xfs_log_recover.c 112 KB
Newer Older
1
/*
2
 * Copyright (c) 2000-2003 Silicon Graphics, Inc.  All Rights Reserved.
3 4 5 6 7 8 9 10 11 12 13
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * 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.
 *
 * Further, this software is distributed without any warranty that it is
 * free of the rightful claim of any third person regarding infringement
Russell Cattelan's avatar
Russell Cattelan committed
14
 * or the like.  Any license provided herein, whether implied or
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
 * otherwise, applies only to this software file.  Patent licenses, if
 * any, provided herein do not apply to combinations of this program with
 * other software, or any other product whatsoever.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write the Free Software Foundation, Inc., 59
 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
 *
 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
 * Mountain View, CA  94043, or:
 *
 * http://www.sgi.com
 *
 * For further information regarding this notice, see:
 *
 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
 */

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65
#include "xfs.h"
#include "xfs_macros.h"
#include "xfs_types.h"
#include "xfs_inum.h"
#include "xfs_log.h"
#include "xfs_ag.h"
#include "xfs_sb.h"
#include "xfs_trans.h"
#include "xfs_dir.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_error.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc.h"
#include "xfs_attr_sf.h"
#include "xfs_dir_sf.h"
#include "xfs_dir2_sf.h"
#include "xfs_dinode.h"
#include "xfs_imap.h"
#include "xfs_inode_item.h"
#include "xfs_inode.h"
#include "xfs_ialloc_btree.h"
#include "xfs_ialloc.h"
#include "xfs_log_priv.h"
#include "xfs_buf_item.h"
#include "xfs_alloc_btree.h"
#include "xfs_log_recover.h"
#include "xfs_extfree_item.h"
#include "xfs_trans_priv.h"
#include "xfs_bit.h"
#include "xfs_quota.h"
#include "xfs_rw.h"
66

67 68
STATIC int	xlog_find_zeroed(xlog_t *, xfs_daddr_t *);
STATIC int	xlog_clear_stale_blocks(xlog_t *, xfs_lsn_t);
69 70 71
STATIC void	xlog_recover_insert_item_backq(xlog_recover_item_t **q,
					       xlog_recover_item_t *item);
#if defined(DEBUG)
72 73
STATIC void	xlog_recover_check_summary(xlog_t *);
STATIC void	xlog_recover_check_ail(xfs_mount_t *, xfs_log_item_t *, int);
74
#else
Russell Cattelan's avatar
Russell Cattelan committed
75 76
#define	xlog_recover_check_summary(log)
#define	xlog_recover_check_ail(mp, lip, gen)
77
#endif
78 79


80 81 82 83 84 85 86 87 88
/*
 * Sector aligned buffer routines for buffer create/read/write/access
 */

#define XLOG_SECTOR_ROUNDUP_BBCOUNT(log, bbs)	\
	( ((log)->l_sectbb_mask && (bbs & (log)->l_sectbb_mask)) ? \
	((bbs + (log)->l_sectbb_mask + 1) & ~(log)->l_sectbb_mask) : (bbs) )
#define XLOG_SECTOR_ROUNDDOWN_BLKNO(log, bno)	((bno) & ~(log)->l_sectbb_mask)

89
xfs_buf_t *
90 91 92
xlog_get_bp(
	xlog_t		*log,
	int		num_bblks)
93 94 95
{
	ASSERT(num_bblks > 0);

96 97 98 99 100
	if (log->l_sectbb_log) {
		if (num_bblks > 1)
			num_bblks += XLOG_SECTOR_ROUNDUP_BBCOUNT(log, 1);
		num_bblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, num_bblks);
	}
101
	return xfs_buf_get_noaddr(BBTOB(num_bblks), log->l_mp->m_logdev_targp);
102
}
103 104

void
105 106
xlog_put_bp(
	xfs_buf_t	*bp)
107
{
108
	xfs_buf_free(bp);
109
}
110 111 112


/*
Russell Cattelan's avatar
Russell Cattelan committed
113
 * nbblks should be uint, but oh well.  Just want to catch that 32-bit length.
114 115
 */
int
116 117 118 119 120
xlog_bread(
	xlog_t		*log,
	xfs_daddr_t	blk_no,
	int		nbblks,
	xfs_buf_t	*bp)
121
{
122 123 124 125 126 127
	int		error;

	if (log->l_sectbb_log) {
		blk_no = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, blk_no);
		nbblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, nbblks);
	}
128 129 130 131 132 133 134 135 136 137 138 139

	ASSERT(nbblks > 0);
	ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));
	ASSERT(bp);

	XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
	XFS_BUF_READ(bp);
	XFS_BUF_BUSY(bp);
	XFS_BUF_SET_COUNT(bp, BBTOB(nbblks));
	XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);

	xfsbdstrat(log->l_mp, bp);
140
	if ((error = xfs_iowait(bp)))
141 142 143
		xfs_ioerror_alert("xlog_bread", log->l_mp,
				  bp, XFS_BUF_ADDR(bp));
	return error;
144
}
145 146 147 148 149 150 151 152

/*
 * 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.
 */
int
xlog_bwrite(
153 154 155
	xlog_t		*log,
	xfs_daddr_t	blk_no,
	int		nbblks,
156 157
	xfs_buf_t	*bp)
{
158 159 160 161 162 163
	int		error;

	if (log->l_sectbb_log) {
		blk_no = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, blk_no);
		nbblks = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, nbblks);
	}
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178

	ASSERT(nbblks > 0);
	ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));

	XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
	XFS_BUF_ZEROFLAGS(bp);
	XFS_BUF_BUSY(bp);
	XFS_BUF_HOLD(bp);
	XFS_BUF_PSEMA(bp, PRIBIO);
	XFS_BUF_SET_COUNT(bp, BBTOB(nbblks));
	XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);

	if ((error = xfs_bwrite(log->l_mp, bp)))
		xfs_ioerror_alert("xlog_bwrite", log->l_mp,
				  bp, XFS_BUF_ADDR(bp));
179 180
	return error;
}
181

182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
xfs_caddr_t
xlog_align(
	xlog_t		*log,
	xfs_daddr_t	blk_no,
	int		nbblks,
	xfs_buf_t	*bp)
{
	xfs_caddr_t	ptr;

	if (!log->l_sectbb_log)
		return XFS_BUF_PTR(bp);

	ptr = XFS_BUF_PTR(bp) + BBTOB((int)blk_no & log->l_sectbb_mask);
	ASSERT(XFS_BUF_SIZE(bp) >=
		BBTOB(nbblks + (blk_no & log->l_sectbb_mask)));
	return ptr;
}
199 200 201

#ifdef DEBUG
/*
202
 * dump debug superblock and log record information
203
 */
204 205 206 207
STATIC void
xlog_header_check_dump(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
208
{
209 210 211 212 213 214 215 216 217 218
	int			b;

	printk("%s:  SB : uuid = ", __FUNCTION__);
	for (b = 0; b < 16; b++)
		printk("%02x",((unsigned char *)&mp->m_sb.sb_uuid)[b]);
	printk(", fmt = %d\n", XLOG_FMT);
	printk("    log : uuid = ");
	for (b = 0; b < 16; b++)
		printk("%02x",((unsigned char *)&head->h_fs_uuid)[b]);
	printk(", fmt = %d\n", INT_GET(head->h_fmt, ARCH_CONVERT));
219
}
220 221
#else
#define xlog_header_check_dump(mp, head)
222 223 224 225 226 227
#endif

/*
 * check log record header for recovery
 */
STATIC int
228 229 230
xlog_header_check_recover(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
231
{
232
	ASSERT(INT_GET(head->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
233

234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
	/*
	 * 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.
	 */
	if (unlikely(INT_GET(head->h_fmt, ARCH_CONVERT) != XLOG_FMT)) {
		xlog_warn(
	"XFS: dirty log written in incompatible format - can't recover");
		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))) {
		xlog_warn(
	"XFS: dirty log entry has mismatched uuid - can't recover");
		xlog_header_check_dump(mp, head);
		XFS_ERROR_REPORT("xlog_header_check_recover(2)",
				 XFS_ERRLEVEL_HIGH, mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	return 0;
255 256 257 258 259 260
}

/*
 * read the head block of the log and check the header
 */
STATIC int
261 262 263
xlog_header_check_mount(
	xfs_mount_t		*mp,
	xlog_rec_header_t	*head)
264
{
265
	ASSERT(INT_GET(head->h_magicno, ARCH_CONVERT) == XLOG_HEADER_MAGIC_NUM);
266

267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
	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.
		 */
		xlog_warn("XFS: nil uuid in log - IRIX style log");
	} else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
		xlog_warn("XFS: log has mismatched uuid - can't recover");
		xlog_header_check_dump(mp, head);
		XFS_ERROR_REPORT("xlog_header_check_mount",
				 XFS_ERRLEVEL_HIGH, mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	return 0;
282 283 284 285 286 287 288
}

STATIC void
xlog_recover_iodone(
	struct xfs_buf	*bp)
{
	xfs_mount_t	*mp;
289

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308
	ASSERT(XFS_BUF_FSPRIVATE(bp, void *));

	if (XFS_BUF_GETERROR(bp)) {
		/*
		 * We're not going to bother about retrying
		 * this during recovery. One strike!
		 */
		mp = XFS_BUF_FSPRIVATE(bp, xfs_mount_t *);
		xfs_ioerror_alert("xlog_recover_iodone",
				  mp, bp, XFS_BUF_ADDR(bp));
		xfs_force_shutdown(mp, XFS_METADATA_IO_ERROR);
	}
	XFS_BUF_SET_FSPRIVATE(bp, NULL);
	XFS_BUF_CLR_IODONE_FUNC(bp);
	xfs_biodone(bp);
}

/*
 * This routine finds (to an approximation) the first block in the physical
Russell Cattelan's avatar
Russell Cattelan committed
309
 * log which contains the given cycle.  It uses a binary search algorithm.
310 311 312 313
 * Note that the algorithm can not be perfect because the disk will not
 * necessarily be perfect.
 */
int
314 315 316 317 318 319
xlog_find_cycle_start(
	xlog_t		*log,
	xfs_buf_t	*bp,
	xfs_daddr_t	first_blk,
	xfs_daddr_t	*last_blk,
	uint		cycle)
320
{
321
	xfs_caddr_t	offset;
322 323 324
	xfs_daddr_t	mid_blk;
	uint		mid_cycle;
	int		error;
325 326 327 328 329

	mid_blk = BLK_AVG(first_blk, *last_blk);
	while (mid_blk != first_blk && mid_blk != *last_blk) {
		if ((error = xlog_bread(log, mid_blk, 1, bp)))
			return error;
330 331
		offset = xlog_align(log, mid_blk, 1, bp);
		mid_cycle = GET_CYCLE(offset, ARCH_CONVERT);
332 333 334 335 336 337 338 339 340 341 342 343 344
		if (mid_cycle == cycle) {
			*last_blk = mid_blk;
			/* last_half_cycle == mid_cycle */
		} else {
			first_blk = mid_blk;
			/* first_half_cycle == mid_cycle */
		}
		mid_blk = BLK_AVG(first_blk, *last_blk);
	}
	ASSERT((mid_blk == first_blk && mid_blk+1 == *last_blk) ||
	       (mid_blk == *last_blk && mid_blk-1 == first_blk));

	return 0;
345
}
346 347 348 349 350 351 352 353 354 355 356 357

/*
 * Check that the range of blocks does not contain the cycle number
 * given.  The scan needs to occur from front to back and the ptr into the
 * region must be updated since a later routine will need to perform another
 * test.  If the region is completely good, we end up returning the same
 * last block number.
 *
 * Set blkno to -1 if we encounter no errors.  This is an invalid block number
 * since we don't ever expect logs to get this large.
 */
STATIC int
358 359 360 361 362 363
xlog_find_verify_cycle(
	xlog_t		*log,
	xfs_daddr_t	start_blk,
	int		nbblks,
	uint		stop_on_cycle_no,
	xfs_daddr_t	*new_blk)
364
{
365 366 367 368 369 370
	xfs_daddr_t	i, j;
	uint		cycle;
	xfs_buf_t	*bp;
	xfs_daddr_t	bufblks;
	xfs_caddr_t	buf = NULL;
	int		error = 0;
371 372 373

	bufblks = 1 << ffs(nbblks);

374
	while (!(bp = xlog_get_bp(log, bufblks))) {
375 376
		/* can't get enough memory to do everything in one big buffer */
		bufblks >>= 1;
377
		if (bufblks <= log->l_sectbb_log)
378 379 380
			return ENOMEM;
	}

381 382
	for (i = start_blk; i < start_blk + nbblks; i += bufblks) {
		int	bcount;
383

384
		bcount = min(bufblks, (start_blk + nbblks - i));
385 386 387 388

		if ((error = xlog_bread(log, i, bcount, bp)))
			goto out;

389
		buf = xlog_align(log, i, bcount, bp);
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405
		for (j = 0; j < bcount; j++) {
			cycle = GET_CYCLE(buf, ARCH_CONVERT);
			if (cycle == stop_on_cycle_no) {
				*new_blk = i+j;
				goto out;
			}

			buf += BBSIZE;
		}
	}

	*new_blk = -1;

out:
	xlog_put_bp(bp);
	return error;
406
}
407 408 409 410 411 412 413 414 415 416 417 418 419 420

/*
 * 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
421 422 423 424 425
xlog_find_verify_log_record(
	xlog_t			*log,
	xfs_daddr_t		start_blk,
	xfs_daddr_t		*last_blk,
	int			extra_bblks)
426
{
427 428 429 430 431 432 433 434
	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;
435

436
	ASSERT(start_blk != 0 || *last_blk != start_blk);
437

438 439 440 441 442 443 444 445 446
	if (!(bp = xlog_get_bp(log, num_blks))) {
		if (!(bp = xlog_get_bp(log, 1)))
			return ENOMEM;
		smallmem = 1;
	} else {
		if ((error = xlog_bread(log, start_blk, num_blks, bp)))
			goto out;
		offset = xlog_align(log, start_blk, num_blks, bp);
		offset += ((num_blks - 1) << BBSHIFT);
447 448
	}

449 450
	for (i = (*last_blk) - 1; i >= 0; i--) {
		if (i < start_blk) {
451
			/* valid log record not found */
452 453 454 455 456 457
			xlog_warn(
		"XFS: Log inconsistent (didn't find previous header)");
			ASSERT(0);
			error = XFS_ERROR(EIO);
			goto out;
		}
458

459 460 461 462 463
		if (smallmem) {
			if ((error = xlog_bread(log, i, 1, bp)))
				goto out;
			offset = xlog_align(log, i, 1, bp);
		}
464

465
		head = (xlog_rec_header_t *)offset;
466

467 468 469
		if (XLOG_HEADER_MAGIC_NUM ==
		    INT_GET(head->h_magicno, ARCH_CONVERT))
			break;
470

471 472 473
		if (!smallmem)
			offset -= BBSIZE;
	}
474

475 476 477 478 479 480 481 482 483
	/*
	 * 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;
	}
484

485 486 487 488 489 490
	/*
	 * 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;
491

492 493 494 495 496 497 498 499 500
	/*
	 * 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.
	 */
	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
		uint	h_size = INT_GET(head->h_size, ARCH_CONVERT);
501

502 503 504 505 506 507
		xhdrs = h_size / XLOG_HEADER_CYCLE_SIZE;
		if (h_size % XLOG_HEADER_CYCLE_SIZE)
			xhdrs++;
	} else {
		xhdrs = 1;
	}
508

509 510 511
	if (*last_blk - i + extra_bblks
			!= BTOBB(INT_GET(head->h_len, ARCH_CONVERT)) + xhdrs)
		*last_blk = i;
512

513 514 515 516
out:
	xlog_put_bp(bp);
	return error;
}
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531

/*
 * 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.
 */
int
532 533 534
xlog_find_head(
	xlog_t 		*log,
	xfs_daddr_t	*return_head_blk)
535
{
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560
	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
			 */
			xlog_warn("XFS: totally zeroed log");
		}

		return 0;
	} else if (error) {
		xlog_warn("XFS: empty log check failed");
		return error;
561 562
	}

563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578
	first_blk = 0;			/* get cycle # of 1st block */
	bp = xlog_get_bp(log, 1);
	if (!bp)
		return ENOMEM;
	if ((error = xlog_bread(log, 0, 1, bp)))
		goto bp_err;
	offset = xlog_align(log, 0, 1, bp);
	first_half_cycle = GET_CYCLE(offset, ARCH_CONVERT);

	last_blk = head_blk = log_bbnum - 1;	/* get cycle # of last block */
	if ((error = xlog_bread(log, last_blk, 1, bp)))
		goto bp_err;
	offset = xlog_align(log, last_blk, 1, bp);
	last_half_cycle = GET_CYCLE(offset, ARCH_CONVERT);
	ASSERT(last_half_cycle != 0);

579
	/*
580 581 582 583
	 * 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,
584
	 * we set it to log_bbnum which is an invalid block number, but this
585 586 587 588
	 * 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.
589
	 */
590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
	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
		 * In this case the head really is somwhere at the end of the
		 * 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
		 *        x + 1 ... | x ...
		 * 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
		 *        x + 1 ... | x | x + 1 | x ...
		 *                               ^ binary search stopped here
		 * or
		 *        x + 1 ... | x ... | x - 1 | x
		 *        <---------> less than scan distance
		 */
		stop_on_cycle = last_half_cycle;
		if ((error = xlog_find_cycle_start(log, bp, first_blk,
						&head_blk, last_half_cycle)))
			goto bp_err;
	}

645
	/*
646 647 648 649 650
	 * 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.
651
	 */
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720
	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.
		 */
		start_blk = log_bbnum - num_scan_bblks + head_blk;
		ASSERT(head_blk <= INT_MAX &&
			(xfs_daddr_t) num_scan_bblks - head_blk >= 0);
		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;
			goto bad_blk;
		}

		/*
		 * 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;
	}

 bad_blk:
721
	/*
722 723
	 * Now we need to make sure head_blk is not pointing to a block in
	 * the middle of a log record.
724
	 */
725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757
	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 */
			start_blk = log_bbnum - num_scan_bblks + head_blk;
			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;
758 759
	}

760 761 762 763 764
	xlog_put_bp(bp);
	if (head_blk == log_bbnum)
		*return_head_blk = 0;
	else
		*return_head_blk = head_blk;
765
	/*
766 767 768 769
	 * 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.
770
	 */
771
	return 0;
772

773
 bp_err:
774 775 776 777 778
	xlog_put_bp(bp);

	if (error)
	    xlog_warn("XFS: failed to find log head");
	return error;
779
}
780 781 782 783 784 785 786

/*
 * 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
Russell Cattelan's avatar
Russell Cattelan committed
787
 * to get a sync block number.  The only concern is to figure out which
788 789 790
 * log record header to believe.
 *
 * The following algorithm uses the log record header with the largest
Russell Cattelan's avatar
Russell Cattelan committed
791
 * lsn.  The entire log record does not need to be valid.  We only care
792 793 794 795 796 797
 * that the header is valid.
 *
 * We could speed up search by using current head_blk buffer, but it is not
 * available.
 */
int
798 799 800 801 802
xlog_find_tail(
	xlog_t			*log,
	xfs_daddr_t		*head_blk,
	xfs_daddr_t		*tail_blk,
	int			readonly)
803 804 805
{
	xlog_rec_header_t	*rhead;
	xlog_op_header_t	*op_head;
806
	xfs_caddr_t		offset = NULL;
807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
	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;

822
	bp = xlog_get_bp(log, 1);
823 824 825 826 827
	if (!bp)
		return ENOMEM;
	if (*head_blk == 0) {				/* special case */
		if ((error = xlog_bread(log, 0, 1, bp)))
			goto bread_err;
828 829
		offset = xlog_align(log, 0, 1, bp);
		if (GET_CYCLE(offset, ARCH_CONVERT) == 0) {
830 831 832 833 834 835 836 837 838 839 840 841 842
			*tail_blk = 0;
			/* leave all other log inited values alone */
			goto exit;
		}
	}

	/*
	 * Search backwards looking for log record header block
	 */
	ASSERT(*head_blk < INT_MAX);
	for (i = (int)(*head_blk) - 1; i >= 0; i--) {
		if ((error = xlog_bread(log, i, 1, bp)))
			goto bread_err;
843
		offset = xlog_align(log, i, 1, bp);
844
		if (XLOG_HEADER_MAGIC_NUM ==
845
		    INT_GET(*(uint *)offset, ARCH_CONVERT)) {
846 847 848 849 850 851 852 853 854 855 856 857 858 859
			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--) {
			if ((error = xlog_bread(log, i, 1, bp)))
				goto bread_err;
860
			offset = xlog_align(log, i, 1, bp);
861
			if (XLOG_HEADER_MAGIC_NUM ==
862
			    INT_GET(*(uint*)offset, ARCH_CONVERT)) {
863 864 865 866 867 868 869 870 871 872 873 874
				found = 2;
				break;
			}
		}
	}
	if (!found) {
		xlog_warn("XFS: xlog_find_tail: couldn't find sync record");
		ASSERT(0);
		return XFS_ERROR(EIO);
	}

	/* find blk_no of tail of log */
875
	rhead = (xlog_rec_header_t *)offset;
876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901
	*tail_blk = BLOCK_LSN(rhead->h_tail_lsn, ARCH_CONVERT);

	/*
	 * 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;
	log->l_curr_cycle = INT_GET(rhead->h_cycle, ARCH_CONVERT);
	if (found == 2)
		log->l_curr_cycle++;
	log->l_tail_lsn = INT_GET(rhead->h_tail_lsn, ARCH_CONVERT);
	log->l_last_sync_lsn = INT_GET(rhead->h_lsn, ARCH_CONVERT);
	log->l_grant_reserve_cycle = log->l_curr_cycle;
	log->l_grant_reserve_bytes = BBTOB(log->l_curr_block);
	log->l_grant_write_cycle = log->l_curr_cycle;
	log->l_grant_write_bytes = BBTOB(log->l_curr_block);

	/*
	 * Look for unmount record.  If we find it, then we know there
Russell Cattelan's avatar
Russell Cattelan committed
902
	 * was a clean unmount.  Since 'i' could be the last block in
903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934
	 * 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.
	 */
	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
		int	h_size = INT_GET(rhead->h_size, ARCH_CONVERT);
		int	h_version = INT_GET(rhead->h_version, ARCH_CONVERT);

		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)
		BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT))) % log->l_logBBsize;
	tail_lsn = log->l_tail_lsn;
	if (*head_blk == after_umount_blk &&
	    INT_GET(rhead->h_num_logops, ARCH_CONVERT) == 1) {
		umount_data_blk = (i + hblks) % log->l_logBBsize;
		if ((error = xlog_bread(log, umount_data_blk, 1, bp))) {
			goto bread_err;
		}
935 936
		offset = xlog_align(log, umount_data_blk, 1, bp);
		op_head = (xlog_op_header_t *)offset;
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
		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.
			 */
			ASSIGN_ANY_LSN(log->l_tail_lsn, log->l_curr_cycle,
					after_umount_blk, ARCH_NOCONVERT);
			ASSIGN_ANY_LSN(log->l_last_sync_lsn, log->l_curr_cycle,
					after_umount_blk, ARCH_NOCONVERT);
			*tail_blk = after_umount_blk;
		}
	}

	/*
	 * 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
961 962 963 964 965 966 967 968
	 *
	 * 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.
969
	 */
970
	if (!xfs_readonly_buftarg(log->l_mp->m_logdev_targp)) {
971
		error = xlog_clear_stale_blocks(log, tail_lsn);
972
	}
973 974 975 976 977 978 979 980

bread_err:
exit:
	xlog_put_bp(bp);

	if (error)
		xlog_warn("XFS: failed to locate log tail");
	return error;
981
}
982 983 984 985 986

/*
 * Is the log zeroed at all?
 *
 * The last binary search should be changed to perform an X block read
Russell Cattelan's avatar
Russell Cattelan committed
987
 * once X becomes small enough.  You can then search linearly through
988 989 990 991 992 993 994 995 996 997 998 999
 * 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
 */
int
1000 1001 1002
xlog_find_zeroed(
	xlog_t		*log,
	xfs_daddr_t	*blk_no)
1003 1004
{
	xfs_buf_t	*bp;
1005
	xfs_caddr_t	offset;
Russell Cattelan's avatar
Russell Cattelan committed
1006
	uint	        first_cycle, last_cycle;
1007
	xfs_daddr_t	new_blk, last_blk, start_blk;
Russell Cattelan's avatar
Russell Cattelan committed
1008 1009
	xfs_daddr_t     num_scan_bblks;
	int	        error, log_bbnum = log->l_logBBsize;
1010 1011

	/* check totally zeroed log */
1012
	bp = xlog_get_bp(log, 1);
1013 1014 1015 1016
	if (!bp)
		return ENOMEM;
	if ((error = xlog_bread(log, 0, 1, bp)))
		goto bp_err;
1017 1018
	offset = xlog_align(log, 0, 1, bp);
	first_cycle = GET_CYCLE(offset, ARCH_CONVERT);
1019 1020 1021 1022 1023 1024 1025 1026 1027
	if (first_cycle == 0) {		/* completely zeroed log */
		*blk_no = 0;
		xlog_put_bp(bp);
		return -1;
	}

	/* check partially zeroed log */
	if ((error = xlog_bread(log, log_bbnum-1, 1, bp)))
		goto bp_err;
1028 1029
	offset = xlog_align(log, log_bbnum-1, 1, bp);
	last_cycle = GET_CYCLE(offset, ARCH_CONVERT);
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
	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.
		 */
		xlog_warn("XFS: Log inconsistent or not a log (last==0, first!=1)");
		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;

	/*
Russell Cattelan's avatar
Russell Cattelan committed
1049
	 * Validate the answer.  Because there is no way to guarantee that
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
	 * 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
Russell Cattelan's avatar
Russell Cattelan committed
1064 1065
	 *        1 ... | 0 | 1 | 0...
	 *                       ^ binary search ends here
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
	 */
	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;
1090
}
1091 1092

/*
1093
 * These are simple subroutines used by xlog_clear_stale_blocks() below
1094 1095 1096
 * to initialize a buffer full of empty log record headers and write
 * them into the log.
 */
1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
STATIC void
xlog_add_record(
	xlog_t			*log,
	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);
	INT_SET(recp->h_magicno, ARCH_CONVERT, XLOG_HEADER_MAGIC_NUM);
	INT_SET(recp->h_cycle, ARCH_CONVERT, cycle);
	INT_SET(recp->h_version, ARCH_CONVERT,
			XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
	ASSIGN_ANY_LSN(recp->h_lsn, cycle, block, ARCH_CONVERT);
	ASSIGN_ANY_LSN(recp->h_tail_lsn, tail_cycle, tail_block, ARCH_CONVERT);
	INT_SET(recp->h_fmt, ARCH_CONVERT, XLOG_FMT);
	memcpy(&recp->h_fs_uuid, &log->l_mp->m_sb.sb_uuid, sizeof(uuid_t));
}

1119 1120
STATIC int
xlog_write_log_records(
1121 1122 1123 1124 1125 1126
	xlog_t		*log,
	int		cycle,
	int		start_block,
	int		blocks,
	int		tail_cycle,
	int		tail_block)
1127
{
1128 1129 1130 1131 1132 1133 1134 1135
	xfs_caddr_t	offset;
	xfs_buf_t	*bp;
	int		balign, ealign;
	int		sectbb = XLOG_SECTOR_ROUNDUP_BBCOUNT(log, 1);
	int		end_block = start_block + blocks;
	int		bufblks;
	int		error = 0;
	int		i, j = 0;
1136 1137

	bufblks = 1 << ffs(blocks);
1138
	while (!(bp = xlog_get_bp(log, bufblks))) {
1139
		bufblks >>= 1;
1140
		if (bufblks <= log->l_sectbb_log)
1141 1142 1143
			return ENOMEM;
	}

1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
	/* 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.
	 */
	balign = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, start_block);
	if (balign != start_block) {
		if ((error = xlog_bread(log, start_block, 1, bp))) {
			xlog_put_bp(bp);
			return error;
		}
		j = start_block - balign;
	}
1156 1157

	for (i = start_block; i < end_block; i += bufblks) {
1158 1159 1160 1161 1162 1163 1164 1165
		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.
1166
		 */
1167 1168 1169 1170 1171 1172 1173 1174
		ealign = XLOG_SECTOR_ROUNDDOWN_BLKNO(log, end_block);
		if (j == 0 && (start_block + endcount > ealign)) {
			offset = XFS_BUF_PTR(bp);
			balign = BBTOB(ealign - start_block);
			XFS_BUF_SET_PTR(bp, offset + balign, BBTOB(sectbb));
			if ((error = xlog_bread(log, ealign, sectbb, bp)))
				break;
			XFS_BUF_SET_PTR(bp, offset, bufblks);
1175
		}
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187

		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;
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
	}
	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(
	xlog_t		*log,
	xfs_lsn_t	tail_lsn)
{
1214 1215 1216 1217 1218
	int		tail_cycle, head_cycle;
	int		tail_block, head_block;
	int		tail_distance, max_distance;
	int		distance;
	int		error;
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238

	tail_cycle = CYCLE_LSN(tail_lsn, ARCH_NOCONVERT);
	tail_block = BLOCK_LSN(tail_lsn, ARCH_NOCONVERT);
	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.
		 */
1239 1240 1241
		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);
1242
			return XFS_ERROR(EFSCORRUPTED);
1243
		}
1244
		tail_distance = tail_block + (log->l_logBBsize - head_block);
1245 1246 1247 1248 1249 1250
	} 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.
		 */
1251 1252 1253
		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);
1254
			return XFS_ERROR(EFSCORRUPTED);
1255
		}
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
		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
Russell Cattelan's avatar
Russell Cattelan committed
1281
		 * wrapping around the end of the log.  Just do it
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332
		 * 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 *
1333 1334 1335
xlog_recover_find_tid(
	xlog_recover_t		*q,
	xlog_tid_t		tid)
1336
{
1337
	xlog_recover_t		*p = q;
1338 1339 1340 1341 1342 1343 1344

	while (p != NULL) {
		if (p->r_log_tid == tid)
		    break;
		p = p->r_next;
	}
	return p;
1345
}
1346 1347

STATIC void
1348 1349 1350
xlog_recover_put_hashq(
	xlog_recover_t		**q,
	xlog_recover_t		*trans)
1351 1352 1353
{
	trans->r_next = *q;
	*q = trans;
1354
}
1355 1356

STATIC void
1357 1358
xlog_recover_add_item(
	xlog_recover_item_t	**itemq)
1359
{
1360
	xlog_recover_item_t	*item;
1361

1362
	item = kmem_zalloc(sizeof(xlog_recover_item_t), KM_SLEEP);
1363
	xlog_recover_insert_item_backq(itemq, item);
1364
}
1365 1366

STATIC int
1367 1368 1369 1370
xlog_recover_add_to_cont_trans(
	xlog_recover_t		*trans,
	xfs_caddr_t		dp,
	int			len)
1371 1372
{
	xlog_recover_item_t	*item;
1373
	xfs_caddr_t		ptr, old_ptr;
1374 1375 1376 1377 1378 1379
	int			old_len;

	item = trans->r_itemq;
	if (item == 0) {
		/* finish copying rest of trans header */
		xlog_recover_add_item(&trans->r_itemq);
1380 1381
		ptr = (xfs_caddr_t) &trans->r_theader +
				sizeof(xfs_trans_header_t) - len;
1382
		memcpy(ptr, dp, len); /* d, s, l */
1383 1384 1385 1386 1387 1388 1389 1390
		return 0;
	}
	item = item->ri_prev;

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

	ptr = kmem_realloc(old_ptr, len+old_len, old_len, 0);
1391
	memcpy(&ptr[old_len], dp, len); /* d, s, l */
1392 1393 1394
	item->ri_buf[item->ri_cnt-1].i_len += len;
	item->ri_buf[item->ri_cnt-1].i_addr = ptr;
	return 0;
1395
}
1396

1397 1398
/*
 * The next region to add is the start of a new region.  It could be
1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
 * 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
1411 1412 1413 1414
xlog_recover_add_to_trans(
	xlog_recover_t		*trans,
	xfs_caddr_t		dp,
	int			len)
1415
{
1416 1417 1418
	xfs_inode_log_format_t	*in_f;			/* any will do */
	xlog_recover_item_t	*item;
	xfs_caddr_t		ptr;
1419 1420 1421 1422 1423 1424 1425 1426

	if (!len)
		return 0;
	item = trans->r_itemq;
	if (item == 0) {
		ASSERT(*(uint *)dp == XFS_TRANS_HEADER_MAGIC);
		if (len == sizeof(xfs_trans_header_t))
			xlog_recover_add_item(&trans->r_itemq);
1427
		memcpy(&trans->r_theader, dp, len); /* d, s, l */
1428 1429
		return 0;
	}
1430

1431
	ptr = kmem_alloc(len, KM_SLEEP);
1432 1433 1434
	memcpy(ptr, dp, len);
	in_f = (xfs_inode_log_format_t *)ptr;

1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
	if (item->ri_prev->ri_total != 0 &&
	     item->ri_prev->ri_total == item->ri_prev->ri_cnt) {
		xlog_recover_add_item(&trans->r_itemq);
	}
	item = trans->r_itemq;
	item = item->ri_prev;

	if (item->ri_total == 0) {		/* first region to be added */
		item->ri_total	= in_f->ilf_size;
		ASSERT(item->ri_total <= XLOG_MAX_REGIONS_IN_ITEM);
		item->ri_buf = kmem_zalloc((item->ri_total *
1446
					    sizeof(xfs_log_iovec_t)), KM_SLEEP);
1447 1448 1449 1450 1451 1452 1453
	}
	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++;
	return 0;
1454
}
1455 1456

STATIC void
1457 1458 1459 1460
xlog_recover_new_tid(
	xlog_recover_t		**q,
	xlog_tid_t		tid,
	xfs_lsn_t		lsn)
1461
{
1462
	xlog_recover_t		*trans;
1463

1464
	trans = kmem_zalloc(sizeof(xlog_recover_t), KM_SLEEP);
1465 1466 1467
	trans->r_log_tid   = tid;
	trans->r_lsn	   = lsn;
	xlog_recover_put_hashq(q, trans);
1468
}
1469 1470

STATIC int
1471 1472 1473
xlog_recover_unlink_tid(
	xlog_recover_t		**q,
	xlog_recover_t		*trans)
1474
{
1475 1476
	xlog_recover_t		*tp;
	int			found = 0;
1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498

	ASSERT(trans != 0);
	if (trans == *q) {
		*q = (*q)->r_next;
	} else {
		tp = *q;
		while (tp != 0) {
			if (tp->r_next == trans) {
				found = 1;
				break;
			}
			tp = tp->r_next;
		}
		if (!found) {
			xlog_warn(
			     "XFS: xlog_recover_unlink_tid: trans not found");
			ASSERT(0);
			return XFS_ERROR(EIO);
		}
		tp->r_next = tp->r_next->r_next;
	}
	return 0;
1499
}
1500 1501

STATIC void
1502 1503 1504
xlog_recover_insert_item_backq(
	xlog_recover_item_t	**q,
	xlog_recover_item_t	*item)
1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
{
	if (*q == 0) {
		item->ri_prev = item->ri_next = item;
		*q = item;
	} else {
		item->ri_next		= *q;
		item->ri_prev		= (*q)->ri_prev;
		(*q)->ri_prev		= item;
		item->ri_prev->ri_next	= item;
	}
1515
}
1516 1517

STATIC void
1518 1519 1520
xlog_recover_insert_item_frontq(
	xlog_recover_item_t	**q,
	xlog_recover_item_t	*item)
1521 1522 1523
{
	xlog_recover_insert_item_backq(q, item);
	*q = item;
1524
}
1525 1526

STATIC int
1527 1528 1529
xlog_recover_reorder_trans(
	xlog_t			*log,
	xlog_recover_t		*trans)
1530
{
1531
	xlog_recover_item_t	*first_item, *itemq, *itemq_next;
1532 1533 1534
	xfs_buf_log_format_t	*buf_f;
	xfs_buf_log_format_v1_t	*obuf_f;
	ushort			flags;
1535

1536 1537 1538 1539
	first_item = itemq = trans->r_itemq;
	trans->r_itemq = NULL;
	do {
		itemq_next = itemq->ri_next;
1540
		buf_f = (xfs_buf_log_format_t *)itemq->ri_buf[0].i_addr;
1541 1542
		switch (ITEM_TYPE(itemq)) {
		case XFS_LI_BUF:
1543 1544
			flags = buf_f->blf_flags;
			break;
1545 1546
		case XFS_LI_6_1_BUF:
		case XFS_LI_5_3_BUF:
1547 1548
			obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
			flags = obuf_f->blf_flags;
1549
			break;
1550 1551 1552 1553 1554 1555
		}

		switch (ITEM_TYPE(itemq)) {
		case XFS_LI_BUF:
		case XFS_LI_6_1_BUF:
		case XFS_LI_5_3_BUF:
Dave Jones's avatar
Dave Jones committed
1556
			if (!(flags & XFS_BLI_CANCEL)) {
1557 1558 1559 1560
				xlog_recover_insert_item_frontq(&trans->r_itemq,
								itemq);
				break;
			}
1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
		case XFS_LI_INODE:
		case XFS_LI_6_1_INODE:
		case XFS_LI_5_3_INODE:
		case XFS_LI_DQUOT:
		case XFS_LI_QUOTAOFF:
		case XFS_LI_EFD:
		case XFS_LI_EFI:
			xlog_recover_insert_item_backq(&trans->r_itemq, itemq);
			break;
		default:
			xlog_warn(
	"XFS: xlog_recover_reorder_trans: unrecognized type of log operation");
			ASSERT(0);
			return XFS_ERROR(EIO);
		}
		itemq = itemq_next;
	} while (first_item != itemq);
	return 0;
}
1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593

/*
 * 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.
 */
STATIC void
1594 1595 1596
xlog_recover_do_buffer_pass1(
	xlog_t			*log,
	xfs_buf_log_format_t	*buf_f)
1597 1598 1599 1600 1601
{
	xfs_buf_cancel_t	*bcp;
	xfs_buf_cancel_t	*nextp;
	xfs_buf_cancel_t	*prevp;
	xfs_buf_cancel_t	**bucket;
Russell Cattelan's avatar
Russell Cattelan committed
1602
	xfs_buf_log_format_v1_t	*obuf_f;
1603 1604 1605
	xfs_daddr_t		blkno = 0;
	uint			len = 0;
	ushort			flags = 0;
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624

	switch (buf_f->blf_type) {
	case XFS_LI_BUF:
		blkno = buf_f->blf_blkno;
		len = buf_f->blf_len;
		flags = buf_f->blf_flags;
		break;
	case XFS_LI_6_1_BUF:
	case XFS_LI_5_3_BUF:
		obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
		blkno = (xfs_daddr_t) obuf_f->blf_blkno;
		len = obuf_f->blf_len;
		flags = obuf_f->blf_flags;
		break;
	}

	/*
	 * If this isn't a cancel buffer item, then just return.
	 */
1625
	if (!(flags & XFS_BLI_CANCEL))
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
		return;

	/*
	 * Insert an xfs_buf_cancel record into the hash table of
	 * them.  If there is already an identical record, bump
	 * its reference count.
	 */
	bucket = &log->l_buf_cancel_table[(__uint64_t)blkno %
					  XLOG_BC_TABLE_SIZE];
	/*
	 * If the hash bucket is empty then just insert a new record into
	 * the bucket.
	 */
	if (*bucket == NULL) {
1640 1641
		bcp = (xfs_buf_cancel_t *)kmem_alloc(sizeof(xfs_buf_cancel_t),
						     KM_SLEEP);
1642 1643 1644 1645 1646 1647 1648 1649 1650 1651
		bcp->bc_blkno = blkno;
		bcp->bc_len = len;
		bcp->bc_refcount = 1;
		bcp->bc_next = NULL;
		*bucket = bcp;
		return;
	}

	/*
	 * The hash bucket is not empty, so search for duplicates of our
Russell Cattelan's avatar
Russell Cattelan committed
1652
	 * record.  If we find one them just bump its refcount.  If not
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
	 * then add us at the end of the list.
	 */
	prevp = NULL;
	nextp = *bucket;
	while (nextp != NULL) {
		if (nextp->bc_blkno == blkno && nextp->bc_len == len) {
			nextp->bc_refcount++;
			return;
		}
		prevp = nextp;
		nextp = nextp->bc_next;
	}
	ASSERT(prevp != NULL);
1666 1667
	bcp = (xfs_buf_cancel_t *)kmem_alloc(sizeof(xfs_buf_cancel_t),
					     KM_SLEEP);
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
	bcp->bc_blkno = blkno;
	bcp->bc_len = len;
	bcp->bc_refcount = 1;
	bcp->bc_next = NULL;
	prevp->bc_next = bcp;
}

/*
 * 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
 * actually a buffer cancel item (XFS_BLI_CANCEL is set), then decrement
 * 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
1689
xlog_check_buffer_cancelled(
1690
	xlog_t			*log,
1691 1692 1693
	xfs_daddr_t		blkno,
	uint			len,
	ushort			flags)
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759
{
	xfs_buf_cancel_t	*bcp;
	xfs_buf_cancel_t	*prevp;
	xfs_buf_cancel_t	**bucket;

	if (log->l_buf_cancel_table == NULL) {
		/*
		 * There is nothing in the table built in pass one,
		 * so this buffer must not be cancelled.
		 */
		ASSERT(!(flags & XFS_BLI_CANCEL));
		return 0;
	}

	bucket = &log->l_buf_cancel_table[(__uint64_t)blkno %
					  XLOG_BC_TABLE_SIZE];
	bcp = *bucket;
	if (bcp == NULL) {
		/*
		 * There is no corresponding entry in the table built
		 * in pass one, so this buffer has not been cancelled.
		 */
		ASSERT(!(flags & XFS_BLI_CANCEL));
		return 0;
	}

	/*
	 * Search for an entry in the buffer cancel table that
	 * matches our buffer.
	 */
	prevp = NULL;
	while (bcp != NULL) {
		if (bcp->bc_blkno == blkno && bcp->bc_len == len) {
			/*
			 * 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_BLI_CANCEL) {
				bcp->bc_refcount--;
				if (bcp->bc_refcount == 0) {
					if (prevp == NULL) {
						*bucket = bcp->bc_next;
					} else {
						prevp->bc_next = bcp->bc_next;
					}
					kmem_free(bcp,
						  sizeof(xfs_buf_cancel_t));
				}
			}
			return 1;
		}
		prevp = bcp;
		bcp = bcp->bc_next;
	}
	/*
	 * We didn't find a corresponding entry in the table, so
	 * return 0 so that the buffer is NOT cancelled.
	 */
	ASSERT(!(flags & XFS_BLI_CANCEL));
	return 0;
}

1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
STATIC int
xlog_recover_do_buffer_pass2(
	xlog_t			*log,
	xfs_buf_log_format_t	*buf_f)
{
	xfs_buf_log_format_v1_t	*obuf_f;
	xfs_daddr_t		blkno = 0;
	ushort			flags = 0;
	uint			len = 0;

	switch (buf_f->blf_type) {
	case XFS_LI_BUF:
		blkno = buf_f->blf_blkno;
		flags = buf_f->blf_flags;
		len = buf_f->blf_len;
		break;
	case XFS_LI_6_1_BUF:
	case XFS_LI_5_3_BUF:
		obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
		blkno = (xfs_daddr_t) obuf_f->blf_blkno;
		flags = obuf_f->blf_flags;
		len = (xfs_daddr_t) obuf_f->blf_len;
		break;
	}

	return xlog_check_buffer_cancelled(log, blkno, len, flags);
}

1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
/*
 * 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_do_inode_trans().
 *
 * 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.
 */
STATIC int
1802 1803 1804 1805 1806
xlog_recover_do_inode_buffer(
	xfs_mount_t		*mp,
	xlog_recover_item_t	*item,
	xfs_buf_t		*bp,
	xfs_buf_log_format_t	*buf_f)
1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
{
	int			i;
	int			item_index;
	int			bit;
	int			nbits;
	int			reg_buf_offset;
	int			reg_buf_bytes;
	int			next_unlinked_offset;
	int			inodes_per_buf;
	xfs_agino_t		*logged_nextp;
	xfs_agino_t		*buffer_nextp;
Russell Cattelan's avatar
Russell Cattelan committed
1818
	xfs_buf_log_format_v1_t	*obuf_f;
1819 1820
	unsigned int		*data_map = NULL;
	unsigned int		map_size = 0;
1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895

	switch (buf_f->blf_type) {
	case XFS_LI_BUF:
		data_map = buf_f->blf_data_map;
		map_size = buf_f->blf_map_size;
		break;
	case XFS_LI_6_1_BUF:
	case XFS_LI_5_3_BUF:
		obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
		data_map = obuf_f->blf_data_map;
		map_size = obuf_f->blf_map_size;
		break;
	}
	/*
	 * Set the variables corresponding to the current region to
	 * 0 so that we'll initialize them on the first pass through
	 * the loop.
	 */
	reg_buf_offset = 0;
	reg_buf_bytes = 0;
	bit = 0;
	nbits = 0;
	item_index = 0;
	inodes_per_buf = XFS_BUF_COUNT(bp) >> mp->m_sb.sb_inodelog;
	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;
			bit = xfs_next_bit(data_map, map_size, bit);

			/*
			 * If there are no more logged regions in the
			 * buffer, then we're done.
			 */
			if (bit == -1) {
				return 0;
			}

			nbits = xfs_contig_bits(data_map, map_size,
							 bit);
			reg_buf_offset = bit << XFS_BLI_SHIFT;
			reg_buf_bytes = nbits << XFS_BLI_SHIFT;
			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.
		 */
		if (next_unlinked_offset < reg_buf_offset) {
			continue;
		}

		ASSERT(item->ri_buf[item_index].i_addr != NULL);
		ASSERT((item->ri_buf[item_index].i_len % XFS_BLI_CHUNK) == 0);
		ASSERT((reg_buf_offset + reg_buf_bytes) <= XFS_BUF_COUNT(bp));

		/*
		 * The current logged region contains a copy of the
		 * current di_next_unlinked field.  Extract its value
		 * and copy it to the buffer copy.
		 */
		logged_nextp = (xfs_agino_t *)
			       ((char *)(item->ri_buf[item_index].i_addr) +
				(next_unlinked_offset - reg_buf_offset));
1896
		if (unlikely(*logged_nextp == 0)) {
1897 1898 1899
			xfs_fs_cmn_err(CE_ALERT, mp,
				"bad inode buffer log record (ptr = 0x%p, bp = 0x%p).  XFS trying to replay bad (0) inode di_next_unlinked field",
				item, bp);
1900 1901
			XFS_ERROR_REPORT("xlog_recover_do_inode_buf",
					 XFS_ERRLEVEL_LOW, mp);
1902 1903 1904 1905 1906 1907 1908 1909 1910
			return XFS_ERROR(EFSCORRUPTED);
		}

		buffer_nextp = (xfs_agino_t *)xfs_buf_offset(bp,
					      next_unlinked_offset);
		INT_SET(*buffer_nextp, ARCH_CONVERT, *logged_nextp);
	}

	return 0;
1911
}
1912 1913

/*
Russell Cattelan's avatar
Russell Cattelan committed
1914
 * Perform a 'normal' buffer recovery.  Each logged region of the
1915 1916 1917 1918 1919 1920
 * 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.
 */
/*ARGSUSED*/
STATIC void
1921 1922 1923 1924 1925
xlog_recover_do_reg_buffer(
	xfs_mount_t		*mp,
	xlog_recover_item_t	*item,
	xfs_buf_t		*bp,
	xfs_buf_log_format_t	*buf_f)
1926 1927 1928 1929
{
	int			i;
	int			bit;
	int			nbits;
Russell Cattelan's avatar
Russell Cattelan committed
1930
	xfs_buf_log_format_v1_t	*obuf_f;
1931 1932
	unsigned int		*data_map = NULL;
	unsigned int		map_size = 0;
Russell Cattelan's avatar
Russell Cattelan committed
1933
	int                     error;
1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947

	switch (buf_f->blf_type) {
	case XFS_LI_BUF:
		data_map = buf_f->blf_data_map;
		map_size = buf_f->blf_map_size;
		break;
	case XFS_LI_6_1_BUF:
	case XFS_LI_5_3_BUF:
		obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
		data_map = obuf_f->blf_data_map;
		map_size = obuf_f->blf_map_size;
		break;
	}
	bit = 0;
Russell Cattelan's avatar
Russell Cattelan committed
1948
	i = 1;  /* 0 is the buf format structure */
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971
	while (1) {
		bit = xfs_next_bit(data_map, map_size, bit);
		if (bit == -1)
			break;
		nbits = xfs_contig_bits(data_map, map_size, bit);
		ASSERT(item->ri_buf[i].i_addr != 0);
		ASSERT(item->ri_buf[i].i_len % XFS_BLI_CHUNK == 0);
		ASSERT(XFS_BUF_COUNT(bp) >=
		       ((uint)bit << XFS_BLI_SHIFT)+(nbits<<XFS_BLI_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_BLI_UDQUOT_BUF|XFS_BLI_GDQUOT_BUF)) {
			error = xfs_qm_dqcheck((xfs_disk_dquot_t *)
					       item->ri_buf[i].i_addr,
					       -1, 0, XFS_QMOPT_DOWARN,
					       "dquot_buf_recover");
		}
		if (!error)
1972
			memcpy(xfs_buf_offset(bp,
1973 1974 1975
				(uint)bit << XFS_BLI_SHIFT),	/* dest */
				item->ri_buf[i].i_addr,		/* source */
				nbits<<XFS_BLI_SHIFT);		/* length */
1976 1977 1978 1979 1980 1981
		i++;
		bit += nbits;
	}

	/* Shouldn't be any more regions */
	ASSERT(i == item->ri_total);
1982
}
1983

1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
/*
 * Do some primitive error checking on ondisk dquot data structures.
 */
int
xfs_qm_dqcheck(
	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.
	 */
	if (INT_GET(ddq->d_magic, ARCH_CONVERT) != XFS_DQUOT_MAGIC) {
		if (flags & XFS_QMOPT_DOWARN)
			cmn_err(CE_ALERT,
			"%s : XFS dquot ID 0x%x, magic 0x%x != 0x%x",
			str, id,
			INT_GET(ddq->d_magic, ARCH_CONVERT), XFS_DQUOT_MAGIC);
		errs++;
	}
	if (INT_GET(ddq->d_version, ARCH_CONVERT) != XFS_DQUOT_VERSION) {
		if (flags & XFS_QMOPT_DOWARN)
			cmn_err(CE_ALERT,
			"%s : XFS dquot ID 0x%x, version 0x%x != 0x%x",
			str, id,
			INT_GET(ddq->d_magic, ARCH_CONVERT), XFS_DQUOT_VERSION);
		errs++;
	}

	if (INT_GET(ddq->d_flags, ARCH_CONVERT) != XFS_DQ_USER &&
	    INT_GET(ddq->d_flags, ARCH_CONVERT) != XFS_DQ_GROUP) {
		if (flags & XFS_QMOPT_DOWARN)
			cmn_err(CE_ALERT,
			"%s : XFS dquot ID 0x%x, unknown flags 0x%x",
			str, id, INT_GET(ddq->d_flags, ARCH_CONVERT));
		errs++;
	}

	if (id != -1 && id != INT_GET(ddq->d_id, ARCH_CONVERT)) {
		if (flags & XFS_QMOPT_DOWARN)
			cmn_err(CE_ALERT,
2042
			"%s : ondisk-dquot 0x%p, ID mismatch: "
2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
			"0x%x expected, found id 0x%x",
			str, ddq, id, INT_GET(ddq->d_id, ARCH_CONVERT));
		errs++;
	}

	if (! errs) {
		if (INT_GET(ddq->d_blk_softlimit, ARCH_CONVERT) &&
		    INT_GET(ddq->d_bcount, ARCH_CONVERT) >=
				INT_GET(ddq->d_blk_softlimit, ARCH_CONVERT)) {
			if (INT_ISZERO(ddq->d_btimer, ARCH_CONVERT) &&
			    !INT_ISZERO(ddq->d_id, ARCH_CONVERT)) {
				if (flags & XFS_QMOPT_DOWARN)
					cmn_err(CE_ALERT,
2056
					"%s : Dquot ID 0x%x (0x%p) "
2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069
					"BLK TIMER NOT STARTED",
					str, (int)
					INT_GET(ddq->d_id, ARCH_CONVERT), ddq);
				errs++;
			}
		}
		if (INT_GET(ddq->d_ino_softlimit, ARCH_CONVERT) &&
		    INT_GET(ddq->d_icount, ARCH_CONVERT) >=
				INT_GET(ddq->d_ino_softlimit, ARCH_CONVERT)) {
			if (INT_ISZERO(ddq->d_itimer, ARCH_CONVERT) &&
			    !INT_ISZERO(ddq->d_id, ARCH_CONVERT)) {
				if (flags & XFS_QMOPT_DOWARN)
					cmn_err(CE_ALERT,
2070
					"%s : Dquot ID 0x%x (0x%p) "
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
					"INODE TIMER NOT STARTED",
					str, (int)
					INT_GET(ddq->d_id, ARCH_CONVERT), ddq);
				errs++;
			}
		}
	}

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

	if (flags & XFS_QMOPT_DOWARN)
		cmn_err(CE_NOTE, "Re-initializing dquot ID 0x%x", id);

	/*
	 * Typically, a repair is only requested by quotacheck.
	 */
	ASSERT(id != -1);
	ASSERT(flags & XFS_QMOPT_DQREPAIR);
	memset(d, 0, sizeof(xfs_dqblk_t));
	INT_SET(d->dd_diskdq.d_magic, ARCH_CONVERT, XFS_DQUOT_MAGIC);
	INT_SET(d->dd_diskdq.d_version, ARCH_CONVERT, XFS_DQUOT_VERSION);
	INT_SET(d->dd_diskdq.d_id, ARCH_CONVERT, id);
	INT_SET(d->dd_diskdq.d_flags, ARCH_CONVERT, type);

	return errs;
}
2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112

/*
 * 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(
	xfs_mount_t		*mp,
	xlog_t			*log,
	xlog_recover_item_t	*item,
	xfs_buf_t		*bp,
	xfs_buf_log_format_t	*buf_f)
{
2113
	uint			type;
2114 2115

	/*
2116
	 * Filesystems are required to send in quota flags at mount time.
2117
	 */
2118
	if (mp->m_qflags == 0) {
2119 2120
		return;
	}
Russell Cattelan's avatar
Russell Cattelan committed
2121

2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159
	type = 0;
	if (buf_f->blf_flags & XFS_BLI_UDQUOT_BUF)
		type |= XFS_DQ_USER;
	if (buf_f->blf_flags & XFS_BLI_GDQUOT_BUF)
		type |= XFS_DQ_GROUP;
	/*
	 * This type of quotas was turned off, so ignore this buffer
	 */
	if (log->l_quotaoffs_flag & type)
		return;

	xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
}

/*
 * 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
 * with the XFS_BLI_CANCEL bit set to indicate that previous copies
 * 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
2160 2161 2162 2163
xlog_recover_do_buffer_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	int			pass)
2164 2165
{
	xfs_buf_log_format_t	*buf_f;
Russell Cattelan's avatar
Russell Cattelan committed
2166
	xfs_buf_log_format_v1_t	*obuf_f;
2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197
	xfs_mount_t		*mp;
	xfs_buf_t		*bp;
	int			error;
	int			cancel;
	xfs_daddr_t		blkno;
	int			len;
	ushort			flags;

	buf_f = (xfs_buf_log_format_t *)item->ri_buf[0].i_addr;

	if (pass == XLOG_RECOVER_PASS1) {
		/*
		 * In this pass we're only looking for buf items
		 * with the XFS_BLI_CANCEL bit set.
		 */
		xlog_recover_do_buffer_pass1(log, buf_f);
		return 0;
	} else {
		/*
		 * In this pass we want to recover all the buffers
		 * which have not been cancelled and are not
		 * cancellation buffers themselves.  The routine
		 * we call here will tell us whether or not to
		 * continue with the replay of this buffer.
		 */
		cancel = xlog_recover_do_buffer_pass2(log, buf_f);
		if (cancel) {
			return 0;
		}
	}
	switch (buf_f->blf_type) {
2198
	case XFS_LI_BUF:
2199 2200 2201 2202
		blkno = buf_f->blf_blkno;
		len = buf_f->blf_len;
		flags = buf_f->blf_flags;
		break;
2203 2204
	case XFS_LI_6_1_BUF:
	case XFS_LI_5_3_BUF:
2205 2206 2207 2208 2209
		obuf_f = (xfs_buf_log_format_v1_t*)buf_f;
		blkno = obuf_f->blf_blkno;
		len = obuf_f->blf_len;
		flags = obuf_f->blf_flags;
		break;
2210
	default:
2211
		xfs_fs_cmn_err(CE_ALERT, log->l_mp,
Stephen Lord's avatar
Stephen Lord committed
2212 2213
			"xfs_log_recover: unknown buffer type 0x%x, dev %s",
			buf_f->blf_type, XFS_BUFTARG_NAME(log->l_targ));
2214 2215
		XFS_ERROR_REPORT("xlog_recover_do_buffer_trans",
				 XFS_ERRLEVEL_LOW, log->l_mp);
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
		return XFS_ERROR(EFSCORRUPTED);
	}

	mp = log->l_mp;
	if (flags & XFS_BLI_INODE_BUF) {
		bp = xfs_buf_read_flags(mp->m_ddev_targp, blkno, len,
								XFS_BUF_LOCK);
	} else {
		bp = xfs_buf_read(mp->m_ddev_targp, blkno, len, 0);
	}
	if (XFS_BUF_ISERROR(bp)) {
		xfs_ioerror_alert("xlog_recover_do..(read#1)", log->l_mp,
				  bp, blkno);
		error = XFS_BUF_GETERROR(bp);
		xfs_buf_relse(bp);
		return error;
	}

	error = 0;
	if (flags & XFS_BLI_INODE_BUF) {
		error = xlog_recover_do_inode_buffer(mp, item, bp, buf_f);
	} else if (flags & (XFS_BLI_UDQUOT_BUF | XFS_BLI_GDQUOT_BUF)) {
		xlog_recover_do_dquot_buffer(mp, log, item, bp, buf_f);
	} else {
		xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
	}
	if (error)
		return XFS_ERROR(error);

	/*
Russell Cattelan's avatar
Russell Cattelan committed
2246
	 * Perform delayed write on the buffer.  Asynchronous writes will be
2247 2248 2249 2250 2251 2252 2253
	 * 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
Russell Cattelan's avatar
Russell Cattelan committed
2254
	 * running with a different inode cluster size.  Regardless, if the
2255 2256 2257 2258 2259
	 * 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.
	 */
2260 2261
	if (XFS_DINODE_MAGIC ==
	    INT_GET(*((__uint16_t *)(xfs_buf_offset(bp, 0))), ARCH_CONVERT) &&
2262 2263
	    (XFS_BUF_COUNT(bp) != MAX(log->l_mp->m_sb.sb_blocksize,
			(__uint32_t)XFS_INODE_CLUSTER_SIZE(log->l_mp)))) {
2264 2265
		XFS_BUF_STALE(bp);
		error = xfs_bwrite(mp, bp);
2266 2267 2268 2269 2270 2271 2272 2273 2274
	} else {
		ASSERT(XFS_BUF_FSPRIVATE(bp, void *) == NULL ||
		       XFS_BUF_FSPRIVATE(bp, xfs_mount_t *) == mp);
		XFS_BUF_SET_FSPRIVATE(bp, mp);
		XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
		xfs_bdwrite(mp, bp);
	}

	return (error);
2275
}
2276 2277

STATIC int
2278 2279 2280 2281
xlog_recover_do_inode_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	int			pass)
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
{
	xfs_inode_log_format_t	*in_f;
	xfs_mount_t		*mp;
	xfs_buf_t		*bp;
	xfs_imap_t		imap;
	xfs_dinode_t		*dip;
	xfs_ino_t		ino;
	int			len;
	xfs_caddr_t		src;
	xfs_caddr_t		dest;
	int			error;
	int			attr_index;
	uint			fields;
	xfs_dinode_core_t	*dicp;

	if (pass == XLOG_RECOVER_PASS1) {
		return 0;
	}

	in_f = (xfs_inode_log_format_t *)item->ri_buf[0].i_addr;
	ino = in_f->ilf_ino;
	mp = log->l_mp;
	if (ITEM_TYPE(item) == XFS_LI_INODE) {
		imap.im_blkno = (xfs_daddr_t)in_f->ilf_blkno;
		imap.im_len = in_f->ilf_len;
		imap.im_boffset = in_f->ilf_boffset;
	} else {
		/*
		 * It's an old inode format record.  We don't know where
		 * its cluster is located on disk, and we can't allow
		 * xfs_imap() to figure it out because the inode btrees
		 * are not ready to be used.  Therefore do not pass the
Russell Cattelan's avatar
Russell Cattelan committed
2314
		 * XFS_IMAP_LOOKUP flag to xfs_imap().  This will give
2315 2316 2317 2318 2319 2320 2321
		 * us only the single block in which the inode lives
		 * rather than its cluster, so we must make sure to
		 * invalidate the buffer when we write it out below.
		 */
		imap.im_blkno = 0;
		xfs_imap(log->l_mp, 0, ino, &imap, 0);
	}
2322 2323 2324 2325 2326 2327 2328 2329

	/*
	 * Inode buffers can be freed, look out for it,
	 * and do not replay the inode.
	 */
	if (xlog_check_buffer_cancelled(log, imap.im_blkno, imap.im_len, 0))
		return 0;

2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
	bp = xfs_buf_read_flags(mp->m_ddev_targp, imap.im_blkno, imap.im_len,
								XFS_BUF_LOCK);
	if (XFS_BUF_ISERROR(bp)) {
		xfs_ioerror_alert("xlog_recover_do..(read#2)", mp,
				  bp, imap.im_blkno);
		error = XFS_BUF_GETERROR(bp);
		xfs_buf_relse(bp);
		return error;
	}
	error = 0;
	ASSERT(in_f->ilf_fields & XFS_ILOG_CORE);
	dip = (xfs_dinode_t *)xfs_buf_offset(bp, imap.im_boffset);

	/*
	 * Make sure the place we're flushing out to really looks
	 * like an inode!
	 */
2347
	if (unlikely(INT_GET(dip->di_core.di_magic, ARCH_CONVERT) != XFS_DINODE_MAGIC)) {
2348 2349 2350 2351
		xfs_buf_relse(bp);
		xfs_fs_cmn_err(CE_ALERT, mp,
			"xfs_inode_recover: Bad inode magic number, dino ptr = 0x%p, dino bp = 0x%p, ino = %Ld",
			dip, bp, ino);
2352 2353
		XFS_ERROR_REPORT("xlog_recover_do_inode_trans(1)",
				 XFS_ERRLEVEL_LOW, mp);
2354 2355 2356
		return XFS_ERROR(EFSCORRUPTED);
	}
	dicp = (xfs_dinode_core_t*)(item->ri_buf[1].i_addr);
2357
	if (unlikely(dicp->di_magic != XFS_DINODE_MAGIC)) {
2358 2359 2360 2361
		xfs_buf_relse(bp);
		xfs_fs_cmn_err(CE_ALERT, mp,
			"xfs_inode_recover: Bad inode log record, rec ptr 0x%p, ino %Ld",
			item, ino);
2362 2363
		XFS_ERROR_REPORT("xlog_recover_do_inode_trans(2)",
				 XFS_ERRLEVEL_LOW, mp);
2364 2365
		return XFS_ERROR(EFSCORRUPTED);
	}
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385

	/* Skip replay when the on disk inode is newer than the log one */
	if (dicp->di_flushiter <
	    INT_GET(dip->di_core.di_flushiter, ARCH_CONVERT)) {
		/*
		 * Deal with the wrap case, DI_MAX_FLUSH is less
		 * than smaller numbers
		 */
		if ((INT_GET(dip->di_core.di_flushiter, ARCH_CONVERT)
							== DI_MAX_FLUSH) &&
		    (dicp->di_flushiter < (DI_MAX_FLUSH>>1))) {
			/* do nothing */
		} else {
			xfs_buf_relse(bp);
			return 0;
		}
	}
	/* Take the opportunity to reset the flush iteration count */
	dicp->di_flushiter = 0;

2386
	if (unlikely((dicp->di_mode & S_IFMT) == S_IFREG)) {
2387 2388
		if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (dicp->di_format != XFS_DINODE_FMT_BTREE)) {
2389 2390
			XFS_CORRUPTION_ERROR("xlog_recover_do_inode_trans(3)",
					 XFS_ERRLEVEL_LOW, mp, dicp);
2391 2392 2393 2394 2395 2396
			xfs_buf_relse(bp);
			xfs_fs_cmn_err(CE_ALERT, mp,
				"xfs_inode_recover: Bad regular inode log record, rec ptr 0x%p, ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
				item, dip, bp, ino);
			return XFS_ERROR(EFSCORRUPTED);
		}
2397
	} else if (unlikely((dicp->di_mode & S_IFMT) == S_IFDIR)) {
2398 2399 2400
		if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
		    (dicp->di_format != XFS_DINODE_FMT_BTREE) &&
		    (dicp->di_format != XFS_DINODE_FMT_LOCAL)) {
2401 2402
			XFS_CORRUPTION_ERROR("xlog_recover_do_inode_trans(4)",
					     XFS_ERRLEVEL_LOW, mp, dicp);
2403 2404 2405 2406 2407 2408 2409
			xfs_buf_relse(bp);
			xfs_fs_cmn_err(CE_ALERT, mp,
				"xfs_inode_recover: Bad dir inode log record, rec ptr 0x%p, ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
				item, dip, bp, ino);
			return XFS_ERROR(EFSCORRUPTED);
		}
	}
2410 2411 2412
	if (unlikely(dicp->di_nextents + dicp->di_anextents > dicp->di_nblocks)){
		XFS_CORRUPTION_ERROR("xlog_recover_do_inode_trans(5)",
				     XFS_ERRLEVEL_LOW, mp, dicp);
2413 2414 2415 2416 2417 2418 2419 2420
		xfs_buf_relse(bp);
		xfs_fs_cmn_err(CE_ALERT, mp,
			"xfs_inode_recover: Bad inode log record, rec ptr 0x%p, dino ptr 0x%p, dino bp 0x%p, ino %Ld, total extents = %d, nblocks = %Ld",
			item, dip, bp, ino,
			dicp->di_nextents + dicp->di_anextents,
			dicp->di_nblocks);
		return XFS_ERROR(EFSCORRUPTED);
	}
2421 2422 2423
	if (unlikely(dicp->di_forkoff > mp->m_sb.sb_inodesize)) {
		XFS_CORRUPTION_ERROR("xlog_recover_do_inode_trans(6)",
				     XFS_ERRLEVEL_LOW, mp, dicp);
2424 2425 2426 2427 2428 2429
		xfs_buf_relse(bp);
		xfs_fs_cmn_err(CE_ALERT, mp,
			"xfs_inode_recover: Bad inode log rec ptr 0x%p, dino ptr 0x%p, dino bp 0x%p, ino %Ld, forkoff 0x%x",
			item, dip, bp, ino, dicp->di_forkoff);
		return XFS_ERROR(EFSCORRUPTED);
	}
2430 2431 2432
	if (unlikely(item->ri_buf[1].i_len > sizeof(xfs_dinode_core_t))) {
		XFS_CORRUPTION_ERROR("xlog_recover_do_inode_trans(7)",
				     XFS_ERRLEVEL_LOW, mp, dicp);
2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445
		xfs_buf_relse(bp);
		xfs_fs_cmn_err(CE_ALERT, mp,
			"xfs_inode_recover: Bad inode log record length %d, rec ptr 0x%p",
			item->ri_buf[1].i_len, item);
		return XFS_ERROR(EFSCORRUPTED);
	}

	/* The core is in in-core format */
	xfs_xlate_dinode_core((xfs_caddr_t)&dip->di_core,
			      (xfs_dinode_core_t*)item->ri_buf[1].i_addr,
			      -1, ARCH_CONVERT);
	/* the rest is in on-disk format */
	if (item->ri_buf[1].i_len > sizeof(xfs_dinode_core_t)) {
2446 2447 2448
		memcpy((xfs_caddr_t) dip + sizeof(xfs_dinode_core_t),
			item->ri_buf[1].i_addr + sizeof(xfs_dinode_core_t),
			item->ri_buf[1].i_len  - sizeof(xfs_dinode_core_t));
2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473
	}

	fields = in_f->ilf_fields;
	switch (fields & (XFS_ILOG_DEV | XFS_ILOG_UUID)) {
	case XFS_ILOG_DEV:
		INT_SET(dip->di_u.di_dev, ARCH_CONVERT, in_f->ilf_u.ilfu_rdev);

		break;
	case XFS_ILOG_UUID:
		dip->di_u.di_muuid = in_f->ilf_u.ilfu_uuid;
		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:
2474
		memcpy(&dip->di_u, src, len);
2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
		break;

	case XFS_ILOG_DBROOT:
		xfs_bmbt_to_bmdr((xfs_bmbt_block_t *)src, len,
				 &(dip->di_u.di_bmbt),
				 XFS_DFORK_DSIZE(dip, mp));
		break;

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

	/*
Russell Cattelan's avatar
Russell Cattelan committed
2492
	 * If we logged any attribute data, recover it.  There may or
2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510
	 * 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));
2511
			memcpy(dest, src, len);
2512 2513 2514 2515 2516 2517 2518 2519 2520 2521
			break;

		case XFS_ILOG_ABROOT:
			dest = XFS_DFORK_APTR(dip);
			xfs_bmbt_to_bmdr((xfs_bmbt_block_t *)src, len,
					 (xfs_bmdr_block_t*)dest,
					 XFS_DFORK_ASIZE(dip, mp));
			break;

		default:
2522
			xlog_warn("XFS: xlog_recover_do_inode_trans: Invalid flag");
2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541
			ASSERT(0);
			xfs_buf_relse(bp);
			return XFS_ERROR(EIO);
		}
	}

write_inode_buffer:
	if (ITEM_TYPE(item) == XFS_LI_INODE) {
		ASSERT(XFS_BUF_FSPRIVATE(bp, void *) == NULL ||
		       XFS_BUF_FSPRIVATE(bp, xfs_mount_t *) == mp);
		XFS_BUF_SET_FSPRIVATE(bp, mp);
		XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
		xfs_bdwrite(mp, bp);
	} else {
		XFS_BUF_STALE(bp);
		error = xfs_bwrite(mp, bp);
	}

	return (error);
2542
}
2543 2544 2545 2546 2547 2548 2549

/*
 * Recover QUOTAOFF records. We simply make a note of it in the xlog_t
 * structure, so that we know not to do any dquot item or dquot buffer recovery,
 * of that type.
 */
STATIC int
2550 2551 2552 2553
xlog_recover_do_quotaoff_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	int			pass)
2554
{
2555
	xfs_qoff_logformat_t	*qoff_f;
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579

	if (pass == XLOG_RECOVER_PASS2) {
		return (0);
	}

	qoff_f = (xfs_qoff_logformat_t *)item->ri_buf[0].i_addr;
	ASSERT(qoff_f);

	/*
	 * The logitem format's flag tells us if this was user quotaoff,
	 * group quotaoff or both.
	 */
	if (qoff_f->qf_flags & XFS_UQUOTA_ACCT)
		log->l_quotaoffs_flag |= XFS_DQ_USER;
	if (qoff_f->qf_flags & XFS_GQUOTA_ACCT)
		log->l_quotaoffs_flag |= XFS_DQ_GROUP;

	return (0);
}

/*
 * Recover a dquot record
 */
STATIC int
2580 2581 2582 2583
xlog_recover_do_dquot_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	int			pass)
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597
{
	xfs_mount_t		*mp;
	xfs_buf_t		*bp;
	struct xfs_disk_dquot	*ddq, *recddq;
	int			error;
	xfs_dq_logformat_t	*dq_f;
	uint			type;

	if (pass == XLOG_RECOVER_PASS1) {
		return 0;
	}
	mp = log->l_mp;

	/*
2598
	 * Filesystems are required to send in quota flags at mount time.
2599
	 */
2600
	if (mp->m_qflags == 0)
2601 2602 2603 2604 2605 2606 2607
		return (0);

	recddq = (xfs_disk_dquot_t *)item->ri_buf[1].i_addr;
	ASSERT(recddq);
	/*
	 * This type of quotas was turned off, so ignore this record.
	 */
2608 2609
	type = INT_GET(recddq->d_flags, ARCH_CONVERT) &
			(XFS_DQ_USER | XFS_DQ_GROUP);
2610 2611 2612 2613 2614
	ASSERT(type);
	if (log->l_quotaoffs_flag & type)
		return (0);

	/*
2615 2616 2617 2618 2619
	 * 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.
2620 2621
	 *
	 * The other possibility, of course, is that the quota subsystem was
2622
	 * removed since the last mount - ENOSYS.
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 2649 2650 2651 2652 2653 2654 2655 2656
	 */
	dq_f = (xfs_dq_logformat_t *)item->ri_buf[0].i_addr;
	ASSERT(dq_f);
	if ((error = xfs_qm_dqcheck(recddq,
			   dq_f->qlf_id,
			   0, XFS_QMOPT_DOWARN,
			   "xlog_recover_do_dquot_trans (log copy)"))) {
		return XFS_ERROR(EIO);
	}
	ASSERT(dq_f->qlf_len == 1);

	error = xfs_read_buf(mp, mp->m_ddev_targp,
			     dq_f->qlf_blkno,
			     XFS_FSB_TO_BB(mp, dq_f->qlf_len),
			     0, &bp);
	if (error) {
		xfs_ioerror_alert("xlog_recover_do..(read#3)", mp,
				  bp, dq_f->qlf_blkno);
		return error;
	}
	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.
	 */
	if (xfs_qm_dqcheck(ddq, dq_f->qlf_id, 0, XFS_QMOPT_DOWARN,
			   "xlog_recover_do_dquot_trans")) {
		xfs_buf_relse(bp);
		return XFS_ERROR(EIO);
	}

2657
	memcpy(ddq, recddq, item->ri_buf[1].i_len);
2658 2659 2660 2661 2662 2663 2664 2665 2666

	ASSERT(dq_f->qlf_size == 2);
	ASSERT(XFS_BUF_FSPRIVATE(bp, void *) == NULL ||
	       XFS_BUF_FSPRIVATE(bp, xfs_mount_t *) == mp);
	XFS_BUF_SET_FSPRIVATE(bp, mp);
	XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
	xfs_bdwrite(mp, bp);

	return (0);
2667
}
2668 2669 2670 2671 2672 2673 2674 2675 2676

/*
 * 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.
 */
STATIC void
2677 2678 2679 2680 2681
xlog_recover_do_efi_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	xfs_lsn_t		lsn,
	int			pass)
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698
{
	xfs_mount_t		*mp;
	xfs_efi_log_item_t	*efip;
	xfs_efi_log_format_t	*efi_formatp;
	SPLDECL(s);

	if (pass == XLOG_RECOVER_PASS1) {
		return;
	}

	efi_formatp = (xfs_efi_log_format_t *)item->ri_buf[0].i_addr;
	ASSERT(item->ri_buf[0].i_len ==
	       (sizeof(xfs_efi_log_format_t) +
		((efi_formatp->efi_nextents - 1) * sizeof(xfs_extent_t))));

	mp = log->l_mp;
	efip = xfs_efi_init(mp, efi_formatp->efi_nextents);
2699
	memcpy((char *)&(efip->efi_format), (char *)efi_formatp,
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709
	      sizeof(xfs_efi_log_format_t) +
	      ((efi_formatp->efi_nextents - 1) * sizeof(xfs_extent_t)));
	efip->efi_next_extent = efi_formatp->efi_nextents;
	efip->efi_flags |= XFS_EFI_COMMITTED;

	AIL_LOCK(mp,s);
	/*
	 * xfs_trans_update_ail() drops the AIL lock.
	 */
	xfs_trans_update_ail(mp, (xfs_log_item_t *)efip, lsn, s);
2710
}
2711 2712 2713 2714


/*
 * This routine is called when an efd format structure is found in
Russell Cattelan's avatar
Russell Cattelan committed
2715
 * a committed transaction in the log.  It's purpose is to cancel
2716 2717 2718 2719 2720 2721
 * 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.
 */
STATIC void
2722 2723 2724 2725
xlog_recover_do_efd_trans(
	xlog_t			*log,
	xlog_recover_item_t	*item,
	int			pass)
2726 2727 2728
{
	xfs_mount_t		*mp;
	xfs_efd_log_format_t	*efd_formatp;
2729
	xfs_efi_log_item_t	*efip = NULL;
2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781
	xfs_log_item_t		*lip;
	int			gen;
	int			nexts;
	__uint64_t		efi_id;
	SPLDECL(s);

	if (pass == XLOG_RECOVER_PASS1) {
		return;
	}

	efd_formatp = (xfs_efd_log_format_t *)item->ri_buf[0].i_addr;
	ASSERT(item->ri_buf[0].i_len ==
	       (sizeof(xfs_efd_log_format_t) +
		((efd_formatp->efd_nextents - 1) * sizeof(xfs_extent_t))));
	efi_id = efd_formatp->efd_efi_id;

	/*
	 * Search for the efi with the id in the efd format structure
	 * in the AIL.
	 */
	mp = log->l_mp;
	AIL_LOCK(mp,s);
	lip = xfs_trans_first_ail(mp, &gen);
	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) {
				/*
				 * xfs_trans_delete_ail() drops the
				 * AIL lock.
				 */
				xfs_trans_delete_ail(mp, lip, s);
				break;
			}
		}
		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
	}
	if (lip == NULL) {
		AIL_UNLOCK(mp, s);
	}

	/*
	 * If we found it, then free it up.  If it wasn't there, it
	 * must have been overwritten in the log.  Oh well.
	 */
	if (lip != NULL) {
		nexts = efip->efi_format.efi_nextents;
		if (nexts > XFS_EFI_MAX_FAST_EXTENTS) {
			kmem_free(lip, sizeof(xfs_efi_log_item_t) +
				  ((nexts - 1) * sizeof(xfs_extent_t)));
		} else {
			kmem_zone_free(xfs_efi_zone, efip);
2782
		}
2783
	}
2784
}
2785 2786 2787 2788 2789 2790 2791 2792

/*
 * 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.
 */
STATIC int
2793 2794 2795 2796
xlog_recover_do_trans(
	xlog_t			*log,
	xlog_recover_t		*trans,
	int			pass)
2797
{
2798 2799
	int			error = 0;
	xlog_recover_item_t	*item, *first_item;
2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848

	if ((error = xlog_recover_reorder_trans(log, trans)))
		return error;
	first_item = item = trans->r_itemq;
	do {
		/*
		 * we don't need to worry about the block number being
		 * truncated in > 1 TB buffers because in user-land,
		 * we're now n32 or 64-bit so xfs_daddr_t is 64-bits so
		 * the blkno's will get through the user-mode buffer
		 * cache properly.  The only bad case is o32 kernels
		 * where xfs_daddr_t is 32-bits but mount will warn us
		 * off a > 1 TB filesystem before we get here.
		 */
		if ((ITEM_TYPE(item) == XFS_LI_BUF) ||
		    (ITEM_TYPE(item) == XFS_LI_6_1_BUF) ||
		    (ITEM_TYPE(item) == XFS_LI_5_3_BUF)) {
			if  ((error = xlog_recover_do_buffer_trans(log, item,
								 pass)))
				break;
		} else if ((ITEM_TYPE(item) == XFS_LI_INODE) ||
			   (ITEM_TYPE(item) == XFS_LI_6_1_INODE) ||
			   (ITEM_TYPE(item) == XFS_LI_5_3_INODE)) {
			if ((error = xlog_recover_do_inode_trans(log, item,
								pass)))
				break;
		} else if (ITEM_TYPE(item) == XFS_LI_EFI) {
			xlog_recover_do_efi_trans(log, item, trans->r_lsn,
						  pass);
		} else if (ITEM_TYPE(item) == XFS_LI_EFD) {
			xlog_recover_do_efd_trans(log, item, pass);
		} else if (ITEM_TYPE(item) == XFS_LI_DQUOT) {
			if ((error = xlog_recover_do_dquot_trans(log, item,
								   pass)))
					break;
		} else if ((ITEM_TYPE(item) == XFS_LI_QUOTAOFF)) {
			if ((error = xlog_recover_do_quotaoff_trans(log, item,
								   pass)))
					break;
		} else {
			xlog_warn("XFS: xlog_recover_do_trans");
			ASSERT(0);
			error = XFS_ERROR(EIO);
			break;
		}
		item = item->ri_next;
	} while (first_item != item);

	return error;
2849
}
2850 2851 2852 2853 2854 2855 2856

/*
 * Free up any resources allocated by the transaction
 *
 * Remember that EFIs, EFDs, and IUNLINKs are handled later.
 */
STATIC void
2857 2858
xlog_recover_free_trans(
	xlog_recover_t		*trans)
2859
{
2860 2861
	xlog_recover_item_t	*first_item, *item, *free_item;
	int			i;
2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878

	item = first_item = trans->r_itemq;
	do {
		free_item = item;
		item = item->ri_next;
		 /* Free the regions in the item. */
		for (i = 0; i < free_item->ri_cnt; i++) {
			kmem_free(free_item->ri_buf[i].i_addr,
				  free_item->ri_buf[i].i_len);
		}
		/* Free the item itself */
		kmem_free(free_item->ri_buf,
			  (free_item->ri_total * sizeof(xfs_log_iovec_t)));
		kmem_free(free_item, sizeof(xlog_recover_item_t));
	} while (first_item != item);
	/* Free the transaction recover structure */
	kmem_free(trans, sizeof(xlog_recover_t));
2879
}
2880 2881

STATIC int
2882 2883 2884 2885 2886
xlog_recover_commit_trans(
	xlog_t			*log,
	xlog_recover_t		**q,
	xlog_recover_t		*trans,
	int			pass)
2887
{
2888
	int			error;
2889 2890 2891 2892 2893 2894 2895

	if ((error = xlog_recover_unlink_tid(q, trans)))
		return error;
	if ((error = xlog_recover_do_trans(log, trans, pass)))
		return error;
	xlog_recover_free_trans(trans);			/* no error */
	return 0;
2896
}
2897 2898

STATIC int
2899 2900
xlog_recover_unmount_trans(
	xlog_recover_t		*trans)
2901 2902 2903
{
	/* Do nothing now */
	xlog_warn("XFS: xlog_recover_unmount_trans: Unmount LR");
2904 2905
	return 0;
}
2906 2907 2908

/*
 * There are two valid states of the r_state field.  0 indicates that the
Russell Cattelan's avatar
Russell Cattelan committed
2909
 * transaction structure is in a normal state.  We have either seen the
2910 2911 2912 2913 2914 2915 2916
 * 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
2917 2918 2919 2920 2921 2922
xlog_recover_process_data(
	xlog_t			*log,
	xlog_recover_t		*rhash[],
	xlog_rec_header_t	*rhead,
	xfs_caddr_t		dp,
	int			pass)
2923
{
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949
	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;

	lp = dp + INT_GET(rhead->h_len, ARCH_CONVERT);
	num_logops = INT_GET(rhead->h_num_logops, ARCH_CONVERT);

	/* 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) {
			xlog_warn(
		"XFS: xlog_recover_process_data: bad clientid");
			ASSERT(0);
			return (XFS_ERROR(EIO));
2950
		}
2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996
		tid = INT_GET(ohead->oh_tid, ARCH_CONVERT);
		hash = XLOG_RHASH(tid);
		trans = xlog_recover_find_tid(rhash[hash], tid);
		if (trans == NULL) {		   /* not found; add new tid */
			if (ohead->oh_flags & XLOG_START_TRANS)
				xlog_recover_new_tid(&rhash[hash], tid,
					INT_GET(rhead->h_lsn, ARCH_CONVERT));
		} else {
			ASSERT(dp+INT_GET(ohead->oh_len, ARCH_CONVERT) <= lp);
			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,
						&rhash[hash], trans, pass);
				break;
			case XLOG_UNMOUNT_TRANS:
				error = xlog_recover_unmount_trans(trans);
				break;
			case XLOG_WAS_CONT_TRANS:
				error = xlog_recover_add_to_cont_trans(trans,
						dp, INT_GET(ohead->oh_len,
							ARCH_CONVERT));
				break;
			case XLOG_START_TRANS:
				xlog_warn(
			"XFS: xlog_recover_process_data: bad transaction");
				ASSERT(0);
				error = XFS_ERROR(EIO);
				break;
			case 0:
			case XLOG_CONTINUE_TRANS:
				error = xlog_recover_add_to_trans(trans,
						dp, INT_GET(ohead->oh_len,
							ARCH_CONVERT));
				break;
			default:
				xlog_warn(
			"XFS: xlog_recover_process_data: bad flag");
				ASSERT(0);
				error = XFS_ERROR(EIO);
				break;
			}
			if (error)
				return error;
2997
		}
2998 2999 3000 3001 3002
		dp += INT_GET(ohead->oh_len, ARCH_CONVERT);
		num_logops--;
	}
	return 0;
}
3003 3004 3005 3006 3007 3008

/*
 * Process an extent free intent item that was recovered from
 * the log.  We need to free the extents that it describes.
 */
STATIC void
3009 3010 3011
xlog_recover_process_efi(
	xfs_mount_t		*mp,
	xfs_efi_log_item_t	*efip)
3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
{
	xfs_efd_log_item_t	*efdp;
	xfs_trans_t		*tp;
	int			i;
	xfs_extent_t		*extp;
	xfs_fsblock_t		startblock_fsb;

	ASSERT(!(efip->efi_flags & XFS_EFI_RECOVERED));

	/*
	 * First check the validity of the extents described by the
Russell Cattelan's avatar
Russell Cattelan committed
3023
	 * EFI.  If any are bad, then assume that all are bad and
3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055
	 * 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.
			 */
			xfs_efi_release(efip, efip->efi_format.efi_nextents);
			return;
		}
	}

	tp = xfs_trans_alloc(mp, 0);
	xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0, 0, 0);
	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]);
		xfs_free_extent(tp, extp->ext_start, extp->ext_len);
		xfs_trans_log_efd_extent(tp, efdp, extp->ext_start,
					 extp->ext_len);
	}

	efip->efi_flags |= XFS_EFI_RECOVERED;
	xfs_trans_commit(tp, 0, NULL);
3056
}
3057 3058 3059 3060 3061 3062 3063

/*
 * Verify that once we've encountered something other than an EFI
 * in the AIL that there are no more EFIs in the AIL.
 */
#if defined(DEBUG)
STATIC void
3064 3065 3066 3067
xlog_recover_check_ail(
	xfs_mount_t		*mp,
	xfs_log_item_t		*lip,
	int			gen)
3068
{
3069
	int			orig_gen = gen;
3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098

	do {
		ASSERT(lip->li_type != XFS_LI_EFI);
		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
		/*
		 * The check will be bogus if we restart from the
		 * beginning of the AIL, so ASSERT that we don't.
		 * We never should since we're holding the AIL lock
		 * the entire time.
		 */
		ASSERT(gen == orig_gen);
	} while (lip != NULL);
}
#endif	/* DEBUG */

/*
 * 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
Russell Cattelan's avatar
Russell Cattelan committed
3099
 * to the AIL.  Since everything added to the AIL must come after
3100 3101 3102 3103
 * everything already in the AIL, we stop processing as soon as
 * we see something other than an EFI in the AIL.
 */
STATIC void
3104 3105
xlog_recover_process_efis(
	xlog_t			*log)
3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140
{
	xfs_log_item_t		*lip;
	xfs_efi_log_item_t	*efip;
	int			gen;
	xfs_mount_t		*mp;
	SPLDECL(s);

	mp = log->l_mp;
	AIL_LOCK(mp,s);

	lip = xfs_trans_first_ail(mp, &gen);
	while (lip != NULL) {
		/*
		 * We're done when we see something other than an EFI.
		 */
		if (lip->li_type != XFS_LI_EFI) {
			xlog_recover_check_ail(mp, lip, gen);
			break;
		}

		/*
		 * Skip EFIs that we've already processed.
		 */
		efip = (xfs_efi_log_item_t *)lip;
		if (efip->efi_flags & XFS_EFI_RECOVERED) {
			lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
			continue;
		}

		AIL_UNLOCK(mp, s);
		xlog_recover_process_efi(mp, efip);
		AIL_LOCK(mp,s);
		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
	}
	AIL_UNLOCK(mp, s);
3141
}
3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161

/*
 * 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);
	xfs_trans_reserve(tp, 0, XFS_CLEAR_AGI_BUCKET_LOG_RES(mp), 0, 0, 0);

3162 3163 3164
	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
				   XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
				   XFS_FSS_TO_BB(mp, 1), 0, &agibp);
3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
	if (error) {
		xfs_trans_cancel(tp, XFS_TRANS_ABORT);
		return;
	}

	agi = XFS_BUF_TO_AGI(agibp);
	if (INT_GET(agi->agi_magicnum, ARCH_CONVERT) != XFS_AGI_MAGIC) {
		xfs_trans_cancel(tp, XFS_TRANS_ABORT);
		return;
	}
	ASSERT(INT_GET(agi->agi_magicnum, ARCH_CONVERT) == XFS_AGI_MAGIC);

	INT_SET(agi->agi_unlinked[bucket], ARCH_CONVERT, NULLAGINO);
	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));

	(void) xfs_trans_commit(tp, 0, NULL);
3184
}
3185 3186 3187 3188 3189 3190 3191 3192 3193

/*
 * 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
Russell Cattelan's avatar
Russell Cattelan committed
3194
 * lists when it has been fully truncated and is freed.  The
3195 3196 3197 3198
 * freeing of the inode and its removal from the list must be
 * atomic.
 */
void
3199 3200
xlog_recover_process_iunlinks(
	xlog_t		*log)
3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226
{
	xfs_mount_t	*mp;
	xfs_agnumber_t	agno;
	xfs_agi_t	*agi;
	xfs_buf_t	*agibp;
	xfs_buf_t	*ibp;
	xfs_dinode_t	*dip;
	xfs_inode_t	*ip;
	xfs_agino_t	agino;
	xfs_ino_t	ino;
	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.
		 */
3227 3228 3229
		agibp = xfs_buf_read(mp->m_ddev_targp,
				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
				XFS_FSS_TO_BB(mp, 1), 0);
3230
		if (XFS_BUF_ISERROR(agibp)) {
3231 3232 3233
			xfs_ioerror_alert("xlog_recover_process_iunlinks(#1)",
				log->l_mp, agibp,
				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)));
3234 3235
		}
		agi = XFS_BUF_TO_AGI(agibp);
3236 3237
		ASSERT(XFS_AGI_MAGIC ==
			INT_GET(agi->agi_magicnum, ARCH_CONVERT));
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275

		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) {

			agino = INT_GET(agi->agi_unlinked[bucket], ARCH_CONVERT);
			while (agino != NULLAGINO) {

				/*
				 * Release the agi buffer so that it can
				 * be acquired in the normal course of the
				 * transaction to truncate and free the inode.
				 */
				xfs_buf_relse(agibp);

				ino = XFS_AGINO_TO_INO(mp, agno, agino);
				error = xfs_iget(mp, NULL, ino, 0, &ip, 0);
				ASSERT(error || (ip != NULL));

				if (!error) {
					/*
					 * Get the on disk inode to find the
					 * next inode in the bucket.
					 */
					error = xfs_itobp(mp, NULL, ip, &dip,
							&ibp, 0);
					ASSERT(error || (dip != NULL));
				}

				if (!error) {
					ASSERT(ip->i_d.di_nlink == 0);

					/* setup for the next pass */
					agino = INT_GET(dip->di_next_unlinked,
							ARCH_CONVERT);
					xfs_buf_relse(ibp);
					/*
					 * Prevent any DMAPI event from
					 * being sent when the
					 * reference on the inode is
3276
					 * dropped.
Russell Cattelan's avatar
Russell Cattelan committed
3277
					 */
3278 3279 3280
					ip->i_d.di_dmevmask = 0;

					/*
3281 3282 3283
					 * If this is a new inode, handle
					 * it specially.  Otherwise,
					 * just drop our reference to the
3284 3285 3286 3287 3288 3289 3290
					 * inode.  If there are no
					 * other references, this will
					 * send the inode to
					 * xfs_inactive() which will
					 * truncate the file and free
					 * the inode.
					 */
3291 3292 3293 3294
					if (ip->i_d.di_mode == 0)
						xfs_iput_new(ip, 0);
					else
						VN_RELE(XFS_ITOV(ip));
3295 3296 3297 3298 3299
				} else {
					/*
					 * We can't read in the inode
					 * this bucket points to, or
					 * this inode is messed up.  Just
Russell Cattelan's avatar
Russell Cattelan committed
3300
					 * ditch this bucket of inodes.  We
3301
					 * will lose some inodes and space,
Russell Cattelan's avatar
Russell Cattelan committed
3302
					 * but at least we won't hang.  Call
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317
					 * 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);

					agino = NULLAGINO;
				}

				/*
				 * Reacquire the agibuffer and continue around
				 * the loop.
				 */
				agibp = xfs_buf_read(mp->m_ddev_targp,
3318 3319 3320
						XFS_AG_DADDR(mp, agno,
							XFS_AGI_DADDR(mp)),
						XFS_FSS_TO_BB(mp, 1), 0);
3321
				if (XFS_BUF_ISERROR(agibp)) {
3322 3323 3324 3325 3326
					xfs_ioerror_alert(
				"xlog_recover_process_iunlinks(#2)",
						log->l_mp, agibp,
						XFS_AG_DADDR(mp, agno,
							XFS_AGI_DADDR(mp)));
3327 3328
				}
				agi = XFS_BUF_TO_AGI(agibp);
3329 3330
				ASSERT(XFS_AGI_MAGIC == INT_GET(
					agi->agi_magicnum, ARCH_CONVERT));
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
			}
		}

		/*
		 * Release the buffer for the current agi so we can
		 * go on to the next one.
		 */
		xfs_buf_relse(agibp);
	}

	mp->m_dmevmask = mp_dmevmask;
3342
}
3343 3344 3345


#ifdef DEBUG
3346 3347 3348 3349 3350 3351 3352 3353 3354
STATIC void
xlog_pack_data_checksum(
	xlog_t		*log,
	xlog_in_core_t	*iclog,
	int		size)
{
	int		i;
	uint		*up;
	uint		chksum = 0;
3355 3356 3357

	up = (uint *)iclog->ic_datap;
	/* divide length by 4 to get # words */
3358
	for (i = 0; i < (size >> 2); i++) {
3359 3360 3361 3362
		chksum ^= INT_GET(*up, ARCH_CONVERT);
		up++;
	}
	INT_SET(iclog->ic_header.h_chksum, ARCH_CONVERT, chksum);
3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382
}
#else
#define xlog_pack_data_checksum(log, iclog, size)
#endif

/*
 * Stamp cycle number in every block
 */
void
xlog_pack_data(
	xlog_t			*log,
	xlog_in_core_t		*iclog)
{
	int			i, j, k;
	int			size = iclog->ic_offset + iclog->ic_roundoff;
	uint			cycle_lsn;
	xfs_caddr_t		dp;
	xlog_in_core_2_t	*xhdr;

	xlog_pack_data_checksum(log, iclog, size);
3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394

	cycle_lsn = CYCLE_LSN_NOCONV(iclog->ic_header.h_lsn, ARCH_CONVERT);

	dp = iclog->ic_datap;
	for (i = 0; i < BTOBB(size) &&
		i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
		iclog->ic_header.h_cycle_data[i] = *(uint *)dp;
		*(uint *)dp = cycle_lsn;
		dp += BBSIZE;
	}

	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
3395
		xhdr = (xlog_in_core_2_t *)&iclog->ic_header;
3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407
		for ( ; i < BTOBB(size); i++) {
			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
			xhdr[j].hic_xheader.xh_cycle_data[k] = *(uint *)dp;
			*(uint *)dp = cycle_lsn;
			dp += BBSIZE;
		}

		for (i = 1; i < log->l_iclog_heads; i++) {
			xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
		}
	}
3408
}
3409

3410
#if defined(DEBUG) && defined(XFS_LOUD_RECOVERY)
3411
STATIC void
3412 3413 3414 3415
xlog_unpack_data_checksum(
	xlog_rec_header_t	*rhead,
	xfs_caddr_t		dp,
	xlog_t			*log)
3416
{
3417 3418
	uint			*up = (uint *)dp;
	uint			chksum = 0;
3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434

	/* divide length by 4 to get # words */
	for (i=0; i < INT_GET(rhead->h_len, ARCH_CONVERT) >> 2; i++) {
		chksum ^= INT_GET(*up, ARCH_CONVERT);
		up++;
	}
	if (chksum != INT_GET(rhead->h_chksum, ARCH_CONVERT)) {
	    if (!INT_ISZERO(rhead->h_chksum, ARCH_CONVERT) ||
		((log->l_flags & XLOG_CHKSUM_MISMATCH) == 0)) {
		    cmn_err(CE_DEBUG,
			"XFS: LogR chksum mismatch: was (0x%x) is (0x%x)",
			    INT_GET(rhead->h_chksum, ARCH_CONVERT), chksum);
		    cmn_err(CE_DEBUG,
"XFS: Disregard message if filesystem was created with non-DEBUG kernel");
		    if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
			    cmn_err(CE_DEBUG,
3435
				"XFS: LogR this is a LogV2 filesystem");
3436 3437 3438 3439
		    }
		    log->l_flags |= XLOG_CHKSUM_MISMATCH;
	    }
	}
3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452
}
#else
#define xlog_unpack_data_checksum(rhead, dp, log)
#endif

STATIC void
xlog_unpack_data(
	xlog_rec_header_t	*rhead,
	xfs_caddr_t		dp,
	xlog_t			*log)
{
	int			i, j, k;
	xlog_in_core_2_t	*xhdr;
3453

3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471
	for (i = 0; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)) &&
		  i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
		*(uint *)dp = *(uint *)&rhead->h_cycle_data[i];
		dp += BBSIZE;
	}

	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
		xhdr = (xlog_in_core_2_t *)rhead;
		for ( ; i < BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT)); i++) {
			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
			*(uint *)dp = xhdr[j].hic_xheader.xh_cycle_data[k];
			dp += BBSIZE;
		}
	}

	xlog_unpack_data_checksum(rhead, dp, log);
}
3472

3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511
STATIC int
xlog_valid_rec_header(
	xlog_t			*log,
	xlog_rec_header_t	*rhead,
	xfs_daddr_t		blkno)
{
	int			bblks;

	if (unlikely(
	    (INT_GET(rhead->h_magicno, ARCH_CONVERT) !=
			XLOG_HEADER_MAGIC_NUM))) {
		XFS_ERROR_REPORT("xlog_valid_rec_header(1)",
				XFS_ERRLEVEL_LOW, log->l_mp);
		return XFS_ERROR(EFSCORRUPTED);
	}
	if (unlikely(
	    (INT_ISZERO(rhead->h_version, ARCH_CONVERT) ||
	    (INT_GET(rhead->h_version, ARCH_CONVERT) &
			(~XLOG_VERSION_OKBITS)) != 0))) {
		xlog_warn("XFS: %s: unrecognised log version (%d).",
			__FUNCTION__, INT_GET(rhead->h_version, ARCH_CONVERT));
		return XFS_ERROR(EIO);
	}

	/* LR body must have data or it wouldn't have been written */
	bblks = INT_GET(rhead->h_len, ARCH_CONVERT);
	if (unlikely( bblks <= 0 || bblks > 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;
}

3512 3513 3514 3515
/*
 * 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
Russell Cattelan's avatar
Russell Cattelan committed
3516
 * the physical log separately.  The pass parameter is passed through
3517 3518 3519 3520
 * to the routines called to process the data and is not looked at
 * here.
 */
STATIC int
3521 3522 3523 3524 3525
xlog_do_recovery_pass(
	xlog_t			*log,
	xfs_daddr_t		head_blk,
	xfs_daddr_t		tail_blk,
	int			pass)
3526
{
3527 3528 3529 3530 3531 3532 3533 3534 3535
	xlog_rec_header_t	*rhead;
	xfs_daddr_t		blk_no;
	xfs_caddr_t		bufaddr, offset;
	xfs_buf_t		*hbp, *dbp;
	int			error = 0, h_size;
	int			bblks, split_bblks;
	int			hblks, split_hblks, wrapped_hblks;
	xlog_recover_t		*rhash[XLOG_RHASH_SIZE];

3536 3537
	ASSERT(head_blk != tail_blk);

3538
	/*
3539 3540
	 * 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.
3541
	 */
3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554
	if (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb)) {
		/*
		 * 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;
		if ((error = xlog_bread(log, tail_blk, 1, hbp)))
			goto bread_err1;
		offset = xlog_align(log, tail_blk, 1, hbp);
		rhead = (xlog_rec_header_t *)offset;
3555 3556
		error = xlog_valid_rec_header(log, rhead, tail_blk);
		if (error)
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569
			goto bread_err1;
		h_size = INT_GET(rhead->h_size, ARCH_CONVERT);
		if ((INT_GET(rhead->h_version, ARCH_CONVERT)
				& XLOG_VERSION_2) &&
		    (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;
		}
3570
	} else {
3571 3572 3573 3574 3575
		ASSERT(log->l_sectbb_log == 0);
		hblks = 1;
		hbp = xlog_get_bp(log, 1);
		h_size = XLOG_BIG_RECORD_BSIZE;
	}
3576

3577 3578 3579 3580 3581 3582
	if (!hbp)
		return ENOMEM;
	dbp = xlog_get_bp(log, BTOBB(h_size));
	if (!dbp) {
		xlog_put_bp(hbp);
		return ENOMEM;
3583 3584
	}

3585 3586 3587 3588 3589 3590 3591
	memset(rhash, 0, sizeof(rhash));
	if (tail_blk <= head_blk) {
		for (blk_no = tail_blk; blk_no < head_blk; ) {
			if ((error = xlog_bread(log, blk_no, hblks, hbp)))
				goto bread_err2;
			offset = xlog_align(log, blk_no, hblks, hbp);
			rhead = (xlog_rec_header_t *)offset;
3592 3593
			error = xlog_valid_rec_header(log, rhead, blk_no);
			if (error)
3594 3595 3596 3597
				goto bread_err2;

			/* blocks in data section */
			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
3598 3599 3600 3601 3602 3603
			error = xlog_bread(log, blk_no + hblks, bblks, dbp);
			if (error)
				goto bread_err2;
			offset = xlog_align(log, blk_no + hblks, bblks, dbp);
			xlog_unpack_data(rhead, offset, log);
			if ((error = xlog_recover_process_data(log,
3604
						rhash, rhead, offset, pass)))
3605 3606
				goto bread_err2;
			blk_no += bblks + hblks;
3607
		}
3608
	} else {
3609
		/*
3610 3611 3612
		 * 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.
3613
		 */
3614 3615 3616 3617 3618
		blk_no = tail_blk;
		while (blk_no < log->l_logBBsize) {
			/*
			 * Check for header wrapping around physical end-of-log
			 */
3619 3620
			offset = NULL;
			split_hblks = 0;
3621
			wrapped_hblks = 0;
3622
			if (blk_no + hblks <= log->l_logBBsize) {
3623
				/* Read header in one read */
3624 3625
				error = xlog_bread(log, blk_no, hblks, hbp);
				if (error)
3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657
					goto bread_err2;
				offset = xlog_align(log, blk_no, hblks, hbp);
			} 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);
					if ((error = xlog_bread(log, blk_no,
							split_hblks, hbp)))
						goto bread_err2;
					offset = xlog_align(log, blk_no,
							split_hblks, hbp);
				}
				/*
				 * 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.
				 */
				bufaddr = XFS_BUF_PTR(hbp);
				XFS_BUF_SET_PTR(hbp,
						bufaddr + BBTOB(split_hblks),
						BBTOB(hblks - split_hblks));
				wrapped_hblks = hblks - split_hblks;
3658 3659
				error = xlog_bread(log, 0, wrapped_hblks, hbp);
				if (error)
3660 3661 3662 3663 3664 3665 3666
					goto bread_err2;
				XFS_BUF_SET_PTR(hbp, bufaddr, hblks);
				if (!offset)
					offset = xlog_align(log, 0,
							wrapped_hblks, hbp);
			}
			rhead = (xlog_rec_header_t *)offset;
3667 3668 3669
			error = xlog_valid_rec_header(log, rhead,
						split_hblks ? blk_no : 0);
			if (error)
3670
				goto bread_err2;
3671 3672 3673

			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
			blk_no += hblks;
3674 3675

			/* Read in data for log record */
3676 3677 3678
			if (blk_no + bblks <= log->l_logBBsize) {
				error = xlog_bread(log, blk_no, bblks, dbp);
				if (error)
3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726
					goto bread_err2;
				offset = xlog_align(log, blk_no, bblks, dbp);
			} else {
				/* This log record is split across the
				 * physical end of log */
				offset = NULL;
				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);
					if ((error = xlog_bread(log, blk_no,
							split_bblks, dbp)))
						goto bread_err2;
					offset = xlog_align(log, blk_no,
							split_bblks, dbp);
				}
				/*
				 * 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.
				 */
				bufaddr = XFS_BUF_PTR(dbp);
				XFS_BUF_SET_PTR(dbp,
						bufaddr + BBTOB(split_bblks),
						BBTOB(bblks - split_bblks));
				if ((error = xlog_bread(log, wrapped_hblks,
						bblks - split_bblks, dbp)))
					goto bread_err2;
				XFS_BUF_SET_PTR(dbp, bufaddr,
						XLOG_BIG_RECORD_BSIZE);
				if (!offset)
					offset = xlog_align(log, wrapped_hblks,
						bblks - split_bblks, dbp);
			}
			xlog_unpack_data(rhead, offset, log);
			if ((error = xlog_recover_process_data(log, rhash,
3727
							rhead, offset, pass)))
3728 3729 3730
				goto bread_err2;
			blk_no += bblks;
		}
3731

3732 3733 3734 3735 3736 3737 3738 3739 3740
		ASSERT(blk_no >= log->l_logBBsize);
		blk_no -= log->l_logBBsize;

		/* read first part of physical log */
		while (blk_no < head_blk) {
			if ((error = xlog_bread(log, blk_no, hblks, hbp)))
				goto bread_err2;
			offset = xlog_align(log, blk_no, hblks, hbp);
			rhead = (xlog_rec_header_t *)offset;
3741 3742 3743
			error = xlog_valid_rec_header(log, rhead, blk_no);
			if (error)
				goto bread_err2;
3744 3745 3746 3747 3748 3749
			bblks = (int)BTOBB(INT_GET(rhead->h_len, ARCH_CONVERT));
			if ((error = xlog_bread(log, blk_no+hblks, bblks, dbp)))
				goto bread_err2;
			offset = xlog_align(log, blk_no+hblks, bblks, dbp);
			xlog_unpack_data(rhead, offset, log);
			if ((error = xlog_recover_process_data(log, rhash,
3750
							rhead, offset, pass)))
3751
				goto bread_err2;
3752
			blk_no += bblks + hblks;
3753 3754
		}
	}
3755

3756 3757 3758 3759 3760
 bread_err2:
	xlog_put_bp(dbp);
 bread_err1:
	xlog_put_bp(hbp);
	return error;
3761 3762 3763
}

/*
Russell Cattelan's avatar
Russell Cattelan committed
3764
 * Do the recovery of the log.  We actually do this in two phases.
3765
 * The two passes are necessary in order to implement the function
Russell Cattelan's avatar
Russell Cattelan committed
3766
 * of cancelling a record written into the log.  The first pass
3767 3768
 * determines those things which have been cancelled, and the
 * second pass replays log items normally except for those which
Russell Cattelan's avatar
Russell Cattelan committed
3769
 * have been cancelled.  The handling of the replay and cancellations
3770 3771 3772 3773 3774 3775 3776
 * 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
3777 3778 3779 3780
xlog_do_log_recovery(
	xlog_t		*log,
	xfs_daddr_t	head_blk,
	xfs_daddr_t	tail_blk)
3781 3782
{
	int		error;
3783

3784
	ASSERT(head_blk != tail_blk);
3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807

	/*
	 * 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.
	 */
	log->l_buf_cancel_table =
		(xfs_buf_cancel_t **)kmem_zalloc(XLOG_BC_TABLE_SIZE *
						 sizeof(xfs_buf_cancel_t*),
						 KM_SLEEP);
	error = xlog_do_recovery_pass(log, head_blk, tail_blk,
				      XLOG_RECOVER_PASS1);
	if (error != 0) {
		kmem_free(log->l_buf_cancel_table,
			  XLOG_BC_TABLE_SIZE * sizeof(xfs_buf_cancel_t*));
		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);
3808 3809 3810 3811 3812 3813
#ifdef DEBUG
	{
		int	i;

		for (i = 0; i < XLOG_BC_TABLE_SIZE; i++)
			ASSERT(log->l_buf_cancel_table[i] == NULL);
3814 3815
	}
#endif	/* DEBUG */
3816

3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827
	kmem_free(log->l_buf_cancel_table,
		  XLOG_BC_TABLE_SIZE * sizeof(xfs_buf_cancel_t*));
	log->l_buf_cancel_table = NULL;

	return error;
}

/*
 * Do the actual recovery
 */
STATIC int
3828 3829 3830 3831
xlog_do_recover(
	xlog_t		*log,
	xfs_daddr_t	head_blk,
	xfs_daddr_t	tail_blk)
3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862
{
	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);
	if (error) {
		return error;
	}

	XFS_bflush(log->l_mp->m_ddev_targp);

	/*
	 * 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.
	 */
3863
	xlog_assign_tail_lsn(log->l_mp);
3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880

	/*
	 * Now that we've finished replaying all buffer and inode
	 * updates, re-read in the superblock.
	 */
	bp = xfs_getsb(log->l_mp, 0);
	XFS_BUF_UNDONE(bp);
	XFS_BUF_READ(bp);
	xfsbdstrat(log->l_mp, bp);
	if ((error = xfs_iowait(bp))) {
		xfs_ioerror_alert("xlog_do_recover",
				  log->l_mp, bp, XFS_BUF_ADDR(bp));
		ASSERT(0);
		xfs_buf_relse(bp);
		return error;
	}

3881 3882
	/* Convert superblock from on-disk format */
	sbp = &log->l_mp->m_sb;
3883 3884 3885 3886 3887 3888 3889 3890 3891 3892
	xfs_xlatesb(XFS_BUF_TO_SBP(bp), sbp, 1, ARCH_CONVERT, XFS_SB_ALL_BITS);
	ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
	ASSERT(XFS_SB_GOOD_VERSION(sbp));
	xfs_buf_relse(bp);

	xlog_recover_check_summary(log);

	/* Normal transactions can now occur */
	log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
	return 0;
3893
}
3894 3895 3896 3897 3898 3899 3900

/*
 * Perform recovery and re-initialize some log variables in xlog_find_tail.
 *
 * Return error or zero.
 */
int
3901 3902 3903
xlog_recover(
	xlog_t		*log,
	int		readonly)
3904
{
3905 3906
	xfs_daddr_t	head_blk, tail_blk;
	int		error;
3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919

	/* find the tail of the log */
	if ((error = xlog_find_tail(log, &head_blk, &tail_blk, readonly)))
		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
3920 3921 3922
		 * 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.
3923
		 */
3924 3925 3926 3927
		if ((error = xfs_dev_is_read_only(log->l_mp,
						"recovery required"))) {
			return error;
		}
3928 3929

		cmn_err(CE_NOTE,
Stephen Lord's avatar
Stephen Lord committed
3930 3931
			"Starting XFS recovery on filesystem: %s (dev: %s)",
			log->l_mp->m_fsname, XFS_BUFTARG_NAME(log->l_targ));
3932

3933 3934 3935 3936
		error = xlog_do_recover(log, head_blk, tail_blk);
		log->l_flags |= XLOG_RECOVERY_NEEDED;
	}
	return error;
3937
}
3938 3939 3940

/*
 * In the first part of recovery we replay inodes and buffers and build
Russell Cattelan's avatar
Russell Cattelan committed
3941
 * up the list of extent free items which need to be processed.  Here
3942
 * we process the extent free items and clean up the on disk unlinked
Russell Cattelan's avatar
Russell Cattelan committed
3943
 * inode lists.  This is separated from the first part of recovery so
3944 3945 3946 3947 3948
 * 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
3949 3950 3951
xlog_recover_finish(
	xlog_t		*log,
	int		mfsi_flags)
3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978
{
	/*
	 * 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) {
		xlog_recover_process_efis(log);
		/*
		 * 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.
		 */
		xfs_log_force(log->l_mp, (xfs_lsn_t)0,
			      (XFS_LOG_FORCE | XFS_LOG_SYNC));

		if ( (mfsi_flags & XFS_MFSI_NOUNLINK) == 0 ) {
			xlog_recover_process_iunlinks(log);
		}

		xlog_recover_check_summary(log);

		cmn_err(CE_NOTE,
3979
			"Ending XFS recovery on filesystem: %s (dev: %s)",
Stephen Lord's avatar
Stephen Lord committed
3980
			log->l_mp->m_fsname, XFS_BUFTARG_NAME(log->l_targ));
3981 3982 3983
		log->l_flags &= ~XLOG_RECOVERY_NEEDED;
	} else {
		cmn_err(CE_DEBUG,
3984
			"!Ending clean XFS mount for filesystem: %s",
3985 3986 3987
			log->l_mp->m_fsname);
	}
	return 0;
3988
}
3989 3990 3991 3992 3993 3994 3995 3996


#if defined(DEBUG)
/*
 * Read all of the agf and agi counters and check that they
 * are consistent with the superblock counters.
 */
void
3997 3998
xlog_recover_check_summary(
	xlog_t		*log)
3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021
{
	xfs_mount_t	*mp;
	xfs_agf_t	*agfp;
	xfs_agi_t	*agip;
	xfs_buf_t	*agfbp;
	xfs_buf_t	*agibp;
	xfs_daddr_t	agfdaddr;
	xfs_daddr_t	agidaddr;
	xfs_buf_t	*sbbp;
#ifdef XFS_LOUD_RECOVERY
	xfs_sb_t	*sbp;
#endif
	xfs_agnumber_t	agno;
	__uint64_t	freeblks;
	__uint64_t	itotal;
	__uint64_t	ifree;

	mp = log->l_mp;

	freeblks = 0LL;
	itotal = 0LL;
	ifree = 0LL;
	for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
4022 4023 4024
		agfdaddr = XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp));
		agfbp = xfs_buf_read(mp->m_ddev_targp, agfdaddr,
				XFS_FSS_TO_BB(mp, 1), 0);
4025 4026
		if (XFS_BUF_ISERROR(agfbp)) {
			xfs_ioerror_alert("xlog_recover_check_summary(agf)",
4027
						mp, agfbp, agfdaddr);
4028 4029
		}
		agfp = XFS_BUF_TO_AGF(agfbp);
4030 4031 4032 4033
		ASSERT(XFS_AGF_MAGIC ==
			INT_GET(agfp->agf_magicnum, ARCH_CONVERT));
		ASSERT(XFS_AGF_GOOD_VERSION(
			INT_GET(agfp->agf_versionnum, ARCH_CONVERT)));
4034 4035 4036 4037 4038 4039
		ASSERT(INT_GET(agfp->agf_seqno, ARCH_CONVERT) == agno);

		freeblks += INT_GET(agfp->agf_freeblks, ARCH_CONVERT) +
			    INT_GET(agfp->agf_flcount, ARCH_CONVERT);
		xfs_buf_relse(agfbp);

4040 4041 4042
		agidaddr = XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp));
		agibp = xfs_buf_read(mp->m_ddev_targp, agidaddr,
				XFS_FSS_TO_BB(mp, 1), 0);
4043 4044 4045 4046 4047
		if (XFS_BUF_ISERROR(agibp)) {
			xfs_ioerror_alert("xlog_recover_check_summary(agi)",
					  log->l_mp, agibp, agidaddr);
		}
		agip = XFS_BUF_TO_AGI(agibp);
4048 4049 4050 4051
		ASSERT(XFS_AGI_MAGIC ==
			INT_GET(agip->agi_magicnum, ARCH_CONVERT));
		ASSERT(XFS_AGI_GOOD_VERSION(
			INT_GET(agip->agi_versionnum, ARCH_CONVERT)));
4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083
		ASSERT(INT_GET(agip->agi_seqno, ARCH_CONVERT) == agno);

		itotal += INT_GET(agip->agi_count, ARCH_CONVERT);
		ifree += INT_GET(agip->agi_freecount, ARCH_CONVERT);
		xfs_buf_relse(agibp);
	}

	sbbp = xfs_getsb(mp, 0);
#ifdef XFS_LOUD_RECOVERY
	sbp = XFS_BUF_TO_SBP(sbbp);
	cmn_err(CE_NOTE,
		"xlog_recover_check_summary: sb_icount %Lu itotal %Lu",
		sbp->sb_icount, itotal);
	cmn_err(CE_NOTE,
		"xlog_recover_check_summary: sb_ifree %Lu itotal %Lu",
		sbp->sb_ifree, ifree);
	cmn_err(CE_NOTE,
		"xlog_recover_check_summary: sb_fdblocks %Lu freeblks %Lu",
		sbp->sb_fdblocks, freeblks);
#if 0
	/*
	 * This is turned off until I account for the allocation
	 * btree blocks which live in free space.
	 */
	ASSERT(sbp->sb_icount == itotal);
	ASSERT(sbp->sb_ifree == ifree);
	ASSERT(sbp->sb_fdblocks == freeblks);
#endif
#endif
	xfs_buf_relse(sbbp);
}
#endif /* DEBUG */