xfs_log_recover.c 110.7 KB
Newer Older
L
Linus Torvalds 已提交
1
/*
2
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3
 * All Rights Reserved.
L
Linus Torvalds 已提交
4
 *
5 6
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
L
Linus Torvalds 已提交
7 8
 * published by the Free Software Foundation.
 *
9 10 11 12
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
L
Linus Torvalds 已提交
13
 *
14 15 16
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
L
Linus Torvalds 已提交
17 18
 */
#include "xfs.h"
19
#include "xfs_fs.h"
L
Linus Torvalds 已提交
20
#include "xfs_types.h"
21
#include "xfs_bit.h"
L
Linus Torvalds 已提交
22
#include "xfs_log.h"
23
#include "xfs_inum.h"
L
Linus Torvalds 已提交
24
#include "xfs_trans.h"
25 26
#include "xfs_sb.h"
#include "xfs_ag.h"
L
Linus Torvalds 已提交
27 28 29
#include "xfs_mount.h"
#include "xfs_error.h"
#include "xfs_bmap_btree.h"
30 31
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
32
#include "xfs_btree.h"
L
Linus Torvalds 已提交
33 34
#include "xfs_dinode.h"
#include "xfs_inode.h"
35 36
#include "xfs_inode_item.h"
#include "xfs_alloc.h"
L
Linus Torvalds 已提交
37 38 39 40 41 42 43
#include "xfs_ialloc.h"
#include "xfs_log_priv.h"
#include "xfs_buf_item.h"
#include "xfs_log_recover.h"
#include "xfs_extfree_item.h"
#include "xfs_trans_priv.h"
#include "xfs_quota.h"
44
#include "xfs_utils.h"
C
Christoph Hellwig 已提交
45
#include "xfs_cksum.h"
C
Christoph Hellwig 已提交
46
#include "xfs_trace.h"
D
Dave Chinner 已提交
47
#include "xfs_icache.h"
48 49

/* Need all the magic numbers and buffer ops structures from these headers */
50
#include "xfs_symlink.h"
51 52 53 54 55
#include "xfs_da_btree.h"
#include "xfs_dir2_format.h"
#include "xfs_dir2_priv.h"
#include "xfs_attr_leaf.h"
#include "xfs_attr_remote.h"
L
Linus Torvalds 已提交
56

M
Mark Tinguely 已提交
57 58 59 60 61 62 63 64
STATIC int
xlog_find_zeroed(
	struct xlog	*,
	xfs_daddr_t	*);
STATIC int
xlog_clear_stale_blocks(
	struct xlog	*,
	xfs_lsn_t);
L
Linus Torvalds 已提交
65
#if defined(DEBUG)
M
Mark Tinguely 已提交
66 67 68
STATIC void
xlog_recover_check_summary(
	struct xlog *);
L
Linus Torvalds 已提交
69 70 71 72
#else
#define	xlog_recover_check_summary(log)
#endif

73 74 75 76 77 78 79 80 81 82 83
/*
 * This structure is used during recovery to record the buf log items which
 * have been canceled and should not be replayed.
 */
struct xfs_buf_cancel {
	xfs_daddr_t		bc_blkno;
	uint			bc_len;
	int			bc_refcount;
	struct list_head	bc_list;
};

L
Linus Torvalds 已提交
84 85 86 87
/*
 * Sector aligned buffer routines for buffer create/read/write/access
 */

88 89 90 91 92 93 94 95
/*
 * Verify the given count of basic blocks is valid number of blocks
 * to specify for an operation involving the given XFS log buffer.
 * Returns nonzero if the count is valid, 0 otherwise.
 */

static inline int
xlog_buf_bbcount_valid(
M
Mark Tinguely 已提交
96
	struct xlog	*log,
97 98 99 100 101
	int		bbcount)
{
	return bbcount > 0 && bbcount <= log->l_logBBsize;
}

102 103 104 105 106
/*
 * Allocate a buffer to hold log data.  The buffer needs to be able
 * to map to a range of nbblks basic blocks at any valid (basic
 * block) offset within the log.
 */
107
STATIC xfs_buf_t *
L
Linus Torvalds 已提交
108
xlog_get_bp(
M
Mark Tinguely 已提交
109
	struct xlog	*log,
110
	int		nbblks)
L
Linus Torvalds 已提交
111
{
112 113
	struct xfs_buf	*bp;

114
	if (!xlog_buf_bbcount_valid(log, nbblks)) {
115
		xfs_warn(log->l_mp, "Invalid block length (0x%x) for buffer",
116 117
			nbblks);
		XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
118 119
		return NULL;
	}
L
Linus Torvalds 已提交
120

121 122 123
	/*
	 * We do log I/O in units of log sectors (a power-of-2
	 * multiple of the basic block size), so we round up the
L
Lucas De Marchi 已提交
124
	 * requested size to accommodate the basic blocks required
125 126 127 128 129 130 131 132 133 134
	 * for complete log sectors.
	 *
	 * In addition, the buffer may be used for a non-sector-
	 * aligned block offset, in which case an I/O of the
	 * requested size could extend beyond the end of the
	 * buffer.  If the requested size is only 1 basic block it
	 * will never straddle a sector boundary, so this won't be
	 * an issue.  Nor will this be a problem if the log I/O is
	 * done in basic blocks (sector size 1).  But otherwise we
	 * extend the buffer by one extra log sector to ensure
L
Lucas De Marchi 已提交
135
	 * there's space to accommodate this possibility.
136
	 */
137 138 139
	if (nbblks > 1 && log->l_sectBBsize > 1)
		nbblks += log->l_sectBBsize;
	nbblks = round_up(nbblks, log->l_sectBBsize);
140

141
	bp = xfs_buf_get_uncached(log->l_mp->m_logdev_targp, nbblks, 0);
142 143 144
	if (bp)
		xfs_buf_unlock(bp);
	return bp;
L
Linus Torvalds 已提交
145 146
}

147
STATIC void
L
Linus Torvalds 已提交
148 149 150 151 152 153
xlog_put_bp(
	xfs_buf_t	*bp)
{
	xfs_buf_free(bp);
}

A
Alex Elder 已提交
154 155 156 157
/*
 * Return the address of the start of the given block number's data
 * in a log buffer.  The buffer covers a log sector-aligned region.
 */
C
Christoph Hellwig 已提交
158 159
STATIC xfs_caddr_t
xlog_align(
M
Mark Tinguely 已提交
160
	struct xlog	*log,
C
Christoph Hellwig 已提交
161 162
	xfs_daddr_t	blk_no,
	int		nbblks,
M
Mark Tinguely 已提交
163
	struct xfs_buf	*bp)
C
Christoph Hellwig 已提交
164
{
C
Christoph Hellwig 已提交
165
	xfs_daddr_t	offset = blk_no & ((xfs_daddr_t)log->l_sectBBsize - 1);
C
Christoph Hellwig 已提交
166

167
	ASSERT(offset + nbblks <= bp->b_length);
168
	return bp->b_addr + BBTOB(offset);
C
Christoph Hellwig 已提交
169 170
}

L
Linus Torvalds 已提交
171 172 173 174

/*
 * nbblks should be uint, but oh well.  Just want to catch that 32-bit length.
 */
C
Christoph Hellwig 已提交
175 176
STATIC int
xlog_bread_noalign(
M
Mark Tinguely 已提交
177
	struct xlog	*log,
L
Linus Torvalds 已提交
178 179
	xfs_daddr_t	blk_no,
	int		nbblks,
M
Mark Tinguely 已提交
180
	struct xfs_buf	*bp)
L
Linus Torvalds 已提交
181 182 183
{
	int		error;

184
	if (!xlog_buf_bbcount_valid(log, nbblks)) {
185
		xfs_warn(log->l_mp, "Invalid block length (0x%x) for buffer",
186 187
			nbblks);
		XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
188 189 190
		return EFSCORRUPTED;
	}

191 192
	blk_no = round_down(blk_no, log->l_sectBBsize);
	nbblks = round_up(nbblks, log->l_sectBBsize);
L
Linus Torvalds 已提交
193 194

	ASSERT(nbblks > 0);
195
	ASSERT(nbblks <= bp->b_length);
L
Linus Torvalds 已提交
196 197 198

	XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
	XFS_BUF_READ(bp);
199
	bp->b_io_length = nbblks;
200
	bp->b_error = 0;
L
Linus Torvalds 已提交
201 202

	xfsbdstrat(log->l_mp, bp);
C
Christoph Hellwig 已提交
203
	error = xfs_buf_iowait(bp);
204
	if (error)
205
		xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
206 207 208
	return error;
}

C
Christoph Hellwig 已提交
209 210
STATIC int
xlog_bread(
M
Mark Tinguely 已提交
211
	struct xlog	*log,
C
Christoph Hellwig 已提交
212 213
	xfs_daddr_t	blk_no,
	int		nbblks,
M
Mark Tinguely 已提交
214
	struct xfs_buf	*bp,
C
Christoph Hellwig 已提交
215 216 217 218 219 220 221 222 223 224 225 226
	xfs_caddr_t	*offset)
{
	int		error;

	error = xlog_bread_noalign(log, blk_no, nbblks, bp);
	if (error)
		return error;

	*offset = xlog_align(log, blk_no, nbblks, bp);
	return 0;
}

227 228 229 230 231 232
/*
 * Read at an offset into the buffer. Returns with the buffer in it's original
 * state regardless of the result of the read.
 */
STATIC int
xlog_bread_offset(
M
Mark Tinguely 已提交
233
	struct xlog	*log,
234 235
	xfs_daddr_t	blk_no,		/* block to read from */
	int		nbblks,		/* blocks to read */
M
Mark Tinguely 已提交
236
	struct xfs_buf	*bp,
237 238
	xfs_caddr_t	offset)
{
239
	xfs_caddr_t	orig_offset = bp->b_addr;
240
	int		orig_len = BBTOB(bp->b_length);
241 242
	int		error, error2;

243
	error = xfs_buf_associate_memory(bp, offset, BBTOB(nbblks));
244 245 246 247 248 249
	if (error)
		return error;

	error = xlog_bread_noalign(log, blk_no, nbblks, bp);

	/* must reset buffer pointer even on error */
250
	error2 = xfs_buf_associate_memory(bp, orig_offset, orig_len);
251 252 253 254 255
	if (error)
		return error;
	return error2;
}

L
Linus Torvalds 已提交
256 257 258 259 260
/*
 * Write out the buffer at the given block for the given number of blocks.
 * The buffer is kept locked across the write and is returned locked.
 * This can only be used for synchronous log writes.
 */
261
STATIC int
L
Linus Torvalds 已提交
262
xlog_bwrite(
M
Mark Tinguely 已提交
263
	struct xlog	*log,
L
Linus Torvalds 已提交
264 265
	xfs_daddr_t	blk_no,
	int		nbblks,
M
Mark Tinguely 已提交
266
	struct xfs_buf	*bp)
L
Linus Torvalds 已提交
267 268 269
{
	int		error;

270
	if (!xlog_buf_bbcount_valid(log, nbblks)) {
271
		xfs_warn(log->l_mp, "Invalid block length (0x%x) for buffer",
272 273
			nbblks);
		XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
274 275 276
		return EFSCORRUPTED;
	}

277 278
	blk_no = round_down(blk_no, log->l_sectBBsize);
	nbblks = round_up(nbblks, log->l_sectBBsize);
L
Linus Torvalds 已提交
279 280

	ASSERT(nbblks > 0);
281
	ASSERT(nbblks <= bp->b_length);
L
Linus Torvalds 已提交
282 283 284

	XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
	XFS_BUF_ZEROFLAGS(bp);
285
	xfs_buf_hold(bp);
286
	xfs_buf_lock(bp);
287
	bp->b_io_length = nbblks;
288
	bp->b_error = 0;
L
Linus Torvalds 已提交
289

290
	error = xfs_bwrite(bp);
291 292
	if (error)
		xfs_buf_ioerror_alert(bp, __func__);
293
	xfs_buf_relse(bp);
L
Linus Torvalds 已提交
294 295 296 297 298 299 300 301 302 303 304 305
	return error;
}

#ifdef DEBUG
/*
 * dump debug superblock and log record information
 */
STATIC void
xlog_header_check_dump(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
{
306
	xfs_debug(mp, "%s:  SB : uuid = %pU, fmt = %d\n",
307
		__func__, &mp->m_sb.sb_uuid, XLOG_FMT);
308
	xfs_debug(mp, "    log : uuid = %pU, fmt = %d\n",
309
		&head->h_fs_uuid, be32_to_cpu(head->h_fmt));
L
Linus Torvalds 已提交
310 311 312 313 314 315 316 317 318 319 320 321 322
}
#else
#define xlog_header_check_dump(mp, head)
#endif

/*
 * check log record header for recovery
 */
STATIC int
xlog_header_check_recover(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
{
323
	ASSERT(head->h_magicno == cpu_to_be32(XLOG_HEADER_MAGIC_NUM));
L
Linus Torvalds 已提交
324 325 326 327 328 329

	/*
	 * IRIX doesn't write the h_fmt field and leaves it zeroed
	 * (XLOG_FMT_UNKNOWN). This stops us from trying to recover
	 * a dirty log created in IRIX.
	 */
330
	if (unlikely(head->h_fmt != cpu_to_be32(XLOG_FMT))) {
331 332
		xfs_warn(mp,
	"dirty log written in incompatible format - can't recover");
L
Linus Torvalds 已提交
333 334 335 336 337
		xlog_header_check_dump(mp, head);
		XFS_ERROR_REPORT("xlog_header_check_recover(1)",
				 XFS_ERRLEVEL_HIGH, mp);
		return XFS_ERROR(EFSCORRUPTED);
	} else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
338 339
		xfs_warn(mp,
	"dirty log entry has mismatched uuid - can't recover");
L
Linus Torvalds 已提交
340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
		xlog_header_check_dump(mp, head);
		XFS_ERROR_REPORT("xlog_header_check_recover(2)",
				 XFS_ERRLEVEL_HIGH, mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	return 0;
}

/*
 * read the head block of the log and check the header
 */
STATIC int
xlog_header_check_mount(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
{
356
	ASSERT(head->h_magicno == cpu_to_be32(XLOG_HEADER_MAGIC_NUM));
L
Linus Torvalds 已提交
357 358 359 360 361 362 363

	if (uuid_is_nil(&head->h_fs_uuid)) {
		/*
		 * IRIX doesn't write the h_fs_uuid or h_fmt fields. If
		 * h_fs_uuid is nil, we assume this log was last mounted
		 * by IRIX and continue.
		 */
364
		xfs_warn(mp, "nil uuid in log - IRIX style log");
L
Linus Torvalds 已提交
365
	} else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
366
		xfs_warn(mp, "log has mismatched uuid - can't recover");
L
Linus Torvalds 已提交
367 368 369 370 371 372 373 374 375 376 377 378
		xlog_header_check_dump(mp, head);
		XFS_ERROR_REPORT("xlog_header_check_mount",
				 XFS_ERRLEVEL_HIGH, mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	return 0;
}

STATIC void
xlog_recover_iodone(
	struct xfs_buf	*bp)
{
379
	if (bp->b_error) {
L
Linus Torvalds 已提交
380 381 382 383
		/*
		 * We're not going to bother about retrying
		 * this during recovery. One strike!
		 */
384
		xfs_buf_ioerror_alert(bp, __func__);
385 386
		xfs_force_shutdown(bp->b_target->bt_mount,
					SHUTDOWN_META_IO_ERROR);
L
Linus Torvalds 已提交
387
	}
388
	bp->b_iodone = NULL;
C
Christoph Hellwig 已提交
389
	xfs_buf_ioend(bp, 0);
L
Linus Torvalds 已提交
390 391 392 393 394 395 396 397
}

/*
 * This routine finds (to an approximation) the first block in the physical
 * log which contains the given cycle.  It uses a binary search algorithm.
 * Note that the algorithm can not be perfect because the disk will not
 * necessarily be perfect.
 */
D
David Chinner 已提交
398
STATIC int
L
Linus Torvalds 已提交
399
xlog_find_cycle_start(
M
Mark Tinguely 已提交
400 401
	struct xlog	*log,
	struct xfs_buf	*bp,
L
Linus Torvalds 已提交
402 403 404 405 406 407
	xfs_daddr_t	first_blk,
	xfs_daddr_t	*last_blk,
	uint		cycle)
{
	xfs_caddr_t	offset;
	xfs_daddr_t	mid_blk;
408
	xfs_daddr_t	end_blk;
L
Linus Torvalds 已提交
409 410 411
	uint		mid_cycle;
	int		error;

412 413 414
	end_blk = *last_blk;
	mid_blk = BLK_AVG(first_blk, end_blk);
	while (mid_blk != first_blk && mid_blk != end_blk) {
C
Christoph Hellwig 已提交
415 416
		error = xlog_bread(log, mid_blk, 1, bp, &offset);
		if (error)
L
Linus Torvalds 已提交
417
			return error;
418
		mid_cycle = xlog_get_cycle(offset);
419 420 421 422 423
		if (mid_cycle == cycle)
			end_blk = mid_blk;   /* last_half_cycle == mid_cycle */
		else
			first_blk = mid_blk; /* first_half_cycle == mid_cycle */
		mid_blk = BLK_AVG(first_blk, end_blk);
L
Linus Torvalds 已提交
424
	}
425 426 427 428
	ASSERT((mid_blk == first_blk && mid_blk+1 == end_blk) ||
	       (mid_blk == end_blk && mid_blk-1 == first_blk));

	*last_blk = end_blk;
L
Linus Torvalds 已提交
429 430 431 432 433

	return 0;
}

/*
434 435 436 437 438 439
 * Check that a range of blocks does not contain stop_on_cycle_no.
 * Fill in *new_blk with the block offset where such a block is
 * found, or with -1 (an invalid block number) if there is no such
 * block in the range.  The scan needs to occur from front to back
 * and the pointer into the region must be updated since a later
 * routine will need to perform another test.
L
Linus Torvalds 已提交
440 441 442
 */
STATIC int
xlog_find_verify_cycle(
M
Mark Tinguely 已提交
443
	struct xlog	*log,
L
Linus Torvalds 已提交
444 445 446 447 448 449 450 451 452 453 454 455
	xfs_daddr_t	start_blk,
	int		nbblks,
	uint		stop_on_cycle_no,
	xfs_daddr_t	*new_blk)
{
	xfs_daddr_t	i, j;
	uint		cycle;
	xfs_buf_t	*bp;
	xfs_daddr_t	bufblks;
	xfs_caddr_t	buf = NULL;
	int		error = 0;

456 457 458 459 460 461
	/*
	 * Greedily allocate a buffer big enough to handle the full
	 * range of basic blocks we'll be examining.  If that fails,
	 * try a smaller size.  We need to be able to read at least
	 * a log sector, or we're out of luck.
	 */
L
Linus Torvalds 已提交
462
	bufblks = 1 << ffs(nbblks);
463 464
	while (bufblks > log->l_logBBsize)
		bufblks >>= 1;
L
Linus Torvalds 已提交
465 466
	while (!(bp = xlog_get_bp(log, bufblks))) {
		bufblks >>= 1;
467
		if (bufblks < log->l_sectBBsize)
L
Linus Torvalds 已提交
468 469 470 471 472 473 474 475
			return ENOMEM;
	}

	for (i = start_blk; i < start_blk + nbblks; i += bufblks) {
		int	bcount;

		bcount = min(bufblks, (start_blk + nbblks - i));

C
Christoph Hellwig 已提交
476 477
		error = xlog_bread(log, i, bcount, bp, &buf);
		if (error)
L
Linus Torvalds 已提交
478 479 480
			goto out;

		for (j = 0; j < bcount; j++) {
481
			cycle = xlog_get_cycle(buf);
L
Linus Torvalds 已提交
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
			if (cycle == stop_on_cycle_no) {
				*new_blk = i+j;
				goto out;
			}

			buf += BBSIZE;
		}
	}

	*new_blk = -1;

out:
	xlog_put_bp(bp);
	return error;
}

/*
 * Potentially backup over partial log record write.
 *
 * In the typical case, last_blk is the number of the block directly after
 * a good log record.  Therefore, we subtract one to get the block number
 * of the last block in the given buffer.  extra_bblks contains the number
 * of blocks we would have read on a previous read.  This happens when the
 * last log record is split over the end of the physical log.
 *
 * extra_bblks is the number of blocks potentially verified on a previous
 * call to this routine.
 */
STATIC int
xlog_find_verify_log_record(
M
Mark Tinguely 已提交
512
	struct xlog		*log,
L
Linus Torvalds 已提交
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
	xfs_daddr_t		start_blk,
	xfs_daddr_t		*last_blk,
	int			extra_bblks)
{
	xfs_daddr_t		i;
	xfs_buf_t		*bp;
	xfs_caddr_t		offset = NULL;
	xlog_rec_header_t	*head = NULL;
	int			error = 0;
	int			smallmem = 0;
	int			num_blks = *last_blk - start_blk;
	int			xhdrs;

	ASSERT(start_blk != 0 || *last_blk != start_blk);

	if (!(bp = xlog_get_bp(log, num_blks))) {
		if (!(bp = xlog_get_bp(log, 1)))
			return ENOMEM;
		smallmem = 1;
	} else {
C
Christoph Hellwig 已提交
533 534
		error = xlog_bread(log, start_blk, num_blks, bp, &offset);
		if (error)
L
Linus Torvalds 已提交
535 536 537 538 539 540 541
			goto out;
		offset += ((num_blks - 1) << BBSHIFT);
	}

	for (i = (*last_blk) - 1; i >= 0; i--) {
		if (i < start_blk) {
			/* valid log record not found */
542 543
			xfs_warn(log->l_mp,
		"Log inconsistent (didn't find previous header)");
L
Linus Torvalds 已提交
544 545 546 547 548 549
			ASSERT(0);
			error = XFS_ERROR(EIO);
			goto out;
		}

		if (smallmem) {
C
Christoph Hellwig 已提交
550 551
			error = xlog_bread(log, i, 1, bp, &offset);
			if (error)
L
Linus Torvalds 已提交
552 553 554 555 556
				goto out;
		}

		head = (xlog_rec_header_t *)offset;

557
		if (head->h_magicno == cpu_to_be32(XLOG_HEADER_MAGIC_NUM))
L
Linus Torvalds 已提交
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
			break;

		if (!smallmem)
			offset -= BBSIZE;
	}

	/*
	 * We hit the beginning of the physical log & still no header.  Return
	 * to caller.  If caller can handle a return of -1, then this routine
	 * will be called again for the end of the physical log.
	 */
	if (i == -1) {
		error = -1;
		goto out;
	}

	/*
	 * We have the final block of the good log (the first block
	 * of the log record _before_ the head. So we check the uuid.
	 */
	if ((error = xlog_header_check_mount(log->l_mp, head)))
		goto out;

	/*
	 * We may have found a log record header before we expected one.
	 * last_blk will be the 1st block # with a given cycle #.  We may end
	 * up reading an entire log record.  In this case, we don't want to
	 * reset last_blk.  Only when last_blk points in the middle of a log
	 * record do we update last_blk.
	 */
588
	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
589
		uint	h_size = be32_to_cpu(head->h_size);
L
Linus Torvalds 已提交
590 591 592 593 594 595 596 597

		xhdrs = h_size / XLOG_HEADER_CYCLE_SIZE;
		if (h_size % XLOG_HEADER_CYCLE_SIZE)
			xhdrs++;
	} else {
		xhdrs = 1;
	}

598 599
	if (*last_blk - i + extra_bblks !=
	    BTOBB(be32_to_cpu(head->h_len)) + xhdrs)
L
Linus Torvalds 已提交
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619
		*last_blk = i;

out:
	xlog_put_bp(bp);
	return error;
}

/*
 * Head is defined to be the point of the log where the next log write
 * write could go.  This means that incomplete LR writes at the end are
 * eliminated when calculating the head.  We aren't guaranteed that previous
 * LR have complete transactions.  We only know that a cycle number of
 * current cycle number -1 won't be present in the log if we start writing
 * from our current block number.
 *
 * last_blk contains the block number of the first block with a given
 * cycle number.
 *
 * Return: zero if normal, non-zero if error.
 */
620
STATIC int
L
Linus Torvalds 已提交
621
xlog_find_head(
M
Mark Tinguely 已提交
622
	struct xlog	*log,
L
Linus Torvalds 已提交
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642
	xfs_daddr_t	*return_head_blk)
{
	xfs_buf_t	*bp;
	xfs_caddr_t	offset;
	xfs_daddr_t	new_blk, first_blk, start_blk, last_blk, head_blk;
	int		num_scan_bblks;
	uint		first_half_cycle, last_half_cycle;
	uint		stop_on_cycle;
	int		error, log_bbnum = log->l_logBBsize;

	/* Is the end of the log device zeroed? */
	if ((error = xlog_find_zeroed(log, &first_blk)) == -1) {
		*return_head_blk = first_blk;

		/* Is the whole lot zeroed? */
		if (!first_blk) {
			/* Linux XFS shouldn't generate totally zeroed logs -
			 * mkfs etc write a dummy unmount record to a fresh
			 * log so we can store the uuid in there
			 */
643
			xfs_warn(log->l_mp, "totally zeroed log");
L
Linus Torvalds 已提交
644 645 646 647
		}

		return 0;
	} else if (error) {
648
		xfs_warn(log->l_mp, "empty log check failed");
L
Linus Torvalds 已提交
649 650 651 652 653 654 655
		return error;
	}

	first_blk = 0;			/* get cycle # of 1st block */
	bp = xlog_get_bp(log, 1);
	if (!bp)
		return ENOMEM;
C
Christoph Hellwig 已提交
656 657 658

	error = xlog_bread(log, 0, 1, bp, &offset);
	if (error)
L
Linus Torvalds 已提交
659
		goto bp_err;
C
Christoph Hellwig 已提交
660

661
	first_half_cycle = xlog_get_cycle(offset);
L
Linus Torvalds 已提交
662 663

	last_blk = head_blk = log_bbnum - 1;	/* get cycle # of last block */
C
Christoph Hellwig 已提交
664 665
	error = xlog_bread(log, last_blk, 1, bp, &offset);
	if (error)
L
Linus Torvalds 已提交
666
		goto bp_err;
C
Christoph Hellwig 已提交
667

668
	last_half_cycle = xlog_get_cycle(offset);
L
Linus Torvalds 已提交
669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
	ASSERT(last_half_cycle != 0);

	/*
	 * If the 1st half cycle number is equal to the last half cycle number,
	 * then the entire log is stamped with the same cycle number.  In this
	 * case, head_blk can't be set to zero (which makes sense).  The below
	 * math doesn't work out properly with head_blk equal to zero.  Instead,
	 * we set it to log_bbnum which is an invalid block number, but this
	 * value makes the math correct.  If head_blk doesn't changed through
	 * all the tests below, *head_blk is set to zero at the very end rather
	 * than log_bbnum.  In a sense, log_bbnum and zero are the same block
	 * in a circular file.
	 */
	if (first_half_cycle == last_half_cycle) {
		/*
		 * In this case we believe that the entire log should have
		 * cycle number last_half_cycle.  We need to scan backwards
		 * from the end verifying that there are no holes still
		 * containing last_half_cycle - 1.  If we find such a hole,
		 * then the start of that hole will be the new head.  The
		 * simple case looks like
		 *        x | x ... | x - 1 | x
		 * Another case that fits this picture would be
		 *        x | x + 1 | x ... | x
693
		 * In this case the head really is somewhere at the end of the
L
Linus Torvalds 已提交
694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
		 * log, as one of the latest writes at the beginning was
		 * incomplete.
		 * One more case is
		 *        x | x + 1 | x ... | x - 1 | x
		 * This is really the combination of the above two cases, and
		 * the head has to end up at the start of the x-1 hole at the
		 * end of the log.
		 *
		 * In the 256k log case, we will read from the beginning to the
		 * end of the log and search for cycle numbers equal to x-1.
		 * We don't worry about the x+1 blocks that we encounter,
		 * because we know that they cannot be the head since the log
		 * started with x.
		 */
		head_blk = log_bbnum;
		stop_on_cycle = last_half_cycle - 1;
	} else {
		/*
		 * In this case we want to find the first block with cycle
		 * number matching last_half_cycle.  We expect the log to be
		 * some variation on
715
		 *        x + 1 ... | x ... | x
L
Linus Torvalds 已提交
716 717 718 719 720 721 722 723 724
		 * The first block with cycle number x (last_half_cycle) will
		 * be where the new head belongs.  First we do a binary search
		 * for the first occurrence of last_half_cycle.  The binary
		 * search may not be totally accurate, so then we scan back
		 * from there looking for occurrences of last_half_cycle before
		 * us.  If that backwards scan wraps around the beginning of
		 * the log, then we look for occurrences of last_half_cycle - 1
		 * at the end of the log.  The cases we're looking for look
		 * like
725 726 727
		 *                               v binary search stopped here
		 *        x + 1 ... | x | x + 1 | x ... | x
		 *                   ^ but we want to locate this spot
L
Linus Torvalds 已提交
728 729
		 * or
		 *        <---------> less than scan distance
730 731
		 *        x + 1 ... | x ... | x - 1 | x
		 *                           ^ we want to locate this spot
L
Linus Torvalds 已提交
732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
		 */
		stop_on_cycle = last_half_cycle;
		if ((error = xlog_find_cycle_start(log, bp, first_blk,
						&head_blk, last_half_cycle)))
			goto bp_err;
	}

	/*
	 * Now validate the answer.  Scan back some number of maximum possible
	 * blocks and make sure each one has the expected cycle number.  The
	 * maximum is determined by the total possible amount of buffering
	 * in the in-core log.  The following number can be made tighter if
	 * we actually look at the block size of the filesystem.
	 */
	num_scan_bblks = XLOG_TOTAL_REC_SHIFT(log);
	if (head_blk >= num_scan_bblks) {
		/*
		 * We are guaranteed that the entire check can be performed
		 * in one buffer.
		 */
		start_blk = head_blk - num_scan_bblks;
		if ((error = xlog_find_verify_cycle(log,
						start_blk, num_scan_bblks,
						stop_on_cycle, &new_blk)))
			goto bp_err;
		if (new_blk != -1)
			head_blk = new_blk;
	} else {		/* need to read 2 parts of log */
		/*
		 * We are going to scan backwards in the log in two parts.
		 * First we scan the physical end of the log.  In this part
		 * of the log, we are looking for blocks with cycle number
		 * last_half_cycle - 1.
		 * If we find one, then we know that the log starts there, as
		 * we've found a hole that didn't get written in going around
		 * the end of the physical log.  The simple case for this is
		 *        x + 1 ... | x ... | x - 1 | x
		 *        <---------> less than scan distance
		 * If all of the blocks at the end of the log have cycle number
		 * last_half_cycle, then we check the blocks at the start of
		 * the log looking for occurrences of last_half_cycle.  If we
		 * find one, then our current estimate for the location of the
		 * first occurrence of last_half_cycle is wrong and we move
		 * back to the hole we've found.  This case looks like
		 *        x + 1 ... | x | x + 1 | x ...
		 *                               ^ binary search stopped here
		 * Another case we need to handle that only occurs in 256k
		 * logs is
		 *        x + 1 ... | x ... | x+1 | x ...
		 *                   ^ binary search stops here
		 * In a 256k log, the scan at the end of the log will see the
		 * x + 1 blocks.  We need to skip past those since that is
		 * certainly not the head of the log.  By searching for
		 * last_half_cycle-1 we accomplish that.
		 */
		ASSERT(head_blk <= INT_MAX &&
788 789
			(xfs_daddr_t) num_scan_bblks >= head_blk);
		start_blk = log_bbnum - (num_scan_bblks - head_blk);
L
Linus Torvalds 已提交
790 791 792 793 794 795
		if ((error = xlog_find_verify_cycle(log, start_blk,
					num_scan_bblks - (int)head_blk,
					(stop_on_cycle - 1), &new_blk)))
			goto bp_err;
		if (new_blk != -1) {
			head_blk = new_blk;
796
			goto validate_head;
L
Linus Torvalds 已提交
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813
		}

		/*
		 * Scan beginning of log now.  The last part of the physical
		 * log is good.  This scan needs to verify that it doesn't find
		 * the last_half_cycle.
		 */
		start_blk = 0;
		ASSERT(head_blk <= INT_MAX);
		if ((error = xlog_find_verify_cycle(log,
					start_blk, (int)head_blk,
					stop_on_cycle, &new_blk)))
			goto bp_err;
		if (new_blk != -1)
			head_blk = new_blk;
	}

814
validate_head:
L
Linus Torvalds 已提交
815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835
	/*
	 * Now we need to make sure head_blk is not pointing to a block in
	 * the middle of a log record.
	 */
	num_scan_bblks = XLOG_REC_SHIFT(log);
	if (head_blk >= num_scan_bblks) {
		start_blk = head_blk - num_scan_bblks; /* don't read head_blk */

		/* start ptr at last block ptr before head_blk */
		if ((error = xlog_find_verify_log_record(log, start_blk,
							&head_blk, 0)) == -1) {
			error = XFS_ERROR(EIO);
			goto bp_err;
		} else if (error)
			goto bp_err;
	} else {
		start_blk = 0;
		ASSERT(head_blk <= INT_MAX);
		if ((error = xlog_find_verify_log_record(log, start_blk,
							&head_blk, 0)) == -1) {
			/* We hit the beginning of the log during our search */
836
			start_blk = log_bbnum - (num_scan_bblks - head_blk);
L
Linus Torvalds 已提交
837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870
			new_blk = log_bbnum;
			ASSERT(start_blk <= INT_MAX &&
				(xfs_daddr_t) log_bbnum-start_blk >= 0);
			ASSERT(head_blk <= INT_MAX);
			if ((error = xlog_find_verify_log_record(log,
							start_blk, &new_blk,
							(int)head_blk)) == -1) {
				error = XFS_ERROR(EIO);
				goto bp_err;
			} else if (error)
				goto bp_err;
			if (new_blk != log_bbnum)
				head_blk = new_blk;
		} else if (error)
			goto bp_err;
	}

	xlog_put_bp(bp);
	if (head_blk == log_bbnum)
		*return_head_blk = 0;
	else
		*return_head_blk = head_blk;
	/*
	 * When returning here, we have a good block number.  Bad block
	 * means that during a previous crash, we didn't have a clean break
	 * from cycle number N to cycle number N-1.  In this case, we need
	 * to find the first block with cycle number N-1.
	 */
	return 0;

 bp_err:
	xlog_put_bp(bp);

	if (error)
871
		xfs_warn(log->l_mp, "failed to find log head");
L
Linus Torvalds 已提交
872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
	return error;
}

/*
 * Find the sync block number or the tail of the log.
 *
 * This will be the block number of the last record to have its
 * associated buffers synced to disk.  Every log record header has
 * a sync lsn embedded in it.  LSNs hold block numbers, so it is easy
 * to get a sync block number.  The only concern is to figure out which
 * log record header to believe.
 *
 * The following algorithm uses the log record header with the largest
 * lsn.  The entire log record does not need to be valid.  We only care
 * that the header is valid.
 *
 * We could speed up search by using current head_blk buffer, but it is not
 * available.
 */
891
STATIC int
L
Linus Torvalds 已提交
892
xlog_find_tail(
M
Mark Tinguely 已提交
893
	struct xlog		*log,
L
Linus Torvalds 已提交
894
	xfs_daddr_t		*head_blk,
895
	xfs_daddr_t		*tail_blk)
L
Linus Torvalds 已提交
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918
{
	xlog_rec_header_t	*rhead;
	xlog_op_header_t	*op_head;
	xfs_caddr_t		offset = NULL;
	xfs_buf_t		*bp;
	int			error, i, found;
	xfs_daddr_t		umount_data_blk;
	xfs_daddr_t		after_umount_blk;
	xfs_lsn_t		tail_lsn;
	int			hblks;

	found = 0;

	/*
	 * Find previous log record
	 */
	if ((error = xlog_find_head(log, head_blk)))
		return error;

	bp = xlog_get_bp(log, 1);
	if (!bp)
		return ENOMEM;
	if (*head_blk == 0) {				/* special case */
C
Christoph Hellwig 已提交
919 920
		error = xlog_bread(log, 0, 1, bp, &offset);
		if (error)
921
			goto done;
C
Christoph Hellwig 已提交
922

923
		if (xlog_get_cycle(offset) == 0) {
L
Linus Torvalds 已提交
924 925
			*tail_blk = 0;
			/* leave all other log inited values alone */
926
			goto done;
L
Linus Torvalds 已提交
927 928 929 930 931 932 933 934
		}
	}

	/*
	 * Search backwards looking for log record header block
	 */
	ASSERT(*head_blk < INT_MAX);
	for (i = (int)(*head_blk) - 1; i >= 0; i--) {
C
Christoph Hellwig 已提交
935 936
		error = xlog_bread(log, i, 1, bp, &offset);
		if (error)
937
			goto done;
C
Christoph Hellwig 已提交
938

939
		if (*(__be32 *)offset == cpu_to_be32(XLOG_HEADER_MAGIC_NUM)) {
L
Linus Torvalds 已提交
940 941 942 943 944 945 946 947 948 949 950 951
			found = 1;
			break;
		}
	}
	/*
	 * If we haven't found the log record header block, start looking
	 * again from the end of the physical log.  XXXmiken: There should be
	 * a check here to make sure we didn't search more than N blocks in
	 * the previous code.
	 */
	if (!found) {
		for (i = log->l_logBBsize - 1; i >= (int)(*head_blk); i--) {
C
Christoph Hellwig 已提交
952 953
			error = xlog_bread(log, i, 1, bp, &offset);
			if (error)
954
				goto done;
C
Christoph Hellwig 已提交
955

956 957
			if (*(__be32 *)offset ==
			    cpu_to_be32(XLOG_HEADER_MAGIC_NUM)) {
L
Linus Torvalds 已提交
958 959 960 961 962 963
				found = 2;
				break;
			}
		}
	}
	if (!found) {
964
		xfs_warn(log->l_mp, "%s: couldn't find sync record", __func__);
L
Linus Torvalds 已提交
965 966 967 968 969 970
		ASSERT(0);
		return XFS_ERROR(EIO);
	}

	/* find blk_no of tail of log */
	rhead = (xlog_rec_header_t *)offset;
971
	*tail_blk = BLOCK_LSN(be64_to_cpu(rhead->h_tail_lsn));
L
Linus Torvalds 已提交
972 973 974 975 976 977 978 979 980 981 982 983 984

	/*
	 * Reset log values according to the state of the log when we
	 * crashed.  In the case where head_blk == 0, we bump curr_cycle
	 * one because the next write starts a new cycle rather than
	 * continuing the cycle of the last good log record.  At this
	 * point we have guaranteed that all partial log records have been
	 * accounted for.  Therefore, we know that the last good log record
	 * written was complete and ended exactly on the end boundary
	 * of the physical log.
	 */
	log->l_prev_block = i;
	log->l_curr_block = (int)*head_blk;
985
	log->l_curr_cycle = be32_to_cpu(rhead->h_cycle);
L
Linus Torvalds 已提交
986 987
	if (found == 2)
		log->l_curr_cycle++;
988
	atomic64_set(&log->l_tail_lsn, be64_to_cpu(rhead->h_tail_lsn));
989
	atomic64_set(&log->l_last_sync_lsn, be64_to_cpu(rhead->h_lsn));
990
	xlog_assign_grant_head(&log->l_reserve_head.grant, log->l_curr_cycle,
991
					BBTOB(log->l_curr_block));
992
	xlog_assign_grant_head(&log->l_write_head.grant, log->l_curr_cycle,
993
					BBTOB(log->l_curr_block));
L
Linus Torvalds 已提交
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005

	/*
	 * Look for unmount record.  If we find it, then we know there
	 * was a clean unmount.  Since 'i' could be the last block in
	 * the physical log, we convert to a log block before comparing
	 * to the head_blk.
	 *
	 * Save the current tail lsn to use to pass to
	 * xlog_clear_stale_blocks() below.  We won't want to clear the
	 * unmount record if there is one, so we pass the lsn of the
	 * unmount record rather than the block after it.
	 */
1006
	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
1007 1008
		int	h_size = be32_to_cpu(rhead->h_size);
		int	h_version = be32_to_cpu(rhead->h_version);
L
Linus Torvalds 已提交
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021

		if ((h_version & XLOG_VERSION_2) &&
		    (h_size > XLOG_HEADER_CYCLE_SIZE)) {
			hblks = h_size / XLOG_HEADER_CYCLE_SIZE;
			if (h_size % XLOG_HEADER_CYCLE_SIZE)
				hblks++;
		} else {
			hblks = 1;
		}
	} else {
		hblks = 1;
	}
	after_umount_blk = (i + hblks + (int)
1022
		BTOBB(be32_to_cpu(rhead->h_len))) % log->l_logBBsize;
1023
	tail_lsn = atomic64_read(&log->l_tail_lsn);
L
Linus Torvalds 已提交
1024
	if (*head_blk == after_umount_blk &&
1025
	    be32_to_cpu(rhead->h_num_logops) == 1) {
L
Linus Torvalds 已提交
1026
		umount_data_blk = (i + hblks) % log->l_logBBsize;
C
Christoph Hellwig 已提交
1027 1028
		error = xlog_bread(log, umount_data_blk, 1, bp, &offset);
		if (error)
1029
			goto done;
C
Christoph Hellwig 已提交
1030

L
Linus Torvalds 已提交
1031 1032 1033 1034 1035 1036 1037
		op_head = (xlog_op_header_t *)offset;
		if (op_head->oh_flags & XLOG_UNMOUNT_TRANS) {
			/*
			 * Set tail and last sync so that newly written
			 * log records will point recovery to after the
			 * current unmount record.
			 */
1038 1039 1040 1041
			xlog_assign_atomic_lsn(&log->l_tail_lsn,
					log->l_curr_cycle, after_umount_blk);
			xlog_assign_atomic_lsn(&log->l_last_sync_lsn,
					log->l_curr_cycle, after_umount_blk);
L
Linus Torvalds 已提交
1042
			*tail_blk = after_umount_blk;
D
David Chinner 已提交
1043 1044 1045 1046 1047 1048 1049 1050

			/*
			 * Note that the unmount was clean. If the unmount
			 * was not clean, we need to know this to rebuild the
			 * superblock counters from the perag headers if we
			 * have a filesystem using non-persistent counters.
			 */
			log->l_mp->m_flags |= XFS_MOUNT_WAS_CLEAN;
L
Linus Torvalds 已提交
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
		}
	}

	/*
	 * Make sure that there are no blocks in front of the head
	 * with the same cycle number as the head.  This can happen
	 * because we allow multiple outstanding log writes concurrently,
	 * and the later writes might make it out before earlier ones.
	 *
	 * We use the lsn from before modifying it so that we'll never
	 * overwrite the unmount record after a clean unmount.
	 *
	 * Do this only if we are going to recover the filesystem
	 *
	 * NOTE: This used to say "if (!readonly)"
	 * However on Linux, we can & do recover a read-only filesystem.
	 * We only skip recovery if NORECOVERY is specified on mount,
	 * in which case we would not be here.
	 *
	 * But... if the -device- itself is readonly, just skip this.
	 * We can't recover this device anyway, so it won't matter.
	 */
1073
	if (!xfs_readonly_buftarg(log->l_mp->m_logdev_targp))
L
Linus Torvalds 已提交
1074 1075
		error = xlog_clear_stale_blocks(log, tail_lsn);

1076
done:
L
Linus Torvalds 已提交
1077 1078 1079
	xlog_put_bp(bp);

	if (error)
1080
		xfs_warn(log->l_mp, "failed to locate log tail");
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
	return error;
}

/*
 * Is the log zeroed at all?
 *
 * The last binary search should be changed to perform an X block read
 * once X becomes small enough.  You can then search linearly through
 * the X blocks.  This will cut down on the number of reads we need to do.
 *
 * If the log is partially zeroed, this routine will pass back the blkno
 * of the first block with cycle number 0.  It won't have a complete LR
 * preceding it.
 *
 * Return:
 *	0  => the log is completely written to
 *	-1 => use *blk_no as the first block of the log
 *	>0 => error has occurred
 */
D
David Chinner 已提交
1100
STATIC int
L
Linus Torvalds 已提交
1101
xlog_find_zeroed(
M
Mark Tinguely 已提交
1102
	struct xlog	*log,
L
Linus Torvalds 已提交
1103 1104 1105 1106 1107 1108 1109 1110 1111
	xfs_daddr_t	*blk_no)
{
	xfs_buf_t	*bp;
	xfs_caddr_t	offset;
	uint	        first_cycle, last_cycle;
	xfs_daddr_t	new_blk, last_blk, start_blk;
	xfs_daddr_t     num_scan_bblks;
	int	        error, log_bbnum = log->l_logBBsize;

1112 1113
	*blk_no = 0;

L
Linus Torvalds 已提交
1114 1115 1116 1117
	/* check totally zeroed log */
	bp = xlog_get_bp(log, 1);
	if (!bp)
		return ENOMEM;
C
Christoph Hellwig 已提交
1118 1119
	error = xlog_bread(log, 0, 1, bp, &offset);
	if (error)
L
Linus Torvalds 已提交
1120
		goto bp_err;
C
Christoph Hellwig 已提交
1121

1122
	first_cycle = xlog_get_cycle(offset);
L
Linus Torvalds 已提交
1123 1124 1125 1126 1127 1128 1129
	if (first_cycle == 0) {		/* completely zeroed log */
		*blk_no = 0;
		xlog_put_bp(bp);
		return -1;
	}

	/* check partially zeroed log */
C
Christoph Hellwig 已提交
1130 1131
	error = xlog_bread(log, log_bbnum-1, 1, bp, &offset);
	if (error)
L
Linus Torvalds 已提交
1132
		goto bp_err;
C
Christoph Hellwig 已提交
1133

1134
	last_cycle = xlog_get_cycle(offset);
L
Linus Torvalds 已提交
1135 1136 1137 1138 1139 1140 1141 1142 1143
	if (last_cycle != 0) {		/* log completely written to */
		xlog_put_bp(bp);
		return 0;
	} else if (first_cycle != 1) {
		/*
		 * If the cycle of the last block is zero, the cycle of
		 * the first block must be 1. If it's not, maybe we're
		 * not looking at a log... Bail out.
		 */
1144 1145
		xfs_warn(log->l_mp,
			"Log inconsistent or not a log (last==0, first!=1)");
L
Linus Torvalds 已提交
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
		return XFS_ERROR(EINVAL);
	}

	/* we have a partially zeroed log */
	last_blk = log_bbnum-1;
	if ((error = xlog_find_cycle_start(log, bp, 0, &last_blk, 0)))
		goto bp_err;

	/*
	 * Validate the answer.  Because there is no way to guarantee that
	 * the entire log is made up of log records which are the same size,
	 * we scan over the defined maximum blocks.  At this point, the maximum
	 * is not chosen to mean anything special.   XXXmiken
	 */
	num_scan_bblks = XLOG_TOTAL_REC_SHIFT(log);
	ASSERT(num_scan_bblks <= INT_MAX);

	if (last_blk < num_scan_bblks)
		num_scan_bblks = last_blk;
	start_blk = last_blk - num_scan_bblks;

	/*
	 * We search for any instances of cycle number 0 that occur before
	 * our current estimate of the head.  What we're trying to detect is
	 *        1 ... | 0 | 1 | 0...
	 *                       ^ binary search ends here
	 */
	if ((error = xlog_find_verify_cycle(log, start_blk,
					 (int)num_scan_bblks, 0, &new_blk)))
		goto bp_err;
	if (new_blk != -1)
		last_blk = new_blk;

	/*
	 * Potentially backup over partial log record write.  We don't need
	 * to search the end of the log because we know it is zero.
	 */
	if ((error = xlog_find_verify_log_record(log, start_blk,
				&last_blk, 0)) == -1) {
	    error = XFS_ERROR(EIO);
	    goto bp_err;
	} else if (error)
	    goto bp_err;

	*blk_no = last_blk;
bp_err:
	xlog_put_bp(bp);
	if (error)
		return error;
	return -1;
}

/*
 * These are simple subroutines used by xlog_clear_stale_blocks() below
 * to initialize a buffer full of empty log record headers and write
 * them into the log.
 */
STATIC void
xlog_add_record(
M
Mark Tinguely 已提交
1205
	struct xlog		*log,
L
Linus Torvalds 已提交
1206 1207 1208 1209 1210 1211 1212 1213 1214
	xfs_caddr_t		buf,
	int			cycle,
	int			block,
	int			tail_cycle,
	int			tail_block)
{
	xlog_rec_header_t	*recp = (xlog_rec_header_t *)buf;

	memset(buf, 0, BBSIZE);
1215 1216 1217
	recp->h_magicno = cpu_to_be32(XLOG_HEADER_MAGIC_NUM);
	recp->h_cycle = cpu_to_be32(cycle);
	recp->h_version = cpu_to_be32(
1218
			xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? 2 : 1);
1219 1220 1221
	recp->h_lsn = cpu_to_be64(xlog_assign_lsn(cycle, block));
	recp->h_tail_lsn = cpu_to_be64(xlog_assign_lsn(tail_cycle, tail_block));
	recp->h_fmt = cpu_to_be32(XLOG_FMT);
L
Linus Torvalds 已提交
1222 1223 1224 1225 1226
	memcpy(&recp->h_fs_uuid, &log->l_mp->m_sb.sb_uuid, sizeof(uuid_t));
}

STATIC int
xlog_write_log_records(
M
Mark Tinguely 已提交
1227
	struct xlog	*log,
L
Linus Torvalds 已提交
1228 1229 1230 1231 1232 1233 1234 1235 1236
	int		cycle,
	int		start_block,
	int		blocks,
	int		tail_cycle,
	int		tail_block)
{
	xfs_caddr_t	offset;
	xfs_buf_t	*bp;
	int		balign, ealign;
1237
	int		sectbb = log->l_sectBBsize;
L
Linus Torvalds 已提交
1238 1239 1240 1241 1242
	int		end_block = start_block + blocks;
	int		bufblks;
	int		error = 0;
	int		i, j = 0;

1243 1244 1245 1246 1247 1248
	/*
	 * Greedily allocate a buffer big enough to handle the full
	 * range of basic blocks to be written.  If that fails, try
	 * a smaller size.  We need to be able to write at least a
	 * log sector, or we're out of luck.
	 */
L
Linus Torvalds 已提交
1249
	bufblks = 1 << ffs(blocks);
1250 1251
	while (bufblks > log->l_logBBsize)
		bufblks >>= 1;
L
Linus Torvalds 已提交
1252 1253
	while (!(bp = xlog_get_bp(log, bufblks))) {
		bufblks >>= 1;
1254
		if (bufblks < sectbb)
L
Linus Torvalds 已提交
1255 1256 1257 1258 1259 1260 1261
			return ENOMEM;
	}

	/* We may need to do a read at the start to fill in part of
	 * the buffer in the starting sector not covered by the first
	 * write below.
	 */
A
Alex Elder 已提交
1262
	balign = round_down(start_block, sectbb);
L
Linus Torvalds 已提交
1263
	if (balign != start_block) {
C
Christoph Hellwig 已提交
1264 1265 1266 1267
		error = xlog_bread_noalign(log, start_block, 1, bp);
		if (error)
			goto out_put_bp;

L
Linus Torvalds 已提交
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
		j = start_block - balign;
	}

	for (i = start_block; i < end_block; i += bufblks) {
		int		bcount, endcount;

		bcount = min(bufblks, end_block - start_block);
		endcount = bcount - j;

		/* We may need to do a read at the end to fill in part of
		 * the buffer in the final sector not covered by the write.
		 * If this is the same sector as the above read, skip it.
		 */
A
Alex Elder 已提交
1281
		ealign = round_down(end_block, sectbb);
L
Linus Torvalds 已提交
1282
		if (j == 0 && (start_block + endcount > ealign)) {
1283
			offset = bp->b_addr + BBTOB(ealign - start_block);
1284 1285
			error = xlog_bread_offset(log, ealign, sectbb,
							bp, offset);
C
Christoph Hellwig 已提交
1286 1287 1288
			if (error)
				break;

L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
		}

		offset = xlog_align(log, start_block, endcount, bp);
		for (; j < endcount; j++) {
			xlog_add_record(log, offset, cycle, i+j,
					tail_cycle, tail_block);
			offset += BBSIZE;
		}
		error = xlog_bwrite(log, start_block, endcount, bp);
		if (error)
			break;
		start_block += endcount;
		j = 0;
	}
C
Christoph Hellwig 已提交
1303 1304

 out_put_bp:
L
Linus Torvalds 已提交
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
	xlog_put_bp(bp);
	return error;
}

/*
 * This routine is called to blow away any incomplete log writes out
 * in front of the log head.  We do this so that we won't become confused
 * if we come up, write only a little bit more, and then crash again.
 * If we leave the partial log records out there, this situation could
 * cause us to think those partial writes are valid blocks since they
 * have the current cycle number.  We get rid of them by overwriting them
 * with empty log records with the old cycle number rather than the
 * current one.
 *
 * The tail lsn is passed in rather than taken from
 * the log so that we will not write over the unmount record after a
 * clean unmount in a 512 block log.  Doing so would leave the log without
 * any valid log records in it until a new one was written.  If we crashed
 * during that time we would not be able to recover.
 */
STATIC int
xlog_clear_stale_blocks(
M
Mark Tinguely 已提交
1327
	struct xlog	*log,
L
Linus Torvalds 已提交
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
	xfs_lsn_t	tail_lsn)
{
	int		tail_cycle, head_cycle;
	int		tail_block, head_block;
	int		tail_distance, max_distance;
	int		distance;
	int		error;

	tail_cycle = CYCLE_LSN(tail_lsn);
	tail_block = BLOCK_LSN(tail_lsn);
	head_cycle = log->l_curr_cycle;
	head_block = log->l_curr_block;

	/*
	 * Figure out the distance between the new head of the log
	 * and the tail.  We want to write over any blocks beyond the
	 * head that we may have written just before the crash, but
	 * we don't want to overwrite the tail of the log.
	 */
	if (head_cycle == tail_cycle) {
		/*
		 * The tail is behind the head in the physical log,
		 * so the distance from the head to the tail is the
		 * distance from the head to the end of the log plus
		 * the distance from the beginning of the log to the
		 * tail.
		 */
		if (unlikely(head_block < tail_block || head_block >= log->l_logBBsize)) {
			XFS_ERROR_REPORT("xlog_clear_stale_blocks(1)",
					 XFS_ERRLEVEL_LOW, log->l_mp);
			return XFS_ERROR(EFSCORRUPTED);
		}
		tail_distance = tail_block + (log->l_logBBsize - head_block);
	} else {
		/*
		 * The head is behind the tail in the physical log,
		 * so the distance from the head to the tail is just
		 * the tail block minus the head block.
		 */
		if (unlikely(head_block >= tail_block || head_cycle != (tail_cycle + 1))){
			XFS_ERROR_REPORT("xlog_clear_stale_blocks(2)",
					 XFS_ERRLEVEL_LOW, log->l_mp);
			return XFS_ERROR(EFSCORRUPTED);
		}
		tail_distance = tail_block - head_block;
	}

	/*
	 * If the head is right up against the tail, we can't clear
	 * anything.
	 */
	if (tail_distance <= 0) {
		ASSERT(tail_distance == 0);
		return 0;
	}

	max_distance = XLOG_TOTAL_REC_SHIFT(log);
	/*
	 * Take the smaller of the maximum amount of outstanding I/O
	 * we could have and the distance to the tail to clear out.
	 * We take the smaller so that we don't overwrite the tail and
	 * we don't waste all day writing from the head to the tail
	 * for no reason.
	 */
	max_distance = MIN(max_distance, tail_distance);

	if ((head_block + max_distance) <= log->l_logBBsize) {
		/*
		 * We can stomp all the blocks we need to without
		 * wrapping around the end of the log.  Just do it
		 * in a single write.  Use the cycle number of the
		 * current cycle minus one so that the log will look like:
		 *     n ... | n - 1 ...
		 */
		error = xlog_write_log_records(log, (head_cycle - 1),
				head_block, max_distance, tail_cycle,
				tail_block);
		if (error)
			return error;
	} else {
		/*
		 * We need to wrap around the end of the physical log in
		 * order to clear all the blocks.  Do it in two separate
		 * I/Os.  The first write should be from the head to the
		 * end of the physical log, and it should use the current
		 * cycle number minus one just like above.
		 */
		distance = log->l_logBBsize - head_block;
		error = xlog_write_log_records(log, (head_cycle - 1),
				head_block, distance, tail_cycle,
				tail_block);

		if (error)
			return error;

		/*
		 * Now write the blocks at the start of the physical log.
		 * This writes the remainder of the blocks we want to clear.
		 * It uses the current cycle number since we're now on the
		 * same cycle as the head so that we get:
		 *    n ... n ... | n - 1 ...
		 *    ^^^^^ blocks we're writing
		 */
		distance = max_distance - (log->l_logBBsize - head_block);
		error = xlog_write_log_records(log, head_cycle, 0, distance,
				tail_cycle, tail_block);
		if (error)
			return error;
	}

	return 0;
}

/******************************************************************************
 *
 *		Log recover routines
 *
 ******************************************************************************
 */

STATIC xlog_recover_t *
xlog_recover_find_tid(
1450
	struct hlist_head	*head,
L
Linus Torvalds 已提交
1451 1452
	xlog_tid_t		tid)
{
1453
	xlog_recover_t		*trans;
L
Linus Torvalds 已提交
1454

1455
	hlist_for_each_entry(trans, head, r_list) {
1456 1457
		if (trans->r_log_tid == tid)
			return trans;
L
Linus Torvalds 已提交
1458
	}
1459
	return NULL;
L
Linus Torvalds 已提交
1460 1461 1462
}

STATIC void
1463 1464 1465 1466
xlog_recover_new_tid(
	struct hlist_head	*head,
	xlog_tid_t		tid,
	xfs_lsn_t		lsn)
L
Linus Torvalds 已提交
1467
{
1468 1469 1470 1471 1472 1473 1474 1475 1476
	xlog_recover_t		*trans;

	trans = kmem_zalloc(sizeof(xlog_recover_t), KM_SLEEP);
	trans->r_log_tid   = tid;
	trans->r_lsn	   = lsn;
	INIT_LIST_HEAD(&trans->r_itemq);

	INIT_HLIST_NODE(&trans->r_list);
	hlist_add_head(&trans->r_list, head);
L
Linus Torvalds 已提交
1477 1478 1479 1480
}

STATIC void
xlog_recover_add_item(
1481
	struct list_head	*head)
L
Linus Torvalds 已提交
1482 1483 1484 1485
{
	xlog_recover_item_t	*item;

	item = kmem_zalloc(sizeof(xlog_recover_item_t), KM_SLEEP);
1486 1487
	INIT_LIST_HEAD(&item->ri_list);
	list_add_tail(&item->ri_list, head);
L
Linus Torvalds 已提交
1488 1489 1490 1491
}

STATIC int
xlog_recover_add_to_cont_trans(
1492 1493
	struct xlog		*log,
	struct xlog_recover	*trans,
L
Linus Torvalds 已提交
1494 1495 1496 1497 1498 1499 1500
	xfs_caddr_t		dp,
	int			len)
{
	xlog_recover_item_t	*item;
	xfs_caddr_t		ptr, old_ptr;
	int			old_len;

1501
	if (list_empty(&trans->r_itemq)) {
L
Linus Torvalds 已提交
1502 1503 1504 1505 1506 1507 1508
		/* finish copying rest of trans header */
		xlog_recover_add_item(&trans->r_itemq);
		ptr = (xfs_caddr_t) &trans->r_theader +
				sizeof(xfs_trans_header_t) - len;
		memcpy(ptr, dp, len); /* d, s, l */
		return 0;
	}
1509 1510
	/* take the tail entry */
	item = list_entry(trans->r_itemq.prev, xlog_recover_item_t, ri_list);
L
Linus Torvalds 已提交
1511 1512 1513 1514

	old_ptr = item->ri_buf[item->ri_cnt-1].i_addr;
	old_len = item->ri_buf[item->ri_cnt-1].i_len;

1515
	ptr = kmem_realloc(old_ptr, len+old_len, old_len, KM_SLEEP);
L
Linus Torvalds 已提交
1516 1517 1518
	memcpy(&ptr[old_len], dp, len); /* d, s, l */
	item->ri_buf[item->ri_cnt-1].i_len += len;
	item->ri_buf[item->ri_cnt-1].i_addr = ptr;
1519
	trace_xfs_log_recover_item_add_cont(log, trans, item, 0);
L
Linus Torvalds 已提交
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
	return 0;
}

/*
 * The next region to add is the start of a new region.  It could be
 * a whole region or it could be the first part of a new region.  Because
 * of this, the assumption here is that the type and size fields of all
 * format structures fit into the first 32 bits of the structure.
 *
 * This works because all regions must be 32 bit aligned.  Therefore, we
 * either have both fields or we have neither field.  In the case we have
 * neither field, the data part of the region is zero length.  We only have
 * a log_op_header and can throw away the header since a new one will appear
 * later.  If we have at least 4 bytes, then we can determine how many regions
 * will appear in the current log item.
 */
STATIC int
xlog_recover_add_to_trans(
1538 1539
	struct xlog		*log,
	struct xlog_recover	*trans,
L
Linus Torvalds 已提交
1540 1541 1542 1543 1544 1545 1546 1547 1548
	xfs_caddr_t		dp,
	int			len)
{
	xfs_inode_log_format_t	*in_f;			/* any will do */
	xlog_recover_item_t	*item;
	xfs_caddr_t		ptr;

	if (!len)
		return 0;
1549
	if (list_empty(&trans->r_itemq)) {
1550 1551
		/* we need to catch log corruptions here */
		if (*(uint *)dp != XFS_TRANS_HEADER_MAGIC) {
1552 1553
			xfs_warn(log->l_mp, "%s: bad header magic number",
				__func__);
1554 1555 1556
			ASSERT(0);
			return XFS_ERROR(EIO);
		}
L
Linus Torvalds 已提交
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
		if (len == sizeof(xfs_trans_header_t))
			xlog_recover_add_item(&trans->r_itemq);
		memcpy(&trans->r_theader, dp, len); /* d, s, l */
		return 0;
	}

	ptr = kmem_alloc(len, KM_SLEEP);
	memcpy(ptr, dp, len);
	in_f = (xfs_inode_log_format_t *)ptr;

1567 1568 1569 1570 1571
	/* take the tail entry */
	item = list_entry(trans->r_itemq.prev, xlog_recover_item_t, ri_list);
	if (item->ri_total != 0 &&
	     item->ri_total == item->ri_cnt) {
		/* tail item is in use, get a new one */
L
Linus Torvalds 已提交
1572
		xlog_recover_add_item(&trans->r_itemq);
1573 1574
		item = list_entry(trans->r_itemq.prev,
					xlog_recover_item_t, ri_list);
L
Linus Torvalds 已提交
1575 1576 1577
	}

	if (item->ri_total == 0) {		/* first region to be added */
1578 1579
		if (in_f->ilf_size == 0 ||
		    in_f->ilf_size > XLOG_MAX_REGIONS_IN_ITEM) {
1580 1581
			xfs_warn(log->l_mp,
		"bad number of regions (%d) in inode log format",
1582 1583 1584 1585 1586 1587 1588 1589 1590
				  in_f->ilf_size);
			ASSERT(0);
			return XFS_ERROR(EIO);
		}

		item->ri_total = in_f->ilf_size;
		item->ri_buf =
			kmem_zalloc(item->ri_total * sizeof(xfs_log_iovec_t),
				    KM_SLEEP);
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595 1596
	}
	ASSERT(item->ri_total > item->ri_cnt);
	/* Description region is ri_buf[0] */
	item->ri_buf[item->ri_cnt].i_addr = ptr;
	item->ri_buf[item->ri_cnt].i_len  = len;
	item->ri_cnt++;
1597
	trace_xfs_log_recover_item_add(log, trans, item, 0);
L
Linus Torvalds 已提交
1598 1599 1600
	return 0;
}

1601 1602 1603 1604 1605 1606
/*
 * Sort the log items in the transaction. Cancelled buffers need
 * to be put first so they are processed before any items that might
 * modify the buffers. If they are cancelled, then the modifications
 * don't need to be replayed.
 */
L
Linus Torvalds 已提交
1607 1608
STATIC int
xlog_recover_reorder_trans(
1609 1610
	struct xlog		*log,
	struct xlog_recover	*trans,
1611
	int			pass)
L
Linus Torvalds 已提交
1612
{
1613 1614 1615 1616 1617
	xlog_recover_item_t	*item, *n;
	LIST_HEAD(sort_list);

	list_splice_init(&trans->r_itemq, &sort_list);
	list_for_each_entry_safe(item, n, &sort_list, ri_list) {
1618
		xfs_buf_log_format_t	*buf_f = item->ri_buf[0].i_addr;
L
Linus Torvalds 已提交
1619

1620
		switch (ITEM_TYPE(item)) {
L
Linus Torvalds 已提交
1621
		case XFS_LI_BUF:
1622
			if (!(buf_f->blf_flags & XFS_BLF_CANCEL)) {
1623 1624
				trace_xfs_log_recover_item_reorder_head(log,
							trans, item, pass);
1625
				list_move(&item->ri_list, &trans->r_itemq);
L
Linus Torvalds 已提交
1626 1627 1628 1629 1630 1631 1632
				break;
			}
		case XFS_LI_INODE:
		case XFS_LI_DQUOT:
		case XFS_LI_QUOTAOFF:
		case XFS_LI_EFD:
		case XFS_LI_EFI:
1633 1634
			trace_xfs_log_recover_item_reorder_tail(log,
							trans, item, pass);
1635
			list_move_tail(&item->ri_list, &trans->r_itemq);
L
Linus Torvalds 已提交
1636 1637
			break;
		default:
1638 1639 1640
			xfs_warn(log->l_mp,
				"%s: unrecognized type of log operation",
				__func__);
L
Linus Torvalds 已提交
1641 1642 1643
			ASSERT(0);
			return XFS_ERROR(EIO);
		}
1644 1645
	}
	ASSERT(list_empty(&sort_list));
L
Linus Torvalds 已提交
1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
	return 0;
}

/*
 * Build up the table of buf cancel records so that we don't replay
 * cancelled data in the second pass.  For buffer records that are
 * not cancel records, there is nothing to do here so we just return.
 *
 * If we get a cancel record which is already in the table, this indicates
 * that the buffer was cancelled multiple times.  In order to ensure
 * that during pass 2 we keep the record in the table until we reach its
 * last occurrence in the log, we keep a reference count in the cancel
 * record in the table to tell us how many times we expect to see this
 * record during the second pass.
 */
1661 1662
STATIC int
xlog_recover_buffer_pass1(
1663 1664
	struct xlog			*log,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
1665
{
1666
	xfs_buf_log_format_t	*buf_f = item->ri_buf[0].i_addr;
1667 1668
	struct list_head	*bucket;
	struct xfs_buf_cancel	*bcp;
L
Linus Torvalds 已提交
1669 1670 1671 1672

	/*
	 * If this isn't a cancel buffer item, then just return.
	 */
1673
	if (!(buf_f->blf_flags & XFS_BLF_CANCEL)) {
1674
		trace_xfs_log_recover_buf_not_cancel(log, buf_f);
1675
		return 0;
1676
	}
L
Linus Torvalds 已提交
1677 1678

	/*
1679 1680
	 * Insert an xfs_buf_cancel record into the hash table of them.
	 * If there is already an identical record, bump its reference count.
L
Linus Torvalds 已提交
1681
	 */
1682 1683 1684 1685 1686
	bucket = XLOG_BUF_CANCEL_BUCKET(log, buf_f->blf_blkno);
	list_for_each_entry(bcp, bucket, bc_list) {
		if (bcp->bc_blkno == buf_f->blf_blkno &&
		    bcp->bc_len == buf_f->blf_len) {
			bcp->bc_refcount++;
1687
			trace_xfs_log_recover_buf_cancel_ref_inc(log, buf_f);
1688
			return 0;
L
Linus Torvalds 已提交
1689
		}
1690 1691 1692 1693 1694
	}

	bcp = kmem_alloc(sizeof(struct xfs_buf_cancel), KM_SLEEP);
	bcp->bc_blkno = buf_f->blf_blkno;
	bcp->bc_len = buf_f->blf_len;
L
Linus Torvalds 已提交
1695
	bcp->bc_refcount = 1;
1696 1697
	list_add_tail(&bcp->bc_list, bucket);

1698
	trace_xfs_log_recover_buf_cancel_add(log, buf_f);
1699
	return 0;
L
Linus Torvalds 已提交
1700 1701 1702 1703 1704 1705
}

/*
 * Check to see whether the buffer being recovered has a corresponding
 * entry in the buffer cancel record table.  If it does then return 1
 * so that it will be cancelled, otherwise return 0.  If the buffer is
1706
 * actually a buffer cancel item (XFS_BLF_CANCEL is set), then decrement
L
Linus Torvalds 已提交
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
 * the refcount on the entry in the table and remove it from the table
 * if this is the last reference.
 *
 * We remove the cancel record from the table when we encounter its
 * last occurrence in the log so that if the same buffer is re-used
 * again after its last cancellation we actually replay the changes
 * made at that point.
 */
STATIC int
xlog_check_buffer_cancelled(
1717
	struct xlog		*log,
L
Linus Torvalds 已提交
1718 1719 1720 1721
	xfs_daddr_t		blkno,
	uint			len,
	ushort			flags)
{
1722 1723
	struct list_head	*bucket;
	struct xfs_buf_cancel	*bcp;
L
Linus Torvalds 已提交
1724 1725 1726 1727 1728 1729

	if (log->l_buf_cancel_table == NULL) {
		/*
		 * There is nothing in the table built in pass one,
		 * so this buffer must not be cancelled.
		 */
1730
		ASSERT(!(flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
1731 1732 1733 1734
		return 0;
	}

	/*
1735
	 * Search for an entry in the  cancel table that matches our buffer.
L
Linus Torvalds 已提交
1736
	 */
1737 1738 1739 1740
	bucket = XLOG_BUF_CANCEL_BUCKET(log, blkno);
	list_for_each_entry(bcp, bucket, bc_list) {
		if (bcp->bc_blkno == blkno && bcp->bc_len == len)
			goto found;
L
Linus Torvalds 已提交
1741
	}
1742

L
Linus Torvalds 已提交
1743
	/*
1744 1745
	 * We didn't find a corresponding entry in the table, so return 0 so
	 * that the buffer is NOT cancelled.
L
Linus Torvalds 已提交
1746
	 */
1747
	ASSERT(!(flags & XFS_BLF_CANCEL));
L
Linus Torvalds 已提交
1748
	return 0;
1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763

found:
	/*
	 * We've go a match, so return 1 so that the recovery of this buffer
	 * is cancelled.  If this buffer is actually a buffer cancel log
	 * item, then decrement the refcount on the one in the table and
	 * remove it if this is the last reference.
	 */
	if (flags & XFS_BLF_CANCEL) {
		if (--bcp->bc_refcount == 0) {
			list_del(&bcp->bc_list);
			kmem_free(bcp);
		}
	}
	return 1;
L
Linus Torvalds 已提交
1764 1765 1766
}

/*
1767 1768 1769 1770 1771
 * Perform recovery for a buffer full of inodes.  In these buffers, the only
 * data which should be recovered is that which corresponds to the
 * di_next_unlinked pointers in the on disk inode structures.  The rest of the
 * data for the inodes is always logged through the inodes themselves rather
 * than the inode buffer and is recovered in xlog_recover_inode_pass2().
L
Linus Torvalds 已提交
1772
 *
1773 1774 1775 1776
 * The only time when buffers full of inodes are fully recovered is when the
 * buffer is full of newly allocated inodes.  In this case the buffer will
 * not be marked as an inode buffer and so will be sent to
 * xlog_recover_do_reg_buffer() below during recovery.
L
Linus Torvalds 已提交
1777 1778 1779
 */
STATIC int
xlog_recover_do_inode_buffer(
1780
	struct xfs_mount	*mp,
L
Linus Torvalds 已提交
1781
	xlog_recover_item_t	*item,
1782
	struct xfs_buf		*bp,
L
Linus Torvalds 已提交
1783 1784 1785
	xfs_buf_log_format_t	*buf_f)
{
	int			i;
1786 1787 1788 1789 1790
	int			item_index = 0;
	int			bit = 0;
	int			nbits = 0;
	int			reg_buf_offset = 0;
	int			reg_buf_bytes = 0;
L
Linus Torvalds 已提交
1791 1792 1793 1794 1795
	int			next_unlinked_offset;
	int			inodes_per_buf;
	xfs_agino_t		*logged_nextp;
	xfs_agino_t		*buffer_nextp;

1796
	trace_xfs_log_recover_buf_inode_buf(mp->m_log, buf_f);
1797
	bp->b_ops = &xfs_inode_buf_ops;
1798

1799
	inodes_per_buf = BBTOB(bp->b_io_length) >> mp->m_sb.sb_inodelog;
L
Linus Torvalds 已提交
1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812
	for (i = 0; i < inodes_per_buf; i++) {
		next_unlinked_offset = (i * mp->m_sb.sb_inodesize) +
			offsetof(xfs_dinode_t, di_next_unlinked);

		while (next_unlinked_offset >=
		       (reg_buf_offset + reg_buf_bytes)) {
			/*
			 * The next di_next_unlinked field is beyond
			 * the current logged region.  Find the next
			 * logged region that contains or is beyond
			 * the current di_next_unlinked field.
			 */
			bit += nbits;
1813 1814
			bit = xfs_next_bit(buf_f->blf_data_map,
					   buf_f->blf_map_size, bit);
L
Linus Torvalds 已提交
1815 1816 1817 1818 1819

			/*
			 * If there are no more logged regions in the
			 * buffer, then we're done.
			 */
1820
			if (bit == -1)
L
Linus Torvalds 已提交
1821 1822
				return 0;

1823 1824
			nbits = xfs_contig_bits(buf_f->blf_data_map,
						buf_f->blf_map_size, bit);
L
Linus Torvalds 已提交
1825
			ASSERT(nbits > 0);
1826 1827
			reg_buf_offset = bit << XFS_BLF_SHIFT;
			reg_buf_bytes = nbits << XFS_BLF_SHIFT;
L
Linus Torvalds 已提交
1828 1829 1830 1831 1832 1833 1834 1835
			item_index++;
		}

		/*
		 * If the current logged region starts after the current
		 * di_next_unlinked field, then move on to the next
		 * di_next_unlinked field.
		 */
1836
		if (next_unlinked_offset < reg_buf_offset)
L
Linus Torvalds 已提交
1837 1838 1839
			continue;

		ASSERT(item->ri_buf[item_index].i_addr != NULL);
1840
		ASSERT((item->ri_buf[item_index].i_len % XFS_BLF_CHUNK) == 0);
1841 1842
		ASSERT((reg_buf_offset + reg_buf_bytes) <=
							BBTOB(bp->b_io_length));
L
Linus Torvalds 已提交
1843 1844 1845 1846 1847 1848

		/*
		 * The current logged region contains a copy of the
		 * current di_next_unlinked field.  Extract its value
		 * and copy it to the buffer copy.
		 */
1849 1850
		logged_nextp = item->ri_buf[item_index].i_addr +
				next_unlinked_offset - reg_buf_offset;
L
Linus Torvalds 已提交
1851
		if (unlikely(*logged_nextp == 0)) {
1852 1853 1854
			xfs_alert(mp,
		"Bad inode buffer log record (ptr = 0x%p, bp = 0x%p). "
		"Trying to replay bad (0) inode di_next_unlinked field.",
L
Linus Torvalds 已提交
1855 1856 1857 1858 1859 1860 1861 1862
				item, bp);
			XFS_ERROR_REPORT("xlog_recover_do_inode_buf",
					 XFS_ERRLEVEL_LOW, mp);
			return XFS_ERROR(EFSCORRUPTED);
		}

		buffer_nextp = (xfs_agino_t *)xfs_buf_offset(bp,
					      next_unlinked_offset);
1863
		*buffer_nextp = *logged_nextp;
L
Linus Torvalds 已提交
1864 1865 1866 1867 1868 1869
	}

	return 0;
}

/*
1870 1871 1872 1873 1874 1875
 * Validate the recovered buffer is of the correct type and attach the
 * appropriate buffer operations to them for writeback. Magic numbers are in a
 * few places:
 *	the first 16 bits of the buffer (inode buffer, dquot buffer),
 *	the first 32 bits of the buffer (most blocks),
 *	inside a struct xfs_da_blkinfo at the start of the buffer.
L
Linus Torvalds 已提交
1876
 */
1877 1878
static void
xlog_recovery_validate_buf_type(
1879
	struct xfs_mount	*mp,
1880
	struct xfs_buf		*bp,
L
Linus Torvalds 已提交
1881 1882
	xfs_buf_log_format_t	*buf_f)
{
1883 1884 1885 1886 1887 1888 1889 1890
	struct xfs_da_blkinfo	*info = bp->b_addr;
	__uint32_t		magic32;
	__uint16_t		magic16;
	__uint16_t		magicda;

	magic32 = be32_to_cpu(*(__be32 *)bp->b_addr);
	magic16 = be16_to_cpu(*(__be16*)bp->b_addr);
	magicda = be16_to_cpu(info->magic);
1891 1892
	switch (xfs_blft_from_flags(buf_f)) {
	case XFS_BLFT_BTREE_BUF:
1893
		switch (magic32) {
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
		case XFS_ABTB_CRC_MAGIC:
		case XFS_ABTC_CRC_MAGIC:
		case XFS_ABTB_MAGIC:
		case XFS_ABTC_MAGIC:
			bp->b_ops = &xfs_allocbt_buf_ops;
			break;
		case XFS_IBT_CRC_MAGIC:
		case XFS_IBT_MAGIC:
			bp->b_ops = &xfs_inobt_buf_ops;
			break;
		case XFS_BMAP_CRC_MAGIC:
		case XFS_BMAP_MAGIC:
			bp->b_ops = &xfs_bmbt_buf_ops;
			break;
		default:
			xfs_warn(mp, "Bad btree block magic!");
			ASSERT(0);
			break;
		}
		break;
1914
	case XFS_BLFT_AGF_BUF:
1915
		if (magic32 != XFS_AGF_MAGIC) {
D
Dave Chinner 已提交
1916 1917 1918 1919 1920 1921
			xfs_warn(mp, "Bad AGF block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_agf_buf_ops;
		break;
1922
	case XFS_BLFT_AGFL_BUF:
1923 1924
		if (!xfs_sb_version_hascrc(&mp->m_sb))
			break;
1925
		if (magic32 != XFS_AGFL_MAGIC) {
1926 1927 1928 1929 1930 1931
			xfs_warn(mp, "Bad AGFL block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_agfl_buf_ops;
		break;
1932
	case XFS_BLFT_AGI_BUF:
1933
		if (magic32 != XFS_AGI_MAGIC) {
D
Dave Chinner 已提交
1934 1935 1936 1937 1938 1939
			xfs_warn(mp, "Bad AGI block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_agi_buf_ops;
		break;
1940 1941 1942
	case XFS_BLFT_UDQUOT_BUF:
	case XFS_BLFT_PDQUOT_BUF:
	case XFS_BLFT_GDQUOT_BUF:
1943
#ifdef CONFIG_XFS_QUOTA
1944
		if (magic16 != XFS_DQUOT_MAGIC) {
1945 1946 1947 1948 1949
			xfs_warn(mp, "Bad DQUOT block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dquot_buf_ops;
1950 1951 1952 1953 1954
#else
		xfs_alert(mp,
	"Trying to recover dquots without QUOTA support built in!");
		ASSERT(0);
#endif
1955
		break;
1956
	case XFS_BLFT_DINO_BUF:
1957 1958 1959 1960
		/*
		 * we get here with inode allocation buffers, not buffers that
		 * track unlinked list changes.
		 */
1961
		if (magic16 != XFS_DINODE_MAGIC) {
1962 1963 1964 1965 1966 1967
			xfs_warn(mp, "Bad INODE block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_inode_buf_ops;
		break;
1968
	case XFS_BLFT_SYMLINK_BUF:
1969
		if (magic32 != XFS_SYMLINK_MAGIC) {
1970 1971 1972 1973 1974 1975
			xfs_warn(mp, "Bad symlink block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_symlink_buf_ops;
		break;
1976
	case XFS_BLFT_DIR_BLOCK_BUF:
1977 1978 1979 1980 1981 1982 1983 1984
		if (magic32 != XFS_DIR2_BLOCK_MAGIC &&
		    magic32 != XFS_DIR3_BLOCK_MAGIC) {
			xfs_warn(mp, "Bad dir block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dir3_block_buf_ops;
		break;
1985
	case XFS_BLFT_DIR_DATA_BUF:
1986 1987 1988 1989 1990 1991 1992 1993
		if (magic32 != XFS_DIR2_DATA_MAGIC &&
		    magic32 != XFS_DIR3_DATA_MAGIC) {
			xfs_warn(mp, "Bad dir data magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dir3_data_buf_ops;
		break;
1994
	case XFS_BLFT_DIR_FREE_BUF:
1995 1996 1997 1998 1999 2000 2001 2002
		if (magic32 != XFS_DIR2_FREE_MAGIC &&
		    magic32 != XFS_DIR3_FREE_MAGIC) {
			xfs_warn(mp, "Bad dir3 free magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dir3_free_buf_ops;
		break;
2003
	case XFS_BLFT_DIR_LEAF1_BUF:
2004 2005 2006 2007 2008 2009 2010 2011
		if (magicda != XFS_DIR2_LEAF1_MAGIC &&
		    magicda != XFS_DIR3_LEAF1_MAGIC) {
			xfs_warn(mp, "Bad dir leaf1 magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dir3_leaf1_buf_ops;
		break;
2012
	case XFS_BLFT_DIR_LEAFN_BUF:
2013 2014 2015 2016 2017 2018 2019 2020
		if (magicda != XFS_DIR2_LEAFN_MAGIC &&
		    magicda != XFS_DIR3_LEAFN_MAGIC) {
			xfs_warn(mp, "Bad dir leafn magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_dir3_leafn_buf_ops;
		break;
2021
	case XFS_BLFT_DA_NODE_BUF:
2022 2023 2024 2025 2026 2027 2028 2029
		if (magicda != XFS_DA_NODE_MAGIC &&
		    magicda != XFS_DA3_NODE_MAGIC) {
			xfs_warn(mp, "Bad da node magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_da3_node_buf_ops;
		break;
2030
	case XFS_BLFT_ATTR_LEAF_BUF:
2031 2032 2033 2034 2035 2036 2037 2038
		if (magicda != XFS_ATTR_LEAF_MAGIC &&
		    magicda != XFS_ATTR3_LEAF_MAGIC) {
			xfs_warn(mp, "Bad attr leaf magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_attr3_leaf_buf_ops;
		break;
2039
	case XFS_BLFT_ATTR_RMT_BUF:
2040 2041 2042 2043 2044 2045 2046 2047 2048
		if (!xfs_sb_version_hascrc(&mp->m_sb))
			break;
		if (magicda != XFS_ATTR3_RMT_MAGIC) {
			xfs_warn(mp, "Bad attr remote magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_attr3_rmt_buf_ops;
		break;
2049 2050 2051 2052 2053 2054 2055 2056
	case XFS_BLFT_SB_BUF:
		if (magic32 != XFS_SB_MAGIC) {
			xfs_warn(mp, "Bad SB block magic!");
			ASSERT(0);
			break;
		}
		bp->b_ops = &xfs_sb_buf_ops;
		break;
2057
	default:
2058 2059
		xfs_warn(mp, "Unknown buffer type %d!",
			 xfs_blft_from_flags(buf_f));
2060 2061
		break;
	}
L
Linus Torvalds 已提交
2062 2063
}

2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140
/*
 * Perform a 'normal' buffer recovery.  Each logged region of the
 * buffer should be copied over the corresponding region in the
 * given buffer.  The bitmap in the buf log format structure indicates
 * where to place the logged data.
 */
STATIC void
xlog_recover_do_reg_buffer(
	struct xfs_mount	*mp,
	xlog_recover_item_t	*item,
	struct xfs_buf		*bp,
	xfs_buf_log_format_t	*buf_f)
{
	int			i;
	int			bit;
	int			nbits;
	int                     error;

	trace_xfs_log_recover_buf_reg_buf(mp->m_log, buf_f);

	bit = 0;
	i = 1;  /* 0 is the buf format structure */
	while (1) {
		bit = xfs_next_bit(buf_f->blf_data_map,
				   buf_f->blf_map_size, bit);
		if (bit == -1)
			break;
		nbits = xfs_contig_bits(buf_f->blf_data_map,
					buf_f->blf_map_size, bit);
		ASSERT(nbits > 0);
		ASSERT(item->ri_buf[i].i_addr != NULL);
		ASSERT(item->ri_buf[i].i_len % XFS_BLF_CHUNK == 0);
		ASSERT(BBTOB(bp->b_io_length) >=
		       ((uint)bit << XFS_BLF_SHIFT) + (nbits << XFS_BLF_SHIFT));

		/*
		 * Do a sanity check if this is a dquot buffer. Just checking
		 * the first dquot in the buffer should do. XXXThis is
		 * probably a good thing to do for other buf types also.
		 */
		error = 0;
		if (buf_f->blf_flags &
		   (XFS_BLF_UDQUOT_BUF|XFS_BLF_PDQUOT_BUF|XFS_BLF_GDQUOT_BUF)) {
			if (item->ri_buf[i].i_addr == NULL) {
				xfs_alert(mp,
					"XFS: NULL dquot in %s.", __func__);
				goto next;
			}
			if (item->ri_buf[i].i_len < sizeof(xfs_disk_dquot_t)) {
				xfs_alert(mp,
					"XFS: dquot too small (%d) in %s.",
					item->ri_buf[i].i_len, __func__);
				goto next;
			}
			error = xfs_qm_dqcheck(mp, item->ri_buf[i].i_addr,
					       -1, 0, XFS_QMOPT_DOWARN,
					       "dquot_buf_recover");
			if (error)
				goto next;
		}

		memcpy(xfs_buf_offset(bp,
			(uint)bit << XFS_BLF_SHIFT),	/* dest */
			item->ri_buf[i].i_addr,		/* source */
			nbits<<XFS_BLF_SHIFT);		/* length */
 next:
		i++;
		bit += nbits;
	}

	/* Shouldn't be any more regions */
	ASSERT(i == item->ri_total);

	xlog_recovery_validate_buf_type(mp, bp, buf_f);

}

L
Linus Torvalds 已提交
2141 2142 2143 2144 2145
/*
 * Do some primitive error checking on ondisk dquot data structures.
 */
int
xfs_qm_dqcheck(
2146
	struct xfs_mount *mp,
L
Linus Torvalds 已提交
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
	xfs_disk_dquot_t *ddq,
	xfs_dqid_t	 id,
	uint		 type,	  /* used only when IO_dorepair is true */
	uint		 flags,
	char		 *str)
{
	xfs_dqblk_t	 *d = (xfs_dqblk_t *)ddq;
	int		errs = 0;

	/*
	 * We can encounter an uninitialized dquot buffer for 2 reasons:
	 * 1. If we crash while deleting the quotainode(s), and those blks got
	 *    used for user data. This is because we take the path of regular
	 *    file deletion; however, the size field of quotainodes is never
	 *    updated, so all the tricks that we play in itruncate_finish
	 *    don't quite matter.
	 *
	 * 2. We don't play the quota buffers when there's a quotaoff logitem.
	 *    But the allocation will be replayed so we'll end up with an
	 *    uninitialized quota block.
	 *
	 * This is all fine; things are still consistent, and we haven't lost
	 * any quota information. Just don't complain about bad dquot blks.
	 */
2171
	if (ddq->d_magic != cpu_to_be16(XFS_DQUOT_MAGIC)) {
L
Linus Torvalds 已提交
2172
		if (flags & XFS_QMOPT_DOWARN)
2173
			xfs_alert(mp,
L
Linus Torvalds 已提交
2174
			"%s : XFS dquot ID 0x%x, magic 0x%x != 0x%x",
2175
			str, id, be16_to_cpu(ddq->d_magic), XFS_DQUOT_MAGIC);
L
Linus Torvalds 已提交
2176 2177
		errs++;
	}
2178
	if (ddq->d_version != XFS_DQUOT_VERSION) {
L
Linus Torvalds 已提交
2179
		if (flags & XFS_QMOPT_DOWARN)
2180
			xfs_alert(mp,
L
Linus Torvalds 已提交
2181
			"%s : XFS dquot ID 0x%x, version 0x%x != 0x%x",
2182
			str, id, ddq->d_version, XFS_DQUOT_VERSION);
L
Linus Torvalds 已提交
2183 2184 2185
		errs++;
	}

2186 2187 2188
	if (ddq->d_flags != XFS_DQ_USER &&
	    ddq->d_flags != XFS_DQ_PROJ &&
	    ddq->d_flags != XFS_DQ_GROUP) {
L
Linus Torvalds 已提交
2189
		if (flags & XFS_QMOPT_DOWARN)
2190
			xfs_alert(mp,
L
Linus Torvalds 已提交
2191
			"%s : XFS dquot ID 0x%x, unknown flags 0x%x",
2192
			str, id, ddq->d_flags);
L
Linus Torvalds 已提交
2193 2194 2195
		errs++;
	}

2196
	if (id != -1 && id != be32_to_cpu(ddq->d_id)) {
L
Linus Torvalds 已提交
2197
		if (flags & XFS_QMOPT_DOWARN)
2198
			xfs_alert(mp,
L
Linus Torvalds 已提交
2199 2200
			"%s : ondisk-dquot 0x%p, ID mismatch: "
			"0x%x expected, found id 0x%x",
2201
			str, ddq, id, be32_to_cpu(ddq->d_id));
L
Linus Torvalds 已提交
2202 2203 2204 2205
		errs++;
	}

	if (!errs && ddq->d_id) {
2206
		if (ddq->d_blk_softlimit &&
2207
		    be64_to_cpu(ddq->d_bcount) >
2208
				be64_to_cpu(ddq->d_blk_softlimit)) {
L
Linus Torvalds 已提交
2209 2210
			if (!ddq->d_btimer) {
				if (flags & XFS_QMOPT_DOWARN)
2211 2212
					xfs_alert(mp,
			"%s : Dquot ID 0x%x (0x%p) BLK TIMER NOT STARTED",
2213
					str, (int)be32_to_cpu(ddq->d_id), ddq);
L
Linus Torvalds 已提交
2214 2215 2216
				errs++;
			}
		}
2217
		if (ddq->d_ino_softlimit &&
2218
		    be64_to_cpu(ddq->d_icount) >
2219
				be64_to_cpu(ddq->d_ino_softlimit)) {
L
Linus Torvalds 已提交
2220 2221
			if (!ddq->d_itimer) {
				if (flags & XFS_QMOPT_DOWARN)
2222 2223
					xfs_alert(mp,
			"%s : Dquot ID 0x%x (0x%p) INODE TIMER NOT STARTED",
2224
					str, (int)be32_to_cpu(ddq->d_id), ddq);
L
Linus Torvalds 已提交
2225 2226 2227
				errs++;
			}
		}
2228
		if (ddq->d_rtb_softlimit &&
2229
		    be64_to_cpu(ddq->d_rtbcount) >
2230
				be64_to_cpu(ddq->d_rtb_softlimit)) {
L
Linus Torvalds 已提交
2231 2232
			if (!ddq->d_rtbtimer) {
				if (flags & XFS_QMOPT_DOWARN)
2233 2234
					xfs_alert(mp,
			"%s : Dquot ID 0x%x (0x%p) RTBLK TIMER NOT STARTED",
2235
					str, (int)be32_to_cpu(ddq->d_id), ddq);
L
Linus Torvalds 已提交
2236 2237 2238 2239 2240 2241 2242 2243 2244
				errs++;
			}
		}
	}

	if (!errs || !(flags & XFS_QMOPT_DQREPAIR))
		return errs;

	if (flags & XFS_QMOPT_DOWARN)
2245
		xfs_notice(mp, "Re-initializing dquot ID 0x%x", id);
L
Linus Torvalds 已提交
2246 2247 2248 2249 2250 2251 2252

	/*
	 * Typically, a repair is only requested by quotacheck.
	 */
	ASSERT(id != -1);
	ASSERT(flags & XFS_QMOPT_DQREPAIR);
	memset(d, 0, sizeof(xfs_dqblk_t));
2253 2254 2255 2256 2257

	d->dd_diskdq.d_magic = cpu_to_be16(XFS_DQUOT_MAGIC);
	d->dd_diskdq.d_version = XFS_DQUOT_VERSION;
	d->dd_diskdq.d_flags = type;
	d->dd_diskdq.d_id = cpu_to_be32(id);
L
Linus Torvalds 已提交
2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269

	return errs;
}

/*
 * Perform a dquot buffer recovery.
 * Simple algorithm: if we have found a QUOTAOFF logitem of the same type
 * (ie. USR or GRP), then just toss this buffer away; don't recover it.
 * Else, treat it as a regular buffer and do recovery.
 */
STATIC void
xlog_recover_do_dquot_buffer(
M
Mark Tinguely 已提交
2270 2271 2272 2273 2274
	struct xfs_mount		*mp,
	struct xlog			*log,
	struct xlog_recover_item	*item,
	struct xfs_buf			*bp,
	struct xfs_buf_log_format	*buf_f)
L
Linus Torvalds 已提交
2275 2276 2277
{
	uint			type;

2278 2279
	trace_xfs_log_recover_buf_dquot_buf(log, buf_f);

L
Linus Torvalds 已提交
2280 2281 2282 2283 2284 2285 2286 2287
	/*
	 * Filesystems are required to send in quota flags at mount time.
	 */
	if (mp->m_qflags == 0) {
		return;
	}

	type = 0;
2288
	if (buf_f->blf_flags & XFS_BLF_UDQUOT_BUF)
L
Linus Torvalds 已提交
2289
		type |= XFS_DQ_USER;
2290
	if (buf_f->blf_flags & XFS_BLF_PDQUOT_BUF)
2291
		type |= XFS_DQ_PROJ;
2292
	if (buf_f->blf_flags & XFS_BLF_GDQUOT_BUF)
L
Linus Torvalds 已提交
2293 2294 2295 2296 2297 2298 2299
		type |= XFS_DQ_GROUP;
	/*
	 * This type of quotas was turned off, so ignore this buffer
	 */
	if (log->l_quotaoffs_flag & type)
		return;

2300
	xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
L
Linus Torvalds 已提交
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
}

/*
 * This routine replays a modification made to a buffer at runtime.
 * There are actually two types of buffer, regular and inode, which
 * are handled differently.  Inode buffers are handled differently
 * in that we only recover a specific set of data from them, namely
 * the inode di_next_unlinked fields.  This is because all other inode
 * data is actually logged via inode records and any data we replay
 * here which overlaps that may be stale.
 *
 * When meta-data buffers are freed at run time we log a buffer item
2313
 * with the XFS_BLF_CANCEL bit set to indicate that previous copies
L
Linus Torvalds 已提交
2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326
 * of the buffer in the log should not be replayed at recovery time.
 * This is so that if the blocks covered by the buffer are reused for
 * file data before we crash we don't end up replaying old, freed
 * meta-data into a user's file.
 *
 * To handle the cancellation of buffer log items, we make two passes
 * over the log during recovery.  During the first we build a table of
 * those buffers which have been cancelled, and during the second we
 * only replay those buffers which do not have corresponding cancel
 * records in the table.  See xlog_recover_do_buffer_pass[1,2] above
 * for more details on the implementation of the table of cancel records.
 */
STATIC int
2327
xlog_recover_buffer_pass2(
M
Mark Tinguely 已提交
2328 2329 2330
	struct xlog			*log,
	struct list_head		*buffer_list,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
2331
{
2332
	xfs_buf_log_format_t	*buf_f = item->ri_buf[0].i_addr;
2333
	xfs_mount_t		*mp = log->l_mp;
L
Linus Torvalds 已提交
2334 2335
	xfs_buf_t		*bp;
	int			error;
2336
	uint			buf_flags;
L
Linus Torvalds 已提交
2337

2338 2339 2340 2341 2342 2343 2344
	/*
	 * In this pass we only want to recover all the buffers which have
	 * not been cancelled and are not cancellation buffers themselves.
	 */
	if (xlog_check_buffer_cancelled(log, buf_f->blf_blkno,
			buf_f->blf_len, buf_f->blf_flags)) {
		trace_xfs_log_recover_buf_cancel(log, buf_f);
L
Linus Torvalds 已提交
2345 2346
		return 0;
	}
2347

2348
	trace_xfs_log_recover_buf_recover(log, buf_f);
L
Linus Torvalds 已提交
2349

D
Dave Chinner 已提交
2350
	buf_flags = 0;
2351 2352
	if (buf_f->blf_flags & XFS_BLF_INODE_BUF)
		buf_flags |= XBF_UNMAPPED;
2353

2354
	bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
2355
			  buf_flags, NULL);
2356 2357
	if (!bp)
		return XFS_ERROR(ENOMEM);
2358
	error = bp->b_error;
2359
	if (error) {
2360
		xfs_buf_ioerror_alert(bp, "xlog_recover_do..(read#1)");
L
Linus Torvalds 已提交
2361 2362 2363 2364
		xfs_buf_relse(bp);
		return error;
	}

2365
	if (buf_f->blf_flags & XFS_BLF_INODE_BUF) {
L
Linus Torvalds 已提交
2366
		error = xlog_recover_do_inode_buffer(mp, item, bp, buf_f);
2367
	} else if (buf_f->blf_flags &
2368
		  (XFS_BLF_UDQUOT_BUF|XFS_BLF_PDQUOT_BUF|XFS_BLF_GDQUOT_BUF)) {
L
Linus Torvalds 已提交
2369 2370
		xlog_recover_do_dquot_buffer(mp, log, item, bp, buf_f);
	} else {
2371
		xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
L
Linus Torvalds 已提交
2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
	}
	if (error)
		return XFS_ERROR(error);

	/*
	 * Perform delayed write on the buffer.  Asynchronous writes will be
	 * slower when taking into account all the buffers to be flushed.
	 *
	 * Also make sure that only inode buffers with good sizes stay in
	 * the buffer cache.  The kernel moves inodes in buffers of 1 block
	 * or XFS_INODE_CLUSTER_SIZE bytes, whichever is bigger.  The inode
	 * buffers in the log can be a different size if the log was generated
	 * by an older kernel using unclustered inode buffers or a newer kernel
	 * running with a different inode cluster size.  Regardless, if the
	 * the inode buffer size isn't MAX(blocksize, XFS_INODE_CLUSTER_SIZE)
	 * for *our* value of XFS_INODE_CLUSTER_SIZE, then we need to keep
	 * the buffer out of the buffer cache so that the buffer won't
	 * overlap with future reads of those inodes.
	 */
	if (XFS_DINODE_MAGIC ==
2392
	    be16_to_cpu(*((__be16 *)xfs_buf_offset(bp, 0))) &&
2393
	    (BBTOB(bp->b_io_length) != MAX(log->l_mp->m_sb.sb_blocksize,
L
Linus Torvalds 已提交
2394
			(__uint32_t)XFS_INODE_CLUSTER_SIZE(log->l_mp)))) {
2395
		xfs_buf_stale(bp);
2396
		error = xfs_bwrite(bp);
L
Linus Torvalds 已提交
2397
	} else {
2398
		ASSERT(bp->b_target->bt_mount == mp);
2399
		bp->b_iodone = xlog_recover_iodone;
2400
		xfs_buf_delwri_queue(bp, buffer_list);
L
Linus Torvalds 已提交
2401 2402
	}

2403 2404
	xfs_buf_relse(bp);
	return error;
L
Linus Torvalds 已提交
2405 2406 2407
}

STATIC int
2408
xlog_recover_inode_pass2(
M
Mark Tinguely 已提交
2409 2410 2411
	struct xlog			*log,
	struct list_head		*buffer_list,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
2412 2413
{
	xfs_inode_log_format_t	*in_f;
2414
	xfs_mount_t		*mp = log->l_mp;
L
Linus Torvalds 已提交
2415 2416 2417 2418 2419 2420 2421 2422
	xfs_buf_t		*bp;
	xfs_dinode_t		*dip;
	int			len;
	xfs_caddr_t		src;
	xfs_caddr_t		dest;
	int			error;
	int			attr_index;
	uint			fields;
2423
	xfs_icdinode_t		*dicp;
2424
	uint			isize;
2425
	int			need_free = 0;
L
Linus Torvalds 已提交
2426

2427
	if (item->ri_buf[0].i_len == sizeof(xfs_inode_log_format_t)) {
2428
		in_f = item->ri_buf[0].i_addr;
2429
	} else {
2430
		in_f = kmem_alloc(sizeof(xfs_inode_log_format_t), KM_SLEEP);
2431 2432 2433 2434 2435
		need_free = 1;
		error = xfs_inode_item_format_convert(&item->ri_buf[0], in_f);
		if (error)
			goto error;
	}
L
Linus Torvalds 已提交
2436 2437 2438 2439 2440

	/*
	 * Inode buffers can be freed, look out for it,
	 * and do not replay the inode.
	 */
2441 2442
	if (xlog_check_buffer_cancelled(log, in_f->ilf_blkno,
					in_f->ilf_len, 0)) {
2443
		error = 0;
2444
		trace_xfs_log_recover_inode_cancel(log, in_f);
2445 2446
		goto error;
	}
2447
	trace_xfs_log_recover_inode_recover(log, in_f);
L
Linus Torvalds 已提交
2448

2449
	bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0,
2450
			  &xfs_inode_buf_ops);
2451 2452 2453 2454
	if (!bp) {
		error = ENOMEM;
		goto error;
	}
2455
	error = bp->b_error;
2456
	if (error) {
2457
		xfs_buf_ioerror_alert(bp, "xlog_recover_do..(read#2)");
L
Linus Torvalds 已提交
2458
		xfs_buf_relse(bp);
2459
		goto error;
L
Linus Torvalds 已提交
2460 2461
	}
	ASSERT(in_f->ilf_fields & XFS_ILOG_CORE);
2462
	dip = (xfs_dinode_t *)xfs_buf_offset(bp, in_f->ilf_boffset);
L
Linus Torvalds 已提交
2463 2464 2465 2466 2467

	/*
	 * Make sure the place we're flushing out to really looks
	 * like an inode!
	 */
2468
	if (unlikely(dip->di_magic != cpu_to_be16(XFS_DINODE_MAGIC))) {
L
Linus Torvalds 已提交
2469
		xfs_buf_relse(bp);
2470 2471 2472
		xfs_alert(mp,
	"%s: Bad inode magic number, dip = 0x%p, dino bp = 0x%p, ino = %Ld",
			__func__, dip, bp, in_f->ilf_ino);
2473
		XFS_ERROR_REPORT("xlog_recover_inode_pass2(1)",
L
Linus Torvalds 已提交
2474
				 XFS_ERRLEVEL_LOW, mp);
2475 2476
		error = EFSCORRUPTED;
		goto error;
L
Linus Torvalds 已提交
2477
	}
2478
	dicp = item->ri_buf[1].i_addr;
L
Linus Torvalds 已提交
2479 2480
	if (unlikely(dicp->di_magic != XFS_DINODE_MAGIC)) {
		xfs_buf_relse(bp);
2481 2482 2483
		xfs_alert(mp,
			"%s: Bad inode log record, rec ptr 0x%p, ino %Ld",
			__func__, item, in_f->ilf_ino);
2484
		XFS_ERROR_REPORT("xlog_recover_inode_pass2(2)",
L
Linus Torvalds 已提交
2485
				 XFS_ERRLEVEL_LOW, mp);
2486 2487
		error = EFSCORRUPTED;
		goto error;
L
Linus Torvalds 已提交
2488 2489 2490
	}

	/* Skip replay when the on disk inode is newer than the log one */
C
Christoph Hellwig 已提交
2491
	if (dicp->di_flushiter < be16_to_cpu(dip->di_flushiter)) {
L
Linus Torvalds 已提交
2492 2493 2494 2495
		/*
		 * Deal with the wrap case, DI_MAX_FLUSH is less
		 * than smaller numbers
		 */
C
Christoph Hellwig 已提交
2496
		if (be16_to_cpu(dip->di_flushiter) == DI_MAX_FLUSH &&
2497
		    dicp->di_flushiter < (DI_MAX_FLUSH >> 1)) {
L
Linus Torvalds 已提交
2498 2499 2500
			/* do nothing */
		} else {
			xfs_buf_relse(bp);
2501
			trace_xfs_log_recover_inode_skip(log, in_f);
2502 2503
			error = 0;
			goto error;
L
Linus Torvalds 已提交
2504 2505 2506 2507 2508
		}
	}
	/* Take the opportunity to reset the flush iteration count */
	dicp->di_flushiter = 0;

2509
	if (unlikely(S_ISREG(dicp->di_mode))) {
L
Linus Torvalds 已提交
2510 2511
		if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (dicp->di_format != XFS_DINODE_FMT_BTREE)) {
2512
			XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(3)",
L
Linus Torvalds 已提交
2513 2514
					 XFS_ERRLEVEL_LOW, mp, dicp);
			xfs_buf_relse(bp);
2515 2516 2517 2518
			xfs_alert(mp,
		"%s: Bad regular inode log record, rec ptr 0x%p, "
		"ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
				__func__, item, dip, bp, in_f->ilf_ino);
2519 2520
			error = EFSCORRUPTED;
			goto error;
L
Linus Torvalds 已提交
2521
		}
2522
	} else if (unlikely(S_ISDIR(dicp->di_mode))) {
L
Linus Torvalds 已提交
2523 2524 2525
		if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (dicp->di_format != XFS_DINODE_FMT_BTREE) &&
		    (dicp->di_format != XFS_DINODE_FMT_LOCAL)) {
2526
			XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(4)",
L
Linus Torvalds 已提交
2527 2528
					     XFS_ERRLEVEL_LOW, mp, dicp);
			xfs_buf_relse(bp);
2529 2530 2531 2532
			xfs_alert(mp,
		"%s: Bad dir inode log record, rec ptr 0x%p, "
		"ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
				__func__, item, dip, bp, in_f->ilf_ino);
2533 2534
			error = EFSCORRUPTED;
			goto error;
L
Linus Torvalds 已提交
2535 2536 2537
		}
	}
	if (unlikely(dicp->di_nextents + dicp->di_anextents > dicp->di_nblocks)){
2538
		XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(5)",
L
Linus Torvalds 已提交
2539 2540
				     XFS_ERRLEVEL_LOW, mp, dicp);
		xfs_buf_relse(bp);
2541 2542 2543 2544
		xfs_alert(mp,
	"%s: Bad inode log record, rec ptr 0x%p, dino ptr 0x%p, "
	"dino bp 0x%p, ino %Ld, total extents = %d, nblocks = %Ld",
			__func__, item, dip, bp, in_f->ilf_ino,
L
Linus Torvalds 已提交
2545 2546
			dicp->di_nextents + dicp->di_anextents,
			dicp->di_nblocks);
2547 2548
		error = EFSCORRUPTED;
		goto error;
L
Linus Torvalds 已提交
2549 2550
	}
	if (unlikely(dicp->di_forkoff > mp->m_sb.sb_inodesize)) {
2551
		XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(6)",
L
Linus Torvalds 已提交
2552 2553
				     XFS_ERRLEVEL_LOW, mp, dicp);
		xfs_buf_relse(bp);
2554 2555 2556
		xfs_alert(mp,
	"%s: Bad inode log record, rec ptr 0x%p, dino ptr 0x%p, "
	"dino bp 0x%p, ino %Ld, forkoff 0x%x", __func__,
2557
			item, dip, bp, in_f->ilf_ino, dicp->di_forkoff);
2558 2559
		error = EFSCORRUPTED;
		goto error;
L
Linus Torvalds 已提交
2560
	}
2561 2562
	isize = xfs_icdinode_size(dicp->di_version);
	if (unlikely(item->ri_buf[1].i_len > isize)) {
2563
		XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(7)",
L
Linus Torvalds 已提交
2564 2565
				     XFS_ERRLEVEL_LOW, mp, dicp);
		xfs_buf_relse(bp);
2566 2567 2568
		xfs_alert(mp,
			"%s: Bad inode log record length %d, rec ptr 0x%p",
			__func__, item->ri_buf[1].i_len, item);
2569 2570
		error = EFSCORRUPTED;
		goto error;
L
Linus Torvalds 已提交
2571 2572 2573
	}

	/* The core is in in-core format */
2574
	xfs_dinode_to_disk(dip, dicp);
L
Linus Torvalds 已提交
2575 2576

	/* the rest is in on-disk format */
2577 2578 2579 2580
	if (item->ri_buf[1].i_len > isize) {
		memcpy((char *)dip + isize,
			item->ri_buf[1].i_addr + isize,
			item->ri_buf[1].i_len - isize);
L
Linus Torvalds 已提交
2581 2582 2583 2584 2585
	}

	fields = in_f->ilf_fields;
	switch (fields & (XFS_ILOG_DEV | XFS_ILOG_UUID)) {
	case XFS_ILOG_DEV:
C
Christoph Hellwig 已提交
2586
		xfs_dinode_put_rdev(dip, in_f->ilf_u.ilfu_rdev);
L
Linus Torvalds 已提交
2587 2588
		break;
	case XFS_ILOG_UUID:
C
Christoph Hellwig 已提交
2589 2590 2591
		memcpy(XFS_DFORK_DPTR(dip),
		       &in_f->ilf_u.ilfu_uuid,
		       sizeof(uuid_t));
L
Linus Torvalds 已提交
2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
		break;
	}

	if (in_f->ilf_size == 2)
		goto write_inode_buffer;
	len = item->ri_buf[2].i_len;
	src = item->ri_buf[2].i_addr;
	ASSERT(in_f->ilf_size <= 4);
	ASSERT((in_f->ilf_size == 3) || (fields & XFS_ILOG_AFORK));
	ASSERT(!(fields & XFS_ILOG_DFORK) ||
	       (len == in_f->ilf_dsize));

	switch (fields & XFS_ILOG_DFORK) {
	case XFS_ILOG_DDATA:
	case XFS_ILOG_DEXT:
C
Christoph Hellwig 已提交
2607
		memcpy(XFS_DFORK_DPTR(dip), src, len);
L
Linus Torvalds 已提交
2608 2609 2610
		break;

	case XFS_ILOG_DBROOT:
2611
		xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src, len,
C
Christoph Hellwig 已提交
2612
				 (xfs_bmdr_block_t *)XFS_DFORK_DPTR(dip),
L
Linus Torvalds 已提交
2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
				 XFS_DFORK_DSIZE(dip, mp));
		break;

	default:
		/*
		 * There are no data fork flags set.
		 */
		ASSERT((fields & XFS_ILOG_DFORK) == 0);
		break;
	}

	/*
	 * If we logged any attribute data, recover it.  There may or
	 * may not have been any other non-core data logged in this
	 * transaction.
	 */
	if (in_f->ilf_fields & XFS_ILOG_AFORK) {
		if (in_f->ilf_fields & XFS_ILOG_DFORK) {
			attr_index = 3;
		} else {
			attr_index = 2;
		}
		len = item->ri_buf[attr_index].i_len;
		src = item->ri_buf[attr_index].i_addr;
		ASSERT(len == in_f->ilf_asize);

		switch (in_f->ilf_fields & XFS_ILOG_AFORK) {
		case XFS_ILOG_ADATA:
		case XFS_ILOG_AEXT:
			dest = XFS_DFORK_APTR(dip);
			ASSERT(len <= XFS_DFORK_ASIZE(dip, mp));
			memcpy(dest, src, len);
			break;

		case XFS_ILOG_ABROOT:
			dest = XFS_DFORK_APTR(dip);
2649 2650
			xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src,
					 len, (xfs_bmdr_block_t*)dest,
L
Linus Torvalds 已提交
2651 2652 2653 2654
					 XFS_DFORK_ASIZE(dip, mp));
			break;

		default:
2655
			xfs_warn(log->l_mp, "%s: Invalid flag", __func__);
L
Linus Torvalds 已提交
2656 2657
			ASSERT(0);
			xfs_buf_relse(bp);
2658 2659
			error = EIO;
			goto error;
L
Linus Torvalds 已提交
2660 2661 2662 2663
		}
	}

write_inode_buffer:
2664 2665 2666
	/* re-generate the checksum. */
	xfs_dinode_calc_crc(log->l_mp, dip);

2667
	ASSERT(bp->b_target->bt_mount == mp);
2668
	bp->b_iodone = xlog_recover_iodone;
2669
	xfs_buf_delwri_queue(bp, buffer_list);
2670
	xfs_buf_relse(bp);
2671 2672
error:
	if (need_free)
2673
		kmem_free(in_f);
2674
	return XFS_ERROR(error);
L
Linus Torvalds 已提交
2675 2676 2677
}

/*
M
Mark Tinguely 已提交
2678
 * Recover QUOTAOFF records. We simply make a note of it in the xlog
L
Linus Torvalds 已提交
2679 2680 2681 2682
 * structure, so that we know not to do any dquot item or dquot buffer recovery,
 * of that type.
 */
STATIC int
2683
xlog_recover_quotaoff_pass1(
M
Mark Tinguely 已提交
2684 2685
	struct xlog			*log,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
2686
{
2687
	xfs_qoff_logformat_t	*qoff_f = item->ri_buf[0].i_addr;
L
Linus Torvalds 已提交
2688 2689 2690 2691
	ASSERT(qoff_f);

	/*
	 * The logitem format's flag tells us if this was user quotaoff,
2692
	 * group/project quotaoff or both.
L
Linus Torvalds 已提交
2693 2694 2695
	 */
	if (qoff_f->qf_flags & XFS_UQUOTA_ACCT)
		log->l_quotaoffs_flag |= XFS_DQ_USER;
2696 2697
	if (qoff_f->qf_flags & XFS_PQUOTA_ACCT)
		log->l_quotaoffs_flag |= XFS_DQ_PROJ;
L
Linus Torvalds 已提交
2698 2699 2700 2701 2702 2703 2704 2705 2706 2707
	if (qoff_f->qf_flags & XFS_GQUOTA_ACCT)
		log->l_quotaoffs_flag |= XFS_DQ_GROUP;

	return (0);
}

/*
 * Recover a dquot record
 */
STATIC int
2708
xlog_recover_dquot_pass2(
M
Mark Tinguely 已提交
2709 2710 2711
	struct xlog			*log,
	struct list_head		*buffer_list,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
2712
{
2713
	xfs_mount_t		*mp = log->l_mp;
L
Linus Torvalds 已提交
2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726
	xfs_buf_t		*bp;
	struct xfs_disk_dquot	*ddq, *recddq;
	int			error;
	xfs_dq_logformat_t	*dq_f;
	uint			type;


	/*
	 * Filesystems are required to send in quota flags at mount time.
	 */
	if (mp->m_qflags == 0)
		return (0);

2727 2728
	recddq = item->ri_buf[1].i_addr;
	if (recddq == NULL) {
2729
		xfs_alert(log->l_mp, "NULL dquot in %s.", __func__);
2730 2731
		return XFS_ERROR(EIO);
	}
2732
	if (item->ri_buf[1].i_len < sizeof(xfs_disk_dquot_t)) {
2733
		xfs_alert(log->l_mp, "dquot too small (%d) in %s.",
2734 2735 2736 2737
			item->ri_buf[1].i_len, __func__);
		return XFS_ERROR(EIO);
	}

L
Linus Torvalds 已提交
2738 2739 2740
	/*
	 * This type of quotas was turned off, so ignore this record.
	 */
2741
	type = recddq->d_flags & (XFS_DQ_USER | XFS_DQ_PROJ | XFS_DQ_GROUP);
L
Linus Torvalds 已提交
2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755
	ASSERT(type);
	if (log->l_quotaoffs_flag & type)
		return (0);

	/*
	 * At this point we know that quota was _not_ turned off.
	 * Since the mount flags are not indicating to us otherwise, this
	 * must mean that quota is on, and the dquot needs to be replayed.
	 * Remember that we may not have fully recovered the superblock yet,
	 * so we can't do the usual trick of looking at the SB quota bits.
	 *
	 * The other possibility, of course, is that the quota subsystem was
	 * removed since the last mount - ENOSYS.
	 */
2756
	dq_f = item->ri_buf[0].i_addr;
L
Linus Torvalds 已提交
2757
	ASSERT(dq_f);
2758 2759 2760
	error = xfs_qm_dqcheck(mp, recddq, dq_f->qlf_id, 0, XFS_QMOPT_DOWARN,
			   "xlog_recover_dquot_pass2 (log copy)");
	if (error)
L
Linus Torvalds 已提交
2761 2762 2763
		return XFS_ERROR(EIO);
	ASSERT(dq_f->qlf_len == 1);

D
Dave Chinner 已提交
2764
	error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dq_f->qlf_blkno,
2765 2766
				   XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp,
				   NULL);
D
Dave Chinner 已提交
2767
	if (error)
L
Linus Torvalds 已提交
2768
		return error;
D
Dave Chinner 已提交
2769

L
Linus Torvalds 已提交
2770 2771 2772 2773 2774 2775 2776 2777
	ASSERT(bp);
	ddq = (xfs_disk_dquot_t *)xfs_buf_offset(bp, dq_f->qlf_boffset);

	/*
	 * At least the magic num portion should be on disk because this
	 * was among a chunk of dquots created earlier, and we did some
	 * minimal initialization then.
	 */
2778 2779 2780
	error = xfs_qm_dqcheck(mp, ddq, dq_f->qlf_id, 0, XFS_QMOPT_DOWARN,
			   "xlog_recover_dquot_pass2");
	if (error) {
L
Linus Torvalds 已提交
2781 2782 2783 2784 2785 2786 2787
		xfs_buf_relse(bp);
		return XFS_ERROR(EIO);
	}

	memcpy(ddq, recddq, item->ri_buf[1].i_len);

	ASSERT(dq_f->qlf_size == 2);
2788
	ASSERT(bp->b_target->bt_mount == mp);
2789
	bp->b_iodone = xlog_recover_iodone;
2790
	xfs_buf_delwri_queue(bp, buffer_list);
2791
	xfs_buf_relse(bp);
L
Linus Torvalds 已提交
2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802

	return (0);
}

/*
 * This routine is called to create an in-core extent free intent
 * item from the efi format structure which was logged on disk.
 * It allocates an in-core efi, copies the extents from the format
 * structure into it, and adds the efi to the AIL with the given
 * LSN.
 */
2803
STATIC int
2804
xlog_recover_efi_pass2(
M
Mark Tinguely 已提交
2805 2806 2807
	struct xlog			*log,
	struct xlog_recover_item	*item,
	xfs_lsn_t			lsn)
L
Linus Torvalds 已提交
2808
{
2809
	int			error;
2810
	xfs_mount_t		*mp = log->l_mp;
L
Linus Torvalds 已提交
2811 2812 2813
	xfs_efi_log_item_t	*efip;
	xfs_efi_log_format_t	*efi_formatp;

2814
	efi_formatp = item->ri_buf[0].i_addr;
L
Linus Torvalds 已提交
2815 2816

	efip = xfs_efi_init(mp, efi_formatp->efi_nextents);
2817 2818 2819 2820 2821
	if ((error = xfs_efi_copy_format(&(item->ri_buf[0]),
					 &(efip->efi_format)))) {
		xfs_efi_item_free(efip);
		return error;
	}
2822
	atomic_set(&efip->efi_next_extent, efi_formatp->efi_nextents);
L
Linus Torvalds 已提交
2823

2824
	spin_lock(&log->l_ailp->xa_lock);
L
Linus Torvalds 已提交
2825
	/*
2826
	 * xfs_trans_ail_update() drops the AIL lock.
L
Linus Torvalds 已提交
2827
	 */
2828
	xfs_trans_ail_update(log->l_ailp, &efip->efi_item, lsn);
2829
	return 0;
L
Linus Torvalds 已提交
2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840
}


/*
 * This routine is called when an efd format structure is found in
 * a committed transaction in the log.  It's purpose is to cancel
 * the corresponding efi if it was still in the log.  To do this
 * it searches the AIL for the efi with an id equal to that in the
 * efd format structure.  If we find it, we remove the efi from the
 * AIL and free it.
 */
2841 2842
STATIC int
xlog_recover_efd_pass2(
M
Mark Tinguely 已提交
2843 2844
	struct xlog			*log,
	struct xlog_recover_item	*item)
L
Linus Torvalds 已提交
2845 2846 2847 2848 2849
{
	xfs_efd_log_format_t	*efd_formatp;
	xfs_efi_log_item_t	*efip = NULL;
	xfs_log_item_t		*lip;
	__uint64_t		efi_id;
2850
	struct xfs_ail_cursor	cur;
2851
	struct xfs_ail		*ailp = log->l_ailp;
L
Linus Torvalds 已提交
2852

2853
	efd_formatp = item->ri_buf[0].i_addr;
2854 2855 2856 2857
	ASSERT((item->ri_buf[0].i_len == (sizeof(xfs_efd_log_format_32_t) +
		((efd_formatp->efd_nextents - 1) * sizeof(xfs_extent_32_t)))) ||
	       (item->ri_buf[0].i_len == (sizeof(xfs_efd_log_format_64_t) +
		((efd_formatp->efd_nextents - 1) * sizeof(xfs_extent_64_t)))));
L
Linus Torvalds 已提交
2858 2859 2860 2861 2862 2863
	efi_id = efd_formatp->efd_efi_id;

	/*
	 * Search for the efi with the id in the efd format structure
	 * in the AIL.
	 */
2864 2865
	spin_lock(&ailp->xa_lock);
	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
L
Linus Torvalds 已提交
2866 2867 2868 2869 2870
	while (lip != NULL) {
		if (lip->li_type == XFS_LI_EFI) {
			efip = (xfs_efi_log_item_t *)lip;
			if (efip->efi_format.efi_id == efi_id) {
				/*
2871
				 * xfs_trans_ail_delete() drops the
L
Linus Torvalds 已提交
2872 2873
				 * AIL lock.
				 */
2874 2875
				xfs_trans_ail_delete(ailp, lip,
						     SHUTDOWN_CORRUPT_INCORE);
2876
				xfs_efi_item_free(efip);
2877
				spin_lock(&ailp->xa_lock);
2878
				break;
L
Linus Torvalds 已提交
2879 2880
			}
		}
2881
		lip = xfs_trans_ail_cursor_next(ailp, &cur);
L
Linus Torvalds 已提交
2882
	}
2883 2884
	xfs_trans_ail_cursor_done(ailp, &cur);
	spin_unlock(&ailp->xa_lock);
2885 2886

	return 0;
L
Linus Torvalds 已提交
2887 2888 2889 2890 2891 2892 2893 2894 2895
}

/*
 * Free up any resources allocated by the transaction
 *
 * Remember that EFIs, EFDs, and IUNLINKs are handled later.
 */
STATIC void
xlog_recover_free_trans(
2896
	struct xlog_recover	*trans)
L
Linus Torvalds 已提交
2897
{
2898
	xlog_recover_item_t	*item, *n;
L
Linus Torvalds 已提交
2899 2900
	int			i;

2901 2902 2903 2904 2905
	list_for_each_entry_safe(item, n, &trans->r_itemq, ri_list) {
		/* Free the regions in the item. */
		list_del(&item->ri_list);
		for (i = 0; i < item->ri_cnt; i++)
			kmem_free(item->ri_buf[i].i_addr);
L
Linus Torvalds 已提交
2906
		/* Free the item itself */
2907 2908 2909
		kmem_free(item->ri_buf);
		kmem_free(item);
	}
L
Linus Torvalds 已提交
2910
	/* Free the transaction recover structure */
2911
	kmem_free(trans);
L
Linus Torvalds 已提交
2912 2913
}

2914
STATIC int
2915
xlog_recover_commit_pass1(
2916 2917 2918
	struct xlog			*log,
	struct xlog_recover		*trans,
	struct xlog_recover_item	*item)
2919
{
2920
	trace_xfs_log_recover_item_recover(log, trans, item, XLOG_RECOVER_PASS1);
2921 2922 2923

	switch (ITEM_TYPE(item)) {
	case XFS_LI_BUF:
2924 2925 2926
		return xlog_recover_buffer_pass1(log, item);
	case XFS_LI_QUOTAOFF:
		return xlog_recover_quotaoff_pass1(log, item);
2927 2928 2929
	case XFS_LI_INODE:
	case XFS_LI_EFI:
	case XFS_LI_EFD:
2930 2931
	case XFS_LI_DQUOT:
		/* nothing to do in pass 1 */
2932
		return 0;
2933
	default:
2934 2935
		xfs_warn(log->l_mp, "%s: invalid item type (%d)",
			__func__, ITEM_TYPE(item));
2936 2937 2938 2939 2940 2941 2942
		ASSERT(0);
		return XFS_ERROR(EIO);
	}
}

STATIC int
xlog_recover_commit_pass2(
2943 2944 2945 2946
	struct xlog			*log,
	struct xlog_recover		*trans,
	struct list_head		*buffer_list,
	struct xlog_recover_item	*item)
2947 2948 2949 2950 2951
{
	trace_xfs_log_recover_item_recover(log, trans, item, XLOG_RECOVER_PASS2);

	switch (ITEM_TYPE(item)) {
	case XFS_LI_BUF:
2952
		return xlog_recover_buffer_pass2(log, buffer_list, item);
2953
	case XFS_LI_INODE:
2954
		return xlog_recover_inode_pass2(log, buffer_list, item);
2955 2956 2957 2958
	case XFS_LI_EFI:
		return xlog_recover_efi_pass2(log, item, trans->r_lsn);
	case XFS_LI_EFD:
		return xlog_recover_efd_pass2(log, item);
2959
	case XFS_LI_DQUOT:
2960
		return xlog_recover_dquot_pass2(log, buffer_list, item);
2961
	case XFS_LI_QUOTAOFF:
2962 2963
		/* nothing to do in pass2 */
		return 0;
2964
	default:
2965 2966
		xfs_warn(log->l_mp, "%s: invalid item type (%d)",
			__func__, ITEM_TYPE(item));
2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977
		ASSERT(0);
		return XFS_ERROR(EIO);
	}
}

/*
 * Perform the transaction.
 *
 * If the transaction modifies a buffer or inode, do it now.  Otherwise,
 * EFIs and EFDs get queued up by adding entries into the AIL for them.
 */
L
Linus Torvalds 已提交
2978 2979
STATIC int
xlog_recover_commit_trans(
2980
	struct xlog		*log,
2981
	struct xlog_recover	*trans,
L
Linus Torvalds 已提交
2982 2983
	int			pass)
{
2984
	int			error = 0, error2;
2985
	xlog_recover_item_t	*item;
2986
	LIST_HEAD		(buffer_list);
L
Linus Torvalds 已提交
2987

2988
	hlist_del(&trans->r_list);
2989 2990 2991

	error = xlog_recover_reorder_trans(log, trans, pass);
	if (error)
L
Linus Torvalds 已提交
2992
		return error;
2993 2994

	list_for_each_entry(item, &trans->r_itemq, ri_list) {
2995 2996
		switch (pass) {
		case XLOG_RECOVER_PASS1:
2997
			error = xlog_recover_commit_pass1(log, trans, item);
2998 2999 3000 3001 3002 3003 3004 3005 3006
			break;
		case XLOG_RECOVER_PASS2:
			error = xlog_recover_commit_pass2(log, trans,
							  &buffer_list, item);
			break;
		default:
			ASSERT(0);
		}

3007
		if (error)
3008
			goto out;
3009 3010 3011
	}

	xlog_recover_free_trans(trans);
3012 3013 3014 3015

out:
	error2 = xfs_buf_delwri_submit(&buffer_list);
	return error ? error : error2;
L
Linus Torvalds 已提交
3016 3017 3018 3019
}

STATIC int
xlog_recover_unmount_trans(
3020 3021
	struct xlog		*log,
	struct xlog_recover	*trans)
L
Linus Torvalds 已提交
3022 3023
{
	/* Do nothing now */
3024
	xfs_warn(log->l_mp, "%s: Unmount LR", __func__);
L
Linus Torvalds 已提交
3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038
	return 0;
}

/*
 * There are two valid states of the r_state field.  0 indicates that the
 * transaction structure is in a normal state.  We have either seen the
 * start of the transaction or the last operation we added was not a partial
 * operation.  If the last operation we added to the transaction was a
 * partial operation, we need to mark r_state with XLOG_WAS_CONT_TRANS.
 *
 * NOTE: skip LRs with 0 data length.
 */
STATIC int
xlog_recover_process_data(
M
Mark Tinguely 已提交
3039
	struct xlog		*log,
3040
	struct hlist_head	rhash[],
M
Mark Tinguely 已提交
3041
	struct xlog_rec_header	*rhead,
L
Linus Torvalds 已提交
3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
	xfs_caddr_t		dp,
	int			pass)
{
	xfs_caddr_t		lp;
	int			num_logops;
	xlog_op_header_t	*ohead;
	xlog_recover_t		*trans;
	xlog_tid_t		tid;
	int			error;
	unsigned long		hash;
	uint			flags;

3054 3055
	lp = dp + be32_to_cpu(rhead->h_len);
	num_logops = be32_to_cpu(rhead->h_num_logops);
L
Linus Torvalds 已提交
3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066

	/* check the log format matches our own - else we can't recover */
	if (xlog_header_check_recover(log->l_mp, rhead))
		return (XFS_ERROR(EIO));

	while ((dp < lp) && num_logops) {
		ASSERT(dp + sizeof(xlog_op_header_t) <= lp);
		ohead = (xlog_op_header_t *)dp;
		dp += sizeof(xlog_op_header_t);
		if (ohead->oh_clientid != XFS_TRANSACTION &&
		    ohead->oh_clientid != XFS_LOG) {
3067 3068
			xfs_warn(log->l_mp, "%s: bad clientid 0x%x",
					__func__, ohead->oh_clientid);
L
Linus Torvalds 已提交
3069 3070 3071
			ASSERT(0);
			return (XFS_ERROR(EIO));
		}
3072
		tid = be32_to_cpu(ohead->oh_tid);
L
Linus Torvalds 已提交
3073
		hash = XLOG_RHASH(tid);
3074
		trans = xlog_recover_find_tid(&rhash[hash], tid);
L
Linus Torvalds 已提交
3075 3076 3077
		if (trans == NULL) {		   /* not found; add new tid */
			if (ohead->oh_flags & XLOG_START_TRANS)
				xlog_recover_new_tid(&rhash[hash], tid,
3078
					be64_to_cpu(rhead->h_lsn));
L
Linus Torvalds 已提交
3079
		} else {
3080
			if (dp + be32_to_cpu(ohead->oh_len) > lp) {
3081 3082
				xfs_warn(log->l_mp, "%s: bad length 0x%x",
					__func__, be32_to_cpu(ohead->oh_len));
3083 3084 3085
				WARN_ON(1);
				return (XFS_ERROR(EIO));
			}
L
Linus Torvalds 已提交
3086 3087 3088 3089 3090 3091
			flags = ohead->oh_flags & ~XLOG_END_TRANS;
			if (flags & XLOG_WAS_CONT_TRANS)
				flags &= ~XLOG_CONTINUE_TRANS;
			switch (flags) {
			case XLOG_COMMIT_TRANS:
				error = xlog_recover_commit_trans(log,
3092
								trans, pass);
L
Linus Torvalds 已提交
3093 3094
				break;
			case XLOG_UNMOUNT_TRANS:
3095
				error = xlog_recover_unmount_trans(log, trans);
L
Linus Torvalds 已提交
3096 3097
				break;
			case XLOG_WAS_CONT_TRANS:
3098 3099 3100
				error = xlog_recover_add_to_cont_trans(log,
						trans, dp,
						be32_to_cpu(ohead->oh_len));
L
Linus Torvalds 已提交
3101 3102
				break;
			case XLOG_START_TRANS:
3103 3104
				xfs_warn(log->l_mp, "%s: bad transaction",
					__func__);
L
Linus Torvalds 已提交
3105 3106 3107 3108 3109
				ASSERT(0);
				error = XFS_ERROR(EIO);
				break;
			case 0:
			case XLOG_CONTINUE_TRANS:
3110
				error = xlog_recover_add_to_trans(log, trans,
3111
						dp, be32_to_cpu(ohead->oh_len));
L
Linus Torvalds 已提交
3112 3113
				break;
			default:
3114 3115
				xfs_warn(log->l_mp, "%s: bad flag 0x%x",
					__func__, flags);
L
Linus Torvalds 已提交
3116 3117 3118 3119 3120 3121 3122
				ASSERT(0);
				error = XFS_ERROR(EIO);
				break;
			}
			if (error)
				return error;
		}
3123
		dp += be32_to_cpu(ohead->oh_len);
L
Linus Torvalds 已提交
3124 3125 3126 3127 3128 3129 3130 3131 3132
		num_logops--;
	}
	return 0;
}

/*
 * Process an extent free intent item that was recovered from
 * the log.  We need to free the extents that it describes.
 */
3133
STATIC int
L
Linus Torvalds 已提交
3134 3135 3136 3137 3138 3139 3140
xlog_recover_process_efi(
	xfs_mount_t		*mp,
	xfs_efi_log_item_t	*efip)
{
	xfs_efd_log_item_t	*efdp;
	xfs_trans_t		*tp;
	int			i;
3141
	int			error = 0;
L
Linus Torvalds 已提交
3142 3143 3144
	xfs_extent_t		*extp;
	xfs_fsblock_t		startblock_fsb;

3145
	ASSERT(!test_bit(XFS_EFI_RECOVERED, &efip->efi_flags));
L
Linus Torvalds 已提交
3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163

	/*
	 * First check the validity of the extents described by the
	 * EFI.  If any are bad, then assume that all are bad and
	 * just toss the EFI.
	 */
	for (i = 0; i < efip->efi_format.efi_nextents; i++) {
		extp = &(efip->efi_format.efi_extents[i]);
		startblock_fsb = XFS_BB_TO_FSB(mp,
				   XFS_FSB_TO_DADDR(mp, extp->ext_start));
		if ((startblock_fsb == 0) ||
		    (extp->ext_len == 0) ||
		    (startblock_fsb >= mp->m_sb.sb_dblocks) ||
		    (extp->ext_len >= mp->m_sb.sb_agblocks)) {
			/*
			 * This will pull the EFI from the AIL and
			 * free the memory associated with it.
			 */
3164
			set_bit(XFS_EFI_RECOVERED, &efip->efi_flags);
L
Linus Torvalds 已提交
3165
			xfs_efi_release(efip, efip->efi_format.efi_nextents);
3166
			return XFS_ERROR(EIO);
L
Linus Torvalds 已提交
3167 3168 3169 3170
		}
	}

	tp = xfs_trans_alloc(mp, 0);
3171
	error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0, 0, 0);
3172 3173
	if (error)
		goto abort_error;
L
Linus Torvalds 已提交
3174 3175 3176 3177
	efdp = xfs_trans_get_efd(tp, efip, efip->efi_format.efi_nextents);

	for (i = 0; i < efip->efi_format.efi_nextents; i++) {
		extp = &(efip->efi_format.efi_extents[i]);
3178 3179 3180
		error = xfs_free_extent(tp, extp->ext_start, extp->ext_len);
		if (error)
			goto abort_error;
L
Linus Torvalds 已提交
3181 3182 3183 3184
		xfs_trans_log_efd_extent(tp, efdp, extp->ext_start,
					 extp->ext_len);
	}

3185
	set_bit(XFS_EFI_RECOVERED, &efip->efi_flags);
3186
	error = xfs_trans_commit(tp, 0);
3187
	return error;
3188 3189 3190 3191

abort_error:
	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
	return error;
L
Linus Torvalds 已提交
3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
}

/*
 * When this is called, all of the EFIs which did not have
 * corresponding EFDs should be in the AIL.  What we do now
 * is free the extents associated with each one.
 *
 * Since we process the EFIs in normal transactions, they
 * will be removed at some point after the commit.  This prevents
 * us from just walking down the list processing each one.
 * We'll use a flag in the EFI to skip those that we've already
 * processed and use the AIL iteration mechanism's generation
 * count to try to speed this up at least a bit.
 *
 * When we start, we know that the EFIs are the only things in
 * the AIL.  As we process them, however, other items are added
 * to the AIL.  Since everything added to the AIL must come after
 * everything already in the AIL, we stop processing as soon as
 * we see something other than an EFI in the AIL.
 */
3212
STATIC int
L
Linus Torvalds 已提交
3213
xlog_recover_process_efis(
M
Mark Tinguely 已提交
3214
	struct xlog	*log)
L
Linus Torvalds 已提交
3215 3216 3217
{
	xfs_log_item_t		*lip;
	xfs_efi_log_item_t	*efip;
3218
	int			error = 0;
3219
	struct xfs_ail_cursor	cur;
3220
	struct xfs_ail		*ailp;
L
Linus Torvalds 已提交
3221

3222 3223 3224
	ailp = log->l_ailp;
	spin_lock(&ailp->xa_lock);
	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
L
Linus Torvalds 已提交
3225 3226 3227
	while (lip != NULL) {
		/*
		 * We're done when we see something other than an EFI.
3228
		 * There should be no EFIs left in the AIL now.
L
Linus Torvalds 已提交
3229 3230
		 */
		if (lip->li_type != XFS_LI_EFI) {
3231
#ifdef DEBUG
3232
			for (; lip; lip = xfs_trans_ail_cursor_next(ailp, &cur))
3233 3234
				ASSERT(lip->li_type != XFS_LI_EFI);
#endif
L
Linus Torvalds 已提交
3235 3236 3237 3238 3239 3240 3241
			break;
		}

		/*
		 * Skip EFIs that we've already processed.
		 */
		efip = (xfs_efi_log_item_t *)lip;
3242
		if (test_bit(XFS_EFI_RECOVERED, &efip->efi_flags)) {
3243
			lip = xfs_trans_ail_cursor_next(ailp, &cur);
L
Linus Torvalds 已提交
3244 3245 3246
			continue;
		}

3247 3248 3249
		spin_unlock(&ailp->xa_lock);
		error = xlog_recover_process_efi(log->l_mp, efip);
		spin_lock(&ailp->xa_lock);
3250 3251
		if (error)
			goto out;
3252
		lip = xfs_trans_ail_cursor_next(ailp, &cur);
L
Linus Torvalds 已提交
3253
	}
3254
out:
3255 3256
	xfs_trans_ail_cursor_done(ailp, &cur);
	spin_unlock(&ailp->xa_lock);
3257
	return error;
L
Linus Torvalds 已提交
3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276
}

/*
 * This routine performs a transaction to null out a bad inode pointer
 * in an agi unlinked inode hash bucket.
 */
STATIC void
xlog_recover_clear_agi_bucket(
	xfs_mount_t	*mp,
	xfs_agnumber_t	agno,
	int		bucket)
{
	xfs_trans_t	*tp;
	xfs_agi_t	*agi;
	xfs_buf_t	*agibp;
	int		offset;
	int		error;

	tp = xfs_trans_alloc(mp, XFS_TRANS_CLEAR_AGI_BUCKET);
3277 3278
	error = xfs_trans_reserve(tp, 0, XFS_CLEAR_AGI_BUCKET_LOG_RES(mp),
				  0, 0, 0);
3279 3280
	if (error)
		goto out_abort;
L
Linus Torvalds 已提交
3281

3282 3283
	error = xfs_read_agi(mp, tp, agno, &agibp);
	if (error)
3284
		goto out_abort;
L
Linus Torvalds 已提交
3285

3286
	agi = XFS_BUF_TO_AGI(agibp);
3287
	agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
L
Linus Torvalds 已提交
3288 3289 3290 3291 3292
	offset = offsetof(xfs_agi_t, agi_unlinked) +
		 (sizeof(xfs_agino_t) * bucket);
	xfs_trans_log_buf(tp, agibp, offset,
			  (offset + sizeof(xfs_agino_t) - 1));

3293 3294 3295 3296 3297 3298 3299 3300
	error = xfs_trans_commit(tp, 0);
	if (error)
		goto out_error;
	return;

out_abort:
	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
out_error:
3301
	xfs_warn(mp, "%s: failed to clear agi %d. Continuing.", __func__, agno);
3302
	return;
L
Linus Torvalds 已提交
3303 3304
}

3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318
STATIC xfs_agino_t
xlog_recover_process_one_iunlink(
	struct xfs_mount		*mp,
	xfs_agnumber_t			agno,
	xfs_agino_t			agino,
	int				bucket)
{
	struct xfs_buf			*ibp;
	struct xfs_dinode		*dip;
	struct xfs_inode		*ip;
	xfs_ino_t			ino;
	int				error;

	ino = XFS_AGINO_TO_INO(mp, agno, agino);
3319
	error = xfs_iget(mp, NULL, ino, 0, 0, &ip);
3320 3321 3322 3323 3324 3325
	if (error)
		goto fail;

	/*
	 * Get the on disk inode to find the next inode in the bucket.
	 */
3326
	error = xfs_imap_to_bp(mp, NULL, &ip->i_imap, &dip, &ibp, 0, 0);
3327
	if (error)
3328
		goto fail_iput;
3329 3330

	ASSERT(ip->i_d.di_nlink == 0);
3331
	ASSERT(ip->i_d.di_mode != 0);
3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342

	/* setup for the next pass */
	agino = be32_to_cpu(dip->di_next_unlinked);
	xfs_buf_relse(ibp);

	/*
	 * Prevent any DMAPI event from being sent when the reference on
	 * the inode is dropped.
	 */
	ip->i_d.di_dmevmask = 0;

3343
	IRELE(ip);
3344 3345
	return agino;

3346 3347
 fail_iput:
	IRELE(ip);
3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360
 fail:
	/*
	 * We can't read in the inode this bucket points to, or this inode
	 * is messed up.  Just ditch this bucket of inodes.  We will lose
	 * some inodes and space, but at least we won't hang.
	 *
	 * Call xlog_recover_clear_agi_bucket() to perform a transaction to
	 * clear the inode pointer in the bucket.
	 */
	xlog_recover_clear_agi_bucket(mp, agno, bucket);
	return NULLAGINO;
}

L
Linus Torvalds 已提交
3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372
/*
 * xlog_iunlink_recover
 *
 * This is called during recovery to process any inodes which
 * we unlinked but not freed when the system crashed.  These
 * inodes will be on the lists in the AGI blocks.  What we do
 * here is scan all the AGIs and fully truncate and free any
 * inodes found on the lists.  Each inode is removed from the
 * lists when it has been fully truncated and is freed.  The
 * freeing of the inode and its removal from the list must be
 * atomic.
 */
3373
STATIC void
L
Linus Torvalds 已提交
3374
xlog_recover_process_iunlinks(
M
Mark Tinguely 已提交
3375
	struct xlog	*log)
L
Linus Torvalds 已提交
3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397
{
	xfs_mount_t	*mp;
	xfs_agnumber_t	agno;
	xfs_agi_t	*agi;
	xfs_buf_t	*agibp;
	xfs_agino_t	agino;
	int		bucket;
	int		error;
	uint		mp_dmevmask;

	mp = log->l_mp;

	/*
	 * Prevent any DMAPI event from being sent while in this function.
	 */
	mp_dmevmask = mp->m_dmevmask;
	mp->m_dmevmask = 0;

	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
		/*
		 * Find the agi for this ag.
		 */
3398 3399 3400 3401 3402 3403 3404 3405 3406
		error = xfs_read_agi(mp, NULL, agno, &agibp);
		if (error) {
			/*
			 * AGI is b0rked. Don't process it.
			 *
			 * We should probably mark the filesystem as corrupt
			 * after we've recovered all the ag's we can....
			 */
			continue;
L
Linus Torvalds 已提交
3407
		}
3408 3409 3410 3411 3412 3413 3414 3415 3416
		/*
		 * Unlock the buffer so that it can be acquired in the normal
		 * course of the transaction to truncate and free each inode.
		 * Because we are not racing with anyone else here for the AGI
		 * buffer, we don't even need to hold it locked to read the
		 * initial unlinked bucket entries out of the buffer. We keep
		 * buffer reference though, so that it stays pinned in memory
		 * while we need the buffer.
		 */
L
Linus Torvalds 已提交
3417
		agi = XFS_BUF_TO_AGI(agibp);
3418
		xfs_buf_unlock(agibp);
L
Linus Torvalds 已提交
3419 3420

		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) {
3421
			agino = be32_to_cpu(agi->agi_unlinked[bucket]);
L
Linus Torvalds 已提交
3422
			while (agino != NULLAGINO) {
3423 3424
				agino = xlog_recover_process_one_iunlink(mp,
							agno, agino, bucket);
L
Linus Torvalds 已提交
3425 3426
			}
		}
3427
		xfs_buf_rele(agibp);
L
Linus Torvalds 已提交
3428 3429 3430 3431 3432 3433
	}

	mp->m_dmevmask = mp_dmevmask;
}

/*
C
Christoph Hellwig 已提交
3434 3435 3436 3437 3438 3439 3440 3441
 * Upack the log buffer data and crc check it. If the check fails, issue a
 * warning if and only if the CRC in the header is non-zero. This makes the
 * check an advisory warning, and the zero CRC check will prevent failure
 * warnings from being emitted when upgrading the kernel from one that does not
 * add CRCs by default.
 *
 * When filesystems are CRC enabled, this CRC mismatch becomes a fatal log
 * corruption failure
L
Linus Torvalds 已提交
3442
 */
C
Christoph Hellwig 已提交
3443 3444 3445 3446 3447
STATIC int
xlog_unpack_data_crc(
	struct xlog_rec_header	*rhead,
	xfs_caddr_t		dp,
	struct xlog		*log)
L
Linus Torvalds 已提交
3448
{
3449
	__le32			crc;
C
Christoph Hellwig 已提交
3450 3451 3452 3453 3454 3455

	crc = xlog_cksum(log, rhead, dp, be32_to_cpu(rhead->h_len));
	if (crc != rhead->h_crc) {
		if (rhead->h_crc || xfs_sb_version_hascrc(&log->l_mp->m_sb)) {
			xfs_alert(log->l_mp,
		"log record CRC mismatch: found 0x%x, expected 0x%x.\n",
3456 3457
					le32_to_cpu(rhead->h_crc),
					le32_to_cpu(crc));
C
Christoph Hellwig 已提交
3458
			xfs_hex_dump(dp, 32);
L
Linus Torvalds 已提交
3459 3460
		}

C
Christoph Hellwig 已提交
3461 3462 3463 3464 3465 3466 3467
		/*
		 * If we've detected a log record corruption, then we can't
		 * recover past this point. Abort recovery if we are enforcing
		 * CRC protection by punting an error back up the stack.
		 */
		if (xfs_sb_version_hascrc(&log->l_mp->m_sb))
			return EFSCORRUPTED;
L
Linus Torvalds 已提交
3468
	}
C
Christoph Hellwig 已提交
3469 3470

	return 0;
L
Linus Torvalds 已提交
3471 3472
}

C
Christoph Hellwig 已提交
3473
STATIC int
L
Linus Torvalds 已提交
3474
xlog_unpack_data(
M
Mark Tinguely 已提交
3475
	struct xlog_rec_header	*rhead,
L
Linus Torvalds 已提交
3476
	xfs_caddr_t		dp,
M
Mark Tinguely 已提交
3477
	struct xlog		*log)
L
Linus Torvalds 已提交
3478 3479
{
	int			i, j, k;
C
Christoph Hellwig 已提交
3480 3481 3482 3483 3484
	int			error;

	error = xlog_unpack_data_crc(rhead, dp, log);
	if (error)
		return error;
L
Linus Torvalds 已提交
3485

3486
	for (i = 0; i < BTOBB(be32_to_cpu(rhead->h_len)) &&
L
Linus Torvalds 已提交
3487
		  i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
3488
		*(__be32 *)dp = *(__be32 *)&rhead->h_cycle_data[i];
L
Linus Torvalds 已提交
3489 3490 3491
		dp += BBSIZE;
	}

3492
	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
3493
		xlog_in_core_2_t *xhdr = (xlog_in_core_2_t *)rhead;
3494
		for ( ; i < BTOBB(be32_to_cpu(rhead->h_len)); i++) {
L
Linus Torvalds 已提交
3495 3496
			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
3497
			*(__be32 *)dp = xhdr[j].hic_xheader.xh_cycle_data[k];
L
Linus Torvalds 已提交
3498 3499 3500
			dp += BBSIZE;
		}
	}
C
Christoph Hellwig 已提交
3501 3502

	return 0;
L
Linus Torvalds 已提交
3503 3504 3505 3506
}

STATIC int
xlog_valid_rec_header(
M
Mark Tinguely 已提交
3507 3508
	struct xlog		*log,
	struct xlog_rec_header	*rhead,
L
Linus Torvalds 已提交
3509 3510 3511 3512
	xfs_daddr_t		blkno)
{
	int			hlen;

3513
	if (unlikely(rhead->h_magicno != cpu_to_be32(XLOG_HEADER_MAGIC_NUM))) {
L
Linus Torvalds 已提交
3514 3515 3516 3517 3518 3519
		XFS_ERROR_REPORT("xlog_valid_rec_header(1)",
				XFS_ERRLEVEL_LOW, log->l_mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	if (unlikely(
	    (!rhead->h_version ||
3520
	    (be32_to_cpu(rhead->h_version) & (~XLOG_VERSION_OKBITS))))) {
3521
		xfs_warn(log->l_mp, "%s: unrecognised log version (%d).",
3522
			__func__, be32_to_cpu(rhead->h_version));
L
Linus Torvalds 已提交
3523 3524 3525 3526
		return XFS_ERROR(EIO);
	}

	/* LR body must have data or it wouldn't have been written */
3527
	hlen = be32_to_cpu(rhead->h_len);
L
Linus Torvalds 已提交
3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550
	if (unlikely( hlen <= 0 || hlen > INT_MAX )) {
		XFS_ERROR_REPORT("xlog_valid_rec_header(2)",
				XFS_ERRLEVEL_LOW, log->l_mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	if (unlikely( blkno > log->l_logBBsize || blkno > INT_MAX )) {
		XFS_ERROR_REPORT("xlog_valid_rec_header(3)",
				XFS_ERRLEVEL_LOW, log->l_mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	return 0;
}

/*
 * Read the log from tail to head and process the log records found.
 * Handle the two cases where the tail and head are in the same cycle
 * and where the active portion of the log wraps around the end of
 * the physical log separately.  The pass parameter is passed through
 * to the routines called to process the data and is not looked at
 * here.
 */
STATIC int
xlog_do_recovery_pass(
M
Mark Tinguely 已提交
3551
	struct xlog		*log,
L
Linus Torvalds 已提交
3552 3553 3554 3555 3556 3557
	xfs_daddr_t		head_blk,
	xfs_daddr_t		tail_blk,
	int			pass)
{
	xlog_rec_header_t	*rhead;
	xfs_daddr_t		blk_no;
3558
	xfs_caddr_t		offset;
L
Linus Torvalds 已提交
3559 3560 3561 3562
	xfs_buf_t		*hbp, *dbp;
	int			error = 0, h_size;
	int			bblks, split_bblks;
	int			hblks, split_hblks, wrapped_hblks;
3563
	struct hlist_head	rhash[XLOG_RHASH_SIZE];
L
Linus Torvalds 已提交
3564 3565 3566 3567 3568 3569 3570

	ASSERT(head_blk != tail_blk);

	/*
	 * Read the header of the tail block and get the iclog buffer size from
	 * h_size.  Use this to tell how many sectors make up the log header.
	 */
3571
	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
L
Linus Torvalds 已提交
3572 3573 3574 3575 3576 3577 3578 3579
		/*
		 * When using variable length iclogs, read first sector of
		 * iclog header and extract the header size from it.  Get a
		 * new hbp that is the correct size.
		 */
		hbp = xlog_get_bp(log, 1);
		if (!hbp)
			return ENOMEM;
C
Christoph Hellwig 已提交
3580 3581 3582

		error = xlog_bread(log, tail_blk, 1, hbp, &offset);
		if (error)
L
Linus Torvalds 已提交
3583
			goto bread_err1;
C
Christoph Hellwig 已提交
3584

L
Linus Torvalds 已提交
3585 3586 3587 3588
		rhead = (xlog_rec_header_t *)offset;
		error = xlog_valid_rec_header(log, rhead, tail_blk);
		if (error)
			goto bread_err1;
3589 3590
		h_size = be32_to_cpu(rhead->h_size);
		if ((be32_to_cpu(rhead->h_version) & XLOG_VERSION_2) &&
L
Linus Torvalds 已提交
3591 3592 3593 3594 3595 3596 3597 3598 3599 3600
		    (h_size > XLOG_HEADER_CYCLE_SIZE)) {
			hblks = h_size / XLOG_HEADER_CYCLE_SIZE;
			if (h_size % XLOG_HEADER_CYCLE_SIZE)
				hblks++;
			xlog_put_bp(hbp);
			hbp = xlog_get_bp(log, hblks);
		} else {
			hblks = 1;
		}
	} else {
3601
		ASSERT(log->l_sectBBsize == 1);
L
Linus Torvalds 已提交
3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617
		hblks = 1;
		hbp = xlog_get_bp(log, 1);
		h_size = XLOG_BIG_RECORD_BSIZE;
	}

	if (!hbp)
		return ENOMEM;
	dbp = xlog_get_bp(log, BTOBB(h_size));
	if (!dbp) {
		xlog_put_bp(hbp);
		return ENOMEM;
	}

	memset(rhash, 0, sizeof(rhash));
	if (tail_blk <= head_blk) {
		for (blk_no = tail_blk; blk_no < head_blk; ) {
C
Christoph Hellwig 已提交
3618 3619
			error = xlog_bread(log, blk_no, hblks, hbp, &offset);
			if (error)
L
Linus Torvalds 已提交
3620
				goto bread_err2;
C
Christoph Hellwig 已提交
3621

L
Linus Torvalds 已提交
3622 3623 3624 3625 3626 3627
			rhead = (xlog_rec_header_t *)offset;
			error = xlog_valid_rec_header(log, rhead, blk_no);
			if (error)
				goto bread_err2;

			/* blocks in data section */
3628
			bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
C
Christoph Hellwig 已提交
3629 3630
			error = xlog_bread(log, blk_no + hblks, bblks, dbp,
					   &offset);
L
Linus Torvalds 已提交
3631 3632
			if (error)
				goto bread_err2;
C
Christoph Hellwig 已提交
3633

C
Christoph Hellwig 已提交
3634 3635 3636 3637 3638 3639 3640
			error = xlog_unpack_data(rhead, offset, log);
			if (error)
				goto bread_err2;

			error = xlog_recover_process_data(log,
						rhash, rhead, offset, pass);
			if (error)
L
Linus Torvalds 已提交
3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
				goto bread_err2;
			blk_no += bblks + hblks;
		}
	} else {
		/*
		 * Perform recovery around the end of the physical log.
		 * When the head is not on the same cycle number as the tail,
		 * we can't do a sequential recovery as above.
		 */
		blk_no = tail_blk;
		while (blk_no < log->l_logBBsize) {
			/*
			 * Check for header wrapping around physical end-of-log
			 */
3655
			offset = hbp->b_addr;
L
Linus Torvalds 已提交
3656 3657 3658 3659
			split_hblks = 0;
			wrapped_hblks = 0;
			if (blk_no + hblks <= log->l_logBBsize) {
				/* Read header in one read */
C
Christoph Hellwig 已提交
3660 3661
				error = xlog_bread(log, blk_no, hblks, hbp,
						   &offset);
L
Linus Torvalds 已提交
3662 3663 3664 3665 3666 3667 3668 3669 3670
				if (error)
					goto bread_err2;
			} else {
				/* This LR is split across physical log end */
				if (blk_no != log->l_logBBsize) {
					/* some data before physical log end */
					ASSERT(blk_no <= INT_MAX);
					split_hblks = log->l_logBBsize - (int)blk_no;
					ASSERT(split_hblks > 0);
C
Christoph Hellwig 已提交
3671 3672 3673 3674
					error = xlog_bread(log, blk_no,
							   split_hblks, hbp,
							   &offset);
					if (error)
L
Linus Torvalds 已提交
3675 3676
						goto bread_err2;
				}
C
Christoph Hellwig 已提交
3677

L
Linus Torvalds 已提交
3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689
				/*
				 * Note: this black magic still works with
				 * large sector sizes (non-512) only because:
				 * - we increased the buffer size originally
				 *   by 1 sector giving us enough extra space
				 *   for the second read;
				 * - the log start is guaranteed to be sector
				 *   aligned;
				 * - we read the log end (LR header start)
				 *   _first_, then the log start (LR header end)
				 *   - order is important.
				 */
3690
				wrapped_hblks = hblks - split_hblks;
3691 3692 3693
				error = xlog_bread_offset(log, 0,
						wrapped_hblks, hbp,
						offset + BBTOB(split_hblks));
L
Linus Torvalds 已提交
3694 3695 3696 3697 3698 3699 3700 3701 3702
				if (error)
					goto bread_err2;
			}
			rhead = (xlog_rec_header_t *)offset;
			error = xlog_valid_rec_header(log, rhead,
						split_hblks ? blk_no : 0);
			if (error)
				goto bread_err2;

3703
			bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
L
Linus Torvalds 已提交
3704 3705 3706 3707
			blk_no += hblks;

			/* Read in data for log record */
			if (blk_no + bblks <= log->l_logBBsize) {
C
Christoph Hellwig 已提交
3708 3709
				error = xlog_bread(log, blk_no, bblks, dbp,
						   &offset);
L
Linus Torvalds 已提交
3710 3711 3712 3713 3714
				if (error)
					goto bread_err2;
			} else {
				/* This log record is split across the
				 * physical end of log */
3715
				offset = dbp->b_addr;
L
Linus Torvalds 已提交
3716 3717 3718 3719 3720 3721 3722 3723 3724
				split_bblks = 0;
				if (blk_no != log->l_logBBsize) {
					/* some data is before the physical
					 * end of log */
					ASSERT(!wrapped_hblks);
					ASSERT(blk_no <= INT_MAX);
					split_bblks =
						log->l_logBBsize - (int)blk_no;
					ASSERT(split_bblks > 0);
C
Christoph Hellwig 已提交
3725 3726 3727 3728
					error = xlog_bread(log, blk_no,
							split_bblks, dbp,
							&offset);
					if (error)
L
Linus Torvalds 已提交
3729 3730
						goto bread_err2;
				}
C
Christoph Hellwig 已提交
3731

L
Linus Torvalds 已提交
3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743
				/*
				 * Note: this black magic still works with
				 * large sector sizes (non-512) only because:
				 * - we increased the buffer size originally
				 *   by 1 sector giving us enough extra space
				 *   for the second read;
				 * - the log start is guaranteed to be sector
				 *   aligned;
				 * - we read the log end (LR header start)
				 *   _first_, then the log start (LR header end)
				 *   - order is important.
				 */
3744
				error = xlog_bread_offset(log, 0,
3745
						bblks - split_bblks, dbp,
3746
						offset + BBTOB(split_bblks));
C
Christoph Hellwig 已提交
3747 3748
				if (error)
					goto bread_err2;
L
Linus Torvalds 已提交
3749
			}
C
Christoph Hellwig 已提交
3750 3751 3752 3753 3754 3755 3756 3757

			error = xlog_unpack_data(rhead, offset, log);
			if (error)
				goto bread_err2;

			error = xlog_recover_process_data(log, rhash,
							rhead, offset, pass);
			if (error)
L
Linus Torvalds 已提交
3758 3759 3760 3761 3762 3763 3764 3765 3766
				goto bread_err2;
			blk_no += bblks;
		}

		ASSERT(blk_no >= log->l_logBBsize);
		blk_no -= log->l_logBBsize;

		/* read first part of physical log */
		while (blk_no < head_blk) {
C
Christoph Hellwig 已提交
3767 3768
			error = xlog_bread(log, blk_no, hblks, hbp, &offset);
			if (error)
L
Linus Torvalds 已提交
3769
				goto bread_err2;
C
Christoph Hellwig 已提交
3770

L
Linus Torvalds 已提交
3771 3772 3773 3774
			rhead = (xlog_rec_header_t *)offset;
			error = xlog_valid_rec_header(log, rhead, blk_no);
			if (error)
				goto bread_err2;
C
Christoph Hellwig 已提交
3775

3776
			bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
C
Christoph Hellwig 已提交
3777 3778 3779
			error = xlog_bread(log, blk_no+hblks, bblks, dbp,
					   &offset);
			if (error)
L
Linus Torvalds 已提交
3780
				goto bread_err2;
C
Christoph Hellwig 已提交
3781

C
Christoph Hellwig 已提交
3782 3783 3784 3785 3786 3787 3788
			error = xlog_unpack_data(rhead, offset, log);
			if (error)
				goto bread_err2;

			error = xlog_recover_process_data(log, rhash,
							rhead, offset, pass);
			if (error)
L
Linus Torvalds 已提交
3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815
				goto bread_err2;
			blk_no += bblks + hblks;
		}
	}

 bread_err2:
	xlog_put_bp(dbp);
 bread_err1:
	xlog_put_bp(hbp);
	return error;
}

/*
 * Do the recovery of the log.  We actually do this in two phases.
 * The two passes are necessary in order to implement the function
 * of cancelling a record written into the log.  The first pass
 * determines those things which have been cancelled, and the
 * second pass replays log items normally except for those which
 * have been cancelled.  The handling of the replay and cancellations
 * takes place in the log item type specific routines.
 *
 * The table of items which have cancel records in the log is allocated
 * and freed at this level, since only here do we know when all of
 * the log recovery has been completed.
 */
STATIC int
xlog_do_log_recovery(
M
Mark Tinguely 已提交
3816
	struct xlog	*log,
L
Linus Torvalds 已提交
3817 3818 3819
	xfs_daddr_t	head_blk,
	xfs_daddr_t	tail_blk)
{
3820
	int		error, i;
L
Linus Torvalds 已提交
3821 3822 3823 3824 3825 3826 3827

	ASSERT(head_blk != tail_blk);

	/*
	 * First do a pass to find all of the cancelled buf log items.
	 * Store them in the buf_cancel_table for use in the second pass.
	 */
3828 3829
	log->l_buf_cancel_table = kmem_zalloc(XLOG_BC_TABLE_SIZE *
						 sizeof(struct list_head),
L
Linus Torvalds 已提交
3830
						 KM_SLEEP);
3831 3832 3833
	for (i = 0; i < XLOG_BC_TABLE_SIZE; i++)
		INIT_LIST_HEAD(&log->l_buf_cancel_table[i]);

L
Linus Torvalds 已提交
3834 3835 3836
	error = xlog_do_recovery_pass(log, head_blk, tail_blk,
				      XLOG_RECOVER_PASS1);
	if (error != 0) {
3837
		kmem_free(log->l_buf_cancel_table);
L
Linus Torvalds 已提交
3838 3839 3840 3841 3842 3843 3844 3845 3846 3847
		log->l_buf_cancel_table = NULL;
		return error;
	}
	/*
	 * Then do a second pass to actually recover the items in the log.
	 * When it is complete free the table of buf cancel items.
	 */
	error = xlog_do_recovery_pass(log, head_blk, tail_blk,
				      XLOG_RECOVER_PASS2);
#ifdef DEBUG
3848
	if (!error) {
L
Linus Torvalds 已提交
3849 3850 3851
		int	i;

		for (i = 0; i < XLOG_BC_TABLE_SIZE; i++)
3852
			ASSERT(list_empty(&log->l_buf_cancel_table[i]));
L
Linus Torvalds 已提交
3853 3854 3855
	}
#endif	/* DEBUG */

3856
	kmem_free(log->l_buf_cancel_table);
L
Linus Torvalds 已提交
3857 3858 3859 3860 3861 3862 3863 3864 3865 3866
	log->l_buf_cancel_table = NULL;

	return error;
}

/*
 * Do the actual recovery
 */
STATIC int
xlog_do_recover(
M
Mark Tinguely 已提交
3867
	struct xlog	*log,
L
Linus Torvalds 已提交
3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
	xfs_daddr_t	head_blk,
	xfs_daddr_t	tail_blk)
{
	int		error;
	xfs_buf_t	*bp;
	xfs_sb_t	*sbp;

	/*
	 * First replay the images in the log.
	 */
	error = xlog_do_log_recovery(log, head_blk, tail_blk);
3879
	if (error)
L
Linus Torvalds 已提交
3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901
		return error;

	/*
	 * If IO errors happened during recovery, bail out.
	 */
	if (XFS_FORCED_SHUTDOWN(log->l_mp)) {
		return (EIO);
	}

	/*
	 * We now update the tail_lsn since much of the recovery has completed
	 * and there may be space available to use.  If there were no extent
	 * or iunlinks, we can free up the entire log and set the tail_lsn to
	 * be the last_sync_lsn.  This was set in xlog_find_tail to be the
	 * lsn of the last known good LR on disk.  If there are extent frees
	 * or iunlinks they will have some entries in the AIL; so we look at
	 * the AIL to determine how to set the tail_lsn.
	 */
	xlog_assign_tail_lsn(log->l_mp);

	/*
	 * Now that we've finished replaying all buffer and inode
3902
	 * updates, re-read in the superblock and reverify it.
L
Linus Torvalds 已提交
3903 3904 3905
	 */
	bp = xfs_getsb(log->l_mp, 0);
	XFS_BUF_UNDONE(bp);
3906
	ASSERT(!(XFS_BUF_ISWRITE(bp)));
L
Linus Torvalds 已提交
3907
	XFS_BUF_READ(bp);
3908
	XFS_BUF_UNASYNC(bp);
3909
	bp->b_ops = &xfs_sb_buf_ops;
L
Linus Torvalds 已提交
3910
	xfsbdstrat(log->l_mp, bp);
C
Christoph Hellwig 已提交
3911
	error = xfs_buf_iowait(bp);
3912
	if (error) {
3913
		xfs_buf_ioerror_alert(bp, __func__);
L
Linus Torvalds 已提交
3914 3915 3916 3917 3918 3919 3920
		ASSERT(0);
		xfs_buf_relse(bp);
		return error;
	}

	/* Convert superblock from on-disk format */
	sbp = &log->l_mp->m_sb;
3921
	xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
L
Linus Torvalds 已提交
3922
	ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
3923
	ASSERT(xfs_sb_good_version(sbp));
L
Linus Torvalds 已提交
3924 3925
	xfs_buf_relse(bp);

3926 3927 3928
	/* We've re-read the superblock so re-initialize per-cpu counters */
	xfs_icsb_reinit_counters(log->l_mp);

L
Linus Torvalds 已提交
3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942
	xlog_recover_check_summary(log);

	/* Normal transactions can now occur */
	log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
	return 0;
}

/*
 * Perform recovery and re-initialize some log variables in xlog_find_tail.
 *
 * Return error or zero.
 */
int
xlog_recover(
M
Mark Tinguely 已提交
3943
	struct xlog	*log)
L
Linus Torvalds 已提交
3944 3945 3946 3947 3948
{
	xfs_daddr_t	head_blk, tail_blk;
	int		error;

	/* find the tail of the log */
3949
	if ((error = xlog_find_tail(log, &head_blk, &tail_blk)))
L
Linus Torvalds 已提交
3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963
		return error;

	if (tail_blk != head_blk) {
		/* There used to be a comment here:
		 *
		 * disallow recovery on read-only mounts.  note -- mount
		 * checks for ENOSPC and turns it into an intelligent
		 * error message.
		 * ...but this is no longer true.  Now, unless you specify
		 * NORECOVERY (in which case this function would never be
		 * called), we just go ahead and recover.  We do this all
		 * under the vfs layer, so we can get away with it unless
		 * the device itself is read-only, in which case we fail.
		 */
3964
		if ((error = xfs_dev_is_read_only(log->l_mp, "recovery"))) {
L
Linus Torvalds 已提交
3965 3966 3967
			return error;
		}

3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986
		/*
		 * Version 5 superblock log feature mask validation. We know the
		 * log is dirty so check if there are any unknown log features
		 * in what we need to recover. If there are unknown features
		 * (e.g. unsupported transactions, then simply reject the
		 * attempt at recovery before touching anything.
		 */
		if (XFS_SB_VERSION_NUM(&log->l_mp->m_sb) == XFS_SB_VERSION_5 &&
		    xfs_sb_has_incompat_log_feature(&log->l_mp->m_sb,
					XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN)) {
			xfs_warn(log->l_mp,
"Superblock has unknown incompatible log features (0x%x) enabled.\n"
"The log can not be fully and/or safely recovered by this kernel.\n"
"Please recover the log on a kernel that supports the unknown features.",
				(log->l_mp->m_sb.sb_features_log_incompat &
					XFS_SB_FEAT_INCOMPAT_LOG_UNKNOWN));
			return EINVAL;
		}

3987 3988 3989
		xfs_notice(log->l_mp, "Starting recovery (logdev: %s)",
				log->l_mp->m_logname ? log->l_mp->m_logname
						     : "internal");
L
Linus Torvalds 已提交
3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007

		error = xlog_do_recover(log, head_blk, tail_blk);
		log->l_flags |= XLOG_RECOVERY_NEEDED;
	}
	return error;
}

/*
 * In the first part of recovery we replay inodes and buffers and build
 * up the list of extent free items which need to be processed.  Here
 * we process the extent free items and clean up the on disk unlinked
 * inode lists.  This is separated from the first part of recovery so
 * that the root and real-time bitmap inodes can be read in from disk in
 * between the two stages.  This is necessary so that we can free space
 * in the real-time portion of the file system.
 */
int
xlog_recover_finish(
M
Mark Tinguely 已提交
4008
	struct xlog	*log)
L
Linus Torvalds 已提交
4009 4010 4011 4012 4013 4014 4015 4016 4017 4018
{
	/*
	 * Now we're ready to do the transactions needed for the
	 * rest of recovery.  Start with completing all the extent
	 * free intent records and then process the unlinked inode
	 * lists.  At this point, we essentially run in normal mode
	 * except that we're still performing recovery actions
	 * rather than accepting new requests.
	 */
	if (log->l_flags & XLOG_RECOVERY_NEEDED) {
4019 4020 4021
		int	error;
		error = xlog_recover_process_efis(log);
		if (error) {
4022
			xfs_alert(log->l_mp, "Failed to recover EFIs");
4023 4024
			return error;
		}
L
Linus Torvalds 已提交
4025 4026 4027 4028 4029 4030
		/*
		 * Sync the log to get all the EFIs out of the AIL.
		 * This isn't absolutely necessary, but it helps in
		 * case the unlink transactions would have problems
		 * pushing the EFIs out of the way.
		 */
4031
		xfs_log_force(log->l_mp, XFS_LOG_SYNC);
L
Linus Torvalds 已提交
4032

C
Christoph Hellwig 已提交
4033
		xlog_recover_process_iunlinks(log);
L
Linus Torvalds 已提交
4034 4035 4036

		xlog_recover_check_summary(log);

4037 4038 4039
		xfs_notice(log->l_mp, "Ending recovery (logdev: %s)",
				log->l_mp->m_logname ? log->l_mp->m_logname
						     : "internal");
L
Linus Torvalds 已提交
4040 4041
		log->l_flags &= ~XLOG_RECOVERY_NEEDED;
	} else {
4042
		xfs_info(log->l_mp, "Ending clean mount");
L
Linus Torvalds 已提交
4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054
	}
	return 0;
}


#if defined(DEBUG)
/*
 * Read all of the agf and agi counters and check that they
 * are consistent with the superblock counters.
 */
void
xlog_recover_check_summary(
M
Mark Tinguely 已提交
4055
	struct xlog	*log)
L
Linus Torvalds 已提交
4056 4057 4058 4059 4060 4061 4062 4063 4064
{
	xfs_mount_t	*mp;
	xfs_agf_t	*agfp;
	xfs_buf_t	*agfbp;
	xfs_buf_t	*agibp;
	xfs_agnumber_t	agno;
	__uint64_t	freeblks;
	__uint64_t	itotal;
	__uint64_t	ifree;
4065
	int		error;
L
Linus Torvalds 已提交
4066 4067 4068 4069 4070 4071 4072

	mp = log->l_mp;

	freeblks = 0LL;
	itotal = 0LL;
	ifree = 0LL;
	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
4073 4074
		error = xfs_read_agf(mp, NULL, agno, 0, &agfbp);
		if (error) {
4075 4076
			xfs_alert(mp, "%s agf read failed agno %d error %d",
						__func__, agno, error);
4077 4078 4079 4080 4081
		} else {
			agfp = XFS_BUF_TO_AGF(agfbp);
			freeblks += be32_to_cpu(agfp->agf_freeblks) +
				    be32_to_cpu(agfp->agf_flcount);
			xfs_buf_relse(agfbp);
L
Linus Torvalds 已提交
4082 4083
		}

4084
		error = xfs_read_agi(mp, NULL, agno, &agibp);
4085 4086 4087 4088
		if (error) {
			xfs_alert(mp, "%s agi read failed agno %d error %d",
						__func__, agno, error);
		} else {
4089
			struct xfs_agi	*agi = XFS_BUF_TO_AGI(agibp);
4090

4091 4092 4093 4094
			itotal += be32_to_cpu(agi->agi_count);
			ifree += be32_to_cpu(agi->agi_freecount);
			xfs_buf_relse(agibp);
		}
L
Linus Torvalds 已提交
4095 4096 4097
	}
}
#endif /* DEBUG */