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 | */ |
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 | */ |
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(s: &gl->gl_stats, index, sample: rtt); /* Local */ |
87 | gfs2_update_stats(s: &lks->lkstats[gltype], index, sample: rtt); /* Global */ |
88 | preempt_enable(); |
89 | |
90 | trace_gfs2_glock_lock_time(gl, tdiff: 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 | |
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(s: &gl->gl_stats, index: GFS2_LKS_SIRT, sample: irt); /* Local */ |
115 | gfs2_update_stats(s: &lks->lkstats[gltype], index: GFS2_LKS_SIRT, sample: irt); /* Global */ |
116 | preempt_enable(); |
117 | } |
118 | |
119 | static void gdlm_ast(void *arg) |
120 | { |
121 | struct gfs2_glock *gl = arg; |
122 | unsigned ret = gl->gl_state; |
123 | |
124 | gfs2_update_reply_times(gl); |
125 | BUG_ON(gl->gl_lksb.sb_flags & DLM_SBF_DEMOTED); |
126 | |
127 | if ((gl->gl_lksb.sb_flags & DLM_SBF_VALNOTVALID) && gl->gl_lksb.sb_lvbptr) |
128 | memset(gl->gl_lksb.sb_lvbptr, 0, GDLM_LVB_SIZE); |
129 | |
130 | switch (gl->gl_lksb.sb_status) { |
131 | case -DLM_EUNLOCK: /* Unlocked, so glock can be freed */ |
132 | if (gl->gl_ops->go_free) |
133 | gl->gl_ops->go_free(gl); |
134 | gfs2_glock_free(gl); |
135 | return; |
136 | case -DLM_ECANCEL: /* Cancel while getting lock */ |
137 | ret |= LM_OUT_CANCELED; |
138 | goto out; |
139 | case -EAGAIN: /* Try lock fails */ |
140 | case -EDEADLK: /* Deadlock detected */ |
141 | goto out; |
142 | case -ETIMEDOUT: /* Canceled due to timeout */ |
143 | ret |= LM_OUT_ERROR; |
144 | goto out; |
145 | case 0: /* Success */ |
146 | break; |
147 | default: /* Something unexpected */ |
148 | BUG(); |
149 | } |
150 | |
151 | ret = gl->gl_req; |
152 | if (gl->gl_lksb.sb_flags & DLM_SBF_ALTMODE) { |
153 | if (gl->gl_req == LM_ST_SHARED) |
154 | ret = LM_ST_DEFERRED; |
155 | else if (gl->gl_req == LM_ST_DEFERRED) |
156 | ret = LM_ST_SHARED; |
157 | else |
158 | BUG(); |
159 | } |
160 | |
161 | set_bit(nr: GLF_INITIAL, addr: &gl->gl_flags); |
162 | gfs2_glock_complete(gl, ret); |
163 | return; |
164 | out: |
165 | if (!test_bit(GLF_INITIAL, &gl->gl_flags)) |
166 | gl->gl_lksb.sb_lkid = 0; |
167 | gfs2_glock_complete(gl, ret); |
168 | } |
169 | |
170 | static void gdlm_bast(void *arg, int mode) |
171 | { |
172 | struct gfs2_glock *gl = arg; |
173 | |
174 | switch (mode) { |
175 | case DLM_LOCK_EX: |
176 | gfs2_glock_cb(gl, LM_ST_UNLOCKED); |
177 | break; |
178 | case DLM_LOCK_CW: |
179 | gfs2_glock_cb(gl, LM_ST_DEFERRED); |
180 | break; |
181 | case DLM_LOCK_PR: |
182 | gfs2_glock_cb(gl, LM_ST_SHARED); |
183 | break; |
184 | default: |
185 | fs_err(gl->gl_name.ln_sbd, "unknown bast mode %d\n" , mode); |
186 | BUG(); |
187 | } |
188 | } |
189 | |
190 | /* convert gfs lock-state to dlm lock-mode */ |
191 | |
192 | static int make_mode(struct gfs2_sbd *sdp, const unsigned int lmstate) |
193 | { |
194 | switch (lmstate) { |
195 | case LM_ST_UNLOCKED: |
196 | return DLM_LOCK_NL; |
197 | case LM_ST_EXCLUSIVE: |
198 | return DLM_LOCK_EX; |
199 | case LM_ST_DEFERRED: |
200 | return DLM_LOCK_CW; |
201 | case LM_ST_SHARED: |
202 | return DLM_LOCK_PR; |
203 | } |
204 | fs_err(sdp, "unknown LM state %d\n" , lmstate); |
205 | BUG(); |
206 | return -1; |
207 | } |
208 | |
209 | static u32 make_flags(struct gfs2_glock *gl, const unsigned int gfs_flags, |
210 | const int req) |
211 | { |
212 | u32 lkf = 0; |
213 | |
214 | if (gl->gl_lksb.sb_lvbptr) |
215 | lkf |= DLM_LKF_VALBLK; |
216 | |
217 | if (gfs_flags & LM_FLAG_TRY) |
218 | lkf |= DLM_LKF_NOQUEUE; |
219 | |
220 | if (gfs_flags & LM_FLAG_TRY_1CB) { |
221 | lkf |= DLM_LKF_NOQUEUE; |
222 | lkf |= DLM_LKF_NOQUEUEBAST; |
223 | } |
224 | |
225 | if (gfs_flags & LM_FLAG_ANY) { |
226 | if (req == DLM_LOCK_PR) |
227 | lkf |= DLM_LKF_ALTCW; |
228 | else if (req == DLM_LOCK_CW) |
229 | lkf |= DLM_LKF_ALTPR; |
230 | else |
231 | BUG(); |
232 | } |
233 | |
234 | if (gl->gl_lksb.sb_lkid != 0) { |
235 | lkf |= DLM_LKF_CONVERT; |
236 | if (test_bit(GLF_BLOCKING, &gl->gl_flags)) |
237 | lkf |= DLM_LKF_QUECVT; |
238 | } |
239 | |
240 | return lkf; |
241 | } |
242 | |
243 | static void gfs2_reverse_hex(char *c, u64 value) |
244 | { |
245 | *c = '0'; |
246 | while (value) { |
247 | *c-- = hex_asc[value & 0x0f]; |
248 | value >>= 4; |
249 | } |
250 | } |
251 | |
252 | static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state, |
253 | unsigned int flags) |
254 | { |
255 | struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; |
256 | int req; |
257 | u32 lkf; |
258 | char strname[GDLM_STRNAME_BYTES] = "" ; |
259 | int error; |
260 | |
261 | req = make_mode(sdp: gl->gl_name.ln_sbd, lmstate: req_state); |
262 | lkf = make_flags(gl, gfs_flags: flags, req); |
263 | gfs2_glstats_inc(gl, which: GFS2_LKS_DCOUNT); |
264 | gfs2_sbstats_inc(gl, which: GFS2_LKS_DCOUNT); |
265 | if (gl->gl_lksb.sb_lkid) { |
266 | gfs2_update_request_times(gl); |
267 | } else { |
268 | memset(strname, ' ', GDLM_STRNAME_BYTES - 1); |
269 | strname[GDLM_STRNAME_BYTES - 1] = '\0'; |
270 | gfs2_reverse_hex(c: strname + 7, value: gl->gl_name.ln_type); |
271 | gfs2_reverse_hex(c: strname + 23, value: gl->gl_name.ln_number); |
272 | gl->gl_dstamp = ktime_get_real(); |
273 | } |
274 | /* |
275 | * Submit the actual lock request. |
276 | */ |
277 | |
278 | again: |
279 | error = dlm_lock(lockspace: ls->ls_dlm, mode: req, lksb: &gl->gl_lksb, flags: lkf, name: strname, |
280 | GDLM_STRNAME_BYTES - 1, parent_lkid: 0, lockast: gdlm_ast, astarg: gl, bast: gdlm_bast); |
281 | if (error == -EBUSY) { |
282 | msleep(msecs: 20); |
283 | goto again; |
284 | } |
285 | return error; |
286 | } |
287 | |
288 | static void gdlm_put_lock(struct gfs2_glock *gl) |
289 | { |
290 | struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
291 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
292 | int error; |
293 | |
294 | if (gl->gl_lksb.sb_lkid == 0) |
295 | goto out_free; |
296 | |
297 | clear_bit(nr: GLF_BLOCKING, addr: &gl->gl_flags); |
298 | gfs2_glstats_inc(gl, which: GFS2_LKS_DCOUNT); |
299 | gfs2_sbstats_inc(gl, which: GFS2_LKS_DCOUNT); |
300 | gfs2_update_request_times(gl); |
301 | |
302 | /* don't want to call dlm if we've unmounted the lock protocol */ |
303 | if (test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) |
304 | goto out_free; |
305 | /* don't want to skip dlm_unlock writing the lvb when lock has one */ |
306 | |
307 | if (test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags) && |
308 | !gl->gl_lksb.sb_lvbptr) |
309 | goto out_free; |
310 | |
311 | again: |
312 | error = dlm_unlock(lockspace: ls->ls_dlm, lkid: gl->gl_lksb.sb_lkid, DLM_LKF_VALBLK, |
313 | NULL, astarg: gl); |
314 | if (error == -EBUSY) { |
315 | msleep(msecs: 20); |
316 | goto again; |
317 | } |
318 | |
319 | if (error) { |
320 | fs_err(sdp, "gdlm_unlock %x,%llx err=%d\n" , |
321 | gl->gl_name.ln_type, |
322 | (unsigned long long)gl->gl_name.ln_number, error); |
323 | } |
324 | return; |
325 | |
326 | out_free: |
327 | gfs2_glock_free(gl); |
328 | } |
329 | |
330 | static void gdlm_cancel(struct gfs2_glock *gl) |
331 | { |
332 | struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; |
333 | dlm_unlock(lockspace: ls->ls_dlm, lkid: gl->gl_lksb.sb_lkid, DLM_LKF_CANCEL, NULL, astarg: gl); |
334 | } |
335 | |
336 | /* |
337 | * dlm/gfs2 recovery coordination using dlm_recover callbacks |
338 | * |
339 | * 0. gfs2 checks for another cluster node withdraw, needing journal replay |
340 | * 1. dlm_controld sees lockspace members change |
341 | * 2. dlm_controld blocks dlm-kernel locking activity |
342 | * 3. dlm_controld within dlm-kernel notifies gfs2 (recover_prep) |
343 | * 4. dlm_controld starts and finishes its own user level recovery |
344 | * 5. dlm_controld starts dlm-kernel dlm_recoverd to do kernel recovery |
345 | * 6. dlm_recoverd notifies gfs2 of failed nodes (recover_slot) |
346 | * 7. dlm_recoverd does its own lock recovery |
347 | * 8. dlm_recoverd unblocks dlm-kernel locking activity |
348 | * 9. dlm_recoverd notifies gfs2 when done (recover_done with new generation) |
349 | * 10. gfs2_control updates control_lock lvb with new generation and jid bits |
350 | * 11. gfs2_control enqueues journals for gfs2_recover to recover (maybe none) |
351 | * 12. gfs2_recover dequeues and recovers journals of failed nodes |
352 | * 13. gfs2_recover provides recovery results to gfs2_control (recovery_result) |
353 | * 14. gfs2_control updates control_lock lvb jid bits for recovered journals |
354 | * 15. gfs2_control unblocks normal locking when all journals are recovered |
355 | * |
356 | * - failures during recovery |
357 | * |
358 | * recover_prep() may set BLOCK_LOCKS (step 3) again before gfs2_control |
359 | * clears BLOCK_LOCKS (step 15), e.g. another node fails while still |
360 | * recovering for a prior failure. gfs2_control needs a way to detect |
361 | * this so it can leave BLOCK_LOCKS set in step 15. This is managed using |
362 | * the recover_block and recover_start values. |
363 | * |
364 | * recover_done() provides a new lockspace generation number each time it |
365 | * is called (step 9). This generation number is saved as recover_start. |
366 | * When recover_prep() is called, it sets BLOCK_LOCKS and sets |
367 | * recover_block = recover_start. So, while recover_block is equal to |
368 | * recover_start, BLOCK_LOCKS should remain set. (recover_spin must |
369 | * be held around the BLOCK_LOCKS/recover_block/recover_start logic.) |
370 | * |
371 | * - more specific gfs2 steps in sequence above |
372 | * |
373 | * 3. recover_prep sets BLOCK_LOCKS and sets recover_block = recover_start |
374 | * 6. recover_slot records any failed jids (maybe none) |
375 | * 9. recover_done sets recover_start = new generation number |
376 | * 10. gfs2_control sets control_lock lvb = new gen + bits for failed jids |
377 | * 12. gfs2_recover does journal recoveries for failed jids identified above |
378 | * 14. gfs2_control clears control_lock lvb bits for recovered jids |
379 | * 15. gfs2_control checks if recover_block == recover_start (step 3 occured |
380 | * again) then do nothing, otherwise if recover_start > recover_block |
381 | * then clear BLOCK_LOCKS. |
382 | * |
383 | * - parallel recovery steps across all nodes |
384 | * |
385 | * All nodes attempt to update the control_lock lvb with the new generation |
386 | * number and jid bits, but only the first to get the control_lock EX will |
387 | * do so; others will see that it's already done (lvb already contains new |
388 | * generation number.) |
389 | * |
390 | * . All nodes get the same recover_prep/recover_slot/recover_done callbacks |
391 | * . All nodes attempt to set control_lock lvb gen + bits for the new gen |
392 | * . One node gets control_lock first and writes the lvb, others see it's done |
393 | * . All nodes attempt to recover jids for which they see control_lock bits set |
394 | * . One node succeeds for a jid, and that one clears the jid bit in the lvb |
395 | * . All nodes will eventually see all lvb bits clear and unblock locks |
396 | * |
397 | * - is there a problem with clearing an lvb bit that should be set |
398 | * and missing a journal recovery? |
399 | * |
400 | * 1. jid fails |
401 | * 2. lvb bit set for step 1 |
402 | * 3. jid recovered for step 1 |
403 | * 4. jid taken again (new mount) |
404 | * 5. jid fails (for step 4) |
405 | * 6. lvb bit set for step 5 (will already be set) |
406 | * 7. lvb bit cleared for step 3 |
407 | * |
408 | * This is not a problem because the failure in step 5 does not |
409 | * require recovery, because the mount in step 4 could not have |
410 | * progressed far enough to unblock locks and access the fs. The |
411 | * control_mount() function waits for all recoveries to be complete |
412 | * for the latest lockspace generation before ever unblocking locks |
413 | * and returning. The mount in step 4 waits until the recovery in |
414 | * step 1 is done. |
415 | * |
416 | * - special case of first mounter: first node to mount the fs |
417 | * |
418 | * The first node to mount a gfs2 fs needs to check all the journals |
419 | * and recover any that need recovery before other nodes are allowed |
420 | * to mount the fs. (Others may begin mounting, but they must wait |
421 | * for the first mounter to be done before taking locks on the fs |
422 | * or accessing the fs.) This has two parts: |
423 | * |
424 | * 1. The mounted_lock tells a node it's the first to mount the fs. |
425 | * Each node holds the mounted_lock in PR while it's mounted. |
426 | * Each node tries to acquire the mounted_lock in EX when it mounts. |
427 | * If a node is granted the mounted_lock EX it means there are no |
428 | * other mounted nodes (no PR locks exist), and it is the first mounter. |
429 | * The mounted_lock is demoted to PR when first recovery is done, so |
430 | * others will fail to get an EX lock, but will get a PR lock. |
431 | * |
432 | * 2. The control_lock blocks others in control_mount() while the first |
433 | * mounter is doing first mount recovery of all journals. |
434 | * A mounting node needs to acquire control_lock in EX mode before |
435 | * it can proceed. The first mounter holds control_lock in EX while doing |
436 | * the first mount recovery, blocking mounts from other nodes, then demotes |
437 | * control_lock to NL when it's done (others_may_mount/first_done), |
438 | * allowing other nodes to continue mounting. |
439 | * |
440 | * first mounter: |
441 | * control_lock EX/NOQUEUE success |
442 | * mounted_lock EX/NOQUEUE success (no other PR, so no other mounters) |
443 | * set first=1 |
444 | * do first mounter recovery |
445 | * mounted_lock EX->PR |
446 | * control_lock EX->NL, write lvb generation |
447 | * |
448 | * other mounter: |
449 | * control_lock EX/NOQUEUE success (if fail -EAGAIN, retry) |
450 | * mounted_lock EX/NOQUEUE fail -EAGAIN (expected due to other mounters PR) |
451 | * mounted_lock PR/NOQUEUE success |
452 | * read lvb generation |
453 | * control_lock EX->NL |
454 | * set first=0 |
455 | * |
456 | * - mount during recovery |
457 | * |
458 | * If a node mounts while others are doing recovery (not first mounter), |
459 | * the mounting node will get its initial recover_done() callback without |
460 | * having seen any previous failures/callbacks. |
461 | * |
462 | * It must wait for all recoveries preceding its mount to be finished |
463 | * before it unblocks locks. It does this by repeating the "other mounter" |
464 | * steps above until the lvb generation number is >= its mount generation |
465 | * number (from initial recover_done) and all lvb bits are clear. |
466 | * |
467 | * - control_lock lvb format |
468 | * |
469 | * 4 bytes generation number: the latest dlm lockspace generation number |
470 | * from recover_done callback. Indicates the jid bitmap has been updated |
471 | * to reflect all slot failures through that generation. |
472 | * 4 bytes unused. |
473 | * GDLM_LVB_SIZE-8 bytes of jid bit map. If bit N is set, it indicates |
474 | * that jid N needs recovery. |
475 | */ |
476 | |
477 | #define JID_BITMAP_OFFSET 8 /* 4 byte generation number + 4 byte unused */ |
478 | |
479 | static void control_lvb_read(struct lm_lockstruct *ls, uint32_t *lvb_gen, |
480 | char *lvb_bits) |
481 | { |
482 | __le32 gen; |
483 | memcpy(lvb_bits, ls->ls_control_lvb, GDLM_LVB_SIZE); |
484 | memcpy(&gen, lvb_bits, sizeof(__le32)); |
485 | *lvb_gen = le32_to_cpu(gen); |
486 | } |
487 | |
488 | static void control_lvb_write(struct lm_lockstruct *ls, uint32_t lvb_gen, |
489 | char *lvb_bits) |
490 | { |
491 | __le32 gen; |
492 | memcpy(ls->ls_control_lvb, lvb_bits, GDLM_LVB_SIZE); |
493 | gen = cpu_to_le32(lvb_gen); |
494 | memcpy(ls->ls_control_lvb, &gen, sizeof(__le32)); |
495 | } |
496 | |
497 | static int all_jid_bits_clear(char *lvb) |
498 | { |
499 | return !memchr_inv(p: lvb + JID_BITMAP_OFFSET, c: 0, |
500 | GDLM_LVB_SIZE - JID_BITMAP_OFFSET); |
501 | } |
502 | |
503 | static void sync_wait_cb(void *arg) |
504 | { |
505 | struct lm_lockstruct *ls = arg; |
506 | complete(&ls->ls_sync_wait); |
507 | } |
508 | |
509 | static int sync_unlock(struct gfs2_sbd *sdp, struct dlm_lksb *lksb, char *name) |
510 | { |
511 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
512 | int error; |
513 | |
514 | error = dlm_unlock(lockspace: ls->ls_dlm, lkid: lksb->sb_lkid, flags: 0, lksb, astarg: ls); |
515 | if (error) { |
516 | fs_err(sdp, "%s lkid %x error %d\n" , |
517 | name, lksb->sb_lkid, error); |
518 | return error; |
519 | } |
520 | |
521 | wait_for_completion(&ls->ls_sync_wait); |
522 | |
523 | if (lksb->sb_status != -DLM_EUNLOCK) { |
524 | fs_err(sdp, "%s lkid %x status %d\n" , |
525 | name, lksb->sb_lkid, lksb->sb_status); |
526 | return -1; |
527 | } |
528 | return 0; |
529 | } |
530 | |
531 | static int sync_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags, |
532 | unsigned int num, struct dlm_lksb *lksb, char *name) |
533 | { |
534 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
535 | char strname[GDLM_STRNAME_BYTES]; |
536 | int error, status; |
537 | |
538 | memset(strname, 0, GDLM_STRNAME_BYTES); |
539 | snprintf(buf: strname, GDLM_STRNAME_BYTES, fmt: "%8x%16x" , LM_TYPE_NONDISK, num); |
540 | |
541 | error = dlm_lock(lockspace: ls->ls_dlm, mode, lksb, flags, |
542 | name: strname, GDLM_STRNAME_BYTES - 1, |
543 | parent_lkid: 0, lockast: sync_wait_cb, astarg: ls, NULL); |
544 | if (error) { |
545 | fs_err(sdp, "%s lkid %x flags %x mode %d error %d\n" , |
546 | name, lksb->sb_lkid, flags, mode, error); |
547 | return error; |
548 | } |
549 | |
550 | wait_for_completion(&ls->ls_sync_wait); |
551 | |
552 | status = lksb->sb_status; |
553 | |
554 | if (status && status != -EAGAIN) { |
555 | fs_err(sdp, "%s lkid %x flags %x mode %d status %d\n" , |
556 | name, lksb->sb_lkid, flags, mode, status); |
557 | } |
558 | |
559 | return status; |
560 | } |
561 | |
562 | static int mounted_unlock(struct gfs2_sbd *sdp) |
563 | { |
564 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
565 | return sync_unlock(sdp, lksb: &ls->ls_mounted_lksb, name: "mounted_lock" ); |
566 | } |
567 | |
568 | static int mounted_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags) |
569 | { |
570 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
571 | return sync_lock(sdp, mode, flags, GFS2_MOUNTED_LOCK, |
572 | lksb: &ls->ls_mounted_lksb, name: "mounted_lock" ); |
573 | } |
574 | |
575 | static int control_unlock(struct gfs2_sbd *sdp) |
576 | { |
577 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
578 | return sync_unlock(sdp, lksb: &ls->ls_control_lksb, name: "control_lock" ); |
579 | } |
580 | |
581 | static int control_lock(struct gfs2_sbd *sdp, int mode, uint32_t flags) |
582 | { |
583 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
584 | return sync_lock(sdp, mode, flags, GFS2_CONTROL_LOCK, |
585 | lksb: &ls->ls_control_lksb, name: "control_lock" ); |
586 | } |
587 | |
588 | /** |
589 | * remote_withdraw - react to a node withdrawing from the file system |
590 | * @sdp: The superblock |
591 | */ |
592 | static void remote_withdraw(struct gfs2_sbd *sdp) |
593 | { |
594 | struct gfs2_jdesc *jd; |
595 | int ret = 0, count = 0; |
596 | |
597 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { |
598 | if (jd->jd_jid == sdp->sd_lockstruct.ls_jid) |
599 | continue; |
600 | ret = gfs2_recover_journal(gfs2_jd: jd, wait: true); |
601 | if (ret) |
602 | break; |
603 | count++; |
604 | } |
605 | |
606 | /* Now drop the additional reference we acquired */ |
607 | fs_err(sdp, "Journals checked: %d, ret = %d.\n" , count, ret); |
608 | } |
609 | |
610 | static void gfs2_control_func(struct work_struct *work) |
611 | { |
612 | struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work); |
613 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
614 | uint32_t block_gen, start_gen, lvb_gen, flags; |
615 | int recover_set = 0; |
616 | int write_lvb = 0; |
617 | int recover_size; |
618 | int i, error; |
619 | |
620 | /* First check for other nodes that may have done a withdraw. */ |
621 | if (test_bit(SDF_REMOTE_WITHDRAW, &sdp->sd_flags)) { |
622 | remote_withdraw(sdp); |
623 | clear_bit(nr: SDF_REMOTE_WITHDRAW, addr: &sdp->sd_flags); |
624 | return; |
625 | } |
626 | |
627 | spin_lock(lock: &ls->ls_recover_spin); |
628 | /* |
629 | * No MOUNT_DONE means we're still mounting; control_mount() |
630 | * will set this flag, after which this thread will take over |
631 | * all further clearing of BLOCK_LOCKS. |
632 | * |
633 | * FIRST_MOUNT means this node is doing first mounter recovery, |
634 | * for which recovery control is handled by |
635 | * control_mount()/control_first_done(), not this thread. |
636 | */ |
637 | if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || |
638 | test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { |
639 | spin_unlock(lock: &ls->ls_recover_spin); |
640 | return; |
641 | } |
642 | block_gen = ls->ls_recover_block; |
643 | start_gen = ls->ls_recover_start; |
644 | spin_unlock(lock: &ls->ls_recover_spin); |
645 | |
646 | /* |
647 | * Equal block_gen and start_gen implies we are between |
648 | * recover_prep and recover_done callbacks, which means |
649 | * dlm recovery is in progress and dlm locking is blocked. |
650 | * There's no point trying to do any work until recover_done. |
651 | */ |
652 | |
653 | if (block_gen == start_gen) |
654 | return; |
655 | |
656 | /* |
657 | * Propagate recover_submit[] and recover_result[] to lvb: |
658 | * dlm_recoverd adds to recover_submit[] jids needing recovery |
659 | * gfs2_recover adds to recover_result[] journal recovery results |
660 | * |
661 | * set lvb bit for jids in recover_submit[] if the lvb has not |
662 | * yet been updated for the generation of the failure |
663 | * |
664 | * clear lvb bit for jids in recover_result[] if the result of |
665 | * the journal recovery is SUCCESS |
666 | */ |
667 | |
668 | error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK); |
669 | if (error) { |
670 | fs_err(sdp, "control lock EX error %d\n" , error); |
671 | return; |
672 | } |
673 | |
674 | control_lvb_read(ls, lvb_gen: &lvb_gen, lvb_bits: ls->ls_lvb_bits); |
675 | |
676 | spin_lock(lock: &ls->ls_recover_spin); |
677 | if (block_gen != ls->ls_recover_block || |
678 | start_gen != ls->ls_recover_start) { |
679 | fs_info(sdp, "recover generation %u block1 %u %u\n" , |
680 | start_gen, block_gen, ls->ls_recover_block); |
681 | spin_unlock(lock: &ls->ls_recover_spin); |
682 | control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); |
683 | return; |
684 | } |
685 | |
686 | recover_size = ls->ls_recover_size; |
687 | |
688 | if (lvb_gen <= start_gen) { |
689 | /* |
690 | * Clear lvb bits for jids we've successfully recovered. |
691 | * Because all nodes attempt to recover failed journals, |
692 | * a journal can be recovered multiple times successfully |
693 | * in succession. Only the first will really do recovery, |
694 | * the others find it clean, but still report a successful |
695 | * recovery. So, another node may have already recovered |
696 | * the jid and cleared the lvb bit for it. |
697 | */ |
698 | for (i = 0; i < recover_size; i++) { |
699 | if (ls->ls_recover_result[i] != LM_RD_SUCCESS) |
700 | continue; |
701 | |
702 | ls->ls_recover_result[i] = 0; |
703 | |
704 | if (!test_bit_le(nr: i, addr: ls->ls_lvb_bits + JID_BITMAP_OFFSET)) |
705 | continue; |
706 | |
707 | __clear_bit_le(nr: i, addr: ls->ls_lvb_bits + JID_BITMAP_OFFSET); |
708 | write_lvb = 1; |
709 | } |
710 | } |
711 | |
712 | if (lvb_gen == start_gen) { |
713 | /* |
714 | * Failed slots before start_gen are already set in lvb. |
715 | */ |
716 | for (i = 0; i < recover_size; i++) { |
717 | if (!ls->ls_recover_submit[i]) |
718 | continue; |
719 | if (ls->ls_recover_submit[i] < lvb_gen) |
720 | ls->ls_recover_submit[i] = 0; |
721 | } |
722 | } else if (lvb_gen < start_gen) { |
723 | /* |
724 | * Failed slots before start_gen are not yet set in lvb. |
725 | */ |
726 | for (i = 0; i < recover_size; i++) { |
727 | if (!ls->ls_recover_submit[i]) |
728 | continue; |
729 | if (ls->ls_recover_submit[i] < start_gen) { |
730 | ls->ls_recover_submit[i] = 0; |
731 | __set_bit_le(nr: i, addr: ls->ls_lvb_bits + JID_BITMAP_OFFSET); |
732 | } |
733 | } |
734 | /* even if there are no bits to set, we need to write the |
735 | latest generation to the lvb */ |
736 | write_lvb = 1; |
737 | } else { |
738 | /* |
739 | * we should be getting a recover_done() for lvb_gen soon |
740 | */ |
741 | } |
742 | spin_unlock(lock: &ls->ls_recover_spin); |
743 | |
744 | if (write_lvb) { |
745 | control_lvb_write(ls, lvb_gen: start_gen, lvb_bits: ls->ls_lvb_bits); |
746 | flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK; |
747 | } else { |
748 | flags = DLM_LKF_CONVERT; |
749 | } |
750 | |
751 | error = control_lock(sdp, DLM_LOCK_NL, flags); |
752 | if (error) { |
753 | fs_err(sdp, "control lock NL error %d\n" , error); |
754 | return; |
755 | } |
756 | |
757 | /* |
758 | * Everyone will see jid bits set in the lvb, run gfs2_recover_set(), |
759 | * and clear a jid bit in the lvb if the recovery is a success. |
760 | * Eventually all journals will be recovered, all jid bits will |
761 | * be cleared in the lvb, and everyone will clear BLOCK_LOCKS. |
762 | */ |
763 | |
764 | for (i = 0; i < recover_size; i++) { |
765 | if (test_bit_le(nr: i, addr: ls->ls_lvb_bits + JID_BITMAP_OFFSET)) { |
766 | fs_info(sdp, "recover generation %u jid %d\n" , |
767 | start_gen, i); |
768 | gfs2_recover_set(sdp, jid: i); |
769 | recover_set++; |
770 | } |
771 | } |
772 | if (recover_set) |
773 | return; |
774 | |
775 | /* |
776 | * No more jid bits set in lvb, all recovery is done, unblock locks |
777 | * (unless a new recover_prep callback has occured blocking locks |
778 | * again while working above) |
779 | */ |
780 | |
781 | spin_lock(lock: &ls->ls_recover_spin); |
782 | if (ls->ls_recover_block == block_gen && |
783 | ls->ls_recover_start == start_gen) { |
784 | clear_bit(nr: DFL_BLOCK_LOCKS, addr: &ls->ls_recover_flags); |
785 | spin_unlock(lock: &ls->ls_recover_spin); |
786 | fs_info(sdp, "recover generation %u done\n" , start_gen); |
787 | gfs2_glock_thaw(sdp); |
788 | } else { |
789 | fs_info(sdp, "recover generation %u block2 %u %u\n" , |
790 | start_gen, block_gen, ls->ls_recover_block); |
791 | spin_unlock(lock: &ls->ls_recover_spin); |
792 | } |
793 | } |
794 | |
795 | static int control_mount(struct gfs2_sbd *sdp) |
796 | { |
797 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
798 | uint32_t start_gen, block_gen, mount_gen, lvb_gen; |
799 | int mounted_mode; |
800 | int retries = 0; |
801 | int error; |
802 | |
803 | memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb)); |
804 | memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb)); |
805 | memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE); |
806 | ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb; |
807 | init_completion(x: &ls->ls_sync_wait); |
808 | |
809 | set_bit(nr: DFL_BLOCK_LOCKS, addr: &ls->ls_recover_flags); |
810 | |
811 | error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK); |
812 | if (error) { |
813 | fs_err(sdp, "control_mount control_lock NL error %d\n" , error); |
814 | return error; |
815 | } |
816 | |
817 | error = mounted_lock(sdp, DLM_LOCK_NL, flags: 0); |
818 | if (error) { |
819 | fs_err(sdp, "control_mount mounted_lock NL error %d\n" , error); |
820 | control_unlock(sdp); |
821 | return error; |
822 | } |
823 | mounted_mode = DLM_LOCK_NL; |
824 | |
825 | restart: |
826 | if (retries++ && signal_pending(current)) { |
827 | error = -EINTR; |
828 | goto fail; |
829 | } |
830 | |
831 | /* |
832 | * We always start with both locks in NL. control_lock is |
833 | * demoted to NL below so we don't need to do it here. |
834 | */ |
835 | |
836 | if (mounted_mode != DLM_LOCK_NL) { |
837 | error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); |
838 | if (error) |
839 | goto fail; |
840 | mounted_mode = DLM_LOCK_NL; |
841 | } |
842 | |
843 | /* |
844 | * Other nodes need to do some work in dlm recovery and gfs2_control |
845 | * before the recover_done and control_lock will be ready for us below. |
846 | * A delay here is not required but often avoids having to retry. |
847 | */ |
848 | |
849 | msleep_interruptible(msecs: 500); |
850 | |
851 | /* |
852 | * Acquire control_lock in EX and mounted_lock in either EX or PR. |
853 | * control_lock lvb keeps track of any pending journal recoveries. |
854 | * mounted_lock indicates if any other nodes have the fs mounted. |
855 | */ |
856 | |
857 | error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK); |
858 | if (error == -EAGAIN) { |
859 | goto restart; |
860 | } else if (error) { |
861 | fs_err(sdp, "control_mount control_lock EX error %d\n" , error); |
862 | goto fail; |
863 | } |
864 | |
865 | /** |
866 | * If we're a spectator, we don't want to take the lock in EX because |
867 | * we cannot do the first-mount responsibility it implies: recovery. |
868 | */ |
869 | if (sdp->sd_args.ar_spectator) |
870 | goto locks_done; |
871 | |
872 | error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE); |
873 | if (!error) { |
874 | mounted_mode = DLM_LOCK_EX; |
875 | goto locks_done; |
876 | } else if (error != -EAGAIN) { |
877 | fs_err(sdp, "control_mount mounted_lock EX error %d\n" , error); |
878 | goto fail; |
879 | } |
880 | |
881 | error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE); |
882 | if (!error) { |
883 | mounted_mode = DLM_LOCK_PR; |
884 | goto locks_done; |
885 | } else { |
886 | /* not even -EAGAIN should happen here */ |
887 | fs_err(sdp, "control_mount mounted_lock PR error %d\n" , error); |
888 | goto fail; |
889 | } |
890 | |
891 | locks_done: |
892 | /* |
893 | * If we got both locks above in EX, then we're the first mounter. |
894 | * If not, then we need to wait for the control_lock lvb to be |
895 | * updated by other mounted nodes to reflect our mount generation. |
896 | * |
897 | * In simple first mounter cases, first mounter will see zero lvb_gen, |
898 | * but in cases where all existing nodes leave/fail before mounting |
899 | * nodes finish control_mount, then all nodes will be mounting and |
900 | * lvb_gen will be non-zero. |
901 | */ |
902 | |
903 | control_lvb_read(ls, lvb_gen: &lvb_gen, lvb_bits: ls->ls_lvb_bits); |
904 | |
905 | if (lvb_gen == 0xFFFFFFFF) { |
906 | /* special value to force mount attempts to fail */ |
907 | fs_err(sdp, "control_mount control_lock disabled\n" ); |
908 | error = -EINVAL; |
909 | goto fail; |
910 | } |
911 | |
912 | if (mounted_mode == DLM_LOCK_EX) { |
913 | /* first mounter, keep both EX while doing first recovery */ |
914 | spin_lock(lock: &ls->ls_recover_spin); |
915 | clear_bit(nr: DFL_BLOCK_LOCKS, addr: &ls->ls_recover_flags); |
916 | set_bit(nr: DFL_MOUNT_DONE, addr: &ls->ls_recover_flags); |
917 | set_bit(nr: DFL_FIRST_MOUNT, addr: &ls->ls_recover_flags); |
918 | spin_unlock(lock: &ls->ls_recover_spin); |
919 | fs_info(sdp, "first mounter control generation %u\n" , lvb_gen); |
920 | return 0; |
921 | } |
922 | |
923 | error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT); |
924 | if (error) |
925 | goto fail; |
926 | |
927 | /* |
928 | * We are not first mounter, now we need to wait for the control_lock |
929 | * lvb generation to be >= the generation from our first recover_done |
930 | * and all lvb bits to be clear (no pending journal recoveries.) |
931 | */ |
932 | |
933 | if (!all_jid_bits_clear(lvb: ls->ls_lvb_bits)) { |
934 | /* journals need recovery, wait until all are clear */ |
935 | fs_info(sdp, "control_mount wait for journal recovery\n" ); |
936 | goto restart; |
937 | } |
938 | |
939 | spin_lock(lock: &ls->ls_recover_spin); |
940 | block_gen = ls->ls_recover_block; |
941 | start_gen = ls->ls_recover_start; |
942 | mount_gen = ls->ls_recover_mount; |
943 | |
944 | if (lvb_gen < mount_gen) { |
945 | /* wait for mounted nodes to update control_lock lvb to our |
946 | generation, which might include new recovery bits set */ |
947 | if (sdp->sd_args.ar_spectator) { |
948 | fs_info(sdp, "Recovery is required. Waiting for a " |
949 | "non-spectator to mount.\n" ); |
950 | msleep_interruptible(msecs: 1000); |
951 | } else { |
952 | fs_info(sdp, "control_mount wait1 block %u start %u " |
953 | "mount %u lvb %u flags %lx\n" , block_gen, |
954 | start_gen, mount_gen, lvb_gen, |
955 | ls->ls_recover_flags); |
956 | } |
957 | spin_unlock(lock: &ls->ls_recover_spin); |
958 | goto restart; |
959 | } |
960 | |
961 | if (lvb_gen != start_gen) { |
962 | /* wait for mounted nodes to update control_lock lvb to the |
963 | latest recovery generation */ |
964 | fs_info(sdp, "control_mount wait2 block %u start %u mount %u " |
965 | "lvb %u flags %lx\n" , block_gen, start_gen, mount_gen, |
966 | lvb_gen, ls->ls_recover_flags); |
967 | spin_unlock(lock: &ls->ls_recover_spin); |
968 | goto restart; |
969 | } |
970 | |
971 | if (block_gen == start_gen) { |
972 | /* dlm recovery in progress, wait for it to finish */ |
973 | fs_info(sdp, "control_mount wait3 block %u start %u mount %u " |
974 | "lvb %u flags %lx\n" , block_gen, start_gen, mount_gen, |
975 | lvb_gen, ls->ls_recover_flags); |
976 | spin_unlock(lock: &ls->ls_recover_spin); |
977 | goto restart; |
978 | } |
979 | |
980 | clear_bit(nr: DFL_BLOCK_LOCKS, addr: &ls->ls_recover_flags); |
981 | set_bit(nr: DFL_MOUNT_DONE, addr: &ls->ls_recover_flags); |
982 | memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t)); |
983 | memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t)); |
984 | spin_unlock(lock: &ls->ls_recover_spin); |
985 | return 0; |
986 | |
987 | fail: |
988 | mounted_unlock(sdp); |
989 | control_unlock(sdp); |
990 | return error; |
991 | } |
992 | |
993 | static int control_first_done(struct gfs2_sbd *sdp) |
994 | { |
995 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
996 | uint32_t start_gen, block_gen; |
997 | int error; |
998 | |
999 | restart: |
1000 | spin_lock(lock: &ls->ls_recover_spin); |
1001 | start_gen = ls->ls_recover_start; |
1002 | block_gen = ls->ls_recover_block; |
1003 | |
1004 | if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) || |
1005 | !test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || |
1006 | !test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { |
1007 | /* sanity check, should not happen */ |
1008 | fs_err(sdp, "control_first_done start %u block %u flags %lx\n" , |
1009 | start_gen, block_gen, ls->ls_recover_flags); |
1010 | spin_unlock(lock: &ls->ls_recover_spin); |
1011 | control_unlock(sdp); |
1012 | return -1; |
1013 | } |
1014 | |
1015 | if (start_gen == block_gen) { |
1016 | /* |
1017 | * Wait for the end of a dlm recovery cycle to switch from |
1018 | * first mounter recovery. We can ignore any recover_slot |
1019 | * callbacks between the recover_prep and next recover_done |
1020 | * because we are still the first mounter and any failed nodes |
1021 | * have not fully mounted, so they don't need recovery. |
1022 | */ |
1023 | spin_unlock(lock: &ls->ls_recover_spin); |
1024 | fs_info(sdp, "control_first_done wait gen %u\n" , start_gen); |
1025 | |
1026 | wait_on_bit(word: &ls->ls_recover_flags, bit: DFL_DLM_RECOVERY, |
1027 | TASK_UNINTERRUPTIBLE); |
1028 | goto restart; |
1029 | } |
1030 | |
1031 | clear_bit(nr: DFL_FIRST_MOUNT, addr: &ls->ls_recover_flags); |
1032 | set_bit(nr: DFL_FIRST_MOUNT_DONE, addr: &ls->ls_recover_flags); |
1033 | memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t)); |
1034 | memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t)); |
1035 | spin_unlock(lock: &ls->ls_recover_spin); |
1036 | |
1037 | memset(ls->ls_lvb_bits, 0, GDLM_LVB_SIZE); |
1038 | control_lvb_write(ls, lvb_gen: start_gen, lvb_bits: ls->ls_lvb_bits); |
1039 | |
1040 | error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT); |
1041 | if (error) |
1042 | fs_err(sdp, "control_first_done mounted PR error %d\n" , error); |
1043 | |
1044 | error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK); |
1045 | if (error) |
1046 | fs_err(sdp, "control_first_done control NL error %d\n" , error); |
1047 | |
1048 | return error; |
1049 | } |
1050 | |
1051 | /* |
1052 | * Expand static jid arrays if necessary (by increments of RECOVER_SIZE_INC) |
1053 | * to accommodate the largest slot number. (NB dlm slot numbers start at 1, |
1054 | * gfs2 jids start at 0, so jid = slot - 1) |
1055 | */ |
1056 | |
1057 | #define RECOVER_SIZE_INC 16 |
1058 | |
1059 | static int set_recover_size(struct gfs2_sbd *sdp, struct dlm_slot *slots, |
1060 | int num_slots) |
1061 | { |
1062 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1063 | uint32_t *submit = NULL; |
1064 | uint32_t *result = NULL; |
1065 | uint32_t old_size, new_size; |
1066 | int i, max_jid; |
1067 | |
1068 | if (!ls->ls_lvb_bits) { |
1069 | ls->ls_lvb_bits = kzalloc(GDLM_LVB_SIZE, GFP_NOFS); |
1070 | if (!ls->ls_lvb_bits) |
1071 | return -ENOMEM; |
1072 | } |
1073 | |
1074 | max_jid = 0; |
1075 | for (i = 0; i < num_slots; i++) { |
1076 | if (max_jid < slots[i].slot - 1) |
1077 | max_jid = slots[i].slot - 1; |
1078 | } |
1079 | |
1080 | old_size = ls->ls_recover_size; |
1081 | new_size = old_size; |
1082 | while (new_size < max_jid + 1) |
1083 | new_size += RECOVER_SIZE_INC; |
1084 | if (new_size == old_size) |
1085 | return 0; |
1086 | |
1087 | submit = kcalloc(n: new_size, size: sizeof(uint32_t), GFP_NOFS); |
1088 | result = kcalloc(n: new_size, size: sizeof(uint32_t), GFP_NOFS); |
1089 | if (!submit || !result) { |
1090 | kfree(objp: submit); |
1091 | kfree(objp: result); |
1092 | return -ENOMEM; |
1093 | } |
1094 | |
1095 | spin_lock(lock: &ls->ls_recover_spin); |
1096 | memcpy(submit, ls->ls_recover_submit, old_size * sizeof(uint32_t)); |
1097 | memcpy(result, ls->ls_recover_result, old_size * sizeof(uint32_t)); |
1098 | kfree(objp: ls->ls_recover_submit); |
1099 | kfree(objp: ls->ls_recover_result); |
1100 | ls->ls_recover_submit = submit; |
1101 | ls->ls_recover_result = result; |
1102 | ls->ls_recover_size = new_size; |
1103 | spin_unlock(lock: &ls->ls_recover_spin); |
1104 | return 0; |
1105 | } |
1106 | |
1107 | static void free_recover_size(struct lm_lockstruct *ls) |
1108 | { |
1109 | kfree(objp: ls->ls_lvb_bits); |
1110 | kfree(objp: ls->ls_recover_submit); |
1111 | kfree(objp: ls->ls_recover_result); |
1112 | ls->ls_recover_submit = NULL; |
1113 | ls->ls_recover_result = NULL; |
1114 | ls->ls_recover_size = 0; |
1115 | ls->ls_lvb_bits = NULL; |
1116 | } |
1117 | |
1118 | /* dlm calls before it does lock recovery */ |
1119 | |
1120 | static void gdlm_recover_prep(void *arg) |
1121 | { |
1122 | struct gfs2_sbd *sdp = arg; |
1123 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1124 | |
1125 | if (gfs2_withdrawing_or_withdrawn(sdp)) { |
1126 | fs_err(sdp, "recover_prep ignored due to withdraw.\n" ); |
1127 | return; |
1128 | } |
1129 | spin_lock(lock: &ls->ls_recover_spin); |
1130 | ls->ls_recover_block = ls->ls_recover_start; |
1131 | set_bit(nr: DFL_DLM_RECOVERY, addr: &ls->ls_recover_flags); |
1132 | |
1133 | if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) || |
1134 | test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { |
1135 | spin_unlock(lock: &ls->ls_recover_spin); |
1136 | return; |
1137 | } |
1138 | set_bit(nr: DFL_BLOCK_LOCKS, addr: &ls->ls_recover_flags); |
1139 | spin_unlock(lock: &ls->ls_recover_spin); |
1140 | } |
1141 | |
1142 | /* dlm calls after recover_prep has been completed on all lockspace members; |
1143 | identifies slot/jid of failed member */ |
1144 | |
1145 | static void gdlm_recover_slot(void *arg, struct dlm_slot *slot) |
1146 | { |
1147 | struct gfs2_sbd *sdp = arg; |
1148 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1149 | int jid = slot->slot - 1; |
1150 | |
1151 | if (gfs2_withdrawing_or_withdrawn(sdp)) { |
1152 | fs_err(sdp, "recover_slot jid %d ignored due to withdraw.\n" , |
1153 | jid); |
1154 | return; |
1155 | } |
1156 | spin_lock(lock: &ls->ls_recover_spin); |
1157 | if (ls->ls_recover_size < jid + 1) { |
1158 | fs_err(sdp, "recover_slot jid %d gen %u short size %d\n" , |
1159 | jid, ls->ls_recover_block, ls->ls_recover_size); |
1160 | spin_unlock(lock: &ls->ls_recover_spin); |
1161 | return; |
1162 | } |
1163 | |
1164 | if (ls->ls_recover_submit[jid]) { |
1165 | fs_info(sdp, "recover_slot jid %d gen %u prev %u\n" , |
1166 | jid, ls->ls_recover_block, ls->ls_recover_submit[jid]); |
1167 | } |
1168 | ls->ls_recover_submit[jid] = ls->ls_recover_block; |
1169 | spin_unlock(lock: &ls->ls_recover_spin); |
1170 | } |
1171 | |
1172 | /* dlm calls after recover_slot and after it completes lock recovery */ |
1173 | |
1174 | static void gdlm_recover_done(void *arg, struct dlm_slot *slots, int num_slots, |
1175 | int our_slot, uint32_t generation) |
1176 | { |
1177 | struct gfs2_sbd *sdp = arg; |
1178 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1179 | |
1180 | if (gfs2_withdrawing_or_withdrawn(sdp)) { |
1181 | fs_err(sdp, "recover_done ignored due to withdraw.\n" ); |
1182 | return; |
1183 | } |
1184 | /* ensure the ls jid arrays are large enough */ |
1185 | set_recover_size(sdp, slots, num_slots); |
1186 | |
1187 | spin_lock(lock: &ls->ls_recover_spin); |
1188 | ls->ls_recover_start = generation; |
1189 | |
1190 | if (!ls->ls_recover_mount) { |
1191 | ls->ls_recover_mount = generation; |
1192 | ls->ls_jid = our_slot - 1; |
1193 | } |
1194 | |
1195 | if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) |
1196 | queue_delayed_work(wq: gfs2_control_wq, dwork: &sdp->sd_control_work, delay: 0); |
1197 | |
1198 | clear_bit(nr: DFL_DLM_RECOVERY, addr: &ls->ls_recover_flags); |
1199 | smp_mb__after_atomic(); |
1200 | wake_up_bit(word: &ls->ls_recover_flags, bit: DFL_DLM_RECOVERY); |
1201 | spin_unlock(lock: &ls->ls_recover_spin); |
1202 | } |
1203 | |
1204 | /* gfs2_recover thread has a journal recovery result */ |
1205 | |
1206 | static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid, |
1207 | unsigned int result) |
1208 | { |
1209 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1210 | |
1211 | if (gfs2_withdrawing_or_withdrawn(sdp)) { |
1212 | fs_err(sdp, "recovery_result jid %d ignored due to withdraw.\n" , |
1213 | jid); |
1214 | return; |
1215 | } |
1216 | if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) |
1217 | return; |
1218 | |
1219 | /* don't care about the recovery of own journal during mount */ |
1220 | if (jid == ls->ls_jid) |
1221 | return; |
1222 | |
1223 | spin_lock(lock: &ls->ls_recover_spin); |
1224 | if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) { |
1225 | spin_unlock(lock: &ls->ls_recover_spin); |
1226 | return; |
1227 | } |
1228 | if (ls->ls_recover_size < jid + 1) { |
1229 | fs_err(sdp, "recovery_result jid %d short size %d\n" , |
1230 | jid, ls->ls_recover_size); |
1231 | spin_unlock(lock: &ls->ls_recover_spin); |
1232 | return; |
1233 | } |
1234 | |
1235 | fs_info(sdp, "recover jid %d result %s\n" , jid, |
1236 | result == LM_RD_GAVEUP ? "busy" : "success" ); |
1237 | |
1238 | ls->ls_recover_result[jid] = result; |
1239 | |
1240 | /* GAVEUP means another node is recovering the journal; delay our |
1241 | next attempt to recover it, to give the other node a chance to |
1242 | finish before trying again */ |
1243 | |
1244 | if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags)) |
1245 | queue_delayed_work(wq: gfs2_control_wq, dwork: &sdp->sd_control_work, |
1246 | delay: result == LM_RD_GAVEUP ? HZ : 0); |
1247 | spin_unlock(lock: &ls->ls_recover_spin); |
1248 | } |
1249 | |
1250 | static const struct dlm_lockspace_ops gdlm_lockspace_ops = { |
1251 | .recover_prep = gdlm_recover_prep, |
1252 | .recover_slot = gdlm_recover_slot, |
1253 | .recover_done = gdlm_recover_done, |
1254 | }; |
1255 | |
1256 | static int gdlm_mount(struct gfs2_sbd *sdp, const char *table) |
1257 | { |
1258 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1259 | char cluster[GFS2_LOCKNAME_LEN]; |
1260 | const char *fsname; |
1261 | uint32_t flags; |
1262 | int error, ops_result; |
1263 | |
1264 | /* |
1265 | * initialize everything |
1266 | */ |
1267 | |
1268 | INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func); |
1269 | spin_lock_init(&ls->ls_recover_spin); |
1270 | ls->ls_recover_flags = 0; |
1271 | ls->ls_recover_mount = 0; |
1272 | ls->ls_recover_start = 0; |
1273 | ls->ls_recover_block = 0; |
1274 | ls->ls_recover_size = 0; |
1275 | ls->ls_recover_submit = NULL; |
1276 | ls->ls_recover_result = NULL; |
1277 | ls->ls_lvb_bits = NULL; |
1278 | |
1279 | error = set_recover_size(sdp, NULL, num_slots: 0); |
1280 | if (error) |
1281 | goto fail; |
1282 | |
1283 | /* |
1284 | * prepare dlm_new_lockspace args |
1285 | */ |
1286 | |
1287 | fsname = strchr(table, ':'); |
1288 | if (!fsname) { |
1289 | fs_info(sdp, "no fsname found\n" ); |
1290 | error = -EINVAL; |
1291 | goto fail_free; |
1292 | } |
1293 | memset(cluster, 0, sizeof(cluster)); |
1294 | memcpy(cluster, table, strlen(table) - strlen(fsname)); |
1295 | fsname++; |
1296 | |
1297 | flags = DLM_LSFL_NEWEXCL; |
1298 | |
1299 | /* |
1300 | * create/join lockspace |
1301 | */ |
1302 | |
1303 | error = dlm_new_lockspace(name: fsname, cluster, flags, GDLM_LVB_SIZE, |
1304 | ops: &gdlm_lockspace_ops, ops_arg: sdp, ops_result: &ops_result, |
1305 | lockspace: &ls->ls_dlm); |
1306 | if (error) { |
1307 | fs_err(sdp, "dlm_new_lockspace error %d\n" , error); |
1308 | goto fail_free; |
1309 | } |
1310 | |
1311 | if (ops_result < 0) { |
1312 | /* |
1313 | * dlm does not support ops callbacks, |
1314 | * old dlm_controld/gfs_controld are used, try without ops. |
1315 | */ |
1316 | fs_info(sdp, "dlm lockspace ops not used\n" ); |
1317 | free_recover_size(ls); |
1318 | set_bit(nr: DFL_NO_DLM_OPS, addr: &ls->ls_recover_flags); |
1319 | return 0; |
1320 | } |
1321 | |
1322 | if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) { |
1323 | fs_err(sdp, "dlm lockspace ops disallow jid preset\n" ); |
1324 | error = -EINVAL; |
1325 | goto fail_release; |
1326 | } |
1327 | |
1328 | /* |
1329 | * control_mount() uses control_lock to determine first mounter, |
1330 | * and for later mounts, waits for any recoveries to be cleared. |
1331 | */ |
1332 | |
1333 | error = control_mount(sdp); |
1334 | if (error) { |
1335 | fs_err(sdp, "mount control error %d\n" , error); |
1336 | goto fail_release; |
1337 | } |
1338 | |
1339 | ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags); |
1340 | clear_bit(nr: SDF_NOJOURNALID, addr: &sdp->sd_flags); |
1341 | smp_mb__after_atomic(); |
1342 | wake_up_bit(word: &sdp->sd_flags, bit: SDF_NOJOURNALID); |
1343 | return 0; |
1344 | |
1345 | fail_release: |
1346 | dlm_release_lockspace(lockspace: ls->ls_dlm, force: 2); |
1347 | fail_free: |
1348 | free_recover_size(ls); |
1349 | fail: |
1350 | return error; |
1351 | } |
1352 | |
1353 | static void gdlm_first_done(struct gfs2_sbd *sdp) |
1354 | { |
1355 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1356 | int error; |
1357 | |
1358 | if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) |
1359 | return; |
1360 | |
1361 | error = control_first_done(sdp); |
1362 | if (error) |
1363 | fs_err(sdp, "mount first_done error %d\n" , error); |
1364 | } |
1365 | |
1366 | static void gdlm_unmount(struct gfs2_sbd *sdp) |
1367 | { |
1368 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; |
1369 | |
1370 | if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags)) |
1371 | goto release; |
1372 | |
1373 | /* wait for gfs2_control_wq to be done with this mount */ |
1374 | |
1375 | spin_lock(lock: &ls->ls_recover_spin); |
1376 | set_bit(nr: DFL_UNMOUNT, addr: &ls->ls_recover_flags); |
1377 | spin_unlock(lock: &ls->ls_recover_spin); |
1378 | flush_delayed_work(dwork: &sdp->sd_control_work); |
1379 | |
1380 | /* mounted_lock and control_lock will be purged in dlm recovery */ |
1381 | release: |
1382 | if (ls->ls_dlm) { |
1383 | dlm_release_lockspace(lockspace: ls->ls_dlm, force: 2); |
1384 | ls->ls_dlm = NULL; |
1385 | } |
1386 | |
1387 | free_recover_size(ls); |
1388 | } |
1389 | |
1390 | static const match_table_t dlm_tokens = { |
1391 | { Opt_jid, "jid=%d" }, |
1392 | { Opt_id, "id=%d" }, |
1393 | { Opt_first, "first=%d" }, |
1394 | { Opt_nodir, "nodir=%d" }, |
1395 | { Opt_err, NULL }, |
1396 | }; |
1397 | |
1398 | const struct lm_lockops gfs2_dlm_ops = { |
1399 | .lm_proto_name = "lock_dlm" , |
1400 | .lm_mount = gdlm_mount, |
1401 | .lm_first_done = gdlm_first_done, |
1402 | .lm_recovery_result = gdlm_recovery_result, |
1403 | .lm_unmount = gdlm_unmount, |
1404 | .lm_put_lock = gdlm_put_lock, |
1405 | .lm_lock = gdlm_lock, |
1406 | .lm_cancel = gdlm_cancel, |
1407 | .lm_tokens = &dlm_tokens, |
1408 | }; |
1409 | |
1410 | |