summaryrefslogtreecommitdiff
path: root/fs/ocfs2/dlm/dlmconvert.c
blob: 965f45dbe17bfd4825b9f1dc43988cb4fc89718a (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
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
// SPDX-License-Identifier: GPL-2.0-or-later
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmconvert.c
 *
 * underlying calls for lock conversion
 *
 * Copyright (C) 2004 Oracle.  All rights reserved.
 */


#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/sysctl.h>
#include <linux/random.h>
#include <linux/blkdev.h>
#include <linux/socket.h>
#include <linux/inet.h>
#include <linux/spinlock.h>


#include "cluster/heartbeat.h"
#include "cluster/nodemanager.h"
#include "cluster/tcp.h"

#include "dlmapi.h"
#include "dlmcommon.h"

#include "dlmconvert.h"

#define MLOG_MASK_PREFIX ML_DLM
#include "cluster/masklog.h"

/* NOTE: __dlmconvert_master is the only function in here that
 * needs a spinlock held on entry (res->spinlock) and it is the
 * only one that holds a lock on exit (res->spinlock).
 * All other functions in here need no locks and drop all of
 * the locks that they acquire. */
static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm,
					   struct dlm_lock_resource *res,
					   struct dlm_lock *lock, int flags,
					   int type, int *call_ast,
					   int *kick_thread);
static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
					   struct dlm_lock_resource *res,
					   struct dlm_lock *lock, int flags, int type);

/*
 * this is only called directly by dlmlock(), and only when the
 * local node is the owner of the lockres
 * locking:
 *   caller needs:  none
 *   taken:         takes and drops res->spinlock
 *   held on exit:  none
 * returns: see __dlmconvert_master
 */
enum dlm_status dlmconvert_master(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *res,
				  struct dlm_lock *lock, int flags, int type)
{
	int call_ast = 0, kick_thread = 0;
	enum dlm_status status;

	spin_lock(&res->spinlock);
	/* we are not in a network handler, this is fine */
	__dlm_wait_on_lockres(res);
	__dlm_lockres_reserve_ast(res);
	res->state |= DLM_LOCK_RES_IN_PROGRESS;

	status = __dlmconvert_master(dlm, res, lock, flags, type,
				     &call_ast, &kick_thread);

	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
	spin_unlock(&res->spinlock);
	wake_up(&res->wq);
	if (status != DLM_NORMAL && status != DLM_NOTQUEUED)
		dlm_error(status);

	/* either queue the ast or release it */
	if (call_ast)
		dlm_queue_ast(dlm, lock);
	else
		dlm_lockres_release_ast(dlm, res);

	if (kick_thread)
		dlm_kick_thread(dlm, res);

	return status;
}

/* performs lock conversion at the lockres master site
 * locking:
 *   caller needs:  res->spinlock
 *   taken:         takes and drops lock->spinlock
 *   held on exit:  res->spinlock
 * returns: DLM_NORMAL, DLM_NOTQUEUED, DLM_DENIED
 *   call_ast: whether ast should be called for this lock
 *   kick_thread: whether dlm_kick_thread should be called
 */
static enum dlm_status __dlmconvert_master(struct dlm_ctxt *dlm,
					   struct dlm_lock_resource *res,
					   struct dlm_lock *lock, int flags,
					   int type, int *call_ast,
					   int *kick_thread)
{
	enum dlm_status status = DLM_NORMAL;
	struct dlm_lock *tmplock=NULL;

	assert_spin_locked(&res->spinlock);

	mlog(0, "type=%d, convert_type=%d, new convert_type=%d\n",
	     lock->ml.type, lock->ml.convert_type, type);

	spin_lock(&lock->spinlock);

	/* already converting? */
	if (lock->ml.convert_type != LKM_IVMODE) {
		mlog(ML_ERROR, "attempted to convert a lock with a lock "
		     "conversion pending\n");
		status = DLM_DENIED;
		goto unlock_exit;
	}

	/* must be on grant queue to convert */
	if (!dlm_lock_on_list(&res->granted, lock)) {
		mlog(ML_ERROR, "attempted to convert a lock not on grant "
		     "queue\n");
		status = DLM_DENIED;
		goto unlock_exit;
	}

	if (flags & LKM_VALBLK) {
		switch (lock->ml.type) {
			case LKM_EXMODE:
				/* EX + LKM_VALBLK + convert == set lvb */
				mlog(0, "will set lvb: converting %s->%s\n",
				     dlm_lock_mode_name(lock->ml.type),
				     dlm_lock_mode_name(type));
				lock->lksb->flags |= DLM_LKSB_PUT_LVB;
				break;
			case LKM_PRMODE:
			case LKM_NLMODE:
				/* refetch if new level is not NL */
				if (type > LKM_NLMODE) {
					mlog(0, "will fetch new value into "
					     "lvb: converting %s->%s\n",
					     dlm_lock_mode_name(lock->ml.type),
					     dlm_lock_mode_name(type));
					lock->lksb->flags |= DLM_LKSB_GET_LVB;
				} else {
					mlog(0, "will NOT fetch new value "
					     "into lvb: converting %s->%s\n",
					     dlm_lock_mode_name(lock->ml.type),
					     dlm_lock_mode_name(type));
					flags &= ~(LKM_VALBLK);
				}
				break;
		}
	}


	/* in-place downconvert? */
	if (type <= lock->ml.type)
		goto grant;

	/* upconvert from here on */
	status = DLM_NORMAL;
	list_for_each_entry(tmplock, &res->granted, list) {
		if (tmplock == lock)
			continue;
		if (!dlm_lock_compatible(tmplock->ml.type, type))
			goto switch_queues;
	}

	list_for_each_entry(tmplock, &res->converting, list) {
		if (!dlm_lock_compatible(tmplock->ml.type, type))
			goto switch_queues;
		/* existing conversion requests take precedence */
		if (!dlm_lock_compatible(tmplock->ml.convert_type, type))
			goto switch_queues;
	}

	/* fall thru to grant */

grant:
	mlog(0, "res %.*s, granting %s lock\n", res->lockname.len,
	     res->lockname.name, dlm_lock_mode_name(type));
	/* immediately grant the new lock type */
	lock->lksb->status = DLM_NORMAL;
	if (lock->ml.node == dlm->node_num)
		mlog(0, "doing in-place convert for nonlocal lock\n");
	lock->ml.type = type;
	if (lock->lksb->flags & DLM_LKSB_PUT_LVB)
		memcpy(res->lvb, lock->lksb->lvb, DLM_LVB_LEN);

	/*
	 * Move the lock to the tail because it may be the only lock which has
	 * an invalid lvb.
	 */
	list_move_tail(&lock->list, &res->granted);

	status = DLM_NORMAL;
	*call_ast = 1;
	goto unlock_exit;

switch_queues:
	if (flags & LKM_NOQUEUE) {
		mlog(0, "failed to convert NOQUEUE lock %.*s from "
		     "%d to %d...\n", res->lockname.len, res->lockname.name,
		     lock->ml.type, type);
		status = DLM_NOTQUEUED;
		goto unlock_exit;
	}
	mlog(0, "res %.*s, queueing...\n", res->lockname.len,
	     res->lockname.name);

	lock->ml.convert_type = type;
	/* do not alter lock refcount.  switching lists. */
	list_move_tail(&lock->list, &res->converting);

unlock_exit:
	spin_unlock(&lock->spinlock);
	if (status == DLM_DENIED) {
		__dlm_print_one_lock_resource(res);
	}
	if (status == DLM_NORMAL)
		*kick_thread = 1;
	return status;
}

void dlm_revert_pending_convert(struct dlm_lock_resource *res,
				struct dlm_lock *lock)
{
	/* do not alter lock refcount.  switching lists. */
	list_move_tail(&lock->list, &res->granted);
	lock->ml.convert_type = LKM_IVMODE;
	lock->lksb->flags &= ~(DLM_LKSB_GET_LVB|DLM_LKSB_PUT_LVB);
}

/* messages the master site to do lock conversion
 * locking:
 *   caller needs:  none
 *   taken:         takes and drops res->spinlock, uses DLM_LOCK_RES_IN_PROGRESS
 *   held on exit:  none
 * returns: DLM_NORMAL, DLM_RECOVERING, status from remote node
 */
enum dlm_status dlmconvert_remote(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *res,
				  struct dlm_lock *lock, int flags, int type)
{
	enum dlm_status status;

	mlog(0, "type=%d, convert_type=%d, busy=%d\n", lock->ml.type,
	     lock->ml.convert_type, res->state & DLM_LOCK_RES_IN_PROGRESS);

	spin_lock(&res->spinlock);
	if (res->state & DLM_LOCK_RES_RECOVERING) {
		mlog(0, "bailing out early since res is RECOVERING "
		     "on secondary queue\n");
		/* __dlm_print_one_lock_resource(res); */
		status = DLM_RECOVERING;
		goto bail;
	}
	/* will exit this call with spinlock held */
	__dlm_wait_on_lockres(res);

	if (lock->ml.convert_type != LKM_IVMODE) {
		__dlm_print_one_lock_resource(res);
		mlog(ML_ERROR, "converting a remote lock that is already "
		     "converting! (cookie=%u:%llu, conv=%d)\n",
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
		     lock->ml.convert_type);
		status = DLM_DENIED;
		goto bail;
	}

	if (lock->ml.type == type && lock->ml.convert_type == LKM_IVMODE) {
		mlog(0, "last convert request returned DLM_RECOVERING, but "
		     "owner has already queued and sent ast to me. res %.*s, "
		     "(cookie=%u:%llu, type=%d, conv=%d)\n",
		     res->lockname.len, res->lockname.name,
		     dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
		     lock->ml.type, lock->ml.convert_type);
		status = DLM_NORMAL;
		goto bail;
	}

	res->state |= DLM_LOCK_RES_IN_PROGRESS;
	/* move lock to local convert queue */
	/* do not alter lock refcount.  switching lists. */
	list_move_tail(&lock->list, &res->converting);
	lock->convert_pending = 1;
	lock->ml.convert_type = type;

	if (flags & LKM_VALBLK) {
		if (lock->ml.type == LKM_EXMODE) {
			flags |= LKM_PUT_LVB;
			lock->lksb->flags |= DLM_LKSB_PUT_LVB;
		} else {
			if (lock->ml.convert_type == LKM_NLMODE)
				flags &= ~LKM_VALBLK;
			else {
				flags |= LKM_GET_LVB;
				lock->lksb->flags |= DLM_LKSB_GET_LVB;
			}
		}
	}
	spin_unlock(&res->spinlock);

	/* no locks held here.
	 * need to wait for a reply as to whether it got queued or not. */
	status = dlm_send_remote_convert_request(dlm, res, lock, flags, type);

	spin_lock(&res->spinlock);
	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
	/* if it failed, move it back to granted queue.
	 * if master returns DLM_NORMAL and then down before sending ast,
	 * it may have already been moved to granted queue, reset to
	 * DLM_RECOVERING and retry convert */
	if (status != DLM_NORMAL) {
		if (status != DLM_NOTQUEUED)
			dlm_error(status);
		dlm_revert_pending_convert(res, lock);
	} else if (!lock->convert_pending) {
		mlog(0, "%s: res %.*s, owner died and lock has been moved back "
				"to granted list, retry convert.\n",
				dlm->name, res->lockname.len, res->lockname.name);
		status = DLM_RECOVERING;
	}

	lock->convert_pending = 0;
bail:
	spin_unlock(&res->spinlock);

	/* TODO: should this be a wake_one? */
	/* wake up any IN_PROGRESS waiters */
	wake_up(&res->wq);

	return status;
}

/* sends DLM_CONVERT_LOCK_MSG to master site
 * locking:
 *   caller needs:  none
 *   taken:         none
 *   held on exit:  none
 * returns: DLM_NOLOCKMGR, status from remote node
 */
static enum dlm_status dlm_send_remote_convert_request(struct dlm_ctxt *dlm,
					   struct dlm_lock_resource *res,
					   struct dlm_lock *lock, int flags, int type)
{
	struct dlm_convert_lock convert;
	int tmpret;
	enum dlm_status ret;
	int status = 0;
	struct kvec vec[2];
	size_t veclen = 1;

	mlog(0, "%.*s\n", res->lockname.len, res->lockname.name);

	memset(&convert, 0, sizeof(struct dlm_convert_lock));
	convert.node_idx = dlm->node_num;
	convert.requested_type = type;
	convert.cookie = lock->ml.cookie;
	convert.namelen = res->lockname.len;
	convert.flags = cpu_to_be32(flags);
	memcpy(convert.name, res->lockname.name, convert.namelen);

	vec[0].iov_len = sizeof(struct dlm_convert_lock);
	vec[0].iov_base = &convert;

	if (flags & LKM_PUT_LVB) {
		/* extra data to send if we are updating lvb */
		vec[1].iov_len = DLM_LVB_LEN;
		vec[1].iov_base = lock->lksb->lvb;
		veclen++;
	}

	tmpret = o2net_send_message_vec(DLM_CONVERT_LOCK_MSG, dlm->key,
					vec, veclen, res->owner, &status);
	if (tmpret >= 0) {
		// successfully sent and received
		ret = status;  // this is already a dlm_status
		if (ret == DLM_RECOVERING) {
			mlog(0, "node %u returned DLM_RECOVERING from convert "
			     "message!\n", res->owner);
		} else if (ret == DLM_MIGRATING) {
			mlog(0, "node %u returned DLM_MIGRATING from convert "
			     "message!\n", res->owner);
		} else if (ret == DLM_FORWARD) {
			mlog(0, "node %u returned DLM_FORWARD from convert "
			     "message!\n", res->owner);
		} else if (ret != DLM_NORMAL && ret != DLM_NOTQUEUED)
			dlm_error(ret);
	} else {
		mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to "
		     "node %u\n", tmpret, DLM_CONVERT_LOCK_MSG, dlm->key,
		     res->owner);
		if (dlm_is_host_down(tmpret)) {
			/* instead of logging the same network error over
			 * and over, sleep here and wait for the heartbeat
			 * to notice the node is dead.  times out after 5s. */
			dlm_wait_for_node_death(dlm, res->owner,
						DLM_NODE_DEATH_WAIT_MAX);
			ret = DLM_RECOVERING;
			mlog(0, "node %u died so returning DLM_RECOVERING "
			     "from convert message!\n", res->owner);
		} else {
			ret = dlm_err_to_dlm_status(tmpret);
		}
	}

	return ret;
}

/* handler for DLM_CONVERT_LOCK_MSG on master site
 * locking:
 *   caller needs:  none
 *   taken:         takes and drop res->spinlock
 *   held on exit:  none
 * returns: DLM_NORMAL, DLM_IVLOCKID, DLM_BADARGS,
 *          status from __dlmconvert_master
 */
int dlm_convert_lock_handler(struct o2net_msg *msg, u32 len, void *data,
			     void **ret_data)
{
	struct dlm_ctxt *dlm = data;
	struct dlm_convert_lock *cnv = (struct dlm_convert_lock *)msg->buf;
	struct dlm_lock_resource *res = NULL;
	struct dlm_lock *lock = NULL;
	struct dlm_lock *tmp_lock;
	struct dlm_lockstatus *lksb;
	enum dlm_status status = DLM_NORMAL;
	u32 flags;
	int call_ast = 0, kick_thread = 0, ast_reserved = 0, wake = 0;

	if (!dlm_grab(dlm)) {
		dlm_error(DLM_REJECTED);
		return DLM_REJECTED;
	}

	mlog_bug_on_msg(!dlm_domain_fully_joined(dlm),
			"Domain %s not fully joined!\n", dlm->name);

	if (cnv->namelen > DLM_LOCKID_NAME_MAX) {
		status = DLM_IVBUFLEN;
		dlm_error(status);
		goto leave;
	}

	flags = be32_to_cpu(cnv->flags);

	if ((flags & (LKM_PUT_LVB|LKM_GET_LVB)) ==
	     (LKM_PUT_LVB|LKM_GET_LVB)) {
		mlog(ML_ERROR, "both PUT and GET lvb specified\n");
		status = DLM_BADARGS;
		goto leave;
	}

	mlog(0, "lvb: %s\n", flags & LKM_PUT_LVB ? "put lvb" :
	     (flags & LKM_GET_LVB ? "get lvb" : "none"));

	status = DLM_IVLOCKID;
	res = dlm_lookup_lockres(dlm, cnv->name, cnv->namelen);
	if (!res) {
		dlm_error(status);
		goto leave;
	}

	spin_lock(&res->spinlock);
	status = __dlm_lockres_state_to_status(res);
	if (status != DLM_NORMAL) {
		spin_unlock(&res->spinlock);
		dlm_error(status);
		goto leave;
	}
	list_for_each_entry(tmp_lock, &res->granted, list) {
		if (tmp_lock->ml.cookie == cnv->cookie &&
		    tmp_lock->ml.node == cnv->node_idx) {
			lock = tmp_lock;
			dlm_lock_get(lock);
			break;
		}
	}
	spin_unlock(&res->spinlock);
	if (!lock) {
		status = DLM_IVLOCKID;
		mlog(ML_ERROR, "did not find lock to convert on grant queue! "
			       "cookie=%u:%llu\n",
		     dlm_get_lock_cookie_node(be64_to_cpu(cnv->cookie)),
		     dlm_get_lock_cookie_seq(be64_to_cpu(cnv->cookie)));
		dlm_print_one_lock_resource(res);
		goto leave;
	}

	/* found the lock */
	lksb = lock->lksb;

	/* see if caller needed to get/put lvb */
	if (flags & LKM_PUT_LVB) {
		BUG_ON(lksb->flags & (DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB));
		lksb->flags |= DLM_LKSB_PUT_LVB;
		memcpy(&lksb->lvb[0], &cnv->lvb[0], DLM_LVB_LEN);
	} else if (flags & LKM_GET_LVB) {
		BUG_ON(lksb->flags & (DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB));
		lksb->flags |= DLM_LKSB_GET_LVB;
	}

	spin_lock(&res->spinlock);
	status = __dlm_lockres_state_to_status(res);
	if (status == DLM_NORMAL) {
		__dlm_lockres_reserve_ast(res);
		ast_reserved = 1;
		res->state |= DLM_LOCK_RES_IN_PROGRESS;
		status = __dlmconvert_master(dlm, res, lock, flags,
					     cnv->requested_type,
					     &call_ast, &kick_thread);
		res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
		wake = 1;
	}
	spin_unlock(&res->spinlock);
	if (wake)
		wake_up(&res->wq);

	if (status != DLM_NORMAL) {
		if (status != DLM_NOTQUEUED)
			dlm_error(status);
		lksb->flags &= ~(DLM_LKSB_GET_LVB|DLM_LKSB_PUT_LVB);
	}

leave:
	if (lock)
		dlm_lock_put(lock);

	/* either queue the ast or release it, if reserved */
	if (call_ast)
		dlm_queue_ast(dlm, lock);
	else if (ast_reserved)
		dlm_lockres_release_ast(dlm, res);

	if (kick_thread)
		dlm_kick_thread(dlm, res);

	if (res)
		dlm_lockres_put(res);

	dlm_put(dlm);

	return status;
}