1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
4 * Copyright 2004-2011 Red Hat, Inc.
5 */
6
7 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8
9 #include <linux/fs.h>
10 #include <linux/dlm.h>
11 #include <linux/slab.h>
12 #include <linux/types.h>
13 #include <linux/delay.h>
14 #include <linux/gfs2_ondisk.h>
15 #include <linux/sched/signal.h>
16
17 #include "incore.h"
18 #include "glock.h"
19 #include "glops.h"
20 #include "recovery.h"
21 #include "util.h"
22 #include "sys.h"
23 #include "trace_gfs2.h"
24
25 /**
26 * gfs2_update_stats - Update time based stats
27 * @s: The stats to update (local or global)
28 * @index: The index inside @s
29 * @sample: New data to include
30 */
gfs2_update_stats(struct gfs2_lkstats * s,unsigned index,s64 sample)31 static inline void gfs2_update_stats(struct gfs2_lkstats *s, unsigned index,
32 s64 sample)
33 {
34 /*
35 * @delta is the difference between the current rtt sample and the
36 * running average srtt. We add 1/8 of that to the srtt in order to
37 * update the current srtt estimate. The variance estimate is a bit
38 * more complicated. We subtract the current variance estimate from
39 * the abs value of the @delta and add 1/4 of that to the running
40 * total. That's equivalent to 3/4 of the current variance
41 * estimate plus 1/4 of the abs of @delta.
42 *
43 * Note that the index points at the array entry containing the
44 * smoothed mean value, and the variance is always in the following
45 * entry
46 *
47 * Reference: TCP/IP Illustrated, vol 2, p. 831,832
48 * All times are in units of integer nanoseconds. Unlike the TCP/IP
49 * case, they are not scaled fixed point.
50 */
51
52 s64 delta = sample - s->stats[index];
53 s->stats[index] += (delta >> 3);
54 index++;
55 s->stats[index] += (s64)(abs(delta) - s->stats[index]) >> 2;
56 }
57
58 /**
59 * gfs2_update_reply_times - Update locking statistics
60 * @gl: The glock to update
61 *
62 * This assumes that gl->gl_dstamp has been set earlier.
63 *
64 * The rtt (lock round trip time) is an estimate of the time
65 * taken to perform a dlm lock request. We update it on each
66 * reply from the dlm.
67 *
68 * The blocking flag is set on the glock for all dlm requests
69 * which may potentially block due to lock requests from other nodes.
70 * DLM requests where the current lock state is exclusive, the
71 * requested state is null (or unlocked) or where the TRY or
72 * TRY_1CB flags are set are classified as non-blocking. All
73 * other DLM requests are counted as (potentially) blocking.
74 */
gfs2_update_reply_times(struct gfs2_glock * gl)75 static inline void gfs2_update_reply_times(struct gfs2_glock *gl)
76 {
77 struct gfs2_pcpu_lkstats *lks;
78 const unsigned gltype = gl->gl_name.ln_type;
79 unsigned index = test_bit(GLF_BLOCKING, &gl->gl_flags) ?
80 GFS2_LKS_SRTTB : GFS2_LKS_SRTT;
81 s64 rtt;
82
83 preempt_disable();
84 rtt = ktime_to_ns(ktime_sub(ktime_get_real(), gl->gl_dstamp));
85 lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats);
86 gfs2_update_stats(&gl->gl_stats, index, rtt); /* Local */
87 gfs2_update_stats(&lks->lkstats[gltype], index, rtt); /* Global */
88 preempt_enable();
89
90 trace_gfs2_glock_lock_time(gl, rtt);
91 }
92
93 /**
94 * gfs2_update_request_times - Update locking statistics
95 * @gl: The glock to update
96 *
97 * The irt (lock inter-request times) measures the average time
98 * between requests to the dlm. It is updated immediately before
99 * each dlm call.
100 */
101
gfs2_update_request_times(struct gfs2_glock * gl)102 static inline void gfs2_update_request_times(struct gfs2_glock *gl)
103 {
104 struct gfs2_pcpu_lkstats *lks;
105 const unsigned gltype = gl->gl_name.ln_type;
106 ktime_t dstamp;
107 s64 irt;
108
109 preempt_disable();
110 dstamp = gl->gl_dstamp;
111 gl->gl_dstamp = ktime_get_real();
112 irt = ktime_to_ns(ktime_sub(gl->gl_dstamp, dstamp));
113 lks = this_cpu_ptr(gl->gl_name.ln_sbd->sd_lkstats);
114 gfs2_update_stats(&gl->gl_stats, GFS2_LKS_SIRT, irt); /* Local */
115 gfs2_update_stats(&lks->lkstats[gltype], GFS2_LKS_SIRT, irt); /* Global */
116 preempt_enable();
117 }
118
gdlm_ast(void * arg)119 static void gdlm_ast(void *arg)
120 {
121 struct gfs2_glock *gl = arg;
122 unsigned ret = gl->gl_state;
123
124 /* If the glock is dead, we only react to a dlm_unlock() reply. */
125 if (__lockref_is_dead(&gl->gl_lockref) &&
126 gl->gl_lksb.sb_status != -DLM_EUNLOCK)
127 return;
128
129 gfs2_update_reply_times(gl);
130 BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED);
131
132 if ((gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) && gl->gl_lksb.sb_lvbptr)
133 memset(gl->gl_lksb.sb_lvbptr, 0, GDLM_LVB_SIZE);
134
135 switch (gl->gl_lksb.sb_status) {
136 case -DLM_EUNLOCK: /* Unlocked, so glock can be freed */
137 if (gl->gl_ops->go_free)
138 gl->gl_ops->go_free(gl);
139 gfs2_glock_free(gl);
140 return;
141 case -DLM_ECANCEL: /* Cancel while getting lock */
142 ret |= LM_OUT_CANCELED;
143 goto out;
144 case -EAGAIN: /* Try lock fails */
145 case -EDEADLK: /* Deadlock detected */
146 goto out;
147 case -ETIMEDOUT: /* Canceled due to timeout */
148 ret |= LM_OUT_ERROR;
149 goto out;
150 case 0: /* Success */
151 break;
152 default: /* Something unexpected */
153 BUG();
154 }
155
156 ret = gl->gl_req;
157 if (gl->gl_lksb.sb_flags & DLM_SBF_ALTMODE) {
158 if (gl->gl_req == LM_ST_SHARED)
159 ret = LM_ST_DEFERRED;
160 else if (gl->gl_req == LM_ST_DEFERRED)
161 ret = LM_ST_SHARED;
162 else
163 BUG();
164 }
165
166 set_bit(GLF_INITIAL, &gl->gl_flags);
167 gfs2_glock_complete(gl, ret);
168 return;
169 out:
170 if (!test_bit(GLF_INITIAL, &gl->gl_flags))
171 gl->gl_lksb.sb_lkid = 0;
172 gfs2_glock_complete(gl, ret);
173 }
174
gdlm_bast(void * arg,int mode)175 static void gdlm_bast(void *arg, int mode)
176 {
177 struct gfs2_glock *gl = arg;
178
179 if (__lockref_is_dead(&gl->gl_lockref))
180 return;
181
182 switch (mode) {
183 case DLM_LOCK_EX:
184 gfs2_glock_cb(gl, LM_ST_UNLOCKED);
185 break;
186 case DLM_LOCK_CW:
187 gfs2_glock_cb(gl, LM_ST_DEFERRED);
188 break;
189 case DLM_LOCK_PR:
190 gfs2_glock_cb(gl, LM_ST_SHARED);
191 break;
192 default:
193 fs_err(gl->gl_name.ln_sbd, "unknown bast mode %d\n", mode);
194 BUG();
195 }
196 }
197
198 /* convert gfs lock-state to dlm lock-mode */
199
make_mode(struct gfs2_sbd * sdp,const unsigned int lmstate)200 static int make_mode(struct gfs2_sbd *sdp, const unsigned int lmstate)
201 {
202 switch (lmstate) {
203 case LM_ST_UNLOCKED:
204 return DLM_LOCK_NL;
205 case LM_ST_EXCLUSIVE:
206 return DLM_LOCK_EX;
207 case LM_ST_DEFERRED:
208 return DLM_LOCK_CW;
209 case LM_ST_SHARED:
210 return DLM_LOCK_PR;
211 }
212 fs_err(sdp, "unknown LM state %d\n", lmstate);
213 BUG();
214 return -1;
215 }
216
make_flags(struct gfs2_glock * gl,const unsigned int gfs_flags,const int req)217 static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags,
218 const int req)
219 {
220 u32 lkf = 0;
221
222 if (gl->gl_lksb.sb_lvbptr)
223 lkf |= DLM_LKF_VALBLK;
224
225 if (gfs_flags & LM_FLAG_TRY)
226 lkf |= DLM_LKF_NOQUEUE;
227
228 if (gfs_flags & LM_FLAG_TRY_1CB) {
229 lkf |= DLM_LKF_NOQUEUE;
230 lkf |= DLM_LKF_NOQUEUEBAST;
231 }
232
233 if (gfs_flags & LM_FLAG_ANY) {
234 if (req == DLM_LOCK_PR)
235 lkf |= DLM_LKF_ALTCW;
236 else if (req == DLM_LOCK_CW)
237 lkf |= DLM_LKF_ALTPR;
238 else
239 BUG();
240 }
241
242 if (gl->gl_lksb.sb_lkid != 0) {
243 lkf |= DLM_LKF_CONVERT;
244 if (test_bit(GLF_BLOCKING, &gl->gl_flags))
245 lkf |= DLM_LKF_QUECVT;
246 }
247
248 return lkf;
249 }
250
gfs2_reverse_hex(char * c,u64 value)251 static void gfs2_reverse_hex(char *c, u64 value)
252 {
253 *c = '0';
254 while (value) {
255 *c-- = hex_asc[value & 0x0f];
256 value >>= 4;
257 }
258 }
259
gdlm_lock(struct gfs2_glock * gl,unsigned int req_state,unsigned int flags)260 static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state,
261 unsigned int flags)
262 {
263 struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
264 int req;
265 u32 lkf;
266 char strname[GDLM_STRNAME_BYTES] = "";
267 int error;
268
269 req = make_mode(gl->gl_name.ln_sbd, req_state);
270 lkf = make_flags(gl, flags, req);
271 gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT);
272 gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
273 if (gl->gl_lksb.sb_lkid) {
274 gfs2_update_request_times(gl);
275 } else {
276 memset(strname, ' ', GDLM_STRNAME_BYTES - 1);
277 strname[GDLM_STRNAME_BYTES - 1] = '\0';
278 gfs2_reverse_hex(strname + 7, gl->gl_name.ln_type);
279 gfs2_reverse_hex(strname + 23, gl->gl_name.ln_number);
280 gl->gl_dstamp = ktime_get_real();
281 }
282 /*
283 * Submit the actual lock request.
284 */
285
286 again:
287 error = dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, strname,
288 GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast);
289 if (error == -EBUSY) {
290 msleep(20);
291 goto again;
292 }
293 return error;
294 }
295
gdlm_put_lock(struct gfs2_glock * gl)296 static void gdlm_put_lock(struct gfs2_glock *gl)
297 {
298 struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
299 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
300 int error;
301
302 BUG_ON(!__lockref_is_dead(&gl->gl_lockref));
303
304 if (gl->gl_lksb.sb_lkid == 0) {
305 gfs2_glock_free(gl);
306 return;
307 }
308
309 clear_bit(GLF_BLOCKING, &gl->gl_flags);
310 gfs2_glstats_inc(gl, GFS2_LKS_DCOUNT);
311 gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
312 gfs2_update_request_times(gl);
313
314 /* don't want to call dlm if we've unmounted the lock protocol */
315 if (test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) {
316 gfs2_glock_free(gl);
317 return;
318 }
319 /* don't want to skip dlm_unlock writing the lvb when lock has one */
320
321 if (test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags) &&
322 !gl->gl_lksb.sb_lvbptr) {
323 gfs2_glock_free_later(gl);
324 return;
325 }
326
327 again:
328 error = dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK,
329 NULL, gl);
330 if (error == -EBUSY) {
331 msleep(20);
332 goto again;
333 }
334
335 if (error) {
336 fs_err(sdp, "gdlm_unlock %x,%llx err=%d\n",
337 gl->gl_name.ln_type,
338 (unsigned long long)gl->gl_name.ln_number, error);
339 }
340 }
341
gdlm_cancel(struct gfs2_glock * gl)342 static void gdlm_cancel(struct gfs2_glock *gl)
343 {
344 struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
345 dlm_unlock(ls->ls_dlm, gl->gl_lksb.sb_lkid, DLM_LKF_CANCEL, NULL, gl);
346 }
347
348 /*
349 * dlm/gfs2 recovery coordination using dlm_recover callbacks
350 *
351 * 0. gfs2 checks for another cluster node withdraw, needing journal replay
352 * 1. dlm_controld sees lockspace members change
353 * 2. dlm_controld blocks dlm-kernel locking activity
354 * 3. dlm_controld within dlm-kernel notifies gfs2 (recover_prep)
355 * 4. dlm_controld starts and finishes its own user level recovery
356 * 5. dlm_controld starts dlm-kernel dlm_recoverd to do kernel recovery
357 * 6. dlm_recoverd notifies gfs2 of failed nodes (recover_slot)
358 * 7. dlm_recoverd does its own lock recovery
359 * 8. dlm_recoverd unblocks dlm-kernel locking activity
360 * 9. dlm_recoverd notifies gfs2 when done (recover_done with new generation)
361 * 10. gfs2_control updates control_lock lvb with new generation and jid bits
362 * 11. gfs2_control enqueues journals for gfs2_recover to recover (maybe none)
363 * 12. gfs2_recover dequeues and recovers journals of failed nodes
364 * 13. gfs2_recover provides recovery results to gfs2_control (recovery_result)
365 * 14. gfs2_control updates control_lock lvb jid bits for recovered journals
366 * 15. gfs2_control unblocks normal locking when all journals are recovered
367 *
368 * - failures during recovery
369 *
370 * recover_prep() may set BLOCK_LOCKS (step 3) again before gfs2_control
371 * clears BLOCK_LOCKS (step 15), e.g. another node fails while still
372 * recovering for a prior failure. gfs2_control needs a way to detect
373 * this so it can leave BLOCK_LOCKS set in step 15. This is managed using
374 * the recover_block and recover_start values.
375 *
376 * recover_done() provides a new lockspace generation number each time it
377 * is called (step 9). This generation number is saved as recover_start.
378 * When recover_prep() is called, it sets BLOCK_LOCKS and sets
379 * recover_block = recover_start. So, while recover_block is equal to
380 * recover_start, BLOCK_LOCKS should remain set. (recover_spin must
381 * be held around the BLOCK_LOCKS/recover_block/recover_start logic.)
382 *
383 * - more specific gfs2 steps in sequence above
384 *
385 * 3. recover_prep sets BLOCK_LOCKS and sets recover_block = recover_start
386 * 6. recover_slot records any failed jids (maybe none)
387 * 9. recover_done sets recover_start = new generation number
388 * 10. gfs2_control sets control_lock lvb = new gen + bits for failed jids
389 * 12. gfs2_recover does journal recoveries for failed jids identified above
390 * 14. gfs2_control clears control_lock lvb bits for recovered jids
391 * 15. gfs2_control checks if recover_block == recover_start (step 3 occured
392 * again) then do nothing, otherwise if recover_start > recover_block
393 * then clear BLOCK_LOCKS.
394 *
395 * - parallel recovery steps across all nodes
396 *
397 * All nodes attempt to update the control_lock lvb with the new generation
398 * number and jid bits, but only the first to get the control_lock EX will
399 * do so; others will see that it's already done (lvb already contains new
400 * generation number.)
401 *
402 * . All nodes get the same recover_prep/recover_slot/recover_done callbacks
403 * . All nodes attempt to set control_lock lvb gen + bits for the new gen
404 * . One node gets control_lock first and writes the lvb, others see it's done
405 * . All nodes attempt to recover jids for which they see control_lock bits set
406 * . One node succeeds for a jid, and that one clears the jid bit in the lvb
407 * . All nodes will eventually see all lvb bits clear and unblock locks
408 *
409 * - is there a problem with clearing an lvb bit that should be set
410 * and missing a journal recovery?
411 *
412 * 1. jid fails
413 * 2. lvb bit set for step 1
414 * 3. jid recovered for step 1
415 * 4. jid taken again (new mount)
416 * 5. jid fails (for step 4)
417 * 6. lvb bit set for step 5 (will already be set)
418 * 7. lvb bit cleared for step 3
419 *
420 * This is not a problem because the failure in step 5 does not
421 * require recovery, because the mount in step 4 could not have
422 * progressed far enough to unblock locks and access the fs. The
423 * control_mount() function waits for all recoveries to be complete
424 * for the latest lockspace generation before ever unblocking locks
425 * and returning. The mount in step 4 waits until the recovery in
426 * step 1 is done.
427 *
428 * - special case of first mounter: first node to mount the fs
429 *
430 * The first node to mount a gfs2 fs needs to check all the journals
431 * and recover any that need recovery before other nodes are allowed
432 * to mount the fs. (Others may begin mounting, but they must wait
433 * for the first mounter to be done before taking locks on the fs
434 * or accessing the fs.) This has two parts:
435 *
436 * 1. The mounted_lock tells a node it's the first to mount the fs.
437 * Each node holds the mounted_lock in PR while it's mounted.
438 * Each node tries to acquire the mounted_lock in EX when it mounts.
439 * If a node is granted the mounted_lock EX it means there are no
440 * other mounted nodes (no PR locks exist), and it is the first mounter.
441 * The mounted_lock is demoted to PR when first recovery is done, so
442 * others will fail to get an EX lock, but will get a PR lock.
443 *
444 * 2. The control_lock blocks others in control_mount() while the first
445 * mounter is doing first mount recovery of all journals.
446 * A mounting node needs to acquire control_lock in EX mode before
447 * it can proceed. The first mounter holds control_lock in EX while doing
448 * the first mount recovery, blocking mounts from other nodes, then demotes
449 * control_lock to NL when it's done (others_may_mount/first_done),
450 * allowing other nodes to continue mounting.
451 *
452 * first mounter:
453 * control_lock EX/NOQUEUE success
454 * mounted_lock EX/NOQUEUE success (no other PR, so no other mounters)
455 * set first=1
456 * do first mounter recovery
457 * mounted_lock EX->PR
458 * control_lock EX->NL, write lvb generation
459 *
460 * other mounter:
461 * control_lock EX/NOQUEUE success (if fail -EAGAIN, retry)
462 * mounted_lock EX/NOQUEUE fail -EAGAIN (expected due to other mounters PR)
463 * mounted_lock PR/NOQUEUE success
464 * read lvb generation
465 * control_lock EX->NL
466 * set first=0
467 *
468 * - mount during recovery
469 *
470 * If a node mounts while others are doing recovery (not first mounter),
471 * the mounting node will get its initial recover_done() callback without
472 * having seen any previous failures/callbacks.
473 *
474 * It must wait for all recoveries preceding its mount to be finished
475 * before it unblocks locks. It does this by repeating the "other mounter"
476 * steps above until the lvb generation number is >= its mount generation
477 * number (from initial recover_done) and all lvb bits are clear.
478 *
479 * - control_lock lvb format
480 *
481 * 4 bytes generation number: the latest dlm lockspace generation number
482 * from recover_done callback. Indicates the jid bitmap has been updated
483 * to reflect all slot failures through that generation.
484 * 4 bytes unused.
485 * GDLM_LVB_SIZE-8 bytes of jid bit map. If bit N is set, it indicates
486 * that jid N needs recovery.
487 */
488
489 #define JID_BITMAP_OFFSET 8 /* 4 byte generation number + 4 byte unused */
490
control_lvb_read(struct lm_lockstruct * ls,uint32_t * lvb_gen,char * lvb_bits)491 static void control_lvb_read(struct lm_lockstruct *ls, uint32_t *lvb_gen,
492 char *lvb_bits)
493 {
494 __le32 gen;
495 memcpy(lvb_bits, ls->ls_control_lvb, GDLM_LVB_SIZE);
496 memcpy(&gen, lvb_bits, sizeof(__le32));
497 *lvb_gen = le32_to_cpu(gen);
498 }
499
control_lvb_write(struct lm_lockstruct * ls,uint32_t lvb_gen,char * lvb_bits)500 static void control_lvb_write(struct lm_lockstruct *ls, uint32_t lvb_gen,
501 char *lvb_bits)
502 {
503 __le32 gen;
504 memcpy(ls->ls_control_lvb, lvb_bits, GDLM_LVB_SIZE);
505 gen = cpu_to_le32(lvb_gen);
506 memcpy(ls->ls_control_lvb, &gen, sizeof(__le32));
507 }
508
all_jid_bits_clear(char * lvb)509 static int all_jid_bits_clear(char *lvb)
510 {
511 return !memchr_inv(lvb + JID_BITMAP_OFFSET, 0,
512 GDLM_LVB_SIZE - JID_BITMAP_OFFSET);
513 }
514
sync_wait_cb(void * arg)515 static void sync_wait_cb(void *arg)
516 {
517 struct lm_lockstruct *ls = arg;
518 complete(&ls->ls_sync_wait);
519 }
520
sync_unlock(struct gfs2_sbd * sdp,struct dlm_lksb * lksb,char * name)521 static int sync_unlock(struct gfs2_sbd *sdp, struct dlm_lksb *lksb, char *name)
522 {
523 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
524 int error;
525
526 error = dlm_unlock(ls->ls_dlm, lksb->sb_lkid, 0, lksb, ls);
527 if (error) {
528 fs_err(sdp, "%s lkid %x error %d\n",
529 name, lksb->sb_lkid, error);
530 return error;
531 }
532
533 wait_for_completion(&ls->ls_sync_wait);
534
535 if (lksb->sb_status != -DLM_EUNLOCK) {
536 fs_err(sdp, "%s lkid %x status %d\n",
537 name, lksb->sb_lkid, lksb->sb_status);
538 return -1;
539 }
540 return 0;
541 }
542
sync_lock(struct gfs2_sbd * sdp,int mode,uint32_t flags,unsigned int num,struct dlm_lksb * lksb,char * name)543 static int sync_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags,
544 unsigned int num, struct dlm_lksb *lksb, char *name)
545 {
546 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
547 char strname[GDLM_STRNAME_BYTES];
548 int error, status;
549
550 memset(strname, 0, GDLM_STRNAME_BYTES);
551 snprintf(strname, GDLM_STRNAME_BYTES, "%8x%16x", LM_TYPE_NONDISK, num);
552
553 error = dlm_lock(ls->ls_dlm, mode, lksb, flags,
554 strname, GDLM_STRNAME_BYTES - 1,
555 0, sync_wait_cb, ls, NULL);
556 if (error) {
557 fs_err(sdp, "%s lkid %x flags %x mode %d error %d\n",
558 name, lksb->sb_lkid, flags, mode, error);
559 return error;
560 }
561
562 wait_for_completion(&ls->ls_sync_wait);
563
564 status = lksb->sb_status;
565
566 if (status && status != -EAGAIN) {
567 fs_err(sdp, "%s lkid %x flags %x mode %d status %d\n",
568 name, lksb->sb_lkid, flags, mode, status);
569 }
570
571 return status;
572 }
573
mounted_unlock(struct gfs2_sbd * sdp)574 static int mounted_unlock(struct gfs2_sbd *sdp)
575 {
576 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
577 return sync_unlock(sdp, &ls->ls_mounted_lksb, "mounted_lock");
578 }
579
mounted_lock(struct gfs2_sbd * sdp,int mode,uint32_t flags)580 static int mounted_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags)
581 {
582 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
583 return sync_lock(sdp, mode, flags, GFS2_MOUNTED_LOCK,
584 &ls->ls_mounted_lksb, "mounted_lock");
585 }
586
control_unlock(struct gfs2_sbd * sdp)587 static int control_unlock(struct gfs2_sbd *sdp)
588 {
589 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
590 return sync_unlock(sdp, &ls->ls_control_lksb, "control_lock");
591 }
592
control_lock(struct gfs2_sbd * sdp,int mode,uint32_t flags)593 static int control_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags)
594 {
595 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
596 return sync_lock(sdp, mode, flags, GFS2_CONTROL_LOCK,
597 &ls->ls_control_lksb, "control_lock");
598 }
599
600 /**
601 * remote_withdraw - react to a node withdrawing from the file system
602 * @sdp: The superblock
603 */
remote_withdraw(struct gfs2_sbd * sdp)604 static void remote_withdraw(struct gfs2_sbd *sdp)
605 {
606 struct gfs2_jdesc *jd;
607 int ret = 0, count = 0;
608
609 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
610 if (jd->jd_jid == sdp->sd_lockstruct.ls_jid)
611 continue;
612 ret = gfs2_recover_journal(jd, true);
613 if (ret)
614 break;
615 count++;
616 }
617
618 /* Now drop the additional reference we acquired */
619 fs_err(sdp, "Journals checked: %d, ret = %d.\n", count, ret);
620 }
621
gfs2_control_func(struct work_struct * work)622 static void gfs2_control_func(struct work_struct *work)
623 {
624 struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work);
625 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
626 uint32_t block_gen, start_gen, lvb_gen, flags;
627 int recover_set = 0;
628 int write_lvb = 0;
629 int recover_size;
630 int i, error;
631
632 /* First check for other nodes that may have done a withdraw. */
633 if (test_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags)) {
634 remote_withdraw(sdp);
635 clear_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags);
636 return;
637 }
638
639 spin_lock(&ls->ls_recover_spin);
640 /*
641 * No MOUNT_DONE means we're still mounting; control_mount()
642 * will set this flag, after which this thread will take over
643 * all further clearing of BLOCK_LOCKS.
644 *
645 * FIRST_MOUNT means this node is doing first mounter recovery,
646 * for which recovery control is handled by
647 * control_mount()/control_first_done(), not this thread.
648 */
649 if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
650 test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
651 spin_unlock(&ls->ls_recover_spin);
652 return;
653 }
654 block_gen = ls->ls_recover_block;
655 start_gen = ls->ls_recover_start;
656 spin_unlock(&ls->ls_recover_spin);
657
658 /*
659 * Equal block_gen and start_gen implies we are between
660 * recover_prep and recover_done callbacks, which means
661 * dlm recovery is in progress and dlm locking is blocked.
662 * There's no point trying to do any work until recover_done.
663 */
664
665 if (block_gen == start_gen)
666 return;
667
668 /*
669 * Propagate recover_submit[] and recover_result[] to lvb:
670 * dlm_recoverd adds to recover_submit[] jids needing recovery
671 * gfs2_recover adds to recover_result[] journal recovery results
672 *
673 * set lvb bit for jids in recover_submit[] if the lvb has not
674 * yet been updated for the generation of the failure
675 *
676 * clear lvb bit for jids in recover_result[] if the result of
677 * the journal recovery is SUCCESS
678 */
679
680 error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
681 if (error) {
682 fs_err(sdp, "control lock EX error %d\n", error);
683 return;
684 }
685
686 control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
687
688 spin_lock(&ls->ls_recover_spin);
689 if (block_gen != ls->ls_recover_block ||
690 start_gen != ls->ls_recover_start) {
691 fs_info(sdp, "recover generation %u block1 %u %u\n",
692 start_gen, block_gen, ls->ls_recover_block);
693 spin_unlock(&ls->ls_recover_spin);
694 control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
695 return;
696 }
697
698 recover_size = ls->ls_recover_size;
699
700 if (lvb_gen <= start_gen) {
701 /*
702 * Clear lvb bits for jids we've successfully recovered.
703 * Because all nodes attempt to recover failed journals,
704 * a journal can be recovered multiple times successfully
705 * in succession. Only the first will really do recovery,
706 * the others find it clean, but still report a successful
707 * recovery. So, another node may have already recovered
708 * the jid and cleared the lvb bit for it.
709 */
710 for (i = 0; i < recover_size; i++) {
711 if (ls->ls_recover_result[i] != LM_RD_SUCCESS)
712 continue;
713
714 ls->ls_recover_result[i] = 0;
715
716 if (!test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET))
717 continue;
718
719 __clear_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
720 write_lvb = 1;
721 }
722 }
723
724 if (lvb_gen == start_gen) {
725 /*
726 * Failed slots before start_gen are already set in lvb.
727 */
728 for (i = 0; i < recover_size; i++) {
729 if (!ls->ls_recover_submit[i])
730 continue;
731 if (ls->ls_recover_submit[i] < lvb_gen)
732 ls->ls_recover_submit[i] = 0;
733 }
734 } else if (lvb_gen < start_gen) {
735 /*
736 * Failed slots before start_gen are not yet set in lvb.
737 */
738 for (i = 0; i < recover_size; i++) {
739 if (!ls->ls_recover_submit[i])
740 continue;
741 if (ls->ls_recover_submit[i] < start_gen) {
742 ls->ls_recover_submit[i] = 0;
743 __set_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
744 }
745 }
746 /* even if there are no bits to set, we need to write the
747 latest generation to the lvb */
748 write_lvb = 1;
749 } else {
750 /*
751 * we should be getting a recover_done() for lvb_gen soon
752 */
753 }
754 spin_unlock(&ls->ls_recover_spin);
755
756 if (write_lvb) {
757 control_lvb_write(ls, start_gen, ls->ls_lvb_bits);
758 flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK;
759 } else {
760 flags = DLM_LKF_CONVERT;
761 }
762
763 error = control_lock(sdp, DLM_LOCK_NL, flags);
764 if (error) {
765 fs_err(sdp, "control lock NL error %d\n", error);
766 return;
767 }
768
769 /*
770 * Everyone will see jid bits set in the lvb, run gfs2_recover_set(),
771 * and clear a jid bit in the lvb if the recovery is a success.
772 * Eventually all journals will be recovered, all jid bits will
773 * be cleared in the lvb, and everyone will clear BLOCK_LOCKS.
774 */
775
776 for (i = 0; i < recover_size; i++) {
777 if (test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET)) {
778 fs_info(sdp, "recover generation %u jid %d\n",
779 start_gen, i);
780 gfs2_recover_set(sdp, i);
781 recover_set++;
782 }
783 }
784 if (recover_set)
785 return;
786
787 /*
788 * No more jid bits set in lvb, all recovery is done, unblock locks
789 * (unless a new recover_prep callback has occured blocking locks
790 * again while working above)
791 */
792
793 spin_lock(&ls->ls_recover_spin);
794 if (ls->ls_recover_block == block_gen &&
795 ls->ls_recover_start == start_gen) {
796 clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
797 spin_unlock(&ls->ls_recover_spin);
798 fs_info(sdp, "recover generation %u done\n", start_gen);
799 gfs2_glock_thaw(sdp);
800 } else {
801 fs_info(sdp, "recover generation %u block2 %u %u\n",
802 start_gen, block_gen, ls->ls_recover_block);
803 spin_unlock(&ls->ls_recover_spin);
804 }
805 }
806
control_mount(struct gfs2_sbd * sdp)807 static int control_mount(struct gfs2_sbd *sdp)
808 {
809 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
810 uint32_t start_gen, block_gen, mount_gen, lvb_gen;
811 int mounted_mode;
812 int retries = 0;
813 int error;
814
815 memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb));
816 memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb));
817 memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE);
818 ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb;
819 init_completion(&ls->ls_sync_wait);
820
821 set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
822
823 error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK);
824 if (error) {
825 fs_err(sdp, "control_mount control_lock NL error %d\n", error);
826 return error;
827 }
828
829 error = mounted_lock(sdp, DLM_LOCK_NL, 0);
830 if (error) {
831 fs_err(sdp, "control_mount mounted_lock NL error %d\n", error);
832 control_unlock(sdp);
833 return error;
834 }
835 mounted_mode = DLM_LOCK_NL;
836
837 restart:
838 if (retries++ && signal_pending(current)) {
839 error = -EINTR;
840 goto fail;
841 }
842
843 /*
844 * We always start with both locks in NL. control_lock is
845 * demoted to NL below so we don't need to do it here.
846 */
847
848 if (mounted_mode != DLM_LOCK_NL) {
849 error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
850 if (error)
851 goto fail;
852 mounted_mode = DLM_LOCK_NL;
853 }
854
855 /*
856 * Other nodes need to do some work in dlm recovery and gfs2_control
857 * before the recover_done and control_lock will be ready for us below.
858 * A delay here is not required but often avoids having to retry.
859 */
860
861 msleep_interruptible(500);
862
863 /*
864 * Acquire control_lock in EX and mounted_lock in either EX or PR.
865 * control_lock lvb keeps track of any pending journal recoveries.
866 * mounted_lock indicates if any other nodes have the fs mounted.
867 */
868
869 error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK);
870 if (error == -EAGAIN) {
871 goto restart;
872 } else if (error) {
873 fs_err(sdp, "control_mount control_lock EX error %d\n", error);
874 goto fail;
875 }
876
877 /**
878 * If we're a spectator, we don't want to take the lock in EX because
879 * we cannot do the first-mount responsibility it implies: recovery.
880 */
881 if (sdp->sd_args.ar_spectator)
882 goto locks_done;
883
884 error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
885 if (!error) {
886 mounted_mode = DLM_LOCK_EX;
887 goto locks_done;
888 } else if (error != -EAGAIN) {
889 fs_err(sdp, "control_mount mounted_lock EX error %d\n", error);
890 goto fail;
891 }
892
893 error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
894 if (!error) {
895 mounted_mode = DLM_LOCK_PR;
896 goto locks_done;
897 } else {
898 /* not even -EAGAIN should happen here */
899 fs_err(sdp, "control_mount mounted_lock PR error %d\n", error);
900 goto fail;
901 }
902
903 locks_done:
904 /*
905 * If we got both locks above in EX, then we're the first mounter.
906 * If not, then we need to wait for the control_lock lvb to be
907 * updated by other mounted nodes to reflect our mount generation.
908 *
909 * In simple first mounter cases, first mounter will see zero lvb_gen,
910 * but in cases where all existing nodes leave/fail before mounting
911 * nodes finish control_mount, then all nodes will be mounting and
912 * lvb_gen will be non-zero.
913 */
914
915 control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
916
917 if (lvb_gen == 0xFFFFFFFF) {
918 /* special value to force mount attempts to fail */
919 fs_err(sdp, "control_mount control_lock disabled\n");
920 error = -EINVAL;
921 goto fail;
922 }
923
924 if (mounted_mode == DLM_LOCK_EX) {
925 /* first mounter, keep both EX while doing first recovery */
926 spin_lock(&ls->ls_recover_spin);
927 clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
928 set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
929 set_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
930 spin_unlock(&ls->ls_recover_spin);
931 fs_info(sdp, "first mounter control generation %u\n", lvb_gen);
932 return 0;
933 }
934
935 error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
936 if (error)
937 goto fail;
938
939 /*
940 * We are not first mounter, now we need to wait for the control_lock
941 * lvb generation to be >= the generation from our first recover_done
942 * and all lvb bits to be clear (no pending journal recoveries.)
943 */
944
945 if (!all_jid_bits_clear(ls->ls_lvb_bits)) {
946 /* journals need recovery, wait until all are clear */
947 fs_info(sdp, "control_mount wait for journal recovery\n");
948 goto restart;
949 }
950
951 spin_lock(&ls->ls_recover_spin);
952 block_gen = ls->ls_recover_block;
953 start_gen = ls->ls_recover_start;
954 mount_gen = ls->ls_recover_mount;
955
956 if (lvb_gen < mount_gen) {
957 /* wait for mounted nodes to update control_lock lvb to our
958 generation, which might include new recovery bits set */
959 if (sdp->sd_args.ar_spectator) {
960 fs_info(sdp, "Recovery is required. Waiting for a "
961 "non-spectator to mount.\n");
962 msleep_interruptible(1000);
963 } else {
964 fs_info(sdp, "control_mount wait1 block %u start %u "
965 "mount %u lvb %u flags %lx\n", block_gen,
966 start_gen, mount_gen, lvb_gen,
967 ls->ls_recover_flags);
968 }
969 spin_unlock(&ls->ls_recover_spin);
970 goto restart;
971 }
972
973 if (lvb_gen != start_gen) {
974 /* wait for mounted nodes to update control_lock lvb to the
975 latest recovery generation */
976 fs_info(sdp, "control_mount wait2 block %u start %u mount %u "
977 "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
978 lvb_gen, ls->ls_recover_flags);
979 spin_unlock(&ls->ls_recover_spin);
980 goto restart;
981 }
982
983 if (block_gen == start_gen) {
984 /* dlm recovery in progress, wait for it to finish */
985 fs_info(sdp, "control_mount wait3 block %u start %u mount %u "
986 "lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
987 lvb_gen, ls->ls_recover_flags);
988 spin_unlock(&ls->ls_recover_spin);
989 goto restart;
990 }
991
992 clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
993 set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
994 memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
995 memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
996 spin_unlock(&ls->ls_recover_spin);
997 return 0;
998
999 fail:
1000 mounted_unlock(sdp);
1001 control_unlock(sdp);
1002 return error;
1003 }
1004
control_first_done(struct gfs2_sbd * sdp)1005 static int control_first_done(struct gfs2_sbd *sdp)
1006 {
1007 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1008 uint32_t start_gen, block_gen;
1009 int error;
1010
1011 restart:
1012 spin_lock(&ls->ls_recover_spin);
1013 start_gen = ls->ls_recover_start;
1014 block_gen = ls->ls_recover_block;
1015
1016 if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) ||
1017 !test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
1018 !test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
1019 /* sanity check, should not happen */
1020 fs_err(sdp, "control_first_done start %u block %u flags %lx\n",
1021 start_gen, block_gen, ls->ls_recover_flags);
1022 spin_unlock(&ls->ls_recover_spin);
1023 control_unlock(sdp);
1024 return -1;
1025 }
1026
1027 if (start_gen == block_gen) {
1028 /*
1029 * Wait for the end of a dlm recovery cycle to switch from
1030 * first mounter recovery. We can ignore any recover_slot
1031 * callbacks between the recover_prep and next recover_done
1032 * because we are still the first mounter and any failed nodes
1033 * have not fully mounted, so they don't need recovery.
1034 */
1035 spin_unlock(&ls->ls_recover_spin);
1036 fs_info(sdp, "control_first_done wait gen %u\n", start_gen);
1037
1038 wait_on_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY,
1039 TASK_UNINTERRUPTIBLE);
1040 goto restart;
1041 }
1042
1043 clear_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
1044 set_bit(DFL_FIRST_MOUNT_DONE, &ls->ls_recover_flags);
1045 memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
1046 memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
1047 spin_unlock(&ls->ls_recover_spin);
1048
1049 memset(ls->ls_lvb_bits, 0, GDLM_LVB_SIZE);
1050 control_lvb_write(ls, start_gen, ls->ls_lvb_bits);
1051
1052 error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT);
1053 if (error)
1054 fs_err(sdp, "control_first_done mounted PR error %d\n", error);
1055
1056 error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
1057 if (error)
1058 fs_err(sdp, "control_first_done control NL error %d\n", error);
1059
1060 return error;
1061 }
1062
1063 /*
1064 * Expand static jid arrays if necessary (by increments of RECOVER_SIZE_INC)
1065 * to accommodate the largest slot number. (NB dlm slot numbers start at 1,
1066 * gfs2 jids start at 0, so jid = slot - 1)
1067 */
1068
1069 #define RECOVER_SIZE_INC 16
1070
set_recover_size(struct gfs2_sbd * sdp,struct dlm_slot * slots,int num_slots)1071 static int set_recover_size(struct gfs2_sbd *sdp, struct dlm_slot *slots,
1072 int num_slots)
1073 {
1074 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1075 uint32_t *submit = NULL;
1076 uint32_t *result = NULL;
1077 uint32_t old_size, new_size;
1078 int i, max_jid;
1079
1080 if (!ls->ls_lvb_bits) {
1081 ls->ls_lvb_bits = kzalloc(GDLM_LVB_SIZE, GFP_NOFS);
1082 if (!ls->ls_lvb_bits)
1083 return -ENOMEM;
1084 }
1085
1086 max_jid = 0;
1087 for (i = 0; i < num_slots; i++) {
1088 if (max_jid < slots[i].slot - 1)
1089 max_jid = slots[i].slot - 1;
1090 }
1091
1092 old_size = ls->ls_recover_size;
1093 new_size = old_size;
1094 while (new_size < max_jid + 1)
1095 new_size += RECOVER_SIZE_INC;
1096 if (new_size == old_size)
1097 return 0;
1098
1099 submit = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
1100 result = kcalloc(new_size, sizeof(uint32_t), GFP_NOFS);
1101 if (!submit || !result) {
1102 kfree(submit);
1103 kfree(result);
1104 return -ENOMEM;
1105 }
1106
1107 spin_lock(&ls->ls_recover_spin);
1108 memcpy(submit, ls->ls_recover_submit, old_size * sizeof(uint32_t));
1109 memcpy(result, ls->ls_recover_result, old_size * sizeof(uint32_t));
1110 kfree(ls->ls_recover_submit);
1111 kfree(ls->ls_recover_result);
1112 ls->ls_recover_submit = submit;
1113 ls->ls_recover_result = result;
1114 ls->ls_recover_size = new_size;
1115 spin_unlock(&ls->ls_recover_spin);
1116 return 0;
1117 }
1118
free_recover_size(struct lm_lockstruct * ls)1119 static void free_recover_size(struct lm_lockstruct *ls)
1120 {
1121 kfree(ls->ls_lvb_bits);
1122 kfree(ls->ls_recover_submit);
1123 kfree(ls->ls_recover_result);
1124 ls->ls_recover_submit = NULL;
1125 ls->ls_recover_result = NULL;
1126 ls->ls_recover_size = 0;
1127 ls->ls_lvb_bits = NULL;
1128 }
1129
1130 /* dlm calls before it does lock recovery */
1131
gdlm_recover_prep(void * arg)1132 static void gdlm_recover_prep(void *arg)
1133 {
1134 struct gfs2_sbd *sdp = arg;
1135 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1136
1137 if (gfs2_withdrawing_or_withdrawn(sdp)) {
1138 fs_err(sdp, "recover_prep ignored due to withdraw.\n");
1139 return;
1140 }
1141 spin_lock(&ls->ls_recover_spin);
1142 ls->ls_recover_block = ls->ls_recover_start;
1143 set_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags);
1144
1145 if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
1146 test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
1147 spin_unlock(&ls->ls_recover_spin);
1148 return;
1149 }
1150 set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
1151 spin_unlock(&ls->ls_recover_spin);
1152 }
1153
1154 /* dlm calls after recover_prep has been completed on all lockspace members;
1155 identifies slot/jid of failed member */
1156
gdlm_recover_slot(void * arg,struct dlm_slot * slot)1157 static void gdlm_recover_slot(void *arg, struct dlm_slot *slot)
1158 {
1159 struct gfs2_sbd *sdp = arg;
1160 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1161 int jid = slot->slot - 1;
1162
1163 if (gfs2_withdrawing_or_withdrawn(sdp)) {
1164 fs_err(sdp, "recover_slot jid %d ignored due to withdraw.\n",
1165 jid);
1166 return;
1167 }
1168 spin_lock(&ls->ls_recover_spin);
1169 if (ls->ls_recover_size < jid + 1) {
1170 fs_err(sdp, "recover_slot jid %d gen %u short size %d\n",
1171 jid, ls->ls_recover_block, ls->ls_recover_size);
1172 spin_unlock(&ls->ls_recover_spin);
1173 return;
1174 }
1175
1176 if (ls->ls_recover_submit[jid]) {
1177 fs_info(sdp, "recover_slot jid %d gen %u prev %u\n",
1178 jid, ls->ls_recover_block, ls->ls_recover_submit[jid]);
1179 }
1180 ls->ls_recover_submit[jid] = ls->ls_recover_block;
1181 spin_unlock(&ls->ls_recover_spin);
1182 }
1183
1184 /* dlm calls after recover_slot and after it completes lock recovery */
1185
gdlm_recover_done(void * arg,struct dlm_slot * slots,int num_slots,int our_slot,uint32_t generation)1186 static void gdlm_recover_done(void *arg, struct dlm_slot *slots, int num_slots,
1187 int our_slot, uint32_t generation)
1188 {
1189 struct gfs2_sbd *sdp = arg;
1190 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1191
1192 if (gfs2_withdrawing_or_withdrawn(sdp)) {
1193 fs_err(sdp, "recover_done ignored due to withdraw.\n");
1194 return;
1195 }
1196 /* ensure the ls jid arrays are large enough */
1197 set_recover_size(sdp, slots, num_slots);
1198
1199 spin_lock(&ls->ls_recover_spin);
1200 ls->ls_recover_start = generation;
1201
1202 if (!ls->ls_recover_mount) {
1203 ls->ls_recover_mount = generation;
1204 ls->ls_jid = our_slot - 1;
1205 }
1206
1207 if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
1208 queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work, 0);
1209
1210 clear_bit(DFL_DLM_RECOVERY, &ls->ls_recover_flags);
1211 smp_mb__after_atomic();
1212 wake_up_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY);
1213 spin_unlock(&ls->ls_recover_spin);
1214 }
1215
1216 /* gfs2_recover thread has a journal recovery result */
1217
gdlm_recovery_result(struct gfs2_sbd * sdp,unsigned int jid,unsigned int result)1218 static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid,
1219 unsigned int result)
1220 {
1221 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1222
1223 if (gfs2_withdrawing_or_withdrawn(sdp)) {
1224 fs_err(sdp, "recovery_result jid %d ignored due to withdraw.\n",
1225 jid);
1226 return;
1227 }
1228 if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1229 return;
1230
1231 /* don't care about the recovery of own journal during mount */
1232 if (jid == ls->ls_jid)
1233 return;
1234
1235 spin_lock(&ls->ls_recover_spin);
1236 if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
1237 spin_unlock(&ls->ls_recover_spin);
1238 return;
1239 }
1240 if (ls->ls_recover_size < jid + 1) {
1241 fs_err(sdp, "recovery_result jid %d short size %d\n",
1242 jid, ls->ls_recover_size);
1243 spin_unlock(&ls->ls_recover_spin);
1244 return;
1245 }
1246
1247 fs_info(sdp, "recover jid %d result %s\n", jid,
1248 result == LM_RD_GAVEUP ? "busy" : "success");
1249
1250 ls->ls_recover_result[jid] = result;
1251
1252 /* GAVEUP means another node is recovering the journal; delay our
1253 next attempt to recover it, to give the other node a chance to
1254 finish before trying again */
1255
1256 if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
1257 queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work,
1258 result == LM_RD_GAVEUP ? HZ : 0);
1259 spin_unlock(&ls->ls_recover_spin);
1260 }
1261
1262 static const struct dlm_lockspace_ops gdlm_lockspace_ops = {
1263 .recover_prep = gdlm_recover_prep,
1264 .recover_slot = gdlm_recover_slot,
1265 .recover_done = gdlm_recover_done,
1266 };
1267
gdlm_mount(struct gfs2_sbd * sdp,const char * table)1268 static int gdlm_mount(struct gfs2_sbd *sdp, const char *table)
1269 {
1270 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1271 char cluster[GFS2_LOCKNAME_LEN];
1272 const char *fsname;
1273 uint32_t flags;
1274 int error, ops_result;
1275
1276 /*
1277 * initialize everything
1278 */
1279
1280 INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func);
1281 spin_lock_init(&ls->ls_recover_spin);
1282 ls->ls_recover_flags = 0;
1283 ls->ls_recover_mount = 0;
1284 ls->ls_recover_start = 0;
1285 ls->ls_recover_block = 0;
1286 ls->ls_recover_size = 0;
1287 ls->ls_recover_submit = NULL;
1288 ls->ls_recover_result = NULL;
1289 ls->ls_lvb_bits = NULL;
1290
1291 error = set_recover_size(sdp, NULL, 0);
1292 if (error)
1293 goto fail;
1294
1295 /*
1296 * prepare dlm_new_lockspace args
1297 */
1298
1299 fsname = strchr(table, ':');
1300 if (!fsname) {
1301 fs_info(sdp, "no fsname found\n");
1302 error = -EINVAL;
1303 goto fail_free;
1304 }
1305 memset(cluster, 0, sizeof(cluster));
1306 memcpy(cluster, table, strlen(table) - strlen(fsname));
1307 fsname++;
1308
1309 flags = DLM_LSFL_NEWEXCL;
1310
1311 /*
1312 * create/join lockspace
1313 */
1314
1315 error = dlm_new_lockspace(fsname, cluster, flags, GDLM_LVB_SIZE,
1316 &gdlm_lockspace_ops, sdp, &ops_result,
1317 &ls->ls_dlm);
1318 if (error) {
1319 fs_err(sdp, "dlm_new_lockspace error %d\n", error);
1320 goto fail_free;
1321 }
1322
1323 if (ops_result < 0) {
1324 /*
1325 * dlm does not support ops callbacks,
1326 * old dlm_controld/gfs_controld are used, try without ops.
1327 */
1328 fs_info(sdp, "dlm lockspace ops not used\n");
1329 free_recover_size(ls);
1330 set_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags);
1331 return 0;
1332 }
1333
1334 if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) {
1335 fs_err(sdp, "dlm lockspace ops disallow jid preset\n");
1336 error = -EINVAL;
1337 goto fail_release;
1338 }
1339
1340 /*
1341 * control_mount() uses control_lock to determine first mounter,
1342 * and for later mounts, waits for any recoveries to be cleared.
1343 */
1344
1345 error = control_mount(sdp);
1346 if (error) {
1347 fs_err(sdp, "mount control error %d\n", error);
1348 goto fail_release;
1349 }
1350
1351 ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
1352 clear_bit(SDF_NOJOURNALID, &sdp->sd_flags);
1353 smp_mb__after_atomic();
1354 wake_up_bit(&sdp->sd_flags, SDF_NOJOURNALID);
1355 return 0;
1356
1357 fail_release:
1358 dlm_release_lockspace(ls->ls_dlm, 2);
1359 fail_free:
1360 free_recover_size(ls);
1361 fail:
1362 return error;
1363 }
1364
gdlm_first_done(struct gfs2_sbd * sdp)1365 static void gdlm_first_done(struct gfs2_sbd *sdp)
1366 {
1367 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1368 int error;
1369
1370 if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1371 return;
1372
1373 error = control_first_done(sdp);
1374 if (error)
1375 fs_err(sdp, "mount first_done error %d\n", error);
1376 }
1377
gdlm_unmount(struct gfs2_sbd * sdp)1378 static void gdlm_unmount(struct gfs2_sbd *sdp)
1379 {
1380 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1381
1382 if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
1383 goto release;
1384
1385 /* wait for gfs2_control_wq to be done with this mount */
1386
1387 spin_lock(&ls->ls_recover_spin);
1388 set_bit(DFL_UNMOUNT, &ls->ls_recover_flags);
1389 spin_unlock(&ls->ls_recover_spin);
1390 flush_delayed_work(&sdp->sd_control_work);
1391
1392 /* mounted_lock and control_lock will be purged in dlm recovery */
1393 release:
1394 if (ls->ls_dlm) {
1395 dlm_release_lockspace(ls->ls_dlm, 2);
1396 ls->ls_dlm = NULL;
1397 }
1398
1399 free_recover_size(ls);
1400 }
1401
1402 static const match_table_t dlm_tokens = {
1403 { Opt_jid, "jid=%d"},
1404 { Opt_id, "id=%d"},
1405 { Opt_first, "first=%d"},
1406 { Opt_nodir, "nodir=%d"},
1407 { Opt_err, NULL },
1408 };
1409
1410 const struct lm_lockops gfs2_dlm_ops = {
1411 .lm_proto_name = "lock_dlm",
1412 .lm_mount = gdlm_mount,
1413 .lm_first_done = gdlm_first_done,
1414 .lm_recovery_result = gdlm_recovery_result,
1415 .lm_unmount = gdlm_unmount,
1416 .lm_put_lock = gdlm_put_lock,
1417 .lm_lock = gdlm_lock,
1418 .lm_cancel = gdlm_cancel,
1419 .lm_tokens = &dlm_tokens,
1420 };
1421
1422