Commit | Line | Data |
---|---|---|
93c68cc4 | 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
b411b363 PR |
2 | /* |
3 | drbd_int.h | |
4 | ||
5 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
6 | ||
7 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
8 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
9 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
10 | ||
b411b363 PR |
11 | |
12 | */ | |
13 | ||
14 | #ifndef _DRBD_INT_H | |
15 | #define _DRBD_INT_H | |
16 | ||
9534d671 | 17 | #include <crypto/hash.h> |
b411b363 PR |
18 | #include <linux/compiler.h> |
19 | #include <linux/types.h> | |
b411b363 | 20 | #include <linux/list.h> |
3f07c014 | 21 | #include <linux/sched/signal.h> |
b411b363 PR |
22 | #include <linux/bitops.h> |
23 | #include <linux/slab.h> | |
132cc538 | 24 | #include <linux/ratelimit.h> |
b411b363 PR |
25 | #include <linux/tcp.h> |
26 | #include <linux/mutex.h> | |
27 | #include <linux/major.h> | |
28 | #include <linux/blkdev.h> | |
66114cad | 29 | #include <linux/backing-dev.h> |
062e879c | 30 | #include <linux/idr.h> |
9d5059c9 | 31 | #include <linux/dynamic_debug.h> |
b411b363 PR |
32 | #include <net/tcp.h> |
33 | #include <linux/lru_cache.h> | |
70c71606 | 34 | #include <linux/prefetch.h> |
3b98c0c2 | 35 | #include <linux/drbd_genl_api.h> |
b8907339 | 36 | #include <linux/drbd.h> |
20f2a34a | 37 | #include <linux/drbd_config.h> |
d9f65229 | 38 | #include "drbd_strings.h" |
b8907339 | 39 | #include "drbd_state.h" |
a3603a6e | 40 | #include "drbd_protocol.h" |
136160c1 | 41 | #include "drbd_polymorph_printk.h" |
b411b363 | 42 | |
183ece30 | 43 | /* shared module parameters, defined in drbd_main.c */ |
b411b363 | 44 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
183ece30 RK |
45 | extern int drbd_enable_faults; |
46 | extern int drbd_fault_rate; | |
b411b363 PR |
47 | #endif |
48 | ||
183ece30 | 49 | extern unsigned int drbd_minor_count; |
8ab761e1 | 50 | extern char drbd_usermode_helper[]; |
183ece30 | 51 | extern int drbd_proc_details; |
b411b363 PR |
52 | |
53 | ||
b411b363 PR |
54 | /* This is used to stop/restart our threads. |
55 | * Cannot use SIGTERM nor SIGKILL, since these | |
56 | * are sent out by init on runlevel changes | |
57 | * I choose SIGHUP for now. | |
58 | */ | |
59 | #define DRBD_SIGKILL SIGHUP | |
60 | ||
b411b363 PR |
61 | #define ID_IN_SYNC (4711ULL) |
62 | #define ID_OUT_OF_SYNC (4712ULL) | |
b411b363 | 63 | #define ID_SYNCER (-1ULL) |
579b57ed | 64 | |
4a23f264 | 65 | #define UUID_NEW_BM_OFFSET ((u64)0x0001000000000000ULL) |
b411b363 | 66 | |
54761697 | 67 | struct drbd_device; |
bde89a9e | 68 | struct drbd_connection; |
8164dd6c | 69 | struct drbd_peer_device; |
b411b363 | 70 | |
b411b363 PR |
71 | /* Defines to control fault insertion */ |
72 | enum { | |
73 | DRBD_FAULT_MD_WR = 0, /* meta data write */ | |
74 | DRBD_FAULT_MD_RD = 1, /* read */ | |
75 | DRBD_FAULT_RS_WR = 2, /* resync */ | |
76 | DRBD_FAULT_RS_RD = 3, | |
77 | DRBD_FAULT_DT_WR = 4, /* data */ | |
78 | DRBD_FAULT_DT_RD = 5, | |
79 | DRBD_FAULT_DT_RA = 6, /* data read ahead */ | |
80 | DRBD_FAULT_BM_ALLOC = 7, /* bitmap allocation */ | |
81 | DRBD_FAULT_AL_EE = 8, /* alloc ee */ | |
6b4388ac | 82 | DRBD_FAULT_RECEIVE = 9, /* Changes some bytes upon receiving a [rs]data block */ |
b411b363 PR |
83 | |
84 | DRBD_FAULT_MAX, | |
85 | }; | |
86 | ||
b411b363 | 87 | extern unsigned int |
b30ab791 | 88 | _drbd_insert_fault(struct drbd_device *device, unsigned int type); |
0cf9d27e | 89 | |
b411b363 | 90 | static inline int |
b30ab791 | 91 | drbd_insert_fault(struct drbd_device *device, unsigned int type) { |
0cf9d27e | 92 | #ifdef CONFIG_DRBD_FAULT_INJECTION |
183ece30 RK |
93 | return drbd_fault_rate && |
94 | (drbd_enable_faults & (1<<type)) && | |
b30ab791 | 95 | _drbd_insert_fault(device, type); |
b411b363 | 96 | #else |
0cf9d27e | 97 | return 0; |
b411b363 | 98 | #endif |
0cf9d27e | 99 | } |
b411b363 PR |
100 | |
101 | /* integer division, round _UP_ to the next integer */ | |
102 | #define div_ceil(A, B) ((A)/(B) + ((A)%(B) ? 1 : 0)) | |
103 | /* usual integer division */ | |
104 | #define div_floor(A, B) ((A)/(B)) | |
105 | ||
b411b363 | 106 | extern struct ratelimit_state drbd_ratelimit_state; |
05a10ec7 | 107 | extern struct idr drbd_devices; /* RCU, updates: genl_lock() */ |
77c556f6 | 108 | extern struct list_head drbd_resources; /* RCU, updates: genl_lock() */ |
b411b363 | 109 | |
d8763023 | 110 | extern const char *cmdname(enum drbd_packet cmd); |
b411b363 PR |
111 | |
112 | /* for sending/receiving the bitmap, | |
113 | * possibly in some encoding scheme */ | |
114 | struct bm_xfer_ctx { | |
115 | /* "const" | |
116 | * stores total bits and long words | |
117 | * of the bitmap, so we don't need to | |
118 | * call the accessor functions over and again. */ | |
119 | unsigned long bm_bits; | |
120 | unsigned long bm_words; | |
121 | /* during xfer, current position within the bitmap */ | |
122 | unsigned long bit_offset; | |
123 | unsigned long word_offset; | |
124 | ||
125 | /* statistics; index: (h->command == P_BITMAP) */ | |
126 | unsigned packets[2]; | |
127 | unsigned bytes[2]; | |
128 | }; | |
129 | ||
5e54c2a6 AG |
130 | extern void INFO_bm_xfer_stats(struct drbd_peer_device *peer_device, |
131 | const char *direction, struct bm_xfer_ctx *c); | |
b411b363 PR |
132 | |
133 | static inline void bm_xfer_ctx_bit_to_word_offset(struct bm_xfer_ctx *c) | |
134 | { | |
135 | /* word_offset counts "native long words" (32 or 64 bit), | |
136 | * aligned at 64 bit. | |
137 | * Encoded packet may end at an unaligned bit offset. | |
138 | * In case a fallback clear text packet is transmitted in | |
139 | * between, we adjust this offset back to the last 64bit | |
140 | * aligned "native long word", which makes coding and decoding | |
141 | * the plain text bitmap much more convenient. */ | |
142 | #if BITS_PER_LONG == 64 | |
143 | c->word_offset = c->bit_offset >> 6; | |
144 | #elif BITS_PER_LONG == 32 | |
145 | c->word_offset = c->bit_offset >> 5; | |
146 | c->word_offset &= ~(1UL); | |
147 | #else | |
148 | # error "unsupported BITS_PER_LONG" | |
149 | #endif | |
150 | } | |
151 | ||
bde89a9e | 152 | extern unsigned int drbd_header_size(struct drbd_connection *connection); |
b411b363 | 153 | |
b411b363 PR |
154 | /**********************************************************************/ |
155 | enum drbd_thread_state { | |
e77a0a5c AG |
156 | NONE, |
157 | RUNNING, | |
158 | EXITING, | |
159 | RESTARTING | |
b411b363 PR |
160 | }; |
161 | ||
162 | struct drbd_thread { | |
163 | spinlock_t t_lock; | |
164 | struct task_struct *task; | |
165 | struct completion stop; | |
166 | enum drbd_thread_state t_state; | |
167 | int (*function) (struct drbd_thread *); | |
2457b6d5 | 168 | struct drbd_resource *resource; |
bde89a9e | 169 | struct drbd_connection *connection; |
b411b363 | 170 | int reset_cpu_mask; |
c60b0251 | 171 | const char *name; |
b411b363 PR |
172 | }; |
173 | ||
174 | static inline enum drbd_thread_state get_t_state(struct drbd_thread *thi) | |
175 | { | |
176 | /* THINK testing the t_state seems to be uncritical in all cases | |
177 | * (but thread_{start,stop}), so we can read it *without* the lock. | |
178 | * --lge */ | |
179 | ||
180 | smp_rmb(); | |
181 | return thi->t_state; | |
182 | } | |
183 | ||
b411b363 PR |
184 | struct drbd_work { |
185 | struct list_head list; | |
309a8348 | 186 | int (*cb)(struct drbd_work *, int cancel); |
84b8c06b AG |
187 | }; |
188 | ||
189 | struct drbd_device_work { | |
190 | struct drbd_work w; | |
191 | struct drbd_device *device; | |
b411b363 PR |
192 | }; |
193 | ||
ace652ac AG |
194 | #include "drbd_interval.h" |
195 | ||
54761697 | 196 | extern int drbd_wait_misc(struct drbd_device *, struct drbd_interval *); |
7be8da07 | 197 | |
28bc3b8c AG |
198 | extern void lock_all_resources(void); |
199 | extern void unlock_all_resources(void); | |
200 | ||
b411b363 PR |
201 | struct drbd_request { |
202 | struct drbd_work w; | |
84b8c06b | 203 | struct drbd_device *device; |
b411b363 PR |
204 | |
205 | /* if local IO is not allowed, will be NULL. | |
206 | * if local IO _is_ allowed, holds the locally submitted bio clone, | |
207 | * or, after local IO completion, the ERR_PTR(error). | |
fcefa62e | 208 | * see drbd_request_endio(). */ |
b411b363 PR |
209 | struct bio *private_bio; |
210 | ||
ace652ac | 211 | struct drbd_interval i; |
b411b363 | 212 | |
b6dd1a89 | 213 | /* epoch: used to check on "completion" whether this req was in |
b411b363 | 214 | * the current epoch, and we therefore have to close it, |
b6dd1a89 LE |
215 | * causing a p_barrier packet to be send, starting a new epoch. |
216 | * | |
217 | * This corresponds to "barrier" in struct p_barrier[_ack], | |
218 | * and to "barrier_nr" in struct drbd_epoch (and various | |
219 | * comments/function parameters/local variable names). | |
b411b363 | 220 | */ |
b6dd1a89 | 221 | unsigned int epoch; |
b411b363 | 222 | |
b411b363 PR |
223 | struct list_head tl_requests; /* ring list in the transfer log */ |
224 | struct bio *master_bio; /* master bio pointer */ | |
e5f891b2 | 225 | |
844a6ae7 LE |
226 | /* see struct drbd_device */ |
227 | struct list_head req_pending_master_completion; | |
228 | struct list_head req_pending_local; | |
229 | ||
e5f891b2 LE |
230 | /* for generic IO accounting */ |
231 | unsigned long start_jif; | |
232 | ||
233 | /* for DRBD internal statistics */ | |
234 | ||
235 | /* Minimal set of time stamps to determine if we wait for activity log | |
236 | * transactions, local disk or peer. 32 bit "jiffies" are good enough, | |
237 | * we don't expect a DRBD request to be stalled for several month. | |
238 | */ | |
239 | ||
240 | /* before actual request processing */ | |
241 | unsigned long in_actlog_jif; | |
242 | ||
243 | /* local disk */ | |
244 | unsigned long pre_submit_jif; | |
245 | ||
246 | /* per connection */ | |
247 | unsigned long pre_send_jif; | |
248 | unsigned long acked_jif; | |
249 | unsigned long net_done_jif; | |
250 | ||
251 | /* Possibly even more detail to track each phase: | |
252 | * master_completion_jif | |
253 | * how long did it take to complete the master bio | |
254 | * (application visible latency) | |
255 | * allocated_jif | |
256 | * how long the master bio was blocked until we finally allocated | |
257 | * a tracking struct | |
258 | * in_actlog_jif | |
259 | * how long did we wait for activity log transactions | |
260 | * | |
261 | * net_queued_jif | |
262 | * when did we finally queue it for sending | |
263 | * pre_send_jif | |
264 | * when did we start sending it | |
265 | * post_send_jif | |
266 | * how long did we block in the network stack trying to send it | |
267 | * acked_jif | |
268 | * when did we receive (or fake, in protocol A) a remote ACK | |
269 | * net_done_jif | |
270 | * when did we receive final acknowledgement (P_BARRIER_ACK), | |
271 | * or decide, e.g. on connection loss, that we do no longer expect | |
272 | * anything from this peer for this request. | |
273 | * | |
274 | * pre_submit_jif | |
275 | * post_sub_jif | |
276 | * when did we start submiting to the lower level device, | |
277 | * and how long did we block in that submit function | |
278 | * local_completion_jif | |
279 | * how long did it take the lower level device to complete this request | |
280 | */ | |
281 | ||
b411b363 | 282 | |
b406777e LE |
283 | /* once it hits 0, we may complete the master_bio */ |
284 | atomic_t completion_ref; | |
285 | /* once it hits 0, we may destroy this drbd_request object */ | |
286 | struct kref kref; | |
b411b363 | 287 | |
a0d856df | 288 | unsigned rq_state; /* see comments above _req_mod() */ |
b411b363 | 289 | }; |
b411b363 PR |
290 | |
291 | struct drbd_epoch { | |
bde89a9e | 292 | struct drbd_connection *connection; |
b411b363 PR |
293 | struct list_head list; |
294 | unsigned int barrier_nr; | |
295 | atomic_t epoch_size; /* increased on every request added. */ | |
296 | atomic_t active; /* increased on every req. added, and dec on every finished. */ | |
297 | unsigned long flags; | |
298 | }; | |
299 | ||
300 | /* drbd_epoch flag bits */ | |
301 | enum { | |
b411b363 | 302 | DE_HAVE_BARRIER_NUMBER, |
b411b363 PR |
303 | }; |
304 | ||
305 | enum epoch_event { | |
306 | EV_PUT, | |
307 | EV_GOT_BARRIER_NR, | |
b411b363 | 308 | EV_BECAME_LAST, |
b411b363 PR |
309 | EV_CLEANUP = 32, /* used as flag */ |
310 | }; | |
311 | ||
b411b363 PR |
312 | struct digest_info { |
313 | int digest_size; | |
314 | void *digest; | |
315 | }; | |
316 | ||
f6ffca9f | 317 | struct drbd_peer_request { |
a8cd15ba AG |
318 | struct drbd_work w; |
319 | struct drbd_peer_device *peer_device; | |
85719573 | 320 | struct drbd_epoch *epoch; /* for writes */ |
45bb912b | 321 | struct page *pages; |
ce668b6d | 322 | blk_opf_t opf; |
45bb912b | 323 | atomic_t pending_bios; |
010f6e67 | 324 | struct drbd_interval i; |
45bb912b LE |
325 | /* see comments on ee flag bits below */ |
326 | unsigned long flags; | |
21ae5d7f | 327 | unsigned long submit_jif; |
85719573 PR |
328 | union { |
329 | u64 block_id; | |
330 | struct digest_info *digest; | |
331 | }; | |
45bb912b LE |
332 | }; |
333 | ||
ce668b6d CB |
334 | /* Equivalent to bio_op and req_op. */ |
335 | #define peer_req_op(peer_req) \ | |
336 | ((peer_req)->opf & REQ_OP_MASK) | |
337 | ||
45bb912b LE |
338 | /* ee flag bits. |
339 | * While corresponding bios are in flight, the only modification will be | |
340 | * set_bit WAS_ERROR, which has to be atomic. | |
341 | * If no bios are in flight yet, or all have been completed, | |
342 | * non-atomic modification to ee->flags is ok. | |
343 | */ | |
b411b363 PR |
344 | enum { |
345 | __EE_CALL_AL_COMPLETE_IO, | |
b411b363 | 346 | __EE_MAY_SET_IN_SYNC, |
45bb912b | 347 | |
9305455a | 348 | /* is this a TRIM aka REQ_OP_DISCARD? */ |
f31e583a LE |
349 | __EE_TRIM, |
350 | /* explicit zero-out requested, or | |
351 | * our lower level cannot handle trim, | |
352 | * and we want to fall back to zeroout instead */ | |
353 | __EE_ZEROOUT, | |
a0fb3c47 | 354 | |
45bb912b LE |
355 | /* In case a barrier failed, |
356 | * we need to resubmit without the barrier flag. */ | |
357 | __EE_RESUBMITTED, | |
358 | ||
6c852bec | 359 | /* we may have several bios per peer request. |
45bb912b LE |
360 | * if any of those fail, we set this flag atomically |
361 | * from the endio callback */ | |
362 | __EE_WAS_ERROR, | |
c36c3ced LE |
363 | |
364 | /* This ee has a pointer to a digest instead of a block id */ | |
365 | __EE_HAS_DIGEST, | |
7be8da07 AG |
366 | |
367 | /* Conflicting local requests need to be restarted after this request */ | |
368 | __EE_RESTART_REQUESTS, | |
303d1448 PR |
369 | |
370 | /* The peer wants a write ACK for this (wire proto C) */ | |
371 | __EE_SEND_WRITE_ACK, | |
302bdeae PR |
372 | |
373 | /* Is set when net_conf had two_primaries set while creating this peer_req */ | |
374 | __EE_IN_INTERVAL_TREE, | |
21ae5d7f LE |
375 | |
376 | /* for debugfs: */ | |
377 | /* has this been submitted, or does it still wait for something else? */ | |
378 | __EE_SUBMITTED, | |
379 | ||
380 | /* this is/was a write request */ | |
381 | __EE_WRITE, | |
382 | ||
9104d31a LE |
383 | /* this is/was a write same request */ |
384 | __EE_WRITE_SAME, | |
385 | ||
21ae5d7f LE |
386 | /* this originates from application on peer |
387 | * (not some resync or verify or other DRBD internal request) */ | |
388 | __EE_APPLICATION, | |
700ca8c0 PR |
389 | |
390 | /* If it contains only 0 bytes, send back P_RS_DEALLOCATED */ | |
391 | __EE_RS_THIN_REQ, | |
b411b363 PR |
392 | }; |
393 | #define EE_CALL_AL_COMPLETE_IO (1<<__EE_CALL_AL_COMPLETE_IO) | |
b411b363 | 394 | #define EE_MAY_SET_IN_SYNC (1<<__EE_MAY_SET_IN_SYNC) |
f31e583a LE |
395 | #define EE_TRIM (1<<__EE_TRIM) |
396 | #define EE_ZEROOUT (1<<__EE_ZEROOUT) | |
a0fb3c47 | 397 | #define EE_RESUBMITTED (1<<__EE_RESUBMITTED) |
45bb912b | 398 | #define EE_WAS_ERROR (1<<__EE_WAS_ERROR) |
c36c3ced | 399 | #define EE_HAS_DIGEST (1<<__EE_HAS_DIGEST) |
7be8da07 | 400 | #define EE_RESTART_REQUESTS (1<<__EE_RESTART_REQUESTS) |
303d1448 | 401 | #define EE_SEND_WRITE_ACK (1<<__EE_SEND_WRITE_ACK) |
302bdeae | 402 | #define EE_IN_INTERVAL_TREE (1<<__EE_IN_INTERVAL_TREE) |
21ae5d7f LE |
403 | #define EE_SUBMITTED (1<<__EE_SUBMITTED) |
404 | #define EE_WRITE (1<<__EE_WRITE) | |
9104d31a | 405 | #define EE_WRITE_SAME (1<<__EE_WRITE_SAME) |
21ae5d7f | 406 | #define EE_APPLICATION (1<<__EE_APPLICATION) |
700ca8c0 | 407 | #define EE_RS_THIN_REQ (1<<__EE_RS_THIN_REQ) |
b411b363 | 408 | |
b30ab791 | 409 | /* flag bits per device */ |
b411b363 | 410 | enum { |
b411b363 PR |
411 | UNPLUG_REMOTE, /* sending a "UnplugRemote" could help */ |
412 | MD_DIRTY, /* current uuids and flags not yet on disk */ | |
b411b363 | 413 | USE_DEGR_WFC_T, /* degr-wfc-timeout instead of wfc-timeout. */ |
b411b363 PR |
414 | CL_ST_CHG_SUCCESS, |
415 | CL_ST_CHG_FAIL, | |
416 | CRASHED_PRIMARY, /* This node was a crashed primary. | |
417 | * Gets cleared when the state.conn | |
418 | * goes into C_CONNECTED state. */ | |
b411b363 PR |
419 | CONSIDER_RESYNC, |
420 | ||
a8a4e51e | 421 | MD_NO_FUA, /* Users wants us to not use FUA/FLUSH on meta data dev */ |
5ab7d2c0 | 422 | |
b411b363 PR |
423 | BITMAP_IO, /* suspend application io; |
424 | once no more io in flight, start bitmap io */ | |
425 | BITMAP_IO_QUEUED, /* Started bitmap IO */ | |
a2a3c74f LE |
426 | WAS_IO_ERROR, /* Local disk failed, returned IO error */ |
427 | WAS_READ_ERROR, /* Local disk READ failed (set additionally to the above) */ | |
383606e0 | 428 | FORCE_DETACH, /* Force-detach from local disk, aborting any pending local IO */ |
b411b363 | 429 | RESYNC_AFTER_NEG, /* Resync after online grow after the attach&negotiate finished. */ |
b411b363 PR |
430 | RESIZE_PENDING, /* Size change detected locally, waiting for the response from |
431 | * the peer, if it changed there as well. */ | |
43a5182c | 432 | NEW_CUR_UUID, /* Create new current UUID when thawing IO */ |
0778286a | 433 | AL_SUSPENDED, /* Activity logging is currently suspended. */ |
370a43e7 | 434 | AHEAD_TO_SYNC_SOURCE, /* Ahead -> SyncSource queued */ |
e64a3294 | 435 | B_RS_H_DONE, /* Before resync handler done (already executed) */ |
08b165ba | 436 | DISCARD_MY_DATA, /* discard_my_data flag per volume */ |
380207d0 | 437 | READ_BALANCE_RR, |
e334f550 | 438 | |
f418815f LE |
439 | FLUSH_PENDING, /* if set, device->flush_jif is when we submitted that flush |
440 | * from drbd_flush_after_epoch() */ | |
441 | ||
e334f550 LE |
442 | /* cleared only after backing device related structures have been destroyed. */ |
443 | GOING_DISKLESS, /* Disk is being detached, because of io-error, or admin request. */ | |
444 | ||
445 | /* to be used in drbd_device_post_work() */ | |
446 | GO_DISKLESS, /* tell worker to schedule cleanup before detach */ | |
447 | DESTROY_DISK, /* tell worker to close backing devices and destroy related structures. */ | |
ac0acb9e LE |
448 | MD_SYNC, /* tell worker to call drbd_md_sync() */ |
449 | RS_START, /* tell worker to start resync/OV */ | |
e334f550 LE |
450 | RS_PROGRESS, /* tell worker that resync made significant progress */ |
451 | RS_DONE, /* tell worker that resync is done */ | |
b411b363 PR |
452 | }; |
453 | ||
54761697 | 454 | struct drbd_bitmap; /* opaque for drbd_device */ |
b411b363 | 455 | |
20ceb2b2 LE |
456 | /* definition of bits in bm_flags to be used in drbd_bm_lock |
457 | * and drbd_bitmap_io and friends. */ | |
458 | enum bm_flag { | |
20ceb2b2 | 459 | /* currently locked for bulk operation */ |
0e8488ad | 460 | BM_LOCKED_MASK = 0xf, |
20ceb2b2 LE |
461 | |
462 | /* in detail, that is: */ | |
463 | BM_DONT_CLEAR = 0x1, | |
464 | BM_DONT_SET = 0x2, | |
465 | BM_DONT_TEST = 0x4, | |
466 | ||
0e8488ad LE |
467 | /* so we can mark it locked for bulk operation, |
468 | * and still allow all non-bulk operations */ | |
469 | BM_IS_LOCKED = 0x8, | |
470 | ||
20ceb2b2 | 471 | /* (test bit, count bit) allowed (common case) */ |
0e8488ad | 472 | BM_LOCKED_TEST_ALLOWED = BM_DONT_CLEAR | BM_DONT_SET | BM_IS_LOCKED, |
20ceb2b2 LE |
473 | |
474 | /* testing bits, as well as setting new bits allowed, but clearing bits | |
475 | * would be unexpected. Used during bitmap receive. Setting new bits | |
476 | * requires sending of "out-of-sync" information, though. */ | |
0e8488ad | 477 | BM_LOCKED_SET_ALLOWED = BM_DONT_CLEAR | BM_IS_LOCKED, |
20ceb2b2 | 478 | |
0e8488ad LE |
479 | /* for drbd_bm_write_copy_pages, everything is allowed, |
480 | * only concurrent bulk operations are locked out. */ | |
481 | BM_LOCKED_CHANGE_ALLOWED = BM_IS_LOCKED, | |
20ceb2b2 LE |
482 | }; |
483 | ||
b411b363 PR |
484 | struct drbd_work_queue { |
485 | struct list_head q; | |
b411b363 | 486 | spinlock_t q_lock; /* to protect the list. */ |
8c0785a5 | 487 | wait_queue_head_t q_wait; |
b411b363 PR |
488 | }; |
489 | ||
490 | struct drbd_socket { | |
b411b363 PR |
491 | struct mutex mutex; |
492 | struct socket *socket; | |
493 | /* this way we get our | |
494 | * send/receive buffers off the stack */ | |
5a87d920 | 495 | void *sbuf; |
e6ef8a5c | 496 | void *rbuf; |
b411b363 PR |
497 | }; |
498 | ||
499 | struct drbd_md { | |
500 | u64 md_offset; /* sector offset to 'super' block */ | |
501 | ||
502 | u64 la_size_sect; /* last agreed size, unit sectors */ | |
9f2247bb | 503 | spinlock_t uuid_lock; |
b411b363 PR |
504 | u64 uuid[UI_SIZE]; |
505 | u64 device_uuid; | |
506 | u32 flags; | |
507 | u32 md_size_sect; | |
508 | ||
ae8bf312 | 509 | s32 al_offset; /* signed relative sector offset to activity log */ |
b411b363 | 510 | s32 bm_offset; /* signed relative sector offset to bitmap */ |
3a4d4eb3 LE |
511 | |
512 | /* cached value of bdev->disk_conf->meta_dev_idx (see below) */ | |
513 | s32 meta_dev_idx; | |
514 | ||
515 | /* see al_tr_number_to_on_disk_sector() */ | |
516 | u32 al_stripes; | |
517 | u32 al_stripe_size_4k; | |
518 | u32 al_size_4k; /* cached product of the above */ | |
b411b363 PR |
519 | }; |
520 | ||
b411b363 PR |
521 | struct drbd_backing_dev { |
522 | struct block_device *backing_bdev; | |
20e6a8d0 | 523 | struct file *backing_bdev_file; |
b411b363 | 524 | struct block_device *md_bdev; |
20e6a8d0 | 525 | struct file *f_md_bdev; |
b411b363 | 526 | struct drbd_md md; |
0500813f | 527 | struct disk_conf *disk_conf; /* RCU, for updates: resource->conf_update */ |
b411b363 PR |
528 | sector_t known_size; /* last known size of that backing device */ |
529 | }; | |
530 | ||
531 | struct drbd_md_io { | |
e37d2438 LE |
532 | struct page *page; |
533 | unsigned long start_jif; /* last call to drbd_md_get_buffer */ | |
534 | unsigned long submit_jif; /* last _drbd_md_sync_page_io() submit */ | |
535 | const char *current_use; | |
536 | atomic_t in_use; | |
0c464425 | 537 | unsigned int done; |
b411b363 PR |
538 | int error; |
539 | }; | |
540 | ||
541 | struct bm_io_work { | |
542 | struct drbd_work w; | |
8164dd6c | 543 | struct drbd_peer_device *peer_device; |
b411b363 | 544 | char *why; |
20ceb2b2 | 545 | enum bm_flag flags; |
8164dd6c | 546 | int (*io_fn)(struct drbd_device *device, struct drbd_peer_device *peer_device); |
b30ab791 | 547 | void (*done)(struct drbd_device *device, int rv); |
b411b363 PR |
548 | }; |
549 | ||
778f271d | 550 | struct fifo_buffer { |
778f271d PR |
551 | unsigned int head_index; |
552 | unsigned int size; | |
9958c857 | 553 | int total; /* sum of all values */ |
150849c5 | 554 | int values[] __counted_by(size); |
778f271d | 555 | }; |
6a365874 | 556 | extern struct fifo_buffer *fifo_alloc(unsigned int fifo_size); |
778f271d | 557 | |
bde89a9e | 558 | /* flag bits per connection */ |
01a311a5 PR |
559 | enum { |
560 | NET_CONGESTED, /* The data socket is congested */ | |
427c0434 | 561 | RESOLVE_CONFLICTS, /* Set on one node, cleared on the peer! */ |
668700b4 | 562 | SEND_PING, |
2a67d8b9 | 563 | GOT_PING_ACK, /* set when we receive a ping_ack packet, ping_wait gets woken */ |
4d0fc3fd | 564 | CONN_WD_ST_CHG_REQ, /* A cluster wide state change on the connection is active */ |
fc3b10a4 PR |
565 | CONN_WD_ST_CHG_OKAY, |
566 | CONN_WD_ST_CHG_FAIL, | |
8169e41b | 567 | CONN_DRY_RUN, /* Expect disconnect after resync handshake. */ |
6936fcb4 | 568 | CREATE_BARRIER, /* next P_DATA is preceded by a P_BARRIER */ |
a1096a6e | 569 | STATE_SENT, /* Do not change state/UUIDs while this is set */ |
6f3465ed LE |
570 | CALLBACK_PENDING, /* Whether we have a call_usermodehelper(, UMH_WAIT_PROC) |
571 | * pending, from drbd worker context. | |
6f3465ed | 572 | */ |
b66623e3 | 573 | DISCONNECT_SENT, |
e334f550 LE |
574 | |
575 | DEVICE_WORK_PENDING, /* tell worker that some device has pending work */ | |
01a311a5 PR |
576 | }; |
577 | ||
a2972846 AG |
578 | enum which_state { NOW, OLD = NOW, NEW }; |
579 | ||
77c556f6 AG |
580 | struct drbd_resource { |
581 | char *name; | |
4d3d5aa8 LE |
582 | #ifdef CONFIG_DEBUG_FS |
583 | struct dentry *debugfs_res; | |
584 | struct dentry *debugfs_res_volumes; | |
585 | struct dentry *debugfs_res_connections; | |
586 | struct dentry *debugfs_res_in_flight_summary; | |
587 | #endif | |
77c556f6 | 588 | struct kref kref; |
803ea134 | 589 | struct idr devices; /* volume number to device mapping */ |
77c556f6 AG |
590 | struct list_head connections; |
591 | struct list_head resources; | |
eb6bea67 | 592 | struct res_opts res_opts; |
0500813f | 593 | struct mutex conf_update; /* mutex for ready-copy-update of net_conf and disk_conf */ |
9e276872 | 594 | struct mutex adm_mutex; /* mutex to serialize administrative requests */ |
0500813f | 595 | spinlock_t req_lock; |
6bbf53ca AG |
596 | |
597 | unsigned susp:1; /* IO suspended by user */ | |
598 | unsigned susp_nod:1; /* IO suspended because no data */ | |
599 | unsigned susp_fen:1; /* IO suspended because fence peer handler runs */ | |
625a6ba2 | 600 | |
e9526580 PR |
601 | enum write_ordering_e write_ordering; |
602 | ||
625a6ba2 | 603 | cpumask_var_t cpu_mask; |
77c556f6 AG |
604 | }; |
605 | ||
944410e9 LE |
606 | struct drbd_thread_timing_details |
607 | { | |
608 | unsigned long start_jif; | |
609 | void *cb_addr; | |
610 | const char *caller_fn; | |
611 | unsigned int line; | |
612 | unsigned int cb_nr; | |
613 | }; | |
614 | ||
77c556f6 AG |
615 | struct drbd_connection { |
616 | struct list_head connections; | |
617 | struct drbd_resource *resource; | |
4d3d5aa8 LE |
618 | #ifdef CONFIG_DEBUG_FS |
619 | struct dentry *debugfs_conn; | |
620 | struct dentry *debugfs_conn_callback_history; | |
621 | struct dentry *debugfs_conn_oldest_requests; | |
622 | #endif | |
9dc9fbb3 | 623 | struct kref kref; |
c06ece6b | 624 | struct idr peer_devices; /* volume number to peer device mapping */ |
8410da8f PR |
625 | enum drbd_conns cstate; /* Only C_STANDALONE to C_WF_REPORT_PARAMS */ |
626 | struct mutex cstate_mutex; /* Protects graceful disconnects */ | |
28e448bb | 627 | unsigned int connect_cnt; /* Inc each time a connection is established */ |
2111438b | 628 | |
062e879c | 629 | unsigned long flags; |
44ed167d | 630 | struct net_conf *net_conf; /* content protected by rcu */ |
91fd4dad | 631 | wait_queue_head_t ping_wait; /* Woken upon reception of a ping, and a state change */ |
e42325a5 | 632 | |
089c075d AG |
633 | struct sockaddr_storage my_addr; |
634 | int my_addr_len; | |
635 | struct sockaddr_storage peer_addr; | |
636 | int peer_addr_len; | |
637 | ||
e42325a5 PR |
638 | struct drbd_socket data; /* data/barrier/cstate/parameter packets */ |
639 | struct drbd_socket meta; /* ping/ack (metadata) packets */ | |
31890f4a | 640 | int agreed_pro_version; /* actually used protocol version */ |
20c68fde | 641 | u32 agreed_features; |
31890f4a PR |
642 | unsigned long last_received; /* in jiffies, either socket */ |
643 | unsigned int ko_count; | |
e6b3ea83 | 644 | |
b6dd1a89 | 645 | struct list_head transfer_log; /* all requests not yet fully processed */ |
87eeee41 | 646 | |
9534d671 | 647 | struct crypto_shash *cram_hmac_tfm; |
3d0e6375 KC |
648 | struct crypto_shash *integrity_tfm; /* checksums we compute, updates protected by connection->data->mutex */ |
649 | struct crypto_shash *peer_integrity_tfm; /* checksums we verify, only accessed from receiver thread */ | |
650 | struct crypto_shash *csums_tfm; | |
651 | struct crypto_shash *verify_tfm; | |
a0638456 PR |
652 | void *int_dig_in; |
653 | void *int_dig_vv; | |
654 | ||
b6dd1a89 | 655 | /* receiver side */ |
12038a3a PR |
656 | struct drbd_epoch *current_epoch; |
657 | spinlock_t epoch_lock; | |
658 | unsigned int epochs; | |
b379c41e | 659 | atomic_t current_tle_nr; /* transfer log epoch number */ |
b6dd1a89 | 660 | unsigned current_tle_writes; /* writes seen within this tl epoch */ |
4b0007c0 | 661 | |
07be15b1 | 662 | unsigned long last_reconnect_jif; |
c51a0ef3 LE |
663 | /* empty member on older kernels without blk_start_plug() */ |
664 | struct blk_plug receiver_plug; | |
e6b3ea83 PR |
665 | struct drbd_thread receiver; |
666 | struct drbd_thread worker; | |
1c03e520 | 667 | struct drbd_thread ack_receiver; |
668700b4 | 668 | struct workqueue_struct *ack_sender; |
b6dd1a89 | 669 | |
7753a4c1 LE |
670 | /* cached pointers, |
671 | * so we can look up the oldest pending requests more quickly. | |
672 | * protected by resource->req_lock */ | |
673 | struct drbd_request *req_next; /* DRBD 9: todo.req_next */ | |
674 | struct drbd_request *req_ack_pending; | |
675 | struct drbd_request *req_not_net_done; | |
676 | ||
b6dd1a89 | 677 | /* sender side */ |
d5b27b01 | 678 | struct drbd_work_queue sender_work; |
b6dd1a89 | 679 | |
944410e9 LE |
680 | #define DRBD_THREAD_DETAILS_HIST 16 |
681 | unsigned int w_cb_nr; /* keeps counting up */ | |
682 | unsigned int r_cb_nr; /* keeps counting up */ | |
683 | struct drbd_thread_timing_details w_timing_details[DRBD_THREAD_DETAILS_HIST]; | |
684 | struct drbd_thread_timing_details r_timing_details[DRBD_THREAD_DETAILS_HIST]; | |
685 | ||
b6dd1a89 | 686 | struct { |
84d34f2f LE |
687 | unsigned long last_sent_barrier_jif; |
688 | ||
b6dd1a89 LE |
689 | /* whether this sender thread |
690 | * has processed a single write yet. */ | |
691 | bool seen_any_write_yet; | |
692 | ||
693 | /* Which barrier number to send with the next P_BARRIER */ | |
694 | int current_epoch_nr; | |
695 | ||
696 | /* how many write requests have been sent | |
697 | * with req->epoch == current_epoch_nr. | |
698 | * If none, no P_BARRIER will be sent. */ | |
699 | unsigned current_epoch_writes; | |
700 | } send; | |
778f271d PR |
701 | }; |
702 | ||
a2972846 AG |
703 | static inline bool has_net_conf(struct drbd_connection *connection) |
704 | { | |
705 | bool has_net_conf; | |
706 | ||
707 | rcu_read_lock(); | |
708 | has_net_conf = rcu_dereference(connection->net_conf); | |
709 | rcu_read_unlock(); | |
710 | ||
711 | return has_net_conf; | |
712 | } | |
713 | ||
944410e9 LE |
714 | void __update_timing_details( |
715 | struct drbd_thread_timing_details *tdp, | |
716 | unsigned int *cb_nr, | |
717 | void *cb, | |
718 | const char *fn, const unsigned int line); | |
719 | ||
720 | #define update_worker_timing_details(c, cb) \ | |
721 | __update_timing_details(c->w_timing_details, &c->w_cb_nr, cb, __func__ , __LINE__ ) | |
722 | #define update_receiver_timing_details(c, cb) \ | |
723 | __update_timing_details(c->r_timing_details, &c->r_cb_nr, cb, __func__ , __LINE__ ) | |
724 | ||
113fef9e LE |
725 | struct submit_worker { |
726 | struct workqueue_struct *wq; | |
727 | struct work_struct worker; | |
728 | ||
844a6ae7 | 729 | /* protected by ..->resource->req_lock */ |
113fef9e LE |
730 | struct list_head writes; |
731 | }; | |
732 | ||
a6b32bc3 AG |
733 | struct drbd_peer_device { |
734 | struct list_head peer_devices; | |
735 | struct drbd_device *device; | |
bde89a9e | 736 | struct drbd_connection *connection; |
668700b4 | 737 | struct work_struct send_acks_work; |
4d3d5aa8 LE |
738 | #ifdef CONFIG_DEBUG_FS |
739 | struct dentry *debugfs_peer_dev; | |
740 | #endif | |
a6b32bc3 AG |
741 | }; |
742 | ||
743 | struct drbd_device { | |
d8628a86 | 744 | struct drbd_resource *resource; |
a6b32bc3 | 745 | struct list_head peer_devices; |
4ce49266 | 746 | struct list_head pending_bitmap_io; |
4d3d5aa8 LE |
747 | |
748 | unsigned long flush_jif; | |
749 | #ifdef CONFIG_DEBUG_FS | |
750 | struct dentry *debugfs_minor; | |
751 | struct dentry *debugfs_vol; | |
752 | struct dentry *debugfs_vol_oldest_requests; | |
753 | struct dentry *debugfs_vol_act_log_extents; | |
754 | struct dentry *debugfs_vol_resync_extents; | |
755 | struct dentry *debugfs_vol_data_gen_id; | |
f5ec0173 | 756 | struct dentry *debugfs_vol_ed_gen_id; |
4d3d5aa8 LE |
757 | #endif |
758 | ||
759 | unsigned int vnr; /* volume number within the connection */ | |
760 | unsigned int minor; /* device minor number */ | |
761 | ||
81fa2e67 | 762 | struct kref kref; |
2111438b | 763 | |
b411b363 PR |
764 | /* things that are stored as / read from meta data on disk */ |
765 | unsigned long flags; | |
766 | ||
767 | /* configured by drbdsetup */ | |
9cf766a4 | 768 | struct drbd_backing_dev *ldev; |
b411b363 PR |
769 | |
770 | sector_t p_size; /* partner's disk size */ | |
771 | struct request_queue *rq_queue; | |
b411b363 PR |
772 | struct gendisk *vdisk; |
773 | ||
07be15b1 | 774 | unsigned long last_reattach_jif; |
84b8c06b AG |
775 | struct drbd_work resync_work; |
776 | struct drbd_work unplug_work; | |
b411b363 PR |
777 | struct timer_list resync_timer; |
778 | struct timer_list md_sync_timer; | |
370a43e7 | 779 | struct timer_list start_resync_timer; |
7fde2be9 | 780 | struct timer_list request_timer; |
b411b363 PR |
781 | |
782 | /* Used after attach while negotiating new disk state. */ | |
783 | union drbd_state new_state_tmp; | |
784 | ||
da9fbc27 | 785 | union drbd_dev_state state; |
b411b363 PR |
786 | wait_queue_head_t misc_wait; |
787 | wait_queue_head_t state_wait; /* upon each state change. */ | |
788 | unsigned int send_cnt; | |
789 | unsigned int recv_cnt; | |
790 | unsigned int read_cnt; | |
791 | unsigned int writ_cnt; | |
792 | unsigned int al_writ_cnt; | |
793 | unsigned int bm_writ_cnt; | |
794 | atomic_t ap_bio_cnt; /* Requests we need to complete */ | |
ad3fee79 | 795 | atomic_t ap_actlog_cnt; /* Requests waiting for activity log */ |
b411b363 PR |
796 | atomic_t ap_pending_cnt; /* AP data packets on the wire, ack expected */ |
797 | atomic_t rs_pending_cnt; /* RS request/data packets on the wire */ | |
d942ae44 | 798 | atomic_t unacked_cnt; /* Need to send replies for */ |
b411b363 | 799 | atomic_t local_cnt; /* Waiting for local completion */ |
7dbb4386 | 800 | atomic_t suspend_cnt; |
b2fb6dbe | 801 | |
dac1389c AG |
802 | /* Interval tree of pending local requests */ |
803 | struct rb_root read_requests; | |
de696716 | 804 | struct rb_root write_requests; |
b411b363 | 805 | |
844a6ae7 LE |
806 | /* for statistics and timeouts */ |
807 | /* [0] read, [1] write */ | |
808 | struct list_head pending_master_completion[2]; | |
809 | struct list_head pending_completion[2]; | |
810 | ||
aaaba345 LE |
811 | /* use checksums for *this* resync */ |
812 | bool use_csums; | |
4b0715f0 | 813 | /* blocks to resync in this run [unit BM_BLOCK_SIZE] */ |
b411b363 | 814 | unsigned long rs_total; |
4b0715f0 | 815 | /* number of resync blocks that failed in this run */ |
b411b363 PR |
816 | unsigned long rs_failed; |
817 | /* Syncer's start time [unit jiffies] */ | |
818 | unsigned long rs_start; | |
819 | /* cumulated time in PausedSyncX state [unit jiffies] */ | |
820 | unsigned long rs_paused; | |
1d7734a0 LE |
821 | /* skipped because csum was equal [unit BM_BLOCK_SIZE] */ |
822 | unsigned long rs_same_csum; | |
823 | #define DRBD_SYNC_MARKS 8 | |
824 | #define DRBD_SYNC_MARK_STEP (3*HZ) | |
b411b363 | 825 | /* block not up-to-date at mark [unit BM_BLOCK_SIZE] */ |
1d7734a0 | 826 | unsigned long rs_mark_left[DRBD_SYNC_MARKS]; |
b411b363 | 827 | /* marks's time [unit jiffies] */ |
1d7734a0 LE |
828 | unsigned long rs_mark_time[DRBD_SYNC_MARKS]; |
829 | /* current index into rs_mark_{left,time} */ | |
830 | int rs_last_mark; | |
328e0f12 | 831 | unsigned long rs_last_bcast; /* [unit jiffies] */ |
b411b363 PR |
832 | |
833 | /* where does the admin want us to start? (sector) */ | |
834 | sector_t ov_start_sector; | |
02b91b55 | 835 | sector_t ov_stop_sector; |
b411b363 PR |
836 | /* where are we now? (sector) */ |
837 | sector_t ov_position; | |
838 | /* Start sector of out of sync range (to merge printk reporting). */ | |
839 | sector_t ov_last_oos_start; | |
840 | /* size of out-of-sync range in sectors. */ | |
841 | sector_t ov_last_oos_size; | |
842 | unsigned long ov_left; /* in bits */ | |
b411b363 | 843 | |
b411b363 PR |
844 | struct drbd_bitmap *bitmap; |
845 | unsigned long bm_resync_fo; /* bit offset for drbd_bm_find_next */ | |
846 | ||
847 | /* Used to track operations of resync... */ | |
848 | struct lru_cache *resync; | |
849 | /* Number of locked elements in resync LRU */ | |
850 | unsigned int resync_locked; | |
851 | /* resync extent number waiting for application requests */ | |
852 | unsigned int resync_wenr; | |
853 | ||
854 | int open_cnt; | |
855 | u64 *p_uuid; | |
4b0007c0 | 856 | |
85719573 PR |
857 | struct list_head active_ee; /* IO in progress (P_DATA gets written to disk) */ |
858 | struct list_head sync_ee; /* IO in progress (P_RS_DATA_REPLY gets written to disk) */ | |
18b75d75 AG |
859 | struct list_head done_ee; /* need to send P_WRITE_ACK */ |
860 | struct list_head read_ee; /* [RS]P_DATA_REQUEST being read */ | |
b411b363 | 861 | struct list_head net_ee; /* zero-copy network send in progress */ |
b411b363 | 862 | |
b411b363 | 863 | struct list_head resync_reads; |
435f0740 LE |
864 | atomic_t pp_in_use; /* allocated from page pool */ |
865 | atomic_t pp_in_use_by_net; /* sendpage()d, still referenced by tcp */ | |
b411b363 | 866 | wait_queue_head_t ee_wait; |
cc94c650 | 867 | struct drbd_md_io md_io; |
b411b363 PR |
868 | spinlock_t al_lock; |
869 | wait_queue_head_t al_wait; | |
870 | struct lru_cache *act_log; /* activity log */ | |
871 | unsigned int al_tr_number; | |
872 | int al_tr_cycle; | |
b411b363 PR |
873 | wait_queue_head_t seq_wait; |
874 | atomic_t packet_seq; | |
875 | unsigned int peer_seq; | |
876 | spinlock_t peer_seq_lock; | |
b411b363 | 877 | unsigned long comm_bm_set; /* communicated number of set bits. */ |
b411b363 PR |
878 | struct bm_io_work bm_io_work; |
879 | u64 ed_uuid; /* UUID of the exposed data */ | |
8410da8f | 880 | struct mutex own_state_mutex; |
a6b32bc3 | 881 | struct mutex *state_mutex; /* either own_state_mutex or first_peer_device(device)->connection->cstate_mutex */ |
b411b363 | 882 | char congestion_reason; /* Why we where congested... */ |
1d7734a0 LE |
883 | atomic_t rs_sect_in; /* for incoming resync data rate, SyncTarget */ |
884 | atomic_t rs_sect_ev; /* for submitted resync data rate, both */ | |
885 | int rs_last_sect_ev; /* counter to compare with */ | |
886 | int rs_last_events; /* counter of read or write "events" (unit sectors) | |
887 | * on the lower level device when we last looked. */ | |
888 | int c_sync_rate; /* current resync rate after syncer throttle magic */ | |
bde89a9e | 889 | struct fifo_buffer *rs_plan_s; /* correction values of resync planer (RCU, connection->conn_update) */ |
778f271d | 890 | int rs_in_flight; /* resync sectors in flight (to proxy, in proxy and from proxy) */ |
759fbdfb | 891 | atomic_t ap_in_flight; /* App sectors in flight (waiting for ack) */ |
db141b2f LE |
892 | unsigned int peer_max_bio_size; |
893 | unsigned int local_max_bio_size; | |
113fef9e LE |
894 | |
895 | /* any requests that would block in drbd_make_request() | |
896 | * are deferred to this single-threaded work queue */ | |
897 | struct submit_worker submit; | |
b411b363 PR |
898 | }; |
899 | ||
4ce49266 LE |
900 | struct drbd_bm_aio_ctx { |
901 | struct drbd_device *device; | |
902 | struct list_head list; /* on device->pending_bitmap_io */; | |
903 | unsigned long start_jif; | |
904 | atomic_t in_flight; | |
905 | unsigned int done; | |
906 | unsigned flags; | |
907 | #define BM_AIO_COPY_PAGES 1 | |
908 | #define BM_AIO_WRITE_HINTED 2 | |
909 | #define BM_AIO_WRITE_ALL_PAGES 4 | |
910 | #define BM_AIO_READ 8 | |
911 | int error; | |
912 | struct kref kref; | |
913 | }; | |
914 | ||
a910b123 LE |
915 | struct drbd_config_context { |
916 | /* assigned from drbd_genlmsghdr */ | |
917 | unsigned int minor; | |
918 | /* assigned from request attributes, if present */ | |
919 | unsigned int volume; | |
920 | #define VOLUME_UNSPECIFIED (-1U) | |
921 | /* pointer into the request skb, | |
922 | * limited lifetime! */ | |
923 | char *resource_name; | |
924 | struct nlattr *my_addr; | |
925 | struct nlattr *peer_addr; | |
926 | ||
927 | /* reply buffer */ | |
928 | struct sk_buff *reply_skb; | |
929 | /* pointer into reply buffer */ | |
930 | struct drbd_genlmsghdr *reply_dh; | |
931 | /* resolved from attributes, if possible */ | |
932 | struct drbd_device *device; | |
933 | struct drbd_resource *resource; | |
934 | struct drbd_connection *connection; | |
935 | }; | |
936 | ||
b30ab791 | 937 | static inline struct drbd_device *minor_to_device(unsigned int minor) |
b411b363 | 938 | { |
05a10ec7 | 939 | return (struct drbd_device *)idr_find(&drbd_devices, minor); |
b411b363 PR |
940 | } |
941 | ||
a6b32bc3 AG |
942 | static inline struct drbd_peer_device *first_peer_device(struct drbd_device *device) |
943 | { | |
ec4a3407 | 944 | return list_first_entry_or_null(&device->peer_devices, struct drbd_peer_device, peer_devices); |
a6b32bc3 AG |
945 | } |
946 | ||
a2972846 AG |
947 | static inline struct drbd_peer_device * |
948 | conn_peer_device(struct drbd_connection *connection, int volume_number) | |
949 | { | |
950 | return idr_find(&connection->peer_devices, volume_number); | |
951 | } | |
952 | ||
77c556f6 AG |
953 | #define for_each_resource(resource, _resources) \ |
954 | list_for_each_entry(resource, _resources, resources) | |
955 | ||
956 | #define for_each_resource_rcu(resource, _resources) \ | |
957 | list_for_each_entry_rcu(resource, _resources, resources) | |
958 | ||
959 | #define for_each_resource_safe(resource, tmp, _resources) \ | |
960 | list_for_each_entry_safe(resource, tmp, _resources, resources) | |
961 | ||
962 | #define for_each_connection(connection, resource) \ | |
963 | list_for_each_entry(connection, &resource->connections, connections) | |
964 | ||
965 | #define for_each_connection_rcu(connection, resource) \ | |
966 | list_for_each_entry_rcu(connection, &resource->connections, connections) | |
967 | ||
968 | #define for_each_connection_safe(connection, tmp, resource) \ | |
969 | list_for_each_entry_safe(connection, tmp, &resource->connections, connections) | |
970 | ||
a6b32bc3 AG |
971 | #define for_each_peer_device(peer_device, device) \ |
972 | list_for_each_entry(peer_device, &device->peer_devices, peer_devices) | |
973 | ||
974 | #define for_each_peer_device_rcu(peer_device, device) \ | |
975 | list_for_each_entry_rcu(peer_device, &device->peer_devices, peer_devices) | |
976 | ||
977 | #define for_each_peer_device_safe(peer_device, tmp, device) \ | |
978 | list_for_each_entry_safe(peer_device, tmp, &device->peer_devices, peer_devices) | |
979 | ||
b30ab791 | 980 | static inline unsigned int device_to_minor(struct drbd_device *device) |
b411b363 | 981 | { |
b30ab791 | 982 | return device->minor; |
b411b363 PR |
983 | } |
984 | ||
b411b363 PR |
985 | /* |
986 | * function declarations | |
987 | *************************/ | |
988 | ||
989 | /* drbd_main.c */ | |
990 | ||
e89b591c PR |
991 | enum dds_flags { |
992 | DDSF_FORCED = 1, | |
993 | DDSF_NO_RESYNC = 2, /* Do not run a resync for the new space */ | |
994 | }; | |
995 | ||
b30ab791 | 996 | extern void drbd_init_set_defaults(struct drbd_device *device); |
b411b363 PR |
997 | extern int drbd_thread_start(struct drbd_thread *thi); |
998 | extern void _drbd_thread_stop(struct drbd_thread *thi, int restart, int wait); | |
999 | #ifdef CONFIG_SMP | |
80822284 | 1000 | extern void drbd_thread_current_set_cpu(struct drbd_thread *thi); |
b411b363 PR |
1001 | #else |
1002 | #define drbd_thread_current_set_cpu(A) ({}) | |
b411b363 | 1003 | #endif |
bde89a9e | 1004 | extern void tl_release(struct drbd_connection *, unsigned int barrier_nr, |
b411b363 | 1005 | unsigned int set_size); |
bde89a9e AG |
1006 | extern void tl_clear(struct drbd_connection *); |
1007 | extern void drbd_free_sock(struct drbd_connection *connection); | |
1008 | extern int drbd_send(struct drbd_connection *connection, struct socket *sock, | |
bedbd2a5 | 1009 | void *buf, size_t size, unsigned msg_flags); |
bde89a9e | 1010 | extern int drbd_send_all(struct drbd_connection *, struct socket *, void *, size_t, |
fb708e40 AG |
1011 | unsigned); |
1012 | ||
bde89a9e AG |
1013 | extern int __drbd_send_protocol(struct drbd_connection *connection, enum drbd_packet cmd); |
1014 | extern int drbd_send_protocol(struct drbd_connection *connection); | |
69a22773 AG |
1015 | extern int drbd_send_uuids(struct drbd_peer_device *); |
1016 | extern int drbd_send_uuids_skip_initial_sync(struct drbd_peer_device *); | |
1017 | extern void drbd_gen_and_send_sync_uuid(struct drbd_peer_device *); | |
1018 | extern int drbd_send_sizes(struct drbd_peer_device *, int trigger_reply, enum dds_flags flags); | |
1019 | extern int drbd_send_state(struct drbd_peer_device *, union drbd_state s); | |
1020 | extern int drbd_send_current_state(struct drbd_peer_device *); | |
1021 | extern int drbd_send_sync_param(struct drbd_peer_device *); | |
bde89a9e | 1022 | extern void drbd_send_b_ack(struct drbd_connection *connection, u32 barrier_nr, |
d4e67d7c | 1023 | u32 set_size); |
69a22773 | 1024 | extern int drbd_send_ack(struct drbd_peer_device *, enum drbd_packet, |
f6ffca9f | 1025 | struct drbd_peer_request *); |
69a22773 | 1026 | extern void drbd_send_ack_rp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 1027 | struct p_block_req *rp); |
69a22773 | 1028 | extern void drbd_send_ack_dp(struct drbd_peer_device *, enum drbd_packet, |
a9a9994d | 1029 | struct p_data *dp, int data_size); |
69a22773 | 1030 | extern int drbd_send_ack_ex(struct drbd_peer_device *, enum drbd_packet, |
b411b363 | 1031 | sector_t sector, int blksize, u64 block_id); |
69a22773 AG |
1032 | extern int drbd_send_out_of_sync(struct drbd_peer_device *, struct drbd_request *); |
1033 | extern int drbd_send_block(struct drbd_peer_device *, enum drbd_packet, | |
f6ffca9f | 1034 | struct drbd_peer_request *); |
69a22773 AG |
1035 | extern int drbd_send_dblock(struct drbd_peer_device *, struct drbd_request *req); |
1036 | extern int drbd_send_drequest(struct drbd_peer_device *, int cmd, | |
b411b363 | 1037 | sector_t sector, int size, u64 block_id); |
69a22773 | 1038 | extern int drbd_send_drequest_csum(struct drbd_peer_device *, sector_t sector, |
d8763023 AG |
1039 | int size, void *digest, int digest_size, |
1040 | enum drbd_packet cmd); | |
69a22773 | 1041 | extern int drbd_send_ov_request(struct drbd_peer_device *, sector_t sector, int size); |
b411b363 | 1042 | |
8164dd6c | 1043 | extern int drbd_send_bitmap(struct drbd_device *device, struct drbd_peer_device *peer_device); |
69a22773 | 1044 | extern void drbd_send_sr_reply(struct drbd_peer_device *, enum drbd_state_rv retcode); |
bde89a9e | 1045 | extern void conn_send_sr_reply(struct drbd_connection *connection, enum drbd_state_rv retcode); |
700ca8c0 | 1046 | extern int drbd_send_rs_deallocated(struct drbd_peer_device *, struct drbd_peer_request *); |
63a7c8ad | 1047 | extern void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev); |
b30ab791 | 1048 | extern void drbd_device_cleanup(struct drbd_device *device); |
c51a0ef3 LE |
1049 | extern void drbd_print_uuids(struct drbd_device *device, const char *text); |
1050 | extern void drbd_queue_unplug(struct drbd_device *device); | |
b411b363 | 1051 | |
bde89a9e | 1052 | extern void conn_md_sync(struct drbd_connection *connection); |
b30ab791 AG |
1053 | extern void drbd_md_write(struct drbd_device *device, void *buffer); |
1054 | extern void drbd_md_sync(struct drbd_device *device); | |
1055 | extern int drbd_md_read(struct drbd_device *device, struct drbd_backing_dev *bdev); | |
1056 | extern void drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1057 | extern void _drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1058 | extern void drbd_uuid_new_current(struct drbd_device *device) __must_hold(local); | |
1059 | extern void drbd_uuid_set_bm(struct drbd_device *device, u64 val) __must_hold(local); | |
1060 | extern void drbd_uuid_move_history(struct drbd_device *device) __must_hold(local); | |
1061 | extern void __drbd_uuid_set(struct drbd_device *device, int idx, u64 val) __must_hold(local); | |
1062 | extern void drbd_md_set_flag(struct drbd_device *device, int flags) __must_hold(local); | |
1063 | extern void drbd_md_clear_flag(struct drbd_device *device, int flags)__must_hold(local); | |
b411b363 | 1064 | extern int drbd_md_test_flag(struct drbd_backing_dev *, int); |
b30ab791 | 1065 | extern void drbd_md_mark_dirty(struct drbd_device *device); |
b30ab791 | 1066 | extern void drbd_queue_bitmap_io(struct drbd_device *device, |
8164dd6c | 1067 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
54761697 | 1068 | void (*done)(struct drbd_device *, int), |
8164dd6c AG |
1069 | char *why, enum bm_flag flags, |
1070 | struct drbd_peer_device *peer_device); | |
b30ab791 | 1071 | extern int drbd_bitmap_io(struct drbd_device *device, |
8164dd6c AG |
1072 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
1073 | char *why, enum bm_flag flags, | |
1074 | struct drbd_peer_device *peer_device); | |
b30ab791 | 1075 | extern int drbd_bitmap_io_from_worker(struct drbd_device *device, |
8164dd6c AG |
1076 | int (*io_fn)(struct drbd_device *, struct drbd_peer_device *), |
1077 | char *why, enum bm_flag flags, | |
1078 | struct drbd_peer_device *peer_device); | |
1079 | extern int drbd_bmio_set_n_write(struct drbd_device *device, | |
1080 | struct drbd_peer_device *peer_device) __must_hold(local); | |
1081 | extern int drbd_bmio_clear_n_write(struct drbd_device *device, | |
1082 | struct drbd_peer_device *peer_device) __must_hold(local); | |
b411b363 | 1083 | |
b411b363 | 1084 | /* Meta data layout |
ae8bf312 LE |
1085 | * |
1086 | * We currently have two possible layouts. | |
1087 | * Offsets in (512 byte) sectors. | |
1088 | * external: | |
1089 | * |----------- md_size_sect ------------------| | |
1090 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
1091 | * | al_offset == 8 | | |
1092 | * | bm_offset = al_offset + X | | |
1093 | * ==> bitmap sectors = md_size_sect - bm_offset | |
1094 | * | |
1095 | * Variants: | |
1096 | * old, indexed fixed size meta data: | |
1097 | * | |
1098 | * internal: | |
1099 | * |----------- md_size_sect ------------------| | |
1100 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ][padding*] | |
1101 | * | al_offset < 0 | | |
1102 | * | bm_offset = al_offset - Y | | |
1103 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
1104 | * | |
1105 | * [padding*] are zero or up to 7 unused 512 Byte sectors to the | |
1106 | * end of the device, so that the [4k superblock] will be 4k aligned. | |
1107 | * | |
1108 | * The activity log consists of 4k transaction blocks, | |
1109 | * which are written in a ring-buffer, or striped ring-buffer like fashion, | |
1110 | * which are writtensize used to be fixed 32kB, | |
1111 | * but is about to become configurable. | |
1112 | */ | |
b411b363 | 1113 | |
ae8bf312 LE |
1114 | /* Our old fixed size meta data layout |
1115 | * allows up to about 3.8TB, so if you want more, | |
7ad651b5 | 1116 | * you need to use the "flexible" meta data format. */ |
ae8bf312 LE |
1117 | #define MD_128MB_SECT (128LLU << 11) /* 128 MB, unit sectors */ |
1118 | #define MD_4kB_SECT 8 | |
1119 | #define MD_32kB_SECT 64 | |
7ad651b5 LE |
1120 | |
1121 | /* One activity log extent represents 4M of storage */ | |
1122 | #define AL_EXTENT_SHIFT 22 | |
b411b363 PR |
1123 | #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT) |
1124 | ||
7ad651b5 LE |
1125 | /* We could make these currently hardcoded constants configurable |
1126 | * variables at create-md time (or even re-configurable at runtime?). | |
1127 | * Which will require some more changes to the DRBD "super block" | |
1128 | * and attach code. | |
1129 | * | |
1130 | * updates per transaction: | |
1131 | * This many changes to the active set can be logged with one transaction. | |
1132 | * This number is arbitrary. | |
1133 | * context per transaction: | |
1134 | * This many context extent numbers are logged with each transaction. | |
1135 | * This number is resulting from the transaction block size (4k), the layout | |
1136 | * of the transaction header, and the number of updates per transaction. | |
1137 | * See drbd_actlog.c:struct al_transaction_on_disk | |
1138 | * */ | |
1139 | #define AL_UPDATES_PER_TRANSACTION 64 // arbitrary | |
1140 | #define AL_CONTEXT_PER_TRANSACTION 919 // (4096 - 36 - 6*64)/4 | |
1141 | ||
b411b363 PR |
1142 | #if BITS_PER_LONG == 32 |
1143 | #define LN2_BPL 5 | |
1144 | #define cpu_to_lel(A) cpu_to_le32(A) | |
1145 | #define lel_to_cpu(A) le32_to_cpu(A) | |
1146 | #elif BITS_PER_LONG == 64 | |
1147 | #define LN2_BPL 6 | |
1148 | #define cpu_to_lel(A) cpu_to_le64(A) | |
1149 | #define lel_to_cpu(A) le64_to_cpu(A) | |
1150 | #else | |
1151 | #error "LN2 of BITS_PER_LONG unknown!" | |
1152 | #endif | |
1153 | ||
1154 | /* resync bitmap */ | |
1155 | /* 16MB sized 'bitmap extent' to track syncer usage */ | |
1156 | struct bm_extent { | |
1157 | int rs_left; /* number of bits set (out of sync) in this extent. */ | |
1158 | int rs_failed; /* number of failed resync requests in this extent. */ | |
1159 | unsigned long flags; | |
1160 | struct lc_element lce; | |
1161 | }; | |
1162 | ||
1163 | #define BME_NO_WRITES 0 /* bm_extent.flags: no more requests on this one! */ | |
1164 | #define BME_LOCKED 1 /* bm_extent.flags: syncer active on this one. */ | |
e3555d85 | 1165 | #define BME_PRIORITY 2 /* finish resync IO on this extent ASAP! App IO waiting! */ |
b411b363 PR |
1166 | |
1167 | /* drbd_bitmap.c */ | |
1168 | /* | |
1169 | * We need to store one bit for a block. | |
1170 | * Example: 1GB disk @ 4096 byte blocks ==> we need 32 KB bitmap. | |
1171 | * Bit 0 ==> local node thinks this block is binary identical on both nodes | |
1172 | * Bit 1 ==> local node thinks this block needs to be synced. | |
1173 | */ | |
1174 | ||
8e26f9cc PR |
1175 | #define SLEEP_TIME (HZ/10) |
1176 | ||
45dfffeb LE |
1177 | /* We do bitmap IO in units of 4k blocks. |
1178 | * We also still have a hardcoded 4k per bit relation. */ | |
1179 | #define BM_BLOCK_SHIFT 12 /* 4k per bit */ | |
b411b363 | 1180 | #define BM_BLOCK_SIZE (1<<BM_BLOCK_SHIFT) |
45dfffeb LE |
1181 | /* mostly arbitrarily set the represented size of one bitmap extent, |
1182 | * aka resync extent, to 16 MiB (which is also 512 Byte worth of bitmap | |
1183 | * at 4k per bit resolution) */ | |
1184 | #define BM_EXT_SHIFT 24 /* 16 MiB per resync extent */ | |
b411b363 PR |
1185 | #define BM_EXT_SIZE (1<<BM_EXT_SHIFT) |
1186 | ||
1187 | #if (BM_EXT_SHIFT != 24) || (BM_BLOCK_SHIFT != 12) | |
1188 | #error "HAVE YOU FIXED drbdmeta AS WELL??" | |
1189 | #endif | |
1190 | ||
1191 | /* thus many _storage_ sectors are described by one bit */ | |
1192 | #define BM_SECT_TO_BIT(x) ((x)>>(BM_BLOCK_SHIFT-9)) | |
1193 | #define BM_BIT_TO_SECT(x) ((sector_t)(x)<<(BM_BLOCK_SHIFT-9)) | |
1194 | #define BM_SECT_PER_BIT BM_BIT_TO_SECT(1) | |
1195 | ||
1196 | /* bit to represented kilo byte conversion */ | |
1197 | #define Bit2KB(bits) ((bits)<<(BM_BLOCK_SHIFT-10)) | |
1198 | ||
1199 | /* in which _bitmap_ extent (resp. sector) the bit for a certain | |
1200 | * _storage_ sector is located in */ | |
1201 | #define BM_SECT_TO_EXT(x) ((x)>>(BM_EXT_SHIFT-9)) | |
5ab7d2c0 | 1202 | #define BM_BIT_TO_EXT(x) ((x) >> (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) |
b411b363 | 1203 | |
5ab7d2c0 | 1204 | /* first storage sector a bitmap extent corresponds to */ |
b411b363 | 1205 | #define BM_EXT_TO_SECT(x) ((sector_t)(x) << (BM_EXT_SHIFT-9)) |
5ab7d2c0 | 1206 | /* how much _storage_ sectors we have per bitmap extent */ |
b411b363 | 1207 | #define BM_SECT_PER_EXT BM_EXT_TO_SECT(1) |
5ab7d2c0 LE |
1208 | /* how many bits are covered by one bitmap extent (resync extent) */ |
1209 | #define BM_BITS_PER_EXT (1UL << (BM_EXT_SHIFT - BM_BLOCK_SHIFT)) | |
1210 | ||
1211 | #define BM_BLOCKS_PER_BM_EXT_MASK (BM_BITS_PER_EXT - 1) | |
1212 | ||
b411b363 PR |
1213 | |
1214 | /* in one sector of the bitmap, we have this many activity_log extents. */ | |
1215 | #define AL_EXT_PER_BM_SECT (1 << (BM_EXT_SHIFT - AL_EXTENT_SHIFT)) | |
b411b363 | 1216 | |
b411b363 PR |
1217 | /* the extent in "PER_EXTENT" below is an activity log extent |
1218 | * we need that many (long words/bytes) to store the bitmap | |
1219 | * of one AL_EXTENT_SIZE chunk of storage. | |
1220 | * we can store the bitmap for that many AL_EXTENTS within | |
1221 | * one sector of the _on_disk_ bitmap: | |
1222 | * bit 0 bit 37 bit 38 bit (512*8)-1 | |
1223 | * ...|........|........|.. // ..|........| | |
1224 | * sect. 0 `296 `304 ^(512*8*8)-1 | |
1225 | * | |
1226 | #define BM_WORDS_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / BITS_PER_LONG ) | |
1227 | #define BM_BYTES_PER_EXT ( (AL_EXT_SIZE/BM_BLOCK_SIZE) / 8 ) // 128 | |
1228 | #define BM_EXT_PER_SECT ( 512 / BM_BYTES_PER_EXTENT ) // 4 | |
1229 | */ | |
1230 | ||
1231 | #define DRBD_MAX_SECTORS_32 (0xffffffffLU) | |
ae8bf312 LE |
1232 | /* we have a certain meta data variant that has a fixed on-disk size of 128 |
1233 | * MiB, of which 4k are our "superblock", and 32k are the fixed size activity | |
1234 | * log, leaving this many sectors for the bitmap. | |
1235 | */ | |
1236 | ||
1237 | #define DRBD_MAX_SECTORS_FIXED_BM \ | |
1238 | ((MD_128MB_SECT - MD_32kB_SECT - MD_4kB_SECT) * (1LL<<(BM_EXT_SHIFT-9))) | |
ae8bf312 | 1239 | #define DRBD_MAX_SECTORS DRBD_MAX_SECTORS_FIXED_BM |
b411b363 PR |
1240 | /* 16 TB in units of sectors */ |
1241 | #if BITS_PER_LONG == 32 | |
1242 | /* adjust by one page worth of bitmap, | |
1243 | * so we won't wrap around in drbd_bm_find_next_bit. | |
1244 | * you should use 64bit OS for that much storage, anyways. */ | |
1245 | #define DRBD_MAX_SECTORS_FLEX BM_BIT_TO_SECT(0xffff7fff) | |
1246 | #else | |
4b0715f0 LE |
1247 | /* we allow up to 1 PiB now on 64bit architecture with "flexible" meta data */ |
1248 | #define DRBD_MAX_SECTORS_FLEX (1UL << 51) | |
1249 | /* corresponds to (1UL << 38) bits right now. */ | |
b411b363 | 1250 | #endif |
b411b363 | 1251 | |
8bf223c2 | 1252 | /* Estimate max bio size as 256 * PAGE_SIZE, |
ea1754a0 | 1253 | * so for typical PAGE_SIZE of 4k, that is (1<<20) Byte. |
23361cf3 LE |
1254 | * Since we may live in a mixed-platform cluster, |
1255 | * we limit us to a platform agnostic constant here for now. | |
1256 | * A followup commit may allow even bigger BIO sizes, | |
1257 | * once we thought that through. */ | |
98683650 | 1258 | #define DRBD_MAX_BIO_SIZE (1U << 20) |
a8affc03 | 1259 | #if DRBD_MAX_BIO_SIZE > (BIO_MAX_VECS << PAGE_SHIFT) |
23361cf3 LE |
1260 | #error Architecture not supported: DRBD_MAX_BIO_SIZE > BIO_MAX_SIZE |
1261 | #endif | |
db141b2f | 1262 | #define DRBD_MAX_BIO_SIZE_SAFE (1U << 12) /* Works always = 4k */ |
b411b363 | 1263 | |
98683650 PR |
1264 | #define DRBD_MAX_SIZE_H80_PACKET (1U << 15) /* Header 80 only allows packets up to 32KiB data */ |
1265 | #define DRBD_MAX_BIO_SIZE_P95 (1U << 17) /* Protocol 95 to 99 allows bios up to 128KiB */ | |
b411b363 | 1266 | |
505675f9 LE |
1267 | /* For now, don't allow more than half of what we can "activate" in one |
1268 | * activity log transaction to be discarded in one go. We may need to rework | |
1269 | * drbd_al_begin_io() to allow for even larger discard ranges */ | |
9104d31a LE |
1270 | #define DRBD_MAX_BATCH_BIO_SIZE (AL_UPDATES_PER_TRANSACTION/2*AL_EXTENT_SIZE) |
1271 | #define DRBD_MAX_BBIO_SECTORS (DRBD_MAX_BATCH_BIO_SIZE >> 9) | |
a0fb3c47 | 1272 | |
b30ab791 AG |
1273 | extern int drbd_bm_init(struct drbd_device *device); |
1274 | extern int drbd_bm_resize(struct drbd_device *device, sector_t sectors, int set_new_bits); | |
1275 | extern void drbd_bm_cleanup(struct drbd_device *device); | |
1276 | extern void drbd_bm_set_all(struct drbd_device *device); | |
1277 | extern void drbd_bm_clear_all(struct drbd_device *device); | |
4b0715f0 | 1278 | /* set/clear/test only a few bits at a time */ |
b411b363 | 1279 | extern int drbd_bm_set_bits( |
b30ab791 | 1280 | struct drbd_device *device, unsigned long s, unsigned long e); |
b411b363 | 1281 | extern int drbd_bm_clear_bits( |
b30ab791 | 1282 | struct drbd_device *device, unsigned long s, unsigned long e); |
4b0715f0 | 1283 | extern int drbd_bm_count_bits( |
b30ab791 | 1284 | struct drbd_device *device, const unsigned long s, const unsigned long e); |
4b0715f0 LE |
1285 | /* bm_set_bits variant for use while holding drbd_bm_lock, |
1286 | * may process the whole bitmap in one go */ | |
b30ab791 | 1287 | extern void _drbd_bm_set_bits(struct drbd_device *device, |
b411b363 | 1288 | const unsigned long s, const unsigned long e); |
b30ab791 AG |
1289 | extern int drbd_bm_test_bit(struct drbd_device *device, unsigned long bitnr); |
1290 | extern int drbd_bm_e_weight(struct drbd_device *device, unsigned long enr); | |
8164dd6c AG |
1291 | extern int drbd_bm_read(struct drbd_device *device, |
1292 | struct drbd_peer_device *peer_device) __must_hold(local); | |
b30ab791 | 1293 | extern void drbd_bm_mark_for_writeout(struct drbd_device *device, int page_nr); |
8164dd6c AG |
1294 | extern int drbd_bm_write(struct drbd_device *device, |
1295 | struct drbd_peer_device *peer_device) __must_hold(local); | |
27ea1d87 | 1296 | extern void drbd_bm_reset_al_hints(struct drbd_device *device) __must_hold(local); |
b30ab791 | 1297 | extern int drbd_bm_write_hinted(struct drbd_device *device) __must_hold(local); |
c7a58db4 | 1298 | extern int drbd_bm_write_lazy(struct drbd_device *device, unsigned upper_idx) __must_hold(local); |
8164dd6c AG |
1299 | extern int drbd_bm_write_all(struct drbd_device *device, |
1300 | struct drbd_peer_device *peer_device) __must_hold(local); | |
1301 | extern int drbd_bm_write_copy_pages(struct drbd_device *device, | |
1302 | struct drbd_peer_device *peer_device) __must_hold(local); | |
b30ab791 AG |
1303 | extern size_t drbd_bm_words(struct drbd_device *device); |
1304 | extern unsigned long drbd_bm_bits(struct drbd_device *device); | |
1305 | extern sector_t drbd_bm_capacity(struct drbd_device *device); | |
4b0715f0 LE |
1306 | |
1307 | #define DRBD_END_OF_BITMAP (~(unsigned long)0) | |
b30ab791 | 1308 | extern unsigned long drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
b411b363 | 1309 | /* bm_find_next variants for use while you hold drbd_bm_lock() */ |
b30ab791 AG |
1310 | extern unsigned long _drbd_bm_find_next(struct drbd_device *device, unsigned long bm_fo); |
1311 | extern unsigned long _drbd_bm_find_next_zero(struct drbd_device *device, unsigned long bm_fo); | |
1312 | extern unsigned long _drbd_bm_total_weight(struct drbd_device *device); | |
1313 | extern unsigned long drbd_bm_total_weight(struct drbd_device *device); | |
b411b363 | 1314 | /* for receive_bitmap */ |
b30ab791 | 1315 | extern void drbd_bm_merge_lel(struct drbd_device *device, size_t offset, |
b411b363 | 1316 | size_t number, unsigned long *buffer); |
19f843aa | 1317 | /* for _drbd_send_bitmap */ |
b30ab791 | 1318 | extern void drbd_bm_get_lel(struct drbd_device *device, size_t offset, |
b411b363 PR |
1319 | size_t number, unsigned long *buffer); |
1320 | ||
b30ab791 AG |
1321 | extern void drbd_bm_lock(struct drbd_device *device, char *why, enum bm_flag flags); |
1322 | extern void drbd_bm_unlock(struct drbd_device *device); | |
b411b363 PR |
1323 | /* drbd_main.c */ |
1324 | ||
1325 | extern struct kmem_cache *drbd_request_cache; | |
6c852bec | 1326 | extern struct kmem_cache *drbd_ee_cache; /* peer requests */ |
b411b363 PR |
1327 | extern struct kmem_cache *drbd_bm_ext_cache; /* bitmap extents */ |
1328 | extern struct kmem_cache *drbd_al_ext_cache; /* activity log extents */ | |
0892fac8 KO |
1329 | extern mempool_t drbd_request_mempool; |
1330 | extern mempool_t drbd_ee_mempool; | |
b411b363 | 1331 | |
4281808f LE |
1332 | /* drbd's page pool, used to buffer data received from the peer, |
1333 | * or data requested by the peer. | |
1334 | * | |
1335 | * This does not have an emergency reserve. | |
1336 | * | |
1337 | * When allocating from this pool, it first takes pages from the pool. | |
1338 | * Only if the pool is depleted will try to allocate from the system. | |
1339 | * | |
1340 | * The assumption is that pages taken from this pool will be processed, | |
1341 | * and given back, "quickly", and then can be recycled, so we can avoid | |
1342 | * frequent calls to alloc_page(), and still will be able to make progress even | |
1343 | * under memory pressure. | |
1344 | */ | |
1345 | extern struct page *drbd_pp_pool; | |
b411b363 PR |
1346 | extern spinlock_t drbd_pp_lock; |
1347 | extern int drbd_pp_vacant; | |
1348 | extern wait_queue_head_t drbd_pp_wait; | |
1349 | ||
4281808f LE |
1350 | /* We also need a standard (emergency-reserve backed) page pool |
1351 | * for meta data IO (activity log, bitmap). | |
1352 | * We can keep it global, as long as it is used as "N pages at a time". | |
1353 | * 128 should be plenty, currently we probably can get away with as few as 1. | |
1354 | */ | |
1355 | #define DRBD_MIN_POOL_PAGES 128 | |
0892fac8 | 1356 | extern mempool_t drbd_md_io_page_pool; |
4281808f | 1357 | |
9476f39d LE |
1358 | /* We also need to make sure we get a bio |
1359 | * when we need it for housekeeping purposes */ | |
0892fac8 | 1360 | extern struct bio_set drbd_md_io_bio_set; |
9476f39d | 1361 | |
8cb0defb | 1362 | /* And a bio_set for cloning */ |
0892fac8 | 1363 | extern struct bio_set drbd_io_bio_set; |
8cb0defb | 1364 | |
28bc3b8c | 1365 | extern struct mutex resources_mutex; |
b411b363 | 1366 | |
bde89a9e | 1367 | extern int conn_lowest_minor(struct drbd_connection *connection); |
a910b123 | 1368 | extern enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsigned int minor); |
05a10ec7 | 1369 | extern void drbd_destroy_device(struct kref *kref); |
a910b123 | 1370 | extern void drbd_delete_device(struct drbd_device *device); |
b411b363 | 1371 | |
77c556f6 AG |
1372 | extern struct drbd_resource *drbd_create_resource(const char *name); |
1373 | extern void drbd_free_resource(struct drbd_resource *resource); | |
1374 | ||
eb6bea67 | 1375 | extern int set_resource_options(struct drbd_resource *resource, struct res_opts *res_opts); |
bde89a9e | 1376 | extern struct drbd_connection *conn_create(const char *name, struct res_opts *res_opts); |
05a10ec7 | 1377 | extern void drbd_destroy_connection(struct kref *kref); |
bde89a9e | 1378 | extern struct drbd_connection *conn_get_by_addrs(void *my_addr, int my_addr_len, |
089c075d | 1379 | void *peer_addr, int peer_addr_len); |
4bc76048 | 1380 | extern struct drbd_resource *drbd_find_resource(const char *name); |
77c556f6 | 1381 | extern void drbd_destroy_resource(struct kref *kref); |
bde89a9e | 1382 | extern void conn_free_crypto(struct drbd_connection *connection); |
b411b363 | 1383 | |
b411b363 | 1384 | /* drbd_req */ |
113fef9e | 1385 | extern void do_submit(struct work_struct *ws); |
370276ba | 1386 | extern void __drbd_make_request(struct drbd_device *, struct bio *); |
3e08773c | 1387 | void drbd_submit_bio(struct bio *bio); |
b411b363 PR |
1388 | |
1389 | /* drbd_nl.c */ | |
a2972846 AG |
1390 | |
1391 | extern struct mutex notification_mutex; | |
1392 | ||
b30ab791 AG |
1393 | extern void drbd_suspend_io(struct drbd_device *device); |
1394 | extern void drbd_resume_io(struct drbd_device *device); | |
b411b363 | 1395 | extern char *ppsize(char *buf, unsigned long long size); |
54761697 | 1396 | extern sector_t drbd_new_dev_size(struct drbd_device *, struct drbd_backing_dev *, sector_t, int); |
e96c9633 | 1397 | enum determine_dev_size { |
d752b269 PR |
1398 | DS_ERROR_SHRINK = -3, |
1399 | DS_ERROR_SPACE_MD = -2, | |
e96c9633 PR |
1400 | DS_ERROR = -1, |
1401 | DS_UNCHANGED = 0, | |
1402 | DS_SHRUNK = 1, | |
57737adc PR |
1403 | DS_GREW = 2, |
1404 | DS_GREW_FROM_ZERO = 3, | |
e96c9633 | 1405 | }; |
d752b269 | 1406 | extern enum determine_dev_size |
54761697 AG |
1407 | drbd_determine_dev_size(struct drbd_device *, enum dds_flags, struct resize_parms *) __must_hold(local); |
1408 | extern void resync_after_online_grow(struct drbd_device *); | |
9104d31a LE |
1409 | extern void drbd_reconsider_queue_parameters(struct drbd_device *device, |
1410 | struct drbd_backing_dev *bdev, struct o_qlim *o); | |
b30ab791 | 1411 | extern enum drbd_state_rv drbd_set_role(struct drbd_device *device, |
bf885f8a AG |
1412 | enum drbd_role new_role, |
1413 | int force); | |
bde89a9e AG |
1414 | extern bool conn_try_outdate_peer(struct drbd_connection *connection); |
1415 | extern void conn_try_outdate_peer_async(struct drbd_connection *connection); | |
7e5fec31 | 1416 | extern enum drbd_peer_state conn_khelper(struct drbd_connection *connection, char *cmd); |
b30ab791 | 1417 | extern int drbd_khelper(struct drbd_device *device, char *cmd); |
b411b363 PR |
1418 | |
1419 | /* drbd_worker.c */ | |
d40e5671 | 1420 | /* bi_end_io handlers */ |
4246a0b6 CH |
1421 | extern void drbd_md_endio(struct bio *bio); |
1422 | extern void drbd_peer_request_endio(struct bio *bio); | |
1423 | extern void drbd_request_endio(struct bio *bio); | |
b411b363 | 1424 | extern int drbd_worker(struct drbd_thread *thi); |
b30ab791 AG |
1425 | enum drbd_ret_code drbd_resync_after_valid(struct drbd_device *device, int o_minor); |
1426 | void drbd_resync_after_changed(struct drbd_device *device); | |
1427 | extern void drbd_start_resync(struct drbd_device *device, enum drbd_conns side); | |
1428 | extern void resume_next_sg(struct drbd_device *device); | |
1429 | extern void suspend_other_sg(struct drbd_device *device); | |
0d11f3cf | 1430 | extern int drbd_resync_finished(struct drbd_peer_device *peer_device); |
b411b363 | 1431 | /* maybe rather drbd_main.c ? */ |
e37d2438 | 1432 | extern void *drbd_md_get_buffer(struct drbd_device *device, const char *intent); |
b30ab791 AG |
1433 | extern void drbd_md_put_buffer(struct drbd_device *device); |
1434 | extern int drbd_md_sync_page_io(struct drbd_device *device, | |
9945172a | 1435 | struct drbd_backing_dev *bdev, sector_t sector, enum req_op op); |
0d11f3cf CB |
1436 | extern void drbd_ov_out_of_sync_found(struct drbd_peer_device *peer_device, |
1437 | sector_t sector, int size); | |
b30ab791 | 1438 | extern void wait_until_done_or_force_detached(struct drbd_device *device, |
44edfb0d | 1439 | struct drbd_backing_dev *bdev, unsigned int *done); |
0d11f3cf | 1440 | extern void drbd_rs_controller_reset(struct drbd_peer_device *peer_device); |
b411b363 | 1441 | |
0d11f3cf | 1442 | static inline void ov_out_of_sync_print(struct drbd_peer_device *peer_device) |
b411b363 | 1443 | { |
0d11f3cf CB |
1444 | struct drbd_device *device = peer_device->device; |
1445 | ||
b30ab791 | 1446 | if (device->ov_last_oos_size) { |
0d11f3cf | 1447 | drbd_err(peer_device, "Out of sync: start=%llu, size=%lu (sectors)\n", |
b30ab791 AG |
1448 | (unsigned long long)device->ov_last_oos_start, |
1449 | (unsigned long)device->ov_last_oos_size); | |
b411b363 | 1450 | } |
b30ab791 | 1451 | device->ov_last_oos_size = 0; |
b411b363 PR |
1452 | } |
1453 | ||
1454 | ||
3d0e6375 KC |
1455 | extern void drbd_csum_bio(struct crypto_shash *, struct bio *, void *); |
1456 | extern void drbd_csum_ee(struct crypto_shash *, struct drbd_peer_request *, | |
1457 | void *); | |
b411b363 | 1458 | /* worker callbacks */ |
99920dc5 AG |
1459 | extern int w_e_end_data_req(struct drbd_work *, int); |
1460 | extern int w_e_end_rsdata_req(struct drbd_work *, int); | |
1461 | extern int w_e_end_csum_rs_req(struct drbd_work *, int); | |
1462 | extern int w_e_end_ov_reply(struct drbd_work *, int); | |
1463 | extern int w_e_end_ov_req(struct drbd_work *, int); | |
1464 | extern int w_ov_finished(struct drbd_work *, int); | |
1465 | extern int w_resync_timer(struct drbd_work *, int); | |
1466 | extern int w_send_write_hint(struct drbd_work *, int); | |
99920dc5 | 1467 | extern int w_send_dblock(struct drbd_work *, int); |
99920dc5 | 1468 | extern int w_send_read_req(struct drbd_work *, int); |
99920dc5 | 1469 | extern int w_restart_disk_io(struct drbd_work *, int); |
8f7bed77 | 1470 | extern int w_send_out_of_sync(struct drbd_work *, int); |
b411b363 | 1471 | |
2bccef39 KC |
1472 | extern void resync_timer_fn(struct timer_list *t); |
1473 | extern void start_resync_timer_fn(struct timer_list *t); | |
b411b363 | 1474 | |
a0fb3c47 LE |
1475 | extern void drbd_endio_write_sec_final(struct drbd_peer_request *peer_req); |
1476 | ||
b411b363 | 1477 | /* drbd_receiver.c */ |
f31e583a LE |
1478 | extern int drbd_issue_discard_or_zero_out(struct drbd_device *device, |
1479 | sector_t start, unsigned int nr_sectors, int flags); | |
753c6191 | 1480 | extern int drbd_receiver(struct drbd_thread *thi); |
1c03e520 | 1481 | extern int drbd_ack_receiver(struct drbd_thread *thi); |
668700b4 | 1482 | extern void drbd_send_acks_wf(struct work_struct *ws); |
e8299874 | 1483 | extern bool drbd_rs_c_min_rate_throttle(struct drbd_device *device); |
0d11f3cf | 1484 | extern bool drbd_rs_should_slow_down(struct drbd_peer_device *peer_device, sector_t sector, |
ad3fee79 | 1485 | bool throttle_if_app_is_waiting); |
ce668b6d | 1486 | extern int drbd_submit_peer_request(struct drbd_peer_request *peer_req); |
54761697 | 1487 | extern int drbd_free_peer_reqs(struct drbd_device *, struct list_head *); |
69a22773 | 1488 | extern struct drbd_peer_request *drbd_alloc_peer_req(struct drbd_peer_device *, u64, |
0db55363 | 1489 | sector_t, unsigned int, |
9104d31a | 1490 | unsigned int, |
0db55363 | 1491 | gfp_t) __must_hold(local); |
54761697 | 1492 | extern void __drbd_free_peer_req(struct drbd_device *, struct drbd_peer_request *, |
3967deb1 AG |
1493 | int); |
1494 | #define drbd_free_peer_req(m,e) __drbd_free_peer_req(m, e, 0) | |
1495 | #define drbd_free_net_peer_req(m,e) __drbd_free_peer_req(m, e, 1) | |
69a22773 | 1496 | extern struct page *drbd_alloc_pages(struct drbd_peer_device *, unsigned int, bool); |
b30ab791 | 1497 | extern void _drbd_clear_done_ee(struct drbd_device *device, struct list_head *to_be_freed); |
69a22773 | 1498 | extern int drbd_connected(struct drbd_peer_device *); |
b411b363 | 1499 | |
d40e5671 | 1500 | /* sets the number of 512 byte sectors of our virtual device */ |
d5412e8d | 1501 | void drbd_set_my_capacity(struct drbd_device *device, sector_t size); |
d40e5671 PR |
1502 | |
1503 | /* | |
1504 | * used to submit our private bio | |
1505 | */ | |
ed00aabd | 1506 | static inline void drbd_submit_bio_noacct(struct drbd_device *device, |
d40e5671 PR |
1507 | int fault_type, struct bio *bio) |
1508 | { | |
1509 | __release(local); | |
309dca30 CH |
1510 | if (!bio->bi_bdev) { |
1511 | drbd_err(device, "drbd_submit_bio_noacct: bio->bi_bdev == NULL\n"); | |
4e4cbee9 | 1512 | bio->bi_status = BLK_STS_IOERR; |
4246a0b6 | 1513 | bio_endio(bio); |
d40e5671 PR |
1514 | return; |
1515 | } | |
1516 | ||
1517 | if (drbd_insert_fault(device, fault_type)) | |
4246a0b6 | 1518 | bio_io_error(bio); |
d40e5671 | 1519 | else |
ed00aabd | 1520 | submit_bio_noacct(bio); |
d40e5671 PR |
1521 | } |
1522 | ||
8fe39aac PR |
1523 | void drbd_bump_write_ordering(struct drbd_resource *resource, struct drbd_backing_dev *bdev, |
1524 | enum write_ordering_e wo); | |
b411b363 PR |
1525 | |
1526 | /* drbd_proc.c */ | |
1527 | extern struct proc_dir_entry *drbd_proc; | |
004fd11d | 1528 | int drbd_seq_show(struct seq_file *seq, void *v); |
b411b363 PR |
1529 | |
1530 | /* drbd_actlog.c */ | |
e4d7d6f4 | 1531 | extern bool drbd_al_begin_io_prepare(struct drbd_device *device, struct drbd_interval *i); |
b30ab791 | 1532 | extern int drbd_al_begin_io_nonblock(struct drbd_device *device, struct drbd_interval *i); |
4dd726f0 | 1533 | extern void drbd_al_begin_io_commit(struct drbd_device *device); |
b30ab791 | 1534 | extern bool drbd_al_begin_io_fastpath(struct drbd_device *device, struct drbd_interval *i); |
4dd726f0 | 1535 | extern void drbd_al_begin_io(struct drbd_device *device, struct drbd_interval *i); |
b30ab791 AG |
1536 | extern void drbd_al_complete_io(struct drbd_device *device, struct drbd_interval *i); |
1537 | extern void drbd_rs_complete_io(struct drbd_device *device, sector_t sector); | |
1538 | extern int drbd_rs_begin_io(struct drbd_device *device, sector_t sector); | |
0d11f3cf | 1539 | extern int drbd_try_rs_begin_io(struct drbd_peer_device *peer_device, sector_t sector); |
b30ab791 AG |
1540 | extern void drbd_rs_cancel_all(struct drbd_device *device); |
1541 | extern int drbd_rs_del_all(struct drbd_device *device); | |
0d11f3cf | 1542 | extern void drbd_rs_failed_io(struct drbd_peer_device *peer_device, |
b411b363 | 1543 | sector_t sector, int size); |
0d11f3cf | 1544 | extern void drbd_advance_rs_marks(struct drbd_peer_device *peer_device, unsigned long still_to_go); |
5ab7d2c0 LE |
1545 | |
1546 | enum update_sync_bits_mode { RECORD_RS_FAILED, SET_OUT_OF_SYNC, SET_IN_SYNC }; | |
0d11f3cf | 1547 | extern int __drbd_change_sync(struct drbd_peer_device *peer_device, sector_t sector, int size, |
179e20b8 | 1548 | enum update_sync_bits_mode mode); |
0d11f3cf CB |
1549 | #define drbd_set_in_sync(peer_device, sector, size) \ |
1550 | __drbd_change_sync(peer_device, sector, size, SET_IN_SYNC) | |
1551 | #define drbd_set_out_of_sync(peer_device, sector, size) \ | |
1552 | __drbd_change_sync(peer_device, sector, size, SET_OUT_OF_SYNC) | |
1553 | #define drbd_rs_failed_io(peer_device, sector, size) \ | |
1554 | __drbd_change_sync(peer_device, sector, size, RECORD_RS_FAILED) | |
b30ab791 | 1555 | extern void drbd_al_shrink(struct drbd_device *device); |
5f7c0124 | 1556 | extern int drbd_al_initialize(struct drbd_device *, void *); |
b411b363 | 1557 | |
b411b363 | 1558 | /* drbd_nl.c */ |
3b98c0c2 LE |
1559 | /* state info broadcast */ |
1560 | struct sib_info { | |
1561 | enum drbd_state_info_bcast_reason sib_reason; | |
1562 | union { | |
1563 | struct { | |
1564 | char *helper_name; | |
1565 | unsigned helper_exit_code; | |
1566 | }; | |
1567 | struct { | |
1568 | union drbd_state os; | |
1569 | union drbd_state ns; | |
1570 | }; | |
1571 | }; | |
1572 | }; | |
b30ab791 | 1573 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib); |
b411b363 | 1574 | |
aadb22ba | 1575 | extern int notify_resource_state(struct sk_buff *, |
a2972846 AG |
1576 | unsigned int, |
1577 | struct drbd_resource *, | |
1578 | struct resource_info *, | |
1579 | enum drbd_notification_type); | |
aadb22ba | 1580 | extern int notify_device_state(struct sk_buff *, |
a2972846 AG |
1581 | unsigned int, |
1582 | struct drbd_device *, | |
1583 | struct device_info *, | |
1584 | enum drbd_notification_type); | |
aadb22ba | 1585 | extern int notify_connection_state(struct sk_buff *, |
a2972846 AG |
1586 | unsigned int, |
1587 | struct drbd_connection *, | |
1588 | struct connection_info *, | |
1589 | enum drbd_notification_type); | |
aadb22ba | 1590 | extern int notify_peer_device_state(struct sk_buff *, |
a2972846 AG |
1591 | unsigned int, |
1592 | struct drbd_peer_device *, | |
1593 | struct peer_device_info *, | |
1594 | enum drbd_notification_type); | |
1595 | extern void notify_helper(enum drbd_notification_type, struct drbd_device *, | |
1596 | struct drbd_connection *, const char *, int); | |
1597 | ||
b411b363 PR |
1598 | /* |
1599 | * inline helper functions | |
1600 | *************************/ | |
1601 | ||
45bb912b LE |
1602 | /* see also page_chain_add and friends in drbd_receiver.c */ |
1603 | static inline struct page *page_chain_next(struct page *page) | |
1604 | { | |
1605 | return (struct page *)page_private(page); | |
1606 | } | |
1607 | #define page_chain_for_each(page) \ | |
1608 | for (; page && ({ prefetch(page_chain_next(page)); 1; }); \ | |
1609 | page = page_chain_next(page)) | |
1610 | #define page_chain_for_each_safe(page, n) \ | |
1611 | for (; page && ({ n = page_chain_next(page); 1; }); page = n) | |
1612 | ||
45bb912b | 1613 | |
045417f7 | 1614 | static inline int drbd_peer_req_has_active_page(struct drbd_peer_request *peer_req) |
45bb912b | 1615 | { |
db830c46 | 1616 | struct page *page = peer_req->pages; |
45bb912b LE |
1617 | page_chain_for_each(page) { |
1618 | if (page_count(page) > 1) | |
1619 | return 1; | |
1620 | } | |
1621 | return 0; | |
1622 | } | |
1623 | ||
b30ab791 | 1624 | static inline union drbd_state drbd_read_state(struct drbd_device *device) |
b411b363 | 1625 | { |
6bbf53ca | 1626 | struct drbd_resource *resource = device->resource; |
78bae59b PR |
1627 | union drbd_state rv; |
1628 | ||
b30ab791 | 1629 | rv.i = device->state.i; |
6bbf53ca AG |
1630 | rv.susp = resource->susp; |
1631 | rv.susp_nod = resource->susp_nod; | |
1632 | rv.susp_fen = resource->susp_fen; | |
78bae59b PR |
1633 | |
1634 | return rv; | |
b411b363 PR |
1635 | } |
1636 | ||
383606e0 | 1637 | enum drbd_force_detach_flags { |
a2a3c74f LE |
1638 | DRBD_READ_ERROR, |
1639 | DRBD_WRITE_ERROR, | |
383606e0 LE |
1640 | DRBD_META_IO_ERROR, |
1641 | DRBD_FORCE_DETACH, | |
1642 | }; | |
1643 | ||
b411b363 | 1644 | #define __drbd_chk_io_error(m,f) __drbd_chk_io_error_(m,f, __func__) |
b30ab791 | 1645 | static inline void __drbd_chk_io_error_(struct drbd_device *device, |
a2a3c74f | 1646 | enum drbd_force_detach_flags df, |
383606e0 | 1647 | const char *where) |
b411b363 | 1648 | { |
daeda1cc PR |
1649 | enum drbd_io_error_p ep; |
1650 | ||
1651 | rcu_read_lock(); | |
b30ab791 | 1652 | ep = rcu_dereference(device->ldev->disk_conf)->on_io_error; |
daeda1cc PR |
1653 | rcu_read_unlock(); |
1654 | switch (ep) { | |
1655 | case EP_PASS_ON: /* FIXME would this be better named "Ignore"? */ | |
a2a3c74f | 1656 | if (df == DRBD_READ_ERROR || df == DRBD_WRITE_ERROR) { |
e3fa02d7 | 1657 | if (drbd_ratelimit()) |
d0180171 | 1658 | drbd_err(device, "Local IO failed in %s.\n", where); |
b30ab791 AG |
1659 | if (device->state.disk > D_INCONSISTENT) |
1660 | _drbd_set_state(_NS(device, disk, D_INCONSISTENT), CS_HARD, NULL); | |
b411b363 PR |
1661 | break; |
1662 | } | |
df561f66 | 1663 | fallthrough; /* for DRBD_META_IO_ERROR or DRBD_FORCE_DETACH */ |
b411b363 PR |
1664 | case EP_DETACH: |
1665 | case EP_CALL_HELPER: | |
a2a3c74f LE |
1666 | /* Remember whether we saw a READ or WRITE error. |
1667 | * | |
1668 | * Recovery of the affected area for WRITE failure is covered | |
1669 | * by the activity log. | |
1670 | * READ errors may fall outside that area though. Certain READ | |
1671 | * errors can be "healed" by writing good data to the affected | |
1672 | * blocks, which triggers block re-allocation in lower layers. | |
1673 | * | |
1674 | * If we can not write the bitmap after a READ error, | |
1675 | * we may need to trigger a full sync (see w_go_diskless()). | |
1676 | * | |
1677 | * Force-detach is not really an IO error, but rather a | |
1678 | * desperate measure to try to deal with a completely | |
1679 | * unresponsive lower level IO stack. | |
1680 | * Still it should be treated as a WRITE error. | |
1681 | * | |
1682 | * Meta IO error is always WRITE error: | |
1683 | * we read meta data only once during attach, | |
1684 | * which will fail in case of errors. | |
1685 | */ | |
b30ab791 | 1686 | set_bit(WAS_IO_ERROR, &device->flags); |
a2a3c74f | 1687 | if (df == DRBD_READ_ERROR) |
b30ab791 | 1688 | set_bit(WAS_READ_ERROR, &device->flags); |
a2a3c74f | 1689 | if (df == DRBD_FORCE_DETACH) |
b30ab791 AG |
1690 | set_bit(FORCE_DETACH, &device->flags); |
1691 | if (device->state.disk > D_FAILED) { | |
1692 | _drbd_set_state(_NS(device, disk, D_FAILED), CS_HARD, NULL); | |
d0180171 | 1693 | drbd_err(device, |
82f59cc6 | 1694 | "Local IO failed in %s. Detaching...\n", where); |
b411b363 PR |
1695 | } |
1696 | break; | |
1697 | } | |
1698 | } | |
1699 | ||
1700 | /** | |
1701 | * drbd_chk_io_error: Handle the on_io_error setting, should be called from all io completion handlers | |
b30ab791 | 1702 | * @device: DRBD device. |
b411b363 PR |
1703 | * @error: Error code passed to the IO completion callback |
1704 | * @forcedetach: Force detach. I.e. the error happened while accessing the meta data | |
1705 | * | |
1706 | * See also drbd_main.c:after_state_ch() if (os.disk > D_FAILED && ns.disk == D_FAILED) | |
1707 | */ | |
1708 | #define drbd_chk_io_error(m,e,f) drbd_chk_io_error_(m,e,f, __func__) | |
b30ab791 | 1709 | static inline void drbd_chk_io_error_(struct drbd_device *device, |
383606e0 | 1710 | int error, enum drbd_force_detach_flags forcedetach, const char *where) |
b411b363 PR |
1711 | { |
1712 | if (error) { | |
1713 | unsigned long flags; | |
0500813f | 1714 | spin_lock_irqsave(&device->resource->req_lock, flags); |
b30ab791 | 1715 | __drbd_chk_io_error_(device, forcedetach, where); |
0500813f | 1716 | spin_unlock_irqrestore(&device->resource->req_lock, flags); |
b411b363 PR |
1717 | } |
1718 | } | |
1719 | ||
1720 | ||
1721 | /** | |
1722 | * drbd_md_first_sector() - Returns the first sector number of the meta data area | |
1723 | * @bdev: Meta data block device. | |
1724 | * | |
1725 | * BTW, for internal meta data, this happens to be the maximum capacity | |
1726 | * we could agree upon with our peer node. | |
1727 | */ | |
68e41a43 | 1728 | static inline sector_t drbd_md_first_sector(struct drbd_backing_dev *bdev) |
b411b363 | 1729 | { |
68e41a43 | 1730 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1731 | case DRBD_MD_INDEX_INTERNAL: |
1732 | case DRBD_MD_INDEX_FLEX_INT: | |
1733 | return bdev->md.md_offset + bdev->md.bm_offset; | |
1734 | case DRBD_MD_INDEX_FLEX_EXT: | |
1735 | default: | |
1736 | return bdev->md.md_offset; | |
1737 | } | |
1738 | } | |
1739 | ||
1740 | /** | |
1741 | * drbd_md_last_sector() - Return the last sector number of the meta data area | |
1742 | * @bdev: Meta data block device. | |
1743 | */ | |
1744 | static inline sector_t drbd_md_last_sector(struct drbd_backing_dev *bdev) | |
1745 | { | |
68e41a43 | 1746 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1747 | case DRBD_MD_INDEX_INTERNAL: |
1748 | case DRBD_MD_INDEX_FLEX_INT: | |
ae8bf312 | 1749 | return bdev->md.md_offset + MD_4kB_SECT -1; |
b411b363 PR |
1750 | case DRBD_MD_INDEX_FLEX_EXT: |
1751 | default: | |
ae8bf312 | 1752 | return bdev->md.md_offset + bdev->md.md_size_sect -1; |
b411b363 PR |
1753 | } |
1754 | } | |
1755 | ||
1756 | /* Returns the number of 512 byte sectors of the device */ | |
1757 | static inline sector_t drbd_get_capacity(struct block_device *bdev) | |
1758 | { | |
da7b3924 | 1759 | return bdev ? bdev_nr_sectors(bdev) : 0; |
b411b363 PR |
1760 | } |
1761 | ||
1762 | /** | |
1763 | * drbd_get_max_capacity() - Returns the capacity we announce to out peer | |
1764 | * @bdev: Meta data block device. | |
1765 | * | |
1766 | * returns the capacity we announce to out peer. we clip ourselves at the | |
1767 | * various MAX_SECTORS, because if we don't, current implementation will | |
1768 | * oops sooner or later | |
1769 | */ | |
1770 | static inline sector_t drbd_get_max_capacity(struct drbd_backing_dev *bdev) | |
1771 | { | |
1772 | sector_t s; | |
daeda1cc | 1773 | |
68e41a43 | 1774 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
1775 | case DRBD_MD_INDEX_INTERNAL: |
1776 | case DRBD_MD_INDEX_FLEX_INT: | |
1777 | s = drbd_get_capacity(bdev->backing_bdev) | |
1778 | ? min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
68e41a43 | 1779 | drbd_md_first_sector(bdev)) |
b411b363 PR |
1780 | : 0; |
1781 | break; | |
1782 | case DRBD_MD_INDEX_FLEX_EXT: | |
1783 | s = min_t(sector_t, DRBD_MAX_SECTORS_FLEX, | |
1784 | drbd_get_capacity(bdev->backing_bdev)); | |
1785 | /* clip at maximum size the meta device can support */ | |
1786 | s = min_t(sector_t, s, | |
1787 | BM_EXT_TO_SECT(bdev->md.md_size_sect | |
1788 | - bdev->md.bm_offset)); | |
1789 | break; | |
1790 | default: | |
1791 | s = min_t(sector_t, DRBD_MAX_SECTORS, | |
1792 | drbd_get_capacity(bdev->backing_bdev)); | |
1793 | } | |
1794 | return s; | |
1795 | } | |
1796 | ||
1797 | /** | |
3a4d4eb3 | 1798 | * drbd_md_ss() - Return the sector number of our meta data super block |
b411b363 PR |
1799 | * @bdev: Meta data block device. |
1800 | */ | |
3a4d4eb3 | 1801 | static inline sector_t drbd_md_ss(struct drbd_backing_dev *bdev) |
b411b363 | 1802 | { |
3a4d4eb3 | 1803 | const int meta_dev_idx = bdev->md.meta_dev_idx; |
daeda1cc | 1804 | |
3a4d4eb3 LE |
1805 | if (meta_dev_idx == DRBD_MD_INDEX_FLEX_EXT) |
1806 | return 0; | |
daeda1cc | 1807 | |
3a4d4eb3 LE |
1808 | /* Since drbd08, internal meta data is always "flexible". |
1809 | * position: last 4k aligned block of 4k size */ | |
1810 | if (meta_dev_idx == DRBD_MD_INDEX_INTERNAL || | |
1811 | meta_dev_idx == DRBD_MD_INDEX_FLEX_INT) | |
ae8bf312 | 1812 | return (drbd_get_capacity(bdev->backing_bdev) & ~7ULL) - 8; |
3a4d4eb3 LE |
1813 | |
1814 | /* external, some index; this is the old fixed size layout */ | |
1815 | return MD_128MB_SECT * bdev->md.meta_dev_idx; | |
b411b363 PR |
1816 | } |
1817 | ||
b411b363 PR |
1818 | static inline void |
1819 | drbd_queue_work(struct drbd_work_queue *q, struct drbd_work *w) | |
1820 | { | |
1821 | unsigned long flags; | |
1822 | spin_lock_irqsave(&q->q_lock, flags); | |
1823 | list_add_tail(&w->list, &q->q); | |
b411b363 | 1824 | spin_unlock_irqrestore(&q->q_lock, flags); |
8c0785a5 | 1825 | wake_up(&q->q_wait); |
b411b363 PR |
1826 | } |
1827 | ||
15e26f6a LE |
1828 | static inline void |
1829 | drbd_queue_work_if_unqueued(struct drbd_work_queue *q, struct drbd_work *w) | |
1830 | { | |
1831 | unsigned long flags; | |
1832 | spin_lock_irqsave(&q->q_lock, flags); | |
1833 | if (list_empty_careful(&w->list)) | |
1834 | list_add_tail(&w->list, &q->q); | |
1835 | spin_unlock_irqrestore(&q->q_lock, flags); | |
1836 | wake_up(&q->q_wait); | |
1837 | } | |
1838 | ||
e334f550 LE |
1839 | static inline void |
1840 | drbd_device_post_work(struct drbd_device *device, int work_bit) | |
1841 | { | |
1842 | if (!test_and_set_bit(work_bit, &device->flags)) { | |
1843 | struct drbd_connection *connection = | |
1844 | first_peer_device(device)->connection; | |
1845 | struct drbd_work_queue *q = &connection->sender_work; | |
1846 | if (!test_and_set_bit(DEVICE_WORK_PENDING, &connection->flags)) | |
1847 | wake_up(&q->q_wait); | |
1848 | } | |
1849 | } | |
1850 | ||
b5043c5e AG |
1851 | extern void drbd_flush_workqueue(struct drbd_work_queue *work_queue); |
1852 | ||
668700b4 PR |
1853 | /* To get the ack_receiver out of the blocking network stack, |
1854 | * so it can change its sk_rcvtimeo from idle- to ping-timeout, | |
1855 | * and send a ping, we need to send a signal. | |
1856 | * Which signal we send is irrelevant. */ | |
1857 | static inline void wake_ack_receiver(struct drbd_connection *connection) | |
b411b363 | 1858 | { |
668700b4 PR |
1859 | struct task_struct *task = connection->ack_receiver.task; |
1860 | if (task && get_t_state(&connection->ack_receiver) == RUNNING) | |
fee10990 | 1861 | send_sig(SIGXCPU, task, 1); |
b411b363 PR |
1862 | } |
1863 | ||
bde89a9e | 1864 | static inline void request_ping(struct drbd_connection *connection) |
b411b363 | 1865 | { |
bde89a9e | 1866 | set_bit(SEND_PING, &connection->flags); |
668700b4 | 1867 | wake_ack_receiver(connection); |
b411b363 PR |
1868 | } |
1869 | ||
bde89a9e | 1870 | extern void *conn_prepare_command(struct drbd_connection *, struct drbd_socket *); |
69a22773 | 1871 | extern void *drbd_prepare_command(struct drbd_peer_device *, struct drbd_socket *); |
bde89a9e | 1872 | extern int conn_send_command(struct drbd_connection *, struct drbd_socket *, |
dba58587 AG |
1873 | enum drbd_packet, unsigned int, void *, |
1874 | unsigned int); | |
69a22773 | 1875 | extern int drbd_send_command(struct drbd_peer_device *, struct drbd_socket *, |
dba58587 AG |
1876 | enum drbd_packet, unsigned int, void *, |
1877 | unsigned int); | |
b411b363 | 1878 | |
bde89a9e AG |
1879 | extern int drbd_send_ping(struct drbd_connection *connection); |
1880 | extern int drbd_send_ping_ack(struct drbd_connection *connection); | |
69a22773 | 1881 | extern int drbd_send_state_req(struct drbd_peer_device *, union drbd_state, union drbd_state); |
bde89a9e | 1882 | extern int conn_send_state_req(struct drbd_connection *, union drbd_state, union drbd_state); |
b411b363 PR |
1883 | |
1884 | static inline void drbd_thread_stop(struct drbd_thread *thi) | |
1885 | { | |
81e84650 | 1886 | _drbd_thread_stop(thi, false, true); |
b411b363 PR |
1887 | } |
1888 | ||
1889 | static inline void drbd_thread_stop_nowait(struct drbd_thread *thi) | |
1890 | { | |
81e84650 | 1891 | _drbd_thread_stop(thi, false, false); |
b411b363 PR |
1892 | } |
1893 | ||
1894 | static inline void drbd_thread_restart_nowait(struct drbd_thread *thi) | |
1895 | { | |
81e84650 | 1896 | _drbd_thread_stop(thi, true, false); |
b411b363 PR |
1897 | } |
1898 | ||
1899 | /* counts how many answer packets packets we expect from our peer, | |
1900 | * for either explicit application requests, | |
1901 | * or implicit barrier packets as necessary. | |
1902 | * increased: | |
1903 | * w_send_barrier | |
8554df1c | 1904 | * _req_mod(req, QUEUE_FOR_NET_WRITE or QUEUE_FOR_NET_READ); |
b411b363 PR |
1905 | * it is much easier and equally valid to count what we queue for the |
1906 | * worker, even before it actually was queued or send. | |
1907 | * (drbd_make_request_common; recovery path on read io-error) | |
1908 | * decreased: | |
1909 | * got_BarrierAck (respective tl_clear, tl_clear_barrier) | |
8554df1c | 1910 | * _req_mod(req, DATA_RECEIVED) |
b411b363 | 1911 | * [from receive_DataReply] |
8554df1c | 1912 | * _req_mod(req, WRITE_ACKED_BY_PEER or RECV_ACKED_BY_PEER or NEG_ACKED) |
b411b363 PR |
1913 | * [from got_BlockAck (P_WRITE_ACK, P_RECV_ACK)] |
1914 | * for some reason it is NOT decreased in got_NegAck, | |
1915 | * but in the resulting cleanup code from report_params. | |
1916 | * we should try to remember the reason for that... | |
8554df1c AG |
1917 | * _req_mod(req, SEND_FAILED or SEND_CANCELED) |
1918 | * _req_mod(req, CONNECTION_LOST_WHILE_PENDING) | |
b411b363 PR |
1919 | * [from tl_clear_barrier] |
1920 | */ | |
b30ab791 | 1921 | static inline void inc_ap_pending(struct drbd_device *device) |
b411b363 | 1922 | { |
b30ab791 | 1923 | atomic_inc(&device->ap_pending_cnt); |
b411b363 PR |
1924 | } |
1925 | ||
33f7d316 AG |
1926 | #define dec_ap_pending(device) ((void)expect((device), __dec_ap_pending(device) >= 0)) |
1927 | static inline int __dec_ap_pending(struct drbd_device *device) | |
49559d87 | 1928 | { |
33f7d316 AG |
1929 | int ap_pending_cnt = atomic_dec_return(&device->ap_pending_cnt); |
1930 | ||
1931 | if (ap_pending_cnt == 0) | |
b30ab791 | 1932 | wake_up(&device->misc_wait); |
33f7d316 | 1933 | return ap_pending_cnt; |
49559d87 | 1934 | } |
b411b363 PR |
1935 | |
1936 | /* counts how many resync-related answers we still expect from the peer | |
1937 | * increase decrease | |
1938 | * C_SYNC_TARGET sends P_RS_DATA_REQUEST (and expects P_RS_DATA_REPLY) | |
25985edc | 1939 | * C_SYNC_SOURCE sends P_RS_DATA_REPLY (and expects P_WRITE_ACK with ID_SYNCER) |
b411b363 PR |
1940 | * (or P_NEG_ACK with ID_SYNCER) |
1941 | */ | |
0d11f3cf | 1942 | static inline void inc_rs_pending(struct drbd_peer_device *peer_device) |
b411b363 | 1943 | { |
0d11f3cf | 1944 | atomic_inc(&peer_device->device->rs_pending_cnt); |
b411b363 PR |
1945 | } |
1946 | ||
0d11f3cf CB |
1947 | #define dec_rs_pending(peer_device) \ |
1948 | ((void)expect((peer_device), __dec_rs_pending(peer_device) >= 0)) | |
1949 | static inline int __dec_rs_pending(struct drbd_peer_device *peer_device) | |
49559d87 | 1950 | { |
0d11f3cf | 1951 | return atomic_dec_return(&peer_device->device->rs_pending_cnt); |
49559d87 | 1952 | } |
b411b363 PR |
1953 | |
1954 | /* counts how many answers we still need to send to the peer. | |
1955 | * increased on | |
1956 | * receive_Data unless protocol A; | |
1957 | * we need to send a P_RECV_ACK (proto B) | |
1958 | * or P_WRITE_ACK (proto C) | |
1959 | * receive_RSDataReply (recv_resync_read) we need to send a P_WRITE_ACK | |
1960 | * receive_DataRequest (receive_RSDataRequest) we need to send back P_DATA | |
1961 | * receive_Barrier_* we need to send a P_BARRIER_ACK | |
1962 | */ | |
b30ab791 | 1963 | static inline void inc_unacked(struct drbd_device *device) |
b411b363 | 1964 | { |
b30ab791 | 1965 | atomic_inc(&device->unacked_cnt); |
b411b363 PR |
1966 | } |
1967 | ||
33f7d316 AG |
1968 | #define dec_unacked(device) ((void)expect(device, __dec_unacked(device) >= 0)) |
1969 | static inline int __dec_unacked(struct drbd_device *device) | |
b411b363 | 1970 | { |
33f7d316 | 1971 | return atomic_dec_return(&device->unacked_cnt); |
b411b363 PR |
1972 | } |
1973 | ||
33f7d316 AG |
1974 | #define sub_unacked(device, n) ((void)expect(device, __sub_unacked(device) >= 0)) |
1975 | static inline int __sub_unacked(struct drbd_device *device, int n) | |
b411b363 | 1976 | { |
33f7d316 | 1977 | return atomic_sub_return(n, &device->unacked_cnt); |
b411b363 PR |
1978 | } |
1979 | ||
5052fee2 LE |
1980 | static inline bool is_sync_target_state(enum drbd_conns connection_state) |
1981 | { | |
1982 | return connection_state == C_SYNC_TARGET || | |
1983 | connection_state == C_PAUSED_SYNC_T; | |
1984 | } | |
1985 | ||
1986 | static inline bool is_sync_source_state(enum drbd_conns connection_state) | |
1987 | { | |
1988 | return connection_state == C_SYNC_SOURCE || | |
1989 | connection_state == C_PAUSED_SYNC_S; | |
1990 | } | |
1991 | ||
5ab7d2c0 LE |
1992 | static inline bool is_sync_state(enum drbd_conns connection_state) |
1993 | { | |
5052fee2 LE |
1994 | return is_sync_source_state(connection_state) || |
1995 | is_sync_target_state(connection_state); | |
5ab7d2c0 LE |
1996 | } |
1997 | ||
b411b363 | 1998 | /** |
b30ab791 | 1999 | * get_ldev() - Increase the ref count on device->ldev. Returns 0 if there is no ldev |
d1b80853 AG |
2000 | * @_device: DRBD device. |
2001 | * @_min_state: Minimum device state required for success. | |
b411b363 | 2002 | * |
b30ab791 | 2003 | * You have to call put_ldev() when finished working with device->ldev. |
b411b363 | 2004 | */ |
d1b80853 AG |
2005 | #define get_ldev_if_state(_device, _min_state) \ |
2006 | (_get_ldev_if_state((_device), (_min_state)) ? \ | |
2007 | ({ __acquire(x); true; }) : false) | |
2008 | #define get_ldev(_device) get_ldev_if_state(_device, D_INCONSISTENT) | |
b411b363 | 2009 | |
b30ab791 | 2010 | static inline void put_ldev(struct drbd_device *device) |
b411b363 | 2011 | { |
11f8b2b6 | 2012 | enum drbd_disk_state disk_state = device->state.disk; |
ba3c6fb8 LE |
2013 | /* We must check the state *before* the atomic_dec becomes visible, |
2014 | * or we have a theoretical race where someone hitting zero, | |
2015 | * while state still D_FAILED, will then see D_DISKLESS in the | |
2016 | * condition below and calling into destroy, where he must not, yet. */ | |
b30ab791 | 2017 | int i = atomic_dec_return(&device->local_cnt); |
9a0d9d03 LE |
2018 | |
2019 | /* This may be called from some endio handler, | |
2020 | * so we must not sleep here. */ | |
2021 | ||
b411b363 | 2022 | __release(local); |
0b0ba1ef | 2023 | D_ASSERT(device, i >= 0); |
e9e6f3ec | 2024 | if (i == 0) { |
11f8b2b6 | 2025 | if (disk_state == D_DISKLESS) |
82f59cc6 | 2026 | /* even internal references gone, safe to destroy */ |
e334f550 | 2027 | drbd_device_post_work(device, DESTROY_DISK); |
11f8b2b6 | 2028 | if (disk_state == D_FAILED) |
82f59cc6 | 2029 | /* all application IO references gone. */ |
e334f550 LE |
2030 | if (!test_and_set_bit(GOING_DISKLESS, &device->flags)) |
2031 | drbd_device_post_work(device, GO_DISKLESS); | |
b30ab791 | 2032 | wake_up(&device->misc_wait); |
e9e6f3ec | 2033 | } |
b411b363 PR |
2034 | } |
2035 | ||
2036 | #ifndef __CHECKER__ | |
b30ab791 | 2037 | static inline int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins) |
b411b363 PR |
2038 | { |
2039 | int io_allowed; | |
2040 | ||
82f59cc6 | 2041 | /* never get a reference while D_DISKLESS */ |
b30ab791 | 2042 | if (device->state.disk == D_DISKLESS) |
82f59cc6 LE |
2043 | return 0; |
2044 | ||
b30ab791 AG |
2045 | atomic_inc(&device->local_cnt); |
2046 | io_allowed = (device->state.disk >= mins); | |
b411b363 | 2047 | if (!io_allowed) |
b30ab791 | 2048 | put_ldev(device); |
b411b363 PR |
2049 | return io_allowed; |
2050 | } | |
2051 | #else | |
b30ab791 | 2052 | extern int _get_ldev_if_state(struct drbd_device *device, enum drbd_disk_state mins); |
b411b363 PR |
2053 | #endif |
2054 | ||
b411b363 PR |
2055 | /* this throttles on-the-fly application requests |
2056 | * according to max_buffers settings; | |
2057 | * maybe re-implement using semaphores? */ | |
b30ab791 | 2058 | static inline int drbd_get_max_buffers(struct drbd_device *device) |
b411b363 | 2059 | { |
44ed167d PR |
2060 | struct net_conf *nc; |
2061 | int mxb; | |
2062 | ||
2063 | rcu_read_lock(); | |
a6b32bc3 | 2064 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
2065 | mxb = nc ? nc->max_buffers : 1000000; /* arbitrary limit on open requests */ |
2066 | rcu_read_unlock(); | |
2067 | ||
b411b363 PR |
2068 | return mxb; |
2069 | } | |
2070 | ||
b30ab791 | 2071 | static inline int drbd_state_is_stable(struct drbd_device *device) |
b411b363 | 2072 | { |
b30ab791 | 2073 | union drbd_dev_state s = device->state; |
b411b363 PR |
2074 | |
2075 | /* DO NOT add a default clause, we want the compiler to warn us | |
2076 | * for any newly introduced state we may have forgotten to add here */ | |
2077 | ||
2078 | switch ((enum drbd_conns)s.conn) { | |
2079 | /* new io only accepted when there is no connection, ... */ | |
2080 | case C_STANDALONE: | |
2081 | case C_WF_CONNECTION: | |
2082 | /* ... or there is a well established connection. */ | |
2083 | case C_CONNECTED: | |
2084 | case C_SYNC_SOURCE: | |
2085 | case C_SYNC_TARGET: | |
2086 | case C_VERIFY_S: | |
2087 | case C_VERIFY_T: | |
2088 | case C_PAUSED_SYNC_S: | |
2089 | case C_PAUSED_SYNC_T: | |
67531718 PR |
2090 | case C_AHEAD: |
2091 | case C_BEHIND: | |
3719094e | 2092 | /* transitional states, IO allowed */ |
b411b363 PR |
2093 | case C_DISCONNECTING: |
2094 | case C_UNCONNECTED: | |
2095 | case C_TIMEOUT: | |
2096 | case C_BROKEN_PIPE: | |
2097 | case C_NETWORK_FAILURE: | |
2098 | case C_PROTOCOL_ERROR: | |
2099 | case C_TEAR_DOWN: | |
2100 | case C_WF_REPORT_PARAMS: | |
2101 | case C_STARTING_SYNC_S: | |
2102 | case C_STARTING_SYNC_T: | |
3719094e PR |
2103 | break; |
2104 | ||
2105 | /* Allow IO in BM exchange states with new protocols */ | |
b411b363 | 2106 | case C_WF_BITMAP_S: |
a6b32bc3 | 2107 | if (first_peer_device(device)->connection->agreed_pro_version < 96) |
3719094e PR |
2108 | return 0; |
2109 | break; | |
2110 | ||
2111 | /* no new io accepted in these states */ | |
b411b363 PR |
2112 | case C_WF_BITMAP_T: |
2113 | case C_WF_SYNC_UUID: | |
2114 | case C_MASK: | |
2115 | /* not "stable" */ | |
2116 | return 0; | |
2117 | } | |
2118 | ||
2119 | switch ((enum drbd_disk_state)s.disk) { | |
2120 | case D_DISKLESS: | |
2121 | case D_INCONSISTENT: | |
2122 | case D_OUTDATED: | |
2123 | case D_CONSISTENT: | |
2124 | case D_UP_TO_DATE: | |
5ca1de03 | 2125 | case D_FAILED: |
b411b363 PR |
2126 | /* disk state is stable as well. */ |
2127 | break; | |
2128 | ||
d942ae44 | 2129 | /* no new io accepted during transitional states */ |
b411b363 | 2130 | case D_ATTACHING: |
b411b363 PR |
2131 | case D_NEGOTIATING: |
2132 | case D_UNKNOWN: | |
2133 | case D_MASK: | |
2134 | /* not "stable" */ | |
2135 | return 0; | |
2136 | } | |
2137 | ||
2138 | return 1; | |
2139 | } | |
2140 | ||
b30ab791 | 2141 | static inline int drbd_suspended(struct drbd_device *device) |
fb22c402 | 2142 | { |
6bbf53ca | 2143 | struct drbd_resource *resource = device->resource; |
8e0af25f | 2144 | |
6bbf53ca | 2145 | return resource->susp || resource->susp_fen || resource->susp_nod; |
fb22c402 PR |
2146 | } |
2147 | ||
b30ab791 | 2148 | static inline bool may_inc_ap_bio(struct drbd_device *device) |
b411b363 | 2149 | { |
b30ab791 | 2150 | int mxb = drbd_get_max_buffers(device); |
b411b363 | 2151 | |
b30ab791 | 2152 | if (drbd_suspended(device)) |
1b881ef7 | 2153 | return false; |
7dbb4386 | 2154 | if (atomic_read(&device->suspend_cnt)) |
1b881ef7 | 2155 | return false; |
b411b363 PR |
2156 | |
2157 | /* to avoid potential deadlock or bitmap corruption, | |
2158 | * in various places, we only allow new application io | |
2159 | * to start during "stable" states. */ | |
2160 | ||
2161 | /* no new io accepted when attaching or detaching the disk */ | |
b30ab791 | 2162 | if (!drbd_state_is_stable(device)) |
1b881ef7 | 2163 | return false; |
b411b363 PR |
2164 | |
2165 | /* since some older kernels don't have atomic_add_unless, | |
2166 | * and we are within the spinlock anyways, we have this workaround. */ | |
b30ab791 | 2167 | if (atomic_read(&device->ap_bio_cnt) > mxb) |
1b881ef7 | 2168 | return false; |
b30ab791 | 2169 | if (test_bit(BITMAP_IO, &device->flags)) |
1b881ef7 AG |
2170 | return false; |
2171 | return true; | |
b411b363 PR |
2172 | } |
2173 | ||
b30ab791 | 2174 | static inline bool inc_ap_bio_cond(struct drbd_device *device) |
b411b363 | 2175 | { |
1b881ef7 | 2176 | bool rv = false; |
8869d683 | 2177 | |
0500813f | 2178 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 2179 | rv = may_inc_ap_bio(device); |
8869d683 | 2180 | if (rv) |
b30ab791 | 2181 | atomic_inc(&device->ap_bio_cnt); |
0500813f | 2182 | spin_unlock_irq(&device->resource->req_lock); |
8869d683 PR |
2183 | |
2184 | return rv; | |
2185 | } | |
b411b363 | 2186 | |
b30ab791 | 2187 | static inline void inc_ap_bio(struct drbd_device *device) |
8869d683 | 2188 | { |
b411b363 PR |
2189 | /* we wait here |
2190 | * as long as the device is suspended | |
2191 | * until the bitmap is no longer on the fly during connection | |
d942ae44 | 2192 | * handshake as long as we would exceed the max_buffer limit. |
b411b363 PR |
2193 | * |
2194 | * to avoid races with the reconnect code, | |
2195 | * we need to atomic_inc within the spinlock. */ | |
2196 | ||
b30ab791 | 2197 | wait_event(device->misc_wait, inc_ap_bio_cond(device)); |
b411b363 PR |
2198 | } |
2199 | ||
b30ab791 | 2200 | static inline void dec_ap_bio(struct drbd_device *device) |
b411b363 | 2201 | { |
b30ab791 AG |
2202 | int mxb = drbd_get_max_buffers(device); |
2203 | int ap_bio = atomic_dec_return(&device->ap_bio_cnt); | |
b411b363 | 2204 | |
0b0ba1ef | 2205 | D_ASSERT(device, ap_bio >= 0); |
7ee1fb93 | 2206 | |
b30ab791 AG |
2207 | if (ap_bio == 0 && test_bit(BITMAP_IO, &device->flags)) { |
2208 | if (!test_and_set_bit(BITMAP_IO_QUEUED, &device->flags)) | |
84b8c06b AG |
2209 | drbd_queue_work(&first_peer_device(device)-> |
2210 | connection->sender_work, | |
2211 | &device->bm_io_work.w); | |
7ee1fb93 LE |
2212 | } |
2213 | ||
b411b363 PR |
2214 | /* this currently does wake_up for every dec_ap_bio! |
2215 | * maybe rather introduce some type of hysteresis? | |
2216 | * e.g. (ap_bio == mxb/2 || ap_bio == 0) ? */ | |
2217 | if (ap_bio < mxb) | |
b30ab791 | 2218 | wake_up(&device->misc_wait); |
b411b363 PR |
2219 | } |
2220 | ||
b30ab791 | 2221 | static inline bool verify_can_do_stop_sector(struct drbd_device *device) |
58ffa580 | 2222 | { |
a6b32bc3 AG |
2223 | return first_peer_device(device)->connection->agreed_pro_version >= 97 && |
2224 | first_peer_device(device)->connection->agreed_pro_version != 100; | |
58ffa580 LE |
2225 | } |
2226 | ||
b30ab791 | 2227 | static inline int drbd_set_ed_uuid(struct drbd_device *device, u64 val) |
b411b363 | 2228 | { |
b30ab791 AG |
2229 | int changed = device->ed_uuid != val; |
2230 | device->ed_uuid = val; | |
62b0da3a | 2231 | return changed; |
b411b363 PR |
2232 | } |
2233 | ||
b30ab791 | 2234 | static inline int drbd_queue_order_type(struct drbd_device *device) |
b411b363 PR |
2235 | { |
2236 | /* sorry, we currently have no working implementation | |
2237 | * of distributed TCQ stuff */ | |
2238 | #ifndef QUEUE_ORDERED_NONE | |
2239 | #define QUEUE_ORDERED_NONE 0 | |
2240 | #endif | |
2241 | return QUEUE_ORDERED_NONE; | |
2242 | } | |
2243 | ||
77c556f6 AG |
2244 | static inline struct drbd_connection *first_connection(struct drbd_resource *resource) |
2245 | { | |
ec4a3407 | 2246 | return list_first_entry_or_null(&resource->connections, |
77c556f6 AG |
2247 | struct drbd_connection, connections); |
2248 | } | |
2249 | ||
b411b363 | 2250 | #endif |