projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drbd: moved receiver, worker and asender from mdev to tconn
[deliverable/linux.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>
31#include <linux/version.h>
32#include <linux/list.h>
33#include <linux/sched.h>
34#include <linux/bitops.h>
35#include <linux/slab.h>
36#include <linux/crypto.h>
132cc538 37#include <linux/ratelimit.h>
b411b363
PR		 38#include <linux/tcp.h>
39#include <linux/mutex.h>
40#include <linux/major.h>
41#include <linux/blkdev.h>
42#include <linux/genhd.h>
43#include <net/tcp.h>
44#include <linux/lru_cache.h>
70c71606 45#include <linux/prefetch.h>
b411b363
PR		 46
47#ifdef __CHECKER__
48# define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr")))
49# define __protected_read_by(x) __attribute__((require_context(x,1,999,"read")))
50# define __protected_write_by(x) __attribute__((require_context(x,1,999,"write")))
51# define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call")))
52#else
53# define __protected_by(x)
54# define __protected_read_by(x)
55# define __protected_write_by(x)
56# define __must_hold(x)
57#endif
58
59#define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0)
60
61/* module parameter, defined in drbd_main.c */
62extern unsigned int minor_count;
63extern int disable_sendpage;
64extern int allow_oos;
65extern unsigned int cn_idx;
66
67#ifdef CONFIG_DRBD_FAULT_INJECTION
68extern int enable_faults;
69extern int fault_rate;
70extern int fault_devs;
71#endif
72
73extern char usermode_helper[];
74
75
b411b363
PR		 76/* I don't remember why XCPU ...
77 * This is used to wake the asender,
78 * and to interrupt sending the sending task
79 * on disconnect.
80 */
81#define DRBD_SIG SIGXCPU
82
83/* This is used to stop/restart our threads.
84 * Cannot use SIGTERM nor SIGKILL, since these
85 * are sent out by init on runlevel changes
86 * I choose SIGHUP for now.
87 */
88#define DRBD_SIGKILL SIGHUP
89
b411b363
PR		 90#define ID_IN_SYNC (4711ULL)
91#define ID_OUT_OF_SYNC (4712ULL)
b411b363 92#define ID_SYNCER (-1ULL)
579b57ed 93
4a23f264 94#define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL)
b411b363
PR		 95
96struct drbd_conf;
2111438b 97struct drbd_tconn;
b411b363
PR		 98
99
100/* to shorten dev_warn(DEV, "msg"); and relatives statements */
101#define DEV (disk_to_dev(mdev->vdisk))
102
103#define D_ASSERT(exp) if (!(exp)) \
104 dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__)
105
841ce241
AG		 106/**
107 * expect - Make an assertion
108 *
109 * Unlike the assert macro, this macro returns a boolean result.
110 */
111#define expect(exp) ({ \
112 bool _bool = (exp); \
113 if (!_bool) \
114 dev_err(DEV, "ASSERTION %s FAILED in %s\n", \
115 #exp, __func__); \
116 _bool; \
117 })
b411b363
PR		 118
119/* Defines to control fault insertion */
120enum {
121 DRBD_FAULT_MD_WR = 0, /* meta data write */
122 DRBD_FAULT_MD_RD = 1, /* read */
123 DRBD_FAULT_RS_WR = 2, /* resync */
124 DRBD_FAULT_RS_RD = 3,
125 DRBD_FAULT_DT_WR = 4, /* data */
126 DRBD_FAULT_DT_RD = 5,
127 DRBD_FAULT_DT_RA = 6, /* data read ahead */
128 DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */
129 DRBD_FAULT_AL_EE = 8, /* alloc ee */
6b4388ac 130 DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */
b411b363
PR		 131
132 DRBD_FAULT_MAX,
133};
134
b411b363
PR		 135extern unsigned int
136_drbd_insert_fault(struct drbd_conf *mdev, unsigned int type);
0cf9d27e 137
b411b363
PR		 138static inline int
139drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) {
0cf9d27e 140#ifdef CONFIG_DRBD_FAULT_INJECTION
b411b363
PR		 141 return fault_rate &&
142 (enable_faults & (1<<type)) &&
143 _drbd_insert_fault(mdev, type);
b411b363 144#else
0cf9d27e 145 return 0;
b411b363 146#endif
0cf9d27e 147}
b411b363
PR		 148
149/* integer division, round _UP_ to the next integer */
150#define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0))
151/* usual integer division */
152#define div_floor(A, B) ((A)/(B))
153
154/* drbd_meta-data.c (still in drbd_main.c) */
155/* 4th incarnation of the disk layout. */
156#define DRBD_MD_MAGIC (DRBD_MAGIC+4)
157
158extern struct drbd_conf **minor_table;
159extern struct ratelimit_state drbd_ratelimit_state;
160
161/* on the wire */
162enum drbd_packets {
163 /* receiver (data socket) */
164 P_DATA = 0x00,
165 P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */
166 P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */
167 P_BARRIER = 0x03,
168 P_BITMAP = 0x04,
169 P_BECOME_SYNC_TARGET = 0x05,
170 P_BECOME_SYNC_SOURCE = 0x06,
171 P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */
172 P_DATA_REQUEST = 0x08, /* Used to ask for a data block */
173 P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */
174 P_SYNC_PARAM = 0x0a,
175 P_PROTOCOL = 0x0b,
176 P_UUIDS = 0x0c,
177 P_SIZES = 0x0d,
178 P_STATE = 0x0e,
179 P_SYNC_UUID = 0x0f,
180 P_AUTH_CHALLENGE = 0x10,
181 P_AUTH_RESPONSE = 0x11,
182 P_STATE_CHG_REQ = 0x12,
183
184 /* asender (meta socket */
185 P_PING = 0x13,
186 P_PING_ACK = 0x14,
187 P_RECV_ACK = 0x15, /* Used in protocol B */
188 P_WRITE_ACK = 0x16, /* Used in protocol C */
189 P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */
190 P_DISCARD_ACK = 0x18, /* Used in proto C, two-primaries conflict detection */
191 P_NEG_ACK = 0x19, /* Sent if local disk is unusable */
192 P_NEG_DREPLY = 0x1a, /* Local disk is broken... */
193 P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */
194 P_BARRIER_ACK = 0x1c,
195 P_STATE_CHG_REPLY = 0x1d,
196
197 /* "new" commands, no longer fitting into the ordering scheme above */
198
199 P_OV_REQUEST = 0x1e, /* data socket */
200 P_OV_REPLY = 0x1f,
201 P_OV_RESULT = 0x20, /* meta socket */
202 P_CSUM_RS_REQUEST = 0x21, /* data socket */
203 P_RS_IS_IN_SYNC = 0x22, /* meta socket */
204 P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */
205 P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */
0ced55a3
PR		 206 /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */
207 /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */
208 P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */
73a01a18 209 P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */
d612d309 210 P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */
b411b363 211
d612d309 212 P_MAX_CMD = 0x2A,
b411b363
PR		 213 P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */
214 P_MAX_OPT_CMD = 0x101,
215
216 /* special command ids for handshake */
217
218 P_HAND_SHAKE_M = 0xfff1, /* First Packet on the MetaSock */
219 P_HAND_SHAKE_S = 0xfff2, /* First Packet on the Socket */
220
221 P_HAND_SHAKE = 0xfffe /* FIXED for the next century! */
222};
223
224static inline const char *cmdname(enum drbd_packets cmd)
225{
226 /* THINK may need to become several global tables
227 * when we want to support more than
228 * one PRO_VERSION */
229 static const char *cmdnames[] = {
230 [P_DATA] = "Data",
231 [P_DATA_REPLY] = "DataReply",
232 [P_RS_DATA_REPLY] = "RSDataReply",
233 [P_BARRIER] = "Barrier",
234 [P_BITMAP] = "ReportBitMap",
235 [P_BECOME_SYNC_TARGET] = "BecomeSyncTarget",
236 [P_BECOME_SYNC_SOURCE] = "BecomeSyncSource",
237 [P_UNPLUG_REMOTE] = "UnplugRemote",
238 [P_DATA_REQUEST] = "DataRequest",
239 [P_RS_DATA_REQUEST] = "RSDataRequest",
240 [P_SYNC_PARAM] = "SyncParam",
241 [P_SYNC_PARAM89] = "SyncParam89",
242 [P_PROTOCOL] = "ReportProtocol",
243 [P_UUIDS] = "ReportUUIDs",
244 [P_SIZES] = "ReportSizes",
245 [P_STATE] = "ReportState",
246 [P_SYNC_UUID] = "ReportSyncUUID",
247 [P_AUTH_CHALLENGE] = "AuthChallenge",
248 [P_AUTH_RESPONSE] = "AuthResponse",
249 [P_PING] = "Ping",
250 [P_PING_ACK] = "PingAck",
251 [P_RECV_ACK] = "RecvAck",
252 [P_WRITE_ACK] = "WriteAck",
253 [P_RS_WRITE_ACK] = "RSWriteAck",
254 [P_DISCARD_ACK] = "DiscardAck",
255 [P_NEG_ACK] = "NegAck",
256 [P_NEG_DREPLY] = "NegDReply",
257 [P_NEG_RS_DREPLY] = "NegRSDReply",
258 [P_BARRIER_ACK] = "BarrierAck",
259 [P_STATE_CHG_REQ] = "StateChgRequest",
260 [P_STATE_CHG_REPLY] = "StateChgReply",
261 [P_OV_REQUEST] = "OVRequest",
262 [P_OV_REPLY] = "OVReply",
263 [P_OV_RESULT] = "OVResult",
c42b6cf4
LE		 264 [P_CSUM_RS_REQUEST] = "CsumRSRequest",
265 [P_RS_IS_IN_SYNC] = "CsumRSIsInSync",
266 [P_COMPRESSED_BITMAP] = "CBitmap",
a8cdfd8d 267 [P_DELAY_PROBE] = "DelayProbe",
73a01a18 268 [P_OUT_OF_SYNC] = "OutOfSync",
b411b363
PR		 269 [P_MAX_CMD] = NULL,
270 };
271
272 if (cmd == P_HAND_SHAKE_M)
273 return "HandShakeM";
274 if (cmd == P_HAND_SHAKE_S)
275 return "HandShakeS";
276 if (cmd == P_HAND_SHAKE)
277 return "HandShake";
278 if (cmd >= P_MAX_CMD)
279 return "Unknown";
280 return cmdnames[cmd];
281}
282
283/* for sending/receiving the bitmap,
284 * possibly in some encoding scheme */
285struct bm_xfer_ctx {
286 /* "const"
287 * stores total bits and long words
288 * of the bitmap, so we don't need to
289 * call the accessor functions over and again. */
290 unsigned long bm_bits;
291 unsigned long bm_words;
292 /* during xfer, current position within the bitmap */
293 unsigned long bit_offset;
294 unsigned long word_offset;
295
296 /* statistics; index: (h->command == P_BITMAP) */
297 unsigned packets[2];
298 unsigned bytes[2];
299};
300
301extern void INFO_bm_xfer_stats(struct drbd_conf *mdev,
302 const char *direction, struct bm_xfer_ctx *c);
303
304static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c)
305{
306 /* word_offset counts "native long words" (32 or 64 bit),
307 * aligned at 64 bit.
308 * Encoded packet may end at an unaligned bit offset.
309 * In case a fallback clear text packet is transmitted in
310 * between, we adjust this offset back to the last 64bit
311 * aligned "native long word", which makes coding and decoding
312 * the plain text bitmap much more convenient. */
313#if BITS_PER_LONG == 64
314 c->word_offset = c->bit_offset >> 6;
315#elif BITS_PER_LONG == 32
316 c->word_offset = c->bit_offset >> 5;
317 c->word_offset &= ~(1UL);
318#else
319# error "unsupported BITS_PER_LONG"
320#endif
321}
322
323#ifndef __packed
324#define __packed __attribute__((packed))
325#endif
326
327/* This is the layout for a packet on the wire.
328 * The byteorder is the network byte order.
329 * (except block_id and barrier fields.
330 * these are pointers to local structs
331 * and have no relevance for the partner,
332 * which just echoes them as received.)
333 *
334 * NOTE that the payload starts at a long aligned offset,
335 * regardless of 32 or 64 bit arch!
336 */
0b70a13d 337struct p_header80 {
b411b363
PR		 338 u32 magic;
339 u16 command;
340 u16 length; /* bytes of data after this header */
341 u8 payload[0];
342} __packed;
0b70a13d
PR		 343
344/* Header for big packets, Used for data packets exceeding 64kB */
345struct p_header95 {
346 u16 magic; /* use DRBD_MAGIC_BIG here */
347 u16 command;
00b42537 348 u32 length; /* Use only 24 bits of that. Ignore the highest 8 bit. */
0b70a13d
PR		 349 u8 payload[0];
350} __packed;
351
352union p_header {
353 struct p_header80 h80;
354 struct p_header95 h95;
355};
b411b363
PR		 356
357/*
358 * short commands, packets without payload, plain p_header:
359 * P_PING
360 * P_PING_ACK
361 * P_BECOME_SYNC_TARGET
362 * P_BECOME_SYNC_SOURCE
363 * P_UNPLUG_REMOTE
364 */
365
366/*
367 * commands with out-of-struct payload:
368 * P_BITMAP (no additional fields)
369 * P_DATA, P_DATA_REPLY (see p_data)
370 * P_COMPRESSED_BITMAP (see receive_compressed_bitmap)
371 */
372
373/* these defines must not be changed without changing the protocol version */
76d2e7ec
PR		 374#define DP_HARDBARRIER 1 /* depricated */
375#define DP_RW_SYNC 2 /* equals REQ_SYNC */
b411b363 376#define DP_MAY_SET_IN_SYNC 4
721a9602 377#define DP_UNPLUG 8 /* not used anymore */
76d2e7ec
PR		 378#define DP_FUA 16 /* equals REQ_FUA */
379#define DP_FLUSH 32 /* equals REQ_FLUSH */
380#define DP_DISCARD 64 /* equals REQ_DISCARD */
b411b363
PR		 381
382struct p_data {
0b70a13d 383 union p_header head;
b411b363
PR		 384 u64 sector; /* 64 bits sector number */
385 u64 block_id; /* to identify the request in protocol B&C */
386 u32 seq_num;
387 u32 dp_flags;
388} __packed;
389
390/*
391 * commands which share a struct:
392 * p_block_ack:
393 * P_RECV_ACK (proto B), P_WRITE_ACK (proto C),
394 * P_DISCARD_ACK (proto C, two-primaries conflict detection)
395 * p_block_req:
396 * P_DATA_REQUEST, P_RS_DATA_REQUEST
397 */
398struct p_block_ack {
0b70a13d 399 struct p_header80 head;
b411b363
PR		 400 u64 sector;
401 u64 block_id;
402 u32 blksize;
403 u32 seq_num;
404} __packed;
405
406
407struct p_block_req {
0b70a13d 408 struct p_header80 head;
b411b363
PR		 409 u64 sector;
410 u64 block_id;
411 u32 blksize;
412 u32 pad; /* to multiple of 8 Byte */
413} __packed;
414
415/*
416 * commands with their own struct for additional fields:
417 * P_HAND_SHAKE
418 * P_BARRIER
419 * P_BARRIER_ACK
420 * P_SYNC_PARAM
421 * ReportParams
422 */
423
424struct p_handshake {
0b70a13d 425 struct p_header80 head; /* 8 bytes */
b411b363
PR		 426 u32 protocol_min;
427 u32 feature_flags;
428 u32 protocol_max;
429
430 /* should be more than enough for future enhancements
431 * for now, feature_flags and the reserverd array shall be zero.
432 */
433
434 u32 _pad;
435 u64 reserverd[7];
436} __packed;
437/* 80 bytes, FIXED for the next century */
438
439struct p_barrier {
0b70a13d 440 struct p_header80 head;
b411b363
PR		 441 u32 barrier; /* barrier number _handle_ only */
442 u32 pad; /* to multiple of 8 Byte */
443} __packed;
444
445struct p_barrier_ack {
0b70a13d 446 struct p_header80 head;
b411b363
PR		 447 u32 barrier;
448 u32 set_size;
449} __packed;
450
451struct p_rs_param {
0b70a13d 452 struct p_header80 head;
b411b363
PR		 453 u32 rate;
454
455 /* Since protocol version 88 and higher. */
456 char verify_alg[0];
457} __packed;
458
459struct p_rs_param_89 {
0b70a13d 460 struct p_header80 head;
b411b363
PR		 461 u32 rate;
462 /* protocol version 89: */
463 char verify_alg[SHARED_SECRET_MAX];
464 char csums_alg[SHARED_SECRET_MAX];
465} __packed;
466
8e26f9cc 467struct p_rs_param_95 {
0b70a13d 468 struct p_header80 head;
8e26f9cc
PR		 469 u32 rate;
470 char verify_alg[SHARED_SECRET_MAX];
471 char csums_alg[SHARED_SECRET_MAX];
472 u32 c_plan_ahead;
473 u32 c_delay_target;
474 u32 c_fill_target;
475 u32 c_max_rate;
476} __packed;
477
cf14c2e9
PR		 478enum drbd_conn_flags {
479 CF_WANT_LOSE = 1,
480 CF_DRY_RUN = 2,
481};
482
b411b363 483struct p_protocol {
0b70a13d 484 struct p_header80 head;
b411b363
PR		 485 u32 protocol;
486 u32 after_sb_0p;
487 u32 after_sb_1p;
488 u32 after_sb_2p;
cf14c2e9 489 u32 conn_flags;
b411b363
PR		 490 u32 two_primaries;
491
492 /* Since protocol version 87 and higher. */
493 char integrity_alg[0];
494
495} __packed;
496
497struct p_uuids {
0b70a13d 498 struct p_header80 head;
b411b363
PR		 499 u64 uuid[UI_EXTENDED_SIZE];
500} __packed;
501
502struct p_rs_uuid {
0b70a13d 503 struct p_header80 head;
b411b363
PR		 504 u64 uuid;
505} __packed;
506
507struct p_sizes {
0b70a13d 508 struct p_header80 head;
b411b363
PR		 509 u64 d_size; /* size of disk */
510 u64 u_size; /* user requested size */
511 u64 c_size; /* current exported size */
1816a2b4 512 u32 max_bio_size; /* Maximal size of a BIO */
e89b591c
PR		 513 u16 queue_order_type; /* not yet implemented in DRBD*/
514 u16 dds_flags; /* use enum dds_flags here. */
b411b363
PR		 515} __packed;
516
517struct p_state {
0b70a13d 518 struct p_header80 head;
b411b363
PR		 519 u32 state;
520} __packed;
521
522struct p_req_state {
0b70a13d 523 struct p_header80 head;
b411b363
PR		 524 u32 mask;
525 u32 val;
526} __packed;
527
528struct p_req_state_reply {
0b70a13d 529 struct p_header80 head;
b411b363
PR		 530 u32 retcode;
531} __packed;
532
533struct p_drbd06_param {
534 u64 size;
535 u32 state;
536 u32 blksize;
537 u32 protocol;
538 u32 version;
539 u32 gen_cnt[5];
540 u32 bit_map_gen[5];
541} __packed;
542
543struct p_discard {
0b70a13d 544 struct p_header80 head;
b411b363
PR		 545 u64 block_id;
546 u32 seq_num;
547 u32 pad;
548} __packed;
549
73a01a18
PR		 550struct p_block_desc {
551 struct p_header80 head;
552 u64 sector;
553 u32 blksize;
554 u32 pad; /* to multiple of 8 Byte */
555} __packed;
556
b411b363
PR		 557/* Valid values for the encoding field.
558 * Bump proto version when changing this. */
559enum drbd_bitmap_code {
560 /* RLE_VLI_Bytes = 0,
561 * and other bit variants had been defined during
562 * algorithm evaluation. */
563 RLE_VLI_Bits = 2,
564};
565
566struct p_compressed_bm {
0b70a13d 567 struct p_header80 head;
b411b363
PR		 568 /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code
569 * (encoding & 0x80): polarity (set/unset) of first runlength
570 * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits
571 * used to pad up to head.length bytes
572 */
573 u8 encoding;
574
575 u8 code[0];
576} __packed;
577
0b70a13d
PR		 578struct p_delay_probe93 {
579 struct p_header80 head;
580 u32 seq_num; /* sequence number to match the two probe packets */
581 u32 offset; /* usecs the probe got sent after the reference time point */
0ced55a3
PR		 582} __packed;
583
b411b363
PR		 584/* DCBP: Drbd Compressed Bitmap Packet ... */
585static inline enum drbd_bitmap_code
586DCBP_get_code(struct p_compressed_bm *p)
587{
588 return (enum drbd_bitmap_code)(p->encoding & 0x0f);
589}
590
591static inline void
592DCBP_set_code(struct p_compressed_bm *p, enum drbd_bitmap_code code)
593{
594 BUG_ON(code & ~0xf);
595 p->encoding = (p->encoding & ~0xf) | code;
596}
597
598static inline int
599DCBP_get_start(struct p_compressed_bm *p)
600{
601 return (p->encoding & 0x80) != 0;
602}
603
604static inline void
605DCBP_set_start(struct p_compressed_bm *p, int set)
606{
607 p->encoding = (p->encoding & ~0x80) | (set ? 0x80 : 0);
608}
609
610static inline int
611DCBP_get_pad_bits(struct p_compressed_bm *p)
612{
613 return (p->encoding >> 4) & 0x7;
614}
615
616static inline void
617DCBP_set_pad_bits(struct p_compressed_bm *p, int n)
618{
619 BUG_ON(n & ~0x7);
620 p->encoding = (p->encoding & (~0x7 << 4)) | (n << 4);
621}
622
623/* one bitmap packet, including the p_header,
624 * should fit within one _architecture independend_ page.
625 * so we need to use the fixed size 4KiB page size
25985edc 626 * most architectures have used for a long time.
b411b363 627 */
0b70a13d 628#define BM_PACKET_PAYLOAD_BYTES (4096 - sizeof(struct p_header80))
b411b363
PR		 629#define BM_PACKET_WORDS (BM_PACKET_PAYLOAD_BYTES/sizeof(long))
630#define BM_PACKET_VLI_BYTES_MAX (4096 - sizeof(struct p_compressed_bm))
631#if (PAGE_SIZE < 4096)
632/* drbd_send_bitmap / receive_bitmap would break horribly */
633#error "PAGE_SIZE too small"
634#endif
635
636union p_polymorph {
02918be2 637 union p_header header;
b411b363
PR		 638 struct p_handshake handshake;
639 struct p_data data;
640 struct p_block_ack block_ack;
641 struct p_barrier barrier;
642 struct p_barrier_ack barrier_ack;
643 struct p_rs_param_89 rs_param_89;
8e26f9cc 644 struct p_rs_param_95 rs_param_95;
b411b363
PR		 645 struct p_protocol protocol;
646 struct p_sizes sizes;
647 struct p_uuids uuids;
648 struct p_state state;
649 struct p_req_state req_state;
650 struct p_req_state_reply req_state_reply;
651 struct p_block_req block_req;
02918be2
PR		 652 struct p_delay_probe93 delay_probe93;
653 struct p_rs_uuid rs_uuid;
73a01a18 654 struct p_block_desc block_desc;
b411b363
PR		 655} __packed;
656
657/**********************************************************************/
658enum drbd_thread_state {
e77a0a5c
AG		 659 NONE,
660 RUNNING,
661 EXITING,
662 RESTARTING
b411b363
PR		 663};
664
665struct drbd_thread {
666 spinlock_t t_lock;
667 struct task_struct *task;
668 struct completion stop;
669 enum drbd_thread_state t_state;
670 int (*function) (struct drbd_thread *);
671 struct drbd_conf *mdev;
672 int reset_cpu_mask;
673};
674
675static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi)
676{
677 /* THINK testing the t_state seems to be uncritical in all cases
678 * (but thread_{start,stop}), so we can read it *without* the lock.
679 * --lge */
680
681 smp_rmb();
682 return thi->t_state;
683}
684
b411b363
PR		 685struct drbd_work;
686typedef int (*drbd_work_cb)(struct drbd_conf *, struct drbd_work *, int cancel);
687struct drbd_work {
688 struct list_head list;
689 drbd_work_cb cb;
690};
691
ace652ac
AG		 692#include "drbd_interval.h"
693
b411b363
PR		 694struct drbd_request {
695 struct drbd_work w;
696 struct drbd_conf *mdev;
697
698 /* if local IO is not allowed, will be NULL.
699 * if local IO _is_ allowed, holds the locally submitted bio clone,
700 * or, after local IO completion, the ERR_PTR(error).
701 * see drbd_endio_pri(). */
702 struct bio *private_bio;
703
ace652ac 704 struct drbd_interval i;
b411b363
PR		 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() */
715 int seq_num;
716 unsigned long start_time;
717};
718
719struct drbd_tl_epoch {
720 struct drbd_work w;
721 struct list_head requests; /* requests before */
722 struct drbd_tl_epoch *next; /* pointer to the next barrier */
723 unsigned int br_number; /* the barriers identifier. */
7e602c0a 724 int n_writes; /* number of requests attached before this barrier */
b411b363
PR		 725};
726
727struct drbd_request;
728
729/* These Tl_epoch_entries may be in one of 6 lists:
730 active_ee .. data packet being written
731 sync_ee .. syncer block being written
732 done_ee .. block written, need to send P_WRITE_ACK
733 read_ee .. [RS]P_DATA_REQUEST being read
734*/
735
736struct drbd_epoch {
737 struct list_head list;
738 unsigned int barrier_nr;
739 atomic_t epoch_size; /* increased on every request added. */
740 atomic_t active; /* increased on every req. added, and dec on every finished. */
741 unsigned long flags;
742};
743
744/* drbd_epoch flag bits */
745enum {
b411b363 746 DE_HAVE_BARRIER_NUMBER,
b411b363
PR		 747};
748
749enum epoch_event {
750 EV_PUT,
751 EV_GOT_BARRIER_NR,
b411b363 752 EV_BECAME_LAST,
b411b363
PR		 753 EV_CLEANUP = 32, /* used as flag */
754};
755
b411b363
PR		 756struct drbd_wq_barrier {
757 struct drbd_work w;
758 struct completion done;
759};
760
761struct digest_info {
762 int digest_size;
763 void *digest;
764};
765
45bb912b
LE		 766struct drbd_epoch_entry {
767 struct drbd_work w;
85719573 768 struct drbd_epoch *epoch; /* for writes */
45bb912b
LE		 769 struct drbd_conf *mdev;
770 struct page *pages;
771 atomic_t pending_bios;
010f6e67 772 struct drbd_interval i;
45bb912b
LE		 773 /* see comments on ee flag bits below */
774 unsigned long flags;
85719573
PR		 775 union {
776 u64 block_id;
777 struct digest_info *digest;
778 };
45bb912b
LE		 779};
780
781/* ee flag bits.
782 * While corresponding bios are in flight, the only modification will be
783 * set_bit WAS_ERROR, which has to be atomic.
784 * If no bios are in flight yet, or all have been completed,
785 * non-atomic modification to ee->flags is ok.
786 */
b411b363
PR		 787enum {
788 __EE_CALL_AL_COMPLETE_IO,
b411b363 789 __EE_MAY_SET_IN_SYNC,
45bb912b 790
45bb912b
LE		 791 /* In case a barrier failed,
792 * we need to resubmit without the barrier flag. */
793 __EE_RESUBMITTED,
794
795 /* we may have several bios per epoch entry.
796 * if any of those fail, we set this flag atomically
797 * from the endio callback */
798 __EE_WAS_ERROR,
c36c3ced
LE		 799
800 /* This ee has a pointer to a digest instead of a block id */
801 __EE_HAS_DIGEST,
b411b363
PR		 802};
803#define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO)
b411b363 804#define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC)
45bb912b
LE		 805#define EE_RESUBMITTED (1<<__EE_RESUBMITTED)
806#define EE_WAS_ERROR (1<<__EE_WAS_ERROR)
c36c3ced 807#define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST)
b411b363
PR		 808
809/* global flag bits */
810enum {
25985edc 811 CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */
b411b363
PR		 812 SIGNAL_ASENDER, /* whether asender wants to be interrupted */
813 SEND_PING, /* whether asender should send a ping asap */
814
b411b363
PR		 815 UNPLUG_QUEUED, /* only relevant with kernel 2.4 */
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. */
19f843aa 826 NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */
b411b363
PR		 827 CONSIDER_RESYNC,
828
a8a4e51e 829 MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */
b411b363
PR		 830 SUSPEND_IO, /* suspend application io */
831 BITMAP_IO, /* suspend application io;
832 once no more io in flight, start bitmap io */
833 BITMAP_IO_QUEUED, /* Started bitmap IO */
82f59cc6
LE		 834 GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */
835 WAS_IO_ERROR, /* Local disk failed returned IO error */
b411b363
PR		 836 RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */
837 NET_CONGESTED, /* The data socket is congested */
838
839 CONFIG_PENDING, /* serialization of (re)configuration requests.
840 * if set, also prevents the device from dying */
841 DEVICE_DYING, /* device became unconfigured,
842 * but worker thread is still handling the cleanup.
843 * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed,
844 * while this is set. */
845 RESIZE_PENDING, /* Size change detected locally, waiting for the response from
846 * the peer, if it changed there as well. */
cf14c2e9 847 CONN_DRY_RUN, /* Expect disconnect after resync handshake. */
309d1608 848 GOT_PING_ACK, /* set when we receive a ping_ack packet, misc wait gets woken */
43a5182c 849 NEW_CUR_UUID, /* Create new current UUID when thawing IO */
0778286a 850 AL_SUSPENDED, /* Activity logging is currently suspended. */
370a43e7 851 AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */
b411b363
PR		 852};
853
854struct drbd_bitmap; /* opaque for drbd_conf */
855
20ceb2b2
LE		 856/* definition of bits in bm_flags to be used in drbd_bm_lock
857 * and drbd_bitmap_io and friends. */
858enum bm_flag {
859 /* do we need to kfree, or vfree bm_pages? */
860 BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */
861
862 /* currently locked for bulk operation */
863 BM_LOCKED_MASK = 0x7,
864
865 /* in detail, that is: */
866 BM_DONT_CLEAR = 0x1,
867 BM_DONT_SET = 0x2,
868 BM_DONT_TEST = 0x4,
869
870 /* (test bit, count bit) allowed (common case) */
871 BM_LOCKED_TEST_ALLOWED = 0x3,
872
873 /* testing bits, as well as setting new bits allowed, but clearing bits
874 * would be unexpected. Used during bitmap receive. Setting new bits
875 * requires sending of "out-of-sync" information, though. */
876 BM_LOCKED_SET_ALLOWED = 0x1,
877
878 /* clear is not expected while bitmap is locked for bulk operation */
879};
880
881
b411b363
PR		 882/* TODO sort members for performance
883 * MAYBE group them further */
884
885/* THINK maybe we actually want to use the default "event/%s" worker threads
886 * or similar in linux 2.6, which uses per cpu data and threads.
b411b363
PR		 887 */
888struct drbd_work_queue {
889 struct list_head q;
890 struct semaphore s; /* producers up it, worker down()s it */
891 spinlock_t q_lock; /* to protect the list. */
892};
893
894struct drbd_socket {
895 struct drbd_work_queue work;
896 struct mutex mutex;
897 struct socket *socket;
898 /* this way we get our
899 * send/receive buffers off the stack */
900 union p_polymorph sbuf;
901 union p_polymorph rbuf;
902};
903
904struct drbd_md {
905 u64 md_offset; /* sector offset to 'super' block */
906
907 u64 la_size_sect; /* last agreed size, unit sectors */
908 u64 uuid[UI_SIZE];
909 u64 device_uuid;
910 u32 flags;
911 u32 md_size_sect;
912
913 s32 al_offset; /* signed relative sector offset to al area */
914 s32 bm_offset; /* signed relative sector offset to bitmap */
915
916 /* u32 al_nr_extents; important for restoring the AL
917 * is stored into sync_conf.al_extents, which in turn
918 * gets applied to act_log->nr_elements
919 */
920};
921
922/* for sync_conf and other types... */
923#define NL_PACKET(name, number, fields) struct name { fields };
924#define NL_INTEGER(pn,pr,member) int member;
925#define NL_INT64(pn,pr,member) __u64 member;
926#define NL_BIT(pn,pr,member) unsigned member:1;
927#define NL_STRING(pn,pr,member,len) unsigned char member[len]; int member ## _len;
928#include "linux/drbd_nl.h"
929
930struct drbd_backing_dev {
931 struct block_device *backing_bdev;
932 struct block_device *md_bdev;
b411b363
PR		 933 struct drbd_md md;
934 struct disk_conf dc; /* The user provided config... */
935 sector_t known_size; /* last known size of that backing device */
936};
937
938struct drbd_md_io {
939 struct drbd_conf *mdev;
940 struct completion event;
941 int error;
942};
943
944struct bm_io_work {
945 struct drbd_work w;
946 char *why;
20ceb2b2 947 enum bm_flag flags;
b411b363
PR		 948 int (*io_fn)(struct drbd_conf *mdev);
949 void (*done)(struct drbd_conf *mdev, int rv);
950};
951
952enum write_ordering_e {
953 WO_none,
954 WO_drain_io,
955 WO_bdev_flush,
b411b363
PR		 956};
957
778f271d
PR		 958struct fifo_buffer {
959 int *values;
960 unsigned int head_index;
961 unsigned int size;
962};
963
2111438b
PR		 964struct drbd_tconn { /* is a resource from the config file */
965 char *name; /* Resource name */
966 struct list_head all_tconn; /* List of all drbd_tconn, prot by global_state_lock */
967 struct drbd_conf *volume0; /* TODO: Remove me again */
968
969 struct net_conf *net_conf; /* protected by get_net_conf() and put_net_conf() */
b2fb6dbe
PR		 970 atomic_t net_cnt; /* Users of net_conf */
971 wait_queue_head_t net_cnt_wait;
e42325a5
PR		 972
973 struct drbd_socket data; /* data/barrier/cstate/parameter packets */
974 struct drbd_socket meta; /* ping/ack (metadata) packets */
e6b3ea83
PR		 975
976 struct drbd_thread receiver;
977 struct drbd_thread worker;
978 struct drbd_thread asender;
2111438b
PR		 979};
980
b411b363 981struct drbd_conf {
2111438b
PR		 982 struct drbd_tconn *tconn;
983 int vnr; /* volume number within the connection */
984
b411b363
PR		 985 /* things that are stored as / read from meta data on disk */
986 unsigned long flags;
987
988 /* configured by drbdsetup */
b411b363
PR		 989 struct syncer_conf sync_conf;
990 struct drbd_backing_dev *ldev __protected_by(local);
991
992 sector_t p_size; /* partner's disk size */
993 struct request_queue *rq_queue;
994 struct block_device *this_bdev;
995 struct gendisk *vdisk;
996
b411b363
PR		 997 int agreed_pro_version; /* actually used protocol version */
998 unsigned long last_received; /* in jiffies, either socket */
999 unsigned int ko_count;
1000 struct drbd_work resync_work,
1001 unplug_work,
e9e6f3ec 1002 go_diskless,
c4752ef1
PR		 1003 md_sync_work,
1004 start_resync_work;
b411b363
PR		 1005 struct timer_list resync_timer;
1006 struct timer_list md_sync_timer;
370a43e7 1007 struct timer_list start_resync_timer;
7fde2be9 1008 struct timer_list request_timer;
ee15b038
LE		 1009#ifdef DRBD_DEBUG_MD_SYNC
1010 struct {
1011 unsigned int line;
1012 const char* func;
1013 } last_md_mark_dirty;
1014#endif
b411b363
PR		 1015
1016 /* Used after attach while negotiating new disk state. */
1017 union drbd_state new_state_tmp;
1018
1019 union drbd_state state;
1020 wait_queue_head_t misc_wait;
1021 wait_queue_head_t state_wait; /* upon each state change. */
1022 unsigned int send_cnt;
1023 unsigned int recv_cnt;
1024 unsigned int read_cnt;
1025 unsigned int writ_cnt;
1026 unsigned int al_writ_cnt;
1027 unsigned int bm_writ_cnt;
1028 atomic_t ap_bio_cnt; /* Requests we need to complete */
1029 atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */
1030 atomic_t rs_pending_cnt; /* RS request/data packets on the wire */
1031 atomic_t unacked_cnt; /* Need to send replys for */
1032 atomic_t local_cnt; /* Waiting for local completion */
b2fb6dbe 1033
b411b363
PR		 1034 spinlock_t req_lock;
1035 struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */
1036 struct drbd_tl_epoch *newest_tle;
1037 struct drbd_tl_epoch *oldest_tle;
1038 struct list_head out_of_sequence_requests;
b411b363 1039
dac1389c
AG		 1040 /* Interval tree of pending local requests */
1041 struct rb_root read_requests;
de696716
AG		 1042 struct rb_root write_requests;
1043
4b0715f0 1044 /* blocks to resync in this run [unit BM_BLOCK_SIZE] */
b411b363 1045 unsigned long rs_total;
4b0715f0 1046 /* number of resync blocks that failed in this run */
b411b363
PR		 1047 unsigned long rs_failed;
1048 /* Syncer's start time [unit jiffies] */
1049 unsigned long rs_start;
1050 /* cumulated time in PausedSyncX state [unit jiffies] */
1051 unsigned long rs_paused;
1d7734a0
LE		 1052 /* skipped because csum was equal [unit BM_BLOCK_SIZE] */
1053 unsigned long rs_same_csum;
1054#define DRBD_SYNC_MARKS 8
1055#define DRBD_SYNC_MARK_STEP (3*HZ)
b411b363 1056 /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */
1d7734a0 1057 unsigned long rs_mark_left[DRBD_SYNC_MARKS];
b411b363 1058 /* marks's time [unit jiffies] */
1d7734a0
LE		 1059 unsigned long rs_mark_time[DRBD_SYNC_MARKS];
1060 /* current index into rs_mark_{left,time} */
1061 int rs_last_mark;
b411b363
PR		 1062
1063 /* where does the admin want us to start? (sector) */
1064 sector_t ov_start_sector;
1065 /* where are we now? (sector) */
1066 sector_t ov_position;
1067 /* Start sector of out of sync range (to merge printk reporting). */
1068 sector_t ov_last_oos_start;
1069 /* size of out-of-sync range in sectors. */
1070 sector_t ov_last_oos_size;
1071 unsigned long ov_left; /* in bits */
1072 struct crypto_hash *csums_tfm;
1073 struct crypto_hash *verify_tfm;
1074
b411b363
PR		 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 1095 struct list_head net_ee; /* zero-copy network send in progress */
b411b363 1096
8b946255
AG		 1097 /* Interval tree of pending remote write requests (struct drbd_epoch_entry) */
1098 struct rb_root epoch_entries;
1099
b411b363
PR		 1100 /* this one is protected by ee_lock, single thread */
1101 struct drbd_epoch_entry *last_write_w_barrier;
1102
1103 int next_barrier_nr;
b411b363 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 */
1110 struct mutex md_io_mutex; /* protects the md_io_buffer */
1111 spinlock_t al_lock;
1112 wait_queue_head_t al_wait;
1113 struct lru_cache *act_log; /* activity log */
1114 unsigned int al_tr_number;
1115 int al_tr_cycle;
1116 int al_tr_pos; /* position of the next transaction in the journal */
1117 struct crypto_hash *cram_hmac_tfm;
1118 struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */
1119 struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */
1120 void *int_dig_out;
1121 void *int_dig_in;
1122 void *int_dig_vv;
1123 wait_queue_head_t seq_wait;
1124 atomic_t packet_seq;
1125 unsigned int peer_seq;
1126 spinlock_t peer_seq_lock;
1127 unsigned int minor;
1128 unsigned long comm_bm_set; /* communicated number of set bits. */
1129 cpumask_var_t cpu_mask;
1130 struct bm_io_work bm_io_work;
1131 u64 ed_uuid; /* UUID of the exposed data */
1132 struct mutex state_mutex;
1133 char congestion_reason; /* Why we where congested... */
1d7734a0
LE		 1134 atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */
1135 atomic_t rs_sect_ev; /* for submitted resync data rate, both */
1136 int rs_last_sect_ev; /* counter to compare with */
1137 int rs_last_events; /* counter of read or write "events" (unit sectors)
1138 * on the lower level device when we last looked. */
1139 int c_sync_rate; /* current resync rate after syncer throttle magic */
778f271d
PR		 1140 struct fifo_buffer rs_plan_s; /* correction values of resync planer */
1141 int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */
25985edc 1142 int rs_planed; /* resync sectors already planned */
759fbdfb 1143 atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */
99432fcc
PR		 1144 int peer_max_bio_size;
1145 int local_max_bio_size;
b411b363
PR		 1146};
1147
1148static inline struct drbd_conf *minor_to_mdev(unsigned int minor)
1149{
1150 struct drbd_conf *mdev;
1151
1152 mdev = minor < minor_count ? minor_table[minor] : NULL;
1153
1154 return mdev;
1155}
1156
1157static inline unsigned int mdev_to_minor(struct drbd_conf *mdev)
1158{
1159 return mdev->minor;
1160}
1161
25985edc 1162/* returns 1 if it was successful,
b411b363
PR		 1163 * returns 0 if there was no data socket.
1164 * so wherever you are going to use the data.socket, e.g. do
1165 * if (!drbd_get_data_sock(mdev))
1166 * return 0;
1167 * CODE();
1168 * drbd_put_data_sock(mdev);
1169 */
1170static inline int drbd_get_data_sock(struct drbd_conf *mdev)
1171{
e42325a5 1172 mutex_lock(&mdev->tconn->data.mutex);
b411b363
PR		 1173 /* drbd_disconnect() could have called drbd_free_sock()
1174 * while we were waiting in down()... */
e42325a5
PR		 1175 if (unlikely(mdev->tconn->data.socket == NULL)) {
1176 mutex_unlock(&mdev->tconn->data.mutex);
b411b363
PR		 1177 return 0;
1178 }
1179 return 1;
1180}
1181
1182static inline void drbd_put_data_sock(struct drbd_conf *mdev)
1183{
e42325a5 1184 mutex_unlock(&mdev->tconn->data.mutex);
b411b363
PR		 1185}
1186
1187/*
1188 * function declarations
1189 *************************/
1190
1191/* drbd_main.c */
1192
1193enum chg_state_flags {
1194 CS_HARD = 1,
1195 CS_VERBOSE = 2,
1196 CS_WAIT_COMPLETE = 4,
1197 CS_SERIALIZE = 8,
1198 CS_ORDERED = CS_WAIT_COMPLETE + CS_SERIALIZE,
1199};
1200
e89b591c
PR		 1201enum dds_flags {
1202 DDSF_FORCED = 1,
1203 DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */
1204};
1205
b411b363 1206extern void drbd_init_set_defaults(struct drbd_conf *mdev);
bf885f8a
AG		 1207extern enum drbd_state_rv drbd_change_state(struct drbd_conf *mdev,
1208 enum chg_state_flags f,
1209 union drbd_state mask,
1210 union drbd_state val);
b411b363
PR		 1211extern void drbd_force_state(struct drbd_conf *, union drbd_state,
1212 union drbd_state);
bf885f8a
AG		 1213extern enum drbd_state_rv _drbd_request_state(struct drbd_conf *,
1214 union drbd_state,
1215 union drbd_state,
1216 enum chg_state_flags);
1217extern enum drbd_state_rv __drbd_set_state(struct drbd_conf *, union drbd_state,
1218 enum chg_state_flags,
1219 struct completion *done);
b411b363
PR		 1220extern void print_st_err(struct drbd_conf *, union drbd_state,
1221 union drbd_state, int);
1222extern int drbd_thread_start(struct drbd_thread *thi);
1223extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait);
1224#ifdef CONFIG_SMP
1225extern void drbd_thread_current_set_cpu(struct drbd_conf *mdev);
1226extern void drbd_calc_cpu_mask(struct drbd_conf *mdev);
1227#else
1228#define drbd_thread_current_set_cpu(A) ({})
1229#define drbd_calc_cpu_mask(A) ({})
1230#endif
1231extern void drbd_free_resources(struct drbd_conf *mdev);
1232extern void tl_release(struct drbd_conf *mdev, unsigned int barrier_nr,
1233 unsigned int set_size);
1234extern void tl_clear(struct drbd_conf *mdev);
1235extern void _tl_add_barrier(struct drbd_conf *, struct drbd_tl_epoch *);
1236extern void drbd_free_sock(struct drbd_conf *mdev);
1237extern int drbd_send(struct drbd_conf *mdev, struct socket *sock,
1238 void *buf, size_t size, unsigned msg_flags);
1239extern int drbd_send_protocol(struct drbd_conf *mdev);
1240extern int drbd_send_uuids(struct drbd_conf *mdev);
1241extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev);
5a22db89 1242extern int drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev);
e89b591c 1243extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags);
b411b363
PR		 1244extern int _drbd_send_state(struct drbd_conf *mdev);
1245extern int drbd_send_state(struct drbd_conf *mdev);
1246extern int _drbd_send_cmd(struct drbd_conf *mdev, struct socket *sock,
0b70a13d 1247 enum drbd_packets cmd, struct p_header80 *h,
b411b363
PR		 1248 size_t size, unsigned msg_flags);
1249#define USE_DATA_SOCKET 1
1250#define USE_META_SOCKET 0
1251extern int drbd_send_cmd(struct drbd_conf *mdev, int use_data_socket,
0b70a13d 1252 enum drbd_packets cmd, struct p_header80 *h,
b411b363
PR		 1253 size_t size);
1254extern int drbd_send_cmd2(struct drbd_conf *mdev, enum drbd_packets cmd,
1255 char *data, size_t size);
1256extern int drbd_send_sync_param(struct drbd_conf *mdev, struct syncer_conf *sc);
1257extern int drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr,
1258 u32 set_size);
1259extern int drbd_send_ack(struct drbd_conf *mdev, enum drbd_packets cmd,
1260 struct drbd_epoch_entry *e);
1261extern int drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packets cmd,
1262 struct p_block_req *rp);
1263extern int drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packets cmd,
2b2bf214 1264 struct p_data *dp, int data_size);
b411b363
PR		 1265extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packets cmd,
1266 sector_t sector, int blksize, u64 block_id);
73a01a18 1267extern int drbd_send_oos(struct drbd_conf *mdev, struct drbd_request *req);
b411b363
PR		 1268extern int drbd_send_block(struct drbd_conf *mdev, enum drbd_packets cmd,
1269 struct drbd_epoch_entry *e);
1270extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req);
b411b363
PR		 1271extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd,
1272 sector_t sector, int size, u64 block_id);
1273extern int drbd_send_drequest_csum(struct drbd_conf *mdev,
1274 sector_t sector,int size,
1275 void *digest, int digest_size,
1276 enum drbd_packets cmd);
1277extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size);
1278
1279extern int drbd_send_bitmap(struct drbd_conf *mdev);
1280extern int _drbd_send_bitmap(struct drbd_conf *mdev);
bf885f8a 1281extern int drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode);
b411b363
PR		 1282extern void drbd_free_bc(struct drbd_backing_dev *ldev);
1283extern void drbd_mdev_cleanup(struct drbd_conf *mdev);
62b0da3a 1284void drbd_print_uuids(struct drbd_conf *mdev, const char *text);
b411b363 1285
b411b363
PR		 1286extern void drbd_md_sync(struct drbd_conf *mdev);
1287extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev);
b411b363
PR		 1288extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1289extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local);
1290extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1291extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local);
1292extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local);
1293extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local);
1294extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local);
1295extern int drbd_md_test_flag(struct drbd_backing_dev *, int);
ee15b038 1296#ifndef DRBD_DEBUG_MD_SYNC
b411b363 1297extern void drbd_md_mark_dirty(struct drbd_conf *mdev);
ee15b038
LE		 1298#else
1299#define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ )
1300extern void drbd_md_mark_dirty_(struct drbd_conf *mdev,
1301 unsigned int line, const char *func);
1302#endif
b411b363
PR		 1303extern void drbd_queue_bitmap_io(struct drbd_conf *mdev,
1304 int (*io_fn)(struct drbd_conf *),
1305 void (*done)(struct drbd_conf *, int),
20ceb2b2
LE		 1306 char *why, enum bm_flag flags);
1307extern int drbd_bitmap_io(struct drbd_conf *mdev,
1308 int (*io_fn)(struct drbd_conf *),
1309 char *why, enum bm_flag flags);
b411b363
PR		 1310extern int drbd_bmio_set_n_write(struct drbd_conf *mdev);
1311extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev);
e9e6f3ec 1312extern void drbd_go_diskless(struct drbd_conf *mdev);
82f59cc6 1313extern void drbd_ldev_destroy(struct drbd_conf *mdev);
b411b363
PR		 1314
1315
1316/* Meta data layout
1317 We reserve a 128MB Block (4k aligned)
1318 * either at the end of the backing device
3ad2f3fb 1319 * or on a separate meta data device. */
b411b363
PR		 1320
1321#define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */
1322/* The following numbers are sectors */
1323#define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */
1324#define MD_AL_MAX_SIZE 64 /* = 32 kb LOG ~ 3776 extents ~ 14 GB Storage */
1325/* Allows up to about 3.8TB */
1326#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
1327
1328/* Since the smalles IO unit is usually 512 byte */
1329#define MD_SECTOR_SHIFT 9
1330#define MD_SECTOR_SIZE (1<<MD_SECTOR_SHIFT)
1331
1332/* activity log */
1333#define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
1334#define AL_EXTENT_SHIFT 22 /* One extent represents 4M Storage */
1335#define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
1336
1337#if BITS_PER_LONG == 32
1338#define LN2_BPL 5
1339#define cpu_to_lel(A) cpu_to_le32(A)
1340#define lel_to_cpu(A) le32_to_cpu(A)
1341#elif BITS_PER_LONG == 64
1342#define LN2_BPL 6
1343#define cpu_to_lel(A) cpu_to_le64(A)
1344#define lel_to_cpu(A) le64_to_cpu(A)
1345#else
1346#error "LN2 of BITS_PER_LONG unknown!"
1347#endif
1348
1349/* resync bitmap */
1350/* 16MB sized 'bitmap extent' to track syncer usage */
1351struct bm_extent {
1352 int rs_left; /* number of bits set (out of sync) in this extent. */
1353 int rs_failed; /* number of failed resync requests in this extent. */
1354 unsigned long flags;
1355 struct lc_element lce;
1356};
1357
1358#define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */
1359#define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */
e3555d85 1360#define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */
b411b363
PR		 1361
1362/* drbd_bitmap.c */
1363/*
1364 * We need to store one bit for a block.
1365 * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap.
1366 * Bit 0 ==> local node thinks this block is binary identical on both nodes
1367 * Bit 1 ==> local node thinks this block needs to be synced.
1368 */
1369
8e26f9cc
PR		 1370#define SLEEP_TIME (HZ/10)
1371
b411b363
PR		 1372#define BM_BLOCK_SHIFT 12 /* 4k per bit */
1373#define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT)
1374/* (9+3) : 512 bytes @ 8 bits; representing 16M storage
1375 * per sector of on disk bitmap */
1376#define BM_EXT_SHIFT (BM_BLOCK_SHIFT + MD_SECTOR_SHIFT + 3) /* = 24 */
1377#define BM_EXT_SIZE (1<<BM_EXT_SHIFT)
1378
1379#if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12)
1380#error "HAVE YOU FIXED drbdmeta AS WELL??"
1381#endif
1382
1383/* thus many _storage_ sectors are described by one bit */
1384#define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9))
1385#define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9))
1386#define BM_SECT_PER_BIT BM_BIT_TO_SECT(1)
1387
1388/* bit to represented kilo byte conversion */
1389#define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10))
1390
1391/* in which _bitmap_ extent (resp. sector) the bit for a certain
1392 * _storage_ sector is located in */
1393#define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9))
1394
1395/* how much _storage_ sectors we have per bitmap sector */
1396#define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9))
1397#define BM_SECT_PER_EXT BM_EXT_TO_SECT(1)
1398
1399/* in one sector of the bitmap, we have this many activity_log extents. */
1400#define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT))
1401#define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL))
1402
1403#define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT)
1404#define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1)
1405
1406/* the extent in "PER_EXTENT" below is an activity log extent
1407 * we need that many (long words/bytes) to store the bitmap
1408 * of one AL_EXTENT_SIZE chunk of storage.
1409 * we can store the bitmap for that many AL_EXTENTS within
1410 * one sector of the _on_disk_ bitmap:
1411 * bit 0 bit 37 bit 38 bit (512*8)-1
1412 * ...|........|........|.. // ..|........|
1413 * sect. 0 `296 `304 ^(512*8*8)-1
1414 *
1415#define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG )
1416#define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128
1417#define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4
1418 */
1419
1420#define DRBD_MAX_SECTORS_32 (0xffffffffLU)
1421#define DRBD_MAX_SECTORS_BM \
1422 ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9)))
1423#if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32
1424#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1425#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM
36bfc7e2 1426#elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32
b411b363
PR		 1427#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32
1428#define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32
1429#else
1430#define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM
1431/* 16 TB in units of sectors */
1432#if BITS_PER_LONG == 32
1433/* adjust by one page worth of bitmap,
1434 * so we won't wrap around in drbd_bm_find_next_bit.
1435 * you should use 64bit OS for that much storage, anyways. */
1436#define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff)
1437#else
4b0715f0
LE		 1438/* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */
1439#define DRBD_MAX_SECTORS_FLEX (1UL << 51)
1440/* corresponds to (1UL << 38) bits right now. */
b411b363
PR		 1441#endif
1442#endif
1443
d5373389 1444#define HT_SHIFT 8
1816a2b4 1445#define DRBD_MAX_BIO_SIZE (1U<<(9+HT_SHIFT))
99432fcc 1446#define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */
b411b363 1447
d5373389 1448#define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */
b411b363
PR		 1449
1450extern int drbd_bm_init(struct drbd_conf *mdev);
02d9a94b 1451extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits);
b411b363
PR		 1452extern void drbd_bm_cleanup(struct drbd_conf *mdev);
1453extern void drbd_bm_set_all(struct drbd_conf *mdev);
1454extern void drbd_bm_clear_all(struct drbd_conf *mdev);
4b0715f0 1455/* set/clear/test only a few bits at a time */
b411b363
PR		 1456extern int drbd_bm_set_bits(
1457 struct drbd_conf *mdev, unsigned long s, unsigned long e);
1458extern int drbd_bm_clear_bits(
1459 struct drbd_conf *mdev, unsigned long s, unsigned long e);
4b0715f0
LE		 1460extern int drbd_bm_count_bits(
1461 struct drbd_conf *mdev, const unsigned long s, const unsigned long e);
1462/* bm_set_bits variant for use while holding drbd_bm_lock,
1463 * may process the whole bitmap in one go */
b411b363
PR		 1464extern void _drbd_bm_set_bits(struct drbd_conf *mdev,
1465 const unsigned long s, const unsigned long e);
1466extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr);
1467extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr);
19f843aa 1468extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local);
b411b363
PR		 1469extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local);
1470extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local);
1471extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev,
1472 unsigned long al_enr);
1473extern size_t drbd_bm_words(struct drbd_conf *mdev);
1474extern unsigned long drbd_bm_bits(struct drbd_conf *mdev);
1475extern sector_t drbd_bm_capacity(struct drbd_conf *mdev);
4b0715f0
LE		 1476
1477#define DRBD_END_OF_BITMAP (~(unsigned long)0)
b411b363
PR		 1478extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1479/* bm_find_next variants for use while you hold drbd_bm_lock() */
1480extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo);
1481extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo);
0778286a 1482extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev);
b411b363
PR		 1483extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev);
1484extern int drbd_bm_rs_done(struct drbd_conf *mdev);
1485/* for receive_bitmap */
1486extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset,
1487 size_t number, unsigned long *buffer);
19f843aa 1488/* for _drbd_send_bitmap */
b411b363
PR		 1489extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset,
1490 size_t number, unsigned long *buffer);
1491
20ceb2b2 1492extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags);
b411b363 1493extern void drbd_bm_unlock(struct drbd_conf *mdev);
b411b363
PR		 1494/* drbd_main.c */
1495
1496extern struct kmem_cache *drbd_request_cache;
1497extern struct kmem_cache *drbd_ee_cache; /* epoch entries */
1498extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */
1499extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */
1500extern mempool_t *drbd_request_mempool;
1501extern mempool_t *drbd_ee_mempool;
1502
1503extern struct page *drbd_pp_pool; /* drbd's page pool */
1504extern spinlock_t drbd_pp_lock;
1505extern int drbd_pp_vacant;
1506extern wait_queue_head_t drbd_pp_wait;
1507
1508extern rwlock_t global_state_lock;
1509
1510extern struct drbd_conf *drbd_new_device(unsigned int minor);
1511extern void drbd_free_mdev(struct drbd_conf *mdev);
1512
2111438b
PR		 1513struct drbd_tconn *drbd_new_tconn(char *name);
1514extern void drbd_free_tconn(struct drbd_tconn *tconn);
1515
b411b363
PR		 1516extern int proc_details;
1517
1518/* drbd_req */
2f58dcfc 1519extern int drbd_make_request(struct request_queue *q, struct bio *bio);
b411b363
PR		 1520extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req);
1521extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec);
1522extern int is_valid_ar_handle(struct drbd_request *, sector_t);
1523
1524
1525/* drbd_nl.c */
1526extern void drbd_suspend_io(struct drbd_conf *mdev);
1527extern void drbd_resume_io(struct drbd_conf *mdev);
1528extern char *ppsize(char *buf, unsigned long long size);
a393db6f 1529extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int);
b411b363 1530enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 };
24c4830c 1531extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local);
b411b363 1532extern void resync_after_online_grow(struct drbd_conf *);
99432fcc 1533extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev);
bf885f8a
AG		 1534extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev,
1535 enum drbd_role new_role,
1536 int force);
87f7be4c
PR		 1537extern enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev);
1538extern void drbd_try_outdate_peer_async(struct drbd_conf *mdev);
b411b363
PR		 1539extern int drbd_khelper(struct drbd_conf *mdev, char *cmd);
1540
1541/* drbd_worker.c */
1542extern int drbd_worker(struct drbd_thread *thi);
1543extern int drbd_alter_sa(struct drbd_conf *mdev, int na);
1544extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side);
1545extern void resume_next_sg(struct drbd_conf *mdev);
1546extern void suspend_other_sg(struct drbd_conf *mdev);
1547extern int drbd_resync_finished(struct drbd_conf *mdev);
1548/* maybe rather drbd_main.c ? */
1549extern int drbd_md_sync_page_io(struct drbd_conf *mdev,
1550 struct drbd_backing_dev *bdev, sector_t sector, int rw);
1551extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int);
9bd28d3c 1552extern void drbd_rs_controller_reset(struct drbd_conf *mdev);
b411b363
PR		 1553
1554static inline void ov_oos_print(struct drbd_conf *mdev)
1555{
1556 if (mdev->ov_last_oos_size) {
1557 dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n",
1558 (unsigned long long)mdev->ov_last_oos_start,
1559 (unsigned long)mdev->ov_last_oos_size);
1560 }
1561 mdev->ov_last_oos_size=0;
1562}
1563
1564
45bb912b
LE		 1565extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *);
1566extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, struct drbd_epoch_entry *, void *);
b411b363
PR		 1567/* worker callbacks */
1568extern int w_req_cancel_conflict(struct drbd_conf *, struct drbd_work *, int);
1569extern int w_read_retry_remote(struct drbd_conf *, struct drbd_work *, int);
1570extern int w_e_end_data_req(struct drbd_conf *, struct drbd_work *, int);
1571extern int w_e_end_rsdata_req(struct drbd_conf *, struct drbd_work *, int);
1572extern int w_e_end_csum_rs_req(struct drbd_conf *, struct drbd_work *, int);
1573extern int w_e_end_ov_reply(struct drbd_conf *, struct drbd_work *, int);
1574extern int w_e_end_ov_req(struct drbd_conf *, struct drbd_work *, int);
1575extern int w_ov_finished(struct drbd_conf *, struct drbd_work *, int);
794abb75 1576extern int w_resync_timer(struct drbd_conf *, struct drbd_work *, int);
b411b363 1577extern int w_resume_next_sg(struct drbd_conf *, struct drbd_work *, int);
b411b363 1578extern int w_send_write_hint(struct drbd_conf *, struct drbd_work *, int);
b411b363
PR		 1579extern int w_send_dblock(struct drbd_conf *, struct drbd_work *, int);
1580extern int w_send_barrier(struct drbd_conf *, struct drbd_work *, int);
1581extern int w_send_read_req(struct drbd_conf *, struct drbd_work *, int);
1582extern int w_prev_work_done(struct drbd_conf *, struct drbd_work *, int);
1583extern int w_e_reissue(struct drbd_conf *, struct drbd_work *, int);
265be2d0 1584extern int w_restart_disk_io(struct drbd_conf *, struct drbd_work *, int);
73a01a18 1585extern int w_send_oos(struct drbd_conf *, struct drbd_work *, int);
c4752ef1 1586extern int w_start_resync(struct drbd_conf *, struct drbd_work *, int);
b411b363
PR		 1587
1588extern void resync_timer_fn(unsigned long data);
370a43e7 1589extern void start_resync_timer_fn(unsigned long data);
b411b363
PR		 1590
1591/* drbd_receiver.c */
e3555d85 1592extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector);
45bb912b
LE		 1593extern int drbd_submit_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1594 const unsigned rw, const int fault_type);
b411b363
PR		 1595extern int drbd_release_ee(struct drbd_conf *mdev, struct list_head *list);
1596extern struct drbd_epoch_entry *drbd_alloc_ee(struct drbd_conf *mdev,
1597 u64 id,
1598 sector_t sector,
1599 unsigned int data_size,
1600 gfp_t gfp_mask) __must_hold(local);
435f0740
LE		 1601extern void drbd_free_some_ee(struct drbd_conf *mdev, struct drbd_epoch_entry *e,
1602 int is_net);
1603#define drbd_free_ee(m,e) drbd_free_some_ee(m, e, 0)
1604#define drbd_free_net_ee(m,e) drbd_free_some_ee(m, e, 1)
b411b363
PR		 1605extern void drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1606 struct list_head *head);
1607extern void _drbd_wait_ee_list_empty(struct drbd_conf *mdev,
1608 struct list_head *head);
1609extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled);
1610extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed);
1611extern void drbd_flush_workqueue(struct drbd_conf *mdev);
1612
1613/* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to
1614 * mess with get_fs/set_fs, we know we are KERNEL_DS always. */
1615static inline int drbd_setsockopt(struct socket *sock, int level, int optname,
1616 char __user *optval, int optlen)
1617{
1618 int err;
1619 if (level == SOL_SOCKET)
1620 err = sock_setsockopt(sock, level, optname, optval, optlen);
1621 else
1622 err = sock->ops->setsockopt(sock, level, optname, optval,
1623 optlen);
1624 return err;
1625}
1626
1627static inline void drbd_tcp_cork(struct socket *sock)
1628{
1629 int __user val = 1;
1630 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1631 (char __user *)&val, sizeof(val));
1632}
1633
1634static inline void drbd_tcp_uncork(struct socket *sock)
1635{
1636 int __user val = 0;
1637 (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK,
1638 (char __user *)&val, sizeof(val));
1639}
1640
1641static inline void drbd_tcp_nodelay(struct socket *sock)
1642{
1643 int __user val = 1;
1644 (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY,
1645 (char __user *)&val, sizeof(val));
1646}
1647
1648static inline void drbd_tcp_quickack(struct socket *sock)
1649{
344fa462 1650 int __user val = 2;
b411b363
PR		 1651 (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK,
1652 (char __user *)&val, sizeof(val));
1653}
1654
1655void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo);
1656
1657/* drbd_proc.c */
1658extern struct proc_dir_entry *drbd_proc;
7d4e9d09 1659extern const struct file_operations drbd_proc_fops;
b411b363
PR		 1660extern const char *drbd_conn_str(enum drbd_conns s);
1661extern const char *drbd_role_str(enum drbd_role s);
1662
1663/* drbd_actlog.c */
1664extern void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector);
1665extern void drbd_al_complete_io(struct drbd_conf *mdev, sector_t sector);
1666extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector);
1667extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1668extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector);
1669extern void drbd_rs_cancel_all(struct drbd_conf *mdev);
1670extern int drbd_rs_del_all(struct drbd_conf *mdev);
1671extern void drbd_rs_failed_io(struct drbd_conf *mdev,
1672 sector_t sector, int size);
1673extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *);
ea5442af 1674extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go);
b411b363
PR		 1675extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector,
1676 int size, const char *file, const unsigned int line);
1677#define drbd_set_in_sync(mdev, sector, size) \
1678 __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__)
73a01a18 1679extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector,
b411b363
PR		 1680 int size, const char *file, const unsigned int line);
1681#define drbd_set_out_of_sync(mdev, sector, size) \
1682 __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__)
1683extern void drbd_al_apply_to_bm(struct drbd_conf *mdev);
b411b363
PR		 1684extern void drbd_al_shrink(struct drbd_conf *mdev);
1685
1686
1687/* drbd_nl.c */
1688
1689void drbd_nl_cleanup(void);
1690int __init drbd_nl_init(void);
1691void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state);
1692void drbd_bcast_sync_progress(struct drbd_conf *mdev);
1693void drbd_bcast_ee(struct drbd_conf *mdev,
1694 const char *reason, const int dgs,
1695 const char* seen_hash, const char* calc_hash,
1696 const struct drbd_epoch_entry* e);
1697
1698
1699/**
1700 * DOC: DRBD State macros
1701 *
1702 * These macros are used to express state changes in easily readable form.
1703 *
1704 * The NS macros expand to a mask and a value, that can be bit ored onto the
1705 * current state as soon as the spinlock (req_lock) was taken.
1706 *
1707 * The _NS macros are used for state functions that get called with the
1708 * spinlock. These macros expand directly to the new state value.
1709 *
1710 * Besides the basic forms NS() and _NS() additional _?NS[23] are defined
1711 * to express state changes that affect more than one aspect of the state.
1712 *
1713 * E.g. NS2(conn, C_CONNECTED, peer, R_SECONDARY)
1714 * Means that the network connection was established and that the peer
1715 * is in secondary role.
1716 */
1717#define role_MASK R_MASK
1718#define peer_MASK R_MASK
1719#define disk_MASK D_MASK
1720#define pdsk_MASK D_MASK
1721#define conn_MASK C_MASK
1722#define susp_MASK 1
1723#define user_isp_MASK 1
1724#define aftr_isp_MASK 1
fb22c402
PR		 1725#define susp_nod_MASK 1
1726#define susp_fen_MASK 1
b411b363
PR		 1727
1728#define NS(T, S) \
1729 ({ union drbd_state mask; mask.i = 0; mask.T = T##_MASK; mask; }), \
1730 ({ union drbd_state val; val.i = 0; val.T = (S); val; })
1731#define NS2(T1, S1, T2, S2) \
1732 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1733 mask.T2 = T2##_MASK; mask; }), \
1734 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1735 val.T2 = (S2); val; })
1736#define NS3(T1, S1, T2, S2, T3, S3) \
1737 ({ union drbd_state mask; mask.i = 0; mask.T1 = T1##_MASK; \
1738 mask.T2 = T2##_MASK; mask.T3 = T3##_MASK; mask; }), \
1739 ({ union drbd_state val; val.i = 0; val.T1 = (S1); \
1740 val.T2 = (S2); val.T3 = (S3); val; })
1741
1742#define _NS(D, T, S) \
1743 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T = (S); __ns; })
1744#define _NS2(D, T1, S1, T2, S2) \
1745 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1746 __ns.T2 = (S2); __ns; })
1747#define _NS3(D, T1, S1, T2, S2, T3, S3) \
1748 D, ({ union drbd_state __ns; __ns.i = D->state.i; __ns.T1 = (S1); \
1749 __ns.T2 = (S2); __ns.T3 = (S3); __ns; })
1750
1751/*
1752 * inline helper functions
1753 *************************/
1754
45bb912b
LE		 1755/* see also page_chain_add and friends in drbd_receiver.c */
1756static inline struct page *page_chain_next(struct page *page)
1757{
1758 return (struct page *)page_private(page);
1759}
1760#define page_chain_for_each(page) \
1761 for (; page && ({ prefetch(page_chain_next(page)); 1; }); \
1762 page = page_chain_next(page))
1763#define page_chain_for_each_safe(page, n) \
1764 for (; page && ({ n = page_chain_next(page); 1; }); page = n)
1765
1766static inline int drbd_bio_has_active_page(struct bio *bio)
1767{
1768 struct bio_vec *bvec;
1769 int i;
1770
1771 __bio_for_each_segment(bvec, bio, i, 0) {
1772 if (page_count(bvec->bv_page) > 1)
1773 return 1;
1774 }
1775
1776 return 0;
1777}
1778
1779static inline int drbd_ee_has_active_page(struct drbd_epoch_entry *e)
1780{
1781 struct page *page = e->pages;
1782 page_chain_for_each(page) {
1783 if (page_count(page) > 1)
1784 return 1;
1785 }
1786 return 0;
1787}
1788
1789
b411b363
PR		 1790static inline void drbd_state_lock(struct drbd_conf *mdev)
1791{
1792 wait_event(mdev->misc_wait,
1793 !test_and_set_bit(CLUSTER_ST_CHANGE, &mdev->flags));
1794}
1795
1796static inline void drbd_state_unlock(struct drbd_conf *mdev)
1797{
1798 clear_bit(CLUSTER_ST_CHANGE, &mdev->flags);
1799 wake_up(&mdev->misc_wait);
1800}
1801
bf885f8a
AG		 1802static inline enum drbd_state_rv
1803_drbd_set_state(struct drbd_conf *mdev, union drbd_state ns,
1804 enum chg_state_flags flags, struct completion *done)
b411b363 1805{
bf885f8a 1806 enum drbd_state_rv rv;
b411b363
PR		 1807
1808 read_lock(&global_state_lock);
1809 rv = __drbd_set_state(mdev, ns, flags, done);
1810 read_unlock(&global_state_lock);
1811
1812 return rv;
1813}
1814
1815/**
1816 * drbd_request_state() - Reqest a state change
1817 * @mdev: DRBD device.
1818 * @mask: mask of state bits to change.
1819 * @val: value of new state bits.
1820 *
1821 * This is the most graceful way of requesting a state change. It is verbose
1822 * quite verbose in case the state change is not possible, and all those
1823 * state changes are globally serialized.
1824 */
1825static inline int drbd_request_state(struct drbd_conf *mdev,
1826 union drbd_state mask,
1827 union drbd_state val)
1828{
1829 return _drbd_request_state(mdev, mask, val, CS_VERBOSE + CS_ORDERED);
1830}
1831
1832#define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__)
1833static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where)
1834{
1835 switch (mdev->ldev->dc.on_io_error) {
1836 case EP_PASS_ON:
1837 if (!forcedetach) {
7383506c 1838 if (__ratelimit(&drbd_ratelimit_state))
82f59cc6 1839 dev_err(DEV, "Local IO failed in %s.\n", where);
d2e17807
PR		 1840 if (mdev->state.disk > D_INCONSISTENT)
1841 _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL);
b411b363
PR		 1842 break;
1843 }
1844 /* NOTE fall through to detach case if forcedetach set */
1845 case EP_DETACH:
1846 case EP_CALL_HELPER:
82f59cc6 1847 set_bit(WAS_IO_ERROR, &mdev->flags);
b411b363
PR		 1848 if (mdev->state.disk > D_FAILED) {
1849 _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL);
82f59cc6
LE		 1850 dev_err(DEV,
1851 "Local IO failed in %s. Detaching...\n", where);
b411b363
PR		 1852 }
1853 break;
1854 }
1855}
1856
1857/**
1858 * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers
1859 * @mdev: DRBD device.
1860 * @error: Error code passed to the IO completion callback
1861 * @forcedetach: Force detach. I.e. the error happened while accessing the meta data
1862 *
1863 * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED)
1864 */
1865#define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__)
1866static inline void drbd_chk_io_error_(struct drbd_conf *mdev,
1867 int error, int forcedetach, const char *where)
1868{
1869 if (error) {
1870 unsigned long flags;
1871 spin_lock_irqsave(&mdev->req_lock, flags);
1872 __drbd_chk_io_error_(mdev, forcedetach, where);
1873 spin_unlock_irqrestore(&mdev->req_lock, flags);
1874 }
1875}
1876
1877
1878/**
1879 * drbd_md_first_sector() - Returns the first sector number of the meta data area
1880 * @bdev: Meta data block device.
1881 *
1882 * BTW, for internal meta data, this happens to be the maximum capacity
1883 * we could agree upon with our peer node.
1884 */
1885static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev)
1886{
1887 switch (bdev->dc.meta_dev_idx) {
1888 case DRBD_MD_INDEX_INTERNAL:
1889 case DRBD_MD_INDEX_FLEX_INT:
1890 return bdev->md.md_offset + bdev->md.bm_offset;
1891 case DRBD_MD_INDEX_FLEX_EXT:
1892 default:
1893 return bdev->md.md_offset;
1894 }
1895}
1896
1897/**
1898 * drbd_md_last_sector() - Return the last sector number of the meta data area
1899 * @bdev: Meta data block device.
1900 */
1901static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev)
1902{
1903 switch (bdev->dc.meta_dev_idx) {
1904 case DRBD_MD_INDEX_INTERNAL:
1905 case DRBD_MD_INDEX_FLEX_INT:
1906 return bdev->md.md_offset + MD_AL_OFFSET - 1;
1907 case DRBD_MD_INDEX_FLEX_EXT:
1908 default:
1909 return bdev->md.md_offset + bdev->md.md_size_sect;
1910 }
1911}
1912
1913/* Returns the number of 512 byte sectors of the device */
1914static inline sector_t drbd_get_capacity(struct block_device *bdev)
1915{
1916 /* return bdev ? get_capacity(bdev->bd_disk) : 0; */
77304d2a 1917 return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0;
b411b363
PR		 1918}
1919
1920/**
1921 * drbd_get_max_capacity() - Returns the capacity we announce to out peer
1922 * @bdev: Meta data block device.
1923 *
1924 * returns the capacity we announce to out peer. we clip ourselves at the
1925 * various MAX_SECTORS, because if we don't, current implementation will
1926 * oops sooner or later
1927 */
1928static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev)
1929{
1930 sector_t s;
1931 switch (bdev->dc.meta_dev_idx) {
1932 case DRBD_MD_INDEX_INTERNAL:
1933 case DRBD_MD_INDEX_FLEX_INT:
1934 s = drbd_get_capacity(bdev->backing_bdev)
1935 ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1936 drbd_md_first_sector(bdev))
1937 : 0;
1938 break;
1939 case DRBD_MD_INDEX_FLEX_EXT:
1940 s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX,
1941 drbd_get_capacity(bdev->backing_bdev));
1942 /* clip at maximum size the meta device can support */
1943 s = min_t(sector_t, s,
1944 BM_EXT_TO_SECT(bdev->md.md_size_sect
1945 - bdev->md.bm_offset));
1946 break;
1947 default:
1948 s = min_t(sector_t, DRBD_MAX_SECTORS,
1949 drbd_get_capacity(bdev->backing_bdev));
1950 }
1951 return s;
1952}
1953
1954/**
1955 * drbd_md_ss__() - Return the sector number of our meta data super block
1956 * @mdev: DRBD device.
1957 * @bdev: Meta data block device.
1958 */
1959static inline sector_t drbd_md_ss__(struct drbd_conf *mdev,
1960 struct drbd_backing_dev *bdev)
1961{
1962 switch (bdev->dc.meta_dev_idx) {
1963 default: /* external, some index */
1964 return MD_RESERVED_SECT * bdev->dc.meta_dev_idx;
1965 case DRBD_MD_INDEX_INTERNAL:
1966 /* with drbd08, internal meta data is always "flexible" */
1967 case DRBD_MD_INDEX_FLEX_INT:
1968 /* sizeof(struct md_on_disk_07) == 4k
1969 * position: last 4k aligned block of 4k size */
1970 if (!bdev->backing_bdev) {
1971 if (__ratelimit(&drbd_ratelimit_state)) {
1972 dev_err(DEV, "bdev->backing_bdev==NULL\n");
1973 dump_stack();
1974 }
1975 return 0;
1976 }
1977 return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL)
1978 - MD_AL_OFFSET;
1979 case DRBD_MD_INDEX_FLEX_EXT:
1980 return 0;
1981 }
1982}
1983
b411b363
PR		 1984static inline void
1985drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w)
1986{
1987 unsigned long flags;
1988 spin_lock_irqsave(&q->q_lock, flags);
1989 list_add(&w->list, &q->q);
1990 up(&q->s); /* within the spinlock,
1991 see comment near end of drbd_worker() */
1992 spin_unlock_irqrestore(&q->q_lock, flags);
1993}
1994
1995static inline void
1996drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w)
1997{
1998 unsigned long flags;
1999 spin_lock_irqsave(&q->q_lock, flags);
2000 list_add_tail(&w->list, &q->q);
2001 up(&q->s); /* within the spinlock,
2002 see comment near end of drbd_worker() */
2003 spin_unlock_irqrestore(&q->q_lock, flags);
2004}
2005
2006static inline void wake_asender(struct drbd_conf *mdev)
2007{
2008 if (test_bit(SIGNAL_ASENDER, &mdev->flags))
e6b3ea83 2009 force_sig(DRBD_SIG, mdev->tconn->asender.task);
b411b363
PR		 2010}
2011
2012static inline void request_ping(struct drbd_conf *mdev)
2013{
2014 set_bit(SEND_PING, &mdev->flags);
2015 wake_asender(mdev);
2016}
2017
2018static inline int drbd_send_short_cmd(struct drbd_conf *mdev,
2019 enum drbd_packets cmd)
2020{
0b70a13d 2021 struct p_header80 h;
b411b363
PR		 2022 return drbd_send_cmd(mdev, USE_DATA_SOCKET, cmd, &h, sizeof(h));
2023}
2024
2025static inline int drbd_send_ping(struct drbd_conf *mdev)
2026{
0b70a13d 2027 struct p_header80 h;
b411b363
PR		 2028 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING, &h, sizeof(h));
2029}
2030
2031static inline int drbd_send_ping_ack(struct drbd_conf *mdev)
2032{
0b70a13d 2033 struct p_header80 h;
b411b363
PR		 2034 return drbd_send_cmd(mdev, USE_META_SOCKET, P_PING_ACK, &h, sizeof(h));
2035}
2036
2037static inline void drbd_thread_stop(struct drbd_thread *thi)
2038{
81e84650 2039 _drbd_thread_stop(thi, false, true);
b411b363
PR		 2040}
2041
2042static inline void drbd_thread_stop_nowait(struct drbd_thread *thi)
2043{
81e84650 2044 _drbd_thread_stop(thi, false, false);
b411b363
PR		 2045}
2046
2047static inline void drbd_thread_restart_nowait(struct drbd_thread *thi)
2048{
81e84650 2049 _drbd_thread_stop(thi, true, false);
b411b363
PR		 2050}
2051
2052/* counts how many answer packets packets we expect from our peer,
2053 * for either explicit application requests,
2054 * or implicit barrier packets as necessary.
2055 * increased:
2056 * w_send_barrier
8554df1c 2057 * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ);
b411b363
PR		 2058 * it is much easier and equally valid to count what we queue for the
2059 * worker, even before it actually was queued or send.
2060 * (drbd_make_request_common; recovery path on read io-error)
2061 * decreased:
2062 * got_BarrierAck (respective tl_clear, tl_clear_barrier)
8554df1c 2063 * _req_mod(req, DATA_RECEIVED)
b411b363 2064 * [from receive_DataReply]
8554df1c 2065 * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED)
b411b363
PR		 2066 * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)]
2067 * for some reason it is NOT decreased in got_NegAck,
2068 * but in the resulting cleanup code from report_params.
2069 * we should try to remember the reason for that...
8554df1c
AG		 2070 * _req_mod(req, SEND_FAILED or SEND_CANCELED)
2071 * _req_mod(req, CONNECTION_LOST_WHILE_PENDING)
b411b363
PR		 2072 * [from tl_clear_barrier]
2073 */
2074static inline void inc_ap_pending(struct drbd_conf *mdev)
2075{
2076 atomic_inc(&mdev->ap_pending_cnt);
2077}
2078
2079#define ERR_IF_CNT_IS_NEGATIVE(which) \
2080 if (atomic_read(&mdev->which) < 0) \
2081 dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \
2082 __func__ , __LINE__ , \
2083 atomic_read(&mdev->which))
2084
2085#define dec_ap_pending(mdev) do { \
2086 typecheck(struct drbd_conf *, mdev); \
2087 if (atomic_dec_and_test(&mdev->ap_pending_cnt)) \
2088 wake_up(&mdev->misc_wait); \
2089 ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt); } while (0)
2090
2091/* counts how many resync-related answers we still expect from the peer
2092 * increase decrease
2093 * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY)
25985edc 2094 * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER)
b411b363
PR		 2095 * (or P_NEG_ACK with ID_SYNCER)
2096 */
2097static inline void inc_rs_pending(struct drbd_conf *mdev)
2098{
2099 atomic_inc(&mdev->rs_pending_cnt);
2100}
2101
2102#define dec_rs_pending(mdev) do { \
2103 typecheck(struct drbd_conf *, mdev); \
2104 atomic_dec(&mdev->rs_pending_cnt); \
2105 ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt); } while (0)
2106
2107/* counts how many answers we still need to send to the peer.
2108 * increased on
2109 * receive_Data unless protocol A;
2110 * we need to send a P_RECV_ACK (proto B)
2111 * or P_WRITE_ACK (proto C)
2112 * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK
2113 * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA
2114 * receive_Barrier_* we need to send a P_BARRIER_ACK
2115 */
2116static inline void inc_unacked(struct drbd_conf *mdev)
2117{
2118 atomic_inc(&mdev->unacked_cnt);
2119}
2120
2121#define dec_unacked(mdev) do { \
2122 typecheck(struct drbd_conf *, mdev); \
2123 atomic_dec(&mdev->unacked_cnt); \
2124 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2125
2126#define sub_unacked(mdev, n) do { \
2127 typecheck(struct drbd_conf *, mdev); \
2128 atomic_sub(n, &mdev->unacked_cnt); \
2129 ERR_IF_CNT_IS_NEGATIVE(unacked_cnt); } while (0)
2130
2131
b2fb6dbe 2132static inline void put_net_conf(struct drbd_tconn *tconn)
b411b363 2133{
b2fb6dbe
PR		 2134 if (atomic_dec_and_test(&tconn->net_cnt))
2135 wake_up(&tconn->net_cnt_wait);
b411b363
PR		 2136}
2137
2138/**
89e58e75 2139 * get_net_conf() - Increase ref count on mdev->tconn->net_conf; Returns 0 if nothing there
b411b363
PR		 2140 * @mdev: DRBD device.
2141 *
89e58e75 2142 * You have to call put_net_conf() when finished working with mdev->tconn->net_conf.
b411b363 2143 */
b2fb6dbe 2144static inline int get_net_conf(struct drbd_tconn *tconn)
b411b363
PR		 2145{
2146 int have_net_conf;
2147
b2fb6dbe
PR		 2148 atomic_inc(&tconn->net_cnt);
2149 have_net_conf = tconn->volume0->state.conn >= C_UNCONNECTED;
b411b363 2150 if (!have_net_conf)
b2fb6dbe 2151 put_net_conf(tconn);
b411b363
PR		 2152 return have_net_conf;
2153}
2154
2155/**
2156 * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev
2157 * @M: DRBD device.
2158 *
2159 * You have to call put_ldev() when finished working with mdev->ldev.
2160 */
2161#define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT))
2162#define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS))
2163
2164static inline void put_ldev(struct drbd_conf *mdev)
2165{
1d7734a0 2166 int i = atomic_dec_return(&mdev->local_cnt);
9a0d9d03
LE		 2167
2168 /* This may be called from some endio handler,
2169 * so we must not sleep here. */
2170
b411b363 2171 __release(local);
1d7734a0 2172 D_ASSERT(i >= 0);
e9e6f3ec 2173 if (i == 0) {
82f59cc6
LE		 2174 if (mdev->state.disk == D_DISKLESS)
2175 /* even internal references gone, safe to destroy */
2176 drbd_ldev_destroy(mdev);
e9e6f3ec 2177 if (mdev->state.disk == D_FAILED)
82f59cc6 2178 /* all application IO references gone. */
e9e6f3ec 2179 drbd_go_diskless(mdev);
b411b363 2180 wake_up(&mdev->misc_wait);
e9e6f3ec 2181 }
b411b363
PR		 2182}
2183
2184#ifndef __CHECKER__
2185static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins)
2186{
2187 int io_allowed;
2188
82f59cc6
LE		 2189 /* never get a reference while D_DISKLESS */
2190 if (mdev->state.disk == D_DISKLESS)
2191 return 0;
2192
b411b363
PR		 2193 atomic_inc(&mdev->local_cnt);
2194 io_allowed = (mdev->state.disk >= mins);
2195 if (!io_allowed)
2196 put_ldev(mdev);
2197 return io_allowed;
2198}
2199#else
2200extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins);
2201#endif
2202
2203/* you must have an "get_ldev" reference */
2204static inline void drbd_get_syncer_progress(struct drbd_conf *mdev,
2205 unsigned long *bits_left, unsigned int *per_mil_done)
2206{
4b0715f0
LE		 2207 /* this is to break it at compile time when we change that, in case we
2208 * want to support more than (1<<32) bits on a 32bit arch. */
b411b363
PR		 2209 typecheck(unsigned long, mdev->rs_total);
2210
2211 /* note: both rs_total and rs_left are in bits, i.e. in
2212 * units of BM_BLOCK_SIZE.
2213 * for the percentage, we don't care. */
2214
439d5953
LE		 2215 if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T)
2216 *bits_left = mdev->ov_left;
2217 else
2218 *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed;
b411b363
PR		 2219 /* >> 10 to prevent overflow,
2220 * +1 to prevent division by zero */
2221 if (*bits_left > mdev->rs_total) {
2222 /* doh. maybe a logic bug somewhere.
2223 * may also be just a race condition
2224 * between this and a disconnect during sync.
2225 * for now, just prevent in-kernel buffer overflow.
2226 */
2227 smp_rmb();
2228 dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n",
2229 drbd_conn_str(mdev->state.conn),
2230 *bits_left, mdev->rs_total, mdev->rs_failed);
2231 *per_mil_done = 0;
2232 } else {
4b0715f0
LE		 2233 /* Make sure the division happens in long context.
2234 * We allow up to one petabyte storage right now,
2235 * at a granularity of 4k per bit that is 2**38 bits.
2236 * After shift right and multiplication by 1000,
2237 * this should still fit easily into a 32bit long,
2238 * so we don't need a 64bit division on 32bit arch.
2239 * Note: currently we don't support such large bitmaps on 32bit
2240 * arch anyways, but no harm done to be prepared for it here.
2241 */
2242 unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10;
2243 unsigned long left = *bits_left >> shift;
2244 unsigned long total = 1UL + (mdev->rs_total >> shift);
2245 unsigned long tmp = 1000UL - left * 1000UL/total;
b411b363
PR		 2246 *per_mil_done = tmp;
2247 }
2248}
2249
2250
2251/* this throttles on-the-fly application requests
2252 * according to max_buffers settings;
2253 * maybe re-implement using semaphores? */
2254static inline int drbd_get_max_buffers(struct drbd_conf *mdev)
2255{
2256 int mxb = 1000000; /* arbitrary limit on open requests */
b2fb6dbe 2257 if (get_net_conf(mdev->tconn)) {
89e58e75 2258 mxb = mdev->tconn->net_conf->max_buffers;
b2fb6dbe 2259 put_net_conf(mdev->tconn);
b411b363
PR		 2260 }
2261 return mxb;
2262}
2263
3719094e 2264static inline int drbd_state_is_stable(struct drbd_conf *mdev)
b411b363 2265{
3719094e 2266 union drbd_state s = mdev->state;
b411b363
PR		 2267
2268 /* DO NOT add a default clause, we want the compiler to warn us
2269 * for any newly introduced state we may have forgotten to add here */
2270
2271 switch ((enum drbd_conns)s.conn) {
2272 /* new io only accepted when there is no connection, ... */
2273 case C_STANDALONE:
2274 case C_WF_CONNECTION:
2275 /* ... or there is a well established connection. */
2276 case C_CONNECTED:
2277 case C_SYNC_SOURCE:
2278 case C_SYNC_TARGET:
2279 case C_VERIFY_S:
2280 case C_VERIFY_T:
2281 case C_PAUSED_SYNC_S:
2282 case C_PAUSED_SYNC_T:
67531718
PR		 2283 case C_AHEAD:
2284 case C_BEHIND:
3719094e 2285 /* transitional states, IO allowed */
b411b363
PR		 2286 case C_DISCONNECTING:
2287 case C_UNCONNECTED:
2288 case C_TIMEOUT:
2289 case C_BROKEN_PIPE:
2290 case C_NETWORK_FAILURE:
2291 case C_PROTOCOL_ERROR:
2292 case C_TEAR_DOWN:
2293 case C_WF_REPORT_PARAMS:
2294 case C_STARTING_SYNC_S:
2295 case C_STARTING_SYNC_T:
3719094e
PR		 2296 break;
2297
2298 /* Allow IO in BM exchange states with new protocols */
b411b363 2299 case C_WF_BITMAP_S:
3719094e
PR		 2300 if (mdev->agreed_pro_version < 96)
2301 return 0;
2302 break;
2303
2304 /* no new io accepted in these states */
b411b363
PR		 2305 case C_WF_BITMAP_T:
2306 case C_WF_SYNC_UUID:
2307 case C_MASK:
2308 /* not "stable" */
2309 return 0;
2310 }
2311
2312 switch ((enum drbd_disk_state)s.disk) {
2313 case D_DISKLESS:
2314 case D_INCONSISTENT:
2315 case D_OUTDATED:
2316 case D_CONSISTENT:
2317 case D_UP_TO_DATE:
2318 /* disk state is stable as well. */
2319 break;
2320
2321 /* no new io accepted during tansitional states */
2322 case D_ATTACHING:
2323 case D_FAILED:
2324 case D_NEGOTIATING:
2325 case D_UNKNOWN:
2326 case D_MASK:
2327 /* not "stable" */
2328 return 0;
2329 }
2330
2331 return 1;
2332}
2333
fb22c402
PR		 2334static inline int is_susp(union drbd_state s)
2335{
2336 return s.susp || s.susp_nod || s.susp_fen;
2337}
2338
1b881ef7 2339static inline bool may_inc_ap_bio(struct drbd_conf *mdev)
b411b363
PR		 2340{
2341 int mxb = drbd_get_max_buffers(mdev);
2342
fb22c402 2343 if (is_susp(mdev->state))
1b881ef7 2344 return false;
b411b363 2345 if (test_bit(SUSPEND_IO, &mdev->flags))
1b881ef7 2346 return false;
b411b363
PR		 2347
2348 /* to avoid potential deadlock or bitmap corruption,
2349 * in various places, we only allow new application io
2350 * to start during "stable" states. */
2351
2352 /* no new io accepted when attaching or detaching the disk */
3719094e 2353 if (!drbd_state_is_stable(mdev))
1b881ef7 2354 return false;
b411b363
PR		 2355
2356 /* since some older kernels don't have atomic_add_unless,
2357 * and we are within the spinlock anyways, we have this workaround. */
2358 if (atomic_read(&mdev->ap_bio_cnt) > mxb)
1b881ef7 2359 return false;
b411b363 2360 if (test_bit(BITMAP_IO, &mdev->flags))
1b881ef7
AG		 2361 return false;
2362 return true;
b411b363
PR		 2363}
2364
1b881ef7 2365static inline bool inc_ap_bio_cond(struct drbd_conf *mdev, int count)
b411b363 2366{
1b881ef7 2367 bool rv = false;
8869d683
PR		 2368
2369 spin_lock_irq(&mdev->req_lock);
1b881ef7 2370 rv = may_inc_ap_bio(mdev);
8869d683
PR		 2371 if (rv)
2372 atomic_add(count, &mdev->ap_bio_cnt);
2373 spin_unlock_irq(&mdev->req_lock);
2374
2375 return rv;
2376}
b411b363 2377
8869d683
PR		 2378static inline void inc_ap_bio(struct drbd_conf *mdev, int count)
2379{
b411b363
PR		 2380 /* we wait here
2381 * as long as the device is suspended
2382 * until the bitmap is no longer on the fly during connection
2383 * handshake as long as we would exeed the max_buffer limit.
2384 *
2385 * to avoid races with the reconnect code,
2386 * we need to atomic_inc within the spinlock. */
2387
1b881ef7 2388 wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev, count));
b411b363
PR		 2389}
2390
2391static inline void dec_ap_bio(struct drbd_conf *mdev)
2392{
2393 int mxb = drbd_get_max_buffers(mdev);
2394 int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt);
2395
2396 D_ASSERT(ap_bio >= 0);
2397 /* this currently does wake_up for every dec_ap_bio!
2398 * maybe rather introduce some type of hysteresis?
2399 * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */
2400 if (ap_bio < mxb)
2401 wake_up(&mdev->misc_wait);
2402 if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) {
2403 if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags))
e42325a5 2404 drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w);
b411b363
PR		 2405 }
2406}
2407
62b0da3a 2408static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val)
b411b363 2409{
62b0da3a 2410 int changed = mdev->ed_uuid != val;
b411b363 2411 mdev->ed_uuid = val;
62b0da3a 2412 return changed;
b411b363
PR		 2413}
2414
2415static inline int seq_cmp(u32 a, u32 b)
2416{
2417 /* we assume wrap around at 32bit.
2418 * for wrap around at 24bit (old atomic_t),
2419 * we'd have to
2420 * a <<= 8; b <<= 8;
2421 */
2422 return (s32)(a) - (s32)(b);
2423}
2424#define seq_lt(a, b) (seq_cmp((a), (b)) < 0)
2425#define seq_gt(a, b) (seq_cmp((a), (b)) > 0)
2426#define seq_ge(a, b) (seq_cmp((a), (b)) >= 0)
2427#define seq_le(a, b) (seq_cmp((a), (b)) <= 0)
2428/* CAUTION: please no side effects in arguments! */
2429#define seq_max(a, b) ((u32)(seq_gt((a), (b)) ? (a) : (b)))
2430
2431static inline void update_peer_seq(struct drbd_conf *mdev, unsigned int new_seq)
2432{
2433 unsigned int m;
2434 spin_lock(&mdev->peer_seq_lock);
2435 m = seq_max(mdev->peer_seq, new_seq);
2436 mdev->peer_seq = m;
2437 spin_unlock(&mdev->peer_seq_lock);
2438 if (m == new_seq)
2439 wake_up(&mdev->seq_wait);
2440}
2441
2442static inline void drbd_update_congested(struct drbd_conf *mdev)
2443{
e42325a5 2444 struct sock *sk = mdev->tconn->data.socket->sk;
b411b363
PR		 2445 if (sk->sk_wmem_queued > sk->sk_sndbuf * 4 / 5)
2446 set_bit(NET_CONGESTED, &mdev->flags);
2447}
2448
2449static inline int drbd_queue_order_type(struct drbd_conf *mdev)
2450{
2451 /* sorry, we currently have no working implementation
2452 * of distributed TCQ stuff */
2453#ifndef QUEUE_ORDERED_NONE
2454#define QUEUE_ORDERED_NONE 0
2455#endif
2456 return QUEUE_ORDERED_NONE;
2457}
2458
b411b363
PR		 2459static inline void drbd_md_flush(struct drbd_conf *mdev)
2460{
2461 int r;
2462
a8a4e51e 2463 if (test_bit(MD_NO_FUA, &mdev->flags))
b411b363
PR		 2464 return;
2465
dd3932ed 2466 r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL);
b411b363 2467 if (r) {
a8a4e51e 2468 set_bit(MD_NO_FUA, &mdev->flags);
b411b363
PR		 2469 dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r);
2470 }
2471}
2472
2473#endif
Linux kernel - Deliverables
This page took 0.22538 seconds and 5 git commands to generate.