projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drbd: log request sector offset and size for IO errors
[linux-2.6-block.git] / drivers / block / drbd / drbd_int.h
CommitLineData
b411b363
PR		 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>
b411b363
PR		 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>
132cc538 36#include <linux/ratelimit.h>
b411b363
PR		 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>
70c71606 44#include <linux/prefetch.h>
b411b363
PR		 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;
90ab5ee9
RR		 62extern bool disable_sendpage;
63extern bool allow_oos;
b411b363
PR		 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
b411b363
PR		 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)
3ad2f3fb 91 * so if it says "cannot dereference null pointer at address 0x00000001",
b411b363
PR		 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)
4a23f264 100#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
b411b363
PR		 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
3beec1d4
LE		 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; \
b411b363
PR		 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 */
6b4388ac 129 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
b411b363
PR		 130
131 DRBD_FAULT_MAX,
132};
133
b411b363
PR		 134extern unsigned int
135_drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
0cf9d27e 136
b411b363
PR		 137static inline int
138drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
0cf9d27e 139#ifdef CONFIG_DRBD_FAULT_INJECTION
b411b363
PR		 140 return fault_rate &&
141 (enable_faults & (1<<type)) &&
142 _drbd_insert_fault(mdev, type);
b411b363 143#else
0cf9d27e 144 return 0;
b411b363 145#endif
0cf9d27e 146}
b411b363
PR		 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 */
0ced55a3
PR		 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 */
73a01a18 208 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
d612d309 209 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
b411b363 210
d612d309 211 P_MAX_CMD = 0x2A,
b411b363
PR		 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",
c42b6cf4
LE		 263 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
264 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
265 [P_COMPRESSED_BITMAP] = "CBitmap",
a8cdfd8d 266 [P_DELAY_PROBE] = "DelayProbe",
73a01a18 267 [P_OUT_OF_SYNC] = "OutOfSync",
b411b363
PR		 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 */
0b70a13d 336struct p_header80 {
b411b363
PR		 337 u32 magic;
338 u16 command;
339 u16 length; /* bytes of data after this header */
340 u8 payload[0];
341} __packed;
0b70a13d
PR		 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;
00b42537 347 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
0b70a13d
PR		 348 u8 payload[0];
349} __packed;
350
351union p_header {
352 struct p_header80 h80;
353 struct p_header95 h95;
354};
b411b363
PR		 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 */
76d2e7ec
PR		 373#define DP_HARDBARRIER 1 /* depricated */
374#define DP_RW_SYNC 2 /* equals REQ_SYNC */
b411b363 375#define DP_MAY_SET_IN_SYNC 4
721a9602 376#define DP_UNPLUG 8 /* not used anymore */
76d2e7ec
PR		 377#define DP_FUA 16 /* equals REQ_FUA */
378#define DP_FLUSH 32 /* equals REQ_FLUSH */
379#define DP_DISCARD 64 /* equals REQ_DISCARD */
b411b363
PR		 380
381struct p_data {
0b70a13d 382 union p_header head;
b411b363
PR		 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 {
0b70a13d 398 struct p_header80 head;
b411b363
PR		 399 u64 sector;
400 u64 block_id;
401 u32 blksize;
402 u32 seq_num;
403} __packed;
404
405
406struct p_block_req {
0b70a13d 407 struct p_header80 head;
b411b363
PR		 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 {
0b70a13d 424 struct p_header80 head; /* 8 bytes */
b411b363
PR		 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 {
0b70a13d 439 struct p_header80 head;
b411b363
PR		 440 u32 barrier; /* barrier number _handle_ only */
441 u32 pad; /* to multiple of 8 Byte */
442} __packed;
443
444struct p_barrier_ack {
0b70a13d 445 struct p_header80 head;
b411b363
PR		 446 u32 barrier;
447 u32 set_size;
448} __packed;
449
450struct p_rs_param {
0b70a13d 451 struct p_header80 head;
b411b363
PR		 452 u32 rate;
453
454 /* Since protocol version 88 and higher. */
455 char verify_alg[0];
456} __packed;
457
458struct p_rs_param_89 {
0b70a13d 459 struct p_header80 head;
b411b363
PR		 460 u32 rate;
461 /* protocol version 89: */
462 char verify_alg[SHARED_SECRET_MAX];
463 char csums_alg[SHARED_SECRET_MAX];
464} __packed;
465
8e26f9cc 466struct p_rs_param_95 {
0b70a13d 467 struct p_header80 head;
8e26f9cc
PR		 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
cf14c2e9
PR		 477enum drbd_conn_flags {
478 CF_WANT_LOSE = 1,
479 CF_DRY_RUN = 2,
480};
481
b411b363 482struct p_protocol {
0b70a13d 483 struct p_header80 head;
b411b363
PR		 484 u32 protocol;
485 u32 after_sb_0p;
486 u32 after_sb_1p;
487 u32 after_sb_2p;
cf14c2e9 488 u32 conn_flags;
b411b363
PR		 489 u32 two_primaries;
490
491 /* Since protocol version 87 and higher. */
492 char integrity_alg[0];
493
494} __packed;
495
496struct p_uuids {
0b70a13d 497 struct p_header80 head;
b411b363
PR		 498 u64 uuid[UI_EXTENDED_SIZE];
499} __packed;
500
501struct p_rs_uuid {
0b70a13d 502 struct p_header80 head;
b411b363
PR		 503 u64 uuid;
504} __packed;
505
506struct p_sizes {
0b70a13d 507 struct p_header80 head;
b411b363
PR		 508 u64 d_size; /* size of disk */
509 u64 u_size; /* user requested size */
510 u64 c_size; /* current exported size */
1816a2b4 511 u32 max_bio_size; /* Maximal size of a BIO */
e89b591c
PR		 512 u16 queue_order_type; /* not yet implemented in DRBD*/
513 u16 dds_flags; /* use enum dds_flags here. */
b411b363
PR		 514} __packed;
515
516struct p_state {
0b70a13d 517 struct p_header80 head;
b411b363
PR		 518 u32 state;
519} __packed;
520
521struct p_req_state {
0b70a13d 522 struct p_header80 head;
b411b363
PR		 523 u32 mask;
524 u32 val;
525} __packed;
526
527struct p_req_state_reply {
0b70a13d 528 struct p_header80 head;
b411b363
PR		 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 {
0b70a13d 543 struct p_header80 head;
b411b363
PR		 544 u64 block_id;
545 u32 seq_num;
546 u32 pad;
547} __packed;
548
73a01a18
PR		 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
b411b363
PR		 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 {
0b70a13d 566 struct p_header80 head;
b411b363
PR		 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
0b70a13d
PR		 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 */
0ced55a3
PR		 581} __packed;
582
b411b363
PR		 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
25985edc 625 * most architectures have used for a long time.
b411b363 626 */
0b70a13d 627#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
b411b363
PR		 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 {
02918be2 636 union p_header header;
b411b363
PR		 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;
8e26f9cc 643 struct p_rs_param_95 rs_param_95;
b411b363
PR		 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;
02918be2
PR		 651 struct p_delay_probe93 delay_probe93;
652 struct p_rs_uuid rs_uuid;
73a01a18 653 struct p_block_desc block_desc;
b411b363
PR		 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
b411b363
PR		 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
24c4830c 702 struct hlist_node collision;
b411b363
PR		 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
b411b363
PR		 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() */
b411b363
PR		 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. */
7e602c0a 723 int n_writes; /* number of requests attached before this barrier */
b411b363
PR		 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 {
b411b363 745 DE_HAVE_BARRIER_NUMBER,
b411b363
PR		 746};
747
748enum epoch_event {
749 EV_PUT,
750 EV_GOT_BARRIER_NR,
b411b363 751 EV_BECAME_LAST,
b411b363
PR		 752 EV_CLEANUP = 32, /* used as flag */
753};
754
b411b363
PR		 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
45bb912b
LE		 765struct drbd_epoch_entry {
766 struct drbd_work w;
24c4830c 767 struct hlist_node collision;
85719573 768 struct drbd_epoch *epoch; /* for writes */
45bb912b
LE		 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;
85719573
PR		 776 union {
777 u64 block_id;
778 struct digest_info *digest;
779 };
45bb912b
LE		 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 */
b411b363
PR		 788enum {
789 __EE_CALL_AL_COMPLETE_IO,
b411b363 790 __EE_MAY_SET_IN_SYNC,
45bb912b 791
45bb912b
LE		 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,
c36c3ced
LE		 800
801 /* This ee has a pointer to a digest instead of a block id */
802 __EE_HAS_DIGEST,
b411b363
PR		 803};
804#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
b411b363 805#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
45bb912b
LE		 806#define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
807#define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
c36c3ced 808#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
b411b363
PR		 809
810/* global flag bits */
06f10adb 811enum drbd_flag {
25985edc 812 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
b411b363
PR		 813 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
814 SEND_PING, /* whether asender should send a ping asap */
815
b411b363
PR		 816 UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */
817 MD_DIRTY, /* current uuids and flags not yet on disk */
818 DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */
819 USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */
820 CLUSTER_ST_CHANGE, /* Cluster wide state change going on... */
821 CL_ST_CHG_SUCCESS,
822 CL_ST_CHG_FAIL,
823 CRASHED_PRIMARY, /* This node was a crashed primary.
824 * Gets cleared when the state.conn
825 * goes into C_CONNECTED state. */
b411b363
PR		 826 CONSIDER_RESYNC,
827
a8a4e51e 828 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
b411b363
PR		 829 SUSPEND_IO, /* suspend application io */
830 BITMAP_IO, /* suspend application io;
831 once no more io in flight, start bitmap io */
832 BITMAP_IO_QUEUED, /* Started bitmap IO */
82f59cc6 833 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
a2a3c74f
LE		 834 WAS_IO_ERROR, /* Local disk failed, returned IO error */
835 WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */
383606e0 836 FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */
b411b363
PR		 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. */
cf14c2e9 848 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
309d1608 849 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
43a5182c 850 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
0778286a 851 AL_SUSPENDED, /* Activity logging is currently suspended. */
370a43e7 852 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
197296ff 853 STATE_SENT, /* Do not change state/UUIDs while this is set */
c2ba686f
LE		 854
855 CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC)
856 * pending, from drbd worker context.
857 * If set, bdi_write_congested() returns true,
858 * so shrink_page_list() would not recurse into,
859 * and potentially deadlock on, this drbd worker.
860 */
599377ac 861 DISCONNECT_SENT, /* Currently the last bit in this 32bit word */
06f10adb
LE		 862
863 /* keep last */
864 DRBD_N_FLAGS,
b411b363
PR		 865};
866
867struct drbd_bitmap; /* opaque for drbd_conf */
868
20ceb2b2
LE		 869/* definition of bits in bm_flags to be used in drbd_bm_lock
870 * and drbd_bitmap_io and friends. */
871enum bm_flag {
872 /* do we need to kfree, or vfree bm_pages? */
873 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
874
875 /* currently locked for bulk operation */
0e8488ad 876 BM_LOCKED_MASK = 0xf,
20ceb2b2
LE		 877
878 /* in detail, that is: */
879 BM_DONT_CLEAR = 0x1,
880 BM_DONT_SET = 0x2,
881 BM_DONT_TEST = 0x4,
882
0e8488ad
LE		 883 /* so we can mark it locked for bulk operation,
884 * and still allow all non-bulk operations */
885 BM_IS_LOCKED = 0x8,
886
20ceb2b2 887 /* (test bit, count bit) allowed (common case) */
0e8488ad 888 BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED,
20ceb2b2
LE		 889
890 /* testing bits, as well as setting new bits allowed, but clearing bits
891 * would be unexpected. Used during bitmap receive. Setting new bits
892 * requires sending of "out-of-sync" information, though. */
0e8488ad 893 BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED,
20ceb2b2 894
0e8488ad
LE		 895 /* for drbd_bm_write_copy_pages, everything is allowed,
896 * only concurrent bulk operations are locked out. */
897 BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED,
20ceb2b2
LE		 898};
899
b411b363
PR		 900struct drbd_work_queue {
901 struct list_head q;
902 struct semaphore s; /* producers up it, worker down()s it */
903 spinlock_t q_lock; /* to protect the list. */
904};
905
906struct drbd_socket {
907 struct drbd_work_queue work;
908 struct mutex mutex;
909 struct socket *socket;
910 /* this way we get our
911 * send/receive buffers off the stack */
912 union p_polymorph sbuf;
913 union p_polymorph rbuf;
914};
915
916struct drbd_md {
917 u64 md_offset; /* sector offset to 'super' block */
918
919 u64 la_size_sect; /* last agreed size, unit sectors */
9f2247bb 920 spinlock_t uuid_lock;
b411b363
PR		 921 u64 uuid[UI_SIZE];
922 u64 device_uuid;
923 u32 flags;
924 u32 md_size_sect;
925
926 s32 al_offset; /* signed relative sector offset to al area */
927 s32 bm_offset; /* signed relative sector offset to bitmap */
928
929 /* u32 al_nr_extents; important for restoring the AL
930 * is stored into sync_conf.al_extents, which in turn
931 * gets applied to act_log->nr_elements
932 */
933};
934
935/* for sync_conf and other types... */
936#define NL_PACKET(name, number, fields) struct name { fields };
937#define NL_INTEGER(pn,pr,member) int member;
938#define NL_INT64(pn,pr,member) __u64 member;
939#define NL_BIT(pn,pr,member) unsigned member:1;
940#define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
1d273b92 941#include <linux/drbd_nl.h>
b411b363
PR		 942
943struct drbd_backing_dev {
944 struct block_device *backing_bdev;
945 struct block_device *md_bdev;
b411b363
PR		 946 struct drbd_md md;
947 struct disk_conf dc; /* The user provided config... */
948 sector_t known_size; /* last known size of that backing device */
949};
950
951struct drbd_md_io {
0c464425 952 unsigned int done;
b411b363
PR		 953 int error;
954};
955
956struct bm_io_work {
957 struct drbd_work w;
958 char *why;
20ceb2b2 959 enum bm_flag flags;
b411b363
PR		 960 int (*io_fn)(struct drbd_conf *mdev);
961 void (*done)(struct drbd_conf *mdev, int rv);
962};
963
964enum write_ordering_e {
965 WO_none,
966 WO_drain_io,
967 WO_bdev_flush,
b411b363
PR		 968};
969
778f271d
PR		 970struct fifo_buffer {
971 int *values;
972 unsigned int head_index;
973 unsigned int size;
974};
975
b411b363 976struct drbd_conf {
06f10adb 977 unsigned long drbd_flags[(DRBD_N_FLAGS + BITS_PER_LONG -1)/BITS_PER_LONG];
b411b363
PR		 978
979 /* configured by drbdsetup */
980 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
981 struct syncer_conf sync_conf;
982 struct drbd_backing_dev *ldev __protected_by(local);
983
984 sector_t p_size; /* partner's disk size */
985 struct request_queue *rq_queue;
986 struct block_device *this_bdev;
987 struct gendisk *vdisk;
988
989 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
990 struct drbd_socket meta; /* ping/ack (metadata) packets */
991 int agreed_pro_version; /* actually used protocol version */
992 unsigned long last_received; /* in jiffies, either socket */
993 unsigned int ko_count;
994 struct drbd_work resync_work,
995 unplug_work,
e9e6f3ec 996 go_diskless,
c4752ef1
PR		 997 md_sync_work,
998 start_resync_work;
b411b363
PR		 999 struct timer_list resync_timer;
1000 struct timer_list md_sync_timer;
370a43e7 1001 struct timer_list start_resync_timer;
7fde2be9 1002 struct timer_list request_timer;
ee15b038
LE		 1003#ifdef DRBD_DEBUG_MD_SYNC
1004 struct {
1005 unsigned int line;
1006 const char* func;
1007 } last_md_mark_dirty;
1008#endif
b411b363
PR		 1009
1010 /* Used after attach while negotiating new disk state. */
1011 union drbd_state new_state_tmp;
1012
1013 union drbd_state state;
1014 wait_queue_head_t misc_wait;
1015 wait_queue_head_t state_wait; /* upon each state change. */
84dfb9f5 1016 wait_queue_head_t net_cnt_wait;
b411b363
PR		 1017 unsigned int send_cnt;
1018 unsigned int recv_cnt;
1019 unsigned int read_cnt;
1020 unsigned int writ_cnt;
1021 unsigned int al_writ_cnt;
1022 unsigned int bm_writ_cnt;
1023 atomic_t ap_bio_cnt; /* Requests we need to complete */
1024 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1025 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1026 atomic_t unacked_cnt; /* Need to send replys for */
1027 atomic_t local_cnt; /* Waiting for local completion */
1028 atomic_t net_cnt; /* Users of net_conf */
1029 spinlock_t req_lock;
1030 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1031 struct drbd_tl_epoch *newest_tle;
1032 struct drbd_tl_epoch *oldest_tle;
1033 struct list_head out_of_sequence_requests;
6d7e32f5 1034 struct list_head barrier_acked_requests;
b411b363
PR		 1035 struct hlist_head *tl_hash;
1036 unsigned int tl_hash_s;
1037
4b0715f0 1038 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
b411b363 1039 unsigned long rs_total;
4b0715f0 1040 /* number of resync blocks that failed in this run */
b411b363
PR		 1041 unsigned long rs_failed;
1042 /* Syncer's start time [unit jiffies] */
1043 unsigned long rs_start;
1044 /* cumulated time in PausedSyncX state [unit jiffies] */
1045 unsigned long rs_paused;
1d7734a0
LE		 1046 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1047 unsigned long rs_same_csum;
1048#define DRBD_SYNC_MARKS 8
1049#define DRBD_SYNC_MARK_STEP (3*HZ)
b411b363 1050 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1d7734a0 1051 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
b411b363 1052 /* marks's time [unit jiffies] */
1d7734a0
LE		 1053 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1054 /* current index into rs_mark_{left,time} */
1055 int rs_last_mark;
b411b363
PR		 1056
1057 /* where does the admin want us to start? (sector) */
1058 sector_t ov_start_sector;
02b91b55 1059 sector_t ov_stop_sector;
b411b363
PR		 1060 /* where are we now? (sector) */
1061 sector_t ov_position;
1062 /* Start sector of out of sync range (to merge printk reporting). */
1063 sector_t ov_last_oos_start;
1064 /* size of out-of-sync range in sectors. */
1065 sector_t ov_last_oos_size;
1066 unsigned long ov_left; /* in bits */
1067 struct crypto_hash *csums_tfm;
1068 struct crypto_hash *verify_tfm;
1069
ba280c09
LE		 1070 unsigned long last_reattach_jif;
1071 unsigned long last_reconnect_jif;
b411b363
PR		 1072 struct drbd_thread receiver;
1073 struct drbd_thread worker;
1074 struct drbd_thread asender;
1075 struct drbd_bitmap *bitmap;
1076 unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */
1077
1078 /* Used to track operations of resync... */
1079 struct lru_cache *resync;
1080 /* Number of locked elements in resync LRU */
1081 unsigned int resync_locked;
1082 /* resync extent number waiting for application requests */
1083 unsigned int resync_wenr;
1084
1085 int open_cnt;
1086 u64 *p_uuid;
1087 struct drbd_epoch *current_epoch;
1088 spinlock_t epoch_lock;
1089 unsigned int epochs;
1090 enum write_ordering_e write_ordering;
85719573
PR		 1091 struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */
1092 struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */
b411b363 1093 struct list_head done_ee; /* send ack */
85719573 1094 struct list_head read_ee; /* IO in progress (any read) */
b411b363
PR		 1095 struct list_head net_ee; /* zero-copy network send in progress */
1096 struct hlist_head *ee_hash; /* is proteced by req_lock! */
1097 unsigned int ee_hash_s;
1098
1099 /* this one is protected by ee_lock, single thread */
1100 struct drbd_epoch_entry *last_write_w_barrier;
1101
1102 int next_barrier_nr;
1103 struct hlist_head *app_reads_hash; /* is proteced by req_lock */
1104 struct list_head resync_reads;
435f0740
LE		 1105 atomic_t pp_in_use; /* allocated from page pool */
1106 atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */
b411b363
PR		 1107 wait_queue_head_t ee_wait;
1108 struct page *md_io_page; /* one page buffer for md_io */
1109 struct page *md_io_tmpp; /* for logical_block_size != 512 */
cc94c650 1110 struct drbd_md_io md_io;
e1711731 1111 atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */
b411b363
PR		 1112 spinlock_t al_lock;
1113 wait_queue_head_t al_wait;
1114 struct lru_cache *act_log; /* activity log */
1115 unsigned int al_tr_number;
1116 int al_tr_cycle;
1117 int al_tr_pos; /* position of the next transaction in the journal */
1118 struct crypto_hash *cram_hmac_tfm;
1119 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1120 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1121 void *int_dig_out;
1122 void *int_dig_in;
1123 void *int_dig_vv;
1124 wait_queue_head_t seq_wait;
1125 atomic_t packet_seq;
1126 unsigned int peer_seq;
1127 spinlock_t peer_seq_lock;
1128 unsigned int minor;
1129 unsigned long comm_bm_set; /* communicated number of set bits. */
1130 cpumask_var_t cpu_mask;
1131 struct bm_io_work bm_io_work;
1132 u64 ed_uuid; /* UUID of the exposed data */
1133 struct mutex state_mutex;
1134 char congestion_reason; /* Why we where congested... */
1d7734a0
LE		 1135 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1136 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1137 int rs_last_sect_ev; /* counter to compare with */
1138 int rs_last_events; /* counter of read or write "events" (unit sectors)
1139 * on the lower level device when we last looked. */
1140 int c_sync_rate; /* current resync rate after syncer throttle magic */
778f271d
PR		 1141 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1142 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
25985edc 1143 int rs_planed; /* resync sectors already planned */
759fbdfb 1144 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
db141b2f
LE		 1145 unsigned int peer_max_bio_size;
1146 unsigned int local_max_bio_size;
b411b363
PR		 1147};
1148
06f10adb
LE		 1149static inline void drbd_set_flag(struct drbd_conf *mdev, enum drbd_flag f)
1150{
1151 set_bit(f, &mdev->drbd_flags[0]);
1152}
1153
1154static inline void drbd_clear_flag(struct drbd_conf *mdev, enum drbd_flag f)
1155{
1156 clear_bit(f, &mdev->drbd_flags[0]);
1157}
1158
1159static inline int drbd_test_flag(struct drbd_conf *mdev, enum drbd_flag f)
1160{
1161 return test_bit(f, &mdev->drbd_flags[0]);
1162}
1163
1164static inline int drbd_test_and_set_flag(struct drbd_conf *mdev, enum drbd_flag f)
1165{
1166 return test_and_set_bit(f, &mdev->drbd_flags[0]);
1167}
1168
1169static inline int drbd_test_and_clear_flag(struct drbd_conf *mdev, enum drbd_flag f)
1170{
1171 return test_and_clear_bit(f, &mdev->drbd_flags[0]);
1172}
1173
b411b363
PR		 1174static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1175{
1176 struct drbd_conf *mdev;
1177
1178 mdev = minor < minor_count ? minor_table[minor] : NULL;
1179
1180 return mdev;
1181}
1182
1183static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1184{
1185 return mdev->minor;
1186}
1187
25985edc 1188/* returns 1 if it was successful,
b411b363
PR		 1189 * returns 0 if there was no data socket.
1190 * so wherever you are going to use the data.socket, e.g. do
1191 * if (!drbd_get_data_sock(mdev))
1192 * return 0;
1193 * CODE();
1194 * drbd_put_data_sock(mdev);
1195 */
1196static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1197{
1198 mutex_lock(&mdev->data.mutex);
1199 /* drbd_disconnect() could have called drbd_free_sock()
1200 * while we were waiting in down()... */
1201 if (unlikely(mdev->data.socket == NULL)) {
1202 mutex_unlock(&mdev->data.mutex);
1203 return 0;
1204 }
1205 return 1;
1206}
1207
1208static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1209{
1210 mutex_unlock(&mdev->data.mutex);
1211}
1212
1213/*
1214 * function declarations
1215 *************************/
1216
1217/* drbd_main.c */
1218
1219enum chg_state_flags {
1220 CS_HARD = 1,
1221 CS_VERBOSE = 2,
1222 CS_WAIT_COMPLETE = 4,
1223 CS_SERIALIZE = 8,
1224 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1225};
1226
e89b591c
PR		 1227enum dds_flags {
1228 DDSF_FORCED = 1,
1229 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1230};
1231
b411b363 1232extern void drbd_init_set_defaults(struct drbd_conf *mdev);
bf885f8a
AG		 1233extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev,
1234 enum chg_state_flags f,
1235 union drbd_state mask,
1236 union drbd_state val);
b411b363
PR		 1237extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1238 union drbd_state);
bf885f8a
AG		 1239extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *,
1240 union drbd_state,
1241 union drbd_state,
1242 enum chg_state_flags);
1243extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state,
1244 enum chg_state_flags,
1245 struct completion *done);
b411b363
PR		 1246extern void print_st_err(struct drbd_conf *, union drbd_state,
1247 union drbd_state, int);
1248extern int drbd_thread_start(struct drbd_thread *thi);
1249extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1250#ifdef CONFIG_SMP
1251extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1252extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1253#else
1254#define drbd_thread_current_set_cpu(A) ({})
1255#define drbd_calc_cpu_mask(A) ({})
1256#endif
1257extern void drbd_free_resources(struct drbd_conf *mdev);
1258extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1259 unsigned int set_size);
1260extern void tl_clear(struct drbd_conf *mdev);
1261extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1262extern void drbd_free_sock(struct drbd_conf *mdev);
1263extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1264 void *buf, size_t size, unsigned msg_flags);
1265extern int drbd_send_protocol(struct drbd_conf *mdev);
1266extern int drbd_send_uuids(struct drbd_conf *mdev);
1267extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
5a22db89 1268extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
e89b591c 1269extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
f479ea06
LE		 1270extern int drbd_send_state(struct drbd_conf *mdev, union drbd_state s);
1271extern int drbd_send_current_state(struct drbd_conf *mdev);
b411b363 1272extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
0b70a13d 1273 enum drbd_packets cmd, struct p_header80 *h,
b411b363
PR		 1274 size_t size, unsigned msg_flags);
1275#define USE_DATA_SOCKET 1
1276#define USE_META_SOCKET 0
1277extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
0b70a13d 1278 enum drbd_packets cmd, struct p_header80 *h,
b411b363
PR		 1279 size_t size);
1280extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1281 char *data, size_t size);
1282extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1283extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1284 u32 set_size);
1285extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1286 struct drbd_epoch_entry *e);
1287extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1288 struct p_block_req *rp);
1289extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
2b2bf214 1290 struct p_data *dp, int data_size);
b411b363
PR		 1291extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1292 sector_t sector, int blksize, u64 block_id);
73a01a18 1293extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
b411b363
PR		 1294extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1295 struct drbd_epoch_entry *e);
1296extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
b411b363
PR		 1297extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1298 sector_t sector, int size, u64 block_id);
1299extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1300 sector_t sector,int size,
1301 void *digest, int digest_size,
1302 enum drbd_packets cmd);
1303extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1304
1305extern int drbd_send_bitmap(struct drbd_conf *mdev);
1306extern int _drbd_send_bitmap(struct drbd_conf *mdev);
bf885f8a 1307extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
b411b363
PR		 1308extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1309extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
62b0da3a 1310void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
b411b363 1311
b411b363
PR		 1312extern void drbd_md_sync(struct drbd_conf *mdev);
1313extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
b411b363
PR		 1314extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1315extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1316extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
b411b363 1317extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
9f2247bb
PR		 1318extern void drbd_uuid_move_history(struct drbd_conf *mdev) __must_hold(local);
1319extern void __drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
b411b363
PR		 1320extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1321extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1322extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
ee15b038 1323#ifndef DRBD_DEBUG_MD_SYNC
b411b363 1324extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
ee15b038
LE		 1325#else
1326#define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1327extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1328 unsigned int line, const char *func);
1329#endif
b411b363
PR		 1330extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1331 int (*io_fn)(struct drbd_conf *),
1332 void (*done)(struct drbd_conf *, int),
20ceb2b2
LE		 1333 char *why, enum bm_flag flags);
1334extern int drbd_bitmap_io(struct drbd_conf *mdev,
1335 int (*io_fn)(struct drbd_conf *),
1336 char *why, enum bm_flag flags);
b411b363
PR		 1337extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1338extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
e9e6f3ec 1339extern void drbd_go_diskless(struct drbd_conf *mdev);
82f59cc6 1340extern void drbd_ldev_destroy(struct drbd_conf *mdev);
b411b363
PR		 1341
1342
1343/* Meta data layout
1344 We reserve a 128MB Block (4k aligned)
1345 * either at the end of the backing device
3ad2f3fb 1346 * or on a separate meta data device. */
b411b363
PR		 1347
1348#define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1349/* The following numbers are sectors */
1350#define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1351#define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1352/* Allows up to about 3.8TB */
1353#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1354
1355/* Since the smalles IO unit is usually 512 byte */
1356#define MD_SECTOR_SHIFT 9
1357#define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1358
1359/* activity log */
1360#define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1361#define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1362#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1363
1364#if BITS_PER_LONG == 32
1365#define LN2_BPL 5
1366#define cpu_to_lel(A) cpu_to_le32(A)
1367#define lel_to_cpu(A) le32_to_cpu(A)
1368#elif BITS_PER_LONG == 64
1369#define LN2_BPL 6
1370#define cpu_to_lel(A) cpu_to_le64(A)
1371#define lel_to_cpu(A) le64_to_cpu(A)
1372#else
1373#error "LN2 of BITS_PER_LONG unknown!"
1374#endif
1375
1376/* resync bitmap */
1377/* 16MB sized 'bitmap extent' to track syncer usage */
1378struct bm_extent {
1379 int rs_left; /* number of bits set (out of sync) in this extent. */
1380 int rs_failed; /* number of failed resync requests in this extent. */
1381 unsigned long flags;
1382 struct lc_element lce;
1383};
1384
1385#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1386#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
e3555d85 1387#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
b411b363
PR		 1388
1389/* drbd_bitmap.c */
1390/*
1391 * We need to store one bit for a block.
1392 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1393 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1394 * Bit 1 ==> local node thinks this block needs to be synced.
1395 */
1396
8e26f9cc
PR		 1397#define SLEEP_TIME (HZ/10)
1398
b411b363
PR		 1399#define BM_BLOCK_SHIFT 12 /* 4k per bit */
1400#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1401/* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1402 * per sector of on disk bitmap */
1403#define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1404#define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1405
1406#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1407#error "HAVE YOU FIXED drbdmeta AS WELL??"
1408#endif
1409
1410/* thus many _storage_ sectors are described by one bit */
1411#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1412#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1413#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1414
1415/* bit to represented kilo byte conversion */
1416#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1417
1418/* in which _bitmap_ extent (resp. sector) the bit for a certain
1419 * _storage_ sector is located in */
1420#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1421
1422/* how much _storage_ sectors we have per bitmap sector */
1423#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1424#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1425
1426/* in one sector of the bitmap, we have this many activity_log extents. */
1427#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1428#define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1429
1430#define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1431#define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1432
1433/* the extent in "PER_EXTENT" below is an activity log extent
1434 * we need that many (long words/bytes) to store the bitmap
1435 * of one AL_EXTENT_SIZE chunk of storage.
1436 * we can store the bitmap for that many AL_EXTENTS within
1437 * one sector of the _on_disk_ bitmap:
1438 * bit 0 bit 37 bit 38 bit (512*8)-1
1439 * ...|........|........|.. // ..|........|
1440 * sect. 0 `296 `304 ^(512*8*8)-1
1441 *
1442#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1443#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1444#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1445 */
1446
1447#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1448#define DRBD_MAX_SECTORS_BM \
1449 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1450#if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1451#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1452#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
36bfc7e2 1453#elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
b411b363
PR		 1454#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1455#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1456#else
1457#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1458/* 16 TB in units of sectors */
1459#if BITS_PER_LONG == 32
1460/* adjust by one page worth of bitmap,
1461 * so we won't wrap around in drbd_bm_find_next_bit.
1462 * you should use 64bit OS for that much storage, anyways. */
1463#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1464#else
4b0715f0
LE		 1465/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1466#define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1467/* corresponds to (1UL << 38) bits right now. */
b411b363
PR		 1468#endif
1469#endif
1470
1471/* Sector shift value for the "hash" functions of tl_hash and ee_hash tables.
d5373389 1472 * With a value of 8 all IO in one 128K block make it to the same slot of the
b411b363 1473 * hash table. */
d5373389 1474#define HT_SHIFT 8
1816a2b4 1475#define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
db141b2f 1476#define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */
b411b363 1477
db141b2f 1478#define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* The old header only allows packets up to 32Kib data */
d5373389 1479
b411b363
PR		 1480/* Number of elements in the app_reads_hash */
1481#define APP_R_HSIZE 15
1482
1483extern int drbd_bm_init(struct drbd_conf *mdev);
02d9a94b 1484extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
b411b363
PR		 1485extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1486extern void drbd_bm_set_all(struct drbd_conf *mdev);
1487extern void drbd_bm_clear_all(struct drbd_conf *mdev);
4b0715f0 1488/* set/clear/test only a few bits at a time */
b411b363
PR		 1489extern int drbd_bm_set_bits(
1490 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1491extern int drbd_bm_clear_bits(
1492 struct drbd_conf *mdev, unsigned long s, unsigned long e);
4b0715f0
LE		 1493extern int drbd_bm_count_bits(
1494 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1495/* bm_set_bits variant for use while holding drbd_bm_lock,
1496 * may process the whole bitmap in one go */
b411b363
PR		 1497extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1498 const unsigned long s, const unsigned long e);
1499extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1500extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
19f843aa 1501extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
b411b363
PR		 1502extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1503extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
d1aa4d04 1504extern int drbd_bm_write_all(struct drbd_conf *mdev) __must_hold(local);
0e8488ad 1505extern int drbd_bm_write_copy_pages(struct drbd_conf *mdev) __must_hold(local);
b411b363
PR		 1506extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1507 unsigned long al_enr);
1508extern size_t drbd_bm_words(struct drbd_conf *mdev);
1509extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1510extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
4b0715f0
LE		 1511
1512#define DRBD_END_OF_BITMAP (~(unsigned long)0)
b411b363
PR		 1513extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1514/* bm_find_next variants for use while you hold drbd_bm_lock() */
1515extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1516extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
0778286a 1517extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
b411b363
PR		 1518extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1519extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1520/* for receive_bitmap */
1521extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1522 size_t number, unsigned long *buffer);
19f843aa 1523/* for _drbd_send_bitmap */
b411b363
PR		 1524extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1525 size_t number, unsigned long *buffer);
1526
20ceb2b2 1527extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
b411b363 1528extern void drbd_bm_unlock(struct drbd_conf *mdev);
b411b363
PR		 1529/* drbd_main.c */
1530
1531extern struct kmem_cache *drbd_request_cache;
1532extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1533extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1534extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1535extern mempool_t *drbd_request_mempool;
1536extern mempool_t *drbd_ee_mempool;
1537
4281808f
LE		 1538/* drbd's page pool, used to buffer data received from the peer,
1539 * or data requested by the peer.
1540 *
1541 * This does not have an emergency reserve.
1542 *
1543 * When allocating from this pool, it first takes pages from the pool.
1544 * Only if the pool is depleted will try to allocate from the system.
1545 *
1546 * The assumption is that pages taken from this pool will be processed,
1547 * and given back, "quickly", and then can be recycled, so we can avoid
1548 * frequent calls to alloc_page(), and still will be able to make progress even
1549 * under memory pressure.
1550 */
1551extern struct page *drbd_pp_pool;
b411b363
PR		 1552extern spinlock_t drbd_pp_lock;
1553extern int drbd_pp_vacant;
1554extern wait_queue_head_t drbd_pp_wait;
1555
4281808f
LE		 1556/* We also need a standard (emergency-reserve backed) page pool
1557 * for meta data IO (activity log, bitmap).
1558 * We can keep it global, as long as it is used as "N pages at a time".
1559 * 128 should be plenty, currently we probably can get away with as few as 1.
1560 */
1561#define DRBD_MIN_POOL_PAGES 128
1562extern mempool_t *drbd_md_io_page_pool;
1563
9476f39d
LE		 1564/* We also need to make sure we get a bio
1565 * when we need it for housekeeping purposes */
1566extern struct bio_set *drbd_md_io_bio_set;
1567/* to allocate from that set */
1568extern struct bio *bio_alloc_drbd(gfp_t gfp_mask);
1569
b411b363
PR		 1570extern rwlock_t global_state_lock;
1571
1572extern struct drbd_conf *drbd_new_device(unsigned int minor);
1573extern void drbd_free_mdev(struct drbd_conf *mdev);
1574
1575extern int proc_details;
1576
1577/* drbd_req */
5a7bbad2 1578extern void drbd_make_request(struct request_queue *q, struct bio *bio);
b411b363
PR		 1579extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1580extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1581extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1582
1583
1584/* drbd_nl.c */
1585extern void drbd_suspend_io(struct drbd_conf *mdev);
1586extern void drbd_resume_io(struct drbd_conf *mdev);
1587extern char *ppsize(char *buf, unsigned long long size);
a393db6f 1588extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
b411b363 1589enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
24c4830c 1590extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
b411b363 1591extern void resync_after_online_grow(struct drbd_conf *);
99432fcc 1592extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
bf885f8a
AG		 1593extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1594 enum drbd_role new_role,
1595 int force);
87f7be4c
PR		 1596extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1597extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
b411b363
PR		 1598extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1599
1600/* drbd_worker.c */
1601extern int drbd_worker(struct drbd_thread *thi);
1602extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1603extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1604extern void resume_next_sg(struct drbd_conf *mdev);
1605extern void suspend_other_sg(struct drbd_conf *mdev);
1606extern int drbd_resync_finished(struct drbd_conf *mdev);
1607/* maybe rather drbd_main.c ? */
e1711731
PR		 1608extern void *drbd_md_get_buffer(struct drbd_conf *mdev);
1609extern void drbd_md_put_buffer(struct drbd_conf *mdev);
b411b363 1610extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
e1711731 1611 struct drbd_backing_dev *bdev, sector_t sector, int rw);
44edfb0d
LE		 1612extern void wait_until_done_or_force_detached(struct drbd_conf *mdev,
1613 struct drbd_backing_dev *bdev, unsigned int *done);
b411b363 1614extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
9bd28d3c 1615extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
b411b363
PR		 1616
1617static inline void ov_oos_print(struct drbd_conf *mdev)
1618{
1619 if (mdev->ov_last_oos_size) {
1620 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1621 (unsigned long long)mdev->ov_last_oos_start,
1622 (unsigned long)mdev->ov_last_oos_size);
1623 }
1624 mdev->ov_last_oos_size=0;
1625}
1626
1627
45bb912b
LE		 1628extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1629extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
b411b363
PR		 1630/* worker callbacks */
1631extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1632extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1633extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1634extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1635extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1636extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1637extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1638extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
794abb75 1639extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int);
b411b363 1640extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
b411b363 1641extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
b411b363
PR		 1642extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1643extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1644extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1645extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1646extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
265be2d0 1647extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
73a01a18 1648extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
c4752ef1 1649extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
b411b363
PR		 1650
1651extern void resync_timer_fn(unsigned long data);
370a43e7 1652extern void start_resync_timer_fn(unsigned long data);
b411b363
PR		 1653
1654/* drbd_receiver.c */
e3555d85 1655extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
45bb912b
LE		 1656extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1657 const unsigned rw, const int fault_type);
b411b363
PR		 1658extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1659extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1660 u64 id,
1661 sector_t sector,
1662 unsigned int data_size,
1663 gfp_t gfp_mask) __must_hold(local);
435f0740
LE		 1664extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1665 int is_net);
1666#define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1667#define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
b411b363
PR		 1668extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1669 struct list_head *head);
1670extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1671 struct list_head *head);
1672extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1673extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1674extern void drbd_flush_workqueue(struct drbd_conf *mdev);
f70b3511 1675extern void drbd_free_tl_hash(struct drbd_conf *mdev);
b411b363
PR		 1676
1677/* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1678 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1679static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1680 char __user *optval, int optlen)
1681{
1682 int err;
1683 if (level == SOL_SOCKET)
1684 err = sock_setsockopt(sock, level, optname, optval, optlen);
1685 else
1686 err = sock->ops->setsockopt(sock, level, optname, optval,
1687 optlen);
1688 return err;
1689}
1690
1691static inline void drbd_tcp_cork(struct socket *sock)
1692{
1693 int __user val = 1;
1694 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1695 (char __user *)&val, sizeof(val));
1696}
1697
1698static inline void drbd_tcp_uncork(struct socket *sock)
1699{
1700 int __user val = 0;
1701 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1702 (char __user *)&val, sizeof(val));
1703}
1704
1705static inline void drbd_tcp_nodelay(struct socket *sock)
1706{
1707 int __user val = 1;
1708 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1709 (char __user *)&val, sizeof(val));
1710}
1711
1712static inline void drbd_tcp_quickack(struct socket *sock)
1713{
344fa462 1714 int __user val = 2;
b411b363
PR		 1715 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1716 (char __user *)&val, sizeof(val));
1717}
1718
1719void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1720
1721/* drbd_proc.c */
1722extern struct proc_dir_entry *drbd_proc;
7d4e9d09 1723extern const struct file_operations drbd_proc_fops;
b411b363
PR		 1724extern const char *drbd_conn_str(enum drbd_conns s);
1725extern const char *drbd_role_str(enum drbd_role s);
1726
1727/* drbd_actlog.c */
1728extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1729extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1730extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1731extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1732extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1733extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1734extern int drbd_rs_del_all(struct drbd_conf *mdev);
1735extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1736 sector_t sector, int size);
1737extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
ea5442af 1738extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
b411b363
PR		 1739extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1740 int size, const char *file, const unsigned int line);
1741#define drbd_set_in_sync(mdev, sector, size) \
1742 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
73a01a18 1743extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
b411b363
PR		 1744 int size, const char *file, const unsigned int line);
1745#define drbd_set_out_of_sync(mdev, sector, size) \
1746 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1747extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
b411b363
PR		 1748extern void drbd_al_shrink(struct drbd_conf *mdev);
1749
1750
1751/* drbd_nl.c */
1752
1753void drbd_nl_cleanup(void);
1754int __init drbd_nl_init(void);
1755void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1756void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1757void drbd_bcast_ee(struct drbd_conf *mdev,
1758 const char *reason, const int dgs,
1759 const char* seen_hash, const char* calc_hash,
1760 const struct drbd_epoch_entry* e);
1761
1762
1763/**
1764 * DOC: DRBD State macros
1765 *
1766 * These macros are used to express state changes in easily readable form.
1767 *
1768 * The NS macros expand to a mask and a value, that can be bit ored onto the
1769 * current state as soon as the spinlock (req_lock) was taken.
1770 *
1771 * The _NS macros are used for state functions that get called with the
1772 * spinlock. These macros expand directly to the new state value.
1773 *
1774 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1775 * to express state changes that affect more than one aspect of the state.
1776 *
1777 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1778 * Means that the network connection was established and that the peer
1779 * is in secondary role.
1780 */
1781#define role_MASK R_MASK
1782#define peer_MASK R_MASK
1783#define disk_MASK D_MASK
1784#define pdsk_MASK D_MASK
1785#define conn_MASK C_MASK
1786#define susp_MASK 1
1787#define user_isp_MASK 1
1788#define aftr_isp_MASK 1
fb22c402
PR		 1789#define susp_nod_MASK 1
1790#define susp_fen_MASK 1
b411b363
PR		 1791
1792#define NS(T, S) \
1793 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1794 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1795#define NS2(T1, S1, T2, S2) \
1796 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1797 mask.T2 = T2##_MASK; mask; }), \
1798 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1799 val.T2 = (S2); val; })
1800#define NS3(T1, S1, T2, S2, T3, S3) \
1801 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1802 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1803 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1804 val.T2 = (S2); val.T3 = (S3); val; })
1805
1806#define _NS(D, T, S) \
1807 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1808#define _NS2(D, T1, S1, T2, S2) \
1809 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1810 __ns.T2 = (S2); __ns; })
1811#define _NS3(D, T1, S1, T2, S2, T3, S3) \
1812 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1813 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1814
1815/*
1816 * inline helper functions
1817 *************************/
1818
45bb912b
LE		 1819/* see also page_chain_add and friends in drbd_receiver.c */
1820static inline struct page *page_chain_next(struct page *page)
1821{
1822 return (struct page *)page_private(page);
1823}
1824#define page_chain_for_each(page) \
1825 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1826 page = page_chain_next(page))
1827#define page_chain_for_each_safe(page, n) \
1828 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1829
45bb912b
LE		 1830static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1831{
1832 struct page *page = e->pages;
1833 page_chain_for_each(page) {
1834 if (page_count(page) > 1)
1835 return 1;
1836 }
1837 return 0;
1838}
1839
b411b363
PR		 1840static inline void drbd_state_lock(struct drbd_conf *mdev)
1841{
1842 wait_event(mdev->misc_wait,
06f10adb 1843 !drbd_test_and_set_flag(mdev, CLUSTER_ST_CHANGE));
b411b363
PR		 1844}
1845
1846static inline void drbd_state_unlock(struct drbd_conf *mdev)
1847{
06f10adb 1848 drbd_clear_flag(mdev, CLUSTER_ST_CHANGE);
b411b363
PR		 1849 wake_up(&mdev->misc_wait);
1850}
1851
bf885f8a
AG		 1852static inline enum drbd_state_rv
1853_drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1854 enum chg_state_flags flags, struct completion *done)
b411b363 1855{
bf885f8a 1856 enum drbd_state_rv rv;
b411b363
PR		 1857
1858 read_lock(&global_state_lock);
1859 rv = __drbd_set_state(mdev, ns, flags, done);
1860 read_unlock(&global_state_lock);
1861
1862 return rv;
1863}
1864
1865/**
1866 * drbd_request_state() - Reqest a state change
1867 * @mdev: DRBD device.
1868 * @mask: mask of state bits to change.
1869 * @val: value of new state bits.
1870 *
1871 * This is the most graceful way of requesting a state change. It is verbose
1872 * quite verbose in case the state change is not possible, and all those
1873 * state changes are globally serialized.
1874 */
1875static inline int drbd_request_state(struct drbd_conf *mdev,
1876 union drbd_state mask,
1877 union drbd_state val)
1878{
1879 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1880}
1881
383606e0 1882enum drbd_force_detach_flags {
a2a3c74f
LE		 1883 DRBD_READ_ERROR,
1884 DRBD_WRITE_ERROR,
383606e0
LE		 1885 DRBD_META_IO_ERROR,
1886 DRBD_FORCE_DETACH,
1887};
1888
b411b363 1889#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
383606e0 1890static inline void __drbd_chk_io_error_(struct drbd_conf *mdev,
a2a3c74f 1891 enum drbd_force_detach_flags df,
383606e0 1892 const char *where)
b411b363
PR		 1893{
1894 switch (mdev->ldev->dc.on_io_error) {
1895 case EP_PASS_ON:
a2a3c74f 1896 if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) {
7383506c 1897 if (__ratelimit(&drbd_ratelimit_state))
82f59cc6 1898 dev_err(DEV, "Local IO failed in %s.\n", where);
d2e17807
PR		 1899 if (mdev->state.disk > D_INCONSISTENT)
1900 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
b411b363
PR		 1901 break;
1902 }
a2a3c74f 1903 /* NOTE fall through for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */
b411b363
PR		 1904 case EP_DETACH:
1905 case EP_CALL_HELPER:
a2a3c74f
LE		 1906 /* Remember whether we saw a READ or WRITE error.
1907 *
1908 * Recovery of the affected area for WRITE failure is covered
1909 * by the activity log.
1910 * READ errors may fall outside that area though. Certain READ
1911 * errors can be "healed" by writing good data to the affected
1912 * blocks, which triggers block re-allocation in lower layers.
1913 *
1914 * If we can not write the bitmap after a READ error,
1915 * we may need to trigger a full sync (see w_go_diskless()).
1916 *
1917 * Force-detach is not really an IO error, but rather a
1918 * desperate measure to try to deal with a completely
1919 * unresponsive lower level IO stack.
1920 * Still it should be treated as a WRITE error.
1921 *
1922 * Meta IO error is always WRITE error:
1923 * we read meta data only once during attach,
1924 * which will fail in case of errors.
1925 */
06f10adb 1926 drbd_set_flag(mdev, WAS_IO_ERROR);
a2a3c74f
LE		 1927 if (df == DRBD_READ_ERROR)
1928 drbd_set_flag(mdev, WAS_READ_ERROR);
1929 if (df == DRBD_FORCE_DETACH)
06f10adb 1930 drbd_set_flag(mdev, FORCE_DETACH);
b411b363
PR		 1931 if (mdev->state.disk > D_FAILED) {
1932 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
82f59cc6
LE		 1933 dev_err(DEV,
1934 "Local IO failed in %s. Detaching...\n", where);
b411b363
PR		 1935 }
1936 break;
1937 }
1938}
1939
1940/**
1941 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1942 * @mdev: DRBD device.
1943 * @error: Error code passed to the IO completion callback
1944 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1945 *
1946 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1947 */
1948#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1949static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
383606e0 1950 int error, enum drbd_force_detach_flags forcedetach, const char *where)
b411b363
PR		 1951{
1952 if (error) {
1953 unsigned long flags;
1954 spin_lock_irqsave(&mdev->req_lock, flags);
1955 __drbd_chk_io_error_(mdev, forcedetach, where);
1956 spin_unlock_irqrestore(&mdev->req_lock, flags);
1957 }
1958}
1959
1960
1961/**
1962 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1963 * @bdev: Meta data block device.
1964 *
1965 * BTW, for internal meta data, this happens to be the maximum capacity
1966 * we could agree upon with our peer node.
1967 */
1968static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1969{
1970 switch (bdev->dc.meta_dev_idx) {
1971 case DRBD_MD_INDEX_INTERNAL:
1972 case DRBD_MD_INDEX_FLEX_INT:
1973 return bdev->md.md_offset + bdev->md.bm_offset;
1974 case DRBD_MD_INDEX_FLEX_EXT:
1975 default:
1976 return bdev->md.md_offset;
1977 }
1978}
1979
1980/**
1981 * drbd_md_last_sector() - Return the last sector number of the meta data area
1982 * @bdev: Meta data block device.
1983 */
1984static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1985{
1986 switch (bdev->dc.meta_dev_idx) {
1987 case DRBD_MD_INDEX_INTERNAL:
1988 case DRBD_MD_INDEX_FLEX_INT:
1989 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1990 case DRBD_MD_INDEX_FLEX_EXT:
1991 default:
1992 return bdev->md.md_offset + bdev->md.md_size_sect;
1993 }
1994}
1995
1996/* Returns the number of 512 byte sectors of the device */
1997static inline sector_t drbd_get_capacity(struct block_device *bdev)
1998{
1999 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
77304d2a 2000 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
b411b363
PR		 2001}
2002
2003/**
2004 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
2005 * @bdev: Meta data block device.
2006 *
2007 * returns the capacity we announce to out peer. we clip ourselves at the
2008 * various MAX_SECTORS, because if we don't, current implementation will
2009 * oops sooner or later
2010 */
2011static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
2012{
2013 sector_t s;
2014 switch (bdev->dc.meta_dev_idx) {
2015 case DRBD_MD_INDEX_INTERNAL:
2016 case DRBD_MD_INDEX_FLEX_INT:
2017 s = drbd_get_capacity(bdev->backing_bdev)
2018 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
2019 drbd_md_first_sector(bdev))
2020 : 0;
2021 break;
2022 case DRBD_MD_INDEX_FLEX_EXT:
2023 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
2024 drbd_get_capacity(bdev->backing_bdev));
2025 /* clip at maximum size the meta device can support */
2026 s = min_t(sector_t, s,
2027 BM_EXT_TO_SECT(bdev->md.md_size_sect
2028 - bdev->md.bm_offset));
2029 break;
2030 default:
2031 s = min_t(sector_t, DRBD_MAX_SECTORS,
2032 drbd_get_capacity(bdev->backing_bdev));
2033 }
2034 return s;
2035}
2036
2037/**
2038 * drbd_md_ss__() - Return the sector number of our meta data super block
2039 * @mdev: DRBD device.
2040 * @bdev: Meta data block device.
2041 */
2042static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
2043 struct drbd_backing_dev *bdev)
2044{
2045 switch (bdev->dc.meta_dev_idx) {
2046 default: /* external, some index */
2047 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
2048 case DRBD_MD_INDEX_INTERNAL:
2049 /* with drbd08, internal meta data is always "flexible" */
2050 case DRBD_MD_INDEX_FLEX_INT:
2051 /* sizeof(struct md_on_disk_07) == 4k
2052 * position: last 4k aligned block of 4k size */
2053 if (!bdev->backing_bdev) {
2054 if (__ratelimit(&drbd_ratelimit_state)) {
2055 dev_err(DEV, "bdev->backing_bdev==NULL\n");
2056 dump_stack();
2057 }
2058 return 0;
2059 }
2060 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
2061 - MD_AL_OFFSET;
2062 case DRBD_MD_INDEX_FLEX_EXT:
2063 return 0;
2064 }
2065}
2066
b411b363
PR		 2067static inline void
2068drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
2069{
2070 unsigned long flags;
2071 spin_lock_irqsave(&q->q_lock, flags);
2072 list_add(&w->list, &q->q);
2073 up(&q->s); /* within the spinlock,
2074 see comment near end of drbd_worker() */
2075 spin_unlock_irqrestore(&q->q_lock, flags);
2076}
2077
2078static inline void
2079drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
2080{
2081 unsigned long flags;
2082 spin_lock_irqsave(&q->q_lock, flags);
2083 list_add_tail(&w->list, &q->q);
2084 up(&q->s); /* within the spinlock,
2085 see comment near end of drbd_worker() */
2086 spin_unlock_irqrestore(&q->q_lock, flags);
2087}
2088
2089static inline void wake_asender(struct drbd_conf *mdev)
2090{
06f10adb 2091 if (drbd_test_flag(mdev, SIGNAL_ASENDER))
b411b363
PR		 2092 force_sig(DRBD_SIG, mdev->asender.task);
2093}
2094
2095static inline void request_ping(struct drbd_conf *mdev)
2096{
06f10adb 2097 drbd_set_flag(mdev, SEND_PING);
b411b363
PR		 2098 wake_asender(mdev);
2099}
2100
2101static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
2102 enum drbd_packets cmd)
2103{
0b70a13d 2104 struct p_header80 h;
b411b363
PR		 2105 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
2106}
2107
2108static inline int drbd_send_ping(struct drbd_conf *mdev)
2109{
0b70a13d 2110 struct p_header80 h;
b411b363
PR		 2111 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
2112}
2113
2114static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
2115{
0b70a13d 2116 struct p_header80 h;
b411b363
PR		 2117 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
2118}
2119
2120static inline void drbd_thread_stop(struct drbd_thread *thi)
2121{
81e84650 2122 _drbd_thread_stop(thi, false, true);
b411b363
PR		 2123}
2124
2125static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2126{
81e84650 2127 _drbd_thread_stop(thi, false, false);
b411b363
PR		 2128}
2129
2130static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2131{
81e84650 2132 _drbd_thread_stop(thi, true, false);
b411b363
PR		 2133}
2134
2135/* counts how many answer packets packets we expect from our peer,
2136 * for either explicit application requests,
2137 * or implicit barrier packets as necessary.
2138 * increased:
2139 * w_send_barrier
2140 * _req_mod(req, queue_for_net_write or queue_for_net_read);
2141 * it is much easier and equally valid to count what we queue for the
2142 * worker, even before it actually was queued or send.
2143 * (drbd_make_request_common; recovery path on read io-error)
2144 * decreased:
2145 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
2146 * _req_mod(req, data_received)
2147 * [from receive_DataReply]
2148 * _req_mod(req, write_acked_by_peer or recv_acked_by_peer or neg_acked)
2149 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2150 * for some reason it is NOT decreased in got_NegAck,
2151 * but in the resulting cleanup code from report_params.
2152 * we should try to remember the reason for that...
2153 * _req_mod(req, send_failed or send_canceled)
2154 * _req_mod(req, connection_lost_while_pending)
2155 * [from tl_clear_barrier]
2156 */
2157static inline void inc_ap_pending(struct drbd_conf *mdev)
2158{
2159 atomic_inc(&mdev->ap_pending_cnt);
2160}
2161
2162#define ERR_IF_CNT_IS_NEGATIVE(which) \
2163 if (atomic_read(&mdev->which) < 0) \
2164 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2165 __func__ , __LINE__ , \
2166 atomic_read(&mdev->which))
2167
2168#define dec_ap_pending(mdev) do { \
2169 typecheck(struct drbd_conf *, mdev); \
2170 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2171 wake_up(&mdev->misc_wait); \
2172 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2173
2174/* counts how many resync-related answers we still expect from the peer
2175 * increase decrease
2176 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
25985edc 2177 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
b411b363
PR		 2178 * (or P_NEG_ACK with ID_SYNCER)
2179 */
2180static inline void inc_rs_pending(struct drbd_conf *mdev)
2181{
2182 atomic_inc(&mdev->rs_pending_cnt);
2183}
2184
2185#define dec_rs_pending(mdev) do { \
2186 typecheck(struct drbd_conf *, mdev); \
2187 atomic_dec(&mdev->rs_pending_cnt); \
2188 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2189
2190/* counts how many answers we still need to send to the peer.
2191 * increased on
2192 * receive_Data unless protocol A;
2193 * we need to send a P_RECV_ACK (proto B)
2194 * or P_WRITE_ACK (proto C)
2195 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2196 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2197 * receive_Barrier_* we need to send a P_BARRIER_ACK
2198 */
2199static inline void inc_unacked(struct drbd_conf *mdev)
2200{
2201 atomic_inc(&mdev->unacked_cnt);
2202}
2203
2204#define dec_unacked(mdev) do { \
2205 typecheck(struct drbd_conf *, mdev); \
2206 atomic_dec(&mdev->unacked_cnt); \
2207 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2208
2209#define sub_unacked(mdev, n) do { \
2210 typecheck(struct drbd_conf *, mdev); \
2211 atomic_sub(n, &mdev->unacked_cnt); \
2212 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2213
2214
2215static inline void put_net_conf(struct drbd_conf *mdev)
2216{
2217 if (atomic_dec_and_test(&mdev->net_cnt))
84dfb9f5 2218 wake_up(&mdev->net_cnt_wait);
b411b363
PR		 2219}
2220
2221/**
2222 * get_net_conf() - Increase ref count on mdev->net_conf; Returns 0 if nothing there
2223 * @mdev: DRBD device.
2224 *
2225 * You have to call put_net_conf() when finished working with mdev->net_conf.
2226 */
2227static inline int get_net_conf(struct drbd_conf *mdev)
2228{
2229 int have_net_conf;
2230
2231 atomic_inc(&mdev->net_cnt);
2232 have_net_conf = mdev->state.conn >= C_UNCONNECTED;
2233 if (!have_net_conf)
2234 put_net_conf(mdev);
2235 return have_net_conf;
2236}
2237
2238/**
2239 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2240 * @M: DRBD device.
2241 *
2242 * You have to call put_ldev() when finished working with mdev->ldev.
2243 */
2244#define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2245#define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2246
2247static inline void put_ldev(struct drbd_conf *mdev)
2248{
1d7734a0 2249 int i = atomic_dec_return(&mdev->local_cnt);
9a0d9d03
LE		 2250
2251 /* This may be called from some endio handler,
2252 * so we must not sleep here. */
2253
b411b363 2254 __release(local);
1d7734a0 2255 D_ASSERT(i >= 0);
e9e6f3ec 2256 if (i == 0) {
82f59cc6
LE		 2257 if (mdev->state.disk == D_DISKLESS)
2258 /* even internal references gone, safe to destroy */
2259 drbd_ldev_destroy(mdev);
e9e6f3ec 2260 if (mdev->state.disk == D_FAILED)
82f59cc6 2261 /* all application IO references gone. */
e9e6f3ec 2262 drbd_go_diskless(mdev);
b411b363 2263 wake_up(&mdev->misc_wait);
e9e6f3ec 2264 }
b411b363
PR		 2265}
2266
2267#ifndef __CHECKER__
2268static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2269{
2270 int io_allowed;
2271
82f59cc6
LE		 2272 /* never get a reference while D_DISKLESS */
2273 if (mdev->state.disk == D_DISKLESS)
2274 return 0;
2275
b411b363
PR		 2276 atomic_inc(&mdev->local_cnt);
2277 io_allowed = (mdev->state.disk >= mins);
2278 if (!io_allowed)
2279 put_ldev(mdev);
2280 return io_allowed;
2281}
2282#else
2283extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2284#endif
2285
2286/* you must have an "get_ldev" reference */
2287static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2288 unsigned long *bits_left, unsigned int *per_mil_done)
2289{
4b0715f0
LE		 2290 /* this is to break it at compile time when we change that, in case we
2291 * want to support more than (1<<32) bits on a 32bit arch. */
b411b363
PR		 2292 typecheck(unsigned long, mdev->rs_total);
2293
2294 /* note: both rs_total and rs_left are in bits, i.e. in
2295 * units of BM_BLOCK_SIZE.
2296 * for the percentage, we don't care. */
2297
439d5953
LE		 2298 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2299 *bits_left = mdev->ov_left;
2300 else
2301 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
b411b363
PR		 2302 /* >> 10 to prevent overflow,
2303 * +1 to prevent division by zero */
2304 if (*bits_left > mdev->rs_total) {
2305 /* doh. maybe a logic bug somewhere.
2306 * may also be just a race condition
2307 * between this and a disconnect during sync.
2308 * for now, just prevent in-kernel buffer overflow.
2309 */
2310 smp_rmb();
2311 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2312 drbd_conn_str(mdev->state.conn),
2313 *bits_left, mdev->rs_total, mdev->rs_failed);
2314 *per_mil_done = 0;
2315 } else {
4b0715f0
LE		 2316 /* Make sure the division happens in long context.
2317 * We allow up to one petabyte storage right now,
2318 * at a granularity of 4k per bit that is 2**38 bits.
2319 * After shift right and multiplication by 1000,
2320 * this should still fit easily into a 32bit long,
2321 * so we don't need a 64bit division on 32bit arch.
2322 * Note: currently we don't support such large bitmaps on 32bit
2323 * arch anyways, but no harm done to be prepared for it here.
2324 */
5f138ce0 2325 unsigned int shift = mdev->rs_total > UINT_MAX ? 16 : 10;
4b0715f0
LE		 2326 unsigned long left = *bits_left >> shift;
2327 unsigned long total = 1UL + (mdev->rs_total >> shift);
2328 unsigned long tmp = 1000UL - left * 1000UL/total;
b411b363
PR		 2329 *per_mil_done = tmp;
2330 }
2331}
2332
2333
2334/* this throttles on-the-fly application requests
2335 * according to max_buffers settings;
2336 * maybe re-implement using semaphores? */
2337static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2338{
2339 int mxb = 1000000; /* arbitrary limit on open requests */
2340 if (get_net_conf(mdev)) {
2341 mxb = mdev->net_conf->max_buffers;
2342 put_net_conf(mdev);
2343 }
2344 return mxb;
2345}
2346
3719094e 2347static inline int drbd_state_is_stable(struct drbd_conf *mdev)
b411b363 2348{
3719094e 2349 union drbd_state s = mdev->state;
b411b363
PR		 2350
2351 /* DO NOT add a default clause, we want the compiler to warn us
2352 * for any newly introduced state we may have forgotten to add here */
2353
2354 switch ((enum drbd_conns)s.conn) {
2355 /* new io only accepted when there is no connection, ... */
2356 case C_STANDALONE:
2357 case C_WF_CONNECTION:
2358 /* ... or there is a well established connection. */
2359 case C_CONNECTED:
2360 case C_SYNC_SOURCE:
2361 case C_SYNC_TARGET:
2362 case C_VERIFY_S:
2363 case C_VERIFY_T:
2364 case C_PAUSED_SYNC_S:
2365 case C_PAUSED_SYNC_T:
67531718
PR		 2366 case C_AHEAD:
2367 case C_BEHIND:
3719094e 2368 /* transitional states, IO allowed */
b411b363
PR		 2369 case C_DISCONNECTING:
2370 case C_UNCONNECTED:
2371 case C_TIMEOUT:
2372 case C_BROKEN_PIPE:
2373 case C_NETWORK_FAILURE:
2374 case C_PROTOCOL_ERROR:
2375 case C_TEAR_DOWN:
2376 case C_WF_REPORT_PARAMS:
2377 case C_STARTING_SYNC_S:
2378 case C_STARTING_SYNC_T:
3719094e
PR		 2379 break;
2380
2381 /* Allow IO in BM exchange states with new protocols */
b411b363 2382 case C_WF_BITMAP_S:
3719094e
PR		 2383 if (mdev->agreed_pro_version < 96)
2384 return 0;
2385 break;
2386
2387 /* no new io accepted in these states */
b411b363
PR		 2388 case C_WF_BITMAP_T:
2389 case C_WF_SYNC_UUID:
2390 case C_MASK:
2391 /* not "stable" */
2392 return 0;
2393 }
2394
2395 switch ((enum drbd_disk_state)s.disk) {
2396 case D_DISKLESS:
2397 case D_INCONSISTENT:
2398 case D_OUTDATED:
2399 case D_CONSISTENT:
2400 case D_UP_TO_DATE:
5ca1de03 2401 case D_FAILED:
b411b363
PR		 2402 /* disk state is stable as well. */
2403 break;
2404
2405 /* no new io accepted during tansitional states */
2406 case D_ATTACHING:
b411b363
PR		 2407 case D_NEGOTIATING:
2408 case D_UNKNOWN:
2409 case D_MASK:
2410 /* not "stable" */
2411 return 0;
2412 }
2413
2414 return 1;
2415}
2416
fb22c402
PR		 2417static inline int is_susp(union drbd_state s)
2418{
2419 return s.susp || s.susp_nod || s.susp_fen;
2420}
2421
1b881ef7 2422static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
b411b363
PR		 2423{
2424 int mxb = drbd_get_max_buffers(mdev);
2425
fb22c402 2426 if (is_susp(mdev->state))
1b881ef7 2427 return false;
06f10adb 2428 if (drbd_test_flag(mdev, SUSPEND_IO))
1b881ef7 2429 return false;
b411b363
PR		 2430
2431 /* to avoid potential deadlock or bitmap corruption,
2432 * in various places, we only allow new application io
2433 * to start during "stable" states. */
2434
2435 /* no new io accepted when attaching or detaching the disk */
3719094e 2436 if (!drbd_state_is_stable(mdev))
1b881ef7 2437 return false;
b411b363
PR		 2438
2439 /* since some older kernels don't have atomic_add_unless,
2440 * and we are within the spinlock anyways, we have this workaround. */
2441 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
1b881ef7 2442 return false;
06f10adb 2443 if (drbd_test_flag(mdev, BITMAP_IO))
1b881ef7
AG		 2444 return false;
2445 return true;
b411b363
PR		 2446}
2447
1b881ef7 2448static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
b411b363 2449{
1b881ef7 2450 bool rv = false;
8869d683
PR		 2451
2452 spin_lock_irq(&mdev->req_lock);
1b881ef7 2453 rv = may_inc_ap_bio(mdev);
8869d683
PR		 2454 if (rv)
2455 atomic_add(count, &mdev->ap_bio_cnt);
2456 spin_unlock_irq(&mdev->req_lock);
2457
2458 return rv;
2459}
b411b363 2460
8869d683
PR		 2461static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2462{
b411b363
PR		 2463 /* we wait here
2464 * as long as the device is suspended
2465 * until the bitmap is no longer on the fly during connection
2466 * handshake as long as we would exeed the max_buffer limit.
2467 *
2468 * to avoid races with the reconnect code,
2469 * we need to atomic_inc within the spinlock. */
2470
1b881ef7 2471 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
b411b363
PR		 2472}
2473
2474static inline void dec_ap_bio(struct drbd_conf *mdev)
2475{
2476 int mxb = drbd_get_max_buffers(mdev);
2477 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2478
2479 D_ASSERT(ap_bio >= 0);
7ee1fb93 2480
06f10adb
LE		 2481 if (ap_bio == 0 && drbd_test_flag(mdev, BITMAP_IO)) {
2482 if (!drbd_test_and_set_flag(mdev, BITMAP_IO_QUEUED))
7ee1fb93
LE		 2483 drbd_queue_work(&mdev->data.work, &mdev->bm_io_work.w);
2484 }
2485
b411b363
PR		 2486 /* this currently does wake_up for every dec_ap_bio!
2487 * maybe rather introduce some type of hysteresis?
2488 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2489 if (ap_bio < mxb)
2490 wake_up(&mdev->misc_wait);
b411b363
PR		 2491}
2492
62b0da3a 2493static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
b411b363 2494{
62b0da3a 2495 int changed = mdev->ed_uuid != val;
b411b363 2496 mdev->ed_uuid = val;
62b0da3a 2497 return changed;
b411b363
PR		 2498}
2499
2500static inline int seq_cmp(u32 a, u32 b)
2501{
2502 /* we assume wrap around at 32bit.
2503 * for wrap around at 24bit (old atomic_t),
2504 * we'd have to
2505 * a <<= 8; b <<= 8;
2506 */
2507 return (s32)(a) - (s32)(b);
2508}
2509#define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2510#define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2511#define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2512#define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2513/* CAUTION: please no side effects in arguments! */
2514#define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2515
2516static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2517{
2518 unsigned int m;
2519 spin_lock(&mdev->peer_seq_lock);
2520 m = seq_max(mdev->peer_seq, new_seq);
2521 mdev->peer_seq = m;
2522 spin_unlock(&mdev->peer_seq_lock);
2523 if (m == new_seq)
2524 wake_up(&mdev->seq_wait);
2525}
2526
2527static inline void drbd_update_congested(struct drbd_conf *mdev)
2528{
2529 struct sock *sk = mdev->data.socket->sk;
2530 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
06f10adb 2531 drbd_set_flag(mdev, NET_CONGESTED);
b411b363
PR		 2532}
2533
2534static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2535{
2536 /* sorry, we currently have no working implementation
2537 * of distributed TCQ stuff */
2538#ifndef QUEUE_ORDERED_NONE
2539#define QUEUE_ORDERED_NONE 0
2540#endif
2541 return QUEUE_ORDERED_NONE;
2542}
2543
b411b363
PR		 2544static inline void drbd_md_flush(struct drbd_conf *mdev)
2545{
2546 int r;
2547
06f10adb 2548 if (drbd_test_flag(mdev, MD_NO_FUA))
b411b363
PR		 2549 return;
2550
8b45a5c8 2551 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_NOIO, NULL);
b411b363 2552 if (r) {
06f10adb 2553 drbd_set_flag(mdev, MD_NO_FUA);
b411b363
PR		 2554 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
2555 }
2556}
2557
06f10adb 2558
b411b363 2559#endif
Linux 6.x block layer and io_uring tree(s)