plock.c 11.6 KB
Newer Older
1
/*
2
 * Copyright (C) 2005-2008 Red Hat, Inc.  All rights reserved.
3 4 5
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
6
 * of the GNU General Public License version 2.
7 8
 */

9
#include <linux/fs.h>
10
#include <linux/miscdevice.h>
11
#include <linux/poll.h>
12 13
#include <linux/dlm.h>
#include <linux/dlm_plock.h>
14
#include <linux/slab.h>
15

16 17
#include "dlm_internal.h"
#include "lockspace.h"
18 19 20 21 22 23 24 25 26 27

static spinlock_t ops_lock;
static struct list_head send_list;
static struct list_head recv_list;
static wait_queue_head_t send_wq;
static wait_queue_head_t recv_wq;

struct plock_op {
	struct list_head list;
	int done;
28
	struct dlm_plock_info info;
29 30
};

Marc Eshel's avatar
Marc Eshel committed
31 32
struct plock_xop {
	struct plock_op xop;
33
	int (*callback)(struct file_lock *fl, int result);
Marc Eshel's avatar
Marc Eshel committed
34 35 36 37 38 39
	void *fl;
	void *file;
	struct file_lock flc;
};


40
static inline void set_version(struct dlm_plock_info *info)
41
{
42 43 44
	info->version[0] = DLM_PLOCK_VERSION_MAJOR;
	info->version[1] = DLM_PLOCK_VERSION_MINOR;
	info->version[2] = DLM_PLOCK_VERSION_PATCH;
45 46
}

47
static int check_version(struct dlm_plock_info *info)
48
{
49 50 51
	if ((DLM_PLOCK_VERSION_MAJOR != info->version[0]) ||
	    (DLM_PLOCK_VERSION_MINOR < info->version[1])) {
		log_print("plock device version mismatch: "
52
			  "kernel (%u.%u.%u), user (%u.%u.%u)",
53 54 55
			  DLM_PLOCK_VERSION_MAJOR,
			  DLM_PLOCK_VERSION_MINOR,
			  DLM_PLOCK_VERSION_PATCH,
56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
			  info->version[0],
			  info->version[1],
			  info->version[2]);
		return -EINVAL;
	}
	return 0;
}

static void send_op(struct plock_op *op)
{
	set_version(&op->info);
	INIT_LIST_HEAD(&op->list);
	spin_lock(&ops_lock);
	list_add_tail(&op->list, &send_list);
	spin_unlock(&ops_lock);
	wake_up(&send_wq);
}

74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94
/* If a process was killed while waiting for the only plock on a file,
   locks_remove_posix will not see any lock on the file so it won't
   send an unlock-close to us to pass on to userspace to clean up the
   abandoned waiter.  So, we have to insert the unlock-close when the
   lock call is interrupted. */

static void do_unlock_close(struct dlm_ls *ls, u64 number,
			    struct file *file, struct file_lock *fl)
{
	struct plock_op *op;

	op = kzalloc(sizeof(*op), GFP_NOFS);
	if (!op)
		return;

	op->info.optype		= DLM_PLOCK_OP_UNLOCK;
	op->info.pid		= fl->fl_pid;
	op->info.fsid		= ls->ls_global_id;
	op->info.number		= number;
	op->info.start		= 0;
	op->info.end		= OFFSET_MAX;
95
	if (fl->fl_lmops && fl->fl_lmops->lm_grant)
96 97 98 99 100 101 102 103
		op->info.owner	= (__u64) fl->fl_pid;
	else
		op->info.owner	= (__u64)(long) fl->fl_owner;

	op->info.flags |= DLM_PLOCK_FL_CLOSE;
	send_op(op);
}

104 105
int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
		   int cmd, struct file_lock *fl)
106
{
107
	struct dlm_ls *ls;
108
	struct plock_op *op;
Marc Eshel's avatar
Marc Eshel committed
109
	struct plock_xop *xop;
110 111
	int rv;

112 113 114 115
	ls = dlm_find_lockspace_local(lockspace);
	if (!ls)
		return -EINVAL;

David Teigland's avatar
David Teigland committed
116
	xop = kzalloc(sizeof(*xop), GFP_NOFS);
117 118 119 120
	if (!xop) {
		rv = -ENOMEM;
		goto out;
	}
121

Marc Eshel's avatar
Marc Eshel committed
122
	op = &xop->xop;
123
	op->info.optype		= DLM_PLOCK_OP_LOCK;
124
	op->info.pid		= fl->fl_pid;
125 126
	op->info.ex		= (fl->fl_type == F_WRLCK);
	op->info.wait		= IS_SETLKW(cmd);
127 128
	op->info.fsid		= ls->ls_global_id;
	op->info.number		= number;
129 130
	op->info.start		= fl->fl_start;
	op->info.end		= fl->fl_end;
131
	if (fl->fl_lmops && fl->fl_lmops->lm_grant) {
132 133 134
		/* fl_owner is lockd which doesn't distinguish
		   processes on the nfs client */
		op->info.owner	= (__u64) fl->fl_pid;
135
		xop->callback	= fl->fl_lmops->lm_grant;
Marc Eshel's avatar
Marc Eshel committed
136 137 138 139
		locks_init_lock(&xop->flc);
		locks_copy_lock(&xop->flc, fl);
		xop->fl		= fl;
		xop->file	= file;
140 141
	} else {
		op->info.owner	= (__u64)(long) fl->fl_owner;
Marc Eshel's avatar
Marc Eshel committed
142
		xop->callback	= NULL;
143
	}
144 145

	send_op(op);
Marc Eshel's avatar
Marc Eshel committed
146

147
	if (xop->callback == NULL) {
148
		rv = wait_event_interruptible(recv_wq, (op->done != 0));
149 150 151 152 153 154 155 156 157 158 159
		if (rv == -ERESTARTSYS) {
			log_debug(ls, "dlm_posix_lock: wait killed %llx",
				  (unsigned long long)number);
			spin_lock(&ops_lock);
			list_del(&op->list);
			spin_unlock(&ops_lock);
			kfree(xop);
			do_unlock_close(ls, number, file, fl);
			goto out;
		}
	} else {
160
		rv = FILE_LOCK_DEFERRED;
161 162
		goto out;
	}
163 164 165

	spin_lock(&ops_lock);
	if (!list_empty(&op->list)) {
166 167
		log_error(ls, "dlm_posix_lock: op on list %llx",
			  (unsigned long long)number);
168 169 170 171 172 173 174
		list_del(&op->list);
	}
	spin_unlock(&ops_lock);

	rv = op->info.rv;

	if (!rv) {
175
		if (locks_lock_file_wait(file, fl) < 0)
176 177
			log_error(ls, "dlm_posix_lock: vfs lock error %llx",
				  (unsigned long long)number);
178 179
	}

Marc Eshel's avatar
Marc Eshel committed
180
	kfree(xop);
181 182
out:
	dlm_put_lockspace(ls);
Marc Eshel's avatar
Marc Eshel committed
183 184
	return rv;
}
185
EXPORT_SYMBOL_GPL(dlm_posix_lock);
Marc Eshel's avatar
Marc Eshel committed
186

187
/* Returns failure iff a successful lock operation should be canceled */
188
static int dlm_plock_callback(struct plock_op *op)
Marc Eshel's avatar
Marc Eshel committed
189 190 191 192
{
	struct file *file;
	struct file_lock *fl;
	struct file_lock *flc;
193
	int (*notify)(struct file_lock *fl, int result) = NULL;
Marc Eshel's avatar
Marc Eshel committed
194 195 196 197 198
	struct plock_xop *xop = (struct plock_xop *)op;
	int rv = 0;

	spin_lock(&ops_lock);
	if (!list_empty(&op->list)) {
199 200
		log_print("dlm_plock_callback: op on list %llx",
			  (unsigned long long)op->info.number);
Marc Eshel's avatar
Marc Eshel committed
201 202 203 204 205 206 207 208 209 210 211
		list_del(&op->list);
	}
	spin_unlock(&ops_lock);

	/* check if the following 2 are still valid or make a copy */
	file = xop->file;
	flc = &xop->flc;
	fl = xop->fl;
	notify = xop->callback;

	if (op->info.rv) {
212
		notify(fl, op->info.rv);
Marc Eshel's avatar
Marc Eshel committed
213 214 215 216 217 218 219 220 221 222
		goto out;
	}

	/* got fs lock; bookkeep locally as well: */
	flc->fl_flags &= ~FL_SLEEP;
	if (posix_lock_file(file, flc, NULL)) {
		/*
		 * This can only happen in the case of kmalloc() failure.
		 * The filesystem's own lock is the authoritative lock,
		 * so a failure to get the lock locally is not a disaster.
223
		 * As long as the fs cannot reliably cancel locks (especially
Marc Eshel's avatar
Marc Eshel committed
224 225 226
		 * in a low-memory situation), we're better off ignoring
		 * this failure than trying to recover.
		 */
227 228
		log_print("dlm_plock_callback: vfs lock error %llx file %p fl %p",
			  (unsigned long long)op->info.number, file, fl);
Marc Eshel's avatar
Marc Eshel committed
229 230
	}

231
	rv = notify(fl, 0);
Marc Eshel's avatar
Marc Eshel committed
232 233
	if (rv) {
		/* XXX: We need to cancel the fs lock here: */
234 235
		log_print("dlm_plock_callback: lock granted after lock request "
			  "failed; dangling lock!\n");
Marc Eshel's avatar
Marc Eshel committed
236 237 238 239 240
		goto out;
	}

out:
	kfree(xop);
241 242 243
	return rv;
}

244 245
int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
		     struct file_lock *fl)
246
{
247
	struct dlm_ls *ls;
248 249
	struct plock_op *op;
	int rv;
250
	unsigned char fl_flags = fl->fl_flags;
251

252 253 254 255
	ls = dlm_find_lockspace_local(lockspace);
	if (!ls)
		return -EINVAL;

David Teigland's avatar
David Teigland committed
256
	op = kzalloc(sizeof(*op), GFP_NOFS);
257 258 259 260
	if (!op) {
		rv = -ENOMEM;
		goto out;
	}
261

262 263 264
	/* cause the vfs unlock to return ENOENT if lock is not found */
	fl->fl_flags |= FL_EXISTS;

265
	rv = locks_lock_file_wait(file, fl);
266 267 268 269 270 271 272 273
	if (rv == -ENOENT) {
		rv = 0;
		goto out_free;
	}
	if (rv < 0) {
		log_error(ls, "dlm_posix_unlock: vfs unlock error %d %llx",
			  rv, (unsigned long long)number);
	}
274

275
	op->info.optype		= DLM_PLOCK_OP_UNLOCK;
276
	op->info.pid		= fl->fl_pid;
277 278
	op->info.fsid		= ls->ls_global_id;
	op->info.number		= number;
279 280
	op->info.start		= fl->fl_start;
	op->info.end		= fl->fl_end;
281
	if (fl->fl_lmops && fl->fl_lmops->lm_grant)
282 283 284
		op->info.owner	= (__u64) fl->fl_pid;
	else
		op->info.owner	= (__u64)(long) fl->fl_owner;
285

286 287 288 289 290 291 292
	if (fl->fl_flags & FL_CLOSE) {
		op->info.flags |= DLM_PLOCK_FL_CLOSE;
		send_op(op);
		rv = 0;
		goto out;
	}

293 294 295 296 297
	send_op(op);
	wait_event(recv_wq, (op->done != 0));

	spin_lock(&ops_lock);
	if (!list_empty(&op->list)) {
298 299
		log_error(ls, "dlm_posix_unlock: op on list %llx",
			  (unsigned long long)number);
300 301 302 303 304 305
		list_del(&op->list);
	}
	spin_unlock(&ops_lock);

	rv = op->info.rv;

Marc Eshel's avatar
Marc Eshel committed
306 307 308
	if (rv == -ENOENT)
		rv = 0;

309
out_free:
310
	kfree(op);
311 312
out:
	dlm_put_lockspace(ls);
313
	fl->fl_flags = fl_flags;
314 315
	return rv;
}
316
EXPORT_SYMBOL_GPL(dlm_posix_unlock);
317

318 319
int dlm_posix_get(dlm_lockspace_t *lockspace, u64 number, struct file *file,
		  struct file_lock *fl)
320
{
321
	struct dlm_ls *ls;
322 323 324
	struct plock_op *op;
	int rv;

325 326 327 328
	ls = dlm_find_lockspace_local(lockspace);
	if (!ls)
		return -EINVAL;

David Teigland's avatar
David Teigland committed
329
	op = kzalloc(sizeof(*op), GFP_NOFS);
330 331 332 333
	if (!op) {
		rv = -ENOMEM;
		goto out;
	}
334

335
	op->info.optype		= DLM_PLOCK_OP_GET;
336
	op->info.pid		= fl->fl_pid;
337
	op->info.ex		= (fl->fl_type == F_WRLCK);
338 339
	op->info.fsid		= ls->ls_global_id;
	op->info.number		= number;
340 341
	op->info.start		= fl->fl_start;
	op->info.end		= fl->fl_end;
342
	if (fl->fl_lmops && fl->fl_lmops->lm_grant)
343 344 345
		op->info.owner	= (__u64) fl->fl_pid;
	else
		op->info.owner	= (__u64)(long) fl->fl_owner;
Marc Eshel's avatar
Marc Eshel committed
346

347 348 349 350 351
	send_op(op);
	wait_event(recv_wq, (op->done != 0));

	spin_lock(&ops_lock);
	if (!list_empty(&op->list)) {
352 353
		log_error(ls, "dlm_posix_get: op on list %llx",
			  (unsigned long long)number);
354 355 356 357
		list_del(&op->list);
	}
	spin_unlock(&ops_lock);

358 359 360
	/* info.rv from userspace is 1 for conflict, 0 for no-conflict,
	   -ENOENT if there are no locks on the file */

361 362
	rv = op->info.rv;

Marc Eshel's avatar
Marc Eshel committed
363 364 365
	fl->fl_type = F_UNLCK;
	if (rv == -ENOENT)
		rv = 0;
366
	else if (rv > 0) {
367
		locks_init_lock(fl);
368
		fl->fl_type = (op->info.ex) ? F_WRLCK : F_RDLCK;
369
		fl->fl_flags = FL_POSIX;
370
		fl->fl_pid = -op->info.pid;
371 372
		fl->fl_start = op->info.start;
		fl->fl_end = op->info.end;
373
		rv = 0;
374 375 376
	}

	kfree(op);
377 378
out:
	dlm_put_lockspace(ls);
379 380
	return rv;
}
381
EXPORT_SYMBOL_GPL(dlm_posix_get);
382 383 384 385 386

/* a read copies out one plock request from the send list */
static ssize_t dev_read(struct file *file, char __user *u, size_t count,
			loff_t *ppos)
{
387
	struct dlm_plock_info info;
388 389 390 391 392 393 394 395
	struct plock_op *op = NULL;

	if (count < sizeof(info))
		return -EINVAL;

	spin_lock(&ops_lock);
	if (!list_empty(&send_list)) {
		op = list_entry(send_list.next, struct plock_op, list);
396 397 398 399
		if (op->info.flags & DLM_PLOCK_FL_CLOSE)
			list_del(&op->list);
		else
			list_move(&op->list, &recv_list);
400 401 402 403 404 405 406
		memcpy(&info, &op->info, sizeof(info));
	}
	spin_unlock(&ops_lock);

	if (!op)
		return -EAGAIN;

407 408 409 410 411 412 413
	/* there is no need to get a reply from userspace for unlocks
	   that were generated by the vfs cleaning up for a close
	   (the process did not make an unlock call). */

	if (op->info.flags & DLM_PLOCK_FL_CLOSE)
		kfree(op);

414 415 416 417 418 419 420 421 422 423
	if (copy_to_user(u, &info, sizeof(info)))
		return -EFAULT;
	return sizeof(info);
}

/* a write copies in one plock result that should match a plock_op
   on the recv list */
static ssize_t dev_write(struct file *file, const char __user *u, size_t count,
			 loff_t *ppos)
{
424
	struct dlm_plock_info info;
425
	struct plock_op *op;
426
	int found = 0, do_callback = 0;
427 428 429 430 431 432 433 434 435 436 437 438

	if (count != sizeof(info))
		return -EINVAL;

	if (copy_from_user(&info, u, sizeof(info)))
		return -EFAULT;

	if (check_version(&info))
		return -EINVAL;

	spin_lock(&ops_lock);
	list_for_each_entry(op, &recv_list, list) {
439 440
		if (op->info.fsid == info.fsid &&
		    op->info.number == info.number &&
441
		    op->info.owner == info.owner) {
442
			struct plock_xop *xop = (struct plock_xop *)op;
443 444
			list_del_init(&op->list);
			memcpy(&op->info, &info, sizeof(info));
445 446 447 448 449
			if (xop->callback)
				do_callback = 1;
			else
				op->done = 1;
			found = 1;
450 451 452 453 454
			break;
		}
	}
	spin_unlock(&ops_lock);

Marc Eshel's avatar
Marc Eshel committed
455
	if (found) {
456
		if (do_callback)
457
			dlm_plock_callback(op);
Marc Eshel's avatar
Marc Eshel committed
458 459 460
		else
			wake_up(&recv_wq);
	} else
461 462
		log_print("dev_write no op %x %llx", info.fsid,
			  (unsigned long long)info.number);
463 464 465 466 467
	return count;
}

static unsigned int dev_poll(struct file *file, poll_table *wait)
{
468 469
	unsigned int mask = 0;

470 471 472
	poll_wait(file, &send_wq, wait);

	spin_lock(&ops_lock);
473 474
	if (!list_empty(&send_list))
		mask = POLLIN | POLLRDNORM;
475
	spin_unlock(&ops_lock);
476 477

	return mask;
478 479
}

480
static const struct file_operations dev_fops = {
481 482 483
	.read    = dev_read,
	.write   = dev_write,
	.poll    = dev_poll,
484 485
	.owner   = THIS_MODULE,
	.llseek  = noop_llseek,
486 487 488 489
};

static struct miscdevice plock_dev_misc = {
	.minor = MISC_DYNAMIC_MINOR,
490
	.name = DLM_PLOCK_MISC_NAME,
491 492 493
	.fops = &dev_fops
};

494
int dlm_plock_init(void)
495 496 497 498 499 500 501 502 503 504 505
{
	int rv;

	spin_lock_init(&ops_lock);
	INIT_LIST_HEAD(&send_list);
	INIT_LIST_HEAD(&recv_list);
	init_waitqueue_head(&send_wq);
	init_waitqueue_head(&recv_wq);

	rv = misc_register(&plock_dev_misc);
	if (rv)
506
		log_print("dlm_plock_init: misc_register failed %d", rv);
507 508 509
	return rv;
}

510
void dlm_plock_exit(void)
511
{
512
	misc_deregister(&plock_dev_misc);
513 514
}