xref: /openbmc/linux/fs/ocfs2/dlm/dlmlock.c (revision 12eb4683)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * dlmlock.c
5  *
6  * underlying calls for lock creation
7  *
8  * Copyright (C) 2004 Oracle.  All rights reserved.
9  *
10  * This program is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU General Public
12  * License as published by the Free Software Foundation; either
13  * version 2 of the License, or (at your option) any later version.
14  *
15  * This program is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18  * General Public License for more details.
19  *
20  * You should have received a copy of the GNU General Public
21  * License along with this program; if not, write to the
22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23  * Boston, MA 021110-1307, USA.
24  *
25  */
26 
27 
28 #include <linux/module.h>
29 #include <linux/fs.h>
30 #include <linux/types.h>
31 #include <linux/slab.h>
32 #include <linux/highmem.h>
33 #include <linux/init.h>
34 #include <linux/sysctl.h>
35 #include <linux/random.h>
36 #include <linux/blkdev.h>
37 #include <linux/socket.h>
38 #include <linux/inet.h>
39 #include <linux/spinlock.h>
40 #include <linux/delay.h>
41 
42 
43 #include "cluster/heartbeat.h"
44 #include "cluster/nodemanager.h"
45 #include "cluster/tcp.h"
46 
47 #include "dlmapi.h"
48 #include "dlmcommon.h"
49 
50 #include "dlmconvert.h"
51 
52 #define MLOG_MASK_PREFIX ML_DLM
53 #include "cluster/masklog.h"
54 
55 static struct kmem_cache *dlm_lock_cache = NULL;
56 
57 static DEFINE_SPINLOCK(dlm_cookie_lock);
58 static u64 dlm_next_cookie = 1;
59 
60 static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
61 					       struct dlm_lock_resource *res,
62 					       struct dlm_lock *lock, int flags);
63 static void dlm_init_lock(struct dlm_lock *newlock, int type,
64 			  u8 node, u64 cookie);
65 static void dlm_lock_release(struct kref *kref);
66 static void dlm_lock_detach_lockres(struct dlm_lock *lock);
67 
68 int dlm_init_lock_cache(void)
69 {
70 	dlm_lock_cache = kmem_cache_create("o2dlm_lock",
71 					   sizeof(struct dlm_lock),
72 					   0, SLAB_HWCACHE_ALIGN, NULL);
73 	if (dlm_lock_cache == NULL)
74 		return -ENOMEM;
75 	return 0;
76 }
77 
78 void dlm_destroy_lock_cache(void)
79 {
80 	if (dlm_lock_cache)
81 		kmem_cache_destroy(dlm_lock_cache);
82 }
83 
84 /* Tell us whether we can grant a new lock request.
85  * locking:
86  *   caller needs:  res->spinlock
87  *   taken:         none
88  *   held on exit:  none
89  * returns: 1 if the lock can be granted, 0 otherwise.
90  */
91 static int dlm_can_grant_new_lock(struct dlm_lock_resource *res,
92 				  struct dlm_lock *lock)
93 {
94 	struct dlm_lock *tmplock;
95 
96 	list_for_each_entry(tmplock, &res->granted, list) {
97 		if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type))
98 			return 0;
99 	}
100 
101 	list_for_each_entry(tmplock, &res->converting, list) {
102 		if (!dlm_lock_compatible(tmplock->ml.type, lock->ml.type))
103 			return 0;
104 		if (!dlm_lock_compatible(tmplock->ml.convert_type,
105 					 lock->ml.type))
106 			return 0;
107 	}
108 
109 	return 1;
110 }
111 
112 /* performs lock creation at the lockres master site
113  * locking:
114  *   caller needs:  none
115  *   taken:         takes and drops res->spinlock
116  *   held on exit:  none
117  * returns: DLM_NORMAL, DLM_NOTQUEUED
118  */
119 static enum dlm_status dlmlock_master(struct dlm_ctxt *dlm,
120 				      struct dlm_lock_resource *res,
121 				      struct dlm_lock *lock, int flags)
122 {
123 	int call_ast = 0, kick_thread = 0;
124 	enum dlm_status status = DLM_NORMAL;
125 
126 	mlog(0, "type=%d\n", lock->ml.type);
127 
128 	spin_lock(&res->spinlock);
129 	/* if called from dlm_create_lock_handler, need to
130 	 * ensure it will not sleep in dlm_wait_on_lockres */
131 	status = __dlm_lockres_state_to_status(res);
132 	if (status != DLM_NORMAL &&
133 	    lock->ml.node != dlm->node_num) {
134 		/* erf.  state changed after lock was dropped. */
135 		spin_unlock(&res->spinlock);
136 		dlm_error(status);
137 		return status;
138 	}
139 	__dlm_wait_on_lockres(res);
140 	__dlm_lockres_reserve_ast(res);
141 
142 	if (dlm_can_grant_new_lock(res, lock)) {
143 		mlog(0, "I can grant this lock right away\n");
144 		/* got it right away */
145 		lock->lksb->status = DLM_NORMAL;
146 		status = DLM_NORMAL;
147 		dlm_lock_get(lock);
148 		list_add_tail(&lock->list, &res->granted);
149 
150 		/* for the recovery lock, we can't allow the ast
151 		 * to be queued since the dlmthread is already
152 		 * frozen.  but the recovery lock is always locked
153 		 * with LKM_NOQUEUE so we do not need the ast in
154 		 * this special case */
155 		if (!dlm_is_recovery_lock(res->lockname.name,
156 					  res->lockname.len)) {
157 			kick_thread = 1;
158 			call_ast = 1;
159 		} else {
160 			mlog(0, "%s: returning DLM_NORMAL to "
161 			     "node %u for reco lock\n", dlm->name,
162 			     lock->ml.node);
163 		}
164 	} else {
165 		/* for NOQUEUE request, unless we get the
166 		 * lock right away, return DLM_NOTQUEUED */
167 		if (flags & LKM_NOQUEUE) {
168 			status = DLM_NOTQUEUED;
169 			if (dlm_is_recovery_lock(res->lockname.name,
170 						 res->lockname.len)) {
171 				mlog(0, "%s: returning NOTQUEUED to "
172 				     "node %u for reco lock\n", dlm->name,
173 				     lock->ml.node);
174 			}
175 		} else {
176 			status = DLM_NORMAL;
177 			dlm_lock_get(lock);
178 			list_add_tail(&lock->list, &res->blocked);
179 			kick_thread = 1;
180 		}
181 	}
182 
183 	spin_unlock(&res->spinlock);
184 	wake_up(&res->wq);
185 
186 	/* either queue the ast or release it */
187 	if (call_ast)
188 		dlm_queue_ast(dlm, lock);
189 	else
190 		dlm_lockres_release_ast(dlm, res);
191 
192 	dlm_lockres_calc_usage(dlm, res);
193 	if (kick_thread)
194 		dlm_kick_thread(dlm, res);
195 
196 	return status;
197 }
198 
199 void dlm_revert_pending_lock(struct dlm_lock_resource *res,
200 			     struct dlm_lock *lock)
201 {
202 	/* remove from local queue if it failed */
203 	list_del_init(&lock->list);
204 	lock->lksb->flags &= ~DLM_LKSB_GET_LVB;
205 }
206 
207 
208 /*
209  * locking:
210  *   caller needs:  none
211  *   taken:         takes and drops res->spinlock
212  *   held on exit:  none
213  * returns: DLM_DENIED, DLM_RECOVERING, or net status
214  */
215 static enum dlm_status dlmlock_remote(struct dlm_ctxt *dlm,
216 				      struct dlm_lock_resource *res,
217 				      struct dlm_lock *lock, int flags)
218 {
219 	enum dlm_status status = DLM_DENIED;
220 	int lockres_changed = 1;
221 
222 	mlog(0, "type=%d, lockres %.*s, flags = 0x%x\n",
223 	     lock->ml.type, res->lockname.len,
224 	     res->lockname.name, flags);
225 
226 	/*
227 	 * Wait if resource is getting recovered, remastered, etc.
228 	 * If the resource was remastered and new owner is self, then exit.
229 	 */
230 	spin_lock(&res->spinlock);
231 	__dlm_wait_on_lockres(res);
232 	if (res->owner == dlm->node_num) {
233 		spin_unlock(&res->spinlock);
234 		return DLM_RECOVERING;
235 	}
236 	res->state |= DLM_LOCK_RES_IN_PROGRESS;
237 
238 	/* add lock to local (secondary) queue */
239 	dlm_lock_get(lock);
240 	list_add_tail(&lock->list, &res->blocked);
241 	lock->lock_pending = 1;
242 	spin_unlock(&res->spinlock);
243 
244 	/* spec seems to say that you will get DLM_NORMAL when the lock
245 	 * has been queued, meaning we need to wait for a reply here. */
246 	status = dlm_send_remote_lock_request(dlm, res, lock, flags);
247 
248 	spin_lock(&res->spinlock);
249 	res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
250 	lock->lock_pending = 0;
251 	if (status != DLM_NORMAL) {
252 		if (status == DLM_RECOVERING &&
253 		    dlm_is_recovery_lock(res->lockname.name,
254 					 res->lockname.len)) {
255 			/* recovery lock was mastered by dead node.
256 			 * we need to have calc_usage shoot down this
257 			 * lockres and completely remaster it. */
258 			mlog(0, "%s: recovery lock was owned by "
259 			     "dead node %u, remaster it now.\n",
260 			     dlm->name, res->owner);
261 		} else if (status != DLM_NOTQUEUED) {
262 			/*
263 			 * DO NOT call calc_usage, as this would unhash
264 			 * the remote lockres before we ever get to use
265 			 * it.  treat as if we never made any change to
266 			 * the lockres.
267 			 */
268 			lockres_changed = 0;
269 			dlm_error(status);
270 		}
271 		dlm_revert_pending_lock(res, lock);
272 		dlm_lock_put(lock);
273 	} else if (dlm_is_recovery_lock(res->lockname.name,
274 					res->lockname.len)) {
275 		/* special case for the $RECOVERY lock.
276 		 * there will never be an AST delivered to put
277 		 * this lock on the proper secondary queue
278 		 * (granted), so do it manually. */
279 		mlog(0, "%s: $RECOVERY lock for this node (%u) is "
280 		     "mastered by %u; got lock, manually granting (no ast)\n",
281 		     dlm->name, dlm->node_num, res->owner);
282 		list_move_tail(&lock->list, &res->granted);
283 	}
284 	spin_unlock(&res->spinlock);
285 
286 	if (lockres_changed)
287 		dlm_lockres_calc_usage(dlm, res);
288 
289 	wake_up(&res->wq);
290 	return status;
291 }
292 
293 
294 /* for remote lock creation.
295  * locking:
296  *   caller needs:  none, but need res->state & DLM_LOCK_RES_IN_PROGRESS
297  *   taken:         none
298  *   held on exit:  none
299  * returns: DLM_NOLOCKMGR, or net status
300  */
301 static enum dlm_status dlm_send_remote_lock_request(struct dlm_ctxt *dlm,
302 					       struct dlm_lock_resource *res,
303 					       struct dlm_lock *lock, int flags)
304 {
305 	struct dlm_create_lock create;
306 	int tmpret, status = 0;
307 	enum dlm_status ret;
308 
309 	memset(&create, 0, sizeof(create));
310 	create.node_idx = dlm->node_num;
311 	create.requested_type = lock->ml.type;
312 	create.cookie = lock->ml.cookie;
313 	create.namelen = res->lockname.len;
314 	create.flags = cpu_to_be32(flags);
315 	memcpy(create.name, res->lockname.name, create.namelen);
316 
317 	tmpret = o2net_send_message(DLM_CREATE_LOCK_MSG, dlm->key, &create,
318 				    sizeof(create), res->owner, &status);
319 	if (tmpret >= 0) {
320 		ret = status;
321 		if (ret == DLM_REJECTED) {
322 			mlog(ML_ERROR, "%s: res %.*s, Stale lockres no longer "
323 			     "owned by node %u. That node is coming back up "
324 			     "currently.\n", dlm->name, create.namelen,
325 			     create.name, res->owner);
326 			dlm_print_one_lock_resource(res);
327 			BUG();
328 		}
329 	} else {
330 		mlog(ML_ERROR, "%s: res %.*s, Error %d send CREATE LOCK to "
331 		     "node %u\n", dlm->name, create.namelen, create.name,
332 		     tmpret, res->owner);
333 		if (dlm_is_host_down(tmpret))
334 			ret = DLM_RECOVERING;
335 		else
336 			ret = dlm_err_to_dlm_status(tmpret);
337 	}
338 
339 	return ret;
340 }
341 
342 void dlm_lock_get(struct dlm_lock *lock)
343 {
344 	kref_get(&lock->lock_refs);
345 }
346 
347 void dlm_lock_put(struct dlm_lock *lock)
348 {
349 	kref_put(&lock->lock_refs, dlm_lock_release);
350 }
351 
352 static void dlm_lock_release(struct kref *kref)
353 {
354 	struct dlm_lock *lock;
355 
356 	lock = container_of(kref, struct dlm_lock, lock_refs);
357 
358 	BUG_ON(!list_empty(&lock->list));
359 	BUG_ON(!list_empty(&lock->ast_list));
360 	BUG_ON(!list_empty(&lock->bast_list));
361 	BUG_ON(lock->ast_pending);
362 	BUG_ON(lock->bast_pending);
363 
364 	dlm_lock_detach_lockres(lock);
365 
366 	if (lock->lksb_kernel_allocated) {
367 		mlog(0, "freeing kernel-allocated lksb\n");
368 		kfree(lock->lksb);
369 	}
370 	kmem_cache_free(dlm_lock_cache, lock);
371 }
372 
373 /* associate a lock with it's lockres, getting a ref on the lockres */
374 void dlm_lock_attach_lockres(struct dlm_lock *lock,
375 			     struct dlm_lock_resource *res)
376 {
377 	dlm_lockres_get(res);
378 	lock->lockres = res;
379 }
380 
381 /* drop ref on lockres, if there is still one associated with lock */
382 static void dlm_lock_detach_lockres(struct dlm_lock *lock)
383 {
384 	struct dlm_lock_resource *res;
385 
386 	res = lock->lockres;
387 	if (res) {
388 		lock->lockres = NULL;
389 		mlog(0, "removing lock's lockres reference\n");
390 		dlm_lockres_put(res);
391 	}
392 }
393 
394 static void dlm_init_lock(struct dlm_lock *newlock, int type,
395 			  u8 node, u64 cookie)
396 {
397 	INIT_LIST_HEAD(&newlock->list);
398 	INIT_LIST_HEAD(&newlock->ast_list);
399 	INIT_LIST_HEAD(&newlock->bast_list);
400 	spin_lock_init(&newlock->spinlock);
401 	newlock->ml.type = type;
402 	newlock->ml.convert_type = LKM_IVMODE;
403 	newlock->ml.highest_blocked = LKM_IVMODE;
404 	newlock->ml.node = node;
405 	newlock->ml.pad1 = 0;
406 	newlock->ml.list = 0;
407 	newlock->ml.flags = 0;
408 	newlock->ast = NULL;
409 	newlock->bast = NULL;
410 	newlock->astdata = NULL;
411 	newlock->ml.cookie = cpu_to_be64(cookie);
412 	newlock->ast_pending = 0;
413 	newlock->bast_pending = 0;
414 	newlock->convert_pending = 0;
415 	newlock->lock_pending = 0;
416 	newlock->unlock_pending = 0;
417 	newlock->cancel_pending = 0;
418 	newlock->lksb_kernel_allocated = 0;
419 
420 	kref_init(&newlock->lock_refs);
421 }
422 
423 struct dlm_lock * dlm_new_lock(int type, u8 node, u64 cookie,
424 			       struct dlm_lockstatus *lksb)
425 {
426 	struct dlm_lock *lock;
427 	int kernel_allocated = 0;
428 
429 	lock = kmem_cache_zalloc(dlm_lock_cache, GFP_NOFS);
430 	if (!lock)
431 		return NULL;
432 
433 	if (!lksb) {
434 		/* zero memory only if kernel-allocated */
435 		lksb = kzalloc(sizeof(*lksb), GFP_NOFS);
436 		if (!lksb) {
437 			kmem_cache_free(dlm_lock_cache, lock);
438 			return NULL;
439 		}
440 		kernel_allocated = 1;
441 	}
442 
443 	dlm_init_lock(lock, type, node, cookie);
444 	if (kernel_allocated)
445 		lock->lksb_kernel_allocated = 1;
446 	lock->lksb = lksb;
447 	lksb->lockid = lock;
448 	return lock;
449 }
450 
451 /* handler for lock creation net message
452  * locking:
453  *   caller needs:  none
454  *   taken:         takes and drops res->spinlock
455  *   held on exit:  none
456  * returns: DLM_NORMAL, DLM_SYSERR, DLM_IVLOCKID, DLM_NOTQUEUED
457  */
458 int dlm_create_lock_handler(struct o2net_msg *msg, u32 len, void *data,
459 			    void **ret_data)
460 {
461 	struct dlm_ctxt *dlm = data;
462 	struct dlm_create_lock *create = (struct dlm_create_lock *)msg->buf;
463 	struct dlm_lock_resource *res = NULL;
464 	struct dlm_lock *newlock = NULL;
465 	struct dlm_lockstatus *lksb = NULL;
466 	enum dlm_status status = DLM_NORMAL;
467 	char *name;
468 	unsigned int namelen;
469 
470 	BUG_ON(!dlm);
471 
472 	if (!dlm_grab(dlm))
473 		return DLM_REJECTED;
474 
475 	name = create->name;
476 	namelen = create->namelen;
477 	status = DLM_REJECTED;
478 	if (!dlm_domain_fully_joined(dlm)) {
479 		mlog(ML_ERROR, "Domain %s not fully joined, but node %u is "
480 		     "sending a create_lock message for lock %.*s!\n",
481 		     dlm->name, create->node_idx, namelen, name);
482 		dlm_error(status);
483 		goto leave;
484 	}
485 
486 	status = DLM_IVBUFLEN;
487 	if (namelen > DLM_LOCKID_NAME_MAX) {
488 		dlm_error(status);
489 		goto leave;
490 	}
491 
492 	status = DLM_SYSERR;
493 	newlock = dlm_new_lock(create->requested_type,
494 			       create->node_idx,
495 			       be64_to_cpu(create->cookie), NULL);
496 	if (!newlock) {
497 		dlm_error(status);
498 		goto leave;
499 	}
500 
501 	lksb = newlock->lksb;
502 
503 	if (be32_to_cpu(create->flags) & LKM_GET_LVB) {
504 		lksb->flags |= DLM_LKSB_GET_LVB;
505 		mlog(0, "set DLM_LKSB_GET_LVB flag\n");
506 	}
507 
508 	status = DLM_IVLOCKID;
509 	res = dlm_lookup_lockres(dlm, name, namelen);
510 	if (!res) {
511 		dlm_error(status);
512 		goto leave;
513 	}
514 
515 	spin_lock(&res->spinlock);
516 	status = __dlm_lockres_state_to_status(res);
517 	spin_unlock(&res->spinlock);
518 
519 	if (status != DLM_NORMAL) {
520 		mlog(0, "lockres recovering/migrating/in-progress\n");
521 		goto leave;
522 	}
523 
524 	dlm_lock_attach_lockres(newlock, res);
525 
526 	status = dlmlock_master(dlm, res, newlock, be32_to_cpu(create->flags));
527 leave:
528 	if (status != DLM_NORMAL)
529 		if (newlock)
530 			dlm_lock_put(newlock);
531 
532 	if (res)
533 		dlm_lockres_put(res);
534 
535 	dlm_put(dlm);
536 
537 	return status;
538 }
539 
540 
541 /* fetch next node-local (u8 nodenum + u56 cookie) into u64 */
542 static inline void dlm_get_next_cookie(u8 node_num, u64 *cookie)
543 {
544 	u64 tmpnode = node_num;
545 
546 	/* shift single byte of node num into top 8 bits */
547 	tmpnode <<= 56;
548 
549 	spin_lock(&dlm_cookie_lock);
550 	*cookie = (dlm_next_cookie | tmpnode);
551 	if (++dlm_next_cookie & 0xff00000000000000ull) {
552 		mlog(0, "This node's cookie will now wrap!\n");
553 		dlm_next_cookie = 1;
554 	}
555 	spin_unlock(&dlm_cookie_lock);
556 }
557 
558 enum dlm_status dlmlock(struct dlm_ctxt *dlm, int mode,
559 			struct dlm_lockstatus *lksb, int flags,
560 			const char *name, int namelen, dlm_astlockfunc_t *ast,
561 			void *data, dlm_bastlockfunc_t *bast)
562 {
563 	enum dlm_status status;
564 	struct dlm_lock_resource *res = NULL;
565 	struct dlm_lock *lock = NULL;
566 	int convert = 0, recovery = 0;
567 
568 	/* yes this function is a mess.
569 	 * TODO: clean this up.  lots of common code in the
570 	 *       lock and convert paths, especially in the retry blocks */
571 	if (!lksb) {
572 		dlm_error(DLM_BADARGS);
573 		return DLM_BADARGS;
574 	}
575 
576 	status = DLM_BADPARAM;
577 	if (mode != LKM_EXMODE && mode != LKM_PRMODE && mode != LKM_NLMODE) {
578 		dlm_error(status);
579 		goto error;
580 	}
581 
582 	if (flags & ~LKM_VALID_FLAGS) {
583 		dlm_error(status);
584 		goto error;
585 	}
586 
587 	convert = (flags & LKM_CONVERT);
588 	recovery = (flags & LKM_RECOVERY);
589 
590 	if (recovery &&
591 	    (!dlm_is_recovery_lock(name, namelen) || convert) ) {
592 		dlm_error(status);
593 		goto error;
594 	}
595 	if (convert && (flags & LKM_LOCAL)) {
596 		mlog(ML_ERROR, "strange LOCAL convert request!\n");
597 		goto error;
598 	}
599 
600 	if (convert) {
601 		/* CONVERT request */
602 
603 		/* if converting, must pass in a valid dlm_lock */
604 		lock = lksb->lockid;
605 		if (!lock) {
606 			mlog(ML_ERROR, "NULL lock pointer in convert "
607 			     "request\n");
608 			goto error;
609 		}
610 
611 		res = lock->lockres;
612 		if (!res) {
613 			mlog(ML_ERROR, "NULL lockres pointer in convert "
614 			     "request\n");
615 			goto error;
616 		}
617 		dlm_lockres_get(res);
618 
619 		/* XXX: for ocfs2 purposes, the ast/bast/astdata/lksb are
620 	 	 * static after the original lock call.  convert requests will
621 		 * ensure that everything is the same, or return DLM_BADARGS.
622 	 	 * this means that DLM_DENIED_NOASTS will never be returned.
623 	 	 */
624 		if (lock->lksb != lksb || lock->ast != ast ||
625 		    lock->bast != bast || lock->astdata != data) {
626 			status = DLM_BADARGS;
627 			mlog(ML_ERROR, "new args:  lksb=%p, ast=%p, bast=%p, "
628 			     "astdata=%p\n", lksb, ast, bast, data);
629 			mlog(ML_ERROR, "orig args: lksb=%p, ast=%p, bast=%p, "
630 			     "astdata=%p\n", lock->lksb, lock->ast,
631 			     lock->bast, lock->astdata);
632 			goto error;
633 		}
634 retry_convert:
635 		dlm_wait_for_recovery(dlm);
636 
637 		if (res->owner == dlm->node_num)
638 			status = dlmconvert_master(dlm, res, lock, flags, mode);
639 		else
640 			status = dlmconvert_remote(dlm, res, lock, flags, mode);
641 		if (status == DLM_RECOVERING || status == DLM_MIGRATING ||
642 		    status == DLM_FORWARD) {
643 			/* for now, see how this works without sleeping
644 			 * and just retry right away.  I suspect the reco
645 			 * or migration will complete fast enough that
646 			 * no waiting will be necessary */
647 			mlog(0, "retrying convert with migration/recovery/"
648 			     "in-progress\n");
649 			msleep(100);
650 			goto retry_convert;
651 		}
652 	} else {
653 		u64 tmpcookie;
654 
655 		/* LOCK request */
656 		status = DLM_BADARGS;
657 		if (!name) {
658 			dlm_error(status);
659 			goto error;
660 		}
661 
662 		status = DLM_IVBUFLEN;
663 		if (namelen > DLM_LOCKID_NAME_MAX || namelen < 1) {
664 			dlm_error(status);
665 			goto error;
666 		}
667 
668 		dlm_get_next_cookie(dlm->node_num, &tmpcookie);
669 		lock = dlm_new_lock(mode, dlm->node_num, tmpcookie, lksb);
670 		if (!lock) {
671 			dlm_error(status);
672 			goto error;
673 		}
674 
675 		if (!recovery)
676 			dlm_wait_for_recovery(dlm);
677 
678 		/* find or create the lock resource */
679 		res = dlm_get_lock_resource(dlm, name, namelen, flags);
680 		if (!res) {
681 			status = DLM_IVLOCKID;
682 			dlm_error(status);
683 			goto error;
684 		}
685 
686 		mlog(0, "type=%d, flags = 0x%x\n", mode, flags);
687 		mlog(0, "creating lock: lock=%p res=%p\n", lock, res);
688 
689 		dlm_lock_attach_lockres(lock, res);
690 		lock->ast = ast;
691 		lock->bast = bast;
692 		lock->astdata = data;
693 
694 retry_lock:
695 		if (flags & LKM_VALBLK) {
696 			mlog(0, "LKM_VALBLK passed by caller\n");
697 
698 			/* LVB requests for non PR, PW or EX locks are
699 			 * ignored. */
700 			if (mode < LKM_PRMODE)
701 				flags &= ~LKM_VALBLK;
702 			else {
703 				flags |= LKM_GET_LVB;
704 				lock->lksb->flags |= DLM_LKSB_GET_LVB;
705 			}
706 		}
707 
708 		if (res->owner == dlm->node_num)
709 			status = dlmlock_master(dlm, res, lock, flags);
710 		else
711 			status = dlmlock_remote(dlm, res, lock, flags);
712 
713 		if (status == DLM_RECOVERING || status == DLM_MIGRATING ||
714 		    status == DLM_FORWARD) {
715 			msleep(100);
716 			if (recovery) {
717 				if (status != DLM_RECOVERING)
718 					goto retry_lock;
719 				/* wait to see the node go down, then
720 				 * drop down and allow the lockres to
721 				 * get cleaned up.  need to remaster. */
722 				dlm_wait_for_node_death(dlm, res->owner,
723 						DLM_NODE_DEATH_WAIT_MAX);
724 			} else {
725 				dlm_wait_for_recovery(dlm);
726 				goto retry_lock;
727 			}
728 		}
729 
730 		/* Inflight taken in dlm_get_lock_resource() is dropped here */
731 		spin_lock(&res->spinlock);
732 		dlm_lockres_drop_inflight_ref(dlm, res);
733 		spin_unlock(&res->spinlock);
734 
735 		dlm_lockres_calc_usage(dlm, res);
736 		dlm_kick_thread(dlm, res);
737 
738 		if (status != DLM_NORMAL) {
739 			lock->lksb->flags &= ~DLM_LKSB_GET_LVB;
740 			if (status != DLM_NOTQUEUED)
741 				dlm_error(status);
742 			goto error;
743 		}
744 	}
745 
746 error:
747 	if (status != DLM_NORMAL) {
748 		if (lock && !convert)
749 			dlm_lock_put(lock);
750 		// this is kind of unnecessary
751 		lksb->status = status;
752 	}
753 
754 	/* put lockres ref from the convert path
755 	 * or from dlm_get_lock_resource */
756 	if (res)
757 		dlm_lockres_put(res);
758 
759 	return status;
760 }
761 EXPORT_SYMBOL_GPL(dlmlock);
762