1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * Copyright (C) 2017-2023 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_log_format.h"
13#include "xfs_trans.h"
14#include "xfs_inode.h"
15#include "xfs_dir2.h"
16#include "xfs_dir2_priv.h"
17#include "xfs_attr_leaf.h"
18#include "scrub/scrub.h"
19#include "scrub/common.h"
20#include "scrub/trace.h"
21#include "scrub/dabtree.h"
22
23/* Directory/Attribute Btree */
24
25/*
26 * Check for da btree operation errors.  See the section about handling
27 * operational errors in common.c.
28 */
29bool
30xchk_da_process_error(
31	struct xchk_da_btree	*ds,
32	int			level,
33	int			*error)
34{
35	struct xfs_scrub	*sc = ds->sc;
36
37	if (*error == 0)
38		return true;
39
40	switch (*error) {
41	case -EDEADLOCK:
42	case -ECHRNG:
43		/* Used to restart an op with deadlock avoidance. */
44		trace_xchk_deadlock_retry(sc->ip, sc->sm, *error);
45		break;
46	case -EFSBADCRC:
47	case -EFSCORRUPTED:
48		/* Note the badness but don't abort. */
49		sc->sm->sm_flags |= XFS_SCRUB_OFLAG_CORRUPT;
50		*error = 0;
51		fallthrough;
52	default:
53		trace_xchk_file_op_error(sc, ds->dargs.whichfork,
54				xfs_dir2_da_to_db(ds->dargs.geo,
55					ds->state->path.blk[level].blkno),
56				*error, __return_address);
57		break;
58	}
59	return false;
60}
61
62/*
63 * Check for da btree corruption.  See the section about handling
64 * operational errors in common.c.
65 */
66void
67xchk_da_set_corrupt(
68	struct xchk_da_btree	*ds,
69	int			level)
70{
71	struct xfs_scrub	*sc = ds->sc;
72
73	sc->sm->sm_flags |= XFS_SCRUB_OFLAG_CORRUPT;
74
75	trace_xchk_fblock_error(sc, ds->dargs.whichfork,
76			xfs_dir2_da_to_db(ds->dargs.geo,
77				ds->state->path.blk[level].blkno),
78			__return_address);
79}
80
81/* Flag a da btree node in need of optimization. */
82void
83xchk_da_set_preen(
84	struct xchk_da_btree	*ds,
85	int			level)
86{
87	struct xfs_scrub	*sc = ds->sc;
88
89	sc->sm->sm_flags |= XFS_SCRUB_OFLAG_PREEN;
90	trace_xchk_fblock_preen(sc, ds->dargs.whichfork,
91			xfs_dir2_da_to_db(ds->dargs.geo,
92				ds->state->path.blk[level].blkno),
93			__return_address);
94}
95
96/* Find an entry at a certain level in a da btree. */
97static struct xfs_da_node_entry *
98xchk_da_btree_node_entry(
99	struct xchk_da_btree		*ds,
100	int				level)
101{
102	struct xfs_da_state_blk		*blk = &ds->state->path.blk[level];
103	struct xfs_da3_icnode_hdr	hdr;
104
105	ASSERT(blk->magic == XFS_DA_NODE_MAGIC);
106
107	xfs_da3_node_hdr_from_disk(ds->sc->mp, &hdr, blk->bp->b_addr);
108	return hdr.btree + blk->index;
109}
110
111/* Scrub a da btree hash (key). */
112int
113xchk_da_btree_hash(
114	struct xchk_da_btree		*ds,
115	int				level,
116	__be32				*hashp)
117{
118	struct xfs_da_node_entry	*entry;
119	xfs_dahash_t			hash;
120	xfs_dahash_t			parent_hash;
121
122	/* Is this hash in order? */
123	hash = be32_to_cpu(*hashp);
124	if (hash < ds->hashes[level])
125		xchk_da_set_corrupt(ds, level);
126	ds->hashes[level] = hash;
127
128	if (level == 0)
129		return 0;
130
131	/* Is this hash no larger than the parent hash? */
132	entry = xchk_da_btree_node_entry(ds, level - 1);
133	parent_hash = be32_to_cpu(entry->hashval);
134	if (parent_hash < hash)
135		xchk_da_set_corrupt(ds, level);
136
137	return 0;
138}
139
140/*
141 * Check a da btree pointer.  Returns true if it's ok to use this
142 * pointer.
143 */
144STATIC bool
145xchk_da_btree_ptr_ok(
146	struct xchk_da_btree	*ds,
147	int			level,
148	xfs_dablk_t		blkno)
149{
150	if (blkno < ds->lowest || (ds->highest != 0 && blkno >= ds->highest)) {
151		xchk_da_set_corrupt(ds, level);
152		return false;
153	}
154
155	return true;
156}
157
158/*
159 * The da btree scrubber can handle leaf1 blocks as a degenerate
160 * form of leafn blocks.  Since the regular da code doesn't handle
161 * leaf1, we must multiplex the verifiers.
162 */
163static void
164xchk_da_btree_read_verify(
165	struct xfs_buf		*bp)
166{
167	struct xfs_da_blkinfo	*info = bp->b_addr;
168
169	switch (be16_to_cpu(info->magic)) {
170	case XFS_DIR2_LEAF1_MAGIC:
171	case XFS_DIR3_LEAF1_MAGIC:
172		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
173		bp->b_ops->verify_read(bp);
174		return;
175	default:
176		/*
177		 * xfs_da3_node_buf_ops already know how to handle
178		 * DA*_NODE, ATTR*_LEAF, and DIR*_LEAFN blocks.
179		 */
180		bp->b_ops = &xfs_da3_node_buf_ops;
181		bp->b_ops->verify_read(bp);
182		return;
183	}
184}
185static void
186xchk_da_btree_write_verify(
187	struct xfs_buf		*bp)
188{
189	struct xfs_da_blkinfo	*info = bp->b_addr;
190
191	switch (be16_to_cpu(info->magic)) {
192	case XFS_DIR2_LEAF1_MAGIC:
193	case XFS_DIR3_LEAF1_MAGIC:
194		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
195		bp->b_ops->verify_write(bp);
196		return;
197	default:
198		/*
199		 * xfs_da3_node_buf_ops already know how to handle
200		 * DA*_NODE, ATTR*_LEAF, and DIR*_LEAFN blocks.
201		 */
202		bp->b_ops = &xfs_da3_node_buf_ops;
203		bp->b_ops->verify_write(bp);
204		return;
205	}
206}
207static void *
208xchk_da_btree_verify(
209	struct xfs_buf		*bp)
210{
211	struct xfs_da_blkinfo	*info = bp->b_addr;
212
213	switch (be16_to_cpu(info->magic)) {
214	case XFS_DIR2_LEAF1_MAGIC:
215	case XFS_DIR3_LEAF1_MAGIC:
216		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
217		return bp->b_ops->verify_struct(bp);
218	default:
219		bp->b_ops = &xfs_da3_node_buf_ops;
220		return bp->b_ops->verify_struct(bp);
221	}
222}
223
224static const struct xfs_buf_ops xchk_da_btree_buf_ops = {
225	.name = "xchk_da_btree",
226	.verify_read = xchk_da_btree_read_verify,
227	.verify_write = xchk_da_btree_write_verify,
228	.verify_struct = xchk_da_btree_verify,
229};
230
231/* Check a block's sibling. */
232STATIC int
233xchk_da_btree_block_check_sibling(
234	struct xchk_da_btree	*ds,
235	int			level,
236	int			direction,
237	xfs_dablk_t		sibling)
238{
239	struct xfs_da_state_path *path = &ds->state->path;
240	struct xfs_da_state_path *altpath = &ds->state->altpath;
241	int			retval;
242	int			plevel;
243	int			error;
244
245	memcpy(altpath, path, sizeof(ds->state->altpath));
246
247	/*
248	 * If the pointer is null, we shouldn't be able to move the upper
249	 * level pointer anywhere.
250	 */
251	if (sibling == 0) {
252		error = xfs_da3_path_shift(ds->state, altpath, direction,
253				false, &retval);
254		if (error == 0 && retval == 0)
255			xchk_da_set_corrupt(ds, level);
256		error = 0;
257		goto out;
258	}
259
260	/* Move the alternate cursor one block in the direction given. */
261	error = xfs_da3_path_shift(ds->state, altpath, direction, false,
262			&retval);
263	if (!xchk_da_process_error(ds, level, &error))
264		goto out;
265	if (retval) {
266		xchk_da_set_corrupt(ds, level);
267		goto out;
268	}
269	if (altpath->blk[level].bp)
270		xchk_buffer_recheck(ds->sc, altpath->blk[level].bp);
271
272	/* Compare upper level pointer to sibling pointer. */
273	if (altpath->blk[level].blkno != sibling)
274		xchk_da_set_corrupt(ds, level);
275
276out:
277	/* Free all buffers in the altpath that aren't referenced from path. */
278	for (plevel = 0; plevel < altpath->active; plevel++) {
279		if (altpath->blk[plevel].bp == NULL ||
280		    (plevel < path->active &&
281		     altpath->blk[plevel].bp == path->blk[plevel].bp))
282			continue;
283
284		xfs_trans_brelse(ds->dargs.trans, altpath->blk[plevel].bp);
285		altpath->blk[plevel].bp = NULL;
286	}
287
288	return error;
289}
290
291/* Check a block's sibling pointers. */
292STATIC int
293xchk_da_btree_block_check_siblings(
294	struct xchk_da_btree	*ds,
295	int			level,
296	struct xfs_da_blkinfo	*hdr)
297{
298	xfs_dablk_t		forw;
299	xfs_dablk_t		back;
300	int			error = 0;
301
302	forw = be32_to_cpu(hdr->forw);
303	back = be32_to_cpu(hdr->back);
304
305	/* Top level blocks should not have sibling pointers. */
306	if (level == 0) {
307		if (forw != 0 || back != 0)
308			xchk_da_set_corrupt(ds, level);
309		return 0;
310	}
311
312	/*
313	 * Check back (left) and forw (right) pointers.  These functions
314	 * absorb error codes for us.
315	 */
316	error = xchk_da_btree_block_check_sibling(ds, level, 0, back);
317	if (error)
318		goto out;
319	error = xchk_da_btree_block_check_sibling(ds, level, 1, forw);
320
321out:
322	memset(&ds->state->altpath, 0, sizeof(ds->state->altpath));
323	return error;
324}
325
326/* Load a dir/attribute block from a btree. */
327STATIC int
328xchk_da_btree_block(
329	struct xchk_da_btree		*ds,
330	int				level,
331	xfs_dablk_t			blkno)
332{
333	struct xfs_da_state_blk		*blk;
334	struct xfs_da_intnode		*node;
335	struct xfs_da_node_entry	*btree;
336	struct xfs_da3_blkinfo		*hdr3;
337	struct xfs_da_args		*dargs = &ds->dargs;
338	struct xfs_inode		*ip = ds->dargs.dp;
339	xfs_failaddr_t			fa;
340	xfs_ino_t			owner;
341	int				*pmaxrecs;
342	struct xfs_da3_icnode_hdr	nodehdr;
343	int				error = 0;
344
345	blk = &ds->state->path.blk[level];
346	ds->state->path.active = level + 1;
347
348	/* Release old block. */
349	if (blk->bp) {
350		xfs_trans_brelse(dargs->trans, blk->bp);
351		blk->bp = NULL;
352	}
353
354	/* Check the pointer. */
355	blk->blkno = blkno;
356	if (!xchk_da_btree_ptr_ok(ds, level, blkno))
357		goto out_nobuf;
358
359	/* Read the buffer. */
360	error = xfs_da_read_buf(dargs->trans, dargs->dp, blk->blkno,
361			XFS_DABUF_MAP_HOLE_OK, &blk->bp, dargs->whichfork,
362			&xchk_da_btree_buf_ops);
363	if (!xchk_da_process_error(ds, level, &error))
364		goto out_nobuf;
365	if (blk->bp)
366		xchk_buffer_recheck(ds->sc, blk->bp);
367
368	/*
369	 * We didn't find a dir btree root block, which means that
370	 * there's no LEAF1/LEAFN tree (at least not where it's supposed
371	 * to be), so jump out now.
372	 */
373	if (ds->dargs.whichfork == XFS_DATA_FORK && level == 0 &&
374			blk->bp == NULL)
375		goto out_nobuf;
376
377	/* It's /not/ ok for attr trees not to have a da btree. */
378	if (blk->bp == NULL) {
379		xchk_da_set_corrupt(ds, level);
380		goto out_nobuf;
381	}
382
383	hdr3 = blk->bp->b_addr;
384	blk->magic = be16_to_cpu(hdr3->hdr.magic);
385	pmaxrecs = &ds->maxrecs[level];
386
387	/* We only started zeroing the header on v5 filesystems. */
388	if (xfs_has_crc(ds->sc->mp) && hdr3->hdr.pad)
389		xchk_da_set_corrupt(ds, level);
390
391	/* Check the owner. */
392	if (xfs_has_crc(ip->i_mount)) {
393		owner = be64_to_cpu(hdr3->owner);
394		if (owner != ip->i_ino)
395			xchk_da_set_corrupt(ds, level);
396	}
397
398	/* Check the siblings. */
399	error = xchk_da_btree_block_check_siblings(ds, level, &hdr3->hdr);
400	if (error)
401		goto out;
402
403	/* Interpret the buffer. */
404	switch (blk->magic) {
405	case XFS_ATTR_LEAF_MAGIC:
406	case XFS_ATTR3_LEAF_MAGIC:
407		xfs_trans_buf_set_type(dargs->trans, blk->bp,
408				XFS_BLFT_ATTR_LEAF_BUF);
409		blk->magic = XFS_ATTR_LEAF_MAGIC;
410		blk->hashval = xfs_attr_leaf_lasthash(blk->bp, pmaxrecs);
411		if (ds->tree_level != 0)
412			xchk_da_set_corrupt(ds, level);
413		break;
414	case XFS_DIR2_LEAFN_MAGIC:
415	case XFS_DIR3_LEAFN_MAGIC:
416		xfs_trans_buf_set_type(dargs->trans, blk->bp,
417				XFS_BLFT_DIR_LEAFN_BUF);
418		blk->magic = XFS_DIR2_LEAFN_MAGIC;
419		blk->hashval = xfs_dir2_leaf_lasthash(ip, blk->bp, pmaxrecs);
420		if (ds->tree_level != 0)
421			xchk_da_set_corrupt(ds, level);
422		break;
423	case XFS_DIR2_LEAF1_MAGIC:
424	case XFS_DIR3_LEAF1_MAGIC:
425		xfs_trans_buf_set_type(dargs->trans, blk->bp,
426				XFS_BLFT_DIR_LEAF1_BUF);
427		blk->magic = XFS_DIR2_LEAF1_MAGIC;
428		blk->hashval = xfs_dir2_leaf_lasthash(ip, blk->bp, pmaxrecs);
429		if (ds->tree_level != 0)
430			xchk_da_set_corrupt(ds, level);
431		break;
432	case XFS_DA_NODE_MAGIC:
433	case XFS_DA3_NODE_MAGIC:
434		xfs_trans_buf_set_type(dargs->trans, blk->bp,
435				XFS_BLFT_DA_NODE_BUF);
436		blk->magic = XFS_DA_NODE_MAGIC;
437		node = blk->bp->b_addr;
438		xfs_da3_node_hdr_from_disk(ip->i_mount, &nodehdr, node);
439		btree = nodehdr.btree;
440		*pmaxrecs = nodehdr.count;
441		blk->hashval = be32_to_cpu(btree[*pmaxrecs - 1].hashval);
442		if (level == 0) {
443			if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH) {
444				xchk_da_set_corrupt(ds, level);
445				goto out_freebp;
446			}
447			ds->tree_level = nodehdr.level;
448		} else {
449			if (ds->tree_level != nodehdr.level) {
450				xchk_da_set_corrupt(ds, level);
451				goto out_freebp;
452			}
453		}
454
455		/* XXX: Check hdr3.pad32 once we know how to fix it. */
456		break;
457	default:
458		xchk_da_set_corrupt(ds, level);
459		goto out_freebp;
460	}
461
462	fa = xfs_da3_header_check(blk->bp, dargs->owner);
463	if (fa) {
464		xchk_da_set_corrupt(ds, level);
465		goto out_freebp;
466	}
467
468	/*
469	 * If we've been handed a block that is below the dabtree root, does
470	 * its hashval match what the parent block expected to see?
471	 */
472	if (level > 0) {
473		struct xfs_da_node_entry	*key;
474
475		key = xchk_da_btree_node_entry(ds, level - 1);
476		if (be32_to_cpu(key->hashval) != blk->hashval) {
477			xchk_da_set_corrupt(ds, level);
478			goto out_freebp;
479		}
480	}
481
482out:
483	return error;
484out_freebp:
485	xfs_trans_brelse(dargs->trans, blk->bp);
486	blk->bp = NULL;
487out_nobuf:
488	blk->blkno = 0;
489	return error;
490}
491
492/* Visit all nodes and leaves of a da btree. */
493int
494xchk_da_btree(
495	struct xfs_scrub		*sc,
496	int				whichfork,
497	xchk_da_btree_rec_fn		scrub_fn,
498	void				*private)
499{
500	struct xchk_da_btree		*ds;
501	struct xfs_mount		*mp = sc->mp;
502	struct xfs_da_state_blk		*blks;
503	struct xfs_da_node_entry	*key;
504	xfs_dablk_t			blkno;
505	int				level;
506	int				error;
507
508	/* Skip short format data structures; no btree to scan. */
509	if (!xfs_ifork_has_extents(xfs_ifork_ptr(sc->ip, whichfork)))
510		return 0;
511
512	/* Set up initial da state. */
513	ds = kzalloc(sizeof(struct xchk_da_btree), XCHK_GFP_FLAGS);
514	if (!ds)
515		return -ENOMEM;
516	ds->dargs.dp = sc->ip;
517	ds->dargs.whichfork = whichfork;
518	ds->dargs.trans = sc->tp;
519	ds->dargs.op_flags = XFS_DA_OP_OKNOENT;
520	ds->dargs.owner = sc->ip->i_ino;
521	ds->state = xfs_da_state_alloc(&ds->dargs);
522	ds->sc = sc;
523	ds->private = private;
524	if (whichfork == XFS_ATTR_FORK) {
525		ds->dargs.geo = mp->m_attr_geo;
526		ds->lowest = 0;
527		ds->highest = 0;
528	} else {
529		ds->dargs.geo = mp->m_dir_geo;
530		ds->lowest = ds->dargs.geo->leafblk;
531		ds->highest = ds->dargs.geo->freeblk;
532	}
533	blkno = ds->lowest;
534	level = 0;
535
536	/* Find the root of the da tree, if present. */
537	blks = ds->state->path.blk;
538	error = xchk_da_btree_block(ds, level, blkno);
539	if (error)
540		goto out_state;
541	/*
542	 * We didn't find a block at ds->lowest, which means that there's
543	 * no LEAF1/LEAFN tree (at least not where it's supposed to be),
544	 * so jump out now.
545	 */
546	if (blks[level].bp == NULL)
547		goto out_state;
548
549	blks[level].index = 0;
550	while (level >= 0 && level < XFS_DA_NODE_MAXDEPTH) {
551		/* Handle leaf block. */
552		if (blks[level].magic != XFS_DA_NODE_MAGIC) {
553			/* End of leaf, pop back towards the root. */
554			if (blks[level].index >= ds->maxrecs[level]) {
555				if (level > 0)
556					blks[level - 1].index++;
557				ds->tree_level++;
558				level--;
559				continue;
560			}
561
562			/* Dispatch record scrubbing. */
563			error = scrub_fn(ds, level);
564			if (error)
565				break;
566			if (xchk_should_terminate(sc, &error) ||
567			    (sc->sm->sm_flags & XFS_SCRUB_OFLAG_CORRUPT))
568				break;
569
570			blks[level].index++;
571			continue;
572		}
573
574
575		/* End of node, pop back towards the root. */
576		if (blks[level].index >= ds->maxrecs[level]) {
577			if (level > 0)
578				blks[level - 1].index++;
579			ds->tree_level++;
580			level--;
581			continue;
582		}
583
584		/* Hashes in order for scrub? */
585		key = xchk_da_btree_node_entry(ds, level);
586		error = xchk_da_btree_hash(ds, level, &key->hashval);
587		if (error)
588			goto out;
589
590		/* Drill another level deeper. */
591		blkno = be32_to_cpu(key->before);
592		level++;
593		if (level >= XFS_DA_NODE_MAXDEPTH) {
594			/* Too deep! */
595			xchk_da_set_corrupt(ds, level - 1);
596			break;
597		}
598		ds->tree_level--;
599		error = xchk_da_btree_block(ds, level, blkno);
600		if (error)
601			goto out;
602		if (blks[level].bp == NULL)
603			goto out;
604
605		blks[level].index = 0;
606	}
607
608out:
609	/* Release all the buffers we're tracking. */
610	for (level = 0; level < XFS_DA_NODE_MAXDEPTH; level++) {
611		if (blks[level].bp == NULL)
612			continue;
613		xfs_trans_brelse(sc->tp, blks[level].bp);
614		blks[level].bp = NULL;
615	}
616
617out_state:
618	xfs_da_state_free(ds->state);
619	kfree(ds);
620	return error;
621}
622