xref: /kernel/linux/linux-6.6/fs/xfs/scrub/repair.c (revision 62306a36)
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (C) 2018-2023 Oracle.  All Rights Reserved.
4 * Author: Darrick J. Wong <djwong@kernel.org>
5 */
6#include "xfs.h"
7#include "xfs_fs.h"
8#include "xfs_shared.h"
9#include "xfs_format.h"
10#include "xfs_trans_resv.h"
11#include "xfs_mount.h"
12#include "xfs_btree.h"
13#include "xfs_log_format.h"
14#include "xfs_trans.h"
15#include "xfs_sb.h"
16#include "xfs_inode.h"
17#include "xfs_alloc.h"
18#include "xfs_alloc_btree.h"
19#include "xfs_ialloc.h"
20#include "xfs_ialloc_btree.h"
21#include "xfs_rmap.h"
22#include "xfs_rmap_btree.h"
23#include "xfs_refcount_btree.h"
24#include "xfs_extent_busy.h"
25#include "xfs_ag.h"
26#include "xfs_ag_resv.h"
27#include "xfs_quota.h"
28#include "xfs_qm.h"
29#include "xfs_defer.h"
30#include "scrub/scrub.h"
31#include "scrub/common.h"
32#include "scrub/trace.h"
33#include "scrub/repair.h"
34#include "scrub/bitmap.h"
35#include "scrub/stats.h"
36
37/*
38 * Attempt to repair some metadata, if the metadata is corrupt and userspace
39 * told us to fix it.  This function returns -EAGAIN to mean "re-run scrub",
40 * and will set *fixed to true if it thinks it repaired anything.
41 */
42int
43xrep_attempt(
44	struct xfs_scrub	*sc,
45	struct xchk_stats_run	*run)
46{
47	u64			repair_start;
48	int			error = 0;
49
50	trace_xrep_attempt(XFS_I(file_inode(sc->file)), sc->sm, error);
51
52	xchk_ag_btcur_free(&sc->sa);
53
54	/* Repair whatever's broken. */
55	ASSERT(sc->ops->repair);
56	run->repair_attempted = true;
57	repair_start = xchk_stats_now();
58	error = sc->ops->repair(sc);
59	trace_xrep_done(XFS_I(file_inode(sc->file)), sc->sm, error);
60	run->repair_ns += xchk_stats_elapsed_ns(repair_start);
61	switch (error) {
62	case 0:
63		/*
64		 * Repair succeeded.  Commit the fixes and perform a second
65		 * scrub so that we can tell userspace if we fixed the problem.
66		 */
67		sc->sm->sm_flags &= ~XFS_SCRUB_FLAGS_OUT;
68		sc->flags |= XREP_ALREADY_FIXED;
69		run->repair_succeeded = true;
70		return -EAGAIN;
71	case -ECHRNG:
72		sc->flags |= XCHK_NEED_DRAIN;
73		run->retries++;
74		return -EAGAIN;
75	case -EDEADLOCK:
76		/* Tell the caller to try again having grabbed all the locks. */
77		if (!(sc->flags & XCHK_TRY_HARDER)) {
78			sc->flags |= XCHK_TRY_HARDER;
79			run->retries++;
80			return -EAGAIN;
81		}
82		/*
83		 * We tried harder but still couldn't grab all the resources
84		 * we needed to fix it.  The corruption has not been fixed,
85		 * so exit to userspace with the scan's output flags unchanged.
86		 */
87		return 0;
88	default:
89		/*
90		 * EAGAIN tells the caller to re-scrub, so we cannot return
91		 * that here.
92		 */
93		ASSERT(error != -EAGAIN);
94		return error;
95	}
96}
97
98/*
99 * Complain about unfixable problems in the filesystem.  We don't log
100 * corruptions when IFLAG_REPAIR wasn't set on the assumption that the driver
101 * program is xfs_scrub, which will call back with IFLAG_REPAIR set if the
102 * administrator isn't running xfs_scrub in no-repairs mode.
103 *
104 * Use this helper function because _ratelimited silently declares a static
105 * structure to track rate limiting information.
106 */
107void
108xrep_failure(
109	struct xfs_mount	*mp)
110{
111	xfs_alert_ratelimited(mp,
112"Corruption not fixed during online repair.  Unmount and run xfs_repair.");
113}
114
115/*
116 * Repair probe -- userspace uses this to probe if we're willing to repair a
117 * given mountpoint.
118 */
119int
120xrep_probe(
121	struct xfs_scrub	*sc)
122{
123	int			error = 0;
124
125	if (xchk_should_terminate(sc, &error))
126		return error;
127
128	return 0;
129}
130
131/*
132 * Roll a transaction, keeping the AG headers locked and reinitializing
133 * the btree cursors.
134 */
135int
136xrep_roll_ag_trans(
137	struct xfs_scrub	*sc)
138{
139	int			error;
140
141	/*
142	 * Keep the AG header buffers locked while we roll the transaction.
143	 * Ensure that both AG buffers are dirty and held when we roll the
144	 * transaction so that they move forward in the log without losing the
145	 * bli (and hence the bli type) when the transaction commits.
146	 *
147	 * Normal code would never hold clean buffers across a roll, but repair
148	 * needs both buffers to maintain a total lock on the AG.
149	 */
150	if (sc->sa.agi_bp) {
151		xfs_ialloc_log_agi(sc->tp, sc->sa.agi_bp, XFS_AGI_MAGICNUM);
152		xfs_trans_bhold(sc->tp, sc->sa.agi_bp);
153	}
154
155	if (sc->sa.agf_bp) {
156		xfs_alloc_log_agf(sc->tp, sc->sa.agf_bp, XFS_AGF_MAGICNUM);
157		xfs_trans_bhold(sc->tp, sc->sa.agf_bp);
158	}
159
160	/*
161	 * Roll the transaction.  We still hold the AG header buffers locked
162	 * regardless of whether or not that succeeds.  On failure, the buffers
163	 * will be released during teardown on our way out of the kernel.  If
164	 * successful, join the buffers to the new transaction and move on.
165	 */
166	error = xfs_trans_roll(&sc->tp);
167	if (error)
168		return error;
169
170	/* Join the AG headers to the new transaction. */
171	if (sc->sa.agi_bp)
172		xfs_trans_bjoin(sc->tp, sc->sa.agi_bp);
173	if (sc->sa.agf_bp)
174		xfs_trans_bjoin(sc->tp, sc->sa.agf_bp);
175
176	return 0;
177}
178
179/* Finish all deferred work attached to the repair transaction. */
180int
181xrep_defer_finish(
182	struct xfs_scrub	*sc)
183{
184	int			error;
185
186	/*
187	 * Keep the AG header buffers locked while we complete deferred work
188	 * items.  Ensure that both AG buffers are dirty and held when we roll
189	 * the transaction so that they move forward in the log without losing
190	 * the bli (and hence the bli type) when the transaction commits.
191	 *
192	 * Normal code would never hold clean buffers across a roll, but repair
193	 * needs both buffers to maintain a total lock on the AG.
194	 */
195	if (sc->sa.agi_bp) {
196		xfs_ialloc_log_agi(sc->tp, sc->sa.agi_bp, XFS_AGI_MAGICNUM);
197		xfs_trans_bhold(sc->tp, sc->sa.agi_bp);
198	}
199
200	if (sc->sa.agf_bp) {
201		xfs_alloc_log_agf(sc->tp, sc->sa.agf_bp, XFS_AGF_MAGICNUM);
202		xfs_trans_bhold(sc->tp, sc->sa.agf_bp);
203	}
204
205	/*
206	 * Finish all deferred work items.  We still hold the AG header buffers
207	 * locked regardless of whether or not that succeeds.  On failure, the
208	 * buffers will be released during teardown on our way out of the
209	 * kernel.  If successful, join the buffers to the new transaction
210	 * and move on.
211	 */
212	error = xfs_defer_finish(&sc->tp);
213	if (error)
214		return error;
215
216	/*
217	 * Release the hold that we set above because defer_finish won't do
218	 * that for us.  The defer roll code redirties held buffers after each
219	 * roll, so the AG header buffers should be ready for logging.
220	 */
221	if (sc->sa.agi_bp)
222		xfs_trans_bhold_release(sc->tp, sc->sa.agi_bp);
223	if (sc->sa.agf_bp)
224		xfs_trans_bhold_release(sc->tp, sc->sa.agf_bp);
225
226	return 0;
227}
228
229/*
230 * Does the given AG have enough space to rebuild a btree?  Neither AG
231 * reservation can be critical, and we must have enough space (factoring
232 * in AG reservations) to construct a whole btree.
233 */
234bool
235xrep_ag_has_space(
236	struct xfs_perag	*pag,
237	xfs_extlen_t		nr_blocks,
238	enum xfs_ag_resv_type	type)
239{
240	return  !xfs_ag_resv_critical(pag, XFS_AG_RESV_RMAPBT) &&
241		!xfs_ag_resv_critical(pag, XFS_AG_RESV_METADATA) &&
242		pag->pagf_freeblks > xfs_ag_resv_needed(pag, type) + nr_blocks;
243}
244
245/*
246 * Figure out how many blocks to reserve for an AG repair.  We calculate the
247 * worst case estimate for the number of blocks we'd need to rebuild one of
248 * any type of per-AG btree.
249 */
250xfs_extlen_t
251xrep_calc_ag_resblks(
252	struct xfs_scrub		*sc)
253{
254	struct xfs_mount		*mp = sc->mp;
255	struct xfs_scrub_metadata	*sm = sc->sm;
256	struct xfs_perag		*pag;
257	struct xfs_buf			*bp;
258	xfs_agino_t			icount = NULLAGINO;
259	xfs_extlen_t			aglen = NULLAGBLOCK;
260	xfs_extlen_t			usedlen;
261	xfs_extlen_t			freelen;
262	xfs_extlen_t			bnobt_sz;
263	xfs_extlen_t			inobt_sz;
264	xfs_extlen_t			rmapbt_sz;
265	xfs_extlen_t			refcbt_sz;
266	int				error;
267
268	if (!(sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR))
269		return 0;
270
271	pag = xfs_perag_get(mp, sm->sm_agno);
272	if (xfs_perag_initialised_agi(pag)) {
273		/* Use in-core icount if possible. */
274		icount = pag->pagi_count;
275	} else {
276		/* Try to get the actual counters from disk. */
277		error = xfs_ialloc_read_agi(pag, NULL, &bp);
278		if (!error) {
279			icount = pag->pagi_count;
280			xfs_buf_relse(bp);
281		}
282	}
283
284	/* Now grab the block counters from the AGF. */
285	error = xfs_alloc_read_agf(pag, NULL, 0, &bp);
286	if (error) {
287		aglen = pag->block_count;
288		freelen = aglen;
289		usedlen = aglen;
290	} else {
291		struct xfs_agf	*agf = bp->b_addr;
292
293		aglen = be32_to_cpu(agf->agf_length);
294		freelen = be32_to_cpu(agf->agf_freeblks);
295		usedlen = aglen - freelen;
296		xfs_buf_relse(bp);
297	}
298
299	/* If the icount is impossible, make some worst-case assumptions. */
300	if (icount == NULLAGINO ||
301	    !xfs_verify_agino(pag, icount)) {
302		icount = pag->agino_max - pag->agino_min + 1;
303	}
304
305	/* If the block counts are impossible, make worst-case assumptions. */
306	if (aglen == NULLAGBLOCK ||
307	    aglen != pag->block_count ||
308	    freelen >= aglen) {
309		aglen = pag->block_count;
310		freelen = aglen;
311		usedlen = aglen;
312	}
313	xfs_perag_put(pag);
314
315	trace_xrep_calc_ag_resblks(mp, sm->sm_agno, icount, aglen,
316			freelen, usedlen);
317
318	/*
319	 * Figure out how many blocks we'd need worst case to rebuild
320	 * each type of btree.  Note that we can only rebuild the
321	 * bnobt/cntbt or inobt/finobt as pairs.
322	 */
323	bnobt_sz = 2 * xfs_allocbt_calc_size(mp, freelen);
324	if (xfs_has_sparseinodes(mp))
325		inobt_sz = xfs_iallocbt_calc_size(mp, icount /
326				XFS_INODES_PER_HOLEMASK_BIT);
327	else
328		inobt_sz = xfs_iallocbt_calc_size(mp, icount /
329				XFS_INODES_PER_CHUNK);
330	if (xfs_has_finobt(mp))
331		inobt_sz *= 2;
332	if (xfs_has_reflink(mp))
333		refcbt_sz = xfs_refcountbt_calc_size(mp, usedlen);
334	else
335		refcbt_sz = 0;
336	if (xfs_has_rmapbt(mp)) {
337		/*
338		 * Guess how many blocks we need to rebuild the rmapbt.
339		 * For non-reflink filesystems we can't have more records than
340		 * used blocks.  However, with reflink it's possible to have
341		 * more than one rmap record per AG block.  We don't know how
342		 * many rmaps there could be in the AG, so we start off with
343		 * what we hope is an generous over-estimation.
344		 */
345		if (xfs_has_reflink(mp))
346			rmapbt_sz = xfs_rmapbt_calc_size(mp,
347					(unsigned long long)aglen * 2);
348		else
349			rmapbt_sz = xfs_rmapbt_calc_size(mp, usedlen);
350	} else {
351		rmapbt_sz = 0;
352	}
353
354	trace_xrep_calc_ag_resblks_btsize(mp, sm->sm_agno, bnobt_sz,
355			inobt_sz, rmapbt_sz, refcbt_sz);
356
357	return max(max(bnobt_sz, inobt_sz), max(rmapbt_sz, refcbt_sz));
358}
359
360/*
361 * Reconstructing per-AG Btrees
362 *
363 * When a space btree is corrupt, we don't bother trying to fix it.  Instead,
364 * we scan secondary space metadata to derive the records that should be in
365 * the damaged btree, initialize a fresh btree root, and insert the records.
366 * Note that for rebuilding the rmapbt we scan all the primary data to
367 * generate the new records.
368 *
369 * However, that leaves the matter of removing all the metadata describing the
370 * old broken structure.  For primary metadata we use the rmap data to collect
371 * every extent with a matching rmap owner (bitmap); we then iterate all other
372 * metadata structures with the same rmap owner to collect the extents that
373 * cannot be removed (sublist).  We then subtract sublist from bitmap to
374 * derive the blocks that were used by the old btree.  These blocks can be
375 * reaped.
376 *
377 * For rmapbt reconstructions we must use different tactics for extent
378 * collection.  First we iterate all primary metadata (this excludes the old
379 * rmapbt, obviously) to generate new rmap records.  The gaps in the rmap
380 * records are collected as bitmap.  The bnobt records are collected as
381 * sublist.  As with the other btrees we subtract sublist from bitmap, and the
382 * result (since the rmapbt lives in the free space) are the blocks from the
383 * old rmapbt.
384 */
385
386/* Ensure the freelist is the correct size. */
387int
388xrep_fix_freelist(
389	struct xfs_scrub	*sc,
390	bool			can_shrink)
391{
392	struct xfs_alloc_arg	args = {0};
393
394	args.mp = sc->mp;
395	args.tp = sc->tp;
396	args.agno = sc->sa.pag->pag_agno;
397	args.alignment = 1;
398	args.pag = sc->sa.pag;
399
400	return xfs_alloc_fix_freelist(&args,
401			can_shrink ? 0 : XFS_ALLOC_FLAG_NOSHRINK);
402}
403
404/*
405 * Finding per-AG Btree Roots for AGF/AGI Reconstruction
406 *
407 * If the AGF or AGI become slightly corrupted, it may be necessary to rebuild
408 * the AG headers by using the rmap data to rummage through the AG looking for
409 * btree roots.  This is not guaranteed to work if the AG is heavily damaged
410 * or the rmap data are corrupt.
411 *
412 * Callers of xrep_find_ag_btree_roots must lock the AGF and AGFL
413 * buffers if the AGF is being rebuilt; or the AGF and AGI buffers if the
414 * AGI is being rebuilt.  It must maintain these locks until it's safe for
415 * other threads to change the btrees' shapes.  The caller provides
416 * information about the btrees to look for by passing in an array of
417 * xrep_find_ag_btree with the (rmap owner, buf_ops, magic) fields set.
418 * The (root, height) fields will be set on return if anything is found.  The
419 * last element of the array should have a NULL buf_ops to mark the end of the
420 * array.
421 *
422 * For every rmapbt record matching any of the rmap owners in btree_info,
423 * read each block referenced by the rmap record.  If the block is a btree
424 * block from this filesystem matching any of the magic numbers and has a
425 * level higher than what we've already seen, remember the block and the
426 * height of the tree required to have such a block.  When the call completes,
427 * we return the highest block we've found for each btree description; those
428 * should be the roots.
429 */
430
431struct xrep_findroot {
432	struct xfs_scrub		*sc;
433	struct xfs_buf			*agfl_bp;
434	struct xfs_agf			*agf;
435	struct xrep_find_ag_btree	*btree_info;
436};
437
438/* See if our block is in the AGFL. */
439STATIC int
440xrep_findroot_agfl_walk(
441	struct xfs_mount	*mp,
442	xfs_agblock_t		bno,
443	void			*priv)
444{
445	xfs_agblock_t		*agbno = priv;
446
447	return (*agbno == bno) ? -ECANCELED : 0;
448}
449
450/* Does this block match the btree information passed in? */
451STATIC int
452xrep_findroot_block(
453	struct xrep_findroot		*ri,
454	struct xrep_find_ag_btree	*fab,
455	uint64_t			owner,
456	xfs_agblock_t			agbno,
457	bool				*done_with_block)
458{
459	struct xfs_mount		*mp = ri->sc->mp;
460	struct xfs_buf			*bp;
461	struct xfs_btree_block		*btblock;
462	xfs_daddr_t			daddr;
463	int				block_level;
464	int				error = 0;
465
466	daddr = XFS_AGB_TO_DADDR(mp, ri->sc->sa.pag->pag_agno, agbno);
467
468	/*
469	 * Blocks in the AGFL have stale contents that might just happen to
470	 * have a matching magic and uuid.  We don't want to pull these blocks
471	 * in as part of a tree root, so we have to filter out the AGFL stuff
472	 * here.  If the AGFL looks insane we'll just refuse to repair.
473	 */
474	if (owner == XFS_RMAP_OWN_AG) {
475		error = xfs_agfl_walk(mp, ri->agf, ri->agfl_bp,
476				xrep_findroot_agfl_walk, &agbno);
477		if (error == -ECANCELED)
478			return 0;
479		if (error)
480			return error;
481	}
482
483	/*
484	 * Read the buffer into memory so that we can see if it's a match for
485	 * our btree type.  We have no clue if it is beforehand, and we want to
486	 * avoid xfs_trans_read_buf's behavior of dumping the DONE state (which
487	 * will cause needless disk reads in subsequent calls to this function)
488	 * and logging metadata verifier failures.
489	 *
490	 * Therefore, pass in NULL buffer ops.  If the buffer was already in
491	 * memory from some other caller it will already have b_ops assigned.
492	 * If it was in memory from a previous unsuccessful findroot_block
493	 * call, the buffer won't have b_ops but it should be clean and ready
494	 * for us to try to verify if the read call succeeds.  The same applies
495	 * if the buffer wasn't in memory at all.
496	 *
497	 * Note: If we never match a btree type with this buffer, it will be
498	 * left in memory with NULL b_ops.  This shouldn't be a problem unless
499	 * the buffer gets written.
500	 */
501	error = xfs_trans_read_buf(mp, ri->sc->tp, mp->m_ddev_targp, daddr,
502			mp->m_bsize, 0, &bp, NULL);
503	if (error)
504		return error;
505
506	/* Ensure the block magic matches the btree type we're looking for. */
507	btblock = XFS_BUF_TO_BLOCK(bp);
508	ASSERT(fab->buf_ops->magic[1] != 0);
509	if (btblock->bb_magic != fab->buf_ops->magic[1])
510		goto out;
511
512	/*
513	 * If the buffer already has ops applied and they're not the ones for
514	 * this btree type, we know this block doesn't match the btree and we
515	 * can bail out.
516	 *
517	 * If the buffer ops match ours, someone else has already validated
518	 * the block for us, so we can move on to checking if this is a root
519	 * block candidate.
520	 *
521	 * If the buffer does not have ops, nobody has successfully validated
522	 * the contents and the buffer cannot be dirty.  If the magic, uuid,
523	 * and structure match this btree type then we'll move on to checking
524	 * if it's a root block candidate.  If there is no match, bail out.
525	 */
526	if (bp->b_ops) {
527		if (bp->b_ops != fab->buf_ops)
528			goto out;
529	} else {
530		ASSERT(!xfs_trans_buf_is_dirty(bp));
531		if (!uuid_equal(&btblock->bb_u.s.bb_uuid,
532				&mp->m_sb.sb_meta_uuid))
533			goto out;
534		/*
535		 * Read verifiers can reference b_ops, so we set the pointer
536		 * here.  If the verifier fails we'll reset the buffer state
537		 * to what it was before we touched the buffer.
538		 */
539		bp->b_ops = fab->buf_ops;
540		fab->buf_ops->verify_read(bp);
541		if (bp->b_error) {
542			bp->b_ops = NULL;
543			bp->b_error = 0;
544			goto out;
545		}
546
547		/*
548		 * Some read verifiers will (re)set b_ops, so we must be
549		 * careful not to change b_ops after running the verifier.
550		 */
551	}
552
553	/*
554	 * This block passes the magic/uuid and verifier tests for this btree
555	 * type.  We don't need the caller to try the other tree types.
556	 */
557	*done_with_block = true;
558
559	/*
560	 * Compare this btree block's level to the height of the current
561	 * candidate root block.
562	 *
563	 * If the level matches the root we found previously, throw away both
564	 * blocks because there can't be two candidate roots.
565	 *
566	 * If level is lower in the tree than the root we found previously,
567	 * ignore this block.
568	 */
569	block_level = xfs_btree_get_level(btblock);
570	if (block_level + 1 == fab->height) {
571		fab->root = NULLAGBLOCK;
572		goto out;
573	} else if (block_level < fab->height) {
574		goto out;
575	}
576
577	/*
578	 * This is the highest block in the tree that we've found so far.
579	 * Update the btree height to reflect what we've learned from this
580	 * block.
581	 */
582	fab->height = block_level + 1;
583
584	/*
585	 * If this block doesn't have sibling pointers, then it's the new root
586	 * block candidate.  Otherwise, the root will be found farther up the
587	 * tree.
588	 */
589	if (btblock->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) &&
590	    btblock->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK))
591		fab->root = agbno;
592	else
593		fab->root = NULLAGBLOCK;
594
595	trace_xrep_findroot_block(mp, ri->sc->sa.pag->pag_agno, agbno,
596			be32_to_cpu(btblock->bb_magic), fab->height - 1);
597out:
598	xfs_trans_brelse(ri->sc->tp, bp);
599	return error;
600}
601
602/*
603 * Do any of the blocks in this rmap record match one of the btrees we're
604 * looking for?
605 */
606STATIC int
607xrep_findroot_rmap(
608	struct xfs_btree_cur		*cur,
609	const struct xfs_rmap_irec	*rec,
610	void				*priv)
611{
612	struct xrep_findroot		*ri = priv;
613	struct xrep_find_ag_btree	*fab;
614	xfs_agblock_t			b;
615	bool				done;
616	int				error = 0;
617
618	/* Ignore anything that isn't AG metadata. */
619	if (!XFS_RMAP_NON_INODE_OWNER(rec->rm_owner))
620		return 0;
621
622	/* Otherwise scan each block + btree type. */
623	for (b = 0; b < rec->rm_blockcount; b++) {
624		done = false;
625		for (fab = ri->btree_info; fab->buf_ops; fab++) {
626			if (rec->rm_owner != fab->rmap_owner)
627				continue;
628			error = xrep_findroot_block(ri, fab,
629					rec->rm_owner, rec->rm_startblock + b,
630					&done);
631			if (error)
632				return error;
633			if (done)
634				break;
635		}
636	}
637
638	return 0;
639}
640
641/* Find the roots of the per-AG btrees described in btree_info. */
642int
643xrep_find_ag_btree_roots(
644	struct xfs_scrub		*sc,
645	struct xfs_buf			*agf_bp,
646	struct xrep_find_ag_btree	*btree_info,
647	struct xfs_buf			*agfl_bp)
648{
649	struct xfs_mount		*mp = sc->mp;
650	struct xrep_findroot		ri;
651	struct xrep_find_ag_btree	*fab;
652	struct xfs_btree_cur		*cur;
653	int				error;
654
655	ASSERT(xfs_buf_islocked(agf_bp));
656	ASSERT(agfl_bp == NULL || xfs_buf_islocked(agfl_bp));
657
658	ri.sc = sc;
659	ri.btree_info = btree_info;
660	ri.agf = agf_bp->b_addr;
661	ri.agfl_bp = agfl_bp;
662	for (fab = btree_info; fab->buf_ops; fab++) {
663		ASSERT(agfl_bp || fab->rmap_owner != XFS_RMAP_OWN_AG);
664		ASSERT(XFS_RMAP_NON_INODE_OWNER(fab->rmap_owner));
665		fab->root = NULLAGBLOCK;
666		fab->height = 0;
667	}
668
669	cur = xfs_rmapbt_init_cursor(mp, sc->tp, agf_bp, sc->sa.pag);
670	error = xfs_rmap_query_all(cur, xrep_findroot_rmap, &ri);
671	xfs_btree_del_cursor(cur, error);
672
673	return error;
674}
675
676/* Force a quotacheck the next time we mount. */
677void
678xrep_force_quotacheck(
679	struct xfs_scrub	*sc,
680	xfs_dqtype_t		type)
681{
682	uint			flag;
683
684	flag = xfs_quota_chkd_flag(type);
685	if (!(flag & sc->mp->m_qflags))
686		return;
687
688	mutex_lock(&sc->mp->m_quotainfo->qi_quotaofflock);
689	sc->mp->m_qflags &= ~flag;
690	spin_lock(&sc->mp->m_sb_lock);
691	sc->mp->m_sb.sb_qflags &= ~flag;
692	spin_unlock(&sc->mp->m_sb_lock);
693	xfs_log_sb(sc->tp);
694	mutex_unlock(&sc->mp->m_quotainfo->qi_quotaofflock);
695}
696
697/*
698 * Attach dquots to this inode, or schedule quotacheck to fix them.
699 *
700 * This function ensures that the appropriate dquots are attached to an inode.
701 * We cannot allow the dquot code to allocate an on-disk dquot block here
702 * because we're already in transaction context with the inode locked.  The
703 * on-disk dquot should already exist anyway.  If the quota code signals
704 * corruption or missing quota information, schedule quotacheck, which will
705 * repair corruptions in the quota metadata.
706 */
707int
708xrep_ino_dqattach(
709	struct xfs_scrub	*sc)
710{
711	int			error;
712
713	error = xfs_qm_dqattach_locked(sc->ip, false);
714	switch (error) {
715	case -EFSBADCRC:
716	case -EFSCORRUPTED:
717	case -ENOENT:
718		xfs_err_ratelimited(sc->mp,
719"inode %llu repair encountered quota error %d, quotacheck forced.",
720				(unsigned long long)sc->ip->i_ino, error);
721		if (XFS_IS_UQUOTA_ON(sc->mp) && !sc->ip->i_udquot)
722			xrep_force_quotacheck(sc, XFS_DQTYPE_USER);
723		if (XFS_IS_GQUOTA_ON(sc->mp) && !sc->ip->i_gdquot)
724			xrep_force_quotacheck(sc, XFS_DQTYPE_GROUP);
725		if (XFS_IS_PQUOTA_ON(sc->mp) && !sc->ip->i_pdquot)
726			xrep_force_quotacheck(sc, XFS_DQTYPE_PROJ);
727		fallthrough;
728	case -ESRCH:
729		error = 0;
730		break;
731	default:
732		break;
733	}
734
735	return error;
736}
737