static void
766xprt_init_autodisconnect(struct timer_list *t)
767{
768 struct rpc_xprt *xprt = from_timer(xprt, t, timer);
769
770 if (!RB_EMPTY_ROOT(&xprt->recv_queue))
771 return;
772 /* Reset xprt->last_used to avoid connect/autodisconnect cycling */
773 xprt->last_used = jiffies;
774 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
775 return;
776 queue_work(xprtiod_workqueue, &xprt->task_cleanup);
777}
778
779bool xprt_lock_connect(struct rpc_xprt *xprt,
780 struct rpc_task *task,
781 void *cookie)
782{
783 bool ret = false;
784
785 spin_lock(&xprt->transport_lock);
786 if (!test_bit(XPRT_LOCKED, &xprt->state))
787 goto out;
788 if (xprt->snd_task != task)
789 goto out;
790 xprt->snd_task = cookie;
791 ret = true;
792out:
793 spin_unlock(&xprt->transport_lock);
794 return ret;
795}
796
797void xprt_unlock_connect(struct rpc_xprt *xprt, void *cookie)
798{
799 spin_lock(&xprt->transport_lock);
800 if (xprt->snd_task != cookie)
801 goto out;
802 if (!test_bit(XPRT_LOCKED, &xprt->state))
803 goto out;
804 xprt->snd_task =NULL;
805 xprt->ops->release_xprt(xprt, NULL);
806 xprt_schedule_autodisconnect(xprt);
807out:
808 spin_unlock(&xprt->transport_lock);
809 wake_up_bit(&xprt->state, XPRT_LOCKED);
810}
811
812/**
813 * xprt_connect - schedule a transport connect operation
814 * @task: RPC task that is requesting the connect
815 *
816 */
817void xprt_connect(struct rpc_task *task)
818{
819 struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
820
821 dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
822 xprt, (xprt_connected(xprt) ? "is" : "is not"));
823
824 if (!xprt_bound(xprt)) {
825 task->tk_status = -EAGAIN;
826 return;
827 }
828 if (!xprt_lock_write(xprt, task))
829 return;
830
831 if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
832 xprt->ops->close(xprt);
833
834 if (!xprt_connected(xprt)) {
835 task->tk_rqstp->rq_connect_cookie = xprt->connect_cookie;
836 rpc_sleep_on_timeout(&xprt->pending, task, NULL,
837 xprt_request_timeout(task->tk_rqstp));
838
839 if (test_bit(XPRT_CLOSING, &xprt->state))
840 return;
841 if (xprt_test_and_set_connecting(xprt))
842 return;
843 /* Race breaker */
844 if (!xprt_connected(xprt)) {
845 xprt->stat.connect_start = jiffies;
846 xprt->ops->connect(xprt, task);
847 } else {
848 xprt_clear_connecting(xprt);
849 task->tk_status = 0;
850 rpc_wake_up_queued_task(&xprt->pending, task);
851 }
852 }
853 xprt_release_write(xprt, task);
854}
855
856/**
857 * xprt_reconnect_delay - compute the wait before scheduling a connect
858 * @xprt: transport instance
859 *
860 */
861unsigned long xprt_reconnect_delay(const struct rpc_xprt *xprt)
862{
863 unsigned long start, now = jiffies;
864
865 start = xprt->stat.connect_start + xprt->reestablish_timeout;
866 if (time_after(start, now))
867 return start - now;
868 return 0;
869}
870EXPORT_SYMBOL_GPL(xprt_reconnect_delay);
871
872/**
873 * xprt_reconnect_backoff - compute the new re-establish timeout
874 * @xprt: transport instance
875 * @init_to: initial reestablish timeout
876 *
877 */
878void xprt_reconnect_backoff(struct rpc_xprt *xprt, unsigned long init_to)
879{
880 xprt->reestablish_timeout <<= 1;
881 if (xprt->reestablish_timeout > xprt->max_reconnect_timeout)
882 xprt->reestablish_timeout = xprt->max_reconnect_timeout;
883 if (xprt->reestablish_timeout < init_to)
884 xprt->reestablish_timeout = init_to;
885}
886EXPORT_SYMBOL_GPL(xprt_reconnect_backoff);
887
888enum xprt_xid_rb_cmp {
889 XID_RB_EQUAL,
890 XID_RB_LEFT,
891 XID_RB_RIGHT,
892};
893static enum xprt_xid_rb_cmp
894xprt_xid_cmp(__be32 xid1, __be32 xid2)
895{
896 if (xid1 == xid2)
897 return XID_RB_EQUAL;
898 if ((__force u32)xid1 < (__force u32)xid2)
899 return XID_RB_LEFT;
900 return XID_RB_RIGHT;
901}
902
903static struct rpc_rqst *
904xprt_request_rb_find(struct rpc_xprt *xprt, __be32 xid)
905{
906 struct rb_node *n = xprt->recv_queue.rb_node;
907 struct rpc_rqst *req;
908
909 while (n != NULL) {
910 req = rb_entry(n, struct rpc_rqst, rq_recv);
911 switch (xprt_xid_cmp(xid, req->rq_xid)) {
912 case XID_RB_LEFT:
913 n = n->rb_left;
914 break;
915 case XID_RB_RIGHT:
916 n = n->rb_right;
917 break;
918 case XID_RB_EQUAL:
919 return req;
920 }
921 }
922 return NULL;
923}
924
925static void
926xprt_request_rb_insert(struct rpc_xprt *xprt, struct rpc_rqst *new)
927{
928 struct rb_node **p = &xprt->recv_queue.rb_node;
929 struct rb_node *n = NULL;
930 struct rpc_rqst *req;
931
932 while (*p != NULL) {
933 n = *p;
934 req = rb_entry(n, struct rpc_rqst, rq_recv);
935 switch(xprt_xid_cmp(new->rq_xid, req->rq_xid)) {
936 case XID_RB_LEFT:
937 p = &n->rb_left;
938 break;
939 case XID_RB_RIGHT:
940 p = &n->rb_right;
941 break;
942 case XID_RB_EQUAL:
943 WARN_ON_ONCE(new != req);
944 return;
945 }
946 }
947 rb_link_node(&new->rq_recv, n, p);
948 rb_insert_color(&new->rq_recv, &xprt->recv_queue);
949}
950
951static void
952xprt_request_rb_remove(struct rpc_xprt *xprt, struct rpc_rqst *req)
953{
954 rb_erase(&req->rq_recv, &xprt->recv_queue);
955}
956
957/**
958 * xprt_lookup_rqst - find an RPC request corresponding to an XID
959 * @xprt: transport on which the original request was transmitted
960 * @xid: RPC XID of incoming reply
961 *
962 * Caller holds xprt->queue_lock.
963 */
964struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
965{
966 struct rpc_rqst *entry;
967
968 entry = xprt_request_rb_find(xprt, xid);
969 if (entry != NULL) {
970 trace_xprt_lookup_rqst(xprt, xid, 0);
971 entry->rq_rtt = ktime_sub(ktime_get(), entry->rq_xtime);
972 return entry;
973 }
974
975 dprintk("RPC: xprt_lookup_rqst did not find xid %08x\n",
976 ntohl(xid));
977 trace_xprt_lookup_rqst(xprt, xid, -ENOENT);
978 xprt->stat.bad_xids++;
979 return NULL;
980}
981EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
982
983static bool
984xprt_is_pinned_rqst(struct rpc_rqst *req)
985{
986 return atomic_read(&req->rq_pin) != 0;
987}
988
989/**
990 * xprt_pin_rqst - Pin a request on the transport receive list
991 * @req: Request to pin
992 *
993 * Caller must ensure this is atomic with the call to xprt_lookup_rqst()
994 * so should be holding xprt->queue_lock.
995 */
996void xprt_pin_rqst(struct rpc_rqst *req)
997{
998 atomic_inc(&req->rq_pin);
999}
1000EXPORT_SYMBOL_GPL(xprt_pin_rqst);
1001
1002/**
1003 * xprt_unpin_rqst - Unpin a request on the transport receive list
1004 * @req: Request to pin
1005 *
1006 * Caller should be holding xprt->queue_lock.
1007 */
1008void xprt_unpin_rqst(struct rpc_rqst *req)
1009{
1010 if (!test_bit(RPC_TASK_MSG_PIN_WAIT, &req->rq_task->tk_runstate)) {
1011 atomic_dec(&req->rq_pin);
1012 return;
1013 }
1014 if (atomic_dec_and_test(&req->rq_pin))
1015 wake_up_var(&req->rq_pin);
1016}
1017EXPORT_SYMBOL_GPL(xprt_unpin_rqst);
1018
1019static void xprt_wait_on_pinned_rqst(struct rpc_rqst *req)
1020{
1021 wait_var_event(&req->rq_pin, !xprt_is_pinned_rqst(req));
1022}
1023
1024static bool
1025xprt_request_data_received(struct rpc_task *task)
1026{
1027 return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1028 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) != 0;
1029}
1030
1031static bool
1032xprt_request_need_enqueue_receive(struct rpc_task *task, struct rpc_rqst *req)
1033{
1034 return !test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) &&
1035 READ_ONCE(task->tk_rqstp->rq_reply_bytes_recvd) == 0;
1036}
1037
1038/**
1039 * xprt_request_enqueue_receive - Add an request to the receive queue
1040 * @task: RPC task
1041 *
1042 */
1043void
1044xprt_request_enqueue_receive(struct rpc_task *task)
1045{
1046 struct rpc_rqst *req = task->tk_rqstp;
1047 struct rpc_xprt *xprt = req->rq_xprt;
1048
1049 if (!xprt_request_need_enqueue_receive(task, req))
1050 return;
1051
1052 xprt_request_prepare(task->tk_rqstp);
1053 spin_lock(&xprt->queue_lock);
1054
1055 /* Update the softirq receive buffer */
1056 memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
1057 sizeof(req->rq_private_buf));
1058
1059 /* Add request to the receive list */
1060 xprt_request_rb_insert(xprt, req);
1061 set_bit(RPC_TASK_NEED_RECV, &task->tk_runstate);
1062 spin_unlock(&xprt->queue_lock);
1063
1064 /* Turn off autodisconnect */
1065 del_singleshot_timer_sync(&xprt->timer);
1066}
1067
1068/**
1069 * xprt_request_dequeue_receive_locked - Remove a request from the receive queue
1070 * @task: RPC task
1071 *
1072 * Caller must hold xprt->queue_lock.
1073 */
1074static void
1075xprt_request_dequeue_receive_locked(struct rpc_task *task)
1076{
1077 struct rpc_rqst *req = task->tk_rqstp;
1078
1079 if (test_and_clear_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1080 xprt_request_rb_remove(req->rq_xprt, req);
1081}
1082
1083/**
1084 * xprt_update_rtt - Update RPC RTT statistics
1085 * @task: RPC request that recently completed
1086 *
1087 * Caller holds xprt->queue_lock.
1088 */
1089void xprt_update_rtt(struct rpc_task *task)
1090{
1091 struct rpc_rqst *req = task->tk_rqstp;
1092 struct rpc_rtt *rtt = task->tk_client->cl_rtt;
1093 unsigned int timer = task->tk_msg.rpc_proc->p_timer;
1094 long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
1095
1096 if (timer) {
1097 if (req->rq_ntrans == 1)
1098 rpc_update_rtt(rtt, timer, m);
1099 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
1100 }
1101}
1102EXPORT_SYMBOL_GPL(xprt_update_rtt);
1103
1104/**
1105 * xprt_complete_rqst - called when reply processing is complete
1106 * @task: RPC request that recently completed
1107 * @copied: actual number of bytes received from the transport
1108 *
1109 * Caller holds xprt->queue_lock.
1110 */
1111void xprt_complete_rqst(struct rpc_task *task, int copied)
1112{
1113 struct rpc_rqst *req = task->tk_rqstp;
1114 struct rpc_xprt *xprt = req->rq_xprt;
1115
1116 dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
1117 task->tk_pid, ntohl(req->rq_xid), copied);
1118 trace_xprt_complete_rqst(xprt, req->rq_xid, copied);
1119
1120 xprt->stat.recvs++;
1121
1122 req->rq_private_buf.len = copied;
1123 /* Ensure all writes are done before we update */
1124 /* req->rq_reply_bytes_recvd */
1125 smp_wmb();
1126 req->rq_reply_bytes_recvd = copied;
1127 xprt_request_dequeue_receive_locked(task);
1128 rpc_wake_up_queued_task(&xprt->pending, task);
1129}
1130EXPORT_SYMBOL_GPL(xprt_complete_rqst);
1131
1132static void xprt_timer(struct rpc_task *task)
1133{
1134 struct rpc_rqst *req = task->tk_rqstp;
1135 struct rpc_xprt *xprt = req->rq_xprt;
1136
1137 if (task->tk_status != -ETIMEDOUT)
1138 return;
1139
1140 trace_xprt_timer(xprt, req->rq_xid, task->tk_status);
1141 if (!req->rq_reply_bytes_recvd) {
1142 if (xprt->ops->timer)
1143 xprt->ops->timer(xprt, task);
1144 } else
1145 task->tk_status = 0;
1146}
1147
1148/**
1149 * xprt_wait_for_reply_request_def - wait for reply
1150 * @task: pointer to rpc_task
1151 *
1152 * Set a request's retransmit timeout based on the transport's
1153 * default timeout parameters. Used by transports that don't adjust
1154 * the retransmit timeout based on round-trip time estimation,
1155 * and put the task to sleep on the pending queue.
1156 */
1157void xprt_wait_for_reply_request_def(struct rpc_task *task)
1158{
1159 struct rpc_rqst *req = task->tk_rqstp;
1160
1161 rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1162 xprt_request_timeout(req));
1163}
1164EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_def);
1165
1166/**
1167 * xprt_wait_for_reply_request_rtt - wait for reply using RTT estimator
1168 * @task: pointer to rpc_task
1169 *
1170 * Set a request's retransmit timeout using the RTT estimator,
1171 * and put the task to sleep on the pending queue.
1172 */
1173void xprt_wait_for_reply_request_rtt(struct rpc_task *task)
1174{
1175 int timer = task->tk_msg.rpc_proc->p_timer;
1176 struct rpc_clnt *clnt = task->tk_client;
1177 struct rpc_rtt *rtt = clnt->cl_rtt;
1178 struct rpc_rqst *req = task->tk_rqstp;
1179 unsigned long max_timeout = clnt->cl_timeout->to_maxval;
1180 unsigned long timeout;
1181
1182 timeout = rpc_calc_rto(rtt, timer);
1183 timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
1184 if (timeout > max_timeout || timeout == 0)
1185 timeout = max_timeout;
1186 rpc_sleep_on_timeout(&req->rq_xprt->pending, task, xprt_timer,
1187 jiffies + timeout);
1188}
1189EXPORT_SYMBOL_GPL(xprt_wait_for_reply_request_rtt);
1190
1191/**
1192 * xprt_request_wait_receive - wait for the reply to an RPC request
1193 * @task: RPC task about to send a request
1194 *
1195 */
1196void xprt_request_wait_receive(struct rpc_task *task)
1197{
1198 struct rpc_rqst *req = task->tk_rqstp;
1199 struct rpc_xprt *xprt = req->rq_xprt;
1200
1201 if (!test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate))
1202 return;
1203 /*
1204 * Sleep on the pending queue if we're expecting a reply.
1205 * The spinlock ensures atomicity between the test of
1206 * req->rq_reply_bytes_recvd, and the call to rpc_sleep_on().
1207 */
1208 spin_lock(&xprt->queue_lock);
1209 if (test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate)) {
1210 xprt->ops->wait_for_reply_request(task);
1211 /*
1212 * Send an extra queue wakeup call if the
1213 * connection was dropped in case the call to
1214 * rpc_sleep_on() raced.
1215 */
1216 if (xprt_request_retransmit_after_disconnect(task))
1217 rpc_wake_up_queued_task_set_status(&xprt->pending,
1218 task, -ENOTCONN);
1219 }
1220 spin_unlock(&xprt->queue_lock);
1221}
1222
1223static bool
1224xprt_request_need_enqueue_transmit(struct rpc_task *task, struct rpc_rqst *req)
1225{
1226 return !test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1227}
1228
1229/**
1230 * xprt_request_enqueue_transmit - queue a task for transmission
1231 * @task: pointer to rpc_task
1232 *
1233 * Add a task to the transmission queue.
1234 */
1235void
1236xprt_request_enqueue_transmit(struct rpc_task *task)
1237{
1238 struct rpc_rqst *pos, *req = task->tk_rqstp;
1239 struct rpc_xprt *xprt = req->rq_xprt;
1240
1241 if (xprt_request_need_enqueue_transmit(task, req)) {
1242 req->rq_bytes_sent = 0;
1243 spin_lock(&xprt->queue_lock);
1244 /*
1245 * Requests that carry congestion control credits are added
1246 * to the head of the list to avoid starvation issues.
1247 */
1248 if (req->rq_cong) {
1249 xprt_clear_congestion_window_wait(xprt);
1250 list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1251 if (pos->rq_cong)
1252 continue;
1253 /* Note: req is added _before_ pos */
1254 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1255 INIT_LIST_HEAD(&req->rq_xmit2);
1256 trace_xprt_enq_xmit(task, 1);
1257 goto out;
1258 }
1259 } else if (RPC_IS_SWAPPER(task)) {
1260 list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1261 if (pos->rq_cong || pos->rq_bytes_sent)
1262 continue;
1263 if (RPC_IS_SWAPPER(pos->rq_task))
1264 continue;
1265 /* Note: req is added _before_ pos */
1266 list_add_tail(&req->rq_xmit, &pos->rq_xmit);
1267 INIT_LIST_HEAD(&req->rq_xmit2);
1268 trace_xprt_enq_xmit(task, 2);
1269 goto out;
1270 }
1271 } else if (!req->rq_seqno) {
1272 list_for_each_entry(pos, &xprt->xmit_queue, rq_xmit) {
1273 if (pos->rq_task->tk_owner != task->tk_owner)
1274 continue;
1275 list_add_tail(&req->rq_xmit2, &pos->rq_xmit2);
1276 INIT_LIST_HEAD(&req->rq_xmit);
1277 trace_xprt_enq_xmit(task, 3);
1278 goto out;
1279 }
1280 }
1281 list_add_tail(&req->rq_xmit, &xprt->xmit_queue);
1282 INIT_LIST_HEAD(&req->rq_xmit2);
1283 trace_xprt_enq_xmit(task, 4);
1284out:
1285 set_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate);
1286 spin_unlock(&xprt->queue_lock);
1287 }
1288}
1289
1290/**
1291 * xprt_request_dequeue_transmit_locked - remove a task from the transmission queue
1292 * @task: pointer to rpc_task
1293 *
1294 * Remove a task from the transmission queue
1295 * Caller must hold xprt->queue_lock
1296 */
1297static void
1298xprt_request_dequeue_transmit_locked(struct rpc_task *task)
1299{
1300 struct rpc_rqst *req = task->tk_rqstp;
1301
1302 if (!test_and_clear_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1303 return;
1304 if (!list_empty(&req->rq_xmit)) {
1305 list_del(&req->rq_xmit);
1306 if (!list_empty(&req->rq_xmit2)) {
1307 struct rpc_rqst *next = list_first_entry(&req->rq_xmit2,
1308 struct rpc_rqst, rq_xmit2);
1309 list_del(&req->rq_xmit2);
1310 list_add_tail(&next->rq_xmit, &next->rq_xprt->xmit_queue);
1311 }
1312 } else
1313 list_del(&req->rq_xmit2);
1314}
1315
1316/**
1317 * xprt_request_dequeue_transmit - remove a task from the transmission queue
1318 * @task: pointer to rpc_task
1319 *
1320 * Remove a task from the transmission queue
1321 */
1322static void
1323xprt_request_dequeue_transmit(struct rpc_task *task)
1324{
1325 struct rpc_rqst *req = task->tk_rqstp;
1326 struct rpc_xprt *xprt = req->rq_xprt;
1327
1328 spin_lock(&xprt->queue_lock);
1329 xprt_request_dequeue_transmit_locked(task);
1330 spin_unlock(&xprt->queue_lock);
1331}
1332
1333/**
1334 * xprt_request_dequeue_xprt - remove a task from the transmit+receive queue
1335 * @task: pointer to rpc_task
1336 *
1337 * Remove a task from the transmit and receive queues, and ensure that
1338 * it is not pinned by the receive work item.
1339 */
1340void
1341xprt_request_dequeue_xprt(struct rpc_task *task)
1342{
1343 struct rpc_rqst *req = task->tk_rqstp;
1344 struct rpc_xprt *xprt = req->rq_xprt;
1345
1346 if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate) ||
1347 test_bit(RPC_TASK_NEED_RECV, &task->tk_runstate) ||
1348 xprt_is_pinned_rqst(req)) {
1349 spin_lock(&xprt->queue_lock);
1350 xprt_request_dequeue_transmit_locked(task);
1351 xprt_request_dequeue_receive_locked(task);
1352 while (xprt_is_pinned_rqst(req)) {
1353 set_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1354 spin_unlock(&xprt->queue_lock);
1355 xprt_wait_on_pinned_rqst(req);
1356 spin_lock(&xprt->queue_lock);
1357 clear_bit(RPC_TASK_MSG_PIN_WAIT, &task->tk_runstate);
1358 }
1359 spin_unlock(&xprt->queue_lock);
1360 }
1361}
1362
1363/**
1364 * xprt_request_prepare - prepare an encoded request for transport
1365 * @req: pointer to rpc_rqst
1366 *
1367 * Calls into the transport layer to do whatever is needed to prepare
1368 * the request for transmission or receive.
1369 */
1370void
1371xprt_request_prepare(struct rpc_rqst *req)
1372{
1373 struct rpc_xprt *xprt = req->rq_xprt;
1374
1375 if (xprt->ops->prepare_request)
1376 xprt->ops->prepare_request(req);
1377}
1378
1379/**
1380 * xprt_request_need_retransmit - Test if a task needs retransmission
1381 * @task: pointer to rpc_task
1382 *
1383 * Test for whether a connection breakage requires the task to retransmit
1384 */
1385bool
1386xprt_request_need_retransmit(struct rpc_task *task)
1387{
1388 return xprt_request_retransmit_after_disconnect(task);
1389}
1390
1391/**
1392 * xprt_prepare_transmit - reserve the transport before sending a request
1393 * @task: RPC task about to send a request
1394 *
1395 */
1396bool xprt_prepare_transmit(struct rpc_task *task)
1397{
1398 struct rpc_rqst *req = task->tk_rqstp;
1399 struct rpc_xprt *xprt = req->rq_xprt;
1400
1401 dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
1402
1403 if (!xprt_lock_write(xprt, task)) {
1404 /* Race breaker: someone may have transmitted us */
1405 if (!test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1406 rpc_wake_up_queued_task_set_status(&xprt->sending,
1407 task, 0);
1408 return false;
1409
1410 }
1411 return true;
1412}
1413
1414void xprt_end_transmit(struct rpc_task *task)
1415{
1416 xprt_release_write(task->tk_rqstp->rq_xprt, task);
1417}
1418
1419/**
1420 * xprt_request_transmit - send an RPC request on a transport
1421 * @req: pointer to request to transmit
1422 * @snd_task: RPC task that owns the transport lock
1423 *
1424 * This performs the transmission of a single request.
1425 * Note that if the request is not the same as snd_task, then it
1426 * does need to be pinned.
1427 * Returns '0' on success.
1428 */
1429static int
1430xprt_request_transmit(struct rpc_rqst *req, struct rpc_task *snd_task)
1431{
1432 struct rpc_xprt *xprt = req->rq_xprt;
1433 struct rpc_task *task = req->rq_task;
1434 unsigned int connect_cookie;
1435 int is_retrans = RPC_WAS_SENT(task);
1436 int status;
1437
1438 if (!req->rq_bytes_sent) {
1439 if (xprt_request_data_received(task)) {
1440 status = 0;
1441 goto out_dequeue;
1442 }
1443 /* Verify that our message lies in the RPCSEC_GSS window */
1444 if (rpcauth_xmit_need_reencode(task)) {
1445 status = -EBADMSG;
1446 goto out_dequeue;
1447 }
1448 if (RPC_SIGNALLED(task)) {
1449 status = -ERESTARTSYS;
1450 goto out_dequeue;
1451 }
1452 }
1453
1454 /*
1455 * Update req->rq_ntrans before transmitting to avoid races with
1456 * xprt_update_rtt(), which needs to know that it is recording a
1457 * reply to the first transmission.
1458 */
1459 req->rq_ntrans++;
1460
1461 connect_cookie = xprt->connect_cookie;
1462 status = xprt->ops->send_request(req);
1463 if (status != 0) {
1464 req->rq_ntrans--;
1465 trace_xprt_transmit(req, status);
1466 return status;
1467 }
1468
1469 if (is_retrans)
1470 task->tk_client->cl_stats->rpcretrans++;
1471
1472 xprt_inject_disconnect(xprt);
1473
1474 task->tk_flags |= RPC_TASK_SENT;
1475 spin_lock(&xprt->transport_lock);
1476
1477 xprt->stat.sends++;
1478 xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
1479 xprt->stat.bklog_u += xprt->backlog.qlen;
1480 xprt->stat.sending_u += xprt->sending.qlen;
1481 xprt->stat.pending_u += xprt->pending.qlen;
1482 spin_unlock(&xprt->transport_lock);
1483
1484 req->rq_connect_cookie = connect_cookie;
1485out_dequeue:
1486 trace_xprt_transmit(req, status);
1487 xprt_request_dequeue_transmit(task);
1488 rpc_wake_up_queued_task_set_status(&xprt->sending, task, status);
1489 return status;
1490}
1491
1492/**
1493 * xprt_transmit - send an RPC request on a transport
1494 * @task: controlling RPC task
1495 *
1496 * Attempts to drain the transmit queue. On exit, either the transport
1497 * signalled an error that needs to be handled before transmission can
1498 * resume, or @task finished transmitting, and detected that it already
1499 * received a reply.
1500 */
1501void
1502xprt_transmit(struct rpc_task *task)
1503{
1504 struct rpc_rqst *next, *req = task->tk_rqstp;
1505 struct rpc_xprt *xprt = req->rq_xprt;
1506 int status;
1507
1508 spin_lock(&xprt->queue_lock);
1509 while (!list_empty(&xprt->xmit_queue)) {
1510 next = list_first_entry(&xprt->xmit_queue,
1511 struct rpc_rqst, rq_xmit);
1512 xprt_pin_rqst(next);
1513 spin_unlock(&xprt->queue_lock);
1514 status = xprt_request_transmit(next, task);
1515 if (status == -EBADMSG && next != req)
1516 status = 0;
1517 cond_resched();
1518 spin_lock(&xprt->queue_lock);
1519 xprt_unpin_rqst(next);
1520 if (status == 0) {
1521 if (!xprt_request_data_received(task) ||
1522 test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1523 continue;
1524 } else if (test_bit(RPC_TASK_NEED_XMIT, &task->tk_runstate))
1525 task->tk_status = status;
1526 break;
1527 }
1528 spin_unlock(&xprt->queue_lock);
1529}
1530
1531static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
1532{
1533 set_bit(XPRT_CONGESTED, &xprt->state);
1534 rpc_sleep_on(&xprt->backlog, task, NULL);
1535}
1536
1537static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
1538{
1539 if (rpc_wake_up_next(&xprt->backlog) == NULL)
1540 clear_bit(XPRT_CONGESTED, &xprt->state);
1541}
1542
1543static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
1544{
1545 bool ret = false;
1546
1547 if (!test_bit(XPRT_CONGESTED, &xprt->state))
1548 goto out;
1549 spin_lock(&xprt->reserve_lock);
1550 if (test_bit(XPRT_CONGESTED, &xprt->state)) {
1551 rpc_sleep_on(&xprt->backlog, task, NULL);
1552 ret = true;
1553 }
1554 spin_unlock(&xprt->reserve_lock);
1555out:
1556 return ret;
1557}
1558
1559static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt)
1560{
1561 struct rpc_rqst *req = ERR_PTR(-EAGAIN);
1562
1563 if (xprt->num_reqs >= xprt->max_reqs)
1564 goto out;
1565 ++xprt->num_reqs;
1566 spin_unlock(&xprt->reserve_lock);
1567 req = kzalloc(sizeof(struct rpc_rqst), GFP_NOFS);
1568 spin_lock(&xprt->reserve_lock);
1569 if (req != NULL)
1570 goto out;
1571 --xprt->num_reqs;
1572 req = ERR_PTR(-ENOMEM);
1573out:
1574 return req;
1575}
1576
1577static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1578{
1579 if (xprt->num_reqs > xprt->min_reqs) {
1580 --xprt->num_reqs;
1581 kfree(req);
1582 return true;
1583 }
1584 return false;
1585}
1586
1587void xprt_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
1588{
1589 struct rpc_rqst *req;
1590
1591 spin_lock(&xprt->reserve_lock);
1592 if (!list_empty(&xprt->free)) {
1593 req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
1594 list_del(&req->rq_list);
1595 goto out_init_req;
1596 }
1597 req = xprt_dynamic_alloc_slot(xprt);
1598 if (!IS_ERR(req))
1599 goto out_init_req;
1600 switch (PTR_ERR(req)) {
1601 case -ENOMEM:
1602 dprintk("RPC: dynamic allocation of request slot "
1603 "failed! Retrying\n");
1604 task->tk_status = -ENOMEM;
1605 break;
1606 case -EAGAIN:
1607 xprt_add_backlog(xprt, task);
1608 dprintk("RPC: waiting for request slot\n");
1609 /* fall through */
1610 default:
1611 task->tk_status = -EAGAIN;
1612 }
1613 spin_unlock(&xprt->reserve_lock);
1614 return;
1615out_init_req:
1616 xprt->stat.max_slots = max_t(unsigned int, xprt->stat.max_slots,
1617 xprt->num_reqs);
1618 spin_unlock(&xprt->reserve_lock);
1619
1620 task->tk_status = 0;
1621 task->tk_rqstp = req;
1622}
1623EXPORT_SYMBOL_GPL(xprt_alloc_slot);
1624
1625void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1626{
1627 spin_lock(&xprt->reserve_lock);
1628 if (!xprt_dynamic_free_slot(xprt, req)) {
1629 memset(req, 0, sizeof(*req)); /* mark unused */
1630 list_add(&req->rq_list, &xprt->free);
1631 }
1632 xprt_wake_up_backlog(xprt);
1633 spin_unlock(&xprt->reserve_lock);
1634}
1635EXPORT_SYMBOL_GPL(xprt_free_slot);
1636
1637static void xprt_free_all_slots(struct rpc_xprt *xprt)
1638{
1639 struct rpc_rqst *req;
1640 while (!list_empty(&xprt->free)) {
1641 req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1642 list_del(&req->rq_list);
1643 kfree(req);
1644 }
1645}
1646
1647struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1648 unsigned int num_prealloc,
1649 unsigned int max_alloc)
1650{
1651 struct rpc_xprt *xprt;
1652 struct rpc_rqst *req;
1653 int i;
1654
1655 xprt = kzalloc(size, GFP_KERNEL);
1656 if (xprt == NULL)
1657 goto out;
1658
1659 xprt_init(xprt, net);
1660
1661 for (i = 0; i < num_prealloc; i++) {
1662 req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1663 if (!req)
1664 goto out_free;
1665 list_add(&req->rq_list, &xprt->free);
1666 }
1667 if (max_alloc > num_prealloc)
1668 xprt->max_reqs = max_alloc;
1669 else
1670 xprt->max_reqs = num_prealloc;
1671 xprt->min_reqs = num_prealloc;
1672 xprt->num_reqs = num_prealloc;
1673
1674 return xprt;
1675
1676out_free:
1677 xprt_free(xprt);
1678out:
1679 return NULL;
1680}
1681EXPORT_SYMBOL_GPL(xprt_alloc);
1682
1683void xprt_free(struct rpc_xprt *xprt)
1684{
1685 put_net(xprt->xprt_net);
1686 xprt_free_all_slots(xprt);
1687 kfree_rcu(xprt, rcu);
1688}
1689EXPORT_SYMBOL_GPL(xprt_free);
1690
1691static void
1692xprt_init_connect_cookie(struct rpc_rqst *req, struct rpc_xprt *xprt)
1693{
1694 req->rq_connect_cookie = xprt_connect_cookie(xprt) - 1;
1695}
1696
1697static __be32
1698xprt_alloc_xid(struct rpc_xprt *xprt)
1699{
1700 __be32 xid;
1701
1702 spin_lock(&xprt->reserve_lock);
1703 xid = (__force __be32)xprt->xid++;
1704 spin_unlock(&xprt->reserve_lock);
1705 return xid;
1706}
1707
1708static void
1709xprt_init_xid(struct rpc_xprt *xprt)
1710{
1711 xprt->xid = prandom_u32();
1712}
1713
1714static void
1715xprt_request_init(struct rpc_task *task)
1716{
1717 struct rpc_xprt *xprt = task->tk_xprt;
1718 struct rpc_rqst *req = task->tk_rqstp;
1719
1720 req->rq_task = task;
1721 req->rq_xprt = xprt;
1722 req->rq_buffer = NULL;
1723 req->rq_xid = xprt_alloc_xid(xprt);
1724 xprt_init_connect_cookie(req, xprt);
1725 req->rq_snd_buf.len = 0;
1726 req->rq_snd_buf.buflen = 0;
1727 req->rq_rcv_buf.len = 0;
1728 req->rq_rcv_buf.buflen = 0;
1729 req->rq_snd_buf.bvec = NULL;
1730 req->rq_rcv_buf.bvec = NULL;
1731 req->rq_release_snd_buf = NULL;
1732 xprt_init_majortimeo(task, req);
1733 dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1734 req, ntohl(req->rq_xid));
1735}
1736
1737static void
1738xprt_do_reserve(struct rpc_xprt *xprt, struct rpc_task *task)
1739{
1740 xprt->ops->alloc_slot(xprt, task);
1741 if (task->tk_rqstp != NULL)
1742 xprt_request_init(task);
1743}
1744
1745/**
1746 * xprt_reserve - allocate an RPC request slot
1747 * @task: RPC task requesting a slot allocation
1748 *
1749 * If the transport is marked as being congested, or if no more
1750 * slots are available, place the task on the transport's
1751 * backlog queue.
1752 */
1753void xprt_reserve(struct rpc_task *task)
1754{
1755 struct rpc_xprt *xprt = task->tk_xprt;
1756
1757 task->tk_status = 0;
1758 if (task->tk_rqstp != NULL)
1759 return;
1760
1761 task->tk_status = -EAGAIN;
1762 if (!xprt_throttle_congested(xprt, task))
1763 xprt_do_reserve(xprt, task);
1764}
1765
1766/**
1767 * xprt_retry_reserve - allocate an RPC request slot
1768 * @task: RPC task requesting a slot allocation
1769 *
1770 * If no more slots are available, place the task on the transport's
1771 * backlog queue.
1772 * Note that the only difference with xprt_reserve is that we now
1773 * ignore the value of the XPRT_CONGESTED flag.
1774 */
1775void xprt_retry_reserve(struct rpc_task *task)
1776{
1777 struct rpc_xprt *xprt = task->tk_xprt;
1778
1779 task->tk_status = 0;
1780 if (task->tk_rqstp != NULL)
1781 return;
1782
1783 task->tk_status = -EAGAIN;
1784 xprt_do_reserve(xprt, task);
1785}
1786
1787/**
1788 * xprt_release - release an RPC request slot
1789 * @task: task which is finished with the slot
1790 *
1791 */
1792void xprt_release(struct rpc_task *task)
1793{
1794 struct rpc_xprt *xprt;
1795 struct rpc_rqst *req = task->tk_rqstp;
1796
1797 if (req == NULL) {
1798 if (task->tk_client) {
1799 xprt = task->tk_xprt;
1800 xprt_release_write(xprt, task);
1801 }
1802 return;
1803 }
1804
1805 xprt = req->rq_xprt;
1806 xprt_request_dequeue_xprt(task);
1807 spin_lock(&xprt->transport_lock);
1808 xprt->ops->release_xprt(xprt, task);
1809 if (xprt->ops->release_request)
1810 xprt->ops->release_request(task);
1811 xprt_schedule_autodisconnect(xprt);
1812 spin_unlock(&xprt->transport_lock);
1813 if (req->rq_buffer)
1814 xprt->ops->buf_free(task);
1815 xprt_inject_disconnect(xprt);
1816 xdr_free_bvec(&req->rq_rcv_buf);
1817 xdr_free_bvec(&req->rq_snd_buf);
1818 if (req->rq_cred != NULL)
1819 put_rpccred(req->rq_cred);
1820 task->tk_rqstp = NULL;
1821 if (req->rq_release_snd_buf)
1822 req->rq_release_snd_buf(req);
1823
1824 dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1825 if (likely(!bc_prealloc(req)))
1826 xprt->ops->free_slot(xprt, req);
1827 else
1828 xprt_free_bc_request(req);
1829}
1830
1831#ifdef CONFIG_SUNRPC_BACKCHANNEL
1832void
1833xprt_init_bc_request(struct rpc_rqst *req, struct rpc_task *task)
1834{
1835 struct xdr_buf *xbufp = &req->rq_snd_buf;
1836
1837 task->tk_rqstp = req;
1838 req->rq_task = task;
1839 xprt_init_connect_cookie(req, req->rq_xprt);
1840 /*
1841 * Set up the xdr_buf length.
1842 * This also indicates that the buffer is XDR encoded already.
1843 */
1844 xbufp->len = xbufp->head[0].iov_len + xbufp->page_len +
1845 xbufp->tail[0].iov_len;
1846}
1847#endif
1848
1849static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1850{
1851 kref_init(&xprt->kref);
1852
1853 spin_lock_init(&xprt->transport_lock);
1854 spin_lock_init(&xprt->reserve_lock);
1855 spin_lock_init(&xprt->queue_lock);
1856
1857 INIT_LIST_HEAD(&xprt->free);
1858 xprt->recv_queue = RB_ROOT;
1859 INIT_LIST_HEAD(&xprt->xmit_queue);
1860#if defined(CONFIG_SUNRPC_BACKCHANNEL)
1861 spin_lock_init(&xprt->bc_pa_lock);
1862 INIT_LIST_HEAD(&xprt->bc_pa_list);
1863#endif /* CONFIG_SUNRPC_BACKCHANNEL */
1864 INIT_LIST_HEAD(&xprt->xprt_switch);
1865
1866 xprt->last_used = jiffies;
1867 xprt->cwnd = RPC_INITCWND;
1868 xprt->bind_index = 0;
1869
1870 rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1871 rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1872 rpc_init_wait_queue(&xprt->sending, "xprt_sending");
1873 rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1874
1875 xprt_init_xid(xprt);
1876
1877 xprt->xprt_net = get_net(net);
1878}
1879
1880/**
1881 * xprt_create_transport - create an RPC transport
1882 * @args: rpc transport creation arguments
1883 *
1884 */
1885struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1886{
1887 struct rpc_xprt *xprt;
1888 struct xprt_class *t;
1889
1890 spin_lock(&xprt_list_lock);
1891 list_for_each_entry(t, &xprt_list, list) {
1892 if (t->ident == args->ident) {
1893 spin_unlock(&xprt_list_lock);
1894 goto found;
1895 }
1896 }
1897 spin_unlock(&xprt_list_lock);
1898 dprintk("RPC: transport (%d) not supported\n", args->ident);
1899 return ERR_PTR(-EIO);
1900
1901found:
1902 xprt = t->setup(args);
1903 if (IS_ERR(xprt)) {
1904 dprintk("RPC: xprt_create_transport: failed, %ld\n",
1905 -PTR_ERR(xprt));
1906 goto out;
1907 }
1908 if (args->flags & XPRT_CREATE_NO_IDLE_TIMEOUT)
1909 xprt->idle_timeout = 0;
1910 INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1911 if (xprt_has_timer(xprt))
1912 timer_setup(&xprt->timer, xprt_init_autodisconnect, 0);
1913 else
1914 timer_setup(&xprt->timer, NULL, 0);
1915
1916 if (strlen(args->servername) > RPC_MAXNETNAMELEN) {
1917 xprt_destroy(xprt);
1918 return ERR_PTR(-EINVAL);
1919 }
1920 xprt->servername = kstrdup(args->servername, GFP_KERNEL);
1921 if (xprt->servername == NULL) {
1922 xprt_destroy(xprt);
1923 return ERR_PTR(-ENOMEM);
1924 }
1925
1926 rpc_xprt_debugfs_register(xprt);
1927
1928 dprintk("RPC: created transport %p with %u slots\n", xprt,
1929 xprt->max_reqs);
1930out:
1931 return xprt;
1932}
1933
1934static void xprt_destroy_cb(struct work_struct *work)
1935{
1936 struct rpc_xprt *xprt =
1937 container_of(work, struct rpc_xprt, task_cleanup);
1938
1939 rpc_xprt_debugfs_unregister(xprt);
1940 rpc_destroy_wait_queue(&xprt->binding);
1941 rpc_destroy_wait_queue(&xprt->pending);
1942 rpc_destroy_wait_queue(&xprt->sending);
1943 rpc_destroy_wait_queue(&xprt->backlog);
1944 kfree(xprt->servername);
1945 /*
1946 * Destroy any existing back channel
1947 */
1948 xprt_destroy_backchannel(xprt, UINT_MAX);
1949
1950 /*
1951 * Tear down transport state and free the rpc_xprt
1952 */
1953 xprt->ops->destroy(xprt);
1954}
1955
1956/**
1957 * xprt_destroy - destroy an RPC transport, killing off all requests.
1958 * @xprt: transport to destroy
1959 *
1960 */
1961static void xprt_destroy(struct rpc_xprt *xprt)
1962{
1963 dprintk("RPC: destroying transport %p\n", xprt);
1964
1965 /*
1966 * Exclude transport connect/disconnect handlers and autoclose
1967 */
1968 wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_UNINTERRUPTIBLE);
1969
1970 del_timer_sync(&xprt->timer);
1971
1972 /*
1973 * Destroy sockets etc from the system workqueue so they can
1974 * safely flush receive work running on rpciod.
1975 */
1976 INIT_WORK(&xprt->task_cleanup, xprt_destroy_cb);
1977 schedule_work(&xprt->task_cleanup);
1978}
1979
1980static void xprt_destroy_kref(struct kref *kref)
1981{
1982 xprt_destroy(container_of(kref, struct rpc_xprt, kref));
1983}
1984
1985/**
1986 * xprt_get - return a reference to an RPC transport.
1987 * @xprt: pointer to the transport
1988 *
1989 */
1990struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1991{
1992 if (xprt != NULL && kref_get_unless_zero(&xprt->kref))
1993 return xprt;
1994 return NULL;
1995}
1996EXPORT_SYMBOL_GPL(xprt_get);
1997
1998/**
1999 * xprt_put - release a reference to an RPC transport.
2000 * @xprt: pointer to the transport
2001 *
2002 */
2003void xprt_put(struct rpc_xprt *xprt)
2004{
2005 if (xprt != NULL)
2006 kref_put(&xprt->kref, xprt_destroy_kref);
2007}
2008EXPORT_SYMBOL_GPL(xprt_put);
1/*
2 * linux/net/sunrpc/xprt.c
3 *
4 * This is a generic RPC call interface supporting congestion avoidance,
5 * and asynchronous calls.
6 *
7 * The interface works like this:
8 *
9 * - When a process places a call, it allocates a request slot if
10 * one is available. Otherwise, it sleeps on the backlog queue
11 * (xprt_reserve).
12 * - Next, the caller puts together the RPC message, stuffs it into
13 * the request struct, and calls xprt_transmit().
14 * - xprt_transmit sends the message and installs the caller on the
15 * transport's wait list. At the same time, if a reply is expected,
16 * it installs a timer that is run after the packet's timeout has
17 * expired.
18 * - When a packet arrives, the data_ready handler walks the list of
19 * pending requests for that transport. If a matching XID is found, the
20 * caller is woken up, and the timer removed.
21 * - When no reply arrives within the timeout interval, the timer is
22 * fired by the kernel and runs xprt_timer(). It either adjusts the
23 * timeout values (minor timeout) or wakes up the caller with a status
24 * of -ETIMEDOUT.
25 * - When the caller receives a notification from RPC that a reply arrived,
26 * it should release the RPC slot, and process the reply.
27 * If the call timed out, it may choose to retry the operation by
28 * adjusting the initial timeout value, and simply calling rpc_call
29 * again.
30 *
31 * Support for async RPC is done through a set of RPC-specific scheduling
32 * primitives that `transparently' work for processes as well as async
33 * tasks that rely on callbacks.
34 *
35 * Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
36 *
37 * Transport switch API copyright (C) 2005, Chuck Lever <cel@netapp.com>
38 */
39
40#include <linux/module.h>
41
42#include <linux/types.h>
43#include <linux/interrupt.h>
44#include <linux/workqueue.h>
45#include <linux/net.h>
46#include <linux/ktime.h>
47
48#include <linux/sunrpc/clnt.h>
49#include <linux/sunrpc/metrics.h>
50#include <linux/sunrpc/bc_xprt.h>
51
52#include "sunrpc.h"
53
54/*
55 * Local variables
56 */
57
58#ifdef RPC_DEBUG
59# define RPCDBG_FACILITY RPCDBG_XPRT
60#endif
61
62/*
63 * Local functions
64 */
65static void xprt_init(struct rpc_xprt *xprt, struct net *net);
66static void xprt_request_init(struct rpc_task *, struct rpc_xprt *);
67static void xprt_connect_status(struct rpc_task *task);
68static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
69static void xprt_destroy(struct rpc_xprt *xprt);
70
71static DEFINE_SPINLOCK(xprt_list_lock);
72static LIST_HEAD(xprt_list);
73
74/*
75 * The transport code maintains an estimate on the maximum number of out-
76 * standing RPC requests, using a smoothed version of the congestion
77 * avoidance implemented in 44BSD. This is basically the Van Jacobson
78 * congestion algorithm: If a retransmit occurs, the congestion window is
79 * halved; otherwise, it is incremented by 1/cwnd when
80 *
81 * - a reply is received and
82 * - a full number of requests are outstanding and
83 * - the congestion window hasn't been updated recently.
84 */
85#define RPC_CWNDSHIFT (8U)
86#define RPC_CWNDSCALE (1U << RPC_CWNDSHIFT)
87#define RPC_INITCWND RPC_CWNDSCALE
88#define RPC_MAXCWND(xprt) ((xprt)->max_reqs << RPC_CWNDSHIFT)
89
90#define RPCXPRT_CONGESTED(xprt) ((xprt)->cong >= (xprt)->cwnd)
91
92/**
93 * xprt_register_transport - register a transport implementation
94 * @transport: transport to register
95 *
96 * If a transport implementation is loaded as a kernel module, it can
97 * call this interface to make itself known to the RPC client.
98 *
99 * Returns:
100 * 0: transport successfully registered
101 * -EEXIST: transport already registered
102 * -EINVAL: transport module being unloaded
103 */
104int xprt_register_transport(struct xprt_class *transport)
105{
106 struct xprt_class *t;
107 int result;
108
109 result = -EEXIST;
110 spin_lock(&xprt_list_lock);
111 list_for_each_entry(t, &xprt_list, list) {
112 /* don't register the same transport class twice */
113 if (t->ident == transport->ident)
114 goto out;
115 }
116
117 list_add_tail(&transport->list, &xprt_list);
118 printk(KERN_INFO "RPC: Registered %s transport module.\n",
119 transport->name);
120 result = 0;
121
122out:
123 spin_unlock(&xprt_list_lock);
124 return result;
125}
126EXPORT_SYMBOL_GPL(xprt_register_transport);
127
128/**
129 * xprt_unregister_transport - unregister a transport implementation
130 * @transport: transport to unregister
131 *
132 * Returns:
133 * 0: transport successfully unregistered
134 * -ENOENT: transport never registered
135 */
136int xprt_unregister_transport(struct xprt_class *transport)
137{
138 struct xprt_class *t;
139 int result;
140
141 result = 0;
142 spin_lock(&xprt_list_lock);
143 list_for_each_entry(t, &xprt_list, list) {
144 if (t == transport) {
145 printk(KERN_INFO
146 "RPC: Unregistered %s transport module.\n",
147 transport->name);
148 list_del_init(&transport->list);
149 goto out;
150 }
151 }
152 result = -ENOENT;
153
154out:
155 spin_unlock(&xprt_list_lock);
156 return result;
157}
158EXPORT_SYMBOL_GPL(xprt_unregister_transport);
159
160/**
161 * xprt_load_transport - load a transport implementation
162 * @transport_name: transport to load
163 *
164 * Returns:
165 * 0: transport successfully loaded
166 * -ENOENT: transport module not available
167 */
168int xprt_load_transport(const char *transport_name)
169{
170 struct xprt_class *t;
171 int result;
172
173 result = 0;
174 spin_lock(&xprt_list_lock);
175 list_for_each_entry(t, &xprt_list, list) {
176 if (strcmp(t->name, transport_name) == 0) {
177 spin_unlock(&xprt_list_lock);
178 goto out;
179 }
180 }
181 spin_unlock(&xprt_list_lock);
182 result = request_module("xprt%s", transport_name);
183out:
184 return result;
185}
186EXPORT_SYMBOL_GPL(xprt_load_transport);
187
188/**
189 * xprt_reserve_xprt - serialize write access to transports
190 * @task: task that is requesting access to the transport
191 * @xprt: pointer to the target transport
192 *
193 * This prevents mixing the payload of separate requests, and prevents
194 * transport connects from colliding with writes. No congestion control
195 * is provided.
196 */
197int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
198{
199 struct rpc_rqst *req = task->tk_rqstp;
200 int priority;
201
202 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
203 if (task == xprt->snd_task)
204 return 1;
205 goto out_sleep;
206 }
207 xprt->snd_task = task;
208 if (req != NULL) {
209 req->rq_bytes_sent = 0;
210 req->rq_ntrans++;
211 }
212
213 return 1;
214
215out_sleep:
216 dprintk("RPC: %5u failed to lock transport %p\n",
217 task->tk_pid, xprt);
218 task->tk_timeout = 0;
219 task->tk_status = -EAGAIN;
220 if (req == NULL)
221 priority = RPC_PRIORITY_LOW;
222 else if (!req->rq_ntrans)
223 priority = RPC_PRIORITY_NORMAL;
224 else
225 priority = RPC_PRIORITY_HIGH;
226 rpc_sleep_on_priority(&xprt->sending, task, NULL, priority);
227 return 0;
228}
229EXPORT_SYMBOL_GPL(xprt_reserve_xprt);
230
231static void xprt_clear_locked(struct rpc_xprt *xprt)
232{
233 xprt->snd_task = NULL;
234 if (!test_bit(XPRT_CLOSE_WAIT, &xprt->state) || xprt->shutdown) {
235 smp_mb__before_clear_bit();
236 clear_bit(XPRT_LOCKED, &xprt->state);
237 smp_mb__after_clear_bit();
238 } else
239 queue_work(rpciod_workqueue, &xprt->task_cleanup);
240}
241
242/*
243 * xprt_reserve_xprt_cong - serialize write access to transports
244 * @task: task that is requesting access to the transport
245 *
246 * Same as xprt_reserve_xprt, but Van Jacobson congestion control is
247 * integrated into the decision of whether a request is allowed to be
248 * woken up and given access to the transport.
249 */
250int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
251{
252 struct rpc_rqst *req = task->tk_rqstp;
253 int priority;
254
255 if (test_and_set_bit(XPRT_LOCKED, &xprt->state)) {
256 if (task == xprt->snd_task)
257 return 1;
258 goto out_sleep;
259 }
260 if (req == NULL) {
261 xprt->snd_task = task;
262 return 1;
263 }
264 if (__xprt_get_cong(xprt, task)) {
265 xprt->snd_task = task;
266 req->rq_bytes_sent = 0;
267 req->rq_ntrans++;
268 return 1;
269 }
270 xprt_clear_locked(xprt);
271out_sleep:
272 dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
273 task->tk_timeout = 0;
274 task->tk_status = -EAGAIN;
275 if (req == NULL)
276 priority = RPC_PRIORITY_LOW;
277 else if (!req->rq_ntrans)
278 priority = RPC_PRIORITY_NORMAL;
279 else
280 priority = RPC_PRIORITY_HIGH;
281 rpc_sleep_on_priority(&xprt->sending, task, NULL, priority);
282 return 0;
283}
284EXPORT_SYMBOL_GPL(xprt_reserve_xprt_cong);
285
286static inline int xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
287{
288 int retval;
289
290 spin_lock_bh(&xprt->transport_lock);
291 retval = xprt->ops->reserve_xprt(xprt, task);
292 spin_unlock_bh(&xprt->transport_lock);
293 return retval;
294}
295
296static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
297{
298 struct rpc_xprt *xprt = data;
299 struct rpc_rqst *req;
300
301 req = task->tk_rqstp;
302 xprt->snd_task = task;
303 if (req) {
304 req->rq_bytes_sent = 0;
305 req->rq_ntrans++;
306 }
307 return true;
308}
309
310static void __xprt_lock_write_next(struct rpc_xprt *xprt)
311{
312 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
313 return;
314
315 if (rpc_wake_up_first(&xprt->sending, __xprt_lock_write_func, xprt))
316 return;
317 xprt_clear_locked(xprt);
318}
319
320static bool __xprt_lock_write_cong_func(struct rpc_task *task, void *data)
321{
322 struct rpc_xprt *xprt = data;
323 struct rpc_rqst *req;
324
325 req = task->tk_rqstp;
326 if (req == NULL) {
327 xprt->snd_task = task;
328 return true;
329 }
330 if (__xprt_get_cong(xprt, task)) {
331 xprt->snd_task = task;
332 req->rq_bytes_sent = 0;
333 req->rq_ntrans++;
334 return true;
335 }
336 return false;
337}
338
339static void __xprt_lock_write_next_cong(struct rpc_xprt *xprt)
340{
341 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
342 return;
343 if (RPCXPRT_CONGESTED(xprt))
344 goto out_unlock;
345 if (rpc_wake_up_first(&xprt->sending, __xprt_lock_write_cong_func, xprt))
346 return;
347out_unlock:
348 xprt_clear_locked(xprt);
349}
350
351/**
352 * xprt_release_xprt - allow other requests to use a transport
353 * @xprt: transport with other tasks potentially waiting
354 * @task: task that is releasing access to the transport
355 *
356 * Note that "task" can be NULL. No congestion control is provided.
357 */
358void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
359{
360 if (xprt->snd_task == task) {
361 xprt_clear_locked(xprt);
362 __xprt_lock_write_next(xprt);
363 }
364}
365EXPORT_SYMBOL_GPL(xprt_release_xprt);
366
367/**
368 * xprt_release_xprt_cong - allow other requests to use a transport
369 * @xprt: transport with other tasks potentially waiting
370 * @task: task that is releasing access to the transport
371 *
372 * Note that "task" can be NULL. Another task is awoken to use the
373 * transport if the transport's congestion window allows it.
374 */
375void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
376{
377 if (xprt->snd_task == task) {
378 xprt_clear_locked(xprt);
379 __xprt_lock_write_next_cong(xprt);
380 }
381}
382EXPORT_SYMBOL_GPL(xprt_release_xprt_cong);
383
384static inline void xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
385{
386 spin_lock_bh(&xprt->transport_lock);
387 xprt->ops->release_xprt(xprt, task);
388 spin_unlock_bh(&xprt->transport_lock);
389}
390
391/*
392 * Van Jacobson congestion avoidance. Check if the congestion window
393 * overflowed. Put the task to sleep if this is the case.
394 */
395static int
396__xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
397{
398 struct rpc_rqst *req = task->tk_rqstp;
399
400 if (req->rq_cong)
401 return 1;
402 dprintk("RPC: %5u xprt_cwnd_limited cong = %lu cwnd = %lu\n",
403 task->tk_pid, xprt->cong, xprt->cwnd);
404 if (RPCXPRT_CONGESTED(xprt))
405 return 0;
406 req->rq_cong = 1;
407 xprt->cong += RPC_CWNDSCALE;
408 return 1;
409}
410
411/*
412 * Adjust the congestion window, and wake up the next task
413 * that has been sleeping due to congestion
414 */
415static void
416__xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
417{
418 if (!req->rq_cong)
419 return;
420 req->rq_cong = 0;
421 xprt->cong -= RPC_CWNDSCALE;
422 __xprt_lock_write_next_cong(xprt);
423}
424
425/**
426 * xprt_release_rqst_cong - housekeeping when request is complete
427 * @task: RPC request that recently completed
428 *
429 * Useful for transports that require congestion control.
430 */
431void xprt_release_rqst_cong(struct rpc_task *task)
432{
433 __xprt_put_cong(task->tk_xprt, task->tk_rqstp);
434}
435EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
436
437/**
438 * xprt_adjust_cwnd - adjust transport congestion window
439 * @task: recently completed RPC request used to adjust window
440 * @result: result code of completed RPC request
441 *
442 * We use a time-smoothed congestion estimator to avoid heavy oscillation.
443 */
444void xprt_adjust_cwnd(struct rpc_task *task, int result)
445{
446 struct rpc_rqst *req = task->tk_rqstp;
447 struct rpc_xprt *xprt = task->tk_xprt;
448 unsigned long cwnd = xprt->cwnd;
449
450 if (result >= 0 && cwnd <= xprt->cong) {
451 /* The (cwnd >> 1) term makes sure
452 * the result gets rounded properly. */
453 cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
454 if (cwnd > RPC_MAXCWND(xprt))
455 cwnd = RPC_MAXCWND(xprt);
456 __xprt_lock_write_next_cong(xprt);
457 } else if (result == -ETIMEDOUT) {
458 cwnd >>= 1;
459 if (cwnd < RPC_CWNDSCALE)
460 cwnd = RPC_CWNDSCALE;
461 }
462 dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
463 xprt->cong, xprt->cwnd, cwnd);
464 xprt->cwnd = cwnd;
465 __xprt_put_cong(xprt, req);
466}
467EXPORT_SYMBOL_GPL(xprt_adjust_cwnd);
468
469/**
470 * xprt_wake_pending_tasks - wake all tasks on a transport's pending queue
471 * @xprt: transport with waiting tasks
472 * @status: result code to plant in each task before waking it
473 *
474 */
475void xprt_wake_pending_tasks(struct rpc_xprt *xprt, int status)
476{
477 if (status < 0)
478 rpc_wake_up_status(&xprt->pending, status);
479 else
480 rpc_wake_up(&xprt->pending);
481}
482EXPORT_SYMBOL_GPL(xprt_wake_pending_tasks);
483
484/**
485 * xprt_wait_for_buffer_space - wait for transport output buffer to clear
486 * @task: task to be put to sleep
487 * @action: function pointer to be executed after wait
488 */
489void xprt_wait_for_buffer_space(struct rpc_task *task, rpc_action action)
490{
491 struct rpc_rqst *req = task->tk_rqstp;
492 struct rpc_xprt *xprt = req->rq_xprt;
493
494 task->tk_timeout = req->rq_timeout;
495 rpc_sleep_on(&xprt->pending, task, action);
496}
497EXPORT_SYMBOL_GPL(xprt_wait_for_buffer_space);
498
499/**
500 * xprt_write_space - wake the task waiting for transport output buffer space
501 * @xprt: transport with waiting tasks
502 *
503 * Can be called in a soft IRQ context, so xprt_write_space never sleeps.
504 */
505void xprt_write_space(struct rpc_xprt *xprt)
506{
507 if (unlikely(xprt->shutdown))
508 return;
509
510 spin_lock_bh(&xprt->transport_lock);
511 if (xprt->snd_task) {
512 dprintk("RPC: write space: waking waiting task on "
513 "xprt %p\n", xprt);
514 rpc_wake_up_queued_task(&xprt->pending, xprt->snd_task);
515 }
516 spin_unlock_bh(&xprt->transport_lock);
517}
518EXPORT_SYMBOL_GPL(xprt_write_space);
519
520/**
521 * xprt_set_retrans_timeout_def - set a request's retransmit timeout
522 * @task: task whose timeout is to be set
523 *
524 * Set a request's retransmit timeout based on the transport's
525 * default timeout parameters. Used by transports that don't adjust
526 * the retransmit timeout based on round-trip time estimation.
527 */
528void xprt_set_retrans_timeout_def(struct rpc_task *task)
529{
530 task->tk_timeout = task->tk_rqstp->rq_timeout;
531}
532EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_def);
533
534/*
535 * xprt_set_retrans_timeout_rtt - set a request's retransmit timeout
536 * @task: task whose timeout is to be set
537 *
538 * Set a request's retransmit timeout using the RTT estimator.
539 */
540void xprt_set_retrans_timeout_rtt(struct rpc_task *task)
541{
542 int timer = task->tk_msg.rpc_proc->p_timer;
543 struct rpc_clnt *clnt = task->tk_client;
544 struct rpc_rtt *rtt = clnt->cl_rtt;
545 struct rpc_rqst *req = task->tk_rqstp;
546 unsigned long max_timeout = clnt->cl_timeout->to_maxval;
547
548 task->tk_timeout = rpc_calc_rto(rtt, timer);
549 task->tk_timeout <<= rpc_ntimeo(rtt, timer) + req->rq_retries;
550 if (task->tk_timeout > max_timeout || task->tk_timeout == 0)
551 task->tk_timeout = max_timeout;
552}
553EXPORT_SYMBOL_GPL(xprt_set_retrans_timeout_rtt);
554
555static void xprt_reset_majortimeo(struct rpc_rqst *req)
556{
557 const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
558
559 req->rq_majortimeo = req->rq_timeout;
560 if (to->to_exponential)
561 req->rq_majortimeo <<= to->to_retries;
562 else
563 req->rq_majortimeo += to->to_increment * to->to_retries;
564 if (req->rq_majortimeo > to->to_maxval || req->rq_majortimeo == 0)
565 req->rq_majortimeo = to->to_maxval;
566 req->rq_majortimeo += jiffies;
567}
568
569/**
570 * xprt_adjust_timeout - adjust timeout values for next retransmit
571 * @req: RPC request containing parameters to use for the adjustment
572 *
573 */
574int xprt_adjust_timeout(struct rpc_rqst *req)
575{
576 struct rpc_xprt *xprt = req->rq_xprt;
577 const struct rpc_timeout *to = req->rq_task->tk_client->cl_timeout;
578 int status = 0;
579
580 if (time_before(jiffies, req->rq_majortimeo)) {
581 if (to->to_exponential)
582 req->rq_timeout <<= 1;
583 else
584 req->rq_timeout += to->to_increment;
585 if (to->to_maxval && req->rq_timeout >= to->to_maxval)
586 req->rq_timeout = to->to_maxval;
587 req->rq_retries++;
588 } else {
589 req->rq_timeout = to->to_initval;
590 req->rq_retries = 0;
591 xprt_reset_majortimeo(req);
592 /* Reset the RTT counters == "slow start" */
593 spin_lock_bh(&xprt->transport_lock);
594 rpc_init_rtt(req->rq_task->tk_client->cl_rtt, to->to_initval);
595 spin_unlock_bh(&xprt->transport_lock);
596 status = -ETIMEDOUT;
597 }
598
599 if (req->rq_timeout == 0) {
600 printk(KERN_WARNING "xprt_adjust_timeout: rq_timeout = 0!\n");
601 req->rq_timeout = 5 * HZ;
602 }
603 return status;
604}
605
606static void xprt_autoclose(struct work_struct *work)
607{
608 struct rpc_xprt *xprt =
609 container_of(work, struct rpc_xprt, task_cleanup);
610
611 xprt->ops->close(xprt);
612 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
613 xprt_release_write(xprt, NULL);
614}
615
616/**
617 * xprt_disconnect_done - mark a transport as disconnected
618 * @xprt: transport to flag for disconnect
619 *
620 */
621void xprt_disconnect_done(struct rpc_xprt *xprt)
622{
623 dprintk("RPC: disconnected transport %p\n", xprt);
624 spin_lock_bh(&xprt->transport_lock);
625 xprt_clear_connected(xprt);
626 xprt_wake_pending_tasks(xprt, -EAGAIN);
627 spin_unlock_bh(&xprt->transport_lock);
628}
629EXPORT_SYMBOL_GPL(xprt_disconnect_done);
630
631/**
632 * xprt_force_disconnect - force a transport to disconnect
633 * @xprt: transport to disconnect
634 *
635 */
636void xprt_force_disconnect(struct rpc_xprt *xprt)
637{
638 /* Don't race with the test_bit() in xprt_clear_locked() */
639 spin_lock_bh(&xprt->transport_lock);
640 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
641 /* Try to schedule an autoclose RPC call */
642 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
643 queue_work(rpciod_workqueue, &xprt->task_cleanup);
644 xprt_wake_pending_tasks(xprt, -EAGAIN);
645 spin_unlock_bh(&xprt->transport_lock);
646}
647
648/**
649 * xprt_conditional_disconnect - force a transport to disconnect
650 * @xprt: transport to disconnect
651 * @cookie: 'connection cookie'
652 *
653 * This attempts to break the connection if and only if 'cookie' matches
654 * the current transport 'connection cookie'. It ensures that we don't
655 * try to break the connection more than once when we need to retransmit
656 * a batch of RPC requests.
657 *
658 */
659void xprt_conditional_disconnect(struct rpc_xprt *xprt, unsigned int cookie)
660{
661 /* Don't race with the test_bit() in xprt_clear_locked() */
662 spin_lock_bh(&xprt->transport_lock);
663 if (cookie != xprt->connect_cookie)
664 goto out;
665 if (test_bit(XPRT_CLOSING, &xprt->state) || !xprt_connected(xprt))
666 goto out;
667 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
668 /* Try to schedule an autoclose RPC call */
669 if (test_and_set_bit(XPRT_LOCKED, &xprt->state) == 0)
670 queue_work(rpciod_workqueue, &xprt->task_cleanup);
671 xprt_wake_pending_tasks(xprt, -EAGAIN);
672out:
673 spin_unlock_bh(&xprt->transport_lock);
674}
675
676static void
677xprt_init_autodisconnect(unsigned long data)
678{
679 struct rpc_xprt *xprt = (struct rpc_xprt *)data;
680
681 spin_lock(&xprt->transport_lock);
682 if (!list_empty(&xprt->recv) || xprt->shutdown)
683 goto out_abort;
684 if (test_and_set_bit(XPRT_LOCKED, &xprt->state))
685 goto out_abort;
686 spin_unlock(&xprt->transport_lock);
687 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
688 queue_work(rpciod_workqueue, &xprt->task_cleanup);
689 return;
690out_abort:
691 spin_unlock(&xprt->transport_lock);
692}
693
694/**
695 * xprt_connect - schedule a transport connect operation
696 * @task: RPC task that is requesting the connect
697 *
698 */
699void xprt_connect(struct rpc_task *task)
700{
701 struct rpc_xprt *xprt = task->tk_xprt;
702
703 dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
704 xprt, (xprt_connected(xprt) ? "is" : "is not"));
705
706 if (!xprt_bound(xprt)) {
707 task->tk_status = -EAGAIN;
708 return;
709 }
710 if (!xprt_lock_write(xprt, task))
711 return;
712
713 if (test_and_clear_bit(XPRT_CLOSE_WAIT, &xprt->state))
714 xprt->ops->close(xprt);
715
716 if (xprt_connected(xprt))
717 xprt_release_write(xprt, task);
718 else {
719 task->tk_rqstp->rq_bytes_sent = 0;
720 task->tk_timeout = task->tk_rqstp->rq_timeout;
721 rpc_sleep_on(&xprt->pending, task, xprt_connect_status);
722
723 if (test_bit(XPRT_CLOSING, &xprt->state))
724 return;
725 if (xprt_test_and_set_connecting(xprt))
726 return;
727 xprt->stat.connect_start = jiffies;
728 xprt->ops->connect(task);
729 }
730}
731
732static void xprt_connect_status(struct rpc_task *task)
733{
734 struct rpc_xprt *xprt = task->tk_xprt;
735
736 if (task->tk_status == 0) {
737 xprt->stat.connect_count++;
738 xprt->stat.connect_time += (long)jiffies - xprt->stat.connect_start;
739 dprintk("RPC: %5u xprt_connect_status: connection established\n",
740 task->tk_pid);
741 return;
742 }
743
744 switch (task->tk_status) {
745 case -EAGAIN:
746 dprintk("RPC: %5u xprt_connect_status: retrying\n", task->tk_pid);
747 break;
748 case -ETIMEDOUT:
749 dprintk("RPC: %5u xprt_connect_status: connect attempt timed "
750 "out\n", task->tk_pid);
751 break;
752 default:
753 dprintk("RPC: %5u xprt_connect_status: error %d connecting to "
754 "server %s\n", task->tk_pid, -task->tk_status,
755 xprt->servername);
756 xprt_release_write(xprt, task);
757 task->tk_status = -EIO;
758 }
759}
760
761/**
762 * xprt_lookup_rqst - find an RPC request corresponding to an XID
763 * @xprt: transport on which the original request was transmitted
764 * @xid: RPC XID of incoming reply
765 *
766 */
767struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, __be32 xid)
768{
769 struct rpc_rqst *entry;
770
771 list_for_each_entry(entry, &xprt->recv, rq_list)
772 if (entry->rq_xid == xid)
773 return entry;
774
775 dprintk("RPC: xprt_lookup_rqst did not find xid %08x\n",
776 ntohl(xid));
777 xprt->stat.bad_xids++;
778 return NULL;
779}
780EXPORT_SYMBOL_GPL(xprt_lookup_rqst);
781
782static void xprt_update_rtt(struct rpc_task *task)
783{
784 struct rpc_rqst *req = task->tk_rqstp;
785 struct rpc_rtt *rtt = task->tk_client->cl_rtt;
786 unsigned int timer = task->tk_msg.rpc_proc->p_timer;
787 long m = usecs_to_jiffies(ktime_to_us(req->rq_rtt));
788
789 if (timer) {
790 if (req->rq_ntrans == 1)
791 rpc_update_rtt(rtt, timer, m);
792 rpc_set_timeo(rtt, timer, req->rq_ntrans - 1);
793 }
794}
795
796/**
797 * xprt_complete_rqst - called when reply processing is complete
798 * @task: RPC request that recently completed
799 * @copied: actual number of bytes received from the transport
800 *
801 * Caller holds transport lock.
802 */
803void xprt_complete_rqst(struct rpc_task *task, int copied)
804{
805 struct rpc_rqst *req = task->tk_rqstp;
806 struct rpc_xprt *xprt = req->rq_xprt;
807
808 dprintk("RPC: %5u xid %08x complete (%d bytes received)\n",
809 task->tk_pid, ntohl(req->rq_xid), copied);
810
811 xprt->stat.recvs++;
812 req->rq_rtt = ktime_sub(ktime_get(), req->rq_xtime);
813 if (xprt->ops->timer != NULL)
814 xprt_update_rtt(task);
815
816 list_del_init(&req->rq_list);
817 req->rq_private_buf.len = copied;
818 /* Ensure all writes are done before we update */
819 /* req->rq_reply_bytes_recvd */
820 smp_wmb();
821 req->rq_reply_bytes_recvd = copied;
822 rpc_wake_up_queued_task(&xprt->pending, task);
823}
824EXPORT_SYMBOL_GPL(xprt_complete_rqst);
825
826static void xprt_timer(struct rpc_task *task)
827{
828 struct rpc_rqst *req = task->tk_rqstp;
829 struct rpc_xprt *xprt = req->rq_xprt;
830
831 if (task->tk_status != -ETIMEDOUT)
832 return;
833 dprintk("RPC: %5u xprt_timer\n", task->tk_pid);
834
835 spin_lock_bh(&xprt->transport_lock);
836 if (!req->rq_reply_bytes_recvd) {
837 if (xprt->ops->timer)
838 xprt->ops->timer(task);
839 } else
840 task->tk_status = 0;
841 spin_unlock_bh(&xprt->transport_lock);
842}
843
844static inline int xprt_has_timer(struct rpc_xprt *xprt)
845{
846 return xprt->idle_timeout != 0;
847}
848
849/**
850 * xprt_prepare_transmit - reserve the transport before sending a request
851 * @task: RPC task about to send a request
852 *
853 */
854int xprt_prepare_transmit(struct rpc_task *task)
855{
856 struct rpc_rqst *req = task->tk_rqstp;
857 struct rpc_xprt *xprt = req->rq_xprt;
858 int err = 0;
859
860 dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
861
862 spin_lock_bh(&xprt->transport_lock);
863 if (req->rq_reply_bytes_recvd && !req->rq_bytes_sent) {
864 err = req->rq_reply_bytes_recvd;
865 goto out_unlock;
866 }
867 if (!xprt->ops->reserve_xprt(xprt, task))
868 err = -EAGAIN;
869out_unlock:
870 spin_unlock_bh(&xprt->transport_lock);
871 return err;
872}
873
874void xprt_end_transmit(struct rpc_task *task)
875{
876 xprt_release_write(task->tk_rqstp->rq_xprt, task);
877}
878
879/**
880 * xprt_transmit - send an RPC request on a transport
881 * @task: controlling RPC task
882 *
883 * We have to copy the iovec because sendmsg fiddles with its contents.
884 */
885void xprt_transmit(struct rpc_task *task)
886{
887 struct rpc_rqst *req = task->tk_rqstp;
888 struct rpc_xprt *xprt = req->rq_xprt;
889 int status, numreqs;
890
891 dprintk("RPC: %5u xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
892
893 if (!req->rq_reply_bytes_recvd) {
894 if (list_empty(&req->rq_list) && rpc_reply_expected(task)) {
895 /*
896 * Add to the list only if we're expecting a reply
897 */
898 spin_lock_bh(&xprt->transport_lock);
899 /* Update the softirq receive buffer */
900 memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
901 sizeof(req->rq_private_buf));
902 /* Add request to the receive list */
903 list_add_tail(&req->rq_list, &xprt->recv);
904 spin_unlock_bh(&xprt->transport_lock);
905 xprt_reset_majortimeo(req);
906 /* Turn off autodisconnect */
907 del_singleshot_timer_sync(&xprt->timer);
908 }
909 } else if (!req->rq_bytes_sent)
910 return;
911
912 req->rq_connect_cookie = xprt->connect_cookie;
913 req->rq_xtime = ktime_get();
914 status = xprt->ops->send_request(task);
915 if (status != 0) {
916 task->tk_status = status;
917 return;
918 }
919
920 dprintk("RPC: %5u xmit complete\n", task->tk_pid);
921 task->tk_flags |= RPC_TASK_SENT;
922 spin_lock_bh(&xprt->transport_lock);
923
924 xprt->ops->set_retrans_timeout(task);
925
926 numreqs = atomic_read(&xprt->num_reqs);
927 if (numreqs > xprt->stat.max_slots)
928 xprt->stat.max_slots = numreqs;
929 xprt->stat.sends++;
930 xprt->stat.req_u += xprt->stat.sends - xprt->stat.recvs;
931 xprt->stat.bklog_u += xprt->backlog.qlen;
932 xprt->stat.sending_u += xprt->sending.qlen;
933 xprt->stat.pending_u += xprt->pending.qlen;
934
935 /* Don't race with disconnect */
936 if (!xprt_connected(xprt))
937 task->tk_status = -ENOTCONN;
938 else if (!req->rq_reply_bytes_recvd && rpc_reply_expected(task)) {
939 /*
940 * Sleep on the pending queue since
941 * we're expecting a reply.
942 */
943 rpc_sleep_on(&xprt->pending, task, xprt_timer);
944 }
945 spin_unlock_bh(&xprt->transport_lock);
946}
947
948static struct rpc_rqst *xprt_dynamic_alloc_slot(struct rpc_xprt *xprt, gfp_t gfp_flags)
949{
950 struct rpc_rqst *req = ERR_PTR(-EAGAIN);
951
952 if (!atomic_add_unless(&xprt->num_reqs, 1, xprt->max_reqs))
953 goto out;
954 req = kzalloc(sizeof(struct rpc_rqst), gfp_flags);
955 if (req != NULL)
956 goto out;
957 atomic_dec(&xprt->num_reqs);
958 req = ERR_PTR(-ENOMEM);
959out:
960 return req;
961}
962
963static bool xprt_dynamic_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
964{
965 if (atomic_add_unless(&xprt->num_reqs, -1, xprt->min_reqs)) {
966 kfree(req);
967 return true;
968 }
969 return false;
970}
971
972static void xprt_alloc_slot(struct rpc_task *task)
973{
974 struct rpc_xprt *xprt = task->tk_xprt;
975 struct rpc_rqst *req;
976
977 if (!list_empty(&xprt->free)) {
978 req = list_entry(xprt->free.next, struct rpc_rqst, rq_list);
979 list_del(&req->rq_list);
980 goto out_init_req;
981 }
982 req = xprt_dynamic_alloc_slot(xprt, GFP_NOWAIT|__GFP_NOWARN);
983 if (!IS_ERR(req))
984 goto out_init_req;
985 switch (PTR_ERR(req)) {
986 case -ENOMEM:
987 dprintk("RPC: dynamic allocation of request slot "
988 "failed! Retrying\n");
989 task->tk_status = -ENOMEM;
990 break;
991 case -EAGAIN:
992 rpc_sleep_on(&xprt->backlog, task, NULL);
993 dprintk("RPC: waiting for request slot\n");
994 default:
995 task->tk_status = -EAGAIN;
996 }
997 return;
998out_init_req:
999 task->tk_status = 0;
1000 task->tk_rqstp = req;
1001 xprt_request_init(task, xprt);
1002}
1003
1004static void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
1005{
1006 spin_lock(&xprt->reserve_lock);
1007 if (!xprt_dynamic_free_slot(xprt, req)) {
1008 memset(req, 0, sizeof(*req)); /* mark unused */
1009 list_add(&req->rq_list, &xprt->free);
1010 }
1011 rpc_wake_up_next(&xprt->backlog);
1012 spin_unlock(&xprt->reserve_lock);
1013}
1014
1015static void xprt_free_all_slots(struct rpc_xprt *xprt)
1016{
1017 struct rpc_rqst *req;
1018 while (!list_empty(&xprt->free)) {
1019 req = list_first_entry(&xprt->free, struct rpc_rqst, rq_list);
1020 list_del(&req->rq_list);
1021 kfree(req);
1022 }
1023}
1024
1025struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
1026 unsigned int num_prealloc,
1027 unsigned int max_alloc)
1028{
1029 struct rpc_xprt *xprt;
1030 struct rpc_rqst *req;
1031 int i;
1032
1033 xprt = kzalloc(size, GFP_KERNEL);
1034 if (xprt == NULL)
1035 goto out;
1036
1037 xprt_init(xprt, net);
1038
1039 for (i = 0; i < num_prealloc; i++) {
1040 req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
1041 if (!req)
1042 break;
1043 list_add(&req->rq_list, &xprt->free);
1044 }
1045 if (i < num_prealloc)
1046 goto out_free;
1047 if (max_alloc > num_prealloc)
1048 xprt->max_reqs = max_alloc;
1049 else
1050 xprt->max_reqs = num_prealloc;
1051 xprt->min_reqs = num_prealloc;
1052 atomic_set(&xprt->num_reqs, num_prealloc);
1053
1054 return xprt;
1055
1056out_free:
1057 xprt_free(xprt);
1058out:
1059 return NULL;
1060}
1061EXPORT_SYMBOL_GPL(xprt_alloc);
1062
1063void xprt_free(struct rpc_xprt *xprt)
1064{
1065 put_net(xprt->xprt_net);
1066 xprt_free_all_slots(xprt);
1067 kfree(xprt);
1068}
1069EXPORT_SYMBOL_GPL(xprt_free);
1070
1071/**
1072 * xprt_reserve - allocate an RPC request slot
1073 * @task: RPC task requesting a slot allocation
1074 *
1075 * If no more slots are available, place the task on the transport's
1076 * backlog queue.
1077 */
1078void xprt_reserve(struct rpc_task *task)
1079{
1080 struct rpc_xprt *xprt = task->tk_xprt;
1081
1082 task->tk_status = 0;
1083 if (task->tk_rqstp != NULL)
1084 return;
1085
1086 /* Note: grabbing the xprt_lock_write() here is not strictly needed,
1087 * but ensures that we throttle new slot allocation if the transport
1088 * is congested (e.g. if reconnecting or if we're out of socket
1089 * write buffer space).
1090 */
1091 task->tk_timeout = 0;
1092 task->tk_status = -EAGAIN;
1093 if (!xprt_lock_write(xprt, task))
1094 return;
1095
1096 spin_lock(&xprt->reserve_lock);
1097 xprt_alloc_slot(task);
1098 spin_unlock(&xprt->reserve_lock);
1099 xprt_release_write(xprt, task);
1100}
1101
1102static inline __be32 xprt_alloc_xid(struct rpc_xprt *xprt)
1103{
1104 return (__force __be32)xprt->xid++;
1105}
1106
1107static inline void xprt_init_xid(struct rpc_xprt *xprt)
1108{
1109 xprt->xid = net_random();
1110}
1111
1112static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
1113{
1114 struct rpc_rqst *req = task->tk_rqstp;
1115
1116 INIT_LIST_HEAD(&req->rq_list);
1117 req->rq_timeout = task->tk_client->cl_timeout->to_initval;
1118 req->rq_task = task;
1119 req->rq_xprt = xprt;
1120 req->rq_buffer = NULL;
1121 req->rq_xid = xprt_alloc_xid(xprt);
1122 req->rq_release_snd_buf = NULL;
1123 xprt_reset_majortimeo(req);
1124 dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
1125 req, ntohl(req->rq_xid));
1126}
1127
1128/**
1129 * xprt_release - release an RPC request slot
1130 * @task: task which is finished with the slot
1131 *
1132 */
1133void xprt_release(struct rpc_task *task)
1134{
1135 struct rpc_xprt *xprt;
1136 struct rpc_rqst *req;
1137
1138 if (!(req = task->tk_rqstp))
1139 return;
1140
1141 xprt = req->rq_xprt;
1142 if (task->tk_ops->rpc_count_stats != NULL)
1143 task->tk_ops->rpc_count_stats(task, task->tk_calldata);
1144 else if (task->tk_client)
1145 rpc_count_iostats(task, task->tk_client->cl_metrics);
1146 spin_lock_bh(&xprt->transport_lock);
1147 xprt->ops->release_xprt(xprt, task);
1148 if (xprt->ops->release_request)
1149 xprt->ops->release_request(task);
1150 if (!list_empty(&req->rq_list))
1151 list_del(&req->rq_list);
1152 xprt->last_used = jiffies;
1153 if (list_empty(&xprt->recv) && xprt_has_timer(xprt))
1154 mod_timer(&xprt->timer,
1155 xprt->last_used + xprt->idle_timeout);
1156 spin_unlock_bh(&xprt->transport_lock);
1157 if (req->rq_buffer)
1158 xprt->ops->buf_free(req->rq_buffer);
1159 if (req->rq_cred != NULL)
1160 put_rpccred(req->rq_cred);
1161 task->tk_rqstp = NULL;
1162 if (req->rq_release_snd_buf)
1163 req->rq_release_snd_buf(req);
1164
1165 dprintk("RPC: %5u release request %p\n", task->tk_pid, req);
1166 if (likely(!bc_prealloc(req)))
1167 xprt_free_slot(xprt, req);
1168 else
1169 xprt_free_bc_request(req);
1170}
1171
1172static void xprt_init(struct rpc_xprt *xprt, struct net *net)
1173{
1174 atomic_set(&xprt->count, 1);
1175
1176 spin_lock_init(&xprt->transport_lock);
1177 spin_lock_init(&xprt->reserve_lock);
1178
1179 INIT_LIST_HEAD(&xprt->free);
1180 INIT_LIST_HEAD(&xprt->recv);
1181#if defined(CONFIG_SUNRPC_BACKCHANNEL)
1182 spin_lock_init(&xprt->bc_pa_lock);
1183 INIT_LIST_HEAD(&xprt->bc_pa_list);
1184#endif /* CONFIG_SUNRPC_BACKCHANNEL */
1185
1186 xprt->last_used = jiffies;
1187 xprt->cwnd = RPC_INITCWND;
1188 xprt->bind_index = 0;
1189
1190 rpc_init_wait_queue(&xprt->binding, "xprt_binding");
1191 rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1192 rpc_init_priority_wait_queue(&xprt->sending, "xprt_sending");
1193 rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1194
1195 xprt_init_xid(xprt);
1196
1197 xprt->xprt_net = get_net(net);
1198}
1199
1200/**
1201 * xprt_create_transport - create an RPC transport
1202 * @args: rpc transport creation arguments
1203 *
1204 */
1205struct rpc_xprt *xprt_create_transport(struct xprt_create *args)
1206{
1207 struct rpc_xprt *xprt;
1208 struct xprt_class *t;
1209
1210 spin_lock(&xprt_list_lock);
1211 list_for_each_entry(t, &xprt_list, list) {
1212 if (t->ident == args->ident) {
1213 spin_unlock(&xprt_list_lock);
1214 goto found;
1215 }
1216 }
1217 spin_unlock(&xprt_list_lock);
1218 printk(KERN_ERR "RPC: transport (%d) not supported\n", args->ident);
1219 return ERR_PTR(-EIO);
1220
1221found:
1222 xprt = t->setup(args);
1223 if (IS_ERR(xprt)) {
1224 dprintk("RPC: xprt_create_transport: failed, %ld\n",
1225 -PTR_ERR(xprt));
1226 goto out;
1227 }
1228 INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
1229 if (xprt_has_timer(xprt))
1230 setup_timer(&xprt->timer, xprt_init_autodisconnect,
1231 (unsigned long)xprt);
1232 else
1233 init_timer(&xprt->timer);
1234
1235 if (strlen(args->servername) > RPC_MAXNETNAMELEN) {
1236 xprt_destroy(xprt);
1237 return ERR_PTR(-EINVAL);
1238 }
1239 xprt->servername = kstrdup(args->servername, GFP_KERNEL);
1240 if (xprt->servername == NULL) {
1241 xprt_destroy(xprt);
1242 return ERR_PTR(-ENOMEM);
1243 }
1244
1245 dprintk("RPC: created transport %p with %u slots\n", xprt,
1246 xprt->max_reqs);
1247out:
1248 return xprt;
1249}
1250
1251/**
1252 * xprt_destroy - destroy an RPC transport, killing off all requests.
1253 * @xprt: transport to destroy
1254 *
1255 */
1256static void xprt_destroy(struct rpc_xprt *xprt)
1257{
1258 dprintk("RPC: destroying transport %p\n", xprt);
1259 xprt->shutdown = 1;
1260 del_timer_sync(&xprt->timer);
1261
1262 rpc_destroy_wait_queue(&xprt->binding);
1263 rpc_destroy_wait_queue(&xprt->pending);
1264 rpc_destroy_wait_queue(&xprt->sending);
1265 rpc_destroy_wait_queue(&xprt->backlog);
1266 cancel_work_sync(&xprt->task_cleanup);
1267 kfree(xprt->servername);
1268 /*
1269 * Tear down transport state and free the rpc_xprt
1270 */
1271 xprt->ops->destroy(xprt);
1272}
1273
1274/**
1275 * xprt_put - release a reference to an RPC transport.
1276 * @xprt: pointer to the transport
1277 *
1278 */
1279void xprt_put(struct rpc_xprt *xprt)
1280{
1281 if (atomic_dec_and_test(&xprt->count))
1282 xprt_destroy(xprt);
1283}
1284
1285/**
1286 * xprt_get - return a reference to an RPC transport.
1287 * @xprt: pointer to the transport
1288 *
1289 */
1290struct rpc_xprt *xprt_get(struct rpc_xprt *xprt)
1291{
1292 if (atomic_inc_not_zero(&xprt->count))
1293 return xprt;
1294 return NULL;
1295}