]>
Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_worker.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 | 26 | #include <linux/module.h> |
b411b363 PR |
27 | #include <linux/drbd.h> |
28 | #include <linux/sched.h> | |
b411b363 PR |
29 | #include <linux/wait.h> |
30 | #include <linux/mm.h> | |
31 | #include <linux/memcontrol.h> | |
32 | #include <linux/mm_inline.h> | |
33 | #include <linux/slab.h> | |
34 | #include <linux/random.h> | |
b411b363 PR |
35 | #include <linux/string.h> |
36 | #include <linux/scatterlist.h> | |
37 | ||
38 | #include "drbd_int.h" | |
39 | #include "drbd_req.h" | |
b411b363 | 40 | |
b411b363 PR |
41 | static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel); |
42 | ||
43 | ||
44 | ||
45 | /* defined here: | |
46 | drbd_md_io_complete | |
45bb912b | 47 | drbd_endio_sec |
b411b363 PR |
48 | drbd_endio_pri |
49 | ||
50 | * more endio handlers: | |
51 | atodb_endio in drbd_actlog.c | |
52 | drbd_bm_async_io_complete in drbd_bitmap.c | |
53 | ||
54 | * For all these callbacks, note the following: | |
55 | * The callbacks will be called in irq context by the IDE drivers, | |
56 | * and in Softirqs/Tasklets/BH context by the SCSI drivers. | |
57 | * Try to get the locking right :) | |
58 | * | |
59 | */ | |
60 | ||
61 | ||
62 | /* About the global_state_lock | |
63 | Each state transition on an device holds a read lock. In case we have | |
64 | to evaluate the sync after dependencies, we grab a write lock, because | |
65 | we need stable states on all devices for that. */ | |
66 | rwlock_t global_state_lock; | |
67 | ||
68 | /* used for synchronous meta data and bitmap IO | |
69 | * submitted by drbd_md_sync_page_io() | |
70 | */ | |
71 | void drbd_md_io_complete(struct bio *bio, int error) | |
72 | { | |
73 | struct drbd_md_io *md_io; | |
74 | ||
75 | md_io = (struct drbd_md_io *)bio->bi_private; | |
76 | md_io->error = error; | |
77 | ||
b411b363 PR |
78 | complete(&md_io->event); |
79 | } | |
80 | ||
81 | /* reads on behalf of the partner, | |
82 | * "submitted" by the receiver | |
83 | */ | |
45bb912b | 84 | void drbd_endio_read_sec_final(struct drbd_epoch_entry *e) __releases(local) |
b411b363 PR |
85 | { |
86 | unsigned long flags = 0; | |
45bb912b | 87 | struct drbd_conf *mdev = e->mdev; |
b411b363 PR |
88 | |
89 | D_ASSERT(e->block_id != ID_VACANT); | |
90 | ||
b411b363 PR |
91 | spin_lock_irqsave(&mdev->req_lock, flags); |
92 | mdev->read_cnt += e->size >> 9; | |
93 | list_del(&e->w.list); | |
94 | if (list_empty(&mdev->read_ee)) | |
95 | wake_up(&mdev->ee_wait); | |
45bb912b LE |
96 | if (test_bit(__EE_WAS_ERROR, &e->flags)) |
97 | __drbd_chk_io_error(mdev, FALSE); | |
b411b363 PR |
98 | spin_unlock_irqrestore(&mdev->req_lock, flags); |
99 | ||
b411b363 PR |
100 | drbd_queue_work(&mdev->data.work, &e->w); |
101 | put_ldev(mdev); | |
b411b363 PR |
102 | } |
103 | ||
104 | /* writes on behalf of the partner, or resync writes, | |
45bb912b LE |
105 | * "submitted" by the receiver, final stage. */ |
106 | static void drbd_endio_write_sec_final(struct drbd_epoch_entry *e) __releases(local) | |
b411b363 PR |
107 | { |
108 | unsigned long flags = 0; | |
45bb912b | 109 | struct drbd_conf *mdev = e->mdev; |
b411b363 PR |
110 | sector_t e_sector; |
111 | int do_wake; | |
112 | int is_syncer_req; | |
113 | int do_al_complete_io; | |
b411b363 | 114 | |
b411b363 PR |
115 | D_ASSERT(e->block_id != ID_VACANT); |
116 | ||
b411b363 PR |
117 | /* after we moved e to done_ee, |
118 | * we may no longer access it, | |
119 | * it may be freed/reused already! | |
120 | * (as soon as we release the req_lock) */ | |
121 | e_sector = e->sector; | |
122 | do_al_complete_io = e->flags & EE_CALL_AL_COMPLETE_IO; | |
45bb912b | 123 | is_syncer_req = is_syncer_block_id(e->block_id); |
b411b363 | 124 | |
45bb912b LE |
125 | spin_lock_irqsave(&mdev->req_lock, flags); |
126 | mdev->writ_cnt += e->size >> 9; | |
b411b363 PR |
127 | list_del(&e->w.list); /* has been on active_ee or sync_ee */ |
128 | list_add_tail(&e->w.list, &mdev->done_ee); | |
129 | ||
b411b363 PR |
130 | /* No hlist_del_init(&e->colision) here, we did not send the Ack yet, |
131 | * neither did we wake possibly waiting conflicting requests. | |
132 | * done from "drbd_process_done_ee" within the appropriate w.cb | |
133 | * (e_end_block/e_end_resync_block) or from _drbd_clear_done_ee */ | |
134 | ||
135 | do_wake = is_syncer_req | |
136 | ? list_empty(&mdev->sync_ee) | |
137 | : list_empty(&mdev->active_ee); | |
138 | ||
45bb912b | 139 | if (test_bit(__EE_WAS_ERROR, &e->flags)) |
b411b363 PR |
140 | __drbd_chk_io_error(mdev, FALSE); |
141 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
142 | ||
143 | if (is_syncer_req) | |
144 | drbd_rs_complete_io(mdev, e_sector); | |
145 | ||
146 | if (do_wake) | |
147 | wake_up(&mdev->ee_wait); | |
148 | ||
149 | if (do_al_complete_io) | |
150 | drbd_al_complete_io(mdev, e_sector); | |
151 | ||
152 | wake_asender(mdev); | |
153 | put_ldev(mdev); | |
45bb912b | 154 | } |
b411b363 | 155 | |
45bb912b LE |
156 | /* writes on behalf of the partner, or resync writes, |
157 | * "submitted" by the receiver. | |
158 | */ | |
159 | void drbd_endio_sec(struct bio *bio, int error) | |
160 | { | |
161 | struct drbd_epoch_entry *e = bio->bi_private; | |
162 | struct drbd_conf *mdev = e->mdev; | |
163 | int uptodate = bio_flagged(bio, BIO_UPTODATE); | |
164 | int is_write = bio_data_dir(bio) == WRITE; | |
165 | ||
166 | if (error) | |
167 | dev_warn(DEV, "%s: error=%d s=%llus\n", | |
168 | is_write ? "write" : "read", error, | |
169 | (unsigned long long)e->sector); | |
170 | if (!error && !uptodate) { | |
171 | dev_warn(DEV, "%s: setting error to -EIO s=%llus\n", | |
172 | is_write ? "write" : "read", | |
173 | (unsigned long long)e->sector); | |
174 | /* strange behavior of some lower level drivers... | |
175 | * fail the request by clearing the uptodate flag, | |
176 | * but do not return any error?! */ | |
177 | error = -EIO; | |
178 | } | |
179 | ||
180 | if (error) | |
181 | set_bit(__EE_WAS_ERROR, &e->flags); | |
182 | ||
183 | bio_put(bio); /* no need for the bio anymore */ | |
184 | if (atomic_dec_and_test(&e->pending_bios)) { | |
185 | if (is_write) | |
186 | drbd_endio_write_sec_final(e); | |
187 | else | |
188 | drbd_endio_read_sec_final(e); | |
189 | } | |
b411b363 PR |
190 | } |
191 | ||
192 | /* read, readA or write requests on R_PRIMARY coming from drbd_make_request | |
193 | */ | |
194 | void drbd_endio_pri(struct bio *bio, int error) | |
195 | { | |
a115413d | 196 | unsigned long flags; |
b411b363 PR |
197 | struct drbd_request *req = bio->bi_private; |
198 | struct drbd_conf *mdev = req->mdev; | |
a115413d | 199 | struct bio_and_error m; |
b411b363 PR |
200 | enum drbd_req_event what; |
201 | int uptodate = bio_flagged(bio, BIO_UPTODATE); | |
202 | ||
b411b363 PR |
203 | if (!error && !uptodate) { |
204 | dev_warn(DEV, "p %s: setting error to -EIO\n", | |
205 | bio_data_dir(bio) == WRITE ? "write" : "read"); | |
206 | /* strange behavior of some lower level drivers... | |
207 | * fail the request by clearing the uptodate flag, | |
208 | * but do not return any error?! */ | |
209 | error = -EIO; | |
210 | } | |
211 | ||
b411b363 PR |
212 | /* to avoid recursion in __req_mod */ |
213 | if (unlikely(error)) { | |
214 | what = (bio_data_dir(bio) == WRITE) | |
215 | ? write_completed_with_error | |
5c3c7e64 | 216 | : (bio_rw(bio) == READ) |
b411b363 PR |
217 | ? read_completed_with_error |
218 | : read_ahead_completed_with_error; | |
219 | } else | |
220 | what = completed_ok; | |
221 | ||
222 | bio_put(req->private_bio); | |
223 | req->private_bio = ERR_PTR(error); | |
224 | ||
a115413d LE |
225 | /* not req_mod(), we need irqsave here! */ |
226 | spin_lock_irqsave(&mdev->req_lock, flags); | |
227 | __req_mod(req, what, &m); | |
228 | spin_unlock_irqrestore(&mdev->req_lock, flags); | |
229 | ||
230 | if (m.bio) | |
231 | complete_master_bio(mdev, &m); | |
b411b363 PR |
232 | } |
233 | ||
b411b363 PR |
234 | int w_read_retry_remote(struct drbd_conf *mdev, struct drbd_work *w, int cancel) |
235 | { | |
236 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
237 | ||
238 | /* We should not detach for read io-error, | |
239 | * but try to WRITE the P_DATA_REPLY to the failed location, | |
240 | * to give the disk the chance to relocate that block */ | |
241 | ||
242 | spin_lock_irq(&mdev->req_lock); | |
d255e5ff LE |
243 | if (cancel || mdev->state.pdsk != D_UP_TO_DATE) { |
244 | _req_mod(req, read_retry_remote_canceled); | |
b411b363 | 245 | spin_unlock_irq(&mdev->req_lock); |
b411b363 PR |
246 | return 1; |
247 | } | |
248 | spin_unlock_irq(&mdev->req_lock); | |
249 | ||
250 | return w_send_read_req(mdev, w, 0); | |
251 | } | |
252 | ||
253 | int w_resync_inactive(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
254 | { | |
255 | ERR_IF(cancel) return 1; | |
256 | dev_err(DEV, "resync inactive, but callback triggered??\n"); | |
257 | return 1; /* Simply ignore this! */ | |
258 | } | |
259 | ||
45bb912b LE |
260 | void drbd_csum_ee(struct drbd_conf *mdev, struct crypto_hash *tfm, struct drbd_epoch_entry *e, void *digest) |
261 | { | |
262 | struct hash_desc desc; | |
263 | struct scatterlist sg; | |
264 | struct page *page = e->pages; | |
265 | struct page *tmp; | |
266 | unsigned len; | |
267 | ||
268 | desc.tfm = tfm; | |
269 | desc.flags = 0; | |
270 | ||
271 | sg_init_table(&sg, 1); | |
272 | crypto_hash_init(&desc); | |
273 | ||
274 | while ((tmp = page_chain_next(page))) { | |
275 | /* all but the last page will be fully used */ | |
276 | sg_set_page(&sg, page, PAGE_SIZE, 0); | |
277 | crypto_hash_update(&desc, &sg, sg.length); | |
278 | page = tmp; | |
279 | } | |
280 | /* and now the last, possibly only partially used page */ | |
281 | len = e->size & (PAGE_SIZE - 1); | |
282 | sg_set_page(&sg, page, len ?: PAGE_SIZE, 0); | |
283 | crypto_hash_update(&desc, &sg, sg.length); | |
284 | crypto_hash_final(&desc, digest); | |
285 | } | |
286 | ||
287 | void drbd_csum_bio(struct drbd_conf *mdev, struct crypto_hash *tfm, struct bio *bio, void *digest) | |
b411b363 PR |
288 | { |
289 | struct hash_desc desc; | |
290 | struct scatterlist sg; | |
291 | struct bio_vec *bvec; | |
292 | int i; | |
293 | ||
294 | desc.tfm = tfm; | |
295 | desc.flags = 0; | |
296 | ||
297 | sg_init_table(&sg, 1); | |
298 | crypto_hash_init(&desc); | |
299 | ||
300 | __bio_for_each_segment(bvec, bio, i, 0) { | |
301 | sg_set_page(&sg, bvec->bv_page, bvec->bv_len, bvec->bv_offset); | |
302 | crypto_hash_update(&desc, &sg, sg.length); | |
303 | } | |
304 | crypto_hash_final(&desc, digest); | |
305 | } | |
306 | ||
307 | static int w_e_send_csum(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
308 | { | |
309 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
310 | int digest_size; | |
311 | void *digest; | |
312 | int ok; | |
313 | ||
314 | D_ASSERT(e->block_id == DRBD_MAGIC + 0xbeef); | |
315 | ||
316 | if (unlikely(cancel)) { | |
317 | drbd_free_ee(mdev, e); | |
318 | return 1; | |
319 | } | |
320 | ||
45bb912b | 321 | if (likely((e->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
322 | digest_size = crypto_hash_digestsize(mdev->csums_tfm); |
323 | digest = kmalloc(digest_size, GFP_NOIO); | |
324 | if (digest) { | |
45bb912b | 325 | drbd_csum_ee(mdev, mdev->csums_tfm, e, digest); |
b411b363 PR |
326 | |
327 | inc_rs_pending(mdev); | |
328 | ok = drbd_send_drequest_csum(mdev, | |
329 | e->sector, | |
330 | e->size, | |
331 | digest, | |
332 | digest_size, | |
333 | P_CSUM_RS_REQUEST); | |
334 | kfree(digest); | |
335 | } else { | |
336 | dev_err(DEV, "kmalloc() of digest failed.\n"); | |
337 | ok = 0; | |
338 | } | |
339 | } else | |
340 | ok = 1; | |
341 | ||
342 | drbd_free_ee(mdev, e); | |
343 | ||
344 | if (unlikely(!ok)) | |
345 | dev_err(DEV, "drbd_send_drequest(..., csum) failed\n"); | |
346 | return ok; | |
347 | } | |
348 | ||
349 | #define GFP_TRY (__GFP_HIGHMEM | __GFP_NOWARN) | |
350 | ||
351 | static int read_for_csum(struct drbd_conf *mdev, sector_t sector, int size) | |
352 | { | |
353 | struct drbd_epoch_entry *e; | |
354 | ||
355 | if (!get_ldev(mdev)) | |
80a40e43 | 356 | return -EIO; |
b411b363 | 357 | |
0f0601f4 LE |
358 | if (drbd_rs_should_slow_down(mdev)) |
359 | goto defer; | |
360 | ||
b411b363 PR |
361 | /* GFP_TRY, because if there is no memory available right now, this may |
362 | * be rescheduled for later. It is "only" background resync, after all. */ | |
363 | e = drbd_alloc_ee(mdev, DRBD_MAGIC+0xbeef, sector, size, GFP_TRY); | |
45bb912b | 364 | if (!e) |
80a40e43 | 365 | goto defer; |
b411b363 | 366 | |
80a40e43 | 367 | e->w.cb = w_e_send_csum; |
b411b363 PR |
368 | spin_lock_irq(&mdev->req_lock); |
369 | list_add(&e->w.list, &mdev->read_ee); | |
370 | spin_unlock_irq(&mdev->req_lock); | |
371 | ||
0f0601f4 | 372 | atomic_add(size >> 9, &mdev->rs_sect_ev); |
45bb912b | 373 | if (drbd_submit_ee(mdev, e, READ, DRBD_FAULT_RS_RD) == 0) |
80a40e43 | 374 | return 0; |
b411b363 | 375 | |
22cc37a9 LE |
376 | /* drbd_submit_ee currently fails for one reason only: |
377 | * not being able to allocate enough bios. | |
378 | * Is dropping the connection going to help? */ | |
379 | spin_lock_irq(&mdev->req_lock); | |
380 | list_del(&e->w.list); | |
381 | spin_unlock_irq(&mdev->req_lock); | |
382 | ||
45bb912b | 383 | drbd_free_ee(mdev, e); |
80a40e43 | 384 | defer: |
45bb912b | 385 | put_ldev(mdev); |
80a40e43 | 386 | return -EAGAIN; |
b411b363 PR |
387 | } |
388 | ||
389 | void resync_timer_fn(unsigned long data) | |
390 | { | |
b411b363 PR |
391 | struct drbd_conf *mdev = (struct drbd_conf *) data; |
392 | int queue; | |
393 | ||
63106d3c PR |
394 | queue = 1; |
395 | switch (mdev->state.conn) { | |
396 | case C_VERIFY_S: | |
397 | mdev->resync_work.cb = w_make_ov_request; | |
398 | break; | |
399 | case C_SYNC_TARGET: | |
400 | mdev->resync_work.cb = w_make_resync_request; | |
401 | break; | |
402 | default: | |
b411b363 PR |
403 | queue = 0; |
404 | mdev->resync_work.cb = w_resync_inactive; | |
405 | } | |
406 | ||
b411b363 PR |
407 | /* harmless race: list_empty outside data.work.q_lock */ |
408 | if (list_empty(&mdev->resync_work.list) && queue) | |
409 | drbd_queue_work(&mdev->data.work, &mdev->resync_work); | |
410 | } | |
411 | ||
778f271d PR |
412 | static void fifo_set(struct fifo_buffer *fb, int value) |
413 | { | |
414 | int i; | |
415 | ||
416 | for (i = 0; i < fb->size; i++) | |
f10f2623 | 417 | fb->values[i] = value; |
778f271d PR |
418 | } |
419 | ||
420 | static int fifo_push(struct fifo_buffer *fb, int value) | |
421 | { | |
422 | int ov; | |
423 | ||
424 | ov = fb->values[fb->head_index]; | |
425 | fb->values[fb->head_index++] = value; | |
426 | ||
427 | if (fb->head_index >= fb->size) | |
428 | fb->head_index = 0; | |
429 | ||
430 | return ov; | |
431 | } | |
432 | ||
433 | static void fifo_add_val(struct fifo_buffer *fb, int value) | |
434 | { | |
435 | int i; | |
436 | ||
437 | for (i = 0; i < fb->size; i++) | |
438 | fb->values[i] += value; | |
439 | } | |
440 | ||
441 | int drbd_rs_controller(struct drbd_conf *mdev) | |
442 | { | |
443 | unsigned int sect_in; /* Number of sectors that came in since the last turn */ | |
444 | unsigned int want; /* The number of sectors we want in the proxy */ | |
445 | int req_sect; /* Number of sectors to request in this turn */ | |
446 | int correction; /* Number of sectors more we need in the proxy*/ | |
447 | int cps; /* correction per invocation of drbd_rs_controller() */ | |
448 | int steps; /* Number of time steps to plan ahead */ | |
449 | int curr_corr; | |
450 | int max_sect; | |
451 | ||
452 | sect_in = atomic_xchg(&mdev->rs_sect_in, 0); /* Number of sectors that came in */ | |
453 | mdev->rs_in_flight -= sect_in; | |
454 | ||
455 | spin_lock(&mdev->peer_seq_lock); /* get an atomic view on mdev->rs_plan_s */ | |
456 | ||
457 | steps = mdev->rs_plan_s.size; /* (mdev->sync_conf.c_plan_ahead * 10 * SLEEP_TIME) / HZ; */ | |
458 | ||
459 | if (mdev->rs_in_flight + sect_in == 0) { /* At start of resync */ | |
460 | want = ((mdev->sync_conf.rate * 2 * SLEEP_TIME) / HZ) * steps; | |
461 | } else { /* normal path */ | |
462 | want = mdev->sync_conf.c_fill_target ? mdev->sync_conf.c_fill_target : | |
463 | sect_in * mdev->sync_conf.c_delay_target * HZ / (SLEEP_TIME * 10); | |
464 | } | |
465 | ||
466 | correction = want - mdev->rs_in_flight - mdev->rs_planed; | |
467 | ||
468 | /* Plan ahead */ | |
469 | cps = correction / steps; | |
470 | fifo_add_val(&mdev->rs_plan_s, cps); | |
471 | mdev->rs_planed += cps * steps; | |
472 | ||
473 | /* What we do in this step */ | |
474 | curr_corr = fifo_push(&mdev->rs_plan_s, 0); | |
475 | spin_unlock(&mdev->peer_seq_lock); | |
476 | mdev->rs_planed -= curr_corr; | |
477 | ||
478 | req_sect = sect_in + curr_corr; | |
479 | if (req_sect < 0) | |
480 | req_sect = 0; | |
481 | ||
482 | max_sect = (mdev->sync_conf.c_max_rate * 2 * SLEEP_TIME) / HZ; | |
483 | if (req_sect > max_sect) | |
484 | req_sect = max_sect; | |
485 | ||
486 | /* | |
487 | dev_warn(DEV, "si=%u if=%d wa=%u co=%d st=%d cps=%d pl=%d cc=%d rs=%d\n", | |
488 | sect_in, mdev->rs_in_flight, want, correction, | |
489 | steps, cps, mdev->rs_planed, curr_corr, req_sect); | |
490 | */ | |
491 | ||
492 | return req_sect; | |
493 | } | |
494 | ||
b411b363 PR |
495 | int w_make_resync_request(struct drbd_conf *mdev, |
496 | struct drbd_work *w, int cancel) | |
497 | { | |
498 | unsigned long bit; | |
499 | sector_t sector; | |
500 | const sector_t capacity = drbd_get_capacity(mdev->this_bdev); | |
bb3d000c | 501 | int max_segment_size; |
0f0601f4 | 502 | int number, rollback_i, size, pe, mx; |
b411b363 | 503 | int align, queued, sndbuf; |
0f0601f4 | 504 | int i = 0; |
b411b363 PR |
505 | |
506 | if (unlikely(cancel)) | |
507 | return 1; | |
508 | ||
509 | if (unlikely(mdev->state.conn < C_CONNECTED)) { | |
510 | dev_err(DEV, "Confused in w_make_resync_request()! cstate < Connected"); | |
511 | return 0; | |
512 | } | |
513 | ||
514 | if (mdev->state.conn != C_SYNC_TARGET) | |
515 | dev_err(DEV, "%s in w_make_resync_request\n", | |
516 | drbd_conn_str(mdev->state.conn)); | |
517 | ||
af85e8e8 LE |
518 | if (mdev->rs_total == 0) { |
519 | /* empty resync? */ | |
520 | drbd_resync_finished(mdev); | |
521 | return 1; | |
522 | } | |
523 | ||
b411b363 PR |
524 | if (!get_ldev(mdev)) { |
525 | /* Since we only need to access mdev->rsync a | |
526 | get_ldev_if_state(mdev,D_FAILED) would be sufficient, but | |
527 | to continue resync with a broken disk makes no sense at | |
528 | all */ | |
529 | dev_err(DEV, "Disk broke down during resync!\n"); | |
530 | mdev->resync_work.cb = w_resync_inactive; | |
531 | return 1; | |
532 | } | |
533 | ||
bb3d000c LE |
534 | /* starting with drbd 8.3.8, we can handle multi-bio EEs, |
535 | * if it should be necessary */ | |
5a75cc7c PR |
536 | max_segment_size = |
537 | mdev->agreed_pro_version < 94 ? queue_max_segment_size(mdev->rq_queue) : | |
538 | mdev->agreed_pro_version < 95 ? DRBD_MAX_SIZE_H80_PACKET : DRBD_MAX_SEGMENT_SIZE; | |
bb3d000c | 539 | |
778f271d PR |
540 | if (mdev->rs_plan_s.size) { /* mdev->sync_conf.c_plan_ahead */ |
541 | number = drbd_rs_controller(mdev) >> (BM_BLOCK_SHIFT - 9); | |
542 | mdev->c_sync_rate = number * HZ * (BM_BLOCK_SIZE / 1024) / SLEEP_TIME; | |
543 | } else { | |
544 | mdev->c_sync_rate = mdev->sync_conf.rate; | |
545 | number = SLEEP_TIME * mdev->c_sync_rate / ((BM_BLOCK_SIZE / 1024) * HZ); | |
546 | } | |
0f0601f4 LE |
547 | |
548 | /* Throttle resync on lower level disk activity, which may also be | |
549 | * caused by application IO on Primary/SyncTarget. | |
550 | * Keep this after the call to drbd_rs_controller, as that assumes | |
551 | * to be called as precisely as possible every SLEEP_TIME, | |
552 | * and would be confused otherwise. */ | |
553 | if (drbd_rs_should_slow_down(mdev)) | |
554 | goto requeue; | |
b411b363 PR |
555 | |
556 | mutex_lock(&mdev->data.mutex); | |
557 | if (mdev->data.socket) | |
558 | mx = mdev->data.socket->sk->sk_rcvbuf / sizeof(struct p_block_req); | |
559 | else | |
560 | mx = 1; | |
561 | mutex_unlock(&mdev->data.mutex); | |
562 | ||
563 | /* For resync rates >160MB/sec, allow more pending RS requests */ | |
564 | if (number > mx) | |
565 | mx = number; | |
566 | ||
567 | /* Limit the number of pending RS requests to no more than the peer's receive buffer */ | |
0f0601f4 | 568 | pe = atomic_read(&mdev->rs_pending_cnt); |
b411b363 PR |
569 | if ((pe + number) > mx) { |
570 | number = mx - pe; | |
571 | } | |
572 | ||
573 | for (i = 0; i < number; i++) { | |
574 | /* Stop generating RS requests, when half of the send buffer is filled */ | |
575 | mutex_lock(&mdev->data.mutex); | |
576 | if (mdev->data.socket) { | |
577 | queued = mdev->data.socket->sk->sk_wmem_queued; | |
578 | sndbuf = mdev->data.socket->sk->sk_sndbuf; | |
579 | } else { | |
580 | queued = 1; | |
581 | sndbuf = 0; | |
582 | } | |
583 | mutex_unlock(&mdev->data.mutex); | |
584 | if (queued > sndbuf / 2) | |
585 | goto requeue; | |
586 | ||
587 | next_sector: | |
588 | size = BM_BLOCK_SIZE; | |
589 | bit = drbd_bm_find_next(mdev, mdev->bm_resync_fo); | |
590 | ||
591 | if (bit == -1UL) { | |
592 | mdev->bm_resync_fo = drbd_bm_bits(mdev); | |
593 | mdev->resync_work.cb = w_resync_inactive; | |
594 | put_ldev(mdev); | |
595 | return 1; | |
596 | } | |
597 | ||
598 | sector = BM_BIT_TO_SECT(bit); | |
599 | ||
600 | if (drbd_try_rs_begin_io(mdev, sector)) { | |
601 | mdev->bm_resync_fo = bit; | |
602 | goto requeue; | |
603 | } | |
604 | mdev->bm_resync_fo = bit + 1; | |
605 | ||
606 | if (unlikely(drbd_bm_test_bit(mdev, bit) == 0)) { | |
607 | drbd_rs_complete_io(mdev, sector); | |
608 | goto next_sector; | |
609 | } | |
610 | ||
611 | #if DRBD_MAX_SEGMENT_SIZE > BM_BLOCK_SIZE | |
612 | /* try to find some adjacent bits. | |
613 | * we stop if we have already the maximum req size. | |
614 | * | |
615 | * Additionally always align bigger requests, in order to | |
616 | * be prepared for all stripe sizes of software RAIDs. | |
b411b363 PR |
617 | */ |
618 | align = 1; | |
d207450c | 619 | rollback_i = i; |
b411b363 PR |
620 | for (;;) { |
621 | if (size + BM_BLOCK_SIZE > max_segment_size) | |
622 | break; | |
623 | ||
624 | /* Be always aligned */ | |
625 | if (sector & ((1<<(align+3))-1)) | |
626 | break; | |
627 | ||
628 | /* do not cross extent boundaries */ | |
629 | if (((bit+1) & BM_BLOCKS_PER_BM_EXT_MASK) == 0) | |
630 | break; | |
631 | /* now, is it actually dirty, after all? | |
632 | * caution, drbd_bm_test_bit is tri-state for some | |
633 | * obscure reason; ( b == 0 ) would get the out-of-band | |
634 | * only accidentally right because of the "oddly sized" | |
635 | * adjustment below */ | |
636 | if (drbd_bm_test_bit(mdev, bit+1) != 1) | |
637 | break; | |
638 | bit++; | |
639 | size += BM_BLOCK_SIZE; | |
640 | if ((BM_BLOCK_SIZE << align) <= size) | |
641 | align++; | |
642 | i++; | |
643 | } | |
644 | /* if we merged some, | |
645 | * reset the offset to start the next drbd_bm_find_next from */ | |
646 | if (size > BM_BLOCK_SIZE) | |
647 | mdev->bm_resync_fo = bit + 1; | |
648 | #endif | |
649 | ||
650 | /* adjust very last sectors, in case we are oddly sized */ | |
651 | if (sector + (size>>9) > capacity) | |
652 | size = (capacity-sector)<<9; | |
653 | if (mdev->agreed_pro_version >= 89 && mdev->csums_tfm) { | |
654 | switch (read_for_csum(mdev, sector, size)) { | |
80a40e43 | 655 | case -EIO: /* Disk failure */ |
b411b363 PR |
656 | put_ldev(mdev); |
657 | return 0; | |
80a40e43 | 658 | case -EAGAIN: /* allocation failed, or ldev busy */ |
b411b363 PR |
659 | drbd_rs_complete_io(mdev, sector); |
660 | mdev->bm_resync_fo = BM_SECT_TO_BIT(sector); | |
d207450c | 661 | i = rollback_i; |
b411b363 | 662 | goto requeue; |
80a40e43 LE |
663 | case 0: |
664 | /* everything ok */ | |
665 | break; | |
666 | default: | |
667 | BUG(); | |
b411b363 PR |
668 | } |
669 | } else { | |
670 | inc_rs_pending(mdev); | |
671 | if (!drbd_send_drequest(mdev, P_RS_DATA_REQUEST, | |
672 | sector, size, ID_SYNCER)) { | |
673 | dev_err(DEV, "drbd_send_drequest() failed, aborting...\n"); | |
674 | dec_rs_pending(mdev); | |
675 | put_ldev(mdev); | |
676 | return 0; | |
677 | } | |
678 | } | |
679 | } | |
680 | ||
681 | if (mdev->bm_resync_fo >= drbd_bm_bits(mdev)) { | |
682 | /* last syncer _request_ was sent, | |
683 | * but the P_RS_DATA_REPLY not yet received. sync will end (and | |
684 | * next sync group will resume), as soon as we receive the last | |
685 | * resync data block, and the last bit is cleared. | |
686 | * until then resync "work" is "inactive" ... | |
687 | */ | |
688 | mdev->resync_work.cb = w_resync_inactive; | |
689 | put_ldev(mdev); | |
690 | return 1; | |
691 | } | |
692 | ||
693 | requeue: | |
778f271d | 694 | mdev->rs_in_flight += (i << (BM_BLOCK_SHIFT - 9)); |
b411b363 PR |
695 | mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME); |
696 | put_ldev(mdev); | |
697 | return 1; | |
698 | } | |
699 | ||
700 | static int w_make_ov_request(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
701 | { | |
702 | int number, i, size; | |
703 | sector_t sector; | |
704 | const sector_t capacity = drbd_get_capacity(mdev->this_bdev); | |
705 | ||
706 | if (unlikely(cancel)) | |
707 | return 1; | |
708 | ||
709 | if (unlikely(mdev->state.conn < C_CONNECTED)) { | |
710 | dev_err(DEV, "Confused in w_make_ov_request()! cstate < Connected"); | |
711 | return 0; | |
712 | } | |
713 | ||
714 | number = SLEEP_TIME*mdev->sync_conf.rate / ((BM_BLOCK_SIZE/1024)*HZ); | |
715 | if (atomic_read(&mdev->rs_pending_cnt) > number) | |
716 | goto requeue; | |
717 | ||
718 | number -= atomic_read(&mdev->rs_pending_cnt); | |
719 | ||
720 | sector = mdev->ov_position; | |
721 | for (i = 0; i < number; i++) { | |
722 | if (sector >= capacity) { | |
723 | mdev->resync_work.cb = w_resync_inactive; | |
724 | return 1; | |
725 | } | |
726 | ||
727 | size = BM_BLOCK_SIZE; | |
728 | ||
729 | if (drbd_try_rs_begin_io(mdev, sector)) { | |
730 | mdev->ov_position = sector; | |
731 | goto requeue; | |
732 | } | |
733 | ||
734 | if (sector + (size>>9) > capacity) | |
735 | size = (capacity-sector)<<9; | |
736 | ||
737 | inc_rs_pending(mdev); | |
738 | if (!drbd_send_ov_request(mdev, sector, size)) { | |
739 | dec_rs_pending(mdev); | |
740 | return 0; | |
741 | } | |
742 | sector += BM_SECT_PER_BIT; | |
743 | } | |
744 | mdev->ov_position = sector; | |
745 | ||
746 | requeue: | |
747 | mod_timer(&mdev->resync_timer, jiffies + SLEEP_TIME); | |
748 | return 1; | |
749 | } | |
750 | ||
751 | ||
752 | int w_ov_finished(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
753 | { | |
754 | kfree(w); | |
755 | ov_oos_print(mdev); | |
756 | drbd_resync_finished(mdev); | |
757 | ||
758 | return 1; | |
759 | } | |
760 | ||
761 | static int w_resync_finished(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
762 | { | |
763 | kfree(w); | |
764 | ||
765 | drbd_resync_finished(mdev); | |
766 | ||
767 | return 1; | |
768 | } | |
769 | ||
af85e8e8 LE |
770 | static void ping_peer(struct drbd_conf *mdev) |
771 | { | |
772 | clear_bit(GOT_PING_ACK, &mdev->flags); | |
773 | request_ping(mdev); | |
774 | wait_event(mdev->misc_wait, | |
775 | test_bit(GOT_PING_ACK, &mdev->flags) || mdev->state.conn < C_CONNECTED); | |
776 | } | |
777 | ||
b411b363 PR |
778 | int drbd_resync_finished(struct drbd_conf *mdev) |
779 | { | |
780 | unsigned long db, dt, dbdt; | |
781 | unsigned long n_oos; | |
782 | union drbd_state os, ns; | |
783 | struct drbd_work *w; | |
784 | char *khelper_cmd = NULL; | |
785 | ||
786 | /* Remove all elements from the resync LRU. Since future actions | |
787 | * might set bits in the (main) bitmap, then the entries in the | |
788 | * resync LRU would be wrong. */ | |
789 | if (drbd_rs_del_all(mdev)) { | |
790 | /* In case this is not possible now, most probably because | |
791 | * there are P_RS_DATA_REPLY Packets lingering on the worker's | |
792 | * queue (or even the read operations for those packets | |
793 | * is not finished by now). Retry in 100ms. */ | |
794 | ||
b411b363 PR |
795 | __set_current_state(TASK_INTERRUPTIBLE); |
796 | schedule_timeout(HZ / 10); | |
797 | w = kmalloc(sizeof(struct drbd_work), GFP_ATOMIC); | |
798 | if (w) { | |
799 | w->cb = w_resync_finished; | |
800 | drbd_queue_work(&mdev->data.work, w); | |
801 | return 1; | |
802 | } | |
803 | dev_err(DEV, "Warn failed to drbd_rs_del_all() and to kmalloc(w).\n"); | |
804 | } | |
805 | ||
806 | dt = (jiffies - mdev->rs_start - mdev->rs_paused) / HZ; | |
807 | if (dt <= 0) | |
808 | dt = 1; | |
809 | db = mdev->rs_total; | |
810 | dbdt = Bit2KB(db/dt); | |
811 | mdev->rs_paused /= HZ; | |
812 | ||
813 | if (!get_ldev(mdev)) | |
814 | goto out; | |
815 | ||
af85e8e8 LE |
816 | ping_peer(mdev); |
817 | ||
b411b363 PR |
818 | spin_lock_irq(&mdev->req_lock); |
819 | os = mdev->state; | |
820 | ||
821 | /* This protects us against multiple calls (that can happen in the presence | |
822 | of application IO), and against connectivity loss just before we arrive here. */ | |
823 | if (os.conn <= C_CONNECTED) | |
824 | goto out_unlock; | |
825 | ||
826 | ns = os; | |
827 | ns.conn = C_CONNECTED; | |
828 | ||
829 | dev_info(DEV, "%s done (total %lu sec; paused %lu sec; %lu K/sec)\n", | |
830 | (os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) ? | |
831 | "Online verify " : "Resync", | |
832 | dt + mdev->rs_paused, mdev->rs_paused, dbdt); | |
833 | ||
834 | n_oos = drbd_bm_total_weight(mdev); | |
835 | ||
836 | if (os.conn == C_VERIFY_S || os.conn == C_VERIFY_T) { | |
837 | if (n_oos) { | |
838 | dev_alert(DEV, "Online verify found %lu %dk block out of sync!\n", | |
839 | n_oos, Bit2KB(1)); | |
840 | khelper_cmd = "out-of-sync"; | |
841 | } | |
842 | } else { | |
843 | D_ASSERT((n_oos - mdev->rs_failed) == 0); | |
844 | ||
845 | if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T) | |
846 | khelper_cmd = "after-resync-target"; | |
847 | ||
848 | if (mdev->csums_tfm && mdev->rs_total) { | |
849 | const unsigned long s = mdev->rs_same_csum; | |
850 | const unsigned long t = mdev->rs_total; | |
851 | const int ratio = | |
852 | (t == 0) ? 0 : | |
853 | (t < 100000) ? ((s*100)/t) : (s/(t/100)); | |
854 | dev_info(DEV, "%u %% had equal check sums, eliminated: %luK; " | |
855 | "transferred %luK total %luK\n", | |
856 | ratio, | |
857 | Bit2KB(mdev->rs_same_csum), | |
858 | Bit2KB(mdev->rs_total - mdev->rs_same_csum), | |
859 | Bit2KB(mdev->rs_total)); | |
860 | } | |
861 | } | |
862 | ||
863 | if (mdev->rs_failed) { | |
864 | dev_info(DEV, " %lu failed blocks\n", mdev->rs_failed); | |
865 | ||
866 | if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T) { | |
867 | ns.disk = D_INCONSISTENT; | |
868 | ns.pdsk = D_UP_TO_DATE; | |
869 | } else { | |
870 | ns.disk = D_UP_TO_DATE; | |
871 | ns.pdsk = D_INCONSISTENT; | |
872 | } | |
873 | } else { | |
874 | ns.disk = D_UP_TO_DATE; | |
875 | ns.pdsk = D_UP_TO_DATE; | |
876 | ||
877 | if (os.conn == C_SYNC_TARGET || os.conn == C_PAUSED_SYNC_T) { | |
878 | if (mdev->p_uuid) { | |
879 | int i; | |
880 | for (i = UI_BITMAP ; i <= UI_HISTORY_END ; i++) | |
881 | _drbd_uuid_set(mdev, i, mdev->p_uuid[i]); | |
882 | drbd_uuid_set(mdev, UI_BITMAP, mdev->ldev->md.uuid[UI_CURRENT]); | |
883 | _drbd_uuid_set(mdev, UI_CURRENT, mdev->p_uuid[UI_CURRENT]); | |
884 | } else { | |
885 | dev_err(DEV, "mdev->p_uuid is NULL! BUG\n"); | |
886 | } | |
887 | } | |
888 | ||
889 | drbd_uuid_set_bm(mdev, 0UL); | |
890 | ||
891 | if (mdev->p_uuid) { | |
892 | /* Now the two UUID sets are equal, update what we | |
893 | * know of the peer. */ | |
894 | int i; | |
895 | for (i = UI_CURRENT ; i <= UI_HISTORY_END ; i++) | |
896 | mdev->p_uuid[i] = mdev->ldev->md.uuid[i]; | |
897 | } | |
898 | } | |
899 | ||
900 | _drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
901 | out_unlock: | |
902 | spin_unlock_irq(&mdev->req_lock); | |
903 | put_ldev(mdev); | |
904 | out: | |
905 | mdev->rs_total = 0; | |
906 | mdev->rs_failed = 0; | |
907 | mdev->rs_paused = 0; | |
908 | mdev->ov_start_sector = 0; | |
909 | ||
13d42685 LE |
910 | drbd_md_sync(mdev); |
911 | ||
b411b363 | 912 | if (test_and_clear_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags)) { |
bc571b8c | 913 | dev_info(DEV, "Writing the whole bitmap\n"); |
b411b363 PR |
914 | drbd_queue_bitmap_io(mdev, &drbd_bm_write, NULL, "write from resync_finished"); |
915 | } | |
916 | ||
917 | if (khelper_cmd) | |
918 | drbd_khelper(mdev, khelper_cmd); | |
919 | ||
920 | return 1; | |
921 | } | |
922 | ||
923 | /* helper */ | |
924 | static void move_to_net_ee_or_free(struct drbd_conf *mdev, struct drbd_epoch_entry *e) | |
925 | { | |
45bb912b | 926 | if (drbd_ee_has_active_page(e)) { |
b411b363 | 927 | /* This might happen if sendpage() has not finished */ |
78db8928 | 928 | int i = (e->size + PAGE_SIZE -1) >> PAGE_SHIFT; |
435f0740 LE |
929 | atomic_add(i, &mdev->pp_in_use_by_net); |
930 | atomic_sub(i, &mdev->pp_in_use); | |
b411b363 PR |
931 | spin_lock_irq(&mdev->req_lock); |
932 | list_add_tail(&e->w.list, &mdev->net_ee); | |
933 | spin_unlock_irq(&mdev->req_lock); | |
435f0740 | 934 | wake_up(&drbd_pp_wait); |
b411b363 PR |
935 | } else |
936 | drbd_free_ee(mdev, e); | |
937 | } | |
938 | ||
939 | /** | |
940 | * w_e_end_data_req() - Worker callback, to send a P_DATA_REPLY packet in response to a P_DATA_REQUEST | |
941 | * @mdev: DRBD device. | |
942 | * @w: work object. | |
943 | * @cancel: The connection will be closed anyways | |
944 | */ | |
945 | int w_e_end_data_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
946 | { | |
947 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
948 | int ok; | |
949 | ||
950 | if (unlikely(cancel)) { | |
951 | drbd_free_ee(mdev, e); | |
952 | dec_unacked(mdev); | |
953 | return 1; | |
954 | } | |
955 | ||
45bb912b | 956 | if (likely((e->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
957 | ok = drbd_send_block(mdev, P_DATA_REPLY, e); |
958 | } else { | |
959 | if (__ratelimit(&drbd_ratelimit_state)) | |
960 | dev_err(DEV, "Sending NegDReply. sector=%llus.\n", | |
961 | (unsigned long long)e->sector); | |
962 | ||
963 | ok = drbd_send_ack(mdev, P_NEG_DREPLY, e); | |
964 | } | |
965 | ||
966 | dec_unacked(mdev); | |
967 | ||
968 | move_to_net_ee_or_free(mdev, e); | |
969 | ||
970 | if (unlikely(!ok)) | |
971 | dev_err(DEV, "drbd_send_block() failed\n"); | |
972 | return ok; | |
973 | } | |
974 | ||
975 | /** | |
976 | * w_e_end_rsdata_req() - Worker callback to send a P_RS_DATA_REPLY packet in response to a P_RS_DATA_REQUESTRS | |
977 | * @mdev: DRBD device. | |
978 | * @w: work object. | |
979 | * @cancel: The connection will be closed anyways | |
980 | */ | |
981 | int w_e_end_rsdata_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
982 | { | |
983 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
984 | int ok; | |
985 | ||
986 | if (unlikely(cancel)) { | |
987 | drbd_free_ee(mdev, e); | |
988 | dec_unacked(mdev); | |
989 | return 1; | |
990 | } | |
991 | ||
992 | if (get_ldev_if_state(mdev, D_FAILED)) { | |
993 | drbd_rs_complete_io(mdev, e->sector); | |
994 | put_ldev(mdev); | |
995 | } | |
996 | ||
45bb912b | 997 | if (likely((e->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
998 | if (likely(mdev->state.pdsk >= D_INCONSISTENT)) { |
999 | inc_rs_pending(mdev); | |
1000 | ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e); | |
1001 | } else { | |
1002 | if (__ratelimit(&drbd_ratelimit_state)) | |
1003 | dev_err(DEV, "Not sending RSDataReply, " | |
1004 | "partner DISKLESS!\n"); | |
1005 | ok = 1; | |
1006 | } | |
1007 | } else { | |
1008 | if (__ratelimit(&drbd_ratelimit_state)) | |
1009 | dev_err(DEV, "Sending NegRSDReply. sector %llus.\n", | |
1010 | (unsigned long long)e->sector); | |
1011 | ||
1012 | ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e); | |
1013 | ||
1014 | /* update resync data with failure */ | |
1015 | drbd_rs_failed_io(mdev, e->sector, e->size); | |
1016 | } | |
1017 | ||
1018 | dec_unacked(mdev); | |
1019 | ||
1020 | move_to_net_ee_or_free(mdev, e); | |
1021 | ||
1022 | if (unlikely(!ok)) | |
1023 | dev_err(DEV, "drbd_send_block() failed\n"); | |
1024 | return ok; | |
1025 | } | |
1026 | ||
1027 | int w_e_end_csum_rs_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1028 | { | |
1029 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
1030 | struct digest_info *di; | |
1031 | int digest_size; | |
1032 | void *digest = NULL; | |
1033 | int ok, eq = 0; | |
1034 | ||
1035 | if (unlikely(cancel)) { | |
1036 | drbd_free_ee(mdev, e); | |
1037 | dec_unacked(mdev); | |
1038 | return 1; | |
1039 | } | |
1040 | ||
1d53f09e LE |
1041 | if (get_ldev(mdev)) { |
1042 | drbd_rs_complete_io(mdev, e->sector); | |
1043 | put_ldev(mdev); | |
1044 | } | |
b411b363 | 1045 | |
85719573 | 1046 | di = e->digest; |
b411b363 | 1047 | |
45bb912b | 1048 | if (likely((e->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
1049 | /* quick hack to try to avoid a race against reconfiguration. |
1050 | * a real fix would be much more involved, | |
1051 | * introducing more locking mechanisms */ | |
1052 | if (mdev->csums_tfm) { | |
1053 | digest_size = crypto_hash_digestsize(mdev->csums_tfm); | |
1054 | D_ASSERT(digest_size == di->digest_size); | |
1055 | digest = kmalloc(digest_size, GFP_NOIO); | |
1056 | } | |
1057 | if (digest) { | |
45bb912b | 1058 | drbd_csum_ee(mdev, mdev->csums_tfm, e, digest); |
b411b363 PR |
1059 | eq = !memcmp(digest, di->digest, digest_size); |
1060 | kfree(digest); | |
1061 | } | |
1062 | ||
1063 | if (eq) { | |
1064 | drbd_set_in_sync(mdev, e->sector, e->size); | |
676396d5 LE |
1065 | /* rs_same_csums unit is BM_BLOCK_SIZE */ |
1066 | mdev->rs_same_csum += e->size >> BM_BLOCK_SHIFT; | |
b411b363 PR |
1067 | ok = drbd_send_ack(mdev, P_RS_IS_IN_SYNC, e); |
1068 | } else { | |
1069 | inc_rs_pending(mdev); | |
204bba99 PR |
1070 | e->block_id = ID_SYNCER; /* By setting block_id, digest pointer becomes invalid! */ |
1071 | e->flags &= ~EE_HAS_DIGEST; /* This e no longer has a digest pointer */ | |
1072 | kfree(di); | |
b411b363 PR |
1073 | ok = drbd_send_block(mdev, P_RS_DATA_REPLY, e); |
1074 | } | |
1075 | } else { | |
1076 | ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e); | |
1077 | if (__ratelimit(&drbd_ratelimit_state)) | |
1078 | dev_err(DEV, "Sending NegDReply. I guess it gets messy.\n"); | |
1079 | } | |
1080 | ||
1081 | dec_unacked(mdev); | |
b411b363 PR |
1082 | move_to_net_ee_or_free(mdev, e); |
1083 | ||
1084 | if (unlikely(!ok)) | |
1085 | dev_err(DEV, "drbd_send_block/ack() failed\n"); | |
1086 | return ok; | |
1087 | } | |
1088 | ||
1089 | int w_e_end_ov_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1090 | { | |
1091 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
1092 | int digest_size; | |
1093 | void *digest; | |
1094 | int ok = 1; | |
1095 | ||
1096 | if (unlikely(cancel)) | |
1097 | goto out; | |
1098 | ||
45bb912b | 1099 | if (unlikely((e->flags & EE_WAS_ERROR) != 0)) |
b411b363 PR |
1100 | goto out; |
1101 | ||
1102 | digest_size = crypto_hash_digestsize(mdev->verify_tfm); | |
1103 | /* FIXME if this allocation fails, online verify will not terminate! */ | |
1104 | digest = kmalloc(digest_size, GFP_NOIO); | |
1105 | if (digest) { | |
45bb912b | 1106 | drbd_csum_ee(mdev, mdev->verify_tfm, e, digest); |
b411b363 PR |
1107 | inc_rs_pending(mdev); |
1108 | ok = drbd_send_drequest_csum(mdev, e->sector, e->size, | |
1109 | digest, digest_size, P_OV_REPLY); | |
1110 | if (!ok) | |
1111 | dec_rs_pending(mdev); | |
1112 | kfree(digest); | |
1113 | } | |
1114 | ||
1115 | out: | |
1116 | drbd_free_ee(mdev, e); | |
1117 | ||
1118 | dec_unacked(mdev); | |
1119 | ||
1120 | return ok; | |
1121 | } | |
1122 | ||
1123 | void drbd_ov_oos_found(struct drbd_conf *mdev, sector_t sector, int size) | |
1124 | { | |
1125 | if (mdev->ov_last_oos_start + mdev->ov_last_oos_size == sector) { | |
1126 | mdev->ov_last_oos_size += size>>9; | |
1127 | } else { | |
1128 | mdev->ov_last_oos_start = sector; | |
1129 | mdev->ov_last_oos_size = size>>9; | |
1130 | } | |
1131 | drbd_set_out_of_sync(mdev, sector, size); | |
1132 | set_bit(WRITE_BM_AFTER_RESYNC, &mdev->flags); | |
1133 | } | |
1134 | ||
1135 | int w_e_end_ov_reply(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1136 | { | |
1137 | struct drbd_epoch_entry *e = container_of(w, struct drbd_epoch_entry, w); | |
1138 | struct digest_info *di; | |
1139 | int digest_size; | |
1140 | void *digest; | |
1141 | int ok, eq = 0; | |
1142 | ||
1143 | if (unlikely(cancel)) { | |
1144 | drbd_free_ee(mdev, e); | |
1145 | dec_unacked(mdev); | |
1146 | return 1; | |
1147 | } | |
1148 | ||
1149 | /* after "cancel", because after drbd_disconnect/drbd_rs_cancel_all | |
1150 | * the resync lru has been cleaned up already */ | |
1d53f09e LE |
1151 | if (get_ldev(mdev)) { |
1152 | drbd_rs_complete_io(mdev, e->sector); | |
1153 | put_ldev(mdev); | |
1154 | } | |
b411b363 | 1155 | |
85719573 | 1156 | di = e->digest; |
b411b363 | 1157 | |
45bb912b | 1158 | if (likely((e->flags & EE_WAS_ERROR) == 0)) { |
b411b363 PR |
1159 | digest_size = crypto_hash_digestsize(mdev->verify_tfm); |
1160 | digest = kmalloc(digest_size, GFP_NOIO); | |
1161 | if (digest) { | |
45bb912b | 1162 | drbd_csum_ee(mdev, mdev->verify_tfm, e, digest); |
b411b363 PR |
1163 | |
1164 | D_ASSERT(digest_size == di->digest_size); | |
1165 | eq = !memcmp(digest, di->digest, digest_size); | |
1166 | kfree(digest); | |
1167 | } | |
1168 | } else { | |
1169 | ok = drbd_send_ack(mdev, P_NEG_RS_DREPLY, e); | |
1170 | if (__ratelimit(&drbd_ratelimit_state)) | |
1171 | dev_err(DEV, "Sending NegDReply. I guess it gets messy.\n"); | |
1172 | } | |
1173 | ||
1174 | dec_unacked(mdev); | |
b411b363 PR |
1175 | if (!eq) |
1176 | drbd_ov_oos_found(mdev, e->sector, e->size); | |
1177 | else | |
1178 | ov_oos_print(mdev); | |
1179 | ||
1180 | ok = drbd_send_ack_ex(mdev, P_OV_RESULT, e->sector, e->size, | |
1181 | eq ? ID_IN_SYNC : ID_OUT_OF_SYNC); | |
1182 | ||
1183 | drbd_free_ee(mdev, e); | |
1184 | ||
1185 | if (--mdev->ov_left == 0) { | |
1186 | ov_oos_print(mdev); | |
1187 | drbd_resync_finished(mdev); | |
1188 | } | |
1189 | ||
1190 | return ok; | |
1191 | } | |
1192 | ||
1193 | int w_prev_work_done(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1194 | { | |
1195 | struct drbd_wq_barrier *b = container_of(w, struct drbd_wq_barrier, w); | |
1196 | complete(&b->done); | |
1197 | return 1; | |
1198 | } | |
1199 | ||
1200 | int w_send_barrier(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1201 | { | |
1202 | struct drbd_tl_epoch *b = container_of(w, struct drbd_tl_epoch, w); | |
1203 | struct p_barrier *p = &mdev->data.sbuf.barrier; | |
1204 | int ok = 1; | |
1205 | ||
1206 | /* really avoid racing with tl_clear. w.cb may have been referenced | |
1207 | * just before it was reassigned and re-queued, so double check that. | |
1208 | * actually, this race was harmless, since we only try to send the | |
1209 | * barrier packet here, and otherwise do nothing with the object. | |
1210 | * but compare with the head of w_clear_epoch */ | |
1211 | spin_lock_irq(&mdev->req_lock); | |
1212 | if (w->cb != w_send_barrier || mdev->state.conn < C_CONNECTED) | |
1213 | cancel = 1; | |
1214 | spin_unlock_irq(&mdev->req_lock); | |
1215 | if (cancel) | |
1216 | return 1; | |
1217 | ||
1218 | if (!drbd_get_data_sock(mdev)) | |
1219 | return 0; | |
1220 | p->barrier = b->br_number; | |
1221 | /* inc_ap_pending was done where this was queued. | |
1222 | * dec_ap_pending will be done in got_BarrierAck | |
1223 | * or (on connection loss) in w_clear_epoch. */ | |
1224 | ok = _drbd_send_cmd(mdev, mdev->data.socket, P_BARRIER, | |
0b70a13d | 1225 | (struct p_header80 *)p, sizeof(*p), 0); |
b411b363 PR |
1226 | drbd_put_data_sock(mdev); |
1227 | ||
1228 | return ok; | |
1229 | } | |
1230 | ||
1231 | int w_send_write_hint(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1232 | { | |
1233 | if (cancel) | |
1234 | return 1; | |
1235 | return drbd_send_short_cmd(mdev, P_UNPLUG_REMOTE); | |
1236 | } | |
1237 | ||
1238 | /** | |
1239 | * w_send_dblock() - Worker callback to send a P_DATA packet in order to mirror a write request | |
1240 | * @mdev: DRBD device. | |
1241 | * @w: work object. | |
1242 | * @cancel: The connection will be closed anyways | |
1243 | */ | |
1244 | int w_send_dblock(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1245 | { | |
1246 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
1247 | int ok; | |
1248 | ||
1249 | if (unlikely(cancel)) { | |
1250 | req_mod(req, send_canceled); | |
1251 | return 1; | |
1252 | } | |
1253 | ||
1254 | ok = drbd_send_dblock(mdev, req); | |
1255 | req_mod(req, ok ? handed_over_to_network : send_failed); | |
1256 | ||
1257 | return ok; | |
1258 | } | |
1259 | ||
1260 | /** | |
1261 | * w_send_read_req() - Worker callback to send a read request (P_DATA_REQUEST) packet | |
1262 | * @mdev: DRBD device. | |
1263 | * @w: work object. | |
1264 | * @cancel: The connection will be closed anyways | |
1265 | */ | |
1266 | int w_send_read_req(struct drbd_conf *mdev, struct drbd_work *w, int cancel) | |
1267 | { | |
1268 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
1269 | int ok; | |
1270 | ||
1271 | if (unlikely(cancel)) { | |
1272 | req_mod(req, send_canceled); | |
1273 | return 1; | |
1274 | } | |
1275 | ||
1276 | ok = drbd_send_drequest(mdev, P_DATA_REQUEST, req->sector, req->size, | |
1277 | (unsigned long)req); | |
1278 | ||
1279 | if (!ok) { | |
1280 | /* ?? we set C_TIMEOUT or C_BROKEN_PIPE in drbd_send(); | |
1281 | * so this is probably redundant */ | |
1282 | if (mdev->state.conn >= C_CONNECTED) | |
1283 | drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE)); | |
1284 | } | |
1285 | req_mod(req, ok ? handed_over_to_network : send_failed); | |
1286 | ||
1287 | return ok; | |
1288 | } | |
1289 | ||
265be2d0 PR |
1290 | int w_restart_disk_io(struct drbd_conf *mdev, struct drbd_work *w, int cancel) |
1291 | { | |
1292 | struct drbd_request *req = container_of(w, struct drbd_request, w); | |
1293 | ||
0778286a | 1294 | if (bio_data_dir(req->master_bio) == WRITE && req->rq_state & RQ_IN_ACT_LOG) |
265be2d0 PR |
1295 | drbd_al_begin_io(mdev, req->sector); |
1296 | /* Calling drbd_al_begin_io() out of the worker might deadlocks | |
1297 | theoretically. Practically it can not deadlock, since this is | |
1298 | only used when unfreezing IOs. All the extents of the requests | |
1299 | that made it into the TL are already active */ | |
1300 | ||
1301 | drbd_req_make_private_bio(req, req->master_bio); | |
1302 | req->private_bio->bi_bdev = mdev->ldev->backing_bdev; | |
1303 | generic_make_request(req->private_bio); | |
1304 | ||
1305 | return 1; | |
1306 | } | |
1307 | ||
b411b363 PR |
1308 | static int _drbd_may_sync_now(struct drbd_conf *mdev) |
1309 | { | |
1310 | struct drbd_conf *odev = mdev; | |
1311 | ||
1312 | while (1) { | |
1313 | if (odev->sync_conf.after == -1) | |
1314 | return 1; | |
1315 | odev = minor_to_mdev(odev->sync_conf.after); | |
1316 | ERR_IF(!odev) return 1; | |
1317 | if ((odev->state.conn >= C_SYNC_SOURCE && | |
1318 | odev->state.conn <= C_PAUSED_SYNC_T) || | |
1319 | odev->state.aftr_isp || odev->state.peer_isp || | |
1320 | odev->state.user_isp) | |
1321 | return 0; | |
1322 | } | |
1323 | } | |
1324 | ||
1325 | /** | |
1326 | * _drbd_pause_after() - Pause resync on all devices that may not resync now | |
1327 | * @mdev: DRBD device. | |
1328 | * | |
1329 | * Called from process context only (admin command and after_state_ch). | |
1330 | */ | |
1331 | static int _drbd_pause_after(struct drbd_conf *mdev) | |
1332 | { | |
1333 | struct drbd_conf *odev; | |
1334 | int i, rv = 0; | |
1335 | ||
1336 | for (i = 0; i < minor_count; i++) { | |
1337 | odev = minor_to_mdev(i); | |
1338 | if (!odev) | |
1339 | continue; | |
1340 | if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS) | |
1341 | continue; | |
1342 | if (!_drbd_may_sync_now(odev)) | |
1343 | rv |= (__drbd_set_state(_NS(odev, aftr_isp, 1), CS_HARD, NULL) | |
1344 | != SS_NOTHING_TO_DO); | |
1345 | } | |
1346 | ||
1347 | return rv; | |
1348 | } | |
1349 | ||
1350 | /** | |
1351 | * _drbd_resume_next() - Resume resync on all devices that may resync now | |
1352 | * @mdev: DRBD device. | |
1353 | * | |
1354 | * Called from process context only (admin command and worker). | |
1355 | */ | |
1356 | static int _drbd_resume_next(struct drbd_conf *mdev) | |
1357 | { | |
1358 | struct drbd_conf *odev; | |
1359 | int i, rv = 0; | |
1360 | ||
1361 | for (i = 0; i < minor_count; i++) { | |
1362 | odev = minor_to_mdev(i); | |
1363 | if (!odev) | |
1364 | continue; | |
1365 | if (odev->state.conn == C_STANDALONE && odev->state.disk == D_DISKLESS) | |
1366 | continue; | |
1367 | if (odev->state.aftr_isp) { | |
1368 | if (_drbd_may_sync_now(odev)) | |
1369 | rv |= (__drbd_set_state(_NS(odev, aftr_isp, 0), | |
1370 | CS_HARD, NULL) | |
1371 | != SS_NOTHING_TO_DO) ; | |
1372 | } | |
1373 | } | |
1374 | return rv; | |
1375 | } | |
1376 | ||
1377 | void resume_next_sg(struct drbd_conf *mdev) | |
1378 | { | |
1379 | write_lock_irq(&global_state_lock); | |
1380 | _drbd_resume_next(mdev); | |
1381 | write_unlock_irq(&global_state_lock); | |
1382 | } | |
1383 | ||
1384 | void suspend_other_sg(struct drbd_conf *mdev) | |
1385 | { | |
1386 | write_lock_irq(&global_state_lock); | |
1387 | _drbd_pause_after(mdev); | |
1388 | write_unlock_irq(&global_state_lock); | |
1389 | } | |
1390 | ||
1391 | static int sync_after_error(struct drbd_conf *mdev, int o_minor) | |
1392 | { | |
1393 | struct drbd_conf *odev; | |
1394 | ||
1395 | if (o_minor == -1) | |
1396 | return NO_ERROR; | |
1397 | if (o_minor < -1 || minor_to_mdev(o_minor) == NULL) | |
1398 | return ERR_SYNC_AFTER; | |
1399 | ||
1400 | /* check for loops */ | |
1401 | odev = minor_to_mdev(o_minor); | |
1402 | while (1) { | |
1403 | if (odev == mdev) | |
1404 | return ERR_SYNC_AFTER_CYCLE; | |
1405 | ||
1406 | /* dependency chain ends here, no cycles. */ | |
1407 | if (odev->sync_conf.after == -1) | |
1408 | return NO_ERROR; | |
1409 | ||
1410 | /* follow the dependency chain */ | |
1411 | odev = minor_to_mdev(odev->sync_conf.after); | |
1412 | } | |
1413 | } | |
1414 | ||
1415 | int drbd_alter_sa(struct drbd_conf *mdev, int na) | |
1416 | { | |
1417 | int changes; | |
1418 | int retcode; | |
1419 | ||
1420 | write_lock_irq(&global_state_lock); | |
1421 | retcode = sync_after_error(mdev, na); | |
1422 | if (retcode == NO_ERROR) { | |
1423 | mdev->sync_conf.after = na; | |
1424 | do { | |
1425 | changes = _drbd_pause_after(mdev); | |
1426 | changes |= _drbd_resume_next(mdev); | |
1427 | } while (changes); | |
1428 | } | |
1429 | write_unlock_irq(&global_state_lock); | |
1430 | return retcode; | |
1431 | } | |
1432 | ||
1433 | /** | |
1434 | * drbd_start_resync() - Start the resync process | |
1435 | * @mdev: DRBD device. | |
1436 | * @side: Either C_SYNC_SOURCE or C_SYNC_TARGET | |
1437 | * | |
1438 | * This function might bring you directly into one of the | |
1439 | * C_PAUSED_SYNC_* states. | |
1440 | */ | |
1441 | void drbd_start_resync(struct drbd_conf *mdev, enum drbd_conns side) | |
1442 | { | |
1443 | union drbd_state ns; | |
1444 | int r; | |
1445 | ||
1446 | if (mdev->state.conn >= C_SYNC_SOURCE) { | |
1447 | dev_err(DEV, "Resync already running!\n"); | |
1448 | return; | |
1449 | } | |
1450 | ||
b411b363 PR |
1451 | /* In case a previous resync run was aborted by an IO error/detach on the peer. */ |
1452 | drbd_rs_cancel_all(mdev); | |
1453 | ||
1454 | if (side == C_SYNC_TARGET) { | |
1455 | /* Since application IO was locked out during C_WF_BITMAP_T and | |
1456 | C_WF_SYNC_UUID we are still unmodified. Before going to C_SYNC_TARGET | |
1457 | we check that we might make the data inconsistent. */ | |
1458 | r = drbd_khelper(mdev, "before-resync-target"); | |
1459 | r = (r >> 8) & 0xff; | |
1460 | if (r > 0) { | |
1461 | dev_info(DEV, "before-resync-target handler returned %d, " | |
1462 | "dropping connection.\n", r); | |
1463 | drbd_force_state(mdev, NS(conn, C_DISCONNECTING)); | |
1464 | return; | |
1465 | } | |
1466 | } | |
1467 | ||
1468 | drbd_state_lock(mdev); | |
1469 | ||
1470 | if (!get_ldev_if_state(mdev, D_NEGOTIATING)) { | |
1471 | drbd_state_unlock(mdev); | |
1472 | return; | |
1473 | } | |
1474 | ||
1475 | if (side == C_SYNC_TARGET) { | |
1476 | mdev->bm_resync_fo = 0; | |
1477 | } else /* side == C_SYNC_SOURCE */ { | |
1478 | u64 uuid; | |
1479 | ||
1480 | get_random_bytes(&uuid, sizeof(u64)); | |
1481 | drbd_uuid_set(mdev, UI_BITMAP, uuid); | |
1482 | drbd_send_sync_uuid(mdev, uuid); | |
1483 | ||
1484 | D_ASSERT(mdev->state.disk == D_UP_TO_DATE); | |
1485 | } | |
1486 | ||
1487 | write_lock_irq(&global_state_lock); | |
1488 | ns = mdev->state; | |
1489 | ||
1490 | ns.aftr_isp = !_drbd_may_sync_now(mdev); | |
1491 | ||
1492 | ns.conn = side; | |
1493 | ||
1494 | if (side == C_SYNC_TARGET) | |
1495 | ns.disk = D_INCONSISTENT; | |
1496 | else /* side == C_SYNC_SOURCE */ | |
1497 | ns.pdsk = D_INCONSISTENT; | |
1498 | ||
1499 | r = __drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
1500 | ns = mdev->state; | |
1501 | ||
1502 | if (ns.conn < C_CONNECTED) | |
1503 | r = SS_UNKNOWN_ERROR; | |
1504 | ||
1505 | if (r == SS_SUCCESS) { | |
1d7734a0 LE |
1506 | unsigned long tw = drbd_bm_total_weight(mdev); |
1507 | unsigned long now = jiffies; | |
1508 | int i; | |
1509 | ||
b411b363 PR |
1510 | mdev->rs_failed = 0; |
1511 | mdev->rs_paused = 0; | |
b411b363 | 1512 | mdev->rs_same_csum = 0; |
0f0601f4 LE |
1513 | mdev->rs_last_events = 0; |
1514 | mdev->rs_last_sect_ev = 0; | |
1d7734a0 LE |
1515 | mdev->rs_total = tw; |
1516 | mdev->rs_start = now; | |
1517 | for (i = 0; i < DRBD_SYNC_MARKS; i++) { | |
1518 | mdev->rs_mark_left[i] = tw; | |
1519 | mdev->rs_mark_time[i] = now; | |
1520 | } | |
b411b363 PR |
1521 | _drbd_pause_after(mdev); |
1522 | } | |
1523 | write_unlock_irq(&global_state_lock); | |
b411b363 PR |
1524 | put_ldev(mdev); |
1525 | ||
1526 | if (r == SS_SUCCESS) { | |
1527 | dev_info(DEV, "Began resync as %s (will sync %lu KB [%lu bits set]).\n", | |
1528 | drbd_conn_str(ns.conn), | |
1529 | (unsigned long) mdev->rs_total << (BM_BLOCK_SHIFT-10), | |
1530 | (unsigned long) mdev->rs_total); | |
1531 | ||
af85e8e8 LE |
1532 | if (mdev->agreed_pro_version < 95 && mdev->rs_total == 0) { |
1533 | /* This still has a race (about when exactly the peers | |
1534 | * detect connection loss) that can lead to a full sync | |
1535 | * on next handshake. In 8.3.9 we fixed this with explicit | |
1536 | * resync-finished notifications, but the fix | |
1537 | * introduces a protocol change. Sleeping for some | |
1538 | * time longer than the ping interval + timeout on the | |
1539 | * SyncSource, to give the SyncTarget the chance to | |
1540 | * detect connection loss, then waiting for a ping | |
1541 | * response (implicit in drbd_resync_finished) reduces | |
1542 | * the race considerably, but does not solve it. */ | |
1543 | if (side == C_SYNC_SOURCE) | |
1544 | schedule_timeout_interruptible( | |
1545 | mdev->net_conf->ping_int * HZ + | |
1546 | mdev->net_conf->ping_timeo*HZ/9); | |
b411b363 | 1547 | drbd_resync_finished(mdev); |
b411b363 PR |
1548 | } |
1549 | ||
778f271d | 1550 | atomic_set(&mdev->rs_sect_in, 0); |
0f0601f4 | 1551 | atomic_set(&mdev->rs_sect_ev, 0); |
778f271d PR |
1552 | mdev->rs_in_flight = 0; |
1553 | mdev->rs_planed = 0; | |
1554 | spin_lock(&mdev->peer_seq_lock); | |
1555 | fifo_set(&mdev->rs_plan_s, 0); | |
1556 | spin_unlock(&mdev->peer_seq_lock); | |
b411b363 PR |
1557 | /* ns.conn may already be != mdev->state.conn, |
1558 | * we may have been paused in between, or become paused until | |
1559 | * the timer triggers. | |
1560 | * No matter, that is handled in resync_timer_fn() */ | |
1561 | if (ns.conn == C_SYNC_TARGET) | |
1562 | mod_timer(&mdev->resync_timer, jiffies); | |
1563 | ||
1564 | drbd_md_sync(mdev); | |
1565 | } | |
d0c3f60f | 1566 | drbd_state_unlock(mdev); |
b411b363 PR |
1567 | } |
1568 | ||
1569 | int drbd_worker(struct drbd_thread *thi) | |
1570 | { | |
1571 | struct drbd_conf *mdev = thi->mdev; | |
1572 | struct drbd_work *w = NULL; | |
1573 | LIST_HEAD(work_list); | |
1574 | int intr = 0, i; | |
1575 | ||
1576 | sprintf(current->comm, "drbd%d_worker", mdev_to_minor(mdev)); | |
1577 | ||
1578 | while (get_t_state(thi) == Running) { | |
1579 | drbd_thread_current_set_cpu(mdev); | |
1580 | ||
1581 | if (down_trylock(&mdev->data.work.s)) { | |
1582 | mutex_lock(&mdev->data.mutex); | |
1583 | if (mdev->data.socket && !mdev->net_conf->no_cork) | |
1584 | drbd_tcp_uncork(mdev->data.socket); | |
1585 | mutex_unlock(&mdev->data.mutex); | |
1586 | ||
1587 | intr = down_interruptible(&mdev->data.work.s); | |
1588 | ||
1589 | mutex_lock(&mdev->data.mutex); | |
1590 | if (mdev->data.socket && !mdev->net_conf->no_cork) | |
1591 | drbd_tcp_cork(mdev->data.socket); | |
1592 | mutex_unlock(&mdev->data.mutex); | |
1593 | } | |
1594 | ||
1595 | if (intr) { | |
1596 | D_ASSERT(intr == -EINTR); | |
1597 | flush_signals(current); | |
1598 | ERR_IF (get_t_state(thi) == Running) | |
1599 | continue; | |
1600 | break; | |
1601 | } | |
1602 | ||
1603 | if (get_t_state(thi) != Running) | |
1604 | break; | |
1605 | /* With this break, we have done a down() but not consumed | |
1606 | the entry from the list. The cleanup code takes care of | |
1607 | this... */ | |
1608 | ||
1609 | w = NULL; | |
1610 | spin_lock_irq(&mdev->data.work.q_lock); | |
1611 | ERR_IF(list_empty(&mdev->data.work.q)) { | |
1612 | /* something terribly wrong in our logic. | |
1613 | * we were able to down() the semaphore, | |
1614 | * but the list is empty... doh. | |
1615 | * | |
1616 | * what is the best thing to do now? | |
1617 | * try again from scratch, restarting the receiver, | |
1618 | * asender, whatnot? could break even more ugly, | |
1619 | * e.g. when we are primary, but no good local data. | |
1620 | * | |
1621 | * I'll try to get away just starting over this loop. | |
1622 | */ | |
1623 | spin_unlock_irq(&mdev->data.work.q_lock); | |
1624 | continue; | |
1625 | } | |
1626 | w = list_entry(mdev->data.work.q.next, struct drbd_work, list); | |
1627 | list_del_init(&w->list); | |
1628 | spin_unlock_irq(&mdev->data.work.q_lock); | |
1629 | ||
1630 | if (!w->cb(mdev, w, mdev->state.conn < C_CONNECTED)) { | |
1631 | /* dev_warn(DEV, "worker: a callback failed! \n"); */ | |
1632 | if (mdev->state.conn >= C_CONNECTED) | |
1633 | drbd_force_state(mdev, | |
1634 | NS(conn, C_NETWORK_FAILURE)); | |
1635 | } | |
1636 | } | |
1637 | D_ASSERT(test_bit(DEVICE_DYING, &mdev->flags)); | |
1638 | D_ASSERT(test_bit(CONFIG_PENDING, &mdev->flags)); | |
1639 | ||
1640 | spin_lock_irq(&mdev->data.work.q_lock); | |
1641 | i = 0; | |
1642 | while (!list_empty(&mdev->data.work.q)) { | |
1643 | list_splice_init(&mdev->data.work.q, &work_list); | |
1644 | spin_unlock_irq(&mdev->data.work.q_lock); | |
1645 | ||
1646 | while (!list_empty(&work_list)) { | |
1647 | w = list_entry(work_list.next, struct drbd_work, list); | |
1648 | list_del_init(&w->list); | |
1649 | w->cb(mdev, w, 1); | |
1650 | i++; /* dead debugging code */ | |
1651 | } | |
1652 | ||
1653 | spin_lock_irq(&mdev->data.work.q_lock); | |
1654 | } | |
1655 | sema_init(&mdev->data.work.s, 0); | |
1656 | /* DANGEROUS race: if someone did queue his work within the spinlock, | |
1657 | * but up() ed outside the spinlock, we could get an up() on the | |
1658 | * semaphore without corresponding list entry. | |
1659 | * So don't do that. | |
1660 | */ | |
1661 | spin_unlock_irq(&mdev->data.work.q_lock); | |
1662 | ||
1663 | D_ASSERT(mdev->state.disk == D_DISKLESS && mdev->state.conn == C_STANDALONE); | |
1664 | /* _drbd_set_state only uses stop_nowait. | |
1665 | * wait here for the Exiting receiver. */ | |
1666 | drbd_thread_stop(&mdev->receiver); | |
1667 | drbd_mdev_cleanup(mdev); | |
1668 | ||
1669 | dev_info(DEV, "worker terminated\n"); | |
1670 | ||
1671 | clear_bit(DEVICE_DYING, &mdev->flags); | |
1672 | clear_bit(CONFIG_PENDING, &mdev->flags); | |
1673 | wake_up(&mdev->state_wait); | |
1674 | ||
1675 | return 0; | |
1676 | } |