Commit | Line | Data |
---|---|---|
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> | |
062e879c | 43 | #include <linux/idr.h> |
b411b363 PR |
44 | #include <net/tcp.h> |
45 | #include <linux/lru_cache.h> | |
70c71606 | 46 | #include <linux/prefetch.h> |
3b98c0c2 | 47 | #include <linux/drbd_genl_api.h> |
b8907339 PR |
48 | #include <linux/drbd.h> |
49 | #include "drbd_state.h" | |
b411b363 PR |
50 | |
51 | #ifdef __CHECKER__ | |
52 | # define __protected_by(x) __attribute__((require_context(x,1,999,"rdwr"))) | |
53 | # define __protected_read_by(x) __attribute__((require_context(x,1,999,"read"))) | |
54 | # define __protected_write_by(x) __attribute__((require_context(x,1,999,"write"))) | |
55 | # define __must_hold(x) __attribute__((context(x,1,1), require_context(x,1,999,"call"))) | |
56 | #else | |
57 | # define __protected_by(x) | |
58 | # define __protected_read_by(x) | |
59 | # define __protected_write_by(x) | |
60 | # define __must_hold(x) | |
61 | #endif | |
62 | ||
63 | #define __no_warn(lock, stmt) do { __acquire(lock); stmt; __release(lock); } while (0) | |
64 | ||
65 | /* module parameter, defined in drbd_main.c */ | |
66 | extern unsigned int minor_count; | |
67 | extern int disable_sendpage; | |
68 | extern int allow_oos; | |
b411b363 PR |
69 | |
70 | #ifdef CONFIG_DRBD_FAULT_INJECTION | |
71 | extern int enable_faults; | |
72 | extern int fault_rate; | |
73 | extern int fault_devs; | |
74 | #endif | |
75 | ||
76 | extern char usermode_helper[]; | |
77 | ||
78 | ||
b411b363 PR |
79 | /* I don't remember why XCPU ... |
80 | * This is used to wake the asender, | |
81 | * and to interrupt sending the sending task | |
82 | * on disconnect. | |
83 | */ | |
84 | #define DRBD_SIG SIGXCPU | |
85 | ||
86 | /* This is used to stop/restart our threads. | |
87 | * Cannot use SIGTERM nor SIGKILL, since these | |
88 | * are sent out by init on runlevel changes | |
89 | * I choose SIGHUP for now. | |
90 | */ | |
91 | #define DRBD_SIGKILL SIGHUP | |
92 | ||
b411b363 PR |
93 | #define ID_IN_SYNC (4711ULL) |
94 | #define ID_OUT_OF_SYNC (4712ULL) | |
b411b363 | 95 | #define ID_SYNCER (-1ULL) |
579b57ed | 96 | |
4a23f264 | 97 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
b411b363 PR |
98 | |
99 | struct drbd_conf; | |
2111438b | 100 | struct drbd_tconn; |
b411b363 PR |
101 | |
102 | ||
103 | /* to shorten dev_warn(DEV, "msg"); and relatives statements */ | |
104 | #define DEV (disk_to_dev(mdev->vdisk)) | |
105 | ||
60ae4966 PR |
106 | #define conn_printk(LEVEL, TCONN, FMT, ARGS...) \ |
107 | printk(LEVEL "d-con %s: " FMT, TCONN->name , ## ARGS) | |
108 | #define conn_alert(TCONN, FMT, ARGS...) conn_printk(KERN_ALERT, TCONN, FMT, ## ARGS) | |
109 | #define conn_crit(TCONN, FMT, ARGS...) conn_printk(KERN_CRIT, TCONN, FMT, ## ARGS) | |
110 | #define conn_err(TCONN, FMT, ARGS...) conn_printk(KERN_ERR, TCONN, FMT, ## ARGS) | |
111 | #define conn_warn(TCONN, FMT, ARGS...) conn_printk(KERN_WARNING, TCONN, FMT, ## ARGS) | |
112 | #define conn_notice(TCONN, FMT, ARGS...) conn_printk(KERN_NOTICE, TCONN, FMT, ## ARGS) | |
113 | #define conn_info(TCONN, FMT, ARGS...) conn_printk(KERN_INFO, TCONN, FMT, ## ARGS) | |
114 | #define conn_dbg(TCONN, FMT, ARGS...) conn_printk(KERN_DEBUG, TCONN, FMT, ## ARGS) | |
115 | ||
b411b363 PR |
116 | #define D_ASSERT(exp) if (!(exp)) \ |
117 | dev_err(DEV, "ASSERT( " #exp " ) in %s:%d\n", __FILE__, __LINE__) | |
118 | ||
841ce241 AG |
119 | /** |
120 | * expect - Make an assertion | |
121 | * | |
122 | * Unlike the assert macro, this macro returns a boolean result. | |
123 | */ | |
124 | #define expect(exp) ({ \ | |
125 | bool _bool = (exp); \ | |
126 | if (!_bool) \ | |
127 | dev_err(DEV, "ASSERTION %s FAILED in %s\n", \ | |
128 | #exp, __func__); \ | |
129 | _bool; \ | |
130 | }) | |
b411b363 PR |
131 | |
132 | /* Defines to control fault insertion */ | |
133 | enum { | |
134 | DRBD_FAULT_MD_WR = 0, /* meta data write */ | |
135 | DRBD_FAULT_MD_RD = 1, /* read */ | |
136 | DRBD_FAULT_RS_WR = 2, /* resync */ | |
137 | DRBD_FAULT_RS_RD = 3, | |
138 | DRBD_FAULT_DT_WR = 4, /* data */ | |
139 | DRBD_FAULT_DT_RD = 5, | |
140 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ | |
141 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ | |
142 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ | |
6b4388ac | 143 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
b411b363 PR |
144 | |
145 | DRBD_FAULT_MAX, | |
146 | }; | |
147 | ||
b411b363 PR |
148 | extern unsigned int |
149 | _drbd_insert_fault(struct drbd_conf *mdev, unsigned int type); | |
0cf9d27e | 150 | |
b411b363 PR |
151 | static inline int |
152 | drbd_insert_fault(struct drbd_conf *mdev, unsigned int type) { | |
0cf9d27e | 153 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
b411b363 PR |
154 | return fault_rate && |
155 | (enable_faults & (1<<type)) && | |
156 | _drbd_insert_fault(mdev, type); | |
b411b363 | 157 | #else |
0cf9d27e | 158 | return 0; |
b411b363 | 159 | #endif |
0cf9d27e | 160 | } |
b411b363 PR |
161 | |
162 | /* integer division, round _UP_ to the next integer */ | |
163 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) | |
164 | /* usual integer division */ | |
165 | #define div_floor(A, B) ((A)/(B)) | |
166 | ||
167 | /* drbd_meta-data.c (still in drbd_main.c) */ | |
168 | /* 4th incarnation of the disk layout. */ | |
169 | #define DRBD_MD_MAGIC (DRBD_MAGIC+4) | |
170 | ||
b411b363 | 171 | extern struct ratelimit_state drbd_ratelimit_state; |
81a5d60e | 172 | extern struct idr minors; |
80883197 | 173 | extern struct list_head drbd_tconns; |
ef356262 | 174 | extern struct rw_semaphore drbd_cfg_rwsem; |
d3fcb490 | 175 | /* drbd_cfg_rwsem protects: drbd_tconns list, minors idr, tconn->volumes idr |
ef356262 | 176 | note: non sleeping iterations over the idrs are protoected by RCU */ |
b411b363 PR |
177 | |
178 | /* on the wire */ | |
d8763023 | 179 | enum drbd_packet { |
b411b363 PR |
180 | /* receiver (data socket) */ |
181 | P_DATA = 0x00, | |
182 | P_DATA_REPLY = 0x01, /* Response to P_DATA_REQUEST */ | |
183 | P_RS_DATA_REPLY = 0x02, /* Response to P_RS_DATA_REQUEST */ | |
184 | P_BARRIER = 0x03, | |
185 | P_BITMAP = 0x04, | |
186 | P_BECOME_SYNC_TARGET = 0x05, | |
187 | P_BECOME_SYNC_SOURCE = 0x06, | |
188 | P_UNPLUG_REMOTE = 0x07, /* Used at various times to hint the peer */ | |
189 | P_DATA_REQUEST = 0x08, /* Used to ask for a data block */ | |
190 | P_RS_DATA_REQUEST = 0x09, /* Used to ask for a data block for resync */ | |
191 | P_SYNC_PARAM = 0x0a, | |
192 | P_PROTOCOL = 0x0b, | |
193 | P_UUIDS = 0x0c, | |
194 | P_SIZES = 0x0d, | |
195 | P_STATE = 0x0e, | |
196 | P_SYNC_UUID = 0x0f, | |
197 | P_AUTH_CHALLENGE = 0x10, | |
198 | P_AUTH_RESPONSE = 0x11, | |
199 | P_STATE_CHG_REQ = 0x12, | |
200 | ||
201 | /* asender (meta socket */ | |
202 | P_PING = 0x13, | |
203 | P_PING_ACK = 0x14, | |
204 | P_RECV_ACK = 0x15, /* Used in protocol B */ | |
205 | P_WRITE_ACK = 0x16, /* Used in protocol C */ | |
206 | P_RS_WRITE_ACK = 0x17, /* Is a P_WRITE_ACK, additionally call set_in_sync(). */ | |
7be8da07 | 207 | P_DISCARD_WRITE = 0x18, /* Used in proto C, two-primaries conflict detection */ |
b411b363 PR |
208 | P_NEG_ACK = 0x19, /* Sent if local disk is unusable */ |
209 | P_NEG_DREPLY = 0x1a, /* Local disk is broken... */ | |
210 | P_NEG_RS_DREPLY = 0x1b, /* Local disk is broken... */ | |
211 | P_BARRIER_ACK = 0x1c, | |
212 | P_STATE_CHG_REPLY = 0x1d, | |
213 | ||
214 | /* "new" commands, no longer fitting into the ordering scheme above */ | |
215 | ||
216 | P_OV_REQUEST = 0x1e, /* data socket */ | |
217 | P_OV_REPLY = 0x1f, | |
218 | P_OV_RESULT = 0x20, /* meta socket */ | |
219 | P_CSUM_RS_REQUEST = 0x21, /* data socket */ | |
220 | P_RS_IS_IN_SYNC = 0x22, /* meta socket */ | |
221 | P_SYNC_PARAM89 = 0x23, /* data socket, protocol version 89 replacement for P_SYNC_PARAM */ | |
222 | P_COMPRESSED_BITMAP = 0x24, /* compressed or otherwise encoded bitmap transfer */ | |
0ced55a3 PR |
223 | /* P_CKPT_FENCE_REQ = 0x25, * currently reserved for protocol D */ |
224 | /* P_CKPT_DISABLE_REQ = 0x26, * currently reserved for protocol D */ | |
225 | P_DELAY_PROBE = 0x27, /* is used on BOTH sockets */ | |
73a01a18 | 226 | P_OUT_OF_SYNC = 0x28, /* Mark as out of sync (Outrunning), data socket */ |
d612d309 | 227 | P_RS_CANCEL = 0x29, /* meta: Used to cancel RS_DATA_REQUEST packet by SyncSource */ |
cf29c9d8 PR |
228 | P_CONN_ST_CHG_REQ = 0x2a, /* data sock: Connection wide state request */ |
229 | P_CONN_ST_CHG_REPLY = 0x2b, /* meta sock: Connection side state req reply */ | |
7be8da07 | 230 | P_RETRY_WRITE = 0x2c, /* Protocol C: retry conflicting write request */ |
b411b363 | 231 | |
b411b363 PR |
232 | P_MAY_IGNORE = 0x100, /* Flag to test if (cmd > P_MAY_IGNORE) ... */ |
233 | P_MAX_OPT_CMD = 0x101, | |
234 | ||
235 | /* special command ids for handshake */ | |
236 | ||
e5d6f33a AG |
237 | P_INITIAL_META = 0xfff1, /* First Packet on the MetaSock */ |
238 | P_INITIAL_DATA = 0xfff2, /* First Packet on the Socket */ | |
b411b363 | 239 | |
6038178e | 240 | P_CONNECTION_FEATURES = 0xfffe /* FIXED for the next century! */ |
b411b363 PR |
241 | }; |
242 | ||
d8763023 | 243 | extern const char *cmdname(enum drbd_packet cmd); |
b411b363 PR |
244 | |
245 | /* for sending/receiving the bitmap, | |
246 | * possibly in some encoding scheme */ | |
247 | struct bm_xfer_ctx { | |
248 | /* "const" | |
249 | * stores total bits and long words | |
250 | * of the bitmap, so we don't need to | |
251 | * call the accessor functions over and again. */ | |
252 | unsigned long bm_bits; | |
253 | unsigned long bm_words; | |
254 | /* during xfer, current position within the bitmap */ | |
255 | unsigned long bit_offset; | |
256 | unsigned long word_offset; | |
257 | ||
258 | /* statistics; index: (h->command == P_BITMAP) */ | |
259 | unsigned packets[2]; | |
260 | unsigned bytes[2]; | |
261 | }; | |
262 | ||
263 | extern void INFO_bm_xfer_stats(struct drbd_conf *mdev, | |
264 | const char *direction, struct bm_xfer_ctx *c); | |
265 | ||
266 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) | |
267 | { | |
268 | /* word_offset counts "native long words" (32 or 64 bit), | |
269 | * aligned at 64 bit. | |
270 | * Encoded packet may end at an unaligned bit offset. | |
271 | * In case a fallback clear text packet is transmitted in | |
272 | * between, we adjust this offset back to the last 64bit | |
273 | * aligned "native long word", which makes coding and decoding | |
274 | * the plain text bitmap much more convenient. */ | |
275 | #if BITS_PER_LONG == 64 | |
276 | c->word_offset = c->bit_offset >> 6; | |
277 | #elif BITS_PER_LONG == 32 | |
278 | c->word_offset = c->bit_offset >> 5; | |
279 | c->word_offset &= ~(1UL); | |
280 | #else | |
281 | # error "unsupported BITS_PER_LONG" | |
282 | #endif | |
283 | } | |
284 | ||
285 | #ifndef __packed | |
286 | #define __packed __attribute__((packed)) | |
287 | #endif | |
288 | ||
289 | /* This is the layout for a packet on the wire. | |
290 | * The byteorder is the network byte order. | |
291 | * (except block_id and barrier fields. | |
292 | * these are pointers to local structs | |
293 | * and have no relevance for the partner, | |
294 | * which just echoes them as received.) | |
295 | * | |
296 | * NOTE that the payload starts at a long aligned offset, | |
297 | * regardless of 32 or 64 bit arch! | |
298 | */ | |
0b70a13d | 299 | struct p_header80 { |
b411b363 PR |
300 | u32 magic; |
301 | u16 command; | |
302 | u16 length; /* bytes of data after this header */ | |
b411b363 | 303 | } __packed; |
0b70a13d PR |
304 | |
305 | /* Header for big packets, Used for data packets exceeding 64kB */ | |
306 | struct p_header95 { | |
307 | u16 magic; /* use DRBD_MAGIC_BIG here */ | |
308 | u16 command; | |
b55d84ba | 309 | u32 length; |
0b70a13d PR |
310 | } __packed; |
311 | ||
0c8e36d9 AG |
312 | struct p_header100 { |
313 | u32 magic; | |
314 | u16 volume; | |
315 | u16 command; | |
316 | u32 length; | |
317 | u32 pad; | |
318 | } __packed; | |
319 | ||
52b061a4 AG |
320 | extern unsigned int drbd_header_size(struct drbd_tconn *tconn); |
321 | ||
b411b363 | 322 | /* these defines must not be changed without changing the protocol version */ |
76d2e7ec PR |
323 | #define DP_HARDBARRIER 1 /* depricated */ |
324 | #define DP_RW_SYNC 2 /* equals REQ_SYNC */ | |
b411b363 | 325 | #define DP_MAY_SET_IN_SYNC 4 |
721a9602 | 326 | #define DP_UNPLUG 8 /* not used anymore */ |
76d2e7ec PR |
327 | #define DP_FUA 16 /* equals REQ_FUA */ |
328 | #define DP_FLUSH 32 /* equals REQ_FLUSH */ | |
329 | #define DP_DISCARD 64 /* equals REQ_DISCARD */ | |
303d1448 PR |
330 | #define DP_SEND_RECEIVE_ACK 128 /* This is a proto B write request */ |
331 | #define DP_SEND_WRITE_ACK 256 /* This is a proto C write request */ | |
b411b363 PR |
332 | |
333 | struct p_data { | |
b411b363 PR |
334 | u64 sector; /* 64 bits sector number */ |
335 | u64 block_id; /* to identify the request in protocol B&C */ | |
336 | u32 seq_num; | |
337 | u32 dp_flags; | |
338 | } __packed; | |
339 | ||
340 | /* | |
341 | * commands which share a struct: | |
342 | * p_block_ack: | |
343 | * P_RECV_ACK (proto B), P_WRITE_ACK (proto C), | |
7be8da07 | 344 | * P_DISCARD_WRITE (proto C, two-primaries conflict detection) |
b411b363 PR |
345 | * p_block_req: |
346 | * P_DATA_REQUEST, P_RS_DATA_REQUEST | |
347 | */ | |
348 | struct p_block_ack { | |
b411b363 PR |
349 | u64 sector; |
350 | u64 block_id; | |
351 | u32 blksize; | |
352 | u32 seq_num; | |
353 | } __packed; | |
354 | ||
b411b363 | 355 | struct p_block_req { |
b411b363 PR |
356 | u64 sector; |
357 | u64 block_id; | |
358 | u32 blksize; | |
359 | u32 pad; /* to multiple of 8 Byte */ | |
360 | } __packed; | |
361 | ||
362 | /* | |
363 | * commands with their own struct for additional fields: | |
6038178e | 364 | * P_CONNECTION_FEATURES |
b411b363 PR |
365 | * P_BARRIER |
366 | * P_BARRIER_ACK | |
367 | * P_SYNC_PARAM | |
368 | * ReportParams | |
369 | */ | |
370 | ||
6038178e | 371 | struct p_connection_features { |
b411b363 PR |
372 | u32 protocol_min; |
373 | u32 feature_flags; | |
374 | u32 protocol_max; | |
375 | ||
376 | /* should be more than enough for future enhancements | |
377 | * for now, feature_flags and the reserverd array shall be zero. | |
378 | */ | |
379 | ||
380 | u32 _pad; | |
381 | u64 reserverd[7]; | |
382 | } __packed; | |
b411b363 PR |
383 | |
384 | struct p_barrier { | |
b411b363 PR |
385 | u32 barrier; /* barrier number _handle_ only */ |
386 | u32 pad; /* to multiple of 8 Byte */ | |
387 | } __packed; | |
388 | ||
389 | struct p_barrier_ack { | |
b411b363 PR |
390 | u32 barrier; |
391 | u32 set_size; | |
392 | } __packed; | |
393 | ||
394 | struct p_rs_param { | |
b411b363 PR |
395 | u32 rate; |
396 | ||
397 | /* Since protocol version 88 and higher. */ | |
398 | char verify_alg[0]; | |
399 | } __packed; | |
400 | ||
401 | struct p_rs_param_89 { | |
b411b363 PR |
402 | u32 rate; |
403 | /* protocol version 89: */ | |
404 | char verify_alg[SHARED_SECRET_MAX]; | |
405 | char csums_alg[SHARED_SECRET_MAX]; | |
406 | } __packed; | |
407 | ||
8e26f9cc | 408 | struct p_rs_param_95 { |
8e26f9cc PR |
409 | u32 rate; |
410 | char verify_alg[SHARED_SECRET_MAX]; | |
411 | char csums_alg[SHARED_SECRET_MAX]; | |
412 | u32 c_plan_ahead; | |
413 | u32 c_delay_target; | |
414 | u32 c_fill_target; | |
415 | u32 c_max_rate; | |
416 | } __packed; | |
417 | ||
cf14c2e9 PR |
418 | enum drbd_conn_flags { |
419 | CF_WANT_LOSE = 1, | |
420 | CF_DRY_RUN = 2, | |
421 | }; | |
422 | ||
b411b363 | 423 | struct p_protocol { |
b411b363 PR |
424 | u32 protocol; |
425 | u32 after_sb_0p; | |
426 | u32 after_sb_1p; | |
427 | u32 after_sb_2p; | |
cf14c2e9 | 428 | u32 conn_flags; |
b411b363 PR |
429 | u32 two_primaries; |
430 | ||
431 | /* Since protocol version 87 and higher. */ | |
432 | char integrity_alg[0]; | |
433 | ||
434 | } __packed; | |
435 | ||
436 | struct p_uuids { | |
b411b363 PR |
437 | u64 uuid[UI_EXTENDED_SIZE]; |
438 | } __packed; | |
439 | ||
440 | struct p_rs_uuid { | |
b411b363 PR |
441 | u64 uuid; |
442 | } __packed; | |
443 | ||
444 | struct p_sizes { | |
b411b363 PR |
445 | u64 d_size; /* size of disk */ |
446 | u64 u_size; /* user requested size */ | |
447 | u64 c_size; /* current exported size */ | |
1816a2b4 | 448 | u32 max_bio_size; /* Maximal size of a BIO */ |
e89b591c PR |
449 | u16 queue_order_type; /* not yet implemented in DRBD*/ |
450 | u16 dds_flags; /* use enum dds_flags here. */ | |
b411b363 PR |
451 | } __packed; |
452 | ||
453 | struct p_state { | |
b411b363 PR |
454 | u32 state; |
455 | } __packed; | |
456 | ||
457 | struct p_req_state { | |
b411b363 PR |
458 | u32 mask; |
459 | u32 val; | |
460 | } __packed; | |
461 | ||
462 | struct p_req_state_reply { | |
b411b363 PR |
463 | u32 retcode; |
464 | } __packed; | |
465 | ||
466 | struct p_drbd06_param { | |
467 | u64 size; | |
468 | u32 state; | |
469 | u32 blksize; | |
470 | u32 protocol; | |
471 | u32 version; | |
472 | u32 gen_cnt[5]; | |
473 | u32 bit_map_gen[5]; | |
474 | } __packed; | |
475 | ||
476 | struct p_discard { | |
b411b363 PR |
477 | u64 block_id; |
478 | u32 seq_num; | |
479 | u32 pad; | |
480 | } __packed; | |
481 | ||
73a01a18 | 482 | struct p_block_desc { |
73a01a18 PR |
483 | u64 sector; |
484 | u32 blksize; | |
485 | u32 pad; /* to multiple of 8 Byte */ | |
486 | } __packed; | |
487 | ||
b411b363 PR |
488 | /* Valid values for the encoding field. |
489 | * Bump proto version when changing this. */ | |
490 | enum drbd_bitmap_code { | |
491 | /* RLE_VLI_Bytes = 0, | |
492 | * and other bit variants had been defined during | |
493 | * algorithm evaluation. */ | |
494 | RLE_VLI_Bits = 2, | |
495 | }; | |
496 | ||
497 | struct p_compressed_bm { | |
b411b363 PR |
498 | /* (encoding & 0x0f): actual encoding, see enum drbd_bitmap_code |
499 | * (encoding & 0x80): polarity (set/unset) of first runlength | |
500 | * ((encoding >> 4) & 0x07): pad_bits, number of trailing zero bits | |
501 | * used to pad up to head.length bytes | |
502 | */ | |
503 | u8 encoding; | |
504 | ||
505 | u8 code[0]; | |
506 | } __packed; | |
507 | ||
0b70a13d | 508 | struct p_delay_probe93 { |
0b70a13d PR |
509 | u32 seq_num; /* sequence number to match the two probe packets */ |
510 | u32 offset; /* usecs the probe got sent after the reference time point */ | |
0ced55a3 PR |
511 | } __packed; |
512 | ||
50d0b1ad AG |
513 | /* |
514 | * Bitmap packets need to fit within a single page on the sender and receiver, | |
515 | * so we are limited to 4 KiB (and not to PAGE_SIZE, which can be bigger). | |
b411b363 | 516 | */ |
e6ef8a5c AG |
517 | #define DRBD_SOCKET_BUFFER_SIZE 4096 |
518 | ||
b411b363 PR |
519 | /**********************************************************************/ |
520 | enum drbd_thread_state { | |
e77a0a5c AG |
521 | NONE, |
522 | RUNNING, | |
523 | EXITING, | |
524 | RESTARTING | |
b411b363 PR |
525 | }; |
526 | ||
527 | struct drbd_thread { | |
528 | spinlock_t t_lock; | |
529 | struct task_struct *task; | |
530 | struct completion stop; | |
531 | enum drbd_thread_state t_state; | |
532 | int (*function) (struct drbd_thread *); | |
392c8801 | 533 | struct drbd_tconn *tconn; |
b411b363 | 534 | int reset_cpu_mask; |
bed879ae | 535 | char name[9]; |
b411b363 PR |
536 | }; |
537 | ||
538 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) | |
539 | { | |
540 | /* THINK testing the t_state seems to be uncritical in all cases | |
541 | * (but thread_{start,stop}), so we can read it *without* the lock. | |
542 | * --lge */ | |
543 | ||
544 | smp_rmb(); | |
545 | return thi->t_state; | |
546 | } | |
547 | ||
b411b363 PR |
548 | struct drbd_work { |
549 | struct list_head list; | |
309a8348 | 550 | int (*cb)(struct drbd_work *, int cancel); |
00d56944 PR |
551 | union { |
552 | struct drbd_conf *mdev; | |
553 | struct drbd_tconn *tconn; | |
554 | }; | |
b411b363 PR |
555 | }; |
556 | ||
ace652ac AG |
557 | #include "drbd_interval.h" |
558 | ||
7be8da07 AG |
559 | extern int drbd_wait_misc(struct drbd_conf *, struct drbd_interval *); |
560 | ||
b411b363 PR |
561 | struct drbd_request { |
562 | struct drbd_work w; | |
b411b363 PR |
563 | |
564 | /* if local IO is not allowed, will be NULL. | |
565 | * if local IO _is_ allowed, holds the locally submitted bio clone, | |
566 | * or, after local IO completion, the ERR_PTR(error). | |
fcefa62e | 567 | * see drbd_request_endio(). */ |
b411b363 PR |
568 | struct bio *private_bio; |
569 | ||
ace652ac | 570 | struct drbd_interval i; |
b411b363 PR |
571 | unsigned int epoch; /* barrier_nr */ |
572 | ||
573 | /* barrier_nr: used to check on "completion" whether this req was in | |
574 | * the current epoch, and we therefore have to close it, | |
575 | * starting a new epoch... | |
576 | */ | |
577 | ||
b411b363 PR |
578 | struct list_head tl_requests; /* ring list in the transfer log */ |
579 | struct bio *master_bio; /* master bio pointer */ | |
580 | unsigned long rq_state; /* see comments above _req_mod() */ | |
581 | int seq_num; | |
582 | unsigned long start_time; | |
583 | }; | |
584 | ||
585 | struct drbd_tl_epoch { | |
586 | struct drbd_work w; | |
587 | struct list_head requests; /* requests before */ | |
588 | struct drbd_tl_epoch *next; /* pointer to the next barrier */ | |
589 | unsigned int br_number; /* the barriers identifier. */ | |
7e602c0a | 590 | int n_writes; /* number of requests attached before this barrier */ |
b411b363 PR |
591 | }; |
592 | ||
b411b363 PR |
593 | struct drbd_epoch { |
594 | struct list_head list; | |
595 | unsigned int barrier_nr; | |
596 | atomic_t epoch_size; /* increased on every request added. */ | |
597 | atomic_t active; /* increased on every req. added, and dec on every finished. */ | |
598 | unsigned long flags; | |
599 | }; | |
600 | ||
601 | /* drbd_epoch flag bits */ | |
602 | enum { | |
b411b363 | 603 | DE_HAVE_BARRIER_NUMBER, |
b411b363 PR |
604 | }; |
605 | ||
606 | enum epoch_event { | |
607 | EV_PUT, | |
608 | EV_GOT_BARRIER_NR, | |
b411b363 | 609 | EV_BECAME_LAST, |
b411b363 PR |
610 | EV_CLEANUP = 32, /* used as flag */ |
611 | }; | |
612 | ||
b411b363 PR |
613 | struct drbd_wq_barrier { |
614 | struct drbd_work w; | |
615 | struct completion done; | |
616 | }; | |
617 | ||
618 | struct digest_info { | |
619 | int digest_size; | |
620 | void *digest; | |
621 | }; | |
622 | ||
f6ffca9f | 623 | struct drbd_peer_request { |
45bb912b | 624 | struct drbd_work w; |
85719573 | 625 | struct drbd_epoch *epoch; /* for writes */ |
45bb912b LE |
626 | struct page *pages; |
627 | atomic_t pending_bios; | |
010f6e67 | 628 | struct drbd_interval i; |
45bb912b LE |
629 | /* see comments on ee flag bits below */ |
630 | unsigned long flags; | |
85719573 PR |
631 | union { |
632 | u64 block_id; | |
633 | struct digest_info *digest; | |
634 | }; | |
45bb912b LE |
635 | }; |
636 | ||
637 | /* ee flag bits. | |
638 | * While corresponding bios are in flight, the only modification will be | |
639 | * set_bit WAS_ERROR, which has to be atomic. | |
640 | * If no bios are in flight yet, or all have been completed, | |
641 | * non-atomic modification to ee->flags is ok. | |
642 | */ | |
b411b363 PR |
643 | enum { |
644 | __EE_CALL_AL_COMPLETE_IO, | |
b411b363 | 645 | __EE_MAY_SET_IN_SYNC, |
45bb912b | 646 | |
45bb912b LE |
647 | /* In case a barrier failed, |
648 | * we need to resubmit without the barrier flag. */ | |
649 | __EE_RESUBMITTED, | |
650 | ||
6c852bec | 651 | /* we may have several bios per peer request. |
45bb912b LE |
652 | * if any of those fail, we set this flag atomically |
653 | * from the endio callback */ | |
654 | __EE_WAS_ERROR, | |
c36c3ced LE |
655 | |
656 | /* This ee has a pointer to a digest instead of a block id */ | |
657 | __EE_HAS_DIGEST, | |
7be8da07 AG |
658 | |
659 | /* Conflicting local requests need to be restarted after this request */ | |
660 | __EE_RESTART_REQUESTS, | |
303d1448 PR |
661 | |
662 | /* The peer wants a write ACK for this (wire proto C) */ | |
663 | __EE_SEND_WRITE_ACK, | |
b411b363 PR |
664 | }; |
665 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) | |
b411b363 | 666 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
45bb912b LE |
667 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) |
668 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) | |
c36c3ced | 669 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
7be8da07 | 670 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
303d1448 | 671 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
b411b363 | 672 | |
01a311a5 | 673 | /* flag bits per mdev */ |
b411b363 | 674 | enum { |
25985edc | 675 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
b411b363 PR |
676 | UNPLUG_QUEUED, /* only relevant with kernel 2.4 */ |
677 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ | |
678 | MD_DIRTY, /* current uuids and flags not yet on disk */ | |
b411b363 | 679 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
b411b363 PR |
680 | CL_ST_CHG_SUCCESS, |
681 | CL_ST_CHG_FAIL, | |
682 | CRASHED_PRIMARY, /* This node was a crashed primary. | |
683 | * Gets cleared when the state.conn | |
684 | * goes into C_CONNECTED state. */ | |
19f843aa | 685 | NO_BARRIER_SUPP, /* underlying block device doesn't implement barriers */ |
b411b363 PR |
686 | CONSIDER_RESYNC, |
687 | ||
a8a4e51e | 688 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
b411b363 PR |
689 | SUSPEND_IO, /* suspend application io */ |
690 | BITMAP_IO, /* suspend application io; | |
691 | once no more io in flight, start bitmap io */ | |
692 | BITMAP_IO_QUEUED, /* Started bitmap IO */ | |
82f59cc6 LE |
693 | GO_DISKLESS, /* Disk is being detached, on io-error or admin request. */ |
694 | WAS_IO_ERROR, /* Local disk failed returned IO error */ | |
b411b363 | 695 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
b411b363 PR |
696 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
697 | * the peer, if it changed there as well. */ | |
43a5182c | 698 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
0778286a | 699 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
370a43e7 | 700 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
e64a3294 | 701 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
b411b363 PR |
702 | }; |
703 | ||
704 | struct drbd_bitmap; /* opaque for drbd_conf */ | |
705 | ||
20ceb2b2 LE |
706 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
707 | * and drbd_bitmap_io and friends. */ | |
708 | enum bm_flag { | |
709 | /* do we need to kfree, or vfree bm_pages? */ | |
710 | BM_P_VMALLOCED = 0x10000, /* internal use only, will be masked out */ | |
711 | ||
712 | /* currently locked for bulk operation */ | |
713 | BM_LOCKED_MASK = 0x7, | |
714 | ||
715 | /* in detail, that is: */ | |
716 | BM_DONT_CLEAR = 0x1, | |
717 | BM_DONT_SET = 0x2, | |
718 | BM_DONT_TEST = 0x4, | |
719 | ||
720 | /* (test bit, count bit) allowed (common case) */ | |
721 | BM_LOCKED_TEST_ALLOWED = 0x3, | |
722 | ||
723 | /* testing bits, as well as setting new bits allowed, but clearing bits | |
724 | * would be unexpected. Used during bitmap receive. Setting new bits | |
725 | * requires sending of "out-of-sync" information, though. */ | |
726 | BM_LOCKED_SET_ALLOWED = 0x1, | |
727 | ||
728 | /* clear is not expected while bitmap is locked for bulk operation */ | |
729 | }; | |
730 | ||
731 | ||
b411b363 PR |
732 | /* TODO sort members for performance |
733 | * MAYBE group them further */ | |
734 | ||
735 | /* THINK maybe we actually want to use the default "event/%s" worker threads | |
736 | * or similar in linux 2.6, which uses per cpu data and threads. | |
b411b363 PR |
737 | */ |
738 | struct drbd_work_queue { | |
739 | struct list_head q; | |
740 | struct semaphore s; /* producers up it, worker down()s it */ | |
741 | spinlock_t q_lock; /* to protect the list. */ | |
742 | }; | |
743 | ||
744 | struct drbd_socket { | |
745 | struct drbd_work_queue work; | |
746 | struct mutex mutex; | |
747 | struct socket *socket; | |
748 | /* this way we get our | |
749 | * send/receive buffers off the stack */ | |
5a87d920 | 750 | void *sbuf; |
e6ef8a5c | 751 | void *rbuf; |
b411b363 PR |
752 | }; |
753 | ||
754 | struct drbd_md { | |
755 | u64 md_offset; /* sector offset to 'super' block */ | |
756 | ||
757 | u64 la_size_sect; /* last agreed size, unit sectors */ | |
758 | u64 uuid[UI_SIZE]; | |
759 | u64 device_uuid; | |
760 | u32 flags; | |
761 | u32 md_size_sect; | |
762 | ||
763 | s32 al_offset; /* signed relative sector offset to al area */ | |
764 | s32 bm_offset; /* signed relative sector offset to bitmap */ | |
765 | ||
766 | /* u32 al_nr_extents; important for restoring the AL | |
f399002e | 767 | * is stored into ldev->dc.al_extents, which in turn |
b411b363 PR |
768 | * gets applied to act_log->nr_elements |
769 | */ | |
770 | }; | |
771 | ||
b411b363 PR |
772 | struct drbd_backing_dev { |
773 | struct block_device *backing_bdev; | |
774 | struct block_device *md_bdev; | |
b411b363 PR |
775 | struct drbd_md md; |
776 | struct disk_conf dc; /* The user provided config... */ | |
777 | sector_t known_size; /* last known size of that backing device */ | |
778 | }; | |
779 | ||
780 | struct drbd_md_io { | |
781 | struct drbd_conf *mdev; | |
782 | struct completion event; | |
783 | int error; | |
784 | }; | |
785 | ||
786 | struct bm_io_work { | |
787 | struct drbd_work w; | |
788 | char *why; | |
20ceb2b2 | 789 | enum bm_flag flags; |
b411b363 PR |
790 | int (*io_fn)(struct drbd_conf *mdev); |
791 | void (*done)(struct drbd_conf *mdev, int rv); | |
792 | }; | |
793 | ||
794 | enum write_ordering_e { | |
795 | WO_none, | |
796 | WO_drain_io, | |
797 | WO_bdev_flush, | |
b411b363 PR |
798 | }; |
799 | ||
778f271d PR |
800 | struct fifo_buffer { |
801 | int *values; | |
802 | unsigned int head_index; | |
803 | unsigned int size; | |
804 | }; | |
805 | ||
01a311a5 PR |
806 | /* flag bits per tconn */ |
807 | enum { | |
808 | NET_CONGESTED, /* The data socket is congested */ | |
25703f83 | 809 | DISCARD_CONCURRENT, /* Set on one node, cleared on the peer! */ |
e43ef195 | 810 | SEND_PING, /* whether asender should send a ping asap */ |
808e37b8 | 811 | SIGNAL_ASENDER, /* whether asender wants to be interrupted */ |
2a67d8b9 | 812 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
fc3b10a4 PR |
813 | CONN_WD_ST_CHG_OKAY, |
814 | CONN_WD_ST_CHG_FAIL, | |
0e29d163 PR |
815 | CONFIG_PENDING, /* serialization of (re)configuration requests. |
816 | * if set, also prevents the device from dying */ | |
817 | OBJECT_DYING, /* device became unconfigured, | |
818 | * but worker thread is still handling the cleanup. | |
819 | * reconfiguring (nl_disk_conf, nl_net_conf) is dissalowed, | |
820 | * while this is set. */ | |
8169e41b | 821 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
01a311a5 PR |
822 | }; |
823 | ||
2111438b PR |
824 | struct drbd_tconn { /* is a resource from the config file */ |
825 | char *name; /* Resource name */ | |
543cc10b | 826 | struct list_head all_tconn; /* linked on global drbd_tconns */ |
8410da8f PR |
827 | struct idr volumes; /* <tconn, vnr> to mdev mapping */ |
828 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ | |
8e0af25f PR |
829 | unsigned susp:1; /* IO suspended by user */ |
830 | unsigned susp_nod:1; /* IO suspended because no data */ | |
831 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ | |
8410da8f | 832 | struct mutex cstate_mutex; /* Protects graceful disconnects */ |
2111438b | 833 | |
062e879c | 834 | unsigned long flags; |
44ed167d | 835 | struct net_conf *net_conf; /* content protected by rcu */ |
91fd4dad PR |
836 | struct mutex net_conf_update; /* mutex for ready-copy-update of net_conf */ |
837 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ | |
f399002e | 838 | struct res_opts res_opts; |
e42325a5 PR |
839 | |
840 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ | |
841 | struct drbd_socket meta; /* ping/ack (metadata) packets */ | |
31890f4a PR |
842 | int agreed_pro_version; /* actually used protocol version */ |
843 | unsigned long last_received; /* in jiffies, either socket */ | |
844 | unsigned int ko_count; | |
e6b3ea83 | 845 | |
87eeee41 PR |
846 | spinlock_t req_lock; |
847 | struct drbd_tl_epoch *unused_spare_tle; /* for pre-allocation */ | |
848 | struct drbd_tl_epoch *newest_tle; | |
849 | struct drbd_tl_epoch *oldest_tle; | |
850 | struct list_head out_of_sequence_requests; | |
851 | ||
a0638456 PR |
852 | struct crypto_hash *cram_hmac_tfm; |
853 | struct crypto_hash *integrity_w_tfm; /* to be used by the worker thread */ | |
854 | struct crypto_hash *integrity_r_tfm; /* to be used by the receiver thread */ | |
f399002e LE |
855 | struct crypto_hash *csums_tfm; |
856 | struct crypto_hash *verify_tfm; | |
a0638456 PR |
857 | void *int_dig_in; |
858 | void *int_dig_vv; | |
859 | ||
e6b3ea83 PR |
860 | struct drbd_thread receiver; |
861 | struct drbd_thread worker; | |
862 | struct drbd_thread asender; | |
80822284 | 863 | cpumask_var_t cpu_mask; |
2111438b PR |
864 | }; |
865 | ||
b411b363 | 866 | struct drbd_conf { |
2111438b PR |
867 | struct drbd_tconn *tconn; |
868 | int vnr; /* volume number within the connection */ | |
869 | ||
b411b363 PR |
870 | /* things that are stored as / read from meta data on disk */ |
871 | unsigned long flags; | |
872 | ||
873 | /* configured by drbdsetup */ | |
b411b363 PR |
874 | struct drbd_backing_dev *ldev __protected_by(local); |
875 | ||
876 | sector_t p_size; /* partner's disk size */ | |
877 | struct request_queue *rq_queue; | |
878 | struct block_device *this_bdev; | |
879 | struct gendisk *vdisk; | |
880 | ||
b411b363 PR |
881 | struct drbd_work resync_work, |
882 | unplug_work, | |
e9e6f3ec | 883 | go_diskless, |
c4752ef1 PR |
884 | md_sync_work, |
885 | start_resync_work; | |
b411b363 PR |
886 | struct timer_list resync_timer; |
887 | struct timer_list md_sync_timer; | |
370a43e7 | 888 | struct timer_list start_resync_timer; |
7fde2be9 | 889 | struct timer_list request_timer; |
ee15b038 LE |
890 | #ifdef DRBD_DEBUG_MD_SYNC |
891 | struct { | |
892 | unsigned int line; | |
893 | const char* func; | |
894 | } last_md_mark_dirty; | |
895 | #endif | |
b411b363 PR |
896 | |
897 | /* Used after attach while negotiating new disk state. */ | |
898 | union drbd_state new_state_tmp; | |
899 | ||
da9fbc27 | 900 | union drbd_dev_state state; |
b411b363 PR |
901 | wait_queue_head_t misc_wait; |
902 | wait_queue_head_t state_wait; /* upon each state change. */ | |
903 | unsigned int send_cnt; | |
904 | unsigned int recv_cnt; | |
905 | unsigned int read_cnt; | |
906 | unsigned int writ_cnt; | |
907 | unsigned int al_writ_cnt; | |
908 | unsigned int bm_writ_cnt; | |
909 | atomic_t ap_bio_cnt; /* Requests we need to complete */ | |
910 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ | |
911 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ | |
912 | atomic_t unacked_cnt; /* Need to send replys for */ | |
913 | atomic_t local_cnt; /* Waiting for local completion */ | |
b2fb6dbe | 914 | |
dac1389c AG |
915 | /* Interval tree of pending local requests */ |
916 | struct rb_root read_requests; | |
de696716 AG |
917 | struct rb_root write_requests; |
918 | ||
4b0715f0 | 919 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
b411b363 | 920 | unsigned long rs_total; |
4b0715f0 | 921 | /* number of resync blocks that failed in this run */ |
b411b363 PR |
922 | unsigned long rs_failed; |
923 | /* Syncer's start time [unit jiffies] */ | |
924 | unsigned long rs_start; | |
925 | /* cumulated time in PausedSyncX state [unit jiffies] */ | |
926 | unsigned long rs_paused; | |
1d7734a0 LE |
927 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
928 | unsigned long rs_same_csum; | |
929 | #define DRBD_SYNC_MARKS 8 | |
930 | #define DRBD_SYNC_MARK_STEP (3*HZ) | |
b411b363 | 931 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
1d7734a0 | 932 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
b411b363 | 933 | /* marks's time [unit jiffies] */ |
1d7734a0 LE |
934 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
935 | /* current index into rs_mark_{left,time} */ | |
936 | int rs_last_mark; | |
b411b363 PR |
937 | |
938 | /* where does the admin want us to start? (sector) */ | |
939 | sector_t ov_start_sector; | |
940 | /* where are we now? (sector) */ | |
941 | sector_t ov_position; | |
942 | /* Start sector of out of sync range (to merge printk reporting). */ | |
943 | sector_t ov_last_oos_start; | |
944 | /* size of out-of-sync range in sectors. */ | |
945 | sector_t ov_last_oos_size; | |
946 | unsigned long ov_left; /* in bits */ | |
b411b363 | 947 | |
b411b363 PR |
948 | struct drbd_bitmap *bitmap; |
949 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ | |
950 | ||
951 | /* Used to track operations of resync... */ | |
952 | struct lru_cache *resync; | |
953 | /* Number of locked elements in resync LRU */ | |
954 | unsigned int resync_locked; | |
955 | /* resync extent number waiting for application requests */ | |
956 | unsigned int resync_wenr; | |
957 | ||
958 | int open_cnt; | |
959 | u64 *p_uuid; | |
960 | struct drbd_epoch *current_epoch; | |
961 | spinlock_t epoch_lock; | |
962 | unsigned int epochs; | |
963 | enum write_ordering_e write_ordering; | |
85719573 PR |
964 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
965 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ | |
18b75d75 AG |
966 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
967 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ | |
b411b363 | 968 | struct list_head net_ee; /* zero-copy network send in progress */ |
b411b363 | 969 | |
b411b363 | 970 | int next_barrier_nr; |
b411b363 | 971 | struct list_head resync_reads; |
435f0740 LE |
972 | atomic_t pp_in_use; /* allocated from page pool */ |
973 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ | |
b411b363 PR |
974 | wait_queue_head_t ee_wait; |
975 | struct page *md_io_page; /* one page buffer for md_io */ | |
b411b363 PR |
976 | struct mutex md_io_mutex; /* protects the md_io_buffer */ |
977 | spinlock_t al_lock; | |
978 | wait_queue_head_t al_wait; | |
979 | struct lru_cache *act_log; /* activity log */ | |
980 | unsigned int al_tr_number; | |
981 | int al_tr_cycle; | |
982 | int al_tr_pos; /* position of the next transaction in the journal */ | |
b411b363 PR |
983 | wait_queue_head_t seq_wait; |
984 | atomic_t packet_seq; | |
985 | unsigned int peer_seq; | |
986 | spinlock_t peer_seq_lock; | |
987 | unsigned int minor; | |
988 | unsigned long comm_bm_set; /* communicated number of set bits. */ | |
b411b363 PR |
989 | struct bm_io_work bm_io_work; |
990 | u64 ed_uuid; /* UUID of the exposed data */ | |
8410da8f PR |
991 | struct mutex own_state_mutex; |
992 | struct mutex *state_mutex; /* either own_state_mutex or mdev->tconn->cstate_mutex */ | |
b411b363 | 993 | char congestion_reason; /* Why we where congested... */ |
1d7734a0 LE |
994 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
995 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ | |
996 | int rs_last_sect_ev; /* counter to compare with */ | |
997 | int rs_last_events; /* counter of read or write "events" (unit sectors) | |
998 | * on the lower level device when we last looked. */ | |
999 | int c_sync_rate; /* current resync rate after syncer throttle magic */ | |
778f271d PR |
1000 | struct fifo_buffer rs_plan_s; /* correction values of resync planer */ |
1001 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ | |
25985edc | 1002 | int rs_planed; /* resync sectors already planned */ |
759fbdfb | 1003 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
99432fcc PR |
1004 | int peer_max_bio_size; |
1005 | int local_max_bio_size; | |
b411b363 PR |
1006 | }; |
1007 | ||
1008 | static inline struct drbd_conf *minor_to_mdev(unsigned int minor) | |
1009 | { | |
81a5d60e | 1010 | return (struct drbd_conf *)idr_find(&minors, minor); |
b411b363 PR |
1011 | } |
1012 | ||
1013 | static inline unsigned int mdev_to_minor(struct drbd_conf *mdev) | |
1014 | { | |
1015 | return mdev->minor; | |
1016 | } | |
1017 | ||
eefc2f7d PR |
1018 | static inline struct drbd_conf *vnr_to_mdev(struct drbd_tconn *tconn, int vnr) |
1019 | { | |
1020 | return (struct drbd_conf *)idr_find(&tconn->volumes, vnr); | |
1021 | } | |
1022 | ||
b411b363 PR |
1023 | /* |
1024 | * function declarations | |
1025 | *************************/ | |
1026 | ||
1027 | /* drbd_main.c */ | |
1028 | ||
e89b591c PR |
1029 | enum dds_flags { |
1030 | DDSF_FORCED = 1, | |
1031 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ | |
1032 | }; | |
1033 | ||
b411b363 | 1034 | extern void drbd_init_set_defaults(struct drbd_conf *mdev); |
b411b363 PR |
1035 | extern int drbd_thread_start(struct drbd_thread *thi); |
1036 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); | |
392c8801 | 1037 | extern char *drbd_task_to_thread_name(struct drbd_tconn *tconn, struct task_struct *task); |
b411b363 | 1038 | #ifdef CONFIG_SMP |
80822284 PR |
1039 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
1040 | extern void drbd_calc_cpu_mask(struct drbd_tconn *tconn); | |
b411b363 | 1041 | #else |
80822284 | 1042 | #define drbd_thread_current_set_cpu(A) ({}) |
b411b363 PR |
1043 | #define drbd_calc_cpu_mask(A) ({}) |
1044 | #endif | |
1045 | extern void drbd_free_resources(struct drbd_conf *mdev); | |
2f5cdd0b | 1046 | extern void tl_release(struct drbd_tconn *, unsigned int barrier_nr, |
b411b363 | 1047 | unsigned int set_size); |
2f5cdd0b PR |
1048 | extern void tl_clear(struct drbd_tconn *); |
1049 | extern void _tl_add_barrier(struct drbd_tconn *, struct drbd_tl_epoch *); | |
360cc740 | 1050 | extern void drbd_free_sock(struct drbd_tconn *tconn); |
bedbd2a5 PR |
1051 | extern int drbd_send(struct drbd_tconn *tconn, struct socket *sock, |
1052 | void *buf, size_t size, unsigned msg_flags); | |
fb708e40 AG |
1053 | extern int drbd_send_all(struct drbd_tconn *, struct socket *, void *, size_t, |
1054 | unsigned); | |
1055 | ||
dc8228d1 | 1056 | extern int drbd_send_protocol(struct drbd_tconn *tconn); |
b411b363 PR |
1057 | extern int drbd_send_uuids(struct drbd_conf *mdev); |
1058 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_conf *mdev); | |
9c1b7f72 | 1059 | extern void drbd_gen_and_send_sync_uuid(struct drbd_conf *mdev); |
e89b591c | 1060 | extern int drbd_send_sizes(struct drbd_conf *mdev, int trigger_reply, enum dds_flags flags); |
cf29c9d8 PR |
1061 | extern int _conn_send_state_req(struct drbd_tconn *, int vnr, enum drbd_packet cmd, |
1062 | union drbd_state, union drbd_state); | |
b411b363 PR |
1063 | extern int _drbd_send_state(struct drbd_conf *mdev); |
1064 | extern int drbd_send_state(struct drbd_conf *mdev); | |
f399002e | 1065 | extern int drbd_send_sync_param(struct drbd_conf *mdev); |
d4e67d7c AG |
1066 | extern void drbd_send_b_ack(struct drbd_conf *mdev, u32 barrier_nr, |
1067 | u32 set_size); | |
f6ffca9f AG |
1068 | extern int drbd_send_ack(struct drbd_conf *, enum drbd_packet, |
1069 | struct drbd_peer_request *); | |
a9a9994d AG |
1070 | extern void drbd_send_ack_rp(struct drbd_conf *mdev, enum drbd_packet cmd, |
1071 | struct p_block_req *rp); | |
1072 | extern void drbd_send_ack_dp(struct drbd_conf *mdev, enum drbd_packet cmd, | |
1073 | struct p_data *dp, int data_size); | |
d8763023 | 1074 | extern int drbd_send_ack_ex(struct drbd_conf *mdev, enum drbd_packet cmd, |
b411b363 | 1075 | sector_t sector, int blksize, u64 block_id); |
8f7bed77 | 1076 | extern int drbd_send_out_of_sync(struct drbd_conf *, struct drbd_request *); |
f6ffca9f AG |
1077 | extern int drbd_send_block(struct drbd_conf *, enum drbd_packet, |
1078 | struct drbd_peer_request *); | |
b411b363 | 1079 | extern int drbd_send_dblock(struct drbd_conf *mdev, struct drbd_request *req); |
b411b363 PR |
1080 | extern int drbd_send_drequest(struct drbd_conf *mdev, int cmd, |
1081 | sector_t sector, int size, u64 block_id); | |
d8763023 AG |
1082 | extern int drbd_send_drequest_csum(struct drbd_conf *mdev, sector_t sector, |
1083 | int size, void *digest, int digest_size, | |
1084 | enum drbd_packet cmd); | |
b411b363 PR |
1085 | extern int drbd_send_ov_request(struct drbd_conf *mdev,sector_t sector,int size); |
1086 | ||
1087 | extern int drbd_send_bitmap(struct drbd_conf *mdev); | |
2f4e7abe | 1088 | extern void drbd_send_sr_reply(struct drbd_conf *mdev, enum drbd_state_rv retcode); |
9f5bdc33 | 1089 | extern void conn_send_sr_reply(struct drbd_tconn *tconn, enum drbd_state_rv retcode); |
b411b363 PR |
1090 | extern void drbd_free_bc(struct drbd_backing_dev *ldev); |
1091 | extern void drbd_mdev_cleanup(struct drbd_conf *mdev); | |
62b0da3a | 1092 | void drbd_print_uuids(struct drbd_conf *mdev, const char *text); |
b411b363 | 1093 | |
b411b363 PR |
1094 | extern void drbd_md_sync(struct drbd_conf *mdev); |
1095 | extern int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev); | |
b411b363 PR |
1096 | extern void drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); |
1097 | extern void _drbd_uuid_set(struct drbd_conf *mdev, int idx, u64 val) __must_hold(local); | |
1098 | extern void drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); | |
1099 | extern void _drbd_uuid_new_current(struct drbd_conf *mdev) __must_hold(local); | |
1100 | extern void drbd_uuid_set_bm(struct drbd_conf *mdev, u64 val) __must_hold(local); | |
1101 | extern void drbd_md_set_flag(struct drbd_conf *mdev, int flags) __must_hold(local); | |
1102 | extern void drbd_md_clear_flag(struct drbd_conf *mdev, int flags)__must_hold(local); | |
1103 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); | |
ee15b038 | 1104 | #ifndef DRBD_DEBUG_MD_SYNC |
b411b363 | 1105 | extern void drbd_md_mark_dirty(struct drbd_conf *mdev); |
ee15b038 LE |
1106 | #else |
1107 | #define drbd_md_mark_dirty(m) drbd_md_mark_dirty_(m, __LINE__ , __func__ ) | |
1108 | extern void drbd_md_mark_dirty_(struct drbd_conf *mdev, | |
1109 | unsigned int line, const char *func); | |
1110 | #endif | |
b411b363 PR |
1111 | extern void drbd_queue_bitmap_io(struct drbd_conf *mdev, |
1112 | int (*io_fn)(struct drbd_conf *), | |
1113 | void (*done)(struct drbd_conf *, int), | |
20ceb2b2 LE |
1114 | char *why, enum bm_flag flags); |
1115 | extern int drbd_bitmap_io(struct drbd_conf *mdev, | |
1116 | int (*io_fn)(struct drbd_conf *), | |
1117 | char *why, enum bm_flag flags); | |
b411b363 PR |
1118 | extern int drbd_bmio_set_n_write(struct drbd_conf *mdev); |
1119 | extern int drbd_bmio_clear_n_write(struct drbd_conf *mdev); | |
e9e6f3ec | 1120 | extern void drbd_go_diskless(struct drbd_conf *mdev); |
82f59cc6 | 1121 | extern void drbd_ldev_destroy(struct drbd_conf *mdev); |
b411b363 | 1122 | |
b411b363 PR |
1123 | /* Meta data layout |
1124 | We reserve a 128MB Block (4k aligned) | |
1125 | * either at the end of the backing device | |
3ad2f3fb | 1126 | * or on a separate meta data device. */ |
b411b363 | 1127 | |
b411b363 | 1128 | /* The following numbers are sectors */ |
7ad651b5 LE |
1129 | /* Allows up to about 3.8TB, so if you want more, |
1130 | * you need to use the "flexible" meta data format. */ | |
1131 | #define MD_RESERVED_SECT (128LU << 11) /* 128 MB, unit sectors */ | |
1132 | #define MD_AL_OFFSET 8 /* 8 Sectors after start of meta area */ | |
1133 | #define MD_AL_SECTORS 64 /* = 32 kB on disk activity log ring buffer */ | |
1134 | #define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS) | |
1135 | ||
1136 | /* we do all meta data IO in 4k blocks */ | |
1137 | #define MD_BLOCK_SHIFT 12 | |
1138 | #define MD_BLOCK_SIZE (1<<MD_BLOCK_SHIFT) | |
1139 | ||
1140 | /* One activity log extent represents 4M of storage */ | |
1141 | #define AL_EXTENT_SHIFT 22 | |
b411b363 PR |
1142 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
1143 | ||
7ad651b5 LE |
1144 | /* We could make these currently hardcoded constants configurable |
1145 | * variables at create-md time (or even re-configurable at runtime?). | |
1146 | * Which will require some more changes to the DRBD "super block" | |
1147 | * and attach code. | |
1148 | * | |
1149 | * updates per transaction: | |
1150 | * This many changes to the active set can be logged with one transaction. | |
1151 | * This number is arbitrary. | |
1152 | * context per transaction: | |
1153 | * This many context extent numbers are logged with each transaction. | |
1154 | * This number is resulting from the transaction block size (4k), the layout | |
1155 | * of the transaction header, and the number of updates per transaction. | |
1156 | * See drbd_actlog.c:struct al_transaction_on_disk | |
1157 | * */ | |
1158 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary | |
1159 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 | |
1160 | ||
b411b363 PR |
1161 | #if BITS_PER_LONG == 32 |
1162 | #define LN2_BPL 5 | |
1163 | #define cpu_to_lel(A) cpu_to_le32(A) | |
1164 | #define lel_to_cpu(A) le32_to_cpu(A) | |
1165 | #elif BITS_PER_LONG == 64 | |
1166 | #define LN2_BPL 6 | |
1167 | #define cpu_to_lel(A) cpu_to_le64(A) | |
1168 | #define lel_to_cpu(A) le64_to_cpu(A) | |
1169 | #else | |
1170 | #error "LN2 of BITS_PER_LONG unknown!" | |
1171 | #endif | |
1172 | ||
1173 | /* resync bitmap */ | |
1174 | /* 16MB sized 'bitmap extent' to track syncer usage */ | |
1175 | struct bm_extent { | |
1176 | int rs_left; /* number of bits set (out of sync) in this extent. */ | |
1177 | int rs_failed; /* number of failed resync requests in this extent. */ | |
1178 | unsigned long flags; | |
1179 | struct lc_element lce; | |
1180 | }; | |
1181 | ||
1182 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ | |
1183 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ | |
e3555d85 | 1184 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
b411b363 PR |
1185 | |
1186 | /* drbd_bitmap.c */ | |
1187 | /* | |
1188 | * We need to store one bit for a block. | |
1189 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. | |
1190 | * Bit 0 ==> local node thinks this block is binary identical on both nodes | |
1191 | * Bit 1 ==> local node thinks this block needs to be synced. | |
1192 | */ | |
1193 | ||
8e26f9cc PR |
1194 | #define SLEEP_TIME (HZ/10) |
1195 | ||
45dfffeb LE |
1196 | /* We do bitmap IO in units of 4k blocks. |
1197 | * We also still have a hardcoded 4k per bit relation. */ | |
1198 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ | |
b411b363 | 1199 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
45dfffeb LE |
1200 | /* mostly arbitrarily set the represented size of one bitmap extent, |
1201 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap | |
1202 | * at 4k per bit resolution) */ | |
1203 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ | |
b411b363 PR |
1204 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
1205 | ||
1206 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) | |
1207 | #error "HAVE YOU FIXED drbdmeta AS WELL??" | |
1208 | #endif | |
1209 | ||
1210 | /* thus many _storage_ sectors are described by one bit */ | |
1211 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) | |
1212 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) | |
1213 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) | |
1214 | ||
1215 | /* bit to represented kilo byte conversion */ | |
1216 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) | |
1217 | ||
1218 | /* in which _bitmap_ extent (resp. sector) the bit for a certain | |
1219 | * _storage_ sector is located in */ | |
1220 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) | |
1221 | ||
1222 | /* how much _storage_ sectors we have per bitmap sector */ | |
1223 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) | |
1224 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) | |
1225 | ||
1226 | /* in one sector of the bitmap, we have this many activity_log extents. */ | |
1227 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) | |
1228 | #define BM_WORDS_PER_AL_EXT (1 << (AL_EXTENT_SHIFT-BM_BLOCK_SHIFT-LN2_BPL)) | |
1229 | ||
1230 | #define BM_BLOCKS_PER_BM_EXT_B (BM_EXT_SHIFT - BM_BLOCK_SHIFT) | |
1231 | #define BM_BLOCKS_PER_BM_EXT_MASK ((1<<BM_BLOCKS_PER_BM_EXT_B) - 1) | |
1232 | ||
1233 | /* the extent in "PER_EXTENT" below is an activity log extent | |
1234 | * we need that many (long words/bytes) to store the bitmap | |
1235 | * of one AL_EXTENT_SIZE chunk of storage. | |
1236 | * we can store the bitmap for that many AL_EXTENTS within | |
1237 | * one sector of the _on_disk_ bitmap: | |
1238 | * bit 0 bit 37 bit 38 bit (512*8)-1 | |
1239 | * ...|........|........|.. // ..|........| | |
1240 | * sect. 0 `296 `304 ^(512*8*8)-1 | |
1241 | * | |
1242 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) | |
1243 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 | |
1244 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 | |
1245 | */ | |
1246 | ||
1247 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) | |
1248 | #define DRBD_MAX_SECTORS_BM \ | |
1249 | ((MD_RESERVED_SECT - MD_BM_OFFSET) * (1LL<<(BM_EXT_SHIFT-9))) | |
1250 | #if DRBD_MAX_SECTORS_BM < DRBD_MAX_SECTORS_32 | |
1251 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM | |
1252 | #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_BM | |
36bfc7e2 | 1253 | #elif !defined(CONFIG_LBDAF) && BITS_PER_LONG == 32 |
b411b363 PR |
1254 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_32 |
1255 | #define DRBD_MAX_SECTORS_FLEX DRBD_MAX_SECTORS_32 | |
1256 | #else | |
1257 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_BM | |
1258 | /* 16 TB in units of sectors */ | |
1259 | #if BITS_PER_LONG == 32 | |
1260 | /* adjust by one page worth of bitmap, | |
1261 | * so we won't wrap around in drbd_bm_find_next_bit. | |
1262 | * you should use 64bit OS for that much storage, anyways. */ | |
1263 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) | |
1264 | #else | |
4b0715f0 LE |
1265 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
1266 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) | |
1267 | /* corresponds to (1UL << 38) bits right now. */ | |
b411b363 PR |
1268 | #endif |
1269 | #endif | |
1270 | ||
23361cf3 LE |
1271 | /* BIO_MAX_SIZE is 256 * PAGE_CACHE_SIZE, |
1272 | * so for typical PAGE_CACHE_SIZE of 4k, that is (1<<20) Byte. | |
1273 | * Since we may live in a mixed-platform cluster, | |
1274 | * we limit us to a platform agnostic constant here for now. | |
1275 | * A followup commit may allow even bigger BIO sizes, | |
1276 | * once we thought that through. */ | |
1277 | #define DRBD_MAX_BIO_SIZE (1 << 20) | |
1278 | #if DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE | |
1279 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE | |
1280 | #endif | |
99432fcc | 1281 | #define DRBD_MAX_BIO_SIZE_SAFE (1 << 12) /* Works always = 4k */ |
b411b363 | 1282 | |
d5373389 | 1283 | #define DRBD_MAX_SIZE_H80_PACKET (1 << 15) /* The old header only allows packets up to 32Kib data */ |
b411b363 PR |
1284 | |
1285 | extern int drbd_bm_init(struct drbd_conf *mdev); | |
02d9a94b | 1286 | extern int drbd_bm_resize(struct drbd_conf *mdev, sector_t sectors, int set_new_bits); |
b411b363 PR |
1287 | extern void drbd_bm_cleanup(struct drbd_conf *mdev); |
1288 | extern void drbd_bm_set_all(struct drbd_conf *mdev); | |
1289 | extern void drbd_bm_clear_all(struct drbd_conf *mdev); | |
4b0715f0 | 1290 | /* set/clear/test only a few bits at a time */ |
b411b363 PR |
1291 | extern int drbd_bm_set_bits( |
1292 | struct drbd_conf *mdev, unsigned long s, unsigned long e); | |
1293 | extern int drbd_bm_clear_bits( | |
1294 | struct drbd_conf *mdev, unsigned long s, unsigned long e); | |
4b0715f0 LE |
1295 | extern int drbd_bm_count_bits( |
1296 | struct drbd_conf *mdev, const unsigned long s, const unsigned long e); | |
1297 | /* bm_set_bits variant for use while holding drbd_bm_lock, | |
1298 | * may process the whole bitmap in one go */ | |
b411b363 PR |
1299 | extern void _drbd_bm_set_bits(struct drbd_conf *mdev, |
1300 | const unsigned long s, const unsigned long e); | |
1301 | extern int drbd_bm_test_bit(struct drbd_conf *mdev, unsigned long bitnr); | |
1302 | extern int drbd_bm_e_weight(struct drbd_conf *mdev, unsigned long enr); | |
19f843aa | 1303 | extern int drbd_bm_write_page(struct drbd_conf *mdev, unsigned int idx) __must_hold(local); |
b411b363 | 1304 | extern int drbd_bm_read(struct drbd_conf *mdev) __must_hold(local); |
45dfffeb | 1305 | extern void drbd_bm_mark_for_writeout(struct drbd_conf *mdev, int page_nr); |
b411b363 | 1306 | extern int drbd_bm_write(struct drbd_conf *mdev) __must_hold(local); |
45dfffeb | 1307 | extern int drbd_bm_write_hinted(struct drbd_conf *mdev) __must_hold(local); |
b411b363 PR |
1308 | extern unsigned long drbd_bm_ALe_set_all(struct drbd_conf *mdev, |
1309 | unsigned long al_enr); | |
1310 | extern size_t drbd_bm_words(struct drbd_conf *mdev); | |
1311 | extern unsigned long drbd_bm_bits(struct drbd_conf *mdev); | |
1312 | extern sector_t drbd_bm_capacity(struct drbd_conf *mdev); | |
4b0715f0 LE |
1313 | |
1314 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) | |
b411b363 PR |
1315 | extern unsigned long drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); |
1316 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ | |
1317 | extern unsigned long _drbd_bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo); | |
1318 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_conf *mdev, unsigned long bm_fo); | |
0778286a | 1319 | extern unsigned long _drbd_bm_total_weight(struct drbd_conf *mdev); |
b411b363 PR |
1320 | extern unsigned long drbd_bm_total_weight(struct drbd_conf *mdev); |
1321 | extern int drbd_bm_rs_done(struct drbd_conf *mdev); | |
1322 | /* for receive_bitmap */ | |
1323 | extern void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, | |
1324 | size_t number, unsigned long *buffer); | |
19f843aa | 1325 | /* for _drbd_send_bitmap */ |
b411b363 PR |
1326 | extern void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, |
1327 | size_t number, unsigned long *buffer); | |
1328 | ||
20ceb2b2 | 1329 | extern void drbd_bm_lock(struct drbd_conf *mdev, char *why, enum bm_flag flags); |
b411b363 | 1330 | extern void drbd_bm_unlock(struct drbd_conf *mdev); |
b411b363 PR |
1331 | /* drbd_main.c */ |
1332 | ||
1333 | extern struct kmem_cache *drbd_request_cache; | |
6c852bec | 1334 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
b411b363 PR |
1335 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
1336 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
1337 | extern mempool_t *drbd_request_mempool; | |
1338 | extern mempool_t *drbd_ee_mempool; | |
1339 | ||
35abf594 LE |
1340 | /* drbd's page pool, used to buffer data received from the peer, |
1341 | * or data requested by the peer. | |
1342 | * | |
1343 | * This does not have an emergency reserve. | |
1344 | * | |
1345 | * When allocating from this pool, it first takes pages from the pool. | |
1346 | * Only if the pool is depleted will try to allocate from the system. | |
1347 | * | |
1348 | * The assumption is that pages taken from this pool will be processed, | |
1349 | * and given back, "quickly", and then can be recycled, so we can avoid | |
1350 | * frequent calls to alloc_page(), and still will be able to make progress even | |
1351 | * under memory pressure. | |
1352 | */ | |
1353 | extern struct page *drbd_pp_pool; | |
b411b363 PR |
1354 | extern spinlock_t drbd_pp_lock; |
1355 | extern int drbd_pp_vacant; | |
1356 | extern wait_queue_head_t drbd_pp_wait; | |
1357 | ||
35abf594 LE |
1358 | /* We also need a standard (emergency-reserve backed) page pool |
1359 | * for meta data IO (activity log, bitmap). | |
1360 | * We can keep it global, as long as it is used as "N pages at a time". | |
1361 | * 128 should be plenty, currently we probably can get away with as few as 1. | |
1362 | */ | |
1363 | #define DRBD_MIN_POOL_PAGES 128 | |
1364 | extern mempool_t *drbd_md_io_page_pool; | |
1365 | ||
da4a75d2 LE |
1366 | /* We also need to make sure we get a bio |
1367 | * when we need it for housekeeping purposes */ | |
1368 | extern struct bio_set *drbd_md_io_bio_set; | |
1369 | /* to allocate from that set */ | |
1370 | extern struct bio *bio_alloc_drbd(gfp_t gfp_mask); | |
1371 | ||
b411b363 PR |
1372 | extern rwlock_t global_state_lock; |
1373 | ||
80883197 | 1374 | extern int conn_lowest_minor(struct drbd_tconn *tconn); |
774b3055 | 1375 | enum drbd_ret_code conn_new_minor(struct drbd_tconn *tconn, unsigned int minor, int vnr); |
ff370e5a | 1376 | extern void drbd_delete_device(struct drbd_conf *mdev); |
b411b363 | 1377 | |
3b98c0c2 | 1378 | struct drbd_tconn *drbd_new_tconn(const char *name); |
2111438b | 1379 | extern void drbd_free_tconn(struct drbd_tconn *tconn); |
1aba4d7f | 1380 | struct drbd_tconn *conn_by_name(const char *name); |
91fd4dad | 1381 | extern void conn_free_crypto(struct drbd_tconn *tconn); |
2111438b | 1382 | |
b411b363 PR |
1383 | extern int proc_details; |
1384 | ||
1385 | /* drbd_req */ | |
7be8da07 | 1386 | extern int __drbd_make_request(struct drbd_conf *, struct bio *, unsigned long); |
2f58dcfc | 1387 | extern int drbd_make_request(struct request_queue *q, struct bio *bio); |
b411b363 PR |
1388 | extern int drbd_read_remote(struct drbd_conf *mdev, struct drbd_request *req); |
1389 | extern int drbd_merge_bvec(struct request_queue *q, struct bvec_merge_data *bvm, struct bio_vec *bvec); | |
1390 | extern int is_valid_ar_handle(struct drbd_request *, sector_t); | |
1391 | ||
1392 | ||
1393 | /* drbd_nl.c */ | |
8432b314 | 1394 | extern int drbd_msg_put_info(const char *info); |
b411b363 PR |
1395 | extern void drbd_suspend_io(struct drbd_conf *mdev); |
1396 | extern void drbd_resume_io(struct drbd_conf *mdev); | |
1397 | extern char *ppsize(char *buf, unsigned long long size); | |
a393db6f | 1398 | extern sector_t drbd_new_dev_size(struct drbd_conf *, struct drbd_backing_dev *, int); |
b411b363 | 1399 | enum determine_dev_size { dev_size_error = -1, unchanged = 0, shrunk = 1, grew = 2 }; |
24c4830c | 1400 | extern enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *, enum dds_flags) __must_hold(local); |
b411b363 | 1401 | extern void resync_after_online_grow(struct drbd_conf *); |
99432fcc | 1402 | extern void drbd_reconsider_max_bio_size(struct drbd_conf *mdev); |
bf885f8a AG |
1403 | extern enum drbd_state_rv drbd_set_role(struct drbd_conf *mdev, |
1404 | enum drbd_role new_role, | |
1405 | int force); | |
cb703454 PR |
1406 | extern bool conn_try_outdate_peer(struct drbd_tconn *tconn); |
1407 | extern void conn_try_outdate_peer_async(struct drbd_tconn *tconn); | |
b411b363 PR |
1408 | extern int drbd_khelper(struct drbd_conf *mdev, char *cmd); |
1409 | ||
1410 | /* drbd_worker.c */ | |
1411 | extern int drbd_worker(struct drbd_thread *thi); | |
1412 | extern int drbd_alter_sa(struct drbd_conf *mdev, int na); | |
1413 | extern void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side); | |
1414 | extern void resume_next_sg(struct drbd_conf *mdev); | |
1415 | extern void suspend_other_sg(struct drbd_conf *mdev); | |
1416 | extern int drbd_resync_finished(struct drbd_conf *mdev); | |
1417 | /* maybe rather drbd_main.c ? */ | |
1418 | extern int drbd_md_sync_page_io(struct drbd_conf *mdev, | |
1419 | struct drbd_backing_dev *bdev, sector_t sector, int rw); | |
8f7bed77 | 1420 | extern void drbd_ov_out_of_sync_found(struct drbd_conf *, sector_t, int); |
9bd28d3c | 1421 | extern void drbd_rs_controller_reset(struct drbd_conf *mdev); |
b411b363 | 1422 | |
8f7bed77 | 1423 | static inline void ov_out_of_sync_print(struct drbd_conf *mdev) |
b411b363 PR |
1424 | { |
1425 | if (mdev->ov_last_oos_size) { | |
1426 | dev_err(DEV, "Out of sync: start=%llu, size=%lu (sectors)\n", | |
1427 | (unsigned long long)mdev->ov_last_oos_start, | |
1428 | (unsigned long)mdev->ov_last_oos_size); | |
1429 | } | |
1430 | mdev->ov_last_oos_size=0; | |
1431 | } | |
1432 | ||
1433 | ||
45bb912b | 1434 | extern void drbd_csum_bio(struct drbd_conf *, struct crypto_hash *, struct bio *, void *); |
f6ffca9f AG |
1435 | extern void drbd_csum_ee(struct drbd_conf *, struct crypto_hash *, |
1436 | struct drbd_peer_request *, void *); | |
b411b363 | 1437 | /* worker callbacks */ |
99920dc5 AG |
1438 | extern int w_read_retry_remote(struct drbd_work *, int); |
1439 | extern int w_e_end_data_req(struct drbd_work *, int); | |
1440 | extern int w_e_end_rsdata_req(struct drbd_work *, int); | |
1441 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); | |
1442 | extern int w_e_end_ov_reply(struct drbd_work *, int); | |
1443 | extern int w_e_end_ov_req(struct drbd_work *, int); | |
1444 | extern int w_ov_finished(struct drbd_work *, int); | |
1445 | extern int w_resync_timer(struct drbd_work *, int); | |
1446 | extern int w_send_write_hint(struct drbd_work *, int); | |
1447 | extern int w_make_resync_request(struct drbd_work *, int); | |
1448 | extern int w_send_dblock(struct drbd_work *, int); | |
1449 | extern int w_send_barrier(struct drbd_work *, int); | |
1450 | extern int w_send_read_req(struct drbd_work *, int); | |
1451 | extern int w_prev_work_done(struct drbd_work *, int); | |
1452 | extern int w_e_reissue(struct drbd_work *, int); | |
1453 | extern int w_restart_disk_io(struct drbd_work *, int); | |
8f7bed77 | 1454 | extern int w_send_out_of_sync(struct drbd_work *, int); |
99920dc5 | 1455 | extern int w_start_resync(struct drbd_work *, int); |
b411b363 PR |
1456 | |
1457 | extern void resync_timer_fn(unsigned long data); | |
370a43e7 | 1458 | extern void start_resync_timer_fn(unsigned long data); |
b411b363 PR |
1459 | |
1460 | /* drbd_receiver.c */ | |
e3555d85 | 1461 | extern int drbd_rs_should_slow_down(struct drbd_conf *mdev, sector_t sector); |
fbe29dec AG |
1462 | extern int drbd_submit_peer_request(struct drbd_conf *, |
1463 | struct drbd_peer_request *, const unsigned, | |
1464 | const int); | |
7721f567 | 1465 | extern int drbd_free_peer_reqs(struct drbd_conf *, struct list_head *); |
0db55363 AG |
1466 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_conf *, u64, |
1467 | sector_t, unsigned int, | |
1468 | gfp_t) __must_hold(local); | |
3967deb1 AG |
1469 | extern void __drbd_free_peer_req(struct drbd_conf *, struct drbd_peer_request *, |
1470 | int); | |
1471 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) | |
1472 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) | |
c37c8ecf | 1473 | extern struct page *drbd_alloc_pages(struct drbd_conf *, unsigned int, bool); |
b411b363 PR |
1474 | extern void drbd_set_recv_tcq(struct drbd_conf *mdev, int tcq_enabled); |
1475 | extern void _drbd_clear_done_ee(struct drbd_conf *mdev, struct list_head *to_be_freed); | |
0e29d163 | 1476 | extern void conn_flush_workqueue(struct drbd_tconn *tconn); |
2325eb66 | 1477 | extern int drbd_connected(int vnr, void *p, void *data); |
0e29d163 PR |
1478 | static inline void drbd_flush_workqueue(struct drbd_conf *mdev) |
1479 | { | |
1480 | conn_flush_workqueue(mdev->tconn); | |
1481 | } | |
b411b363 PR |
1482 | |
1483 | /* yes, there is kernel_setsockopt, but only since 2.6.18. we don't need to | |
1484 | * mess with get_fs/set_fs, we know we are KERNEL_DS always. */ | |
1485 | static inline int drbd_setsockopt(struct socket *sock, int level, int optname, | |
1486 | char __user *optval, int optlen) | |
1487 | { | |
1488 | int err; | |
1489 | if (level == SOL_SOCKET) | |
1490 | err = sock_setsockopt(sock, level, optname, optval, optlen); | |
1491 | else | |
1492 | err = sock->ops->setsockopt(sock, level, optname, optval, | |
1493 | optlen); | |
1494 | return err; | |
1495 | } | |
1496 | ||
1497 | static inline void drbd_tcp_cork(struct socket *sock) | |
1498 | { | |
1499 | int __user val = 1; | |
1500 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, | |
1501 | (char __user *)&val, sizeof(val)); | |
1502 | } | |
1503 | ||
1504 | static inline void drbd_tcp_uncork(struct socket *sock) | |
1505 | { | |
1506 | int __user val = 0; | |
1507 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_CORK, | |
1508 | (char __user *)&val, sizeof(val)); | |
1509 | } | |
1510 | ||
1511 | static inline void drbd_tcp_nodelay(struct socket *sock) | |
1512 | { | |
1513 | int __user val = 1; | |
1514 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_NODELAY, | |
1515 | (char __user *)&val, sizeof(val)); | |
1516 | } | |
1517 | ||
1518 | static inline void drbd_tcp_quickack(struct socket *sock) | |
1519 | { | |
344fa462 | 1520 | int __user val = 2; |
b411b363 PR |
1521 | (void) drbd_setsockopt(sock, SOL_TCP, TCP_QUICKACK, |
1522 | (char __user *)&val, sizeof(val)); | |
1523 | } | |
1524 | ||
1525 | void drbd_bump_write_ordering(struct drbd_conf *mdev, enum write_ordering_e wo); | |
1526 | ||
1527 | /* drbd_proc.c */ | |
1528 | extern struct proc_dir_entry *drbd_proc; | |
7d4e9d09 | 1529 | extern const struct file_operations drbd_proc_fops; |
b411b363 PR |
1530 | extern const char *drbd_conn_str(enum drbd_conns s); |
1531 | extern const char *drbd_role_str(enum drbd_role s); | |
1532 | ||
1533 | /* drbd_actlog.c */ | |
181286ad LE |
1534 | extern void drbd_al_begin_io(struct drbd_conf *mdev, struct drbd_interval *i); |
1535 | extern void drbd_al_complete_io(struct drbd_conf *mdev, struct drbd_interval *i); | |
b411b363 PR |
1536 | extern void drbd_rs_complete_io(struct drbd_conf *mdev, sector_t sector); |
1537 | extern int drbd_rs_begin_io(struct drbd_conf *mdev, sector_t sector); | |
1538 | extern int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector); | |
1539 | extern void drbd_rs_cancel_all(struct drbd_conf *mdev); | |
1540 | extern int drbd_rs_del_all(struct drbd_conf *mdev); | |
1541 | extern void drbd_rs_failed_io(struct drbd_conf *mdev, | |
1542 | sector_t sector, int size); | |
1543 | extern int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *); | |
ea5442af | 1544 | extern void drbd_advance_rs_marks(struct drbd_conf *mdev, unsigned long still_to_go); |
b411b363 PR |
1545 | extern void __drbd_set_in_sync(struct drbd_conf *mdev, sector_t sector, |
1546 | int size, const char *file, const unsigned int line); | |
1547 | #define drbd_set_in_sync(mdev, sector, size) \ | |
1548 | __drbd_set_in_sync(mdev, sector, size, __FILE__, __LINE__) | |
73a01a18 | 1549 | extern int __drbd_set_out_of_sync(struct drbd_conf *mdev, sector_t sector, |
b411b363 PR |
1550 | int size, const char *file, const unsigned int line); |
1551 | #define drbd_set_out_of_sync(mdev, sector, size) \ | |
1552 | __drbd_set_out_of_sync(mdev, sector, size, __FILE__, __LINE__) | |
1553 | extern void drbd_al_apply_to_bm(struct drbd_conf *mdev); | |
b411b363 PR |
1554 | extern void drbd_al_shrink(struct drbd_conf *mdev); |
1555 | ||
b411b363 | 1556 | /* drbd_nl.c */ |
3b98c0c2 LE |
1557 | /* state info broadcast */ |
1558 | struct sib_info { | |
1559 | enum drbd_state_info_bcast_reason sib_reason; | |
1560 | union { | |
1561 | struct { | |
1562 | char *helper_name; | |
1563 | unsigned helper_exit_code; | |
1564 | }; | |
1565 | struct { | |
1566 | union drbd_state os; | |
1567 | union drbd_state ns; | |
1568 | }; | |
1569 | }; | |
1570 | }; | |
1571 | void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib); | |
b411b363 | 1572 | |
b411b363 PR |
1573 | /* |
1574 | * inline helper functions | |
1575 | *************************/ | |
1576 | ||
45bb912b LE |
1577 | /* see also page_chain_add and friends in drbd_receiver.c */ |
1578 | static inline struct page *page_chain_next(struct page *page) | |
1579 | { | |
1580 | return (struct page *)page_private(page); | |
1581 | } | |
1582 | #define page_chain_for_each(page) \ | |
1583 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ | |
1584 | page = page_chain_next(page)) | |
1585 | #define page_chain_for_each_safe(page, n) \ | |
1586 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) | |
1587 | ||
1588 | static inline int drbd_bio_has_active_page(struct bio *bio) | |
1589 | { | |
1590 | struct bio_vec *bvec; | |
1591 | int i; | |
1592 | ||
1593 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1594 | if (page_count(bvec->bv_page) > 1) | |
1595 | return 1; | |
1596 | } | |
1597 | ||
1598 | return 0; | |
1599 | } | |
1600 | ||
045417f7 | 1601 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
45bb912b | 1602 | { |
db830c46 | 1603 | struct page *page = peer_req->pages; |
45bb912b LE |
1604 | page_chain_for_each(page) { |
1605 | if (page_count(page) > 1) | |
1606 | return 1; | |
1607 | } | |
1608 | return 0; | |
1609 | } | |
1610 | ||
bf885f8a AG |
1611 | static inline enum drbd_state_rv |
1612 | _drbd_set_state(struct drbd_conf *mdev, union drbd_state ns, | |
1613 | enum chg_state_flags flags, struct completion *done) | |
b411b363 | 1614 | { |
bf885f8a | 1615 | enum drbd_state_rv rv; |
b411b363 PR |
1616 | |
1617 | read_lock(&global_state_lock); | |
1618 | rv = __drbd_set_state(mdev, ns, flags, done); | |
1619 | read_unlock(&global_state_lock); | |
1620 | ||
1621 | return rv; | |
1622 | } | |
1623 | ||
78bae59b PR |
1624 | static inline union drbd_state drbd_read_state(struct drbd_conf *mdev) |
1625 | { | |
1626 | union drbd_state rv; | |
1627 | ||
da9fbc27 | 1628 | rv.i = mdev->state.i; |
8e0af25f PR |
1629 | rv.susp = mdev->tconn->susp; |
1630 | rv.susp_nod = mdev->tconn->susp_nod; | |
1631 | rv.susp_fen = mdev->tconn->susp_fen; | |
78bae59b PR |
1632 | |
1633 | return rv; | |
1634 | } | |
1635 | ||
b411b363 PR |
1636 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
1637 | static inline void __drbd_chk_io_error_(struct drbd_conf *mdev, int forcedetach, const char *where) | |
1638 | { | |
1639 | switch (mdev->ldev->dc.on_io_error) { | |
1640 | case EP_PASS_ON: | |
1641 | if (!forcedetach) { | |
7383506c | 1642 | if (__ratelimit(&drbd_ratelimit_state)) |
82f59cc6 | 1643 | dev_err(DEV, "Local IO failed in %s.\n", where); |
d2e17807 PR |
1644 | if (mdev->state.disk > D_INCONSISTENT) |
1645 | _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_HARD, NULL); | |
b411b363 PR |
1646 | break; |
1647 | } | |
1648 | /* NOTE fall through to detach case if forcedetach set */ | |
1649 | case EP_DETACH: | |
1650 | case EP_CALL_HELPER: | |
82f59cc6 | 1651 | set_bit(WAS_IO_ERROR, &mdev->flags); |
b411b363 PR |
1652 | if (mdev->state.disk > D_FAILED) { |
1653 | _drbd_set_state(_NS(mdev, disk, D_FAILED), CS_HARD, NULL); | |
82f59cc6 LE |
1654 | dev_err(DEV, |
1655 | "Local IO failed in %s. Detaching...\n", where); | |
b411b363 PR |
1656 | } |
1657 | break; | |
1658 | } | |
1659 | } | |
1660 | ||
1661 | /** | |
1662 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers | |
1663 | * @mdev: DRBD device. | |
1664 | * @error: Error code passed to the IO completion callback | |
1665 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data | |
1666 | * | |
1667 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) | |
1668 | */ | |
1669 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) | |
1670 | static inline void drbd_chk_io_error_(struct drbd_conf *mdev, | |
1671 | int error, int forcedetach, const char *where) | |
1672 | { | |
1673 | if (error) { | |
1674 | unsigned long flags; | |
87eeee41 | 1675 | spin_lock_irqsave(&mdev->tconn->req_lock, flags); |
b411b363 | 1676 | __drbd_chk_io_error_(mdev, forcedetach, where); |
87eeee41 | 1677 | spin_unlock_irqrestore(&mdev->tconn->req_lock, flags); |
b411b363 PR |
1678 | } |
1679 | } | |
1680 | ||
1681 | ||
1682 | /** | |
1683 | * drbd_md_first_sector() - Returns the first sector number of the meta data area | |
1684 | * @bdev: Meta data block device. | |
1685 | * | |
1686 | * BTW, for internal meta data, this happens to be the maximum capacity | |
1687 | * we could agree upon with our peer node. | |
1688 | */ | |
1689 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) | |
1690 | { | |
1691 | switch (bdev->dc.meta_dev_idx) { | |
1692 | case DRBD_MD_INDEX_INTERNAL: | |
1693 | case DRBD_MD_INDEX_FLEX_INT: | |
1694 | return bdev->md.md_offset + bdev->md.bm_offset; | |
1695 | case DRBD_MD_INDEX_FLEX_EXT: | |
1696 | default: | |
1697 | return bdev->md.md_offset; | |
1698 | } | |
1699 | } | |
1700 | ||
1701 | /** | |
1702 | * drbd_md_last_sector() - Return the last sector number of the meta data area | |
1703 | * @bdev: Meta data block device. | |
1704 | */ | |
1705 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) | |
1706 | { | |
1707 | switch (bdev->dc.meta_dev_idx) { | |
1708 | case DRBD_MD_INDEX_INTERNAL: | |
1709 | case DRBD_MD_INDEX_FLEX_INT: | |
1710 | return bdev->md.md_offset + MD_AL_OFFSET - 1; | |
1711 | case DRBD_MD_INDEX_FLEX_EXT: | |
1712 | default: | |
1713 | return bdev->md.md_offset + bdev->md.md_size_sect; | |
1714 | } | |
1715 | } | |
1716 | ||
1717 | /* Returns the number of 512 byte sectors of the device */ | |
1718 | static inline sector_t drbd_get_capacity(struct block_device *bdev) | |
1719 | { | |
1720 | /* return bdev ? get_capacity(bdev->bd_disk) : 0; */ | |
77304d2a | 1721 | return bdev ? i_size_read(bdev->bd_inode) >> 9 : 0; |
b411b363 PR |
1722 | } |
1723 | ||
1724 | /** | |
1725 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer | |
1726 | * @bdev: Meta data block device. | |
1727 | * | |
1728 | * returns the capacity we announce to out peer. we clip ourselves at the | |
1729 | * various MAX_SECTORS, because if we don't, current implementation will | |
1730 | * oops sooner or later | |
1731 | */ | |
1732 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) | |
1733 | { | |
1734 | sector_t s; | |
1735 | switch (bdev->dc.meta_dev_idx) { | |
1736 | case DRBD_MD_INDEX_INTERNAL: | |
1737 | case DRBD_MD_INDEX_FLEX_INT: | |
1738 | s = drbd_get_capacity(bdev->backing_bdev) | |
1739 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1740 | drbd_md_first_sector(bdev)) | |
1741 | : 0; | |
1742 | break; | |
1743 | case DRBD_MD_INDEX_FLEX_EXT: | |
1744 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1745 | drbd_get_capacity(bdev->backing_bdev)); | |
1746 | /* clip at maximum size the meta device can support */ | |
1747 | s = min_t(sector_t, s, | |
1748 | BM_EXT_TO_SECT(bdev->md.md_size_sect | |
1749 | - bdev->md.bm_offset)); | |
1750 | break; | |
1751 | default: | |
1752 | s = min_t(sector_t, DRBD_MAX_SECTORS, | |
1753 | drbd_get_capacity(bdev->backing_bdev)); | |
1754 | } | |
1755 | return s; | |
1756 | } | |
1757 | ||
1758 | /** | |
1759 | * drbd_md_ss__() - Return the sector number of our meta data super block | |
1760 | * @mdev: DRBD device. | |
1761 | * @bdev: Meta data block device. | |
1762 | */ | |
1763 | static inline sector_t drbd_md_ss__(struct drbd_conf *mdev, | |
1764 | struct drbd_backing_dev *bdev) | |
1765 | { | |
1766 | switch (bdev->dc.meta_dev_idx) { | |
1767 | default: /* external, some index */ | |
1768 | return MD_RESERVED_SECT * bdev->dc.meta_dev_idx; | |
1769 | case DRBD_MD_INDEX_INTERNAL: | |
1770 | /* with drbd08, internal meta data is always "flexible" */ | |
1771 | case DRBD_MD_INDEX_FLEX_INT: | |
1772 | /* sizeof(struct md_on_disk_07) == 4k | |
1773 | * position: last 4k aligned block of 4k size */ | |
1774 | if (!bdev->backing_bdev) { | |
1775 | if (__ratelimit(&drbd_ratelimit_state)) { | |
1776 | dev_err(DEV, "bdev->backing_bdev==NULL\n"); | |
1777 | dump_stack(); | |
1778 | } | |
1779 | return 0; | |
1780 | } | |
1781 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) | |
1782 | - MD_AL_OFFSET; | |
1783 | case DRBD_MD_INDEX_FLEX_EXT: | |
1784 | return 0; | |
1785 | } | |
1786 | } | |
1787 | ||
b411b363 PR |
1788 | static inline void |
1789 | drbd_queue_work_front(struct drbd_work_queue *q, struct drbd_work *w) | |
1790 | { | |
1791 | unsigned long flags; | |
1792 | spin_lock_irqsave(&q->q_lock, flags); | |
1793 | list_add(&w->list, &q->q); | |
1794 | up(&q->s); /* within the spinlock, | |
1795 | see comment near end of drbd_worker() */ | |
1796 | spin_unlock_irqrestore(&q->q_lock, flags); | |
1797 | } | |
1798 | ||
1799 | static inline void | |
1800 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) | |
1801 | { | |
1802 | unsigned long flags; | |
1803 | spin_lock_irqsave(&q->q_lock, flags); | |
1804 | list_add_tail(&w->list, &q->q); | |
1805 | up(&q->s); /* within the spinlock, | |
1806 | see comment near end of drbd_worker() */ | |
1807 | spin_unlock_irqrestore(&q->q_lock, flags); | |
1808 | } | |
1809 | ||
0625ac19 | 1810 | static inline void wake_asender(struct drbd_tconn *tconn) |
b411b363 | 1811 | { |
0625ac19 PR |
1812 | if (test_bit(SIGNAL_ASENDER, &tconn->flags)) |
1813 | force_sig(DRBD_SIG, tconn->asender.task); | |
b411b363 PR |
1814 | } |
1815 | ||
0625ac19 | 1816 | static inline void request_ping(struct drbd_tconn *tconn) |
b411b363 | 1817 | { |
0625ac19 PR |
1818 | set_bit(SEND_PING, &tconn->flags); |
1819 | wake_asender(tconn); | |
b411b363 PR |
1820 | } |
1821 | ||
dba58587 AG |
1822 | extern void *conn_prepare_command(struct drbd_tconn *, struct drbd_socket *); |
1823 | extern void *drbd_prepare_command(struct drbd_conf *, struct drbd_socket *); | |
1824 | extern int conn_send_command(struct drbd_tconn *, struct drbd_socket *, | |
1825 | enum drbd_packet, unsigned int, void *, | |
1826 | unsigned int); | |
1827 | extern int drbd_send_command(struct drbd_conf *, struct drbd_socket *, | |
1828 | enum drbd_packet, unsigned int, void *, | |
1829 | unsigned int); | |
1830 | ||
e307f352 AG |
1831 | extern int drbd_send_ping(struct drbd_tconn *tconn); |
1832 | extern int drbd_send_ping_ack(struct drbd_tconn *tconn); | |
9f5bdc33 AG |
1833 | extern int drbd_send_state_req(struct drbd_conf *, union drbd_state, union drbd_state); |
1834 | extern int conn_send_state_req(struct drbd_tconn *, union drbd_state, union drbd_state); | |
cf29c9d8 | 1835 | |
b411b363 PR |
1836 | static inline void drbd_thread_stop(struct drbd_thread *thi) |
1837 | { | |
81e84650 | 1838 | _drbd_thread_stop(thi, false, true); |
b411b363 PR |
1839 | } |
1840 | ||
1841 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) | |
1842 | { | |
81e84650 | 1843 | _drbd_thread_stop(thi, false, false); |
b411b363 PR |
1844 | } |
1845 | ||
1846 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) | |
1847 | { | |
81e84650 | 1848 | _drbd_thread_stop(thi, true, false); |
b411b363 PR |
1849 | } |
1850 | ||
1851 | /* counts how many answer packets packets we expect from our peer, | |
1852 | * for either explicit application requests, | |
1853 | * or implicit barrier packets as necessary. | |
1854 | * increased: | |
1855 | * w_send_barrier | |
8554df1c | 1856 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
b411b363 PR |
1857 | * it is much easier and equally valid to count what we queue for the |
1858 | * worker, even before it actually was queued or send. | |
1859 | * (drbd_make_request_common; recovery path on read io-error) | |
1860 | * decreased: | |
1861 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) | |
8554df1c | 1862 | * _req_mod(req, DATA_RECEIVED) |
b411b363 | 1863 | * [from receive_DataReply] |
8554df1c | 1864 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
b411b363 PR |
1865 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
1866 | * for some reason it is NOT decreased in got_NegAck, | |
1867 | * but in the resulting cleanup code from report_params. | |
1868 | * we should try to remember the reason for that... | |
8554df1c AG |
1869 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
1870 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) | |
b411b363 PR |
1871 | * [from tl_clear_barrier] |
1872 | */ | |
1873 | static inline void inc_ap_pending(struct drbd_conf *mdev) | |
1874 | { | |
1875 | atomic_inc(&mdev->ap_pending_cnt); | |
1876 | } | |
1877 | ||
49559d87 PR |
1878 | #define ERR_IF_CNT_IS_NEGATIVE(which, func, line) \ |
1879 | if (atomic_read(&mdev->which) < 0) \ | |
b411b363 | 1880 | dev_err(DEV, "in %s:%d: " #which " = %d < 0 !\n", \ |
49559d87 PR |
1881 | func, line, \ |
1882 | atomic_read(&mdev->which)) | |
b411b363 | 1883 | |
49559d87 PR |
1884 | #define dec_ap_pending(mdev) _dec_ap_pending(mdev, __FUNCTION__, __LINE__) |
1885 | static inline void _dec_ap_pending(struct drbd_conf *mdev, const char *func, int line) | |
1886 | { | |
1887 | if (atomic_dec_and_test(&mdev->ap_pending_cnt)) | |
1888 | wake_up(&mdev->misc_wait); | |
1889 | ERR_IF_CNT_IS_NEGATIVE(ap_pending_cnt, func, line); | |
1890 | } | |
b411b363 PR |
1891 | |
1892 | /* counts how many resync-related answers we still expect from the peer | |
1893 | * increase decrease | |
1894 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) | |
25985edc | 1895 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
b411b363 PR |
1896 | * (or P_NEG_ACK with ID_SYNCER) |
1897 | */ | |
1898 | static inline void inc_rs_pending(struct drbd_conf *mdev) | |
1899 | { | |
1900 | atomic_inc(&mdev->rs_pending_cnt); | |
1901 | } | |
1902 | ||
49559d87 PR |
1903 | #define dec_rs_pending(mdev) _dec_rs_pending(mdev, __FUNCTION__, __LINE__) |
1904 | static inline void _dec_rs_pending(struct drbd_conf *mdev, const char *func, int line) | |
1905 | { | |
1906 | atomic_dec(&mdev->rs_pending_cnt); | |
1907 | ERR_IF_CNT_IS_NEGATIVE(rs_pending_cnt, func, line); | |
1908 | } | |
b411b363 PR |
1909 | |
1910 | /* counts how many answers we still need to send to the peer. | |
1911 | * increased on | |
1912 | * receive_Data unless protocol A; | |
1913 | * we need to send a P_RECV_ACK (proto B) | |
1914 | * or P_WRITE_ACK (proto C) | |
1915 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK | |
1916 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA | |
1917 | * receive_Barrier_* we need to send a P_BARRIER_ACK | |
1918 | */ | |
1919 | static inline void inc_unacked(struct drbd_conf *mdev) | |
1920 | { | |
1921 | atomic_inc(&mdev->unacked_cnt); | |
1922 | } | |
1923 | ||
49559d87 PR |
1924 | #define dec_unacked(mdev) _dec_unacked(mdev, __FUNCTION__, __LINE__) |
1925 | static inline void _dec_unacked(struct drbd_conf *mdev, const char *func, int line) | |
1926 | { | |
1927 | atomic_dec(&mdev->unacked_cnt); | |
1928 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); | |
1929 | } | |
b411b363 | 1930 | |
49559d87 PR |
1931 | #define sub_unacked(mdev, n) _sub_unacked(mdev, n, __FUNCTION__, __LINE__) |
1932 | static inline void _sub_unacked(struct drbd_conf *mdev, int n, const char *func, int line) | |
1933 | { | |
1934 | atomic_sub(n, &mdev->unacked_cnt); | |
1935 | ERR_IF_CNT_IS_NEGATIVE(unacked_cnt, func, line); | |
1936 | } | |
b411b363 | 1937 | |
b411b363 PR |
1938 | /** |
1939 | * get_ldev() - Increase the ref count on mdev->ldev. Returns 0 if there is no ldev | |
1940 | * @M: DRBD device. | |
1941 | * | |
1942 | * You have to call put_ldev() when finished working with mdev->ldev. | |
1943 | */ | |
1944 | #define get_ldev(M) __cond_lock(local, _get_ldev_if_state(M,D_INCONSISTENT)) | |
1945 | #define get_ldev_if_state(M,MINS) __cond_lock(local, _get_ldev_if_state(M,MINS)) | |
1946 | ||
1947 | static inline void put_ldev(struct drbd_conf *mdev) | |
1948 | { | |
1d7734a0 | 1949 | int i = atomic_dec_return(&mdev->local_cnt); |
9a0d9d03 LE |
1950 | |
1951 | /* This may be called from some endio handler, | |
1952 | * so we must not sleep here. */ | |
1953 | ||
b411b363 | 1954 | __release(local); |
1d7734a0 | 1955 | D_ASSERT(i >= 0); |
e9e6f3ec | 1956 | if (i == 0) { |
82f59cc6 LE |
1957 | if (mdev->state.disk == D_DISKLESS) |
1958 | /* even internal references gone, safe to destroy */ | |
1959 | drbd_ldev_destroy(mdev); | |
e9e6f3ec | 1960 | if (mdev->state.disk == D_FAILED) |
82f59cc6 | 1961 | /* all application IO references gone. */ |
e9e6f3ec | 1962 | drbd_go_diskless(mdev); |
b411b363 | 1963 | wake_up(&mdev->misc_wait); |
e9e6f3ec | 1964 | } |
b411b363 PR |
1965 | } |
1966 | ||
1967 | #ifndef __CHECKER__ | |
1968 | static inline int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins) | |
1969 | { | |
1970 | int io_allowed; | |
1971 | ||
82f59cc6 LE |
1972 | /* never get a reference while D_DISKLESS */ |
1973 | if (mdev->state.disk == D_DISKLESS) | |
1974 | return 0; | |
1975 | ||
b411b363 PR |
1976 | atomic_inc(&mdev->local_cnt); |
1977 | io_allowed = (mdev->state.disk >= mins); | |
1978 | if (!io_allowed) | |
1979 | put_ldev(mdev); | |
1980 | return io_allowed; | |
1981 | } | |
1982 | #else | |
1983 | extern int _get_ldev_if_state(struct drbd_conf *mdev, enum drbd_disk_state mins); | |
1984 | #endif | |
1985 | ||
1986 | /* you must have an "get_ldev" reference */ | |
1987 | static inline void drbd_get_syncer_progress(struct drbd_conf *mdev, | |
1988 | unsigned long *bits_left, unsigned int *per_mil_done) | |
1989 | { | |
4b0715f0 LE |
1990 | /* this is to break it at compile time when we change that, in case we |
1991 | * want to support more than (1<<32) bits on a 32bit arch. */ | |
b411b363 PR |
1992 | typecheck(unsigned long, mdev->rs_total); |
1993 | ||
1994 | /* note: both rs_total and rs_left are in bits, i.e. in | |
1995 | * units of BM_BLOCK_SIZE. | |
1996 | * for the percentage, we don't care. */ | |
1997 | ||
439d5953 LE |
1998 | if (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T) |
1999 | *bits_left = mdev->ov_left; | |
2000 | else | |
2001 | *bits_left = drbd_bm_total_weight(mdev) - mdev->rs_failed; | |
b411b363 PR |
2002 | /* >> 10 to prevent overflow, |
2003 | * +1 to prevent division by zero */ | |
2004 | if (*bits_left > mdev->rs_total) { | |
2005 | /* doh. maybe a logic bug somewhere. | |
2006 | * may also be just a race condition | |
2007 | * between this and a disconnect during sync. | |
2008 | * for now, just prevent in-kernel buffer overflow. | |
2009 | */ | |
2010 | smp_rmb(); | |
2011 | dev_warn(DEV, "cs:%s rs_left=%lu > rs_total=%lu (rs_failed %lu)\n", | |
2012 | drbd_conn_str(mdev->state.conn), | |
2013 | *bits_left, mdev->rs_total, mdev->rs_failed); | |
2014 | *per_mil_done = 0; | |
2015 | } else { | |
4b0715f0 LE |
2016 | /* Make sure the division happens in long context. |
2017 | * We allow up to one petabyte storage right now, | |
2018 | * at a granularity of 4k per bit that is 2**38 bits. | |
2019 | * After shift right and multiplication by 1000, | |
2020 | * this should still fit easily into a 32bit long, | |
2021 | * so we don't need a 64bit division on 32bit arch. | |
2022 | * Note: currently we don't support such large bitmaps on 32bit | |
2023 | * arch anyways, but no harm done to be prepared for it here. | |
2024 | */ | |
2025 | unsigned int shift = mdev->rs_total >= (1ULL << 32) ? 16 : 10; | |
2026 | unsigned long left = *bits_left >> shift; | |
2027 | unsigned long total = 1UL + (mdev->rs_total >> shift); | |
2028 | unsigned long tmp = 1000UL - left * 1000UL/total; | |
b411b363 PR |
2029 | *per_mil_done = tmp; |
2030 | } | |
2031 | } | |
2032 | ||
2033 | ||
2034 | /* this throttles on-the-fly application requests | |
2035 | * according to max_buffers settings; | |
2036 | * maybe re-implement using semaphores? */ | |
2037 | static inline int drbd_get_max_buffers(struct drbd_conf *mdev) | |
2038 | { | |
44ed167d PR |
2039 | struct net_conf *nc; |
2040 | int mxb; | |
2041 | ||
2042 | rcu_read_lock(); | |
2043 | nc = rcu_dereference(mdev->tconn->net_conf); | |
2044 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ | |
2045 | rcu_read_unlock(); | |
2046 | ||
b411b363 PR |
2047 | return mxb; |
2048 | } | |
2049 | ||
3719094e | 2050 | static inline int drbd_state_is_stable(struct drbd_conf *mdev) |
b411b363 | 2051 | { |
da9fbc27 | 2052 | union drbd_dev_state s = mdev->state; |
b411b363 PR |
2053 | |
2054 | /* DO NOT add a default clause, we want the compiler to warn us | |
2055 | * for any newly introduced state we may have forgotten to add here */ | |
2056 | ||
2057 | switch ((enum drbd_conns)s.conn) { | |
2058 | /* new io only accepted when there is no connection, ... */ | |
2059 | case C_STANDALONE: | |
2060 | case C_WF_CONNECTION: | |
2061 | /* ... or there is a well established connection. */ | |
2062 | case C_CONNECTED: | |
2063 | case C_SYNC_SOURCE: | |
2064 | case C_SYNC_TARGET: | |
2065 | case C_VERIFY_S: | |
2066 | case C_VERIFY_T: | |
2067 | case C_PAUSED_SYNC_S: | |
2068 | case C_PAUSED_SYNC_T: | |
67531718 PR |
2069 | case C_AHEAD: |
2070 | case C_BEHIND: | |
3719094e | 2071 | /* transitional states, IO allowed */ |
b411b363 PR |
2072 | case C_DISCONNECTING: |
2073 | case C_UNCONNECTED: | |
2074 | case C_TIMEOUT: | |
2075 | case C_BROKEN_PIPE: | |
2076 | case C_NETWORK_FAILURE: | |
2077 | case C_PROTOCOL_ERROR: | |
2078 | case C_TEAR_DOWN: | |
2079 | case C_WF_REPORT_PARAMS: | |
2080 | case C_STARTING_SYNC_S: | |
2081 | case C_STARTING_SYNC_T: | |
3719094e PR |
2082 | break; |
2083 | ||
2084 | /* Allow IO in BM exchange states with new protocols */ | |
b411b363 | 2085 | case C_WF_BITMAP_S: |
31890f4a | 2086 | if (mdev->tconn->agreed_pro_version < 96) |
3719094e PR |
2087 | return 0; |
2088 | break; | |
2089 | ||
2090 | /* no new io accepted in these states */ | |
b411b363 PR |
2091 | case C_WF_BITMAP_T: |
2092 | case C_WF_SYNC_UUID: | |
2093 | case C_MASK: | |
2094 | /* not "stable" */ | |
2095 | return 0; | |
2096 | } | |
2097 | ||
2098 | switch ((enum drbd_disk_state)s.disk) { | |
2099 | case D_DISKLESS: | |
2100 | case D_INCONSISTENT: | |
2101 | case D_OUTDATED: | |
2102 | case D_CONSISTENT: | |
2103 | case D_UP_TO_DATE: | |
2104 | /* disk state is stable as well. */ | |
2105 | break; | |
2106 | ||
2107 | /* no new io accepted during tansitional states */ | |
2108 | case D_ATTACHING: | |
2109 | case D_FAILED: | |
2110 | case D_NEGOTIATING: | |
2111 | case D_UNKNOWN: | |
2112 | case D_MASK: | |
2113 | /* not "stable" */ | |
2114 | return 0; | |
2115 | } | |
2116 | ||
2117 | return 1; | |
2118 | } | |
2119 | ||
2aebfabb | 2120 | static inline int drbd_suspended(struct drbd_conf *mdev) |
fb22c402 | 2121 | { |
8e0af25f PR |
2122 | struct drbd_tconn *tconn = mdev->tconn; |
2123 | ||
2124 | return tconn->susp || tconn->susp_fen || tconn->susp_nod; | |
fb22c402 PR |
2125 | } |
2126 | ||
1b881ef7 | 2127 | static inline bool may_inc_ap_bio(struct drbd_conf *mdev) |
b411b363 PR |
2128 | { |
2129 | int mxb = drbd_get_max_buffers(mdev); | |
2130 | ||
2aebfabb | 2131 | if (drbd_suspended(mdev)) |
1b881ef7 | 2132 | return false; |
b411b363 | 2133 | if (test_bit(SUSPEND_IO, &mdev->flags)) |
1b881ef7 | 2134 | return false; |
b411b363 PR |
2135 | |
2136 | /* to avoid potential deadlock or bitmap corruption, | |
2137 | * in various places, we only allow new application io | |
2138 | * to start during "stable" states. */ | |
2139 | ||
2140 | /* no new io accepted when attaching or detaching the disk */ | |
3719094e | 2141 | if (!drbd_state_is_stable(mdev)) |
1b881ef7 | 2142 | return false; |
b411b363 PR |
2143 | |
2144 | /* since some older kernels don't have atomic_add_unless, | |
2145 | * and we are within the spinlock anyways, we have this workaround. */ | |
2146 | if (atomic_read(&mdev->ap_bio_cnt) > mxb) | |
1b881ef7 | 2147 | return false; |
b411b363 | 2148 | if (test_bit(BITMAP_IO, &mdev->flags)) |
1b881ef7 AG |
2149 | return false; |
2150 | return true; | |
b411b363 PR |
2151 | } |
2152 | ||
23361cf3 | 2153 | static inline bool inc_ap_bio_cond(struct drbd_conf *mdev) |
b411b363 | 2154 | { |
1b881ef7 | 2155 | bool rv = false; |
8869d683 | 2156 | |
87eeee41 | 2157 | spin_lock_irq(&mdev->tconn->req_lock); |
1b881ef7 | 2158 | rv = may_inc_ap_bio(mdev); |
8869d683 | 2159 | if (rv) |
23361cf3 | 2160 | atomic_inc(&mdev->ap_bio_cnt); |
87eeee41 | 2161 | spin_unlock_irq(&mdev->tconn->req_lock); |
8869d683 PR |
2162 | |
2163 | return rv; | |
2164 | } | |
b411b363 | 2165 | |
23361cf3 | 2166 | static inline void inc_ap_bio(struct drbd_conf *mdev) |
8869d683 | 2167 | { |
b411b363 PR |
2168 | /* we wait here |
2169 | * as long as the device is suspended | |
2170 | * until the bitmap is no longer on the fly during connection | |
2171 | * handshake as long as we would exeed the max_buffer limit. | |
2172 | * | |
2173 | * to avoid races with the reconnect code, | |
2174 | * we need to atomic_inc within the spinlock. */ | |
2175 | ||
23361cf3 | 2176 | wait_event(mdev->misc_wait, inc_ap_bio_cond(mdev)); |
b411b363 PR |
2177 | } |
2178 | ||
2179 | static inline void dec_ap_bio(struct drbd_conf *mdev) | |
2180 | { | |
2181 | int mxb = drbd_get_max_buffers(mdev); | |
2182 | int ap_bio = atomic_dec_return(&mdev->ap_bio_cnt); | |
2183 | ||
2184 | D_ASSERT(ap_bio >= 0); | |
2185 | /* this currently does wake_up for every dec_ap_bio! | |
2186 | * maybe rather introduce some type of hysteresis? | |
2187 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ | |
2188 | if (ap_bio < mxb) | |
2189 | wake_up(&mdev->misc_wait); | |
2190 | if (ap_bio == 0 && test_bit(BITMAP_IO, &mdev->flags)) { | |
2191 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &mdev->flags)) | |
e42325a5 | 2192 | drbd_queue_work(&mdev->tconn->data.work, &mdev->bm_io_work.w); |
b411b363 PR |
2193 | } |
2194 | } | |
2195 | ||
62b0da3a | 2196 | static inline int drbd_set_ed_uuid(struct drbd_conf *mdev, u64 val) |
b411b363 | 2197 | { |
62b0da3a | 2198 | int changed = mdev->ed_uuid != val; |
b411b363 | 2199 | mdev->ed_uuid = val; |
62b0da3a | 2200 | return changed; |
b411b363 PR |
2201 | } |
2202 | ||
b411b363 PR |
2203 | static inline int drbd_queue_order_type(struct drbd_conf *mdev) |
2204 | { | |
2205 | /* sorry, we currently have no working implementation | |
2206 | * of distributed TCQ stuff */ | |
2207 | #ifndef QUEUE_ORDERED_NONE | |
2208 | #define QUEUE_ORDERED_NONE 0 | |
2209 | #endif | |
2210 | return QUEUE_ORDERED_NONE; | |
2211 | } | |
2212 | ||
b411b363 PR |
2213 | static inline void drbd_md_flush(struct drbd_conf *mdev) |
2214 | { | |
2215 | int r; | |
2216 | ||
a8a4e51e | 2217 | if (test_bit(MD_NO_FUA, &mdev->flags)) |
b411b363 PR |
2218 | return; |
2219 | ||
dd3932ed | 2220 | r = blkdev_issue_flush(mdev->ldev->md_bdev, GFP_KERNEL, NULL); |
b411b363 | 2221 | if (r) { |
a8a4e51e | 2222 | set_bit(MD_NO_FUA, &mdev->flags); |
b411b363 PR |
2223 | dev_err(DEV, "meta data flush failed with status %d, disabling md-flushes\n", r); |
2224 | } | |
2225 | } | |
2226 | ||
2227 | #endif |