Loading...
1/*
2 drbd_int.h
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
9
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
14
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
19
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
23
24*/
25
26#ifndef _DRBD_INT_H
27#define _DRBD_INT_H
28
29#include <linux/compiler.h>
30#include <linux/types.h>
31#include <linux/list.h>
32#include <linux/sched.h>
33#include <linux/bitops.h>
34#include <linux/slab.h>
35#include <linux/crypto.h>
36#include <linux/ratelimit.h>
37#include <linux/tcp.h>
38#include <linux/mutex.h>
39#include <linux/major.h>
40#include <linux/blkdev.h>
41#include <linux/genhd.h>
42#include <net/tcp.h>
43#include <linux/lru_cache.h>
44#include <linux/prefetch.h>
45
46#ifdef __CHECKER__
47# define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
48# define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
49# define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
50# define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
51#else
52# define __protected_by(x)
53# define __protected_read_by(x)
54# define __protected_write_by(x)
55# define __must_hold(x)
56#endif
57
58#define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
59
60/* module parameter, defined in drbd_main.c */
61extern unsigned int minor_count;
62extern bool disable_sendpage;
63extern bool allow_oos;
64extern unsigned int cn_idx;
65
66#ifdef CONFIG_DRBD_FAULT_INJECTION
67extern int enable_faults;
68extern int fault_rate;
69extern int fault_devs;
70#endif
71
72extern char usermode_helper[];
73
74
75/* I don't remember why XCPU ...
76 * This is used to wake the asender,
77 * and to interrupt sending the sending task
78 * on disconnect.
79 */
80#define DRBD_SIG SIGXCPU
81
82/* This is used to stop/restart our threads.
83 * Cannot use SIGTERM nor SIGKILL, since these
84 * are sent out by init on runlevel changes
85 * I choose SIGHUP for now.
86 */
87#define DRBD_SIGKILL SIGHUP
88
89/* All EEs on the free list should have ID_VACANT (== 0)
90 * freshly allocated EEs get !ID_VACANT (== 1)
91 * so if it says "cannot dereference null pointer at address 0x00000001",
92 * it is most likely one of these :( */
93
94#define ID_IN_SYNC (4711ULL)
95#define ID_OUT_OF_SYNC (4712ULL)
96
97#define ID_SYNCER (-1ULL)
98#define ID_VACANT 0
99#define is_syncer_block_id(id) ((id) == ID_SYNCER)
100#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
101
102struct drbd_conf;
103
104
105/* to shorten dev_warn(DEV, "msg"); and relatives statements */
106#define DEV (disk_to_dev(mdev->vdisk))
107
108#define D_ASSERT(exp) if (!(exp)) \
109 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
110
111#define ERR_IF(exp) if (({ \
112 int _b = (exp) != 0; \
113 if (_b) dev_err(DEV, "ASSERT FAILED: %s: (%s) in %s:%d\n", \
114 __func__, #exp, __FILE__, __LINE__); \
115 _b; \
116 }))
117
118/* Defines to control fault insertion */
119enum {
120 DRBD_FAULT_MD_WR = 0, /* meta data write */
121 DRBD_FAULT_MD_RD = 1, /* read */
122 DRBD_FAULT_RS_WR = 2, /* resync */
123 DRBD_FAULT_RS_RD = 3,
124 DRBD_FAULT_DT_WR = 4, /* data */
125 DRBD_FAULT_DT_RD = 5,
126 DRBD_FAULT_DT_RA = 6, /* data read ahead */
127 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
128 DRBD_FAULT_AL_EE = 8, /* alloc ee */
129 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
130
131 DRBD_FAULT_MAX,
132};
133
134extern unsigned int
135_drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
136
137static inline int
138drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
139#ifdef CONFIG_DRBD_FAULT_INJECTION
140 return fault_rate &&
141 (enable_faults & (1<<type)) &&
142 _drbd_insert_fault(mdev, type);
143#else
144 return 0;
145#endif
146}
147
148/* integer division, round _UP_ to the next integer */
149#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
150/* usual integer division */
151#define div_floor(A, B) ((A)/(B))
152
153/* drbd_meta-data.c (still in drbd_main.c) */
154/* 4th incarnation of the disk layout. */
155#define DRBD_MD_MAGIC (DRBD_MAGIC+4)
156
157extern struct drbd_conf **minor_table;
158extern struct ratelimit_state drbd_ratelimit_state;
159
160/* on the wire */
161enum drbd_packets {
162 /* receiver (data socket) */
163 P_DATA = 0x00,
164 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
165 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
166 P_BARRIER = 0x03,
167 P_BITMAP = 0x04,
168 P_BECOME_SYNC_TARGET = 0x05,
169 P_BECOME_SYNC_SOURCE = 0x06,
170 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
171 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
172 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
173 P_SYNC_PARAM = 0x0a,
174 P_PROTOCOL = 0x0b,
175 P_UUIDS = 0x0c,
176 P_SIZES = 0x0d,
177 P_STATE = 0x0e,
178 P_SYNC_UUID = 0x0f,
179 P_AUTH_CHALLENGE = 0x10,
180 P_AUTH_RESPONSE = 0x11,
181 P_STATE_CHG_REQ = 0x12,
182
183 /* asender (meta socket */
184 P_PING = 0x13,
185 P_PING_ACK = 0x14,
186 P_RECV_ACK = 0x15, /* Used in protocol B */
187 P_WRITE_ACK = 0x16, /* Used in protocol C */
188 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
189 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
190 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
191 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
192 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
193 P_BARRIER_ACK = 0x1c,
194 P_STATE_CHG_REPLY = 0x1d,
195
196 /* "new" commands, no longer fitting into the ordering scheme above */
197
198 P_OV_REQUEST = 0x1e, /* data socket */
199 P_OV_REPLY = 0x1f,
200 P_OV_RESULT = 0x20, /* meta socket */
201 P_CSUM_RS_REQUEST = 0x21, /* data socket */
202 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
203 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
204 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
205 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
206 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
207 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
208 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
209 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
210
211 P_MAX_CMD = 0x2A,
212 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
213 P_MAX_OPT_CMD = 0x101,
214
215 /* special command ids for handshake */
216
217 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
218 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
219
220 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
221};
222
223static inline const char *cmdname(enum drbd_packets cmd)
224{
225 /* THINK may need to become several global tables
226 * when we want to support more than
227 * one PRO_VERSION */
228 static const char *cmdnames[] = {
229 [P_DATA] = "Data",
230 [P_DATA_REPLY] = "DataReply",
231 [P_RS_DATA_REPLY] = "RSDataReply",
232 [P_BARRIER] = "Barrier",
233 [P_BITMAP] = "ReportBitMap",
234 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
235 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
236 [P_UNPLUG_REMOTE] = "UnplugRemote",
237 [P_DATA_REQUEST] = "DataRequest",
238 [P_RS_DATA_REQUEST] = "RSDataRequest",
239 [P_SYNC_PARAM] = "SyncParam",
240 [P_SYNC_PARAM89] = "SyncParam89",
241 [P_PROTOCOL] = "ReportProtocol",
242 [P_UUIDS] = "ReportUUIDs",
243 [P_SIZES] = "ReportSizes",
244 [P_STATE] = "ReportState",
245 [P_SYNC_UUID] = "ReportSyncUUID",
246 [P_AUTH_CHALLENGE] = "AuthChallenge",
247 [P_AUTH_RESPONSE] = "AuthResponse",
248 [P_PING] = "Ping",
249 [P_PING_ACK] = "PingAck",
250 [P_RECV_ACK] = "RecvAck",
251 [P_WRITE_ACK] = "WriteAck",
252 [P_RS_WRITE_ACK] = "RSWriteAck",
253 [P_DISCARD_ACK] = "DiscardAck",
254 [P_NEG_ACK] = "NegAck",
255 [P_NEG_DREPLY] = "NegDReply",
256 [P_NEG_RS_DREPLY] = "NegRSDReply",
257 [P_BARRIER_ACK] = "BarrierAck",
258 [P_STATE_CHG_REQ] = "StateChgRequest",
259 [P_STATE_CHG_REPLY] = "StateChgReply",
260 [P_OV_REQUEST] = "OVRequest",
261 [P_OV_REPLY] = "OVReply",
262 [P_OV_RESULT] = "OVResult",
263 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
264 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
265 [P_COMPRESSED_BITMAP] = "CBitmap",
266 [P_DELAY_PROBE] = "DelayProbe",
267 [P_OUT_OF_SYNC] = "OutOfSync",
268 [P_MAX_CMD] = NULL,
269 };
270
271 if (cmd == P_HAND_SHAKE_M)
272 return "HandShakeM";
273 if (cmd == P_HAND_SHAKE_S)
274 return "HandShakeS";
275 if (cmd == P_HAND_SHAKE)
276 return "HandShake";
277 if (cmd >= P_MAX_CMD)
278 return "Unknown";
279 return cmdnames[cmd];
280}
281
282/* for sending/receiving the bitmap,
283 * possibly in some encoding scheme */
284struct bm_xfer_ctx {
285 /* "const"
286 * stores total bits and long words
287 * of the bitmap, so we don't need to
288 * call the accessor functions over and again. */
289 unsigned long bm_bits;
290 unsigned long bm_words;
291 /* during xfer, current position within the bitmap */
292 unsigned long bit_offset;
293 unsigned long word_offset;
294
295 /* statistics; index: (h->command == P_BITMAP) */
296 unsigned packets[2];
297 unsigned bytes[2];
298};
299
300extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
301 const char *direction, struct bm_xfer_ctx *c);
302
303static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
304{
305 /* word_offset counts "native long words" (32 or 64 bit),
306 * aligned at 64 bit.
307 * Encoded packet may end at an unaligned bit offset.
308 * In case a fallback clear text packet is transmitted in
309 * between, we adjust this offset back to the last 64bit
310 * aligned "native long word", which makes coding and decoding
311 * the plain text bitmap much more convenient. */
312#if BITS_PER_LONG == 64
313 c->word_offset = c->bit_offset >> 6;
314#elif BITS_PER_LONG == 32
315 c->word_offset = c->bit_offset >> 5;
316 c->word_offset &= ~(1UL);
317#else
318# error "unsupported BITS_PER_LONG"
319#endif
320}
321
322#ifndef __packed
323#define __packed __attribute__((packed))
324#endif
325
326/* This is the layout for a packet on the wire.
327 * The byteorder is the network byte order.
328 * (except block_id and barrier fields.
329 * these are pointers to local structs
330 * and have no relevance for the partner,
331 * which just echoes them as received.)
332 *
333 * NOTE that the payload starts at a long aligned offset,
334 * regardless of 32 or 64 bit arch!
335 */
336struct p_header80 {
337 u32 magic;
338 u16 command;
339 u16 length; /* bytes of data after this header */
340 u8 payload[0];
341} __packed;
342
343/* Header for big packets, Used for data packets exceeding 64kB */
344struct p_header95 {
345 u16 magic; /* use DRBD_MAGIC_BIG here */
346 u16 command;
347 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
348 u8 payload[0];
349} __packed;
350
351union p_header {
352 struct p_header80 h80;
353 struct p_header95 h95;
354};
355
356/*
357 * short commands, packets without payload, plain p_header:
358 * P_PING
359 * P_PING_ACK
360 * P_BECOME_SYNC_TARGET
361 * P_BECOME_SYNC_SOURCE
362 * P_UNPLUG_REMOTE
363 */
364
365/*
366 * commands with out-of-struct payload:
367 * P_BITMAP (no additional fields)
368 * P_DATA, P_DATA_REPLY (see p_data)
369 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
370 */
371
372/* these defines must not be changed without changing the protocol version */
373#define DP_HARDBARRIER 1 /* depricated */
374#define DP_RW_SYNC 2 /* equals REQ_SYNC */
375#define DP_MAY_SET_IN_SYNC 4
376#define DP_UNPLUG 8 /* not used anymore */
377#define DP_FUA 16 /* equals REQ_FUA */
378#define DP_FLUSH 32 /* equals REQ_FLUSH */
379#define DP_DISCARD 64 /* equals REQ_DISCARD */
380
381struct p_data {
382 union p_header head;
383 u64 sector; /* 64 bits sector number */
384 u64 block_id; /* to identify the request in protocol B&C */
385 u32 seq_num;
386 u32 dp_flags;
387} __packed;
388
389/*
390 * commands which share a struct:
391 * p_block_ack:
392 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
393 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
394 * p_block_req:
395 * P_DATA_REQUEST, P_RS_DATA_REQUEST
396 */
397struct p_block_ack {
398 struct p_header80 head;
399 u64 sector;
400 u64 block_id;
401 u32 blksize;
402 u32 seq_num;
403} __packed;
404
405
406struct p_block_req {
407 struct p_header80 head;
408 u64 sector;
409 u64 block_id;
410 u32 blksize;
411 u32 pad; /* to multiple of 8 Byte */
412} __packed;
413
414/*
415 * commands with their own struct for additional fields:
416 * P_HAND_SHAKE
417 * P_BARRIER
418 * P_BARRIER_ACK
419 * P_SYNC_PARAM
420 * ReportParams
421 */
422
423struct p_handshake {
424 struct p_header80 head; /* 8 bytes */
425 u32 protocol_min;
426 u32 feature_flags;
427 u32 protocol_max;
428
429 /* should be more than enough for future enhancements
430 * for now, feature_flags and the reserverd array shall be zero.
431 */
432
433 u32 _pad;
434 u64 reserverd[7];
435} __packed;
436/* 80 bytes, FIXED for the next century */
437
438struct p_barrier {
439 struct p_header80 head;
440 u32 barrier; /* barrier number _handle_ only */
441 u32 pad; /* to multiple of 8 Byte */
442} __packed;
443
444struct p_barrier_ack {
445 struct p_header80 head;
446 u32 barrier;
447 u32 set_size;
448} __packed;
449
450struct p_rs_param {
451 struct p_header80 head;
452 u32 rate;
453
454 /* Since protocol version 88 and higher. */
455 char verify_alg[0];
456} __packed;
457
458struct p_rs_param_89 {
459 struct p_header80 head;
460 u32 rate;
461 /* protocol version 89: */
462 char verify_alg[SHARED_SECRET_MAX];
463 char csums_alg[SHARED_SECRET_MAX];
464} __packed;
465
466struct p_rs_param_95 {
467 struct p_header80 head;
468 u32 rate;
469 char verify_alg[SHARED_SECRET_MAX];
470 char csums_alg[SHARED_SECRET_MAX];
471 u32 c_plan_ahead;
472 u32 c_delay_target;
473 u32 c_fill_target;
474 u32 c_max_rate;
475} __packed;
476
477enum drbd_conn_flags {
478 CF_WANT_LOSE = 1,
479 CF_DRY_RUN = 2,
480};
481
482struct p_protocol {
483 struct p_header80 head;
484 u32 protocol;
485 u32 after_sb_0p;
486 u32 after_sb_1p;
487 u32 after_sb_2p;
488 u32 conn_flags;
489 u32 two_primaries;
490
491 /* Since protocol version 87 and higher. */
492 char integrity_alg[0];
493
494} __packed;
495
496struct p_uuids {
497 struct p_header80 head;
498 u64 uuid[UI_EXTENDED_SIZE];
499} __packed;
500
501struct p_rs_uuid {
502 struct p_header80 head;
503 u64 uuid;
504} __packed;
505
506struct p_sizes {
507 struct p_header80 head;
508 u64 d_size; /* size of disk */
509 u64 u_size; /* user requested size */
510 u64 c_size; /* current exported size */
511 u32 max_bio_size; /* Maximal size of a BIO */
512 u16 queue_order_type; /* not yet implemented in DRBD*/
513 u16 dds_flags; /* use enum dds_flags here. */
514} __packed;
515
516struct p_state {
517 struct p_header80 head;
518 u32 state;
519} __packed;
520
521struct p_req_state {
522 struct p_header80 head;
523 u32 mask;
524 u32 val;
525} __packed;
526
527struct p_req_state_reply {
528 struct p_header80 head;
529 u32 retcode;
530} __packed;
531
532struct p_drbd06_param {
533 u64 size;
534 u32 state;
535 u32 blksize;
536 u32 protocol;
537 u32 version;
538 u32 gen_cnt[5];
539 u32 bit_map_gen[5];
540} __packed;
541
542struct p_discard {
543 struct p_header80 head;
544 u64 block_id;
545 u32 seq_num;
546 u32 pad;
547} __packed;
548
549struct p_block_desc {
550 struct p_header80 head;
551 u64 sector;
552 u32 blksize;
553 u32 pad; /* to multiple of 8 Byte */
554} __packed;
555
556/* Valid values for the encoding field.
557 * Bump proto version when changing this. */
558enum drbd_bitmap_code {
559 /* RLE_VLI_Bytes = 0,
560 * and other bit variants had been defined during
561 * algorithm evaluation. */
562 RLE_VLI_Bits = 2,
563};
564
565struct p_compressed_bm {
566 struct p_header80 head;
567 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
568 * (encoding & 0x80): polarity (set/unset) of first runlength
569 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
570 * used to pad up to head.length bytes
571 */
572 u8 encoding;
573
574 u8 code[0];
575} __packed;
576
577struct p_delay_probe93 {
578 struct p_header80 head;
579 u32 seq_num; /* sequence number to match the two probe packets */
580 u32 offset; /* usecs the probe got sent after the reference time point */
581} __packed;
582
583/* DCBP: Drbd Compressed Bitmap Packet ... */
584static inline enum drbd_bitmap_code
585DCBP_get_code(struct p_compressed_bm *p)
586{
587 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
588}
589
590static inline void
591DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
592{
593 BUG_ON(code & ~0xf);
594 p->encoding = (p->encoding & ~0xf) | code;
595}
596
597static inline int
598DCBP_get_start(struct p_compressed_bm *p)
599{
600 return (p->encoding & 0x80) != 0;
601}
602
603static inline void
604DCBP_set_start(struct p_compressed_bm *p, int set)
605{
606 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
607}
608
609static inline int
610DCBP_get_pad_bits(struct p_compressed_bm *p)
611{
612 return (p->encoding >> 4) & 0x7;
613}
614
615static inline void
616DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
617{
618 BUG_ON(n & ~0x7);
619 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
620}
621
622/* one bitmap packet, including the p_header,
623 * should fit within one _architecture independend_ page.
624 * so we need to use the fixed size 4KiB page size
625 * most architectures have used for a long time.
626 */
627#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
628#define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
629#define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
630#if (PAGE_SIZE < 4096)
631/* drbd_send_bitmap / receive_bitmap would break horribly */
632#error "PAGE_SIZE too small"
633#endif
634
635union p_polymorph {
636 union p_header header;
637 struct p_handshake handshake;
638 struct p_data data;
639 struct p_block_ack block_ack;
640 struct p_barrier barrier;
641 struct p_barrier_ack barrier_ack;
642 struct p_rs_param_89 rs_param_89;
643 struct p_rs_param_95 rs_param_95;
644 struct p_protocol protocol;
645 struct p_sizes sizes;
646 struct p_uuids uuids;
647 struct p_state state;
648 struct p_req_state req_state;
649 struct p_req_state_reply req_state_reply;
650 struct p_block_req block_req;
651 struct p_delay_probe93 delay_probe93;
652 struct p_rs_uuid rs_uuid;
653 struct p_block_desc block_desc;
654} __packed;
655
656/**********************************************************************/
657enum drbd_thread_state {
658 None,
659 Running,
660 Exiting,
661 Restarting
662};
663
664struct drbd_thread {
665 spinlock_t t_lock;
666 struct task_struct *task;
667 struct completion stop;
668 enum drbd_thread_state t_state;
669 int (*function) (struct drbd_thread *);
670 struct drbd_conf *mdev;
671 int reset_cpu_mask;
672};
673
674static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
675{
676 /* THINK testing the t_state seems to be uncritical in all cases
677 * (but thread_{start,stop}), so we can read it *without* the lock.
678 * --lge */
679
680 smp_rmb();
681 return thi->t_state;
682}
683
684struct drbd_work;
685typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
686struct drbd_work {
687 struct list_head list;
688 drbd_work_cb cb;
689};
690
691struct drbd_tl_epoch;
692struct drbd_request {
693 struct drbd_work w;
694 struct drbd_conf *mdev;
695
696 /* if local IO is not allowed, will be NULL.
697 * if local IO _is_ allowed, holds the locally submitted bio clone,
698 * or, after local IO completion, the ERR_PTR(error).
699 * see drbd_endio_pri(). */
700 struct bio *private_bio;
701
702 struct hlist_node collision;
703 sector_t sector;
704 unsigned int size;
705 unsigned int epoch; /* barrier_nr */
706
707 /* barrier_nr: used to check on "completion" whether this req was in
708 * the current epoch, and we therefore have to close it,
709 * starting a new epoch...
710 */
711
712 struct list_head tl_requests; /* ring list in the transfer log */
713 struct bio *master_bio; /* master bio pointer */
714 unsigned long rq_state; /* see comments above _req_mod() */
715 unsigned long start_time;
716};
717
718struct drbd_tl_epoch {
719 struct drbd_work w;
720 struct list_head requests; /* requests before */
721 struct drbd_tl_epoch *next; /* pointer to the next barrier */
722 unsigned int br_number; /* the barriers identifier. */
723 int n_writes; /* number of requests attached before this barrier */
724};
725
726struct drbd_request;
727
728/* These Tl_epoch_entries may be in one of 6 lists:
729 active_ee .. data packet being written
730 sync_ee .. syncer block being written
731 done_ee .. block written, need to send P_WRITE_ACK
732 read_ee .. [RS]P_DATA_REQUEST being read
733*/
734
735struct drbd_epoch {
736 struct list_head list;
737 unsigned int barrier_nr;
738 atomic_t epoch_size; /* increased on every request added. */
739 atomic_t active; /* increased on every req. added, and dec on every finished. */
740 unsigned long flags;
741};
742
743/* drbd_epoch flag bits */
744enum {
745 DE_HAVE_BARRIER_NUMBER,
746};
747
748enum epoch_event {
749 EV_PUT,
750 EV_GOT_BARRIER_NR,
751 EV_BECAME_LAST,
752 EV_CLEANUP = 32, /* used as flag */
753};
754
755struct drbd_wq_barrier {
756 struct drbd_work w;
757 struct completion done;
758};
759
760struct digest_info {
761 int digest_size;
762 void *digest;
763};
764
765struct drbd_epoch_entry {
766 struct drbd_work w;
767 struct hlist_node collision;
768 struct drbd_epoch *epoch; /* for writes */
769 struct drbd_conf *mdev;
770 struct page *pages;
771 atomic_t pending_bios;
772 unsigned int size;
773 /* see comments on ee flag bits below */
774 unsigned long flags;
775 sector_t sector;
776 union {
777 u64 block_id;
778 struct digest_info *digest;
779 };
780};
781
782/* ee flag bits.
783 * While corresponding bios are in flight, the only modification will be
784 * set_bit WAS_ERROR, which has to be atomic.
785 * If no bios are in flight yet, or all have been completed,
786 * non-atomic modification to ee->flags is ok.
787 */
788enum {
789 __EE_CALL_AL_COMPLETE_IO,
790 __EE_MAY_SET_IN_SYNC,
791
792 /* In case a barrier failed,
793 * we need to resubmit without the barrier flag. */
794 __EE_RESUBMITTED,
795
796 /* we may have several bios per epoch entry.
797 * if any of those fail, we set this flag atomically
798 * from the endio callback */
799 __EE_WAS_ERROR,
800
801 /* This ee has a pointer to a digest instead of a block id */
802 __EE_HAS_DIGEST,
803};
804#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
805#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
806#define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
807#define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
808#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
809
810/* global flag bits */
811enum {
812 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
813 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
814 SEND_PING, /* whether asender should send a ping asap */
815
816 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
817 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
818 MD_DIRTY, /* current uuids and flags not yet on disk */
819 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
820 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
821 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
822 CL_ST_CHG_SUCCESS,
823 CL_ST_CHG_FAIL,
824 CRASHED_PRIMARY, /* This node was a crashed primary.
825 * Gets cleared when the state.conn
826 * goes into C_CONNECTED state. */
827 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
828 CONSIDER_RESYNC,
829
830 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
831 SUSPEND_IO, /* suspend application io */
832 BITMAP_IO, /* suspend application io;
833 once no more io in flight, start bitmap io */
834 BITMAP_IO_QUEUED, /* Started bitmap IO */
835 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
836 WAS_IO_ERROR, /* Local disk failed returned IO error */
837 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
838 NET_CONGESTED, /* The data socket is congested */
839
840 CONFIG_PENDING, /* serialization of (re)configuration requests.
841 * if set, also prevents the device from dying */
842 DEVICE_DYING, /* device became unconfigured,
843 * but worker thread is still handling the cleanup.
844 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
845 * while this is set. */
846 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
847 * the peer, if it changed there as well. */
848 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
849 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
850 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
851 AL_SUSPENDED, /* Activity logging is currently suspended. */
852 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
853 STATE_SENT, /* Do not change state/UUIDs while this is set */
854};
855
856struct drbd_bitmap; /* opaque for drbd_conf */
857
858/* definition of bits in bm_flags to be used in drbd_bm_lock
859 * and drbd_bitmap_io and friends. */
860enum bm_flag {
861 /* do we need to kfree, or vfree bm_pages? */
862 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
863
864 /* currently locked for bulk operation */
865 BM_LOCKED_MASK = 0xf,
866
867 /* in detail, that is: */
868 BM_DONT_CLEAR = 0x1,
869 BM_DONT_SET = 0x2,
870 BM_DONT_TEST = 0x4,
871
872 /* so we can mark it locked for bulk operation,
873 * and still allow all non-bulk operations */
874 BM_IS_LOCKED = 0x8,
875
876 /* (test bit, count bit) allowed (common case) */
877 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
878
879 /* testing bits, as well as setting new bits allowed, but clearing bits
880 * would be unexpected. Used during bitmap receive. Setting new bits
881 * requires sending of "out-of-sync" information, though. */
882 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
883
884 /* for drbd_bm_write_copy_pages, everything is allowed,
885 * only concurrent bulk operations are locked out. */
886 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
887};
888
889struct drbd_work_queue {
890 struct list_head q;
891 struct semaphore s; /* producers up it, worker down()s it */
892 spinlock_t q_lock; /* to protect the list. */
893};
894
895struct drbd_socket {
896 struct drbd_work_queue work;
897 struct mutex mutex;
898 struct socket *socket;
899 /* this way we get our
900 * send/receive buffers off the stack */
901 union p_polymorph sbuf;
902 union p_polymorph rbuf;
903};
904
905struct drbd_md {
906 u64 md_offset; /* sector offset to 'super' block */
907
908 u64 la_size_sect; /* last agreed size, unit sectors */
909 u64 uuid[UI_SIZE];
910 u64 device_uuid;
911 u32 flags;
912 u32 md_size_sect;
913
914 s32 al_offset; /* signed relative sector offset to al area */
915 s32 bm_offset; /* signed relative sector offset to bitmap */
916
917 /* u32 al_nr_extents; important for restoring the AL
918 * is stored into sync_conf.al_extents, which in turn
919 * gets applied to act_log->nr_elements
920 */
921};
922
923/* for sync_conf and other types... */
924#define NL_PACKET(name, number, fields) struct name { fields };
925#define NL_INTEGER(pn,pr,member) int member;
926#define NL_INT64(pn,pr,member) __u64 member;
927#define NL_BIT(pn,pr,member) unsigned member:1;
928#define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
929#include <linux/drbd_nl.h>
930
931struct drbd_backing_dev {
932 struct block_device *backing_bdev;
933 struct block_device *md_bdev;
934 struct drbd_md md;
935 struct disk_conf dc; /* The user provided config... */
936 sector_t known_size; /* last known size of that backing device */
937};
938
939struct drbd_md_io {
940 unsigned int done;
941 int error;
942};
943
944struct bm_io_work {
945 struct drbd_work w;
946 char *why;
947 enum bm_flag flags;
948 int (*io_fn)(struct drbd_conf *mdev);
949 void (*done)(struct drbd_conf *mdev, int rv);
950};
951
952enum write_ordering_e {
953 WO_none,
954 WO_drain_io,
955 WO_bdev_flush,
956};
957
958struct fifo_buffer {
959 int *values;
960 unsigned int head_index;
961 unsigned int size;
962};
963
964struct drbd_conf {
965 /* things that are stored as / read from meta data on disk */
966 unsigned long flags;
967
968 /* configured by drbdsetup */
969 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
970 struct syncer_conf sync_conf;
971 struct drbd_backing_dev *ldev __protected_by(local);
972
973 sector_t p_size; /* partner's disk size */
974 struct request_queue *rq_queue;
975 struct block_device *this_bdev;
976 struct gendisk *vdisk;
977
978 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
979 struct drbd_socket meta; /* ping/ack (metadata) packets */
980 int agreed_pro_version; /* actually used protocol version */
981 unsigned long last_received; /* in jiffies, either socket */
982 unsigned int ko_count;
983 struct drbd_work resync_work,
984 unplug_work,
985 go_diskless,
986 md_sync_work,
987 start_resync_work;
988 struct timer_list resync_timer;
989 struct timer_list md_sync_timer;
990 struct timer_list start_resync_timer;
991 struct timer_list request_timer;
992#ifdef DRBD_DEBUG_MD_SYNC
993 struct {
994 unsigned int line;
995 const char* func;
996 } last_md_mark_dirty;
997#endif
998
999 /* Used after attach while negotiating new disk state. */
1000 union drbd_state new_state_tmp;
1001
1002 union drbd_state state;
1003 wait_queue_head_t misc_wait;
1004 wait_queue_head_t state_wait; /* upon each state change. */
1005 wait_queue_head_t net_cnt_wait;
1006 unsigned int send_cnt;
1007 unsigned int recv_cnt;
1008 unsigned int read_cnt;
1009 unsigned int writ_cnt;
1010 unsigned int al_writ_cnt;
1011 unsigned int bm_writ_cnt;
1012 atomic_t ap_bio_cnt; /* Requests we need to complete */
1013 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1014 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1015 atomic_t unacked_cnt; /* Need to send replys for */
1016 atomic_t local_cnt; /* Waiting for local completion */
1017 atomic_t net_cnt; /* Users of net_conf */
1018 spinlock_t req_lock;
1019 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1020 struct drbd_tl_epoch *newest_tle;
1021 struct drbd_tl_epoch *oldest_tle;
1022 struct list_head out_of_sequence_requests;
1023 struct list_head barrier_acked_requests;
1024 struct hlist_head *tl_hash;
1025 unsigned int tl_hash_s;
1026
1027 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
1028 unsigned long rs_total;
1029 /* number of resync blocks that failed in this run */
1030 unsigned long rs_failed;
1031 /* Syncer's start time [unit jiffies] */
1032 unsigned long rs_start;
1033 /* cumulated time in PausedSyncX state [unit jiffies] */
1034 unsigned long rs_paused;
1035 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1036 unsigned long rs_same_csum;
1037#define DRBD_SYNC_MARKS 8
1038#define DRBD_SYNC_MARK_STEP (3*HZ)
1039 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1040 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
1041 /* marks's time [unit jiffies] */
1042 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1043 /* current index into rs_mark_{left,time} */
1044 int rs_last_mark;
1045
1046 /* where does the admin want us to start? (sector) */
1047 sector_t ov_start_sector;
1048 /* where are we now? (sector) */
1049 sector_t ov_position;
1050 /* Start sector of out of sync range (to merge printk reporting). */
1051 sector_t ov_last_oos_start;
1052 /* size of out-of-sync range in sectors. */
1053 sector_t ov_last_oos_size;
1054 unsigned long ov_left; /* in bits */
1055 struct crypto_hash *csums_tfm;
1056 struct crypto_hash *verify_tfm;
1057
1058 unsigned long last_reattach_jif;
1059 unsigned long last_reconnect_jif;
1060 struct drbd_thread receiver;
1061 struct drbd_thread worker;
1062 struct drbd_thread asender;
1063 struct drbd_bitmap *bitmap;
1064 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1065
1066 /* Used to track operations of resync... */
1067 struct lru_cache *resync;
1068 /* Number of locked elements in resync LRU */
1069 unsigned int resync_locked;
1070 /* resync extent number waiting for application requests */
1071 unsigned int resync_wenr;
1072
1073 int open_cnt;
1074 u64 *p_uuid;
1075 struct drbd_epoch *current_epoch;
1076 spinlock_t epoch_lock;
1077 unsigned int epochs;
1078 enum write_ordering_e write_ordering;
1079 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1080 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
1081 struct list_head done_ee; /* send ack */
1082 struct list_head read_ee; /* IO in progress (any read) */
1083 struct list_head net_ee; /* zero-copy network send in progress */
1084 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1085 unsigned int ee_hash_s;
1086
1087 /* this one is protected by ee_lock, single thread */
1088 struct drbd_epoch_entry *last_write_w_barrier;
1089
1090 int next_barrier_nr;
1091 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1092 struct list_head resync_reads;
1093 atomic_t pp_in_use; /* allocated from page pool */
1094 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
1095 wait_queue_head_t ee_wait;
1096 struct page *md_io_page; /* one page buffer for md_io */
1097 struct page *md_io_tmpp; /* for logical_block_size != 512 */
1098 struct drbd_md_io md_io;
1099 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */
1100 spinlock_t al_lock;
1101 wait_queue_head_t al_wait;
1102 struct lru_cache *act_log; /* activity log */
1103 unsigned int al_tr_number;
1104 int al_tr_cycle;
1105 int al_tr_pos; /* position of the next transaction in the journal */
1106 struct crypto_hash *cram_hmac_tfm;
1107 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1108 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1109 void *int_dig_out;
1110 void *int_dig_in;
1111 void *int_dig_vv;
1112 wait_queue_head_t seq_wait;
1113 atomic_t packet_seq;
1114 unsigned int peer_seq;
1115 spinlock_t peer_seq_lock;
1116 unsigned int minor;
1117 unsigned long comm_bm_set; /* communicated number of set bits. */
1118 cpumask_var_t cpu_mask;
1119 struct bm_io_work bm_io_work;
1120 u64 ed_uuid; /* UUID of the exposed data */
1121 struct mutex state_mutex;
1122 char congestion_reason; /* Why we where congested... */
1123 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1124 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1125 int rs_last_sect_ev; /* counter to compare with */
1126 int rs_last_events; /* counter of read or write "events" (unit sectors)
1127 * on the lower level device when we last looked. */
1128 int c_sync_rate; /* current resync rate after syncer throttle magic */
1129 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1130 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
1131 int rs_planed; /* resync sectors already planned */
1132 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
1133 int peer_max_bio_size;
1134 int local_max_bio_size;
1135};
1136
1137static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1138{
1139 struct drbd_conf *mdev;
1140
1141 mdev = minor < minor_count ? minor_table[minor] : NULL;
1142
1143 return mdev;
1144}
1145
1146static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1147{
1148 return mdev->minor;
1149}
1150
1151/* returns 1 if it was successful,
1152 * returns 0 if there was no data socket.
1153 * so wherever you are going to use the data.socket, e.g. do
1154 * if (!drbd_get_data_sock(mdev))
1155 * return 0;
1156 * CODE();
1157 * drbd_put_data_sock(mdev);
1158 */
1159static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1160{
1161 mutex_lock(&mdev->data.mutex);
1162 /* drbd_disconnect() could have called drbd_free_sock()
1163 * while we were waiting in down()... */
1164 if (unlikely(mdev->data.socket == NULL)) {
1165 mutex_unlock(&mdev->data.mutex);
1166 return 0;
1167 }
1168 return 1;
1169}
1170
1171static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1172{
1173 mutex_unlock(&mdev->data.mutex);
1174}
1175
1176/*
1177 * function declarations
1178 *************************/
1179
1180/* drbd_main.c */
1181
1182enum chg_state_flags {
1183 CS_HARD = 1,
1184 CS_VERBOSE = 2,
1185 CS_WAIT_COMPLETE = 4,
1186 CS_SERIALIZE = 8,
1187 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1188};
1189
1190enum dds_flags {
1191 DDSF_FORCED = 1,
1192 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1193};
1194
1195extern void drbd_init_set_defaults(struct drbd_conf *mdev);
1196extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev,
1197 enum chg_state_flags f,
1198 union drbd_state mask,
1199 union drbd_state val);
1200extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1201 union drbd_state);
1202extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *,
1203 union drbd_state,
1204 union drbd_state,
1205 enum chg_state_flags);
1206extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state,
1207 enum chg_state_flags,
1208 struct completion *done);
1209extern void print_st_err(struct drbd_conf *, union drbd_state,
1210 union drbd_state, int);
1211extern int drbd_thread_start(struct drbd_thread *thi);
1212extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1213#ifdef CONFIG_SMP
1214extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1215extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1216#else
1217#define drbd_thread_current_set_cpu(A) ({})
1218#define drbd_calc_cpu_mask(A) ({})
1219#endif
1220extern void drbd_free_resources(struct drbd_conf *mdev);
1221extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1222 unsigned int set_size);
1223extern void tl_clear(struct drbd_conf *mdev);
1224extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1225extern void drbd_free_sock(struct drbd_conf *mdev);
1226extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1227 void *buf, size_t size, unsigned msg_flags);
1228extern int drbd_send_protocol(struct drbd_conf *mdev);
1229extern int drbd_send_uuids(struct drbd_conf *mdev);
1230extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
1231extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
1232extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
1233extern int drbd_send_state(struct drbd_conf *mdev, union drbd_state s);
1234extern int drbd_send_current_state(struct drbd_conf *mdev);
1235extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
1236 enum drbd_packets cmd, struct p_header80 *h,
1237 size_t size, unsigned msg_flags);
1238#define USE_DATA_SOCKET 1
1239#define USE_META_SOCKET 0
1240extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
1241 enum drbd_packets cmd, struct p_header80 *h,
1242 size_t size);
1243extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1244 char *data, size_t size);
1245extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1246extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1247 u32 set_size);
1248extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1249 struct drbd_epoch_entry *e);
1250extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1251 struct p_block_req *rp);
1252extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
1253 struct p_data *dp, int data_size);
1254extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1255 sector_t sector, int blksize, u64 block_id);
1256extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
1257extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1258 struct drbd_epoch_entry *e);
1259extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
1260extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1261 sector_t sector, int size, u64 block_id);
1262extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1263 sector_t sector,int size,
1264 void *digest, int digest_size,
1265 enum drbd_packets cmd);
1266extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1267
1268extern int drbd_send_bitmap(struct drbd_conf *mdev);
1269extern int _drbd_send_bitmap(struct drbd_conf *mdev);
1270extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
1271extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1272extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
1273void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
1274
1275extern void drbd_md_sync(struct drbd_conf *mdev);
1276extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
1277extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1278extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1279extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1280extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1281extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1282extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1283extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1284extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
1285#ifndef DRBD_DEBUG_MD_SYNC
1286extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
1287#else
1288#define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1289extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1290 unsigned int line, const char *func);
1291#endif
1292extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1293 int (*io_fn)(struct drbd_conf *),
1294 void (*done)(struct drbd_conf *, int),
1295 char *why, enum bm_flag flags);
1296extern int drbd_bitmap_io(struct drbd_conf *mdev,
1297 int (*io_fn)(struct drbd_conf *),
1298 char *why, enum bm_flag flags);
1299extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1300extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
1301extern void drbd_go_diskless(struct drbd_conf *mdev);
1302extern void drbd_ldev_destroy(struct drbd_conf *mdev);
1303
1304
1305/* Meta data layout
1306 We reserve a 128MB Block (4k aligned)
1307 * either at the end of the backing device
1308 * or on a separate meta data device. */
1309
1310#define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1311/* The following numbers are sectors */
1312#define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1313#define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1314/* Allows up to about 3.8TB */
1315#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1316
1317/* Since the smalles IO unit is usually 512 byte */
1318#define MD_SECTOR_SHIFT 9
1319#define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1320
1321/* activity log */
1322#define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1323#define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1324#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1325
1326#if BITS_PER_LONG == 32
1327#define LN2_BPL 5
1328#define cpu_to_lel(A) cpu_to_le32(A)
1329#define lel_to_cpu(A) le32_to_cpu(A)
1330#elif BITS_PER_LONG == 64
1331#define LN2_BPL 6
1332#define cpu_to_lel(A) cpu_to_le64(A)
1333#define lel_to_cpu(A) le64_to_cpu(A)
1334#else
1335#error "LN2 of BITS_PER_LONG unknown!"
1336#endif
1337
1338/* resync bitmap */
1339/* 16MB sized 'bitmap extent' to track syncer usage */
1340struct bm_extent {
1341 int rs_left; /* number of bits set (out of sync) in this extent. */
1342 int rs_failed; /* number of failed resync requests in this extent. */
1343 unsigned long flags;
1344 struct lc_element lce;
1345};
1346
1347#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1348#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
1349#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
1350
1351/* drbd_bitmap.c */
1352/*
1353 * We need to store one bit for a block.
1354 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1355 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1356 * Bit 1 ==> local node thinks this block needs to be synced.
1357 */
1358
1359#define SLEEP_TIME (HZ/10)
1360
1361#define BM_BLOCK_SHIFT 12 /* 4k per bit */
1362#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1363/* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1364 * per sector of on disk bitmap */
1365#define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1366#define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1367
1368#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1369#error "HAVE YOU FIXED drbdmeta AS WELL??"
1370#endif
1371
1372/* thus many _storage_ sectors are described by one bit */
1373#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1374#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1375#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1376
1377/* bit to represented kilo byte conversion */
1378#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1379
1380/* in which _bitmap_ extent (resp. sector) the bit for a certain
1381 * _storage_ sector is located in */
1382#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1383
1384/* how much _storage_ sectors we have per bitmap sector */
1385#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1386#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1387
1388/* in one sector of the bitmap, we have this many activity_log extents. */
1389#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1390#define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1391
1392#define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1393#define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1394
1395/* the extent in "PER_EXTENT" below is an activity log extent
1396 * we need that many (long words/bytes) to store the bitmap
1397 * of one AL_EXTENT_SIZE chunk of storage.
1398 * we can store the bitmap for that many AL_EXTENTS within
1399 * one sector of the _on_disk_ bitmap:
1400 * bit 0 bit 37 bit 38 bit (512*8)-1
1401 * ...|........|........|.. // ..|........|
1402 * sect. 0 `296 `304 ^(512*8*8)-1
1403 *
1404#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1405#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1406#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1407 */
1408
1409#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1410#define DRBD_MAX_SECTORS_BM \
1411 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1412#if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1413#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1414#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
1415#elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
1416#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1417#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1418#else
1419#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1420/* 16 TB in units of sectors */
1421#if BITS_PER_LONG == 32
1422/* adjust by one page worth of bitmap,
1423 * so we won't wrap around in drbd_bm_find_next_bit.
1424 * you should use 64bit OS for that much storage, anyways. */
1425#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1426#else
1427/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1428#define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1429/* corresponds to (1UL << 38) bits right now. */
1430#endif
1431#endif
1432
1433/* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
1434 * With a value of 8 all IO in one 128K block make it to the same slot of the
1435 * hash table. */
1436#define HT_SHIFT 8
1437#define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
1438#define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
1439
1440#define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
1441
1442/* Number of elements in the app_reads_hash */
1443#define APP_R_HSIZE 15
1444
1445extern int drbd_bm_init(struct drbd_conf *mdev);
1446extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
1447extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1448extern void drbd_bm_set_all(struct drbd_conf *mdev);
1449extern void drbd_bm_clear_all(struct drbd_conf *mdev);
1450/* set/clear/test only a few bits at a time */
1451extern int drbd_bm_set_bits(
1452 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1453extern int drbd_bm_clear_bits(
1454 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1455extern int drbd_bm_count_bits(
1456 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1457/* bm_set_bits variant for use while holding drbd_bm_lock,
1458 * may process the whole bitmap in one go */
1459extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1460 const unsigned long s, const unsigned long e);
1461extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1462extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
1463extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
1464extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1465extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1466extern int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local);
1467extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1468 unsigned long al_enr);
1469extern size_t drbd_bm_words(struct drbd_conf *mdev);
1470extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1471extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
1472
1473#define DRBD_END_OF_BITMAP (~(unsigned long)0)
1474extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1475/* bm_find_next variants for use while you hold drbd_bm_lock() */
1476extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1477extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
1478extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
1479extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1480extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1481/* for receive_bitmap */
1482extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1483 size_t number, unsigned long *buffer);
1484/* for _drbd_send_bitmap */
1485extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1486 size_t number, unsigned long *buffer);
1487
1488extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
1489extern void drbd_bm_unlock(struct drbd_conf *mdev);
1490/* drbd_main.c */
1491
1492extern struct kmem_cache *drbd_request_cache;
1493extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1494extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1495extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1496extern mempool_t *drbd_request_mempool;
1497extern mempool_t *drbd_ee_mempool;
1498
1499/* drbd's page pool, used to buffer data received from the peer,
1500 * or data requested by the peer.
1501 *
1502 * This does not have an emergency reserve.
1503 *
1504 * When allocating from this pool, it first takes pages from the pool.
1505 * Only if the pool is depleted will try to allocate from the system.
1506 *
1507 * The assumption is that pages taken from this pool will be processed,
1508 * and given back, "quickly", and then can be recycled, so we can avoid
1509 * frequent calls to alloc_page(), and still will be able to make progress even
1510 * under memory pressure.
1511 */
1512extern struct page *drbd_pp_pool;
1513extern spinlock_t drbd_pp_lock;
1514extern int drbd_pp_vacant;
1515extern wait_queue_head_t drbd_pp_wait;
1516
1517/* We also need a standard (emergency-reserve backed) page pool
1518 * for meta data IO (activity log, bitmap).
1519 * We can keep it global, as long as it is used as "N pages at a time".
1520 * 128 should be plenty, currently we probably can get away with as few as 1.
1521 */
1522#define DRBD_MIN_POOL_PAGES 128
1523extern mempool_t *drbd_md_io_page_pool;
1524
1525/* We also need to make sure we get a bio
1526 * when we need it for housekeeping purposes */
1527extern struct bio_set *drbd_md_io_bio_set;
1528/* to allocate from that set */
1529extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1530
1531extern rwlock_t global_state_lock;
1532
1533extern struct drbd_conf *drbd_new_device(unsigned int minor);
1534extern void drbd_free_mdev(struct drbd_conf *mdev);
1535
1536extern int proc_details;
1537
1538/* drbd_req */
1539extern void drbd_make_request(struct request_queue *q, struct bio *bio);
1540extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1541extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1542extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1543
1544
1545/* drbd_nl.c */
1546extern void drbd_suspend_io(struct drbd_conf *mdev);
1547extern void drbd_resume_io(struct drbd_conf *mdev);
1548extern char *ppsize(char *buf, unsigned long long size);
1549extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
1550enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
1551extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
1552extern void resync_after_online_grow(struct drbd_conf *);
1553extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
1554extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1555 enum drbd_role new_role,
1556 int force);
1557extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1558extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
1559extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1560
1561/* drbd_worker.c */
1562extern int drbd_worker(struct drbd_thread *thi);
1563extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1564extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1565extern void resume_next_sg(struct drbd_conf *mdev);
1566extern void suspend_other_sg(struct drbd_conf *mdev);
1567extern int drbd_resync_finished(struct drbd_conf *mdev);
1568/* maybe rather drbd_main.c ? */
1569extern void *drbd_md_get_buffer(struct drbd_conf *mdev);
1570extern void drbd_md_put_buffer(struct drbd_conf *mdev);
1571extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1572 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1573extern void wait_until_done_or_disk_failure(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
1574 unsigned int *done);
1575extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
1576extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
1577
1578static inline void ov_oos_print(struct drbd_conf *mdev)
1579{
1580 if (mdev->ov_last_oos_size) {
1581 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1582 (unsigned long long)mdev->ov_last_oos_start,
1583 (unsigned long)mdev->ov_last_oos_size);
1584 }
1585 mdev->ov_last_oos_size=0;
1586}
1587
1588
1589extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1590extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
1591/* worker callbacks */
1592extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1593extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1594extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1595extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1596extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1597extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1598extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1599extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
1600extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int);
1601extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
1602extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
1603extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1604extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1605extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1606extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1607extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
1608extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
1609extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
1610extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
1611
1612extern void resync_timer_fn(unsigned long data);
1613extern void start_resync_timer_fn(unsigned long data);
1614
1615/* drbd_receiver.c */
1616extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
1617extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1618 const unsigned rw, const int fault_type);
1619extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1620extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1621 u64 id,
1622 sector_t sector,
1623 unsigned int data_size,
1624 gfp_t gfp_mask) __must_hold(local);
1625extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1626 int is_net);
1627#define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1628#define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
1629extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1630 struct list_head *head);
1631extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1632 struct list_head *head);
1633extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1634extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1635extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1636extern void drbd_free_tl_hash(struct drbd_conf *mdev);
1637
1638/* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1639 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1640static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1641 char __user *optval, int optlen)
1642{
1643 int err;
1644 if (level == SOL_SOCKET)
1645 err = sock_setsockopt(sock, level, optname, optval, optlen);
1646 else
1647 err = sock->ops->setsockopt(sock, level, optname, optval,
1648 optlen);
1649 return err;
1650}
1651
1652static inline void drbd_tcp_cork(struct socket *sock)
1653{
1654 int __user val = 1;
1655 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1656 (char __user *)&val, sizeof(val));
1657}
1658
1659static inline void drbd_tcp_uncork(struct socket *sock)
1660{
1661 int __user val = 0;
1662 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1663 (char __user *)&val, sizeof(val));
1664}
1665
1666static inline void drbd_tcp_nodelay(struct socket *sock)
1667{
1668 int __user val = 1;
1669 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1670 (char __user *)&val, sizeof(val));
1671}
1672
1673static inline void drbd_tcp_quickack(struct socket *sock)
1674{
1675 int __user val = 2;
1676 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1677 (char __user *)&val, sizeof(val));
1678}
1679
1680void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1681
1682/* drbd_proc.c */
1683extern struct proc_dir_entry *drbd_proc;
1684extern const struct file_operations drbd_proc_fops;
1685extern const char *drbd_conn_str(enum drbd_conns s);
1686extern const char *drbd_role_str(enum drbd_role s);
1687
1688/* drbd_actlog.c */
1689extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1690extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1691extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1692extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1693extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1694extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1695extern int drbd_rs_del_all(struct drbd_conf *mdev);
1696extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1697 sector_t sector, int size);
1698extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
1699extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
1700extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1701 int size, const char *file, const unsigned int line);
1702#define drbd_set_in_sync(mdev, sector, size) \
1703 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
1704extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
1705 int size, const char *file, const unsigned int line);
1706#define drbd_set_out_of_sync(mdev, sector, size) \
1707 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1708extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
1709extern void drbd_al_shrink(struct drbd_conf *mdev);
1710
1711
1712/* drbd_nl.c */
1713
1714void drbd_nl_cleanup(void);
1715int __init drbd_nl_init(void);
1716void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1717void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1718void drbd_bcast_ee(struct drbd_conf *mdev,
1719 const char *reason, const int dgs,
1720 const char* seen_hash, const char* calc_hash,
1721 const struct drbd_epoch_entry* e);
1722
1723
1724/**
1725 * DOC: DRBD State macros
1726 *
1727 * These macros are used to express state changes in easily readable form.
1728 *
1729 * The NS macros expand to a mask and a value, that can be bit ored onto the
1730 * current state as soon as the spinlock (req_lock) was taken.
1731 *
1732 * The _NS macros are used for state functions that get called with the
1733 * spinlock. These macros expand directly to the new state value.
1734 *
1735 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1736 * to express state changes that affect more than one aspect of the state.
1737 *
1738 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1739 * Means that the network connection was established and that the peer
1740 * is in secondary role.
1741 */
1742#define role_MASK R_MASK
1743#define peer_MASK R_MASK
1744#define disk_MASK D_MASK
1745#define pdsk_MASK D_MASK
1746#define conn_MASK C_MASK
1747#define susp_MASK 1
1748#define user_isp_MASK 1
1749#define aftr_isp_MASK 1
1750#define susp_nod_MASK 1
1751#define susp_fen_MASK 1
1752
1753#define NS(T, S) \
1754 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1755 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1756#define NS2(T1, S1, T2, S2) \
1757 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1758 mask.T2 = T2##_MASK; mask; }), \
1759 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1760 val.T2 = (S2); val; })
1761#define NS3(T1, S1, T2, S2, T3, S3) \
1762 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1763 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1764 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1765 val.T2 = (S2); val.T3 = (S3); val; })
1766
1767#define _NS(D, T, S) \
1768 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1769#define _NS2(D, T1, S1, T2, S2) \
1770 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1771 __ns.T2 = (S2); __ns; })
1772#define _NS3(D, T1, S1, T2, S2, T3, S3) \
1773 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1774 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1775
1776/*
1777 * inline helper functions
1778 *************************/
1779
1780/* see also page_chain_add and friends in drbd_receiver.c */
1781static inline struct page *page_chain_next(struct page *page)
1782{
1783 return (struct page *)page_private(page);
1784}
1785#define page_chain_for_each(page) \
1786 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1787 page = page_chain_next(page))
1788#define page_chain_for_each_safe(page, n) \
1789 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1790
1791static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1792{
1793 struct page *page = e->pages;
1794 page_chain_for_each(page) {
1795 if (page_count(page) > 1)
1796 return 1;
1797 }
1798 return 0;
1799}
1800
1801static inline void drbd_state_lock(struct drbd_conf *mdev)
1802{
1803 wait_event(mdev->misc_wait,
1804 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1805}
1806
1807static inline void drbd_state_unlock(struct drbd_conf *mdev)
1808{
1809 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1810 wake_up(&mdev->misc_wait);
1811}
1812
1813static inline enum drbd_state_rv
1814_drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1815 enum chg_state_flags flags, struct completion *done)
1816{
1817 enum drbd_state_rv rv;
1818
1819 read_lock(&global_state_lock);
1820 rv = __drbd_set_state(mdev, ns, flags, done);
1821 read_unlock(&global_state_lock);
1822
1823 return rv;
1824}
1825
1826/**
1827 * drbd_request_state() - Reqest a state change
1828 * @mdev: DRBD device.
1829 * @mask: mask of state bits to change.
1830 * @val: value of new state bits.
1831 *
1832 * This is the most graceful way of requesting a state change. It is verbose
1833 * quite verbose in case the state change is not possible, and all those
1834 * state changes are globally serialized.
1835 */
1836static inline int drbd_request_state(struct drbd_conf *mdev,
1837 union drbd_state mask,
1838 union drbd_state val)
1839{
1840 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1841}
1842
1843#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1844static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1845{
1846 switch (mdev->ldev->dc.on_io_error) {
1847 case EP_PASS_ON:
1848 if (!forcedetach) {
1849 if (__ratelimit(&drbd_ratelimit_state))
1850 dev_err(DEV, "Local IO failed in %s.\n", where);
1851 if (mdev->state.disk > D_INCONSISTENT)
1852 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
1853 break;
1854 }
1855 /* NOTE fall through to detach case if forcedetach set */
1856 case EP_DETACH:
1857 case EP_CALL_HELPER:
1858 set_bit(WAS_IO_ERROR, &mdev->flags);
1859 if (mdev->state.disk > D_FAILED) {
1860 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
1861 dev_err(DEV,
1862 "Local IO failed in %s. Detaching...\n", where);
1863 }
1864 break;
1865 }
1866}
1867
1868/**
1869 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1870 * @mdev: DRBD device.
1871 * @error: Error code passed to the IO completion callback
1872 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1873 *
1874 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1875 */
1876#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1877static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1878 int error, int forcedetach, const char *where)
1879{
1880 if (error) {
1881 unsigned long flags;
1882 spin_lock_irqsave(&mdev->req_lock, flags);
1883 __drbd_chk_io_error_(mdev, forcedetach, where);
1884 spin_unlock_irqrestore(&mdev->req_lock, flags);
1885 }
1886}
1887
1888
1889/**
1890 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1891 * @bdev: Meta data block device.
1892 *
1893 * BTW, for internal meta data, this happens to be the maximum capacity
1894 * we could agree upon with our peer node.
1895 */
1896static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1897{
1898 switch (bdev->dc.meta_dev_idx) {
1899 case DRBD_MD_INDEX_INTERNAL:
1900 case DRBD_MD_INDEX_FLEX_INT:
1901 return bdev->md.md_offset + bdev->md.bm_offset;
1902 case DRBD_MD_INDEX_FLEX_EXT:
1903 default:
1904 return bdev->md.md_offset;
1905 }
1906}
1907
1908/**
1909 * drbd_md_last_sector() - Return the last sector number of the meta data area
1910 * @bdev: Meta data block device.
1911 */
1912static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1913{
1914 switch (bdev->dc.meta_dev_idx) {
1915 case DRBD_MD_INDEX_INTERNAL:
1916 case DRBD_MD_INDEX_FLEX_INT:
1917 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1918 case DRBD_MD_INDEX_FLEX_EXT:
1919 default:
1920 return bdev->md.md_offset + bdev->md.md_size_sect;
1921 }
1922}
1923
1924/* Returns the number of 512 byte sectors of the device */
1925static inline sector_t drbd_get_capacity(struct block_device *bdev)
1926{
1927 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
1928 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
1929}
1930
1931/**
1932 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1933 * @bdev: Meta data block device.
1934 *
1935 * returns the capacity we announce to out peer. we clip ourselves at the
1936 * various MAX_SECTORS, because if we don't, current implementation will
1937 * oops sooner or later
1938 */
1939static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1940{
1941 sector_t s;
1942 switch (bdev->dc.meta_dev_idx) {
1943 case DRBD_MD_INDEX_INTERNAL:
1944 case DRBD_MD_INDEX_FLEX_INT:
1945 s = drbd_get_capacity(bdev->backing_bdev)
1946 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1947 drbd_md_first_sector(bdev))
1948 : 0;
1949 break;
1950 case DRBD_MD_INDEX_FLEX_EXT:
1951 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1952 drbd_get_capacity(bdev->backing_bdev));
1953 /* clip at maximum size the meta device can support */
1954 s = min_t(sector_t, s,
1955 BM_EXT_TO_SECT(bdev->md.md_size_sect
1956 - bdev->md.bm_offset));
1957 break;
1958 default:
1959 s = min_t(sector_t, DRBD_MAX_SECTORS,
1960 drbd_get_capacity(bdev->backing_bdev));
1961 }
1962 return s;
1963}
1964
1965/**
1966 * drbd_md_ss__() - Return the sector number of our meta data super block
1967 * @mdev: DRBD device.
1968 * @bdev: Meta data block device.
1969 */
1970static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1971 struct drbd_backing_dev *bdev)
1972{
1973 switch (bdev->dc.meta_dev_idx) {
1974 default: /* external, some index */
1975 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1976 case DRBD_MD_INDEX_INTERNAL:
1977 /* with drbd08, internal meta data is always "flexible" */
1978 case DRBD_MD_INDEX_FLEX_INT:
1979 /* sizeof(struct md_on_disk_07) == 4k
1980 * position: last 4k aligned block of 4k size */
1981 if (!bdev->backing_bdev) {
1982 if (__ratelimit(&drbd_ratelimit_state)) {
1983 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1984 dump_stack();
1985 }
1986 return 0;
1987 }
1988 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1989 - MD_AL_OFFSET;
1990 case DRBD_MD_INDEX_FLEX_EXT:
1991 return 0;
1992 }
1993}
1994
1995static inline void
1996drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1997{
1998 unsigned long flags;
1999 spin_lock_irqsave(&q->q_lock, flags);
2000 list_add(&w->list, &q->q);
2001 up(&q->s); /* within the spinlock,
2002 see comment near end of drbd_worker() */
2003 spin_unlock_irqrestore(&q->q_lock, flags);
2004}
2005
2006static inline void
2007drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
2008{
2009 unsigned long flags;
2010 spin_lock_irqsave(&q->q_lock, flags);
2011 list_add_tail(&w->list, &q->q);
2012 up(&q->s); /* within the spinlock,
2013 see comment near end of drbd_worker() */
2014 spin_unlock_irqrestore(&q->q_lock, flags);
2015}
2016
2017static inline void wake_asender(struct drbd_conf *mdev)
2018{
2019 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
2020 force_sig(DRBD_SIG, mdev->asender.task);
2021}
2022
2023static inline void request_ping(struct drbd_conf *mdev)
2024{
2025 set_bit(SEND_PING, &mdev->flags);
2026 wake_asender(mdev);
2027}
2028
2029static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
2030 enum drbd_packets cmd)
2031{
2032 struct p_header80 h;
2033 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
2034}
2035
2036static inline int drbd_send_ping(struct drbd_conf *mdev)
2037{
2038 struct p_header80 h;
2039 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
2040}
2041
2042static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
2043{
2044 struct p_header80 h;
2045 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
2046}
2047
2048static inline void drbd_thread_stop(struct drbd_thread *thi)
2049{
2050 _drbd_thread_stop(thi, false, true);
2051}
2052
2053static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2054{
2055 _drbd_thread_stop(thi, false, false);
2056}
2057
2058static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2059{
2060 _drbd_thread_stop(thi, true, false);
2061}
2062
2063/* counts how many answer packets packets we expect from our peer,
2064 * for either explicit application requests,
2065 * or implicit barrier packets as necessary.
2066 * increased:
2067 * w_send_barrier
2068 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2069 * it is much easier and equally valid to count what we queue for the
2070 * worker, even before it actually was queued or send.
2071 * (drbd_make_request_common; recovery path on read io-error)
2072 * decreased:
2073 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2074 * _req_mod(req, data_received)
2075 * [from receive_DataReply]
2076 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2077 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2078 * for some reason it is NOT decreased in got_NegAck,
2079 * but in the resulting cleanup code from report_params.
2080 * we should try to remember the reason for that...
2081 * _req_mod(req, send_failed or send_canceled)
2082 * _req_mod(req, connection_lost_while_pending)
2083 * [from tl_clear_barrier]
2084 */
2085static inline void inc_ap_pending(struct drbd_conf *mdev)
2086{
2087 atomic_inc(&mdev->ap_pending_cnt);
2088}
2089
2090#define ERR_IF_CNT_IS_NEGATIVE(which) \
2091 if (atomic_read(&mdev->which) < 0) \
2092 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2093 __func__ , __LINE__ , \
2094 atomic_read(&mdev->which))
2095
2096#define dec_ap_pending(mdev) do { \
2097 typecheck(struct drbd_conf *, mdev); \
2098 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2099 wake_up(&mdev->misc_wait); \
2100 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2101
2102/* counts how many resync-related answers we still expect from the peer
2103 * increase decrease
2104 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
2105 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
2106 * (or P_NEG_ACK with ID_SYNCER)
2107 */
2108static inline void inc_rs_pending(struct drbd_conf *mdev)
2109{
2110 atomic_inc(&mdev->rs_pending_cnt);
2111}
2112
2113#define dec_rs_pending(mdev) do { \
2114 typecheck(struct drbd_conf *, mdev); \
2115 atomic_dec(&mdev->rs_pending_cnt); \
2116 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2117
2118/* counts how many answers we still need to send to the peer.
2119 * increased on
2120 * receive_Data unless protocol A;
2121 * we need to send a P_RECV_ACK (proto B)
2122 * or P_WRITE_ACK (proto C)
2123 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2124 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2125 * receive_Barrier_* we need to send a P_BARRIER_ACK
2126 */
2127static inline void inc_unacked(struct drbd_conf *mdev)
2128{
2129 atomic_inc(&mdev->unacked_cnt);
2130}
2131
2132#define dec_unacked(mdev) do { \
2133 typecheck(struct drbd_conf *, mdev); \
2134 atomic_dec(&mdev->unacked_cnt); \
2135 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2136
2137#define sub_unacked(mdev, n) do { \
2138 typecheck(struct drbd_conf *, mdev); \
2139 atomic_sub(n, &mdev->unacked_cnt); \
2140 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2141
2142
2143static inline void put_net_conf(struct drbd_conf *mdev)
2144{
2145 if (atomic_dec_and_test(&mdev->net_cnt))
2146 wake_up(&mdev->net_cnt_wait);
2147}
2148
2149/**
2150 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2151 * @mdev: DRBD device.
2152 *
2153 * You have to call put_net_conf() when finished working with mdev->net_conf.
2154 */
2155static inline int get_net_conf(struct drbd_conf *mdev)
2156{
2157 int have_net_conf;
2158
2159 atomic_inc(&mdev->net_cnt);
2160 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2161 if (!have_net_conf)
2162 put_net_conf(mdev);
2163 return have_net_conf;
2164}
2165
2166/**
2167 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2168 * @M: DRBD device.
2169 *
2170 * You have to call put_ldev() when finished working with mdev->ldev.
2171 */
2172#define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2173#define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2174
2175static inline void put_ldev(struct drbd_conf *mdev)
2176{
2177 int i = atomic_dec_return(&mdev->local_cnt);
2178
2179 /* This may be called from some endio handler,
2180 * so we must not sleep here. */
2181
2182 __release(local);
2183 D_ASSERT(i >= 0);
2184 if (i == 0) {
2185 if (mdev->state.disk == D_DISKLESS)
2186 /* even internal references gone, safe to destroy */
2187 drbd_ldev_destroy(mdev);
2188 if (mdev->state.disk == D_FAILED)
2189 /* all application IO references gone. */
2190 drbd_go_diskless(mdev);
2191 wake_up(&mdev->misc_wait);
2192 }
2193}
2194
2195#ifndef __CHECKER__
2196static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2197{
2198 int io_allowed;
2199
2200 /* never get a reference while D_DISKLESS */
2201 if (mdev->state.disk == D_DISKLESS)
2202 return 0;
2203
2204 atomic_inc(&mdev->local_cnt);
2205 io_allowed = (mdev->state.disk >= mins);
2206 if (!io_allowed)
2207 put_ldev(mdev);
2208 return io_allowed;
2209}
2210#else
2211extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2212#endif
2213
2214/* you must have an "get_ldev" reference */
2215static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2216 unsigned long *bits_left, unsigned int *per_mil_done)
2217{
2218 /* this is to break it at compile time when we change that, in case we
2219 * want to support more than (1<<32) bits on a 32bit arch. */
2220 typecheck(unsigned long, mdev->rs_total);
2221
2222 /* note: both rs_total and rs_left are in bits, i.e. in
2223 * units of BM_BLOCK_SIZE.
2224 * for the percentage, we don't care. */
2225
2226 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2227 *bits_left = mdev->ov_left;
2228 else
2229 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
2230 /* >> 10 to prevent overflow,
2231 * +1 to prevent division by zero */
2232 if (*bits_left > mdev->rs_total) {
2233 /* doh. maybe a logic bug somewhere.
2234 * may also be just a race condition
2235 * between this and a disconnect during sync.
2236 * for now, just prevent in-kernel buffer overflow.
2237 */
2238 smp_rmb();
2239 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2240 drbd_conn_str(mdev->state.conn),
2241 *bits_left, mdev->rs_total, mdev->rs_failed);
2242 *per_mil_done = 0;
2243 } else {
2244 /* Make sure the division happens in long context.
2245 * We allow up to one petabyte storage right now,
2246 * at a granularity of 4k per bit that is 2**38 bits.
2247 * After shift right and multiplication by 1000,
2248 * this should still fit easily into a 32bit long,
2249 * so we don't need a 64bit division on 32bit arch.
2250 * Note: currently we don't support such large bitmaps on 32bit
2251 * arch anyways, but no harm done to be prepared for it here.
2252 */
2253 unsigned int shift = mdev->rs_total > UINT_MAX ? 16 : 10;
2254 unsigned long left = *bits_left >> shift;
2255 unsigned long total = 1UL + (mdev->rs_total >> shift);
2256 unsigned long tmp = 1000UL - left * 1000UL/total;
2257 *per_mil_done = tmp;
2258 }
2259}
2260
2261
2262/* this throttles on-the-fly application requests
2263 * according to max_buffers settings;
2264 * maybe re-implement using semaphores? */
2265static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2266{
2267 int mxb = 1000000; /* arbitrary limit on open requests */
2268 if (get_net_conf(mdev)) {
2269 mxb = mdev->net_conf->max_buffers;
2270 put_net_conf(mdev);
2271 }
2272 return mxb;
2273}
2274
2275static inline int drbd_state_is_stable(struct drbd_conf *mdev)
2276{
2277 union drbd_state s = mdev->state;
2278
2279 /* DO NOT add a default clause, we want the compiler to warn us
2280 * for any newly introduced state we may have forgotten to add here */
2281
2282 switch ((enum drbd_conns)s.conn) {
2283 /* new io only accepted when there is no connection, ... */
2284 case C_STANDALONE:
2285 case C_WF_CONNECTION:
2286 /* ... or there is a well established connection. */
2287 case C_CONNECTED:
2288 case C_SYNC_SOURCE:
2289 case C_SYNC_TARGET:
2290 case C_VERIFY_S:
2291 case C_VERIFY_T:
2292 case C_PAUSED_SYNC_S:
2293 case C_PAUSED_SYNC_T:
2294 case C_AHEAD:
2295 case C_BEHIND:
2296 /* transitional states, IO allowed */
2297 case C_DISCONNECTING:
2298 case C_UNCONNECTED:
2299 case C_TIMEOUT:
2300 case C_BROKEN_PIPE:
2301 case C_NETWORK_FAILURE:
2302 case C_PROTOCOL_ERROR:
2303 case C_TEAR_DOWN:
2304 case C_WF_REPORT_PARAMS:
2305 case C_STARTING_SYNC_S:
2306 case C_STARTING_SYNC_T:
2307 break;
2308
2309 /* Allow IO in BM exchange states with new protocols */
2310 case C_WF_BITMAP_S:
2311 if (mdev->agreed_pro_version < 96)
2312 return 0;
2313 break;
2314
2315 /* no new io accepted in these states */
2316 case C_WF_BITMAP_T:
2317 case C_WF_SYNC_UUID:
2318 case C_MASK:
2319 /* not "stable" */
2320 return 0;
2321 }
2322
2323 switch ((enum drbd_disk_state)s.disk) {
2324 case D_DISKLESS:
2325 case D_INCONSISTENT:
2326 case D_OUTDATED:
2327 case D_CONSISTENT:
2328 case D_UP_TO_DATE:
2329 case D_FAILED:
2330 /* disk state is stable as well. */
2331 break;
2332
2333 /* no new io accepted during tansitional states */
2334 case D_ATTACHING:
2335 case D_NEGOTIATING:
2336 case D_UNKNOWN:
2337 case D_MASK:
2338 /* not "stable" */
2339 return 0;
2340 }
2341
2342 return 1;
2343}
2344
2345static inline int is_susp(union drbd_state s)
2346{
2347 return s.susp || s.susp_nod || s.susp_fen;
2348}
2349
2350static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
2351{
2352 int mxb = drbd_get_max_buffers(mdev);
2353
2354 if (is_susp(mdev->state))
2355 return false;
2356 if (test_bit(SUSPEND_IO, &mdev->flags))
2357 return false;
2358
2359 /* to avoid potential deadlock or bitmap corruption,
2360 * in various places, we only allow new application io
2361 * to start during "stable" states. */
2362
2363 /* no new io accepted when attaching or detaching the disk */
2364 if (!drbd_state_is_stable(mdev))
2365 return false;
2366
2367 /* since some older kernels don't have atomic_add_unless,
2368 * and we are within the spinlock anyways, we have this workaround. */
2369 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
2370 return false;
2371 if (test_bit(BITMAP_IO, &mdev->flags))
2372 return false;
2373 return true;
2374}
2375
2376static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
2377{
2378 bool rv = false;
2379
2380 spin_lock_irq(&mdev->req_lock);
2381 rv = may_inc_ap_bio(mdev);
2382 if (rv)
2383 atomic_add(count, &mdev->ap_bio_cnt);
2384 spin_unlock_irq(&mdev->req_lock);
2385
2386 return rv;
2387}
2388
2389static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2390{
2391 /* we wait here
2392 * as long as the device is suspended
2393 * until the bitmap is no longer on the fly during connection
2394 * handshake as long as we would exeed the max_buffer limit.
2395 *
2396 * to avoid races with the reconnect code,
2397 * we need to atomic_inc within the spinlock. */
2398
2399 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
2400}
2401
2402static inline void dec_ap_bio(struct drbd_conf *mdev)
2403{
2404 int mxb = drbd_get_max_buffers(mdev);
2405 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2406
2407 D_ASSERT(ap_bio >= 0);
2408 /* this currently does wake_up for every dec_ap_bio!
2409 * maybe rather introduce some type of hysteresis?
2410 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2411 if (ap_bio < mxb)
2412 wake_up(&mdev->misc_wait);
2413 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2414 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
2415 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2416 }
2417}
2418
2419static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
2420{
2421 int changed = mdev->ed_uuid != val;
2422 mdev->ed_uuid = val;
2423 return changed;
2424}
2425
2426static inline int seq_cmp(u32 a, u32 b)
2427{
2428 /* we assume wrap around at 32bit.
2429 * for wrap around at 24bit (old atomic_t),
2430 * we'd have to
2431 * a <<= 8; b <<= 8;
2432 */
2433 return (s32)(a) - (s32)(b);
2434}
2435#define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2436#define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2437#define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2438#define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2439/* CAUTION: please no side effects in arguments! */
2440#define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2441
2442static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2443{
2444 unsigned int m;
2445 spin_lock(&mdev->peer_seq_lock);
2446 m = seq_max(mdev->peer_seq, new_seq);
2447 mdev->peer_seq = m;
2448 spin_unlock(&mdev->peer_seq_lock);
2449 if (m == new_seq)
2450 wake_up(&mdev->seq_wait);
2451}
2452
2453static inline void drbd_update_congested(struct drbd_conf *mdev)
2454{
2455 struct sock *sk = mdev->data.socket->sk;
2456 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2457 set_bit(NET_CONGESTED, &mdev->flags);
2458}
2459
2460static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2461{
2462 /* sorry, we currently have no working implementation
2463 * of distributed TCQ stuff */
2464#ifndef QUEUE_ORDERED_NONE
2465#define QUEUE_ORDERED_NONE 0
2466#endif
2467 return QUEUE_ORDERED_NONE;
2468}
2469
2470static inline void drbd_md_flush(struct drbd_conf *mdev)
2471{
2472 int r;
2473
2474 if (test_bit(MD_NO_FUA, &mdev->flags))
2475 return;
2476
2477 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
2478 if (r) {
2479 set_bit(MD_NO_FUA, &mdev->flags);
2480 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
2481 }
2482}
2483
2484#endif