summaryrefslogtreecommitdiff
path: root/fs/xfs/scrub/scrub.c
blob: 02d278b7d20b48b173b7e42287f4c8bdc09e2334 (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
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
// SPDX-License-Identifier: GPL-2.0+
/*
 * Copyright (C) 2017 Oracle.  All Rights Reserved.
 * Author: Darrick J. Wong <darrick.wong@oracle.com>
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_trans_resv.h"
#include "xfs_mount.h"
#include "xfs_defer.h"
#include "xfs_btree.h"
#include "xfs_bit.h"
#include "xfs_log_format.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_inode.h"
#include "xfs_icache.h"
#include "xfs_itable.h"
#include "xfs_alloc.h"
#include "xfs_alloc_btree.h"
#include "xfs_bmap.h"
#include "xfs_bmap_btree.h"
#include "xfs_ialloc.h"
#include "xfs_ialloc_btree.h"
#include "xfs_refcount.h"
#include "xfs_refcount_btree.h"
#include "xfs_rmap.h"
#include "xfs_rmap_btree.h"
#include "xfs_quota.h"
#include "xfs_qm.h"
#include "xfs_errortag.h"
#include "xfs_error.h"
#include "xfs_log.h"
#include "xfs_trans_priv.h"
#include "scrub/xfs_scrub.h"
#include "scrub/scrub.h"
#include "scrub/common.h"
#include "scrub/trace.h"
#include "scrub/btree.h"
#include "scrub/repair.h"

/*
 * Online Scrub and Repair
 *
 * Traditionally, XFS (the kernel driver) did not know how to check or
 * repair on-disk data structures.  That task was left to the xfs_check
 * and xfs_repair tools, both of which require taking the filesystem
 * offline for a thorough but time consuming examination.  Online
 * scrub & repair, on the other hand, enables us to check the metadata
 * for obvious errors while carefully stepping around the filesystem's
 * ongoing operations, locking rules, etc.
 *
 * Given that most XFS metadata consist of records stored in a btree,
 * most of the checking functions iterate the btree blocks themselves
 * looking for irregularities.  When a record block is encountered, each
 * record can be checked for obviously bad values.  Record values can
 * also be cross-referenced against other btrees to look for potential
 * misunderstandings between pieces of metadata.
 *
 * It is expected that the checkers responsible for per-AG metadata
 * structures will lock the AG headers (AGI, AGF, AGFL), iterate the
 * metadata structure, and perform any relevant cross-referencing before
 * unlocking the AG and returning the results to userspace.  These
 * scrubbers must not keep an AG locked for too long to avoid tying up
 * the block and inode allocators.
 *
 * Block maps and b-trees rooted in an inode present a special challenge
 * because they can involve extents from any AG.  The general scrubber
 * structure of lock -> check -> xref -> unlock still holds, but AG
 * locking order rules /must/ be obeyed to avoid deadlocks.  The
 * ordering rule, of course, is that we must lock in increasing AG
 * order.  Helper functions are provided to track which AG headers we've
 * already locked.  If we detect an imminent locking order violation, we
 * can signal a potential deadlock, in which case the scrubber can jump
 * out to the top level, lock all the AGs in order, and retry the scrub.
 *
 * For file data (directories, extended attributes, symlinks) scrub, we
 * can simply lock the inode and walk the data.  For btree data
 * (directories and attributes) we follow the same btree-scrubbing
 * strategy outlined previously to check the records.
 *
 * We use a bit of trickery with transactions to avoid buffer deadlocks
 * if there is a cycle in the metadata.  The basic problem is that
 * travelling down a btree involves locking the current buffer at each
 * tree level.  If a pointer should somehow point back to a buffer that
 * we've already examined, we will deadlock due to the second buffer
 * locking attempt.  Note however that grabbing a buffer in transaction
 * context links the locked buffer to the transaction.  If we try to
 * re-grab the buffer in the context of the same transaction, we avoid
 * the second lock attempt and continue.  Between the verifier and the
 * scrubber, something will notice that something is amiss and report
 * the corruption.  Therefore, each scrubber will allocate an empty
 * transaction, attach buffers to it, and cancel the transaction at the
 * end of the scrub run.  Cancelling a non-dirty transaction simply
 * unlocks the buffers.
 *
 * There are four pieces of data that scrub can communicate to
 * userspace.  The first is the error code (errno), which can be used to
 * communicate operational errors in performing the scrub.  There are
 * also three flags that can be set in the scrub context.  If the data
 * structure itself is corrupt, the CORRUPT flag will be set.  If
 * the metadata is correct but otherwise suboptimal, the PREEN flag
 * will be set.
 *
 * We perform secondary validation of filesystem metadata by
 * cross-referencing every record with all other available metadata.
 * For example, for block mapping extents, we verify that there are no
 * records in the free space and inode btrees corresponding to that
 * space extent and that there is a corresponding entry in the reverse
 * mapping btree.  Inconsistent metadata is noted by setting the
 * XCORRUPT flag; btree query function errors are noted by setting the
 * XFAIL flag and deleting the cursor to prevent further attempts to
 * cross-reference with a defective btree.
 *
 * If a piece of metadata proves corrupt or suboptimal, the userspace
 * program can ask the kernel to apply some tender loving care (TLC) to
 * the metadata object by setting the REPAIR flag and re-calling the
 * scrub ioctl.  "Corruption" is defined by metadata violating the
 * on-disk specification; operations cannot continue if the violation is
 * left untreated.  It is possible for XFS to continue if an object is
 * "suboptimal", however performance may be degraded.  Repairs are
 * usually performed by rebuilding the metadata entirely out of
 * redundant metadata.  Optimizing, on the other hand, can sometimes be
 * done without rebuilding entire structures.
 *
 * Generally speaking, the repair code has the following code structure:
 * Lock -> scrub -> repair -> commit -> re-lock -> re-scrub -> unlock.
 * The first check helps us figure out if we need to rebuild or simply
 * optimize the structure so that the rebuild knows what to do.  The
 * second check evaluates the completeness of the repair; that is what
 * is reported to userspace.
 *
 * A quick note on symbol prefixes:
 * - "xfs_" are general XFS symbols.
 * - "xchk_" are symbols related to metadata checking.
 * - "xrep_" are symbols related to metadata repair.
 * - "xfs_scrub_" are symbols that tie online fsck to the rest of XFS.
 */

/*
 * Scrub probe -- userspace uses this to probe if we're willing to scrub
 * or repair a given mountpoint.  This will be used by xfs_scrub to
 * probe the kernel's abilities to scrub (and repair) the metadata.  We
 * do this by validating the ioctl inputs from userspace, preparing the
 * filesystem for a scrub (or a repair) operation, and immediately
 * returning to userspace.  Userspace can use the returned errno and
 * structure state to decide (in broad terms) if scrub/repair are
 * supported by the running kernel.
 */
static int
xchk_probe(
	struct xfs_scrub	*sc)
{
	int			error = 0;

	if (xchk_should_terminate(sc, &error))
		return error;

	return 0;
}

/* Scrub setup and teardown */

/* Free all the resources and finish the transactions. */
STATIC int
xchk_teardown(
	struct xfs_scrub	*sc,
	struct xfs_inode	*ip_in,
	int			error)
{
	xchk_ag_free(sc, &sc->sa);
	if (sc->tp) {
		if (error == 0 && (sc->sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR))
			error = xfs_trans_commit(sc->tp);
		else
			xfs_trans_cancel(sc->tp);
		sc->tp = NULL;
	}
	if (sc->ip) {
		if (sc->ilock_flags)
			xfs_iunlock(sc->ip, sc->ilock_flags);
		if (sc->ip != ip_in &&
		    !xfs_internal_inum(sc->mp, sc->ip->i_ino))
			xfs_irele(sc->ip);
		sc->ip = NULL;
	}
	if (sc->flags & XCHK_HAS_QUOTAOFFLOCK) {
		mutex_unlock(&sc->mp->m_quotainfo->qi_quotaofflock);
		sc->flags &= ~XCHK_HAS_QUOTAOFFLOCK;
	}
	if (sc->buf) {
		kmem_free(sc->buf);
		sc->buf = NULL;
	}
	return error;
}

/* Scrubbing dispatch. */

static const struct xchk_meta_ops meta_scrub_ops[] = {
	[XFS_SCRUB_TYPE_PROBE] = {	/* ioctl presence test */
		.type	= ST_NONE,
		.setup	= xchk_setup_fs,
		.scrub	= xchk_probe,
		.repair = xrep_probe,
	},
	[XFS_SCRUB_TYPE_SB] = {		/* superblock */
		.type	= ST_PERAG,
		.setup	= xchk_setup_fs,
		.scrub	= xchk_superblock,
		.repair	= xrep_superblock,
	},
	[XFS_SCRUB_TYPE_AGF] = {	/* agf */
		.type	= ST_PERAG,
		.setup	= xchk_setup_fs,
		.scrub	= xchk_agf,
		.repair	= xrep_agf,
	},
	[XFS_SCRUB_TYPE_AGFL]= {	/* agfl */
		.type	= ST_PERAG,
		.setup	= xchk_setup_fs,
		.scrub	= xchk_agfl,
		.repair	= xrep_agfl,
	},
	[XFS_SCRUB_TYPE_AGI] = {	/* agi */
		.type	= ST_PERAG,
		.setup	= xchk_setup_fs,
		.scrub	= xchk_agi,
		.repair	= xrep_agi,
	},
	[XFS_SCRUB_TYPE_BNOBT] = {	/* bnobt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_allocbt,
		.scrub	= xchk_bnobt,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_CNTBT] = {	/* cntbt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_allocbt,
		.scrub	= xchk_cntbt,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_INOBT] = {	/* inobt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_iallocbt,
		.scrub	= xchk_inobt,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_FINOBT] = {	/* finobt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_iallocbt,
		.scrub	= xchk_finobt,
		.has	= xfs_sb_version_hasfinobt,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_RMAPBT] = {	/* rmapbt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_rmapbt,
		.scrub	= xchk_rmapbt,
		.has	= xfs_sb_version_hasrmapbt,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_REFCNTBT] = {	/* refcountbt */
		.type	= ST_PERAG,
		.setup	= xchk_setup_ag_refcountbt,
		.scrub	= xchk_refcountbt,
		.has	= xfs_sb_version_hasreflink,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_INODE] = {	/* inode record */
		.type	= ST_INODE,
		.setup	= xchk_setup_inode,
		.scrub	= xchk_inode,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_BMBTD] = {	/* inode data fork */
		.type	= ST_INODE,
		.setup	= xchk_setup_inode_bmap,
		.scrub	= xchk_bmap_data,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_BMBTA] = {	/* inode attr fork */
		.type	= ST_INODE,
		.setup	= xchk_setup_inode_bmap,
		.scrub	= xchk_bmap_attr,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_BMBTC] = {	/* inode CoW fork */
		.type	= ST_INODE,
		.setup	= xchk_setup_inode_bmap,
		.scrub	= xchk_bmap_cow,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_DIR] = {	/* directory */
		.type	= ST_INODE,
		.setup	= xchk_setup_directory,
		.scrub	= xchk_directory,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_XATTR] = {	/* extended attributes */
		.type	= ST_INODE,
		.setup	= xchk_setup_xattr,
		.scrub	= xchk_xattr,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_SYMLINK] = {	/* symbolic link */
		.type	= ST_INODE,
		.setup	= xchk_setup_symlink,
		.scrub	= xchk_symlink,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_PARENT] = {	/* parent pointers */
		.type	= ST_INODE,
		.setup	= xchk_setup_parent,
		.scrub	= xchk_parent,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_RTBITMAP] = {	/* realtime bitmap */
		.type	= ST_FS,
		.setup	= xchk_setup_rt,
		.scrub	= xchk_rtbitmap,
		.has	= xfs_sb_version_hasrealtime,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_RTSUM] = {	/* realtime summary */
		.type	= ST_FS,
		.setup	= xchk_setup_rt,
		.scrub	= xchk_rtsummary,
		.has	= xfs_sb_version_hasrealtime,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_UQUOTA] = {	/* user quota */
		.type	= ST_FS,
		.setup	= xchk_setup_quota,
		.scrub	= xchk_quota,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_GQUOTA] = {	/* group quota */
		.type	= ST_FS,
		.setup	= xchk_setup_quota,
		.scrub	= xchk_quota,
		.repair	= xrep_notsupported,
	},
	[XFS_SCRUB_TYPE_PQUOTA] = {	/* project quota */
		.type	= ST_FS,
		.setup	= xchk_setup_quota,
		.scrub	= xchk_quota,
		.repair	= xrep_notsupported,
	},
};

/* This isn't a stable feature, warn once per day. */
static inline void
xchk_experimental_warning(
	struct xfs_mount	*mp)
{
	static struct ratelimit_state scrub_warning = RATELIMIT_STATE_INIT(
			"xchk_warning", 86400 * HZ, 1);
	ratelimit_set_flags(&scrub_warning, RATELIMIT_MSG_ON_RELEASE);

	if (__ratelimit(&scrub_warning))
		xfs_alert(mp,
"EXPERIMENTAL online scrub feature in use. Use at your own risk!");
}

static int
xchk_validate_inputs(
	struct xfs_mount		*mp,
	struct xfs_scrub_metadata	*sm)
{
	int				error;
	const struct xchk_meta_ops	*ops;

	error = -EINVAL;
	/* Check our inputs. */
	sm->sm_flags &= ~XFS_SCRUB_FLAGS_OUT;
	if (sm->sm_flags & ~XFS_SCRUB_FLAGS_IN)
		goto out;
	/* sm_reserved[] must be zero */
	if (memchr_inv(sm->sm_reserved, 0, sizeof(sm->sm_reserved)))
		goto out;

	error = -ENOENT;
	/* Do we know about this type of metadata? */
	if (sm->sm_type >= XFS_SCRUB_TYPE_NR)
		goto out;
	ops = &meta_scrub_ops[sm->sm_type];
	if (ops->setup == NULL || ops->scrub == NULL)
		goto out;
	/* Does this fs even support this type of metadata? */
	if (ops->has && !ops->has(&mp->m_sb))
		goto out;

	error = -EINVAL;
	/* restricting fields must be appropriate for type */
	switch (ops->type) {
	case ST_NONE:
	case ST_FS:
		if (sm->sm_ino || sm->sm_gen || sm->sm_agno)
			goto out;
		break;
	case ST_PERAG:
		if (sm->sm_ino || sm->sm_gen ||
		    sm->sm_agno >= mp->m_sb.sb_agcount)
			goto out;
		break;
	case ST_INODE:
		if (sm->sm_agno || (sm->sm_gen && !sm->sm_ino))
			goto out;
		break;
	default:
		goto out;
	}

	/*
	 * We only want to repair read-write v5+ filesystems.  Defer the check
	 * for ops->repair until after our scrub confirms that we need to
	 * perform repairs so that we avoid failing due to not supporting
	 * repairing an object that doesn't need repairs.
	 */
	if (sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR) {
		error = -EOPNOTSUPP;
		if (!xfs_sb_version_hascrc(&mp->m_sb))
			goto out;

		error = -EROFS;
		if (mp->m_flags & XFS_MOUNT_RDONLY)
			goto out;
	}

	error = 0;
out:
	return error;
}

#ifdef CONFIG_XFS_ONLINE_REPAIR
static inline void xchk_postmortem(struct xfs_scrub *sc)
{
	/*
	 * Userspace asked us to repair something, we repaired it, rescanned
	 * it, and the rescan says it's still broken.  Scream about this in
	 * the system logs.
	 */
	if ((sc->sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR) &&
	    (sc->sm->sm_flags & (XFS_SCRUB_OFLAG_CORRUPT |
				 XFS_SCRUB_OFLAG_XCORRUPT)))
		xrep_failure(sc->mp);
}
#else
static inline void xchk_postmortem(struct xfs_scrub *sc)
{
	/*
	 * Userspace asked us to scrub something, it's broken, and we have no
	 * way of fixing it.  Scream in the logs.
	 */
	if (sc->sm->sm_flags & (XFS_SCRUB_OFLAG_CORRUPT |
				XFS_SCRUB_OFLAG_XCORRUPT))
		xfs_alert_ratelimited(sc->mp,
				"Corruption detected during scrub.");
}
#endif /* CONFIG_XFS_ONLINE_REPAIR */

/* Dispatch metadata scrubbing. */
int
xfs_scrub_metadata(
	struct xfs_inode		*ip,
	struct xfs_scrub_metadata	*sm)
{
	struct xfs_scrub		sc = {
		.mp			= ip->i_mount,
		.sm			= sm,
		.sa			= {
			.agno		= NULLAGNUMBER,
		},
	};
	struct xfs_mount		*mp = ip->i_mount;
	int				error = 0;

	BUILD_BUG_ON(sizeof(meta_scrub_ops) !=
		(sizeof(struct xchk_meta_ops) * XFS_SCRUB_TYPE_NR));

	trace_xchk_start(ip, sm, error);

	/* Forbidden if we are shut down or mounted norecovery. */
	error = -ESHUTDOWN;
	if (XFS_FORCED_SHUTDOWN(mp))
		goto out;
	error = -ENOTRECOVERABLE;
	if (mp->m_flags & XFS_MOUNT_NORECOVERY)
		goto out;

	error = xchk_validate_inputs(mp, sm);
	if (error)
		goto out;

	xchk_experimental_warning(mp);

	sc.ops = &meta_scrub_ops[sm->sm_type];
retry_op:
	/* Set up for the operation. */
	error = sc.ops->setup(&sc, ip);
	if (error)
		goto out_teardown;

	/* Scrub for errors. */
	error = sc.ops->scrub(&sc);
	if (!(sc.flags & XCHK_TRY_HARDER) && error == -EDEADLOCK) {
		/*
		 * Scrubbers return -EDEADLOCK to mean 'try harder'.
		 * Tear down everything we hold, then set up again with
		 * preparation for worst-case scenarios.
		 */
		error = xchk_teardown(&sc, ip, 0);
		if (error)
			goto out;
		sc.flags |= XCHK_TRY_HARDER;
		goto retry_op;
	} else if (error)
		goto out_teardown;

	if ((sc.sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR) &&
	    !(sc.flags & XREP_ALREADY_FIXED)) {
		bool needs_fix;

		/* Let debug users force us into the repair routines. */
		if (XFS_TEST_ERROR(false, mp, XFS_ERRTAG_FORCE_SCRUB_REPAIR))
			sc.sm->sm_flags |= XFS_SCRUB_OFLAG_CORRUPT;

		needs_fix = (sc.sm->sm_flags & (XFS_SCRUB_OFLAG_CORRUPT |
						XFS_SCRUB_OFLAG_XCORRUPT |
						XFS_SCRUB_OFLAG_PREEN));
		/*
		 * If userspace asked for a repair but it wasn't necessary,
		 * report that back to userspace.
		 */
		if (!needs_fix) {
			sc.sm->sm_flags |= XFS_SCRUB_OFLAG_NO_REPAIR_NEEDED;
			goto out_nofix;
		}

		/*
		 * If it's broken, userspace wants us to fix it, and we haven't
		 * already tried to fix it, then attempt a repair.
		 */
		error = xrep_attempt(ip, &sc);
		if (error == -EAGAIN) {
			/*
			 * Either the repair function succeeded or it couldn't
			 * get all the resources it needs; either way, we go
			 * back to the beginning and call the scrub function.
			 */
			error = xchk_teardown(&sc, ip, 0);
			if (error) {
				xrep_failure(mp);
				goto out;
			}
			goto retry_op;
		}
	}

out_nofix:
	xchk_postmortem(&sc);
out_teardown:
	error = xchk_teardown(&sc, ip, error);
out:
	trace_xchk_done(ip, sm, error);
	if (error == -EFSCORRUPTED || error == -EFSBADCRC) {
		sm->sm_flags |= XFS_SCRUB_OFLAG_CORRUPT;
		error = 0;
	}
	return error;
}