summaryrefslogtreecommitdiff
path: root/fs/gfs2/locking/dlm/lock.c
blob: c305255bfe8ade71b4e53b8d0b7c5389da0623a0 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
 * Copyright (C) 2004-2005 Red Hat, Inc.  All rights reserved.
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
 * of the GNU General Public License version 2.
 */

#include "lock_dlm.h"

static char junk_lvb[GDLM_LVB_SIZE];

static void queue_complete(struct gdlm_lock *lp)
{
	struct gdlm_ls *ls = lp->ls;

	clear_bit(LFL_ACTIVE, &lp->flags);

	spin_lock(&ls->async_lock);
	list_add_tail(&lp->clist, &ls->complete);
	spin_unlock(&ls->async_lock);
	wake_up(&ls->thread_wait);
}

static inline void gdlm_ast(void *astarg)
{
	queue_complete(astarg);
}

static inline void gdlm_bast(void *astarg, int mode)
{
	struct gdlm_lock *lp = astarg;
	struct gdlm_ls *ls = lp->ls;

	if (!mode) {
		printk(KERN_INFO "lock_dlm: bast mode zero %x,%llx\n",
			lp->lockname.ln_type,
			(unsigned long long)lp->lockname.ln_number);
		return;
	}

	spin_lock(&ls->async_lock);
	if (!lp->bast_mode) {
		list_add_tail(&lp->blist, &ls->blocking);
		lp->bast_mode = mode;
	} else if (lp->bast_mode < mode)
		lp->bast_mode = mode;
	spin_unlock(&ls->async_lock);
	wake_up(&ls->thread_wait);
}

void gdlm_queue_delayed(struct gdlm_lock *lp)
{
	struct gdlm_ls *ls = lp->ls;

	spin_lock(&ls->async_lock);
	list_add_tail(&lp->delay_list, &ls->delayed);
	spin_unlock(&ls->async_lock);
}

/* convert gfs lock-state to dlm lock-mode */

static s16 make_mode(s16 lmstate)
{
	switch (lmstate) {
	case LM_ST_UNLOCKED:
		return DLM_LOCK_NL;
	case LM_ST_EXCLUSIVE:
		return DLM_LOCK_EX;
	case LM_ST_DEFERRED:
		return DLM_LOCK_CW;
	case LM_ST_SHARED:
		return DLM_LOCK_PR;
	}
	gdlm_assert(0, "unknown LM state %d", lmstate);
	return -1;
}

/* convert dlm lock-mode to gfs lock-state */

s16 gdlm_make_lmstate(s16 dlmmode)
{
	switch (dlmmode) {
	case DLM_LOCK_IV:
	case DLM_LOCK_NL:
		return LM_ST_UNLOCKED;
	case DLM_LOCK_EX:
		return LM_ST_EXCLUSIVE;
	case DLM_LOCK_CW:
		return LM_ST_DEFERRED;
	case DLM_LOCK_PR:
		return LM_ST_SHARED;
	}
	gdlm_assert(0, "unknown DLM mode %d", dlmmode);
	return -1;
}

/* verify agreement with GFS on the current lock state, NB: DLM_LOCK_NL and
   DLM_LOCK_IV are both considered LM_ST_UNLOCKED by GFS. */

static void check_cur_state(struct gdlm_lock *lp, unsigned int cur_state)
{
	s16 cur = make_mode(cur_state);
	if (lp->cur != DLM_LOCK_IV)
		gdlm_assert(lp->cur == cur, "%d, %d", lp->cur, cur);
}

static inline unsigned int make_flags(struct gdlm_lock *lp,
				      unsigned int gfs_flags,
				      s16 cur, s16 req)
{
	unsigned int lkf = 0;

	if (gfs_flags & LM_FLAG_TRY)
		lkf |= DLM_LKF_NOQUEUE;

	if (gfs_flags & LM_FLAG_TRY_1CB) {
		lkf |= DLM_LKF_NOQUEUE;
		lkf |= DLM_LKF_NOQUEUEBAST;
	}

	if (gfs_flags & LM_FLAG_PRIORITY) {
		lkf |= DLM_LKF_NOORDER;
		lkf |= DLM_LKF_HEADQUE;
	}

	if (gfs_flags & LM_FLAG_ANY) {
		if (req == DLM_LOCK_PR)
			lkf |= DLM_LKF_ALTCW;
		else if (req == DLM_LOCK_CW)
			lkf |= DLM_LKF_ALTPR;
	}

	if (lp->lksb.sb_lkid != 0) {
		lkf |= DLM_LKF_CONVERT;

		/* Conversion deadlock avoidance by DLM */

		if (!test_bit(LFL_FORCE_PROMOTE, &lp->flags) &&
		    !(lkf & DLM_LKF_NOQUEUE) &&
		    cur > DLM_LOCK_NL && req > DLM_LOCK_NL && cur != req)
			lkf |= DLM_LKF_CONVDEADLK;
	}

	if (lp->lvb)
		lkf |= DLM_LKF_VALBLK;

	return lkf;
}

/* make_strname - convert GFS lock numbers to a string */

static inline void make_strname(const struct lm_lockname *lockname,
				struct gdlm_strname *str)
{
	sprintf(str->name, "%8x%16llx", lockname->ln_type,
		(unsigned long long)lockname->ln_number);
	str->namelen = GDLM_STRNAME_BYTES;
}

static int gdlm_create_lp(struct gdlm_ls *ls, struct lm_lockname *name,
			  struct gdlm_lock **lpp)
{
	struct gdlm_lock *lp;

	lp = kzalloc(sizeof(struct gdlm_lock), GFP_KERNEL);
	if (!lp)
		return -ENOMEM;

	lp->lockname = *name;
	make_strname(name, &lp->strname);
	lp->ls = ls;
	lp->cur = DLM_LOCK_IV;
	lp->lvb = NULL;
	lp->hold_null = NULL;
	init_completion(&lp->ast_wait);
	INIT_LIST_HEAD(&lp->clist);
	INIT_LIST_HEAD(&lp->blist);
	INIT_LIST_HEAD(&lp->delay_list);

	spin_lock(&ls->async_lock);
	list_add(&lp->all_list, &ls->all_locks);
	ls->all_locks_count++;
	spin_unlock(&ls->async_lock);

	*lpp = lp;
	return 0;
}

void gdlm_delete_lp(struct gdlm_lock *lp)
{
	struct gdlm_ls *ls = lp->ls;

	spin_lock(&ls->async_lock);
	if (!list_empty(&lp->clist))
		list_del_init(&lp->clist);
	if (!list_empty(&lp->blist))
		list_del_init(&lp->blist);
	if (!list_empty(&lp->delay_list))
		list_del_init(&lp->delay_list);
	gdlm_assert(!list_empty(&lp->all_list), "%x,%llx", lp->lockname.ln_type,
		    (unsigned long long)lp->lockname.ln_number);
	list_del_init(&lp->all_list);
	ls->all_locks_count--;
	spin_unlock(&ls->async_lock);

	kfree(lp);
}

int gdlm_get_lock(void *lockspace, struct lm_lockname *name,
		  void **lockp)
{
	struct gdlm_lock *lp;
	int error;

	error = gdlm_create_lp(lockspace, name, &lp);

	*lockp = lp;
	return error;
}

void gdlm_put_lock(void *lock)
{
	gdlm_delete_lp(lock);
}

unsigned int gdlm_do_lock(struct gdlm_lock *lp)
{
	struct gdlm_ls *ls = lp->ls;
	int error, bast = 1;

	/*
	 * When recovery is in progress, delay lock requests for submission
	 * once recovery is done.  Requests for recovery (NOEXP) and unlocks
	 * can pass.
	 */

	if (test_bit(DFL_BLOCK_LOCKS, &ls->flags) &&
	    !test_bit(LFL_NOBLOCK, &lp->flags) && lp->req != DLM_LOCK_NL) {
		gdlm_queue_delayed(lp);
		return LM_OUT_ASYNC;
	}

	/*
	 * Submit the actual lock request.
	 */

	if (test_bit(LFL_NOBAST, &lp->flags))
		bast = 0;

	set_bit(LFL_ACTIVE, &lp->flags);

	log_debug("lk %x,%llx id %x %d,%d %x", lp->lockname.ln_type,
		  (unsigned long long)lp->lockname.ln_number, lp->lksb.sb_lkid,
		  lp->cur, lp->req, lp->lkf);

	error = dlm_lock(ls->dlm_lockspace, lp->req, &lp->lksb, lp->lkf,
			 lp->strname.name, lp->strname.namelen, 0, gdlm_ast,
			 lp, bast ? gdlm_bast : NULL);

	if ((error == -EAGAIN) && (lp->lkf & DLM_LKF_NOQUEUE)) {
		lp->lksb.sb_status = -EAGAIN;
		queue_complete(lp);
		error = 0;
	}

	if (error) {
		log_error("%s: gdlm_lock %x,%llx err=%d cur=%d req=%d lkf=%x "
			  "flags=%lx", ls->fsname, lp->lockname.ln_type,
			  (unsigned long long)lp->lockname.ln_number, error,
			  lp->cur, lp->req, lp->lkf, lp->flags);
		return LM_OUT_ERROR;
	}
	return LM_OUT_ASYNC;
}

static unsigned int gdlm_do_unlock(struct gdlm_lock *lp)
{
	struct gdlm_ls *ls = lp->ls;
	unsigned int lkf = 0;
	int error;

	set_bit(LFL_DLM_UNLOCK, &lp->flags);
	set_bit(LFL_ACTIVE, &lp->flags);

	if (lp->lvb)
		lkf = DLM_LKF_VALBLK;

	log_debug("un %x,%llx %x %d %x", lp->lockname.ln_type,
		  (unsigned long long)lp->lockname.ln_number,
		  lp->lksb.sb_lkid, lp->cur, lkf);

	error = dlm_unlock(ls->dlm_lockspace, lp->lksb.sb_lkid, lkf, NULL, lp);

	if (error) {
		log_error("%s: gdlm_unlock %x,%llx err=%d cur=%d req=%d lkf=%x "
			  "flags=%lx", ls->fsname, lp->lockname.ln_type,
			  (unsigned long long)lp->lockname.ln_number, error,
			  lp->cur, lp->req, lp->lkf, lp->flags);
		return LM_OUT_ERROR;
	}
	return LM_OUT_ASYNC;
}

unsigned int gdlm_lock(void *lock, unsigned int cur_state,
		       unsigned int req_state, unsigned int flags)
{
	struct gdlm_lock *lp = lock;

	clear_bit(LFL_DLM_CANCEL, &lp->flags);
	if (flags & LM_FLAG_NOEXP)
		set_bit(LFL_NOBLOCK, &lp->flags);

	check_cur_state(lp, cur_state);
	lp->req = make_mode(req_state);
	lp->lkf = make_flags(lp, flags, lp->cur, lp->req);

	return gdlm_do_lock(lp);
}

unsigned int gdlm_unlock(void *lock, unsigned int cur_state)
{
	struct gdlm_lock *lp = lock;

	clear_bit(LFL_DLM_CANCEL, &lp->flags);
	if (lp->cur == DLM_LOCK_IV)
		return 0;
	return gdlm_do_unlock(lp);
}

void gdlm_cancel(void *lock)
{
	struct gdlm_lock *lp = lock;
	struct gdlm_ls *ls = lp->ls;
	int error, delay_list = 0;

	if (test_bit(LFL_DLM_CANCEL, &lp->flags))
		return;

	log_info("gdlm_cancel %x,%llx flags %lx", lp->lockname.ln_type,
		 (unsigned long long)lp->lockname.ln_number, lp->flags);

	spin_lock(&ls->async_lock);
	if (!list_empty(&lp->delay_list)) {
		list_del_init(&lp->delay_list);
		delay_list = 1;
	}
	spin_unlock(&ls->async_lock);

	if (delay_list) {
		set_bit(LFL_CANCEL, &lp->flags);
		set_bit(LFL_ACTIVE, &lp->flags);
		queue_complete(lp);
		return;
	}

	if (!test_bit(LFL_ACTIVE, &lp->flags) ||
	    test_bit(LFL_DLM_UNLOCK, &lp->flags)) {
		log_info("gdlm_cancel skip %x,%llx flags %lx",
		 	 lp->lockname.ln_type,
			 (unsigned long long)lp->lockname.ln_number, lp->flags);
		return;
	}

	/* the lock is blocked in the dlm */

	set_bit(LFL_DLM_CANCEL, &lp->flags);
	set_bit(LFL_ACTIVE, &lp->flags);

	error = dlm_unlock(ls->dlm_lockspace, lp->lksb.sb_lkid, DLM_LKF_CANCEL,
			   NULL, lp);

	log_info("gdlm_cancel rv %d %x,%llx flags %lx", error,
		 lp->lockname.ln_type,
		 (unsigned long long)lp->lockname.ln_number, lp->flags);

	if (error == -EBUSY)
		clear_bit(LFL_DLM_CANCEL, &lp->flags);
}

static int gdlm_add_lvb(struct gdlm_lock *lp)
{
	char *lvb;

	lvb = kzalloc(GDLM_LVB_SIZE, GFP_KERNEL);
	if (!lvb)
		return -ENOMEM;

	lp->lksb.sb_lvbptr = lvb;
	lp->lvb = lvb;
	return 0;
}

static void gdlm_del_lvb(struct gdlm_lock *lp)
{
	kfree(lp->lvb);
	lp->lvb = NULL;
	lp->lksb.sb_lvbptr = NULL;
}

/* This can do a synchronous dlm request (requiring a lock_dlm thread to get
   the completion) because gfs won't call hold_lvb() during a callback (from
   the context of a lock_dlm thread). */

static int hold_null_lock(struct gdlm_lock *lp)
{
	struct gdlm_lock *lpn = NULL;
	int error;

	if (lp->hold_null) {
		printk(KERN_INFO "lock_dlm: lvb already held\n");
		return 0;
	}

	error = gdlm_create_lp(lp->ls, &lp->lockname, &lpn);
	if (error)
		goto out;

	lpn->lksb.sb_lvbptr = junk_lvb;
	lpn->lvb = junk_lvb;

	lpn->req = DLM_LOCK_NL;
	lpn->lkf = DLM_LKF_VALBLK | DLM_LKF_EXPEDITE;
	set_bit(LFL_NOBAST, &lpn->flags);
	set_bit(LFL_INLOCK, &lpn->flags);

	init_completion(&lpn->ast_wait);
	gdlm_do_lock(lpn);
	wait_for_completion(&lpn->ast_wait);
	error = lpn->lksb.sb_status;
	if (error) {
		printk(KERN_INFO "lock_dlm: hold_null_lock dlm error %d\n",
		       error);
		gdlm_delete_lp(lpn);
		lpn = NULL;
	}
out:
	lp->hold_null = lpn;
	return error;
}

/* This cannot do a synchronous dlm request (requiring a lock_dlm thread to get
   the completion) because gfs may call unhold_lvb() during a callback (from
   the context of a lock_dlm thread) which could cause a deadlock since the
   other lock_dlm thread could be engaged in recovery. */

static void unhold_null_lock(struct gdlm_lock *lp)
{
	struct gdlm_lock *lpn = lp->hold_null;

	gdlm_assert(lpn, "%x,%llx", lp->lockname.ln_type,
		    (unsigned long long)lp->lockname.ln_number);
	lpn->lksb.sb_lvbptr = NULL;
	lpn->lvb = NULL;
	set_bit(LFL_UNLOCK_DELETE, &lpn->flags);
	gdlm_do_unlock(lpn);
	lp->hold_null = NULL;
}

/* Acquire a NL lock because gfs requires the value block to remain
   intact on the resource while the lvb is "held" even if it's holding no locks
   on the resource. */

int gdlm_hold_lvb(void *lock, char **lvbp)
{
	struct gdlm_lock *lp = lock;
	int error;

	error = gdlm_add_lvb(lp);
	if (error)
		return error;

	*lvbp = lp->lvb;

	error = hold_null_lock(lp);
	if (error)
		gdlm_del_lvb(lp);

	return error;
}

void gdlm_unhold_lvb(void *lock, char *lvb)
{
	struct gdlm_lock *lp = lock;

	unhold_null_lock(lp);
	gdlm_del_lvb(lp);
}

void gdlm_submit_delayed(struct gdlm_ls *ls)
{
	struct gdlm_lock *lp, *safe;

	spin_lock(&ls->async_lock);
	list_for_each_entry_safe(lp, safe, &ls->delayed, delay_list) {
		list_del_init(&lp->delay_list);
		list_add_tail(&lp->delay_list, &ls->submit);
	}
	spin_unlock(&ls->async_lock);
	wake_up(&ls->thread_wait);
}

int gdlm_release_all_locks(struct gdlm_ls *ls)
{
	struct gdlm_lock *lp, *safe;
	int count = 0;

	spin_lock(&ls->async_lock);
	list_for_each_entry_safe(lp, safe, &ls->all_locks, all_list) {
		list_del_init(&lp->all_list);

		if (lp->lvb && lp->lvb != junk_lvb)
			kfree(lp->lvb);
		kfree(lp);
		count++;
	}
	spin_unlock(&ls->async_lock);

	return count;
}