1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_types.h"
21#include "xfs_bit.h"
22#include "xfs_log.h"
23#include "xfs_inum.h"
24#include "xfs_trans.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_mount.h"
28#include "xfs_error.h"
29#include "xfs_bmap_btree.h"
30#include "xfs_alloc_btree.h"
31#include "xfs_ialloc_btree.h"
32#include "xfs_dinode.h"
33#include "xfs_inode.h"
34#include "xfs_inode_item.h"
35#include "xfs_alloc.h"
36#include "xfs_ialloc.h"
37#include "xfs_log_priv.h"
38#include "xfs_buf_item.h"
39#include "xfs_log_recover.h"
40#include "xfs_extfree_item.h"
41#include "xfs_trans_priv.h"
42#include "xfs_quota.h"
43#include "xfs_rw.h"
44#include "xfs_utils.h"
45#include "xfs_trace.h"
46
47STATIC int xlog_find_zeroed(xlog_t *, xfs_daddr_t *);
48STATIC int xlog_clear_stale_blocks(xlog_t *, xfs_lsn_t);
49#if defined(DEBUG)
50STATIC void xlog_recover_check_summary(xlog_t *);
51#else
52#define xlog_recover_check_summary(log)
53#endif
54
55
56
57
58
59struct xfs_buf_cancel {
60 xfs_daddr_t bc_blkno;
61 uint bc_len;
62 int bc_refcount;
63 struct list_head bc_list;
64};
65
66
67
68
69
70
71
72
73
74
75
76static inline int
77xlog_buf_bbcount_valid(
78 xlog_t *log,
79 int bbcount)
80{
81 return bbcount > 0 && bbcount <= log->l_logBBsize;
82}
83
84
85
86
87
88
89STATIC xfs_buf_t *
90xlog_get_bp(
91 xlog_t *log,
92 int nbblks)
93{
94 if (!xlog_buf_bbcount_valid(log, nbblks)) {
95 xlog_warn("XFS: Invalid block length (0x%x) given for buffer",
96 nbblks);
97 XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
98 return NULL;
99 }
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117 if (nbblks > 1 && log->l_sectBBsize > 1)
118 nbblks += log->l_sectBBsize;
119 nbblks = round_up(nbblks, log->l_sectBBsize);
120
121 return xfs_buf_get_uncached(log->l_mp->m_logdev_targp,
122 BBTOB(nbblks), 0);
123}
124
125STATIC void
126xlog_put_bp(
127 xfs_buf_t *bp)
128{
129 xfs_buf_free(bp);
130}
131
132
133
134
135
136STATIC xfs_caddr_t
137xlog_align(
138 xlog_t *log,
139 xfs_daddr_t blk_no,
140 int nbblks,
141 xfs_buf_t *bp)
142{
143 xfs_daddr_t offset = blk_no & ((xfs_daddr_t)log->l_sectBBsize - 1);
144
145 ASSERT(BBTOB(offset + nbblks) <= XFS_BUF_SIZE(bp));
146 return XFS_BUF_PTR(bp) + BBTOB(offset);
147}
148
149
150
151
152
153STATIC int
154xlog_bread_noalign(
155 xlog_t *log,
156 xfs_daddr_t blk_no,
157 int nbblks,
158 xfs_buf_t *bp)
159{
160 int error;
161
162 if (!xlog_buf_bbcount_valid(log, nbblks)) {
163 xlog_warn("XFS: Invalid block length (0x%x) given for buffer",
164 nbblks);
165 XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
166 return EFSCORRUPTED;
167 }
168
169 blk_no = round_down(blk_no, log->l_sectBBsize);
170 nbblks = round_up(nbblks, log->l_sectBBsize);
171
172 ASSERT(nbblks > 0);
173 ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));
174
175 XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
176 XFS_BUF_READ(bp);
177 XFS_BUF_BUSY(bp);
178 XFS_BUF_SET_COUNT(bp, BBTOB(nbblks));
179 XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);
180
181 xfsbdstrat(log->l_mp, bp);
182 error = xfs_buf_iowait(bp);
183 if (error)
184 xfs_ioerror_alert("xlog_bread", log->l_mp,
185 bp, XFS_BUF_ADDR(bp));
186 return error;
187}
188
189STATIC int
190xlog_bread(
191 xlog_t *log,
192 xfs_daddr_t blk_no,
193 int nbblks,
194 xfs_buf_t *bp,
195 xfs_caddr_t *offset)
196{
197 int error;
198
199 error = xlog_bread_noalign(log, blk_no, nbblks, bp);
200 if (error)
201 return error;
202
203 *offset = xlog_align(log, blk_no, nbblks, bp);
204 return 0;
205}
206
207
208
209
210
211
212STATIC int
213xlog_bwrite(
214 xlog_t *log,
215 xfs_daddr_t blk_no,
216 int nbblks,
217 xfs_buf_t *bp)
218{
219 int error;
220
221 if (!xlog_buf_bbcount_valid(log, nbblks)) {
222 xlog_warn("XFS: Invalid block length (0x%x) given for buffer",
223 nbblks);
224 XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_HIGH, log->l_mp);
225 return EFSCORRUPTED;
226 }
227
228 blk_no = round_down(blk_no, log->l_sectBBsize);
229 nbblks = round_up(nbblks, log->l_sectBBsize);
230
231 ASSERT(nbblks > 0);
232 ASSERT(BBTOB(nbblks) <= XFS_BUF_SIZE(bp));
233
234 XFS_BUF_SET_ADDR(bp, log->l_logBBstart + blk_no);
235 XFS_BUF_ZEROFLAGS(bp);
236 XFS_BUF_BUSY(bp);
237 XFS_BUF_HOLD(bp);
238 XFS_BUF_PSEMA(bp, PRIBIO);
239 XFS_BUF_SET_COUNT(bp, BBTOB(nbblks));
240 XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);
241
242 if ((error = xfs_bwrite(log->l_mp, bp)))
243 xfs_ioerror_alert("xlog_bwrite", log->l_mp,
244 bp, XFS_BUF_ADDR(bp));
245 return error;
246}
247
248#ifdef DEBUG
249
250
251
252STATIC void
253xlog_header_check_dump(
254 xfs_mount_t *mp,
255 xlog_rec_header_t *head)
256{
257 cmn_err(CE_DEBUG, "%s: SB : uuid = %pU, fmt = %d\n",
258 __func__, &mp->m_sb.sb_uuid, XLOG_FMT);
259 cmn_err(CE_DEBUG, " log : uuid = %pU, fmt = %d\n",
260 &head->h_fs_uuid, be32_to_cpu(head->h_fmt));
261}
262#else
263#define xlog_header_check_dump(mp, head)
264#endif
265
266
267
268
269STATIC int
270xlog_header_check_recover(
271 xfs_mount_t *mp,
272 xlog_rec_header_t *head)
273{
274 ASSERT(be32_to_cpu(head->h_magicno) == XLOG_HEADER_MAGIC_NUM);
275
276
277
278
279
280
281 if (unlikely(be32_to_cpu(head->h_fmt) != XLOG_FMT)) {
282 xlog_warn(
283 "XFS: dirty log written in incompatible format - can't recover");
284 xlog_header_check_dump(mp, head);
285 XFS_ERROR_REPORT("xlog_header_check_recover(1)",
286 XFS_ERRLEVEL_HIGH, mp);
287 return XFS_ERROR(EFSCORRUPTED);
288 } else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
289 xlog_warn(
290 "XFS: dirty log entry has mismatched uuid - can't recover");
291 xlog_header_check_dump(mp, head);
292 XFS_ERROR_REPORT("xlog_header_check_recover(2)",
293 XFS_ERRLEVEL_HIGH, mp);
294 return XFS_ERROR(EFSCORRUPTED);
295 }
296 return 0;
297}
298
299
300
301
302STATIC int
303xlog_header_check_mount(
304 xfs_mount_t *mp,
305 xlog_rec_header_t *head)
306{
307 ASSERT(be32_to_cpu(head->h_magicno) == XLOG_HEADER_MAGIC_NUM);
308
309 if (uuid_is_nil(&head->h_fs_uuid)) {
310
311
312
313
314
315 xlog_warn("XFS: nil uuid in log - IRIX style log");
316 } else if (unlikely(!uuid_equal(&mp->m_sb.sb_uuid, &head->h_fs_uuid))) {
317 xlog_warn("XFS: log has mismatched uuid - can't recover");
318 xlog_header_check_dump(mp, head);
319 XFS_ERROR_REPORT("xlog_header_check_mount",
320 XFS_ERRLEVEL_HIGH, mp);
321 return XFS_ERROR(EFSCORRUPTED);
322 }
323 return 0;
324}
325
326STATIC void
327xlog_recover_iodone(
328 struct xfs_buf *bp)
329{
330 if (XFS_BUF_GETERROR(bp)) {
331
332
333
334
335 xfs_ioerror_alert("xlog_recover_iodone",
336 bp->b_target->bt_mount, bp,
337 XFS_BUF_ADDR(bp));
338 xfs_force_shutdown(bp->b_target->bt_mount,
339 SHUTDOWN_META_IO_ERROR);
340 }
341 XFS_BUF_CLR_IODONE_FUNC(bp);
342 xfs_buf_ioend(bp, 0);
343}
344
345
346
347
348
349
350
351STATIC int
352xlog_find_cycle_start(
353 xlog_t *log,
354 xfs_buf_t *bp,
355 xfs_daddr_t first_blk,
356 xfs_daddr_t *last_blk,
357 uint cycle)
358{
359 xfs_caddr_t offset;
360 xfs_daddr_t mid_blk;
361 xfs_daddr_t end_blk;
362 uint mid_cycle;
363 int error;
364
365 end_blk = *last_blk;
366 mid_blk = BLK_AVG(first_blk, end_blk);
367 while (mid_blk != first_blk && mid_blk != end_blk) {
368 error = xlog_bread(log, mid_blk, 1, bp, &offset);
369 if (error)
370 return error;
371 mid_cycle = xlog_get_cycle(offset);
372 if (mid_cycle == cycle)
373 end_blk = mid_blk;
374 else
375 first_blk = mid_blk;
376 mid_blk = BLK_AVG(first_blk, end_blk);
377 }
378 ASSERT((mid_blk == first_blk && mid_blk+1 == end_blk) ||
379 (mid_blk == end_blk && mid_blk-1 == first_blk));
380
381 *last_blk = end_blk;
382
383 return 0;
384}
385
386
387
388
389
390
391
392
393
394STATIC int
395xlog_find_verify_cycle(
396 xlog_t *log,
397 xfs_daddr_t start_blk,
398 int nbblks,
399 uint stop_on_cycle_no,
400 xfs_daddr_t *new_blk)
401{
402 xfs_daddr_t i, j;
403 uint cycle;
404 xfs_buf_t *bp;
405 xfs_daddr_t bufblks;
406 xfs_caddr_t buf = NULL;
407 int error = 0;
408
409
410
411
412
413
414
415 bufblks = 1 << ffs(nbblks);
416 while (!(bp = xlog_get_bp(log, bufblks))) {
417 bufblks >>= 1;
418 if (bufblks < log->l_sectBBsize)
419 return ENOMEM;
420 }
421
422 for (i = start_blk; i < start_blk + nbblks; i += bufblks) {
423 int bcount;
424
425 bcount = min(bufblks, (start_blk + nbblks - i));
426
427 error = xlog_bread(log, i, bcount, bp, &buf);
428 if (error)
429 goto out;
430
431 for (j = 0; j < bcount; j++) {
432 cycle = xlog_get_cycle(buf);
433 if (cycle == stop_on_cycle_no) {
434 *new_blk = i+j;
435 goto out;
436 }
437
438 buf += BBSIZE;
439 }
440 }
441
442 *new_blk = -1;
443
444out:
445 xlog_put_bp(bp);
446 return error;
447}
448
449
450
451
452
453
454
455
456
457
458
459
460
461STATIC int
462xlog_find_verify_log_record(
463 xlog_t *log,
464 xfs_daddr_t start_blk,
465 xfs_daddr_t *last_blk,
466 int extra_bblks)
467{
468 xfs_daddr_t i;
469 xfs_buf_t *bp;
470 xfs_caddr_t offset = NULL;
471 xlog_rec_header_t *head = NULL;
472 int error = 0;
473 int smallmem = 0;
474 int num_blks = *last_blk - start_blk;
475 int xhdrs;
476
477 ASSERT(start_blk != 0 || *last_blk != start_blk);
478
479 if (!(bp = xlog_get_bp(log, num_blks))) {
480 if (!(bp = xlog_get_bp(log, 1)))
481 return ENOMEM;
482 smallmem = 1;
483 } else {
484 error = xlog_bread(log, start_blk, num_blks, bp, &offset);
485 if (error)
486 goto out;
487 offset += ((num_blks - 1) << BBSHIFT);
488 }
489
490 for (i = (*last_blk) - 1; i >= 0; i--) {
491 if (i < start_blk) {
492
493 xlog_warn(
494 "XFS: Log inconsistent (didn't find previous header)");
495 ASSERT(0);
496 error = XFS_ERROR(EIO);
497 goto out;
498 }
499
500 if (smallmem) {
501 error = xlog_bread(log, i, 1, bp, &offset);
502 if (error)
503 goto out;
504 }
505
506 head = (xlog_rec_header_t *)offset;
507
508 if (XLOG_HEADER_MAGIC_NUM == be32_to_cpu(head->h_magicno))
509 break;
510
511 if (!smallmem)
512 offset -= BBSIZE;
513 }
514
515
516
517
518
519
520 if (i == -1) {
521 error = -1;
522 goto out;
523 }
524
525
526
527
528
529 if ((error = xlog_header_check_mount(log->l_mp, head)))
530 goto out;
531
532
533
534
535
536
537
538
539 if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
540 uint h_size = be32_to_cpu(head->h_size);
541
542 xhdrs = h_size / XLOG_HEADER_CYCLE_SIZE;
543 if (h_size % XLOG_HEADER_CYCLE_SIZE)
544 xhdrs++;
545 } else {
546 xhdrs = 1;
547 }
548
549 if (*last_blk - i + extra_bblks !=
550 BTOBB(be32_to_cpu(head->h_len)) + xhdrs)
551 *last_blk = i;
552
553out:
554 xlog_put_bp(bp);
555 return error;
556}
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571STATIC int
572xlog_find_head(
573 xlog_t *log,
574 xfs_daddr_t *return_head_blk)
575{
576 xfs_buf_t *bp;
577 xfs_caddr_t offset;
578 xfs_daddr_t new_blk, first_blk, start_blk, last_blk, head_blk;
579 int num_scan_bblks;
580 uint first_half_cycle, last_half_cycle;
581 uint stop_on_cycle;
582 int error, log_bbnum = log->l_logBBsize;
583
584
585 if ((error = xlog_find_zeroed(log, &first_blk)) == -1) {
586 *return_head_blk = first_blk;
587
588
589 if (!first_blk) {
590
591
592
593
594 xlog_warn("XFS: totally zeroed log");
595 }
596
597 return 0;
598 } else if (error) {
599 xlog_warn("XFS: empty log check failed");
600 return error;
601 }
602
603 first_blk = 0;
604 bp = xlog_get_bp(log, 1);
605 if (!bp)
606 return ENOMEM;
607
608 error = xlog_bread(log, 0, 1, bp, &offset);
609 if (error)
610 goto bp_err;
611
612 first_half_cycle = xlog_get_cycle(offset);
613
614 last_blk = head_blk = log_bbnum - 1;
615 error = xlog_bread(log, last_blk, 1, bp, &offset);
616 if (error)
617 goto bp_err;
618
619 last_half_cycle = xlog_get_cycle(offset);
620 ASSERT(last_half_cycle != 0);
621
622
623
624
625
626
627
628
629
630
631
632
633 if (first_half_cycle == last_half_cycle) {
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659 head_blk = log_bbnum;
660 stop_on_cycle = last_half_cycle - 1;
661 } else {
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684 stop_on_cycle = last_half_cycle;
685 if ((error = xlog_find_cycle_start(log, bp, first_blk,
686 &head_blk, last_half_cycle)))
687 goto bp_err;
688 }
689
690
691
692
693
694
695
696
697 num_scan_bblks = XLOG_TOTAL_REC_SHIFT(log);
698 if (head_blk >= num_scan_bblks) {
699
700
701
702
703 start_blk = head_blk - num_scan_bblks;
704 if ((error = xlog_find_verify_cycle(log,
705 start_blk, num_scan_bblks,
706 stop_on_cycle, &new_blk)))
707 goto bp_err;
708 if (new_blk != -1)
709 head_blk = new_blk;
710 } else {
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738 ASSERT(head_blk <= INT_MAX &&
739 (xfs_daddr_t) num_scan_bblks >= head_blk);
740 start_blk = log_bbnum - (num_scan_bblks - head_blk);
741 if ((error = xlog_find_verify_cycle(log, start_blk,
742 num_scan_bblks - (int)head_blk,
743 (stop_on_cycle - 1), &new_blk)))
744 goto bp_err;
745 if (new_blk != -1) {
746 head_blk = new_blk;
747 goto validate_head;
748 }
749
750
751
752
753
754
755 start_blk = 0;
756 ASSERT(head_blk <= INT_MAX);
757 if ((error = xlog_find_verify_cycle(log,
758 start_blk, (int)head_blk,
759 stop_on_cycle, &new_blk)))
760 goto bp_err;
761 if (new_blk != -1)
762 head_blk = new_blk;
763 }
764
765validate_head:
766
767
768
769
770 num_scan_bblks = XLOG_REC_SHIFT(log);
771 if (head_blk >= num_scan_bblks) {
772 start_blk = head_blk - num_scan_bblks;
773
774
775 if ((error = xlog_find_verify_log_record(log, start_blk,
776 &head_blk, 0)) == -1) {
777 error = XFS_ERROR(EIO);
778 goto bp_err;
779 } else if (error)
780 goto bp_err;
781 } else {
782 start_blk = 0;
783 ASSERT(head_blk <= INT_MAX);
784 if ((error = xlog_find_verify_log_record(log, start_blk,
785 &head_blk, 0)) == -1) {
786
787 start_blk = log_bbnum - (num_scan_bblks - head_blk);
788 new_blk = log_bbnum;
789 ASSERT(start_blk <= INT_MAX &&
790 (xfs_daddr_t) log_bbnum-start_blk >= 0);
791 ASSERT(head_blk <= INT_MAX);
792 if ((error = xlog_find_verify_log_record(log,
793 start_blk, &new_blk,
794 (int)head_blk)) == -1) {
795 error = XFS_ERROR(EIO);
796 goto bp_err;
797 } else if (error)
798 goto bp_err;
799 if (new_blk != log_bbnum)
800 head_blk = new_blk;
801 } else if (error)
802 goto bp_err;
803 }
804
805 xlog_put_bp(bp);
806 if (head_blk == log_bbnum)
807 *return_head_blk = 0;
808 else
809 *return_head_blk = head_blk;
810
811
812
813
814
815
816 return 0;
817
818 bp_err:
819 xlog_put_bp(bp);
820
821 if (error)
822 xlog_warn("XFS: failed to find log head");
823 return error;
824}
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842STATIC int
843xlog_find_tail(
844 xlog_t *log,
845 xfs_daddr_t *head_blk,
846 xfs_daddr_t *tail_blk)
847{
848 xlog_rec_header_t *rhead;
849 xlog_op_header_t *op_head;
850 xfs_caddr_t offset = NULL;
851 xfs_buf_t *bp;
852 int error, i, found;
853 xfs_daddr_t umount_data_blk;
854 xfs_daddr_t after_umount_blk;
855 xfs_lsn_t tail_lsn;
856 int hblks;
857
858 found = 0;
859
860
861
862
863 if ((error = xlog_find_head(log, head_blk)))
864 return error;
865
866 bp = xlog_get_bp(log, 1);
867 if (!bp)
868 return ENOMEM;
869 if (*head_blk == 0) {
870 error = xlog_bread(log, 0, 1, bp, &offset);
871 if (error)
872 goto done;
873
874 if (xlog_get_cycle(offset) == 0) {
875 *tail_blk = 0;
876
877 goto done;
878 }
879 }
880
881
882
883
884 ASSERT(*head_blk < INT_MAX);
885 for (i = (int)(*head_blk) - 1; i >= 0; i--) {
886 error = xlog_bread(log, i, 1, bp, &offset);
887 if (error)
888 goto done;
889
890 if (XLOG_HEADER_MAGIC_NUM == be32_to_cpu(*(__be32 *)offset)) {
891 found = 1;
892 break;
893 }
894 }
895
896
897
898
899
900
901 if (!found) {
902 for (i = log->l_logBBsize - 1; i >= (int)(*head_blk); i--) {
903 error = xlog_bread(log, i, 1, bp, &offset);
904 if (error)
905 goto done;
906
907 if (XLOG_HEADER_MAGIC_NUM ==
908 be32_to_cpu(*(__be32 *)offset)) {
909 found = 2;
910 break;
911 }
912 }
913 }
914 if (!found) {
915 xlog_warn("XFS: xlog_find_tail: couldn't find sync record");
916 ASSERT(0);
917 return XFS_ERROR(EIO);
918 }
919
920
921 rhead = (xlog_rec_header_t *)offset;
922 *tail_blk = BLOCK_LSN(be64_to_cpu(rhead->h_tail_lsn));
923
924
925
926
927
928
929
930
931
932
933
934 log->l_prev_block = i;
935 log->l_curr_block = (int)*head_blk;
936 log->l_curr_cycle = be32_to_cpu(rhead->h_cycle);
937 if (found == 2)
938 log->l_curr_cycle++;
939 atomic64_set(&log->l_tail_lsn, be64_to_cpu(rhead->h_tail_lsn));
940 atomic64_set(&log->l_last_sync_lsn, be64_to_cpu(rhead->h_lsn));
941 xlog_assign_grant_head(&log->l_grant_reserve_head, log->l_curr_cycle,
942 BBTOB(log->l_curr_block));
943 xlog_assign_grant_head(&log->l_grant_write_head, log->l_curr_cycle,
944 BBTOB(log->l_curr_block));
945
946
947
948
949
950
951
952
953
954
955
956
957 if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
958 int h_size = be32_to_cpu(rhead->h_size);
959 int h_version = be32_to_cpu(rhead->h_version);
960
961 if ((h_version & XLOG_VERSION_2) &&
962 (h_size > XLOG_HEADER_CYCLE_SIZE)) {
963 hblks = h_size / XLOG_HEADER_CYCLE_SIZE;
964 if (h_size % XLOG_HEADER_CYCLE_SIZE)
965 hblks++;
966 } else {
967 hblks = 1;
968 }
969 } else {
970 hblks = 1;
971 }
972 after_umount_blk = (i + hblks + (int)
973 BTOBB(be32_to_cpu(rhead->h_len))) % log->l_logBBsize;
974 tail_lsn = atomic64_read(&log->l_tail_lsn);
975 if (*head_blk == after_umount_blk &&
976 be32_to_cpu(rhead->h_num_logops) == 1) {
977 umount_data_blk = (i + hblks) % log->l_logBBsize;
978 error = xlog_bread(log, umount_data_blk, 1, bp, &offset);
979 if (error)
980 goto done;
981
982 op_head = (xlog_op_header_t *)offset;
983 if (op_head->oh_flags & XLOG_UNMOUNT_TRANS) {
984
985
986
987
988
989 xlog_assign_atomic_lsn(&log->l_tail_lsn,
990 log->l_curr_cycle, after_umount_blk);
991 xlog_assign_atomic_lsn(&log->l_last_sync_lsn,
992 log->l_curr_cycle, after_umount_blk);
993 *tail_blk = after_umount_blk;
994
995
996
997
998
999
1000
1001 log->l_mp->m_flags |= XFS_MOUNT_WAS_CLEAN;
1002 }
1003 }
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024 if (!xfs_readonly_buftarg(log->l_mp->m_logdev_targp))
1025 error = xlog_clear_stale_blocks(log, tail_lsn);
1026
1027done:
1028 xlog_put_bp(bp);
1029
1030 if (error)
1031 xlog_warn("XFS: failed to locate log tail");
1032 return error;
1033}
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051STATIC int
1052xlog_find_zeroed(
1053 xlog_t *log,
1054 xfs_daddr_t *blk_no)
1055{
1056 xfs_buf_t *bp;
1057 xfs_caddr_t offset;
1058 uint first_cycle, last_cycle;
1059 xfs_daddr_t new_blk, last_blk, start_blk;
1060 xfs_daddr_t num_scan_bblks;
1061 int error, log_bbnum = log->l_logBBsize;
1062
1063 *blk_no = 0;
1064
1065
1066 bp = xlog_get_bp(log, 1);
1067 if (!bp)
1068 return ENOMEM;
1069 error = xlog_bread(log, 0, 1, bp, &offset);
1070 if (error)
1071 goto bp_err;
1072
1073 first_cycle = xlog_get_cycle(offset);
1074 if (first_cycle == 0) {
1075 *blk_no = 0;
1076 xlog_put_bp(bp);
1077 return -1;
1078 }
1079
1080
1081 error = xlog_bread(log, log_bbnum-1, 1, bp, &offset);
1082 if (error)
1083 goto bp_err;
1084
1085 last_cycle = xlog_get_cycle(offset);
1086 if (last_cycle != 0) {
1087 xlog_put_bp(bp);
1088 return 0;
1089 } else if (first_cycle != 1) {
1090
1091
1092
1093
1094
1095 xlog_warn("XFS: Log inconsistent or not a log (last==0, first!=1)");
1096 return XFS_ERROR(EINVAL);
1097 }
1098
1099
1100 last_blk = log_bbnum-1;
1101 if ((error = xlog_find_cycle_start(log, bp, 0, &last_blk, 0)))
1102 goto bp_err;
1103
1104
1105
1106
1107
1108
1109
1110 num_scan_bblks = XLOG_TOTAL_REC_SHIFT(log);
1111 ASSERT(num_scan_bblks <= INT_MAX);
1112
1113 if (last_blk < num_scan_bblks)
1114 num_scan_bblks = last_blk;
1115 start_blk = last_blk - num_scan_bblks;
1116
1117
1118
1119
1120
1121
1122
1123 if ((error = xlog_find_verify_cycle(log, start_blk,
1124 (int)num_scan_bblks, 0, &new_blk)))
1125 goto bp_err;
1126 if (new_blk != -1)
1127 last_blk = new_blk;
1128
1129
1130
1131
1132
1133 if ((error = xlog_find_verify_log_record(log, start_blk,
1134 &last_blk, 0)) == -1) {
1135 error = XFS_ERROR(EIO);
1136 goto bp_err;
1137 } else if (error)
1138 goto bp_err;
1139
1140 *blk_no = last_blk;
1141bp_err:
1142 xlog_put_bp(bp);
1143 if (error)
1144 return error;
1145 return -1;
1146}
1147
1148
1149
1150
1151
1152
1153STATIC void
1154xlog_add_record(
1155 xlog_t *log,
1156 xfs_caddr_t buf,
1157 int cycle,
1158 int block,
1159 int tail_cycle,
1160 int tail_block)
1161{
1162 xlog_rec_header_t *recp = (xlog_rec_header_t *)buf;
1163
1164 memset(buf, 0, BBSIZE);
1165 recp->h_magicno = cpu_to_be32(XLOG_HEADER_MAGIC_NUM);
1166 recp->h_cycle = cpu_to_be32(cycle);
1167 recp->h_version = cpu_to_be32(
1168 xfs_sb_version_haslogv2(&log->l_mp->m_sb) ? 2 : 1);
1169 recp->h_lsn = cpu_to_be64(xlog_assign_lsn(cycle, block));
1170 recp->h_tail_lsn = cpu_to_be64(xlog_assign_lsn(tail_cycle, tail_block));
1171 recp->h_fmt = cpu_to_be32(XLOG_FMT);
1172 memcpy(&recp->h_fs_uuid, &log->l_mp->m_sb.sb_uuid, sizeof(uuid_t));
1173}
1174
1175STATIC int
1176xlog_write_log_records(
1177 xlog_t *log,
1178 int cycle,
1179 int start_block,
1180 int blocks,
1181 int tail_cycle,
1182 int tail_block)
1183{
1184 xfs_caddr_t offset;
1185 xfs_buf_t *bp;
1186 int balign, ealign;
1187 int sectbb = log->l_sectBBsize;
1188 int end_block = start_block + blocks;
1189 int bufblks;
1190 int error = 0;
1191 int i, j = 0;
1192
1193
1194
1195
1196
1197
1198
1199 bufblks = 1 << ffs(blocks);
1200 while (!(bp = xlog_get_bp(log, bufblks))) {
1201 bufblks >>= 1;
1202 if (bufblks < sectbb)
1203 return ENOMEM;
1204 }
1205
1206
1207
1208
1209
1210 balign = round_down(start_block, sectbb);
1211 if (balign != start_block) {
1212 error = xlog_bread_noalign(log, start_block, 1, bp);
1213 if (error)
1214 goto out_put_bp;
1215
1216 j = start_block - balign;
1217 }
1218
1219 for (i = start_block; i < end_block; i += bufblks) {
1220 int bcount, endcount;
1221
1222 bcount = min(bufblks, end_block - start_block);
1223 endcount = bcount - j;
1224
1225
1226
1227
1228
1229 ealign = round_down(end_block, sectbb);
1230 if (j == 0 && (start_block + endcount > ealign)) {
1231 offset = XFS_BUF_PTR(bp);
1232 balign = BBTOB(ealign - start_block);
1233 error = XFS_BUF_SET_PTR(bp, offset + balign,
1234 BBTOB(sectbb));
1235 if (error)
1236 break;
1237
1238 error = xlog_bread_noalign(log, ealign, sectbb, bp);
1239 if (error)
1240 break;
1241
1242 error = XFS_BUF_SET_PTR(bp, offset, bufblks);
1243 if (error)
1244 break;
1245 }
1246
1247 offset = xlog_align(log, start_block, endcount, bp);
1248 for (; j < endcount; j++) {
1249 xlog_add_record(log, offset, cycle, i+j,
1250 tail_cycle, tail_block);
1251 offset += BBSIZE;
1252 }
1253 error = xlog_bwrite(log, start_block, endcount, bp);
1254 if (error)
1255 break;
1256 start_block += endcount;
1257 j = 0;
1258 }
1259
1260 out_put_bp:
1261 xlog_put_bp(bp);
1262 return error;
1263}
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281STATIC int
1282xlog_clear_stale_blocks(
1283 xlog_t *log,
1284 xfs_lsn_t tail_lsn)
1285{
1286 int tail_cycle, head_cycle;
1287 int tail_block, head_block;
1288 int tail_distance, max_distance;
1289 int distance;
1290 int error;
1291
1292 tail_cycle = CYCLE_LSN(tail_lsn);
1293 tail_block = BLOCK_LSN(tail_lsn);
1294 head_cycle = log->l_curr_cycle;
1295 head_block = log->l_curr_block;
1296
1297
1298
1299
1300
1301
1302
1303 if (head_cycle == tail_cycle) {
1304
1305
1306
1307
1308
1309
1310
1311 if (unlikely(head_block < tail_block || head_block >= log->l_logBBsize)) {
1312 XFS_ERROR_REPORT("xlog_clear_stale_blocks(1)",
1313 XFS_ERRLEVEL_LOW, log->l_mp);
1314 return XFS_ERROR(EFSCORRUPTED);
1315 }
1316 tail_distance = tail_block + (log->l_logBBsize - head_block);
1317 } else {
1318
1319
1320
1321
1322
1323 if (unlikely(head_block >= tail_block || head_cycle != (tail_cycle + 1))){
1324 XFS_ERROR_REPORT("xlog_clear_stale_blocks(2)",
1325 XFS_ERRLEVEL_LOW, log->l_mp);
1326 return XFS_ERROR(EFSCORRUPTED);
1327 }
1328 tail_distance = tail_block - head_block;
1329 }
1330
1331
1332
1333
1334
1335 if (tail_distance <= 0) {
1336 ASSERT(tail_distance == 0);
1337 return 0;
1338 }
1339
1340 max_distance = XLOG_TOTAL_REC_SHIFT(log);
1341
1342
1343
1344
1345
1346
1347
1348 max_distance = MIN(max_distance, tail_distance);
1349
1350 if ((head_block + max_distance) <= log->l_logBBsize) {
1351
1352
1353
1354
1355
1356
1357
1358 error = xlog_write_log_records(log, (head_cycle - 1),
1359 head_block, max_distance, tail_cycle,
1360 tail_block);
1361 if (error)
1362 return error;
1363 } else {
1364
1365
1366
1367
1368
1369
1370
1371 distance = log->l_logBBsize - head_block;
1372 error = xlog_write_log_records(log, (head_cycle - 1),
1373 head_block, distance, tail_cycle,
1374 tail_block);
1375
1376 if (error)
1377 return error;
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387 distance = max_distance - (log->l_logBBsize - head_block);
1388 error = xlog_write_log_records(log, head_cycle, 0, distance,
1389 tail_cycle, tail_block);
1390 if (error)
1391 return error;
1392 }
1393
1394 return 0;
1395}
1396
1397
1398
1399
1400
1401
1402
1403
1404STATIC xlog_recover_t *
1405xlog_recover_find_tid(
1406 struct hlist_head *head,
1407 xlog_tid_t tid)
1408{
1409 xlog_recover_t *trans;
1410 struct hlist_node *n;
1411
1412 hlist_for_each_entry(trans, n, head, r_list) {
1413 if (trans->r_log_tid == tid)
1414 return trans;
1415 }
1416 return NULL;
1417}
1418
1419STATIC void
1420xlog_recover_new_tid(
1421 struct hlist_head *head,
1422 xlog_tid_t tid,
1423 xfs_lsn_t lsn)
1424{
1425 xlog_recover_t *trans;
1426
1427 trans = kmem_zalloc(sizeof(xlog_recover_t), KM_SLEEP);
1428 trans->r_log_tid = tid;
1429 trans->r_lsn = lsn;
1430 INIT_LIST_HEAD(&trans->r_itemq);
1431
1432 INIT_HLIST_NODE(&trans->r_list);
1433 hlist_add_head(&trans->r_list, head);
1434}
1435
1436STATIC void
1437xlog_recover_add_item(
1438 struct list_head *head)
1439{
1440 xlog_recover_item_t *item;
1441
1442 item = kmem_zalloc(sizeof(xlog_recover_item_t), KM_SLEEP);
1443 INIT_LIST_HEAD(&item->ri_list);
1444 list_add_tail(&item->ri_list, head);
1445}
1446
1447STATIC int
1448xlog_recover_add_to_cont_trans(
1449 struct log *log,
1450 xlog_recover_t *trans,
1451 xfs_caddr_t dp,
1452 int len)
1453{
1454 xlog_recover_item_t *item;
1455 xfs_caddr_t ptr, old_ptr;
1456 int old_len;
1457
1458 if (list_empty(&trans->r_itemq)) {
1459
1460 xlog_recover_add_item(&trans->r_itemq);
1461 ptr = (xfs_caddr_t) &trans->r_theader +
1462 sizeof(xfs_trans_header_t) - len;
1463 memcpy(ptr, dp, len);
1464 return 0;
1465 }
1466
1467 item = list_entry(trans->r_itemq.prev, xlog_recover_item_t, ri_list);
1468
1469 old_ptr = item->ri_buf[item->ri_cnt-1].i_addr;
1470 old_len = item->ri_buf[item->ri_cnt-1].i_len;
1471
1472 ptr = kmem_realloc(old_ptr, len+old_len, old_len, 0u);
1473 memcpy(&ptr[old_len], dp, len);
1474 item->ri_buf[item->ri_cnt-1].i_len += len;
1475 item->ri_buf[item->ri_cnt-1].i_addr = ptr;
1476 trace_xfs_log_recover_item_add_cont(log, trans, item, 0);
1477 return 0;
1478}
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493STATIC int
1494xlog_recover_add_to_trans(
1495 struct log *log,
1496 xlog_recover_t *trans,
1497 xfs_caddr_t dp,
1498 int len)
1499{
1500 xfs_inode_log_format_t *in_f;
1501 xlog_recover_item_t *item;
1502 xfs_caddr_t ptr;
1503
1504 if (!len)
1505 return 0;
1506 if (list_empty(&trans->r_itemq)) {
1507
1508 if (*(uint *)dp != XFS_TRANS_HEADER_MAGIC) {
1509 xlog_warn("XFS: xlog_recover_add_to_trans: "
1510 "bad header magic number");
1511 ASSERT(0);
1512 return XFS_ERROR(EIO);
1513 }
1514 if (len == sizeof(xfs_trans_header_t))
1515 xlog_recover_add_item(&trans->r_itemq);
1516 memcpy(&trans->r_theader, dp, len);
1517 return 0;
1518 }
1519
1520 ptr = kmem_alloc(len, KM_SLEEP);
1521 memcpy(ptr, dp, len);
1522 in_f = (xfs_inode_log_format_t *)ptr;
1523
1524
1525 item = list_entry(trans->r_itemq.prev, xlog_recover_item_t, ri_list);
1526 if (item->ri_total != 0 &&
1527 item->ri_total == item->ri_cnt) {
1528
1529 xlog_recover_add_item(&trans->r_itemq);
1530 item = list_entry(trans->r_itemq.prev,
1531 xlog_recover_item_t, ri_list);
1532 }
1533
1534 if (item->ri_total == 0) {
1535 if (in_f->ilf_size == 0 ||
1536 in_f->ilf_size > XLOG_MAX_REGIONS_IN_ITEM) {
1537 xlog_warn(
1538 "XFS: bad number of regions (%d) in inode log format",
1539 in_f->ilf_size);
1540 ASSERT(0);
1541 return XFS_ERROR(EIO);
1542 }
1543
1544 item->ri_total = in_f->ilf_size;
1545 item->ri_buf =
1546 kmem_zalloc(item->ri_total * sizeof(xfs_log_iovec_t),
1547 KM_SLEEP);
1548 }
1549 ASSERT(item->ri_total > item->ri_cnt);
1550
1551 item->ri_buf[item->ri_cnt].i_addr = ptr;
1552 item->ri_buf[item->ri_cnt].i_len = len;
1553 item->ri_cnt++;
1554 trace_xfs_log_recover_item_add(log, trans, item, 0);
1555 return 0;
1556}
1557
1558
1559
1560
1561
1562
1563
1564STATIC int
1565xlog_recover_reorder_trans(
1566 struct log *log,
1567 xlog_recover_t *trans,
1568 int pass)
1569{
1570 xlog_recover_item_t *item, *n;
1571 LIST_HEAD(sort_list);
1572
1573 list_splice_init(&trans->r_itemq, &sort_list);
1574 list_for_each_entry_safe(item, n, &sort_list, ri_list) {
1575 xfs_buf_log_format_t *buf_f = item->ri_buf[0].i_addr;
1576
1577 switch (ITEM_TYPE(item)) {
1578 case XFS_LI_BUF:
1579 if (!(buf_f->blf_flags & XFS_BLF_CANCEL)) {
1580 trace_xfs_log_recover_item_reorder_head(log,
1581 trans, item, pass);
1582 list_move(&item->ri_list, &trans->r_itemq);
1583 break;
1584 }
1585 case XFS_LI_INODE:
1586 case XFS_LI_DQUOT:
1587 case XFS_LI_QUOTAOFF:
1588 case XFS_LI_EFD:
1589 case XFS_LI_EFI:
1590 trace_xfs_log_recover_item_reorder_tail(log,
1591 trans, item, pass);
1592 list_move_tail(&item->ri_list, &trans->r_itemq);
1593 break;
1594 default:
1595 xlog_warn(
1596 "XFS: xlog_recover_reorder_trans: unrecognized type of log operation");
1597 ASSERT(0);
1598 return XFS_ERROR(EIO);
1599 }
1600 }
1601 ASSERT(list_empty(&sort_list));
1602 return 0;
1603}
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617STATIC int
1618xlog_recover_buffer_pass1(
1619 struct log *log,
1620 xlog_recover_item_t *item)
1621{
1622 xfs_buf_log_format_t *buf_f = item->ri_buf[0].i_addr;
1623 struct list_head *bucket;
1624 struct xfs_buf_cancel *bcp;
1625
1626
1627
1628
1629 if (!(buf_f->blf_flags & XFS_BLF_CANCEL)) {
1630 trace_xfs_log_recover_buf_not_cancel(log, buf_f);
1631 return 0;
1632 }
1633
1634
1635
1636
1637
1638 bucket = XLOG_BUF_CANCEL_BUCKET(log, buf_f->blf_blkno);
1639 list_for_each_entry(bcp, bucket, bc_list) {
1640 if (bcp->bc_blkno == buf_f->blf_blkno &&
1641 bcp->bc_len == buf_f->blf_len) {
1642 bcp->bc_refcount++;
1643 trace_xfs_log_recover_buf_cancel_ref_inc(log, buf_f);
1644 return 0;
1645 }
1646 }
1647
1648 bcp = kmem_alloc(sizeof(struct xfs_buf_cancel), KM_SLEEP);
1649 bcp->bc_blkno = buf_f->blf_blkno;
1650 bcp->bc_len = buf_f->blf_len;
1651 bcp->bc_refcount = 1;
1652 list_add_tail(&bcp->bc_list, bucket);
1653
1654 trace_xfs_log_recover_buf_cancel_add(log, buf_f);
1655 return 0;
1656}
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671STATIC int
1672xlog_check_buffer_cancelled(
1673 struct log *log,
1674 xfs_daddr_t blkno,
1675 uint len,
1676 ushort flags)
1677{
1678 struct list_head *bucket;
1679 struct xfs_buf_cancel *bcp;
1680
1681 if (log->l_buf_cancel_table == NULL) {
1682
1683
1684
1685
1686 ASSERT(!(flags & XFS_BLF_CANCEL));
1687 return 0;
1688 }
1689
1690
1691
1692
1693 bucket = XLOG_BUF_CANCEL_BUCKET(log, blkno);
1694 list_for_each_entry(bcp, bucket, bc_list) {
1695 if (bcp->bc_blkno == blkno && bcp->bc_len == len)
1696 goto found;
1697 }
1698
1699
1700
1701
1702
1703 ASSERT(!(flags & XFS_BLF_CANCEL));
1704 return 0;
1705
1706found:
1707
1708
1709
1710
1711
1712
1713 if (flags & XFS_BLF_CANCEL) {
1714 if (--bcp->bc_refcount == 0) {
1715 list_del(&bcp->bc_list);
1716 kmem_free(bcp);
1717 }
1718 }
1719 return 1;
1720}
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734STATIC int
1735xlog_recover_do_inode_buffer(
1736 struct xfs_mount *mp,
1737 xlog_recover_item_t *item,
1738 struct xfs_buf *bp,
1739 xfs_buf_log_format_t *buf_f)
1740{
1741 int i;
1742 int item_index = 0;
1743 int bit = 0;
1744 int nbits = 0;
1745 int reg_buf_offset = 0;
1746 int reg_buf_bytes = 0;
1747 int next_unlinked_offset;
1748 int inodes_per_buf;
1749 xfs_agino_t *logged_nextp;
1750 xfs_agino_t *buffer_nextp;
1751
1752 trace_xfs_log_recover_buf_inode_buf(mp->m_log, buf_f);
1753
1754 inodes_per_buf = XFS_BUF_COUNT(bp) >> mp->m_sb.sb_inodelog;
1755 for (i = 0; i < inodes_per_buf; i++) {
1756 next_unlinked_offset = (i * mp->m_sb.sb_inodesize) +
1757 offsetof(xfs_dinode_t, di_next_unlinked);
1758
1759 while (next_unlinked_offset >=
1760 (reg_buf_offset + reg_buf_bytes)) {
1761
1762
1763
1764
1765
1766
1767 bit += nbits;
1768 bit = xfs_next_bit(buf_f->blf_data_map,
1769 buf_f->blf_map_size, bit);
1770
1771
1772
1773
1774
1775 if (bit == -1)
1776 return 0;
1777
1778 nbits = xfs_contig_bits(buf_f->blf_data_map,
1779 buf_f->blf_map_size, bit);
1780 ASSERT(nbits > 0);
1781 reg_buf_offset = bit << XFS_BLF_SHIFT;
1782 reg_buf_bytes = nbits << XFS_BLF_SHIFT;
1783 item_index++;
1784 }
1785
1786
1787
1788
1789
1790
1791 if (next_unlinked_offset < reg_buf_offset)
1792 continue;
1793
1794 ASSERT(item->ri_buf[item_index].i_addr != NULL);
1795 ASSERT((item->ri_buf[item_index].i_len % XFS_BLF_CHUNK) == 0);
1796 ASSERT((reg_buf_offset + reg_buf_bytes) <= XFS_BUF_COUNT(bp));
1797
1798
1799
1800
1801
1802
1803 logged_nextp = item->ri_buf[item_index].i_addr +
1804 next_unlinked_offset - reg_buf_offset;
1805 if (unlikely(*logged_nextp == 0)) {
1806 xfs_fs_cmn_err(CE_ALERT, mp,
1807 "bad inode buffer log record (ptr = 0x%p, bp = 0x%p). XFS trying to replay bad (0) inode di_next_unlinked field",
1808 item, bp);
1809 XFS_ERROR_REPORT("xlog_recover_do_inode_buf",
1810 XFS_ERRLEVEL_LOW, mp);
1811 return XFS_ERROR(EFSCORRUPTED);
1812 }
1813
1814 buffer_nextp = (xfs_agino_t *)xfs_buf_offset(bp,
1815 next_unlinked_offset);
1816 *buffer_nextp = *logged_nextp;
1817 }
1818
1819 return 0;
1820}
1821
1822
1823
1824
1825
1826
1827
1828STATIC void
1829xlog_recover_do_reg_buffer(
1830 struct xfs_mount *mp,
1831 xlog_recover_item_t *item,
1832 struct xfs_buf *bp,
1833 xfs_buf_log_format_t *buf_f)
1834{
1835 int i;
1836 int bit;
1837 int nbits;
1838 int error;
1839
1840 trace_xfs_log_recover_buf_reg_buf(mp->m_log, buf_f);
1841
1842 bit = 0;
1843 i = 1;
1844 while (1) {
1845 bit = xfs_next_bit(buf_f->blf_data_map,
1846 buf_f->blf_map_size, bit);
1847 if (bit == -1)
1848 break;
1849 nbits = xfs_contig_bits(buf_f->blf_data_map,
1850 buf_f->blf_map_size, bit);
1851 ASSERT(nbits > 0);
1852 ASSERT(item->ri_buf[i].i_addr != NULL);
1853 ASSERT(item->ri_buf[i].i_len % XFS_BLF_CHUNK == 0);
1854 ASSERT(XFS_BUF_COUNT(bp) >=
1855 ((uint)bit << XFS_BLF_SHIFT)+(nbits<<XFS_BLF_SHIFT));
1856
1857
1858
1859
1860
1861
1862 error = 0;
1863 if (buf_f->blf_flags &
1864 (XFS_BLF_UDQUOT_BUF|XFS_BLF_PDQUOT_BUF|XFS_BLF_GDQUOT_BUF)) {
1865 if (item->ri_buf[i].i_addr == NULL) {
1866 cmn_err(CE_ALERT,
1867 "XFS: NULL dquot in %s.", __func__);
1868 goto next;
1869 }
1870 if (item->ri_buf[i].i_len < sizeof(xfs_disk_dquot_t)) {
1871 cmn_err(CE_ALERT,
1872 "XFS: dquot too small (%d) in %s.",
1873 item->ri_buf[i].i_len, __func__);
1874 goto next;
1875 }
1876 error = xfs_qm_dqcheck(item->ri_buf[i].i_addr,
1877 -1, 0, XFS_QMOPT_DOWARN,
1878 "dquot_buf_recover");
1879 if (error)
1880 goto next;
1881 }
1882
1883 memcpy(xfs_buf_offset(bp,
1884 (uint)bit << XFS_BLF_SHIFT),
1885 item->ri_buf[i].i_addr,
1886 nbits<<XFS_BLF_SHIFT);
1887 next:
1888 i++;
1889 bit += nbits;
1890 }
1891
1892
1893 ASSERT(i == item->ri_total);
1894}
1895
1896
1897
1898
1899int
1900xfs_qm_dqcheck(
1901 xfs_disk_dquot_t *ddq,
1902 xfs_dqid_t id,
1903 uint type,
1904 uint flags,
1905 char *str)
1906{
1907 xfs_dqblk_t *d = (xfs_dqblk_t *)ddq;
1908 int errs = 0;
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925 if (be16_to_cpu(ddq->d_magic) != XFS_DQUOT_MAGIC) {
1926 if (flags & XFS_QMOPT_DOWARN)
1927 cmn_err(CE_ALERT,
1928 "%s : XFS dquot ID 0x%x, magic 0x%x != 0x%x",
1929 str, id, be16_to_cpu(ddq->d_magic), XFS_DQUOT_MAGIC);
1930 errs++;
1931 }
1932 if (ddq->d_version != XFS_DQUOT_VERSION) {
1933 if (flags & XFS_QMOPT_DOWARN)
1934 cmn_err(CE_ALERT,
1935 "%s : XFS dquot ID 0x%x, version 0x%x != 0x%x",
1936 str, id, ddq->d_version, XFS_DQUOT_VERSION);
1937 errs++;
1938 }
1939
1940 if (ddq->d_flags != XFS_DQ_USER &&
1941 ddq->d_flags != XFS_DQ_PROJ &&
1942 ddq->d_flags != XFS_DQ_GROUP) {
1943 if (flags & XFS_QMOPT_DOWARN)
1944 cmn_err(CE_ALERT,
1945 "%s : XFS dquot ID 0x%x, unknown flags 0x%x",
1946 str, id, ddq->d_flags);
1947 errs++;
1948 }
1949
1950 if (id != -1 && id != be32_to_cpu(ddq->d_id)) {
1951 if (flags & XFS_QMOPT_DOWARN)
1952 cmn_err(CE_ALERT,
1953 "%s : ondisk-dquot 0x%p, ID mismatch: "
1954 "0x%x expected, found id 0x%x",
1955 str, ddq, id, be32_to_cpu(ddq->d_id));
1956 errs++;
1957 }
1958
1959 if (!errs && ddq->d_id) {
1960 if (ddq->d_blk_softlimit &&
1961 be64_to_cpu(ddq->d_bcount) >=
1962 be64_to_cpu(ddq->d_blk_softlimit)) {
1963 if (!ddq->d_btimer) {
1964 if (flags & XFS_QMOPT_DOWARN)
1965 cmn_err(CE_ALERT,
1966 "%s : Dquot ID 0x%x (0x%p) "
1967 "BLK TIMER NOT STARTED",
1968 str, (int)be32_to_cpu(ddq->d_id), ddq);
1969 errs++;
1970 }
1971 }
1972 if (ddq->d_ino_softlimit &&
1973 be64_to_cpu(ddq->d_icount) >=
1974 be64_to_cpu(ddq->d_ino_softlimit)) {
1975 if (!ddq->d_itimer) {
1976 if (flags & XFS_QMOPT_DOWARN)
1977 cmn_err(CE_ALERT,
1978 "%s : Dquot ID 0x%x (0x%p) "
1979 "INODE TIMER NOT STARTED",
1980 str, (int)be32_to_cpu(ddq->d_id), ddq);
1981 errs++;
1982 }
1983 }
1984 if (ddq->d_rtb_softlimit &&
1985 be64_to_cpu(ddq->d_rtbcount) >=
1986 be64_to_cpu(ddq->d_rtb_softlimit)) {
1987 if (!ddq->d_rtbtimer) {
1988 if (flags & XFS_QMOPT_DOWARN)
1989 cmn_err(CE_ALERT,
1990 "%s : Dquot ID 0x%x (0x%p) "
1991 "RTBLK TIMER NOT STARTED",
1992 str, (int)be32_to_cpu(ddq->d_id), ddq);
1993 errs++;
1994 }
1995 }
1996 }
1997
1998 if (!errs || !(flags & XFS_QMOPT_DQREPAIR))
1999 return errs;
2000
2001 if (flags & XFS_QMOPT_DOWARN)
2002 cmn_err(CE_NOTE, "Re-initializing dquot ID 0x%x", id);
2003
2004
2005
2006
2007 ASSERT(id != -1);
2008 ASSERT(flags & XFS_QMOPT_DQREPAIR);
2009 memset(d, 0, sizeof(xfs_dqblk_t));
2010
2011 d->dd_diskdq.d_magic = cpu_to_be16(XFS_DQUOT_MAGIC);
2012 d->dd_diskdq.d_version = XFS_DQUOT_VERSION;
2013 d->dd_diskdq.d_flags = type;
2014 d->dd_diskdq.d_id = cpu_to_be32(id);
2015
2016 return errs;
2017}
2018
2019
2020
2021
2022
2023
2024
2025STATIC void
2026xlog_recover_do_dquot_buffer(
2027 xfs_mount_t *mp,
2028 xlog_t *log,
2029 xlog_recover_item_t *item,
2030 xfs_buf_t *bp,
2031 xfs_buf_log_format_t *buf_f)
2032{
2033 uint type;
2034
2035 trace_xfs_log_recover_buf_dquot_buf(log, buf_f);
2036
2037
2038
2039
2040 if (mp->m_qflags == 0) {
2041 return;
2042 }
2043
2044 type = 0;
2045 if (buf_f->blf_flags & XFS_BLF_UDQUOT_BUF)
2046 type |= XFS_DQ_USER;
2047 if (buf_f->blf_flags & XFS_BLF_PDQUOT_BUF)
2048 type |= XFS_DQ_PROJ;
2049 if (buf_f->blf_flags & XFS_BLF_GDQUOT_BUF)
2050 type |= XFS_DQ_GROUP;
2051
2052
2053
2054 if (log->l_quotaoffs_flag & type)
2055 return;
2056
2057 xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
2058}
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083STATIC int
2084xlog_recover_buffer_pass2(
2085 xlog_t *log,
2086 xlog_recover_item_t *item)
2087{
2088 xfs_buf_log_format_t *buf_f = item->ri_buf[0].i_addr;
2089 xfs_mount_t *mp = log->l_mp;
2090 xfs_buf_t *bp;
2091 int error;
2092 uint buf_flags;
2093
2094
2095
2096
2097
2098 if (xlog_check_buffer_cancelled(log, buf_f->blf_blkno,
2099 buf_f->blf_len, buf_f->blf_flags)) {
2100 trace_xfs_log_recover_buf_cancel(log, buf_f);
2101 return 0;
2102 }
2103
2104 trace_xfs_log_recover_buf_recover(log, buf_f);
2105
2106 buf_flags = XBF_LOCK;
2107 if (!(buf_f->blf_flags & XFS_BLF_INODE_BUF))
2108 buf_flags |= XBF_MAPPED;
2109
2110 bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
2111 buf_flags);
2112 if (XFS_BUF_ISERROR(bp)) {
2113 xfs_ioerror_alert("xlog_recover_do..(read#1)", mp,
2114 bp, buf_f->blf_blkno);
2115 error = XFS_BUF_GETERROR(bp);
2116 xfs_buf_relse(bp);
2117 return error;
2118 }
2119
2120 error = 0;
2121 if (buf_f->blf_flags & XFS_BLF_INODE_BUF) {
2122 error = xlog_recover_do_inode_buffer(mp, item, bp, buf_f);
2123 } else if (buf_f->blf_flags &
2124 (XFS_BLF_UDQUOT_BUF|XFS_BLF_PDQUOT_BUF|XFS_BLF_GDQUOT_BUF)) {
2125 xlog_recover_do_dquot_buffer(mp, log, item, bp, buf_f);
2126 } else {
2127 xlog_recover_do_reg_buffer(mp, item, bp, buf_f);
2128 }
2129 if (error)
2130 return XFS_ERROR(error);
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147 if (XFS_DINODE_MAGIC ==
2148 be16_to_cpu(*((__be16 *)xfs_buf_offset(bp, 0))) &&
2149 (XFS_BUF_COUNT(bp) != MAX(log->l_mp->m_sb.sb_blocksize,
2150 (__uint32_t)XFS_INODE_CLUSTER_SIZE(log->l_mp)))) {
2151 XFS_BUF_STALE(bp);
2152 error = xfs_bwrite(mp, bp);
2153 } else {
2154 ASSERT(bp->b_target->bt_mount == mp);
2155 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2156 xfs_bdwrite(mp, bp);
2157 }
2158
2159 return (error);
2160}
2161
2162STATIC int
2163xlog_recover_inode_pass2(
2164 xlog_t *log,
2165 xlog_recover_item_t *item)
2166{
2167 xfs_inode_log_format_t *in_f;
2168 xfs_mount_t *mp = log->l_mp;
2169 xfs_buf_t *bp;
2170 xfs_dinode_t *dip;
2171 int len;
2172 xfs_caddr_t src;
2173 xfs_caddr_t dest;
2174 int error;
2175 int attr_index;
2176 uint fields;
2177 xfs_icdinode_t *dicp;
2178 int need_free = 0;
2179
2180 if (item->ri_buf[0].i_len == sizeof(xfs_inode_log_format_t)) {
2181 in_f = item->ri_buf[0].i_addr;
2182 } else {
2183 in_f = kmem_alloc(sizeof(xfs_inode_log_format_t), KM_SLEEP);
2184 need_free = 1;
2185 error = xfs_inode_item_format_convert(&item->ri_buf[0], in_f);
2186 if (error)
2187 goto error;
2188 }
2189
2190
2191
2192
2193
2194 if (xlog_check_buffer_cancelled(log, in_f->ilf_blkno,
2195 in_f->ilf_len, 0)) {
2196 error = 0;
2197 trace_xfs_log_recover_inode_cancel(log, in_f);
2198 goto error;
2199 }
2200 trace_xfs_log_recover_inode_recover(log, in_f);
2201
2202 bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len,
2203 XBF_LOCK);
2204 if (XFS_BUF_ISERROR(bp)) {
2205 xfs_ioerror_alert("xlog_recover_do..(read#2)", mp,
2206 bp, in_f->ilf_blkno);
2207 error = XFS_BUF_GETERROR(bp);
2208 xfs_buf_relse(bp);
2209 goto error;
2210 }
2211 error = 0;
2212 ASSERT(in_f->ilf_fields & XFS_ILOG_CORE);
2213 dip = (xfs_dinode_t *)xfs_buf_offset(bp, in_f->ilf_boffset);
2214
2215
2216
2217
2218
2219 if (unlikely(be16_to_cpu(dip->di_magic) != XFS_DINODE_MAGIC)) {
2220 xfs_buf_relse(bp);
2221 xfs_fs_cmn_err(CE_ALERT, mp,
2222 "xfs_inode_recover: Bad inode magic number, dino ptr = 0x%p, dino bp = 0x%p, ino = %Ld",
2223 dip, bp, in_f->ilf_ino);
2224 XFS_ERROR_REPORT("xlog_recover_inode_pass2(1)",
2225 XFS_ERRLEVEL_LOW, mp);
2226 error = EFSCORRUPTED;
2227 goto error;
2228 }
2229 dicp = item->ri_buf[1].i_addr;
2230 if (unlikely(dicp->di_magic != XFS_DINODE_MAGIC)) {
2231 xfs_buf_relse(bp);
2232 xfs_fs_cmn_err(CE_ALERT, mp,
2233 "xfs_inode_recover: Bad inode log record, rec ptr 0x%p, ino %Ld",
2234 item, in_f->ilf_ino);
2235 XFS_ERROR_REPORT("xlog_recover_inode_pass2(2)",
2236 XFS_ERRLEVEL_LOW, mp);
2237 error = EFSCORRUPTED;
2238 goto error;
2239 }
2240
2241
2242 if (dicp->di_flushiter < be16_to_cpu(dip->di_flushiter)) {
2243
2244
2245
2246
2247 if (be16_to_cpu(dip->di_flushiter) == DI_MAX_FLUSH &&
2248 dicp->di_flushiter < (DI_MAX_FLUSH >> 1)) {
2249
2250 } else {
2251 xfs_buf_relse(bp);
2252 trace_xfs_log_recover_inode_skip(log, in_f);
2253 error = 0;
2254 goto error;
2255 }
2256 }
2257
2258 dicp->di_flushiter = 0;
2259
2260 if (unlikely((dicp->di_mode & S_IFMT) == S_IFREG)) {
2261 if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
2262 (dicp->di_format != XFS_DINODE_FMT_BTREE)) {
2263 XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(3)",
2264 XFS_ERRLEVEL_LOW, mp, dicp);
2265 xfs_buf_relse(bp);
2266 xfs_fs_cmn_err(CE_ALERT, mp,
2267 "xfs_inode_recover: Bad regular inode log record, rec ptr 0x%p, ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
2268 item, dip, bp, in_f->ilf_ino);
2269 error = EFSCORRUPTED;
2270 goto error;
2271 }
2272 } else if (unlikely((dicp->di_mode & S_IFMT) == S_IFDIR)) {
2273 if ((dicp->di_format != XFS_DINODE_FMT_EXTENTS) &&
2274 (dicp->di_format != XFS_DINODE_FMT_BTREE) &&
2275 (dicp->di_format != XFS_DINODE_FMT_LOCAL)) {
2276 XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(4)",
2277 XFS_ERRLEVEL_LOW, mp, dicp);
2278 xfs_buf_relse(bp);
2279 xfs_fs_cmn_err(CE_ALERT, mp,
2280 "xfs_inode_recover: Bad dir inode log record, rec ptr 0x%p, ino ptr = 0x%p, ino bp = 0x%p, ino %Ld",
2281 item, dip, bp, in_f->ilf_ino);
2282 error = EFSCORRUPTED;
2283 goto error;
2284 }
2285 }
2286 if (unlikely(dicp->di_nextents + dicp->di_anextents > dicp->di_nblocks)){
2287 XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(5)",
2288 XFS_ERRLEVEL_LOW, mp, dicp);
2289 xfs_buf_relse(bp);
2290 xfs_fs_cmn_err(CE_ALERT, mp,
2291 "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",
2292 item, dip, bp, in_f->ilf_ino,
2293 dicp->di_nextents + dicp->di_anextents,
2294 dicp->di_nblocks);
2295 error = EFSCORRUPTED;
2296 goto error;
2297 }
2298 if (unlikely(dicp->di_forkoff > mp->m_sb.sb_inodesize)) {
2299 XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(6)",
2300 XFS_ERRLEVEL_LOW, mp, dicp);
2301 xfs_buf_relse(bp);
2302 xfs_fs_cmn_err(CE_ALERT, mp,
2303 "xfs_inode_recover: Bad inode log rec ptr 0x%p, dino ptr 0x%p, dino bp 0x%p, ino %Ld, forkoff 0x%x",
2304 item, dip, bp, in_f->ilf_ino, dicp->di_forkoff);
2305 error = EFSCORRUPTED;
2306 goto error;
2307 }
2308 if (unlikely(item->ri_buf[1].i_len > sizeof(struct xfs_icdinode))) {
2309 XFS_CORRUPTION_ERROR("xlog_recover_inode_pass2(7)",
2310 XFS_ERRLEVEL_LOW, mp, dicp);
2311 xfs_buf_relse(bp);
2312 xfs_fs_cmn_err(CE_ALERT, mp,
2313 "xfs_inode_recover: Bad inode log record length %d, rec ptr 0x%p",
2314 item->ri_buf[1].i_len, item);
2315 error = EFSCORRUPTED;
2316 goto error;
2317 }
2318
2319
2320 xfs_dinode_to_disk(dip, item->ri_buf[1].i_addr);
2321
2322
2323 if (item->ri_buf[1].i_len > sizeof(struct xfs_icdinode)) {
2324 memcpy((xfs_caddr_t) dip + sizeof(struct xfs_icdinode),
2325 item->ri_buf[1].i_addr + sizeof(struct xfs_icdinode),
2326 item->ri_buf[1].i_len - sizeof(struct xfs_icdinode));
2327 }
2328
2329 fields = in_f->ilf_fields;
2330 switch (fields & (XFS_ILOG_DEV | XFS_ILOG_UUID)) {
2331 case XFS_ILOG_DEV:
2332 xfs_dinode_put_rdev(dip, in_f->ilf_u.ilfu_rdev);
2333 break;
2334 case XFS_ILOG_UUID:
2335 memcpy(XFS_DFORK_DPTR(dip),
2336 &in_f->ilf_u.ilfu_uuid,
2337 sizeof(uuid_t));
2338 break;
2339 }
2340
2341 if (in_f->ilf_size == 2)
2342 goto write_inode_buffer;
2343 len = item->ri_buf[2].i_len;
2344 src = item->ri_buf[2].i_addr;
2345 ASSERT(in_f->ilf_size <= 4);
2346 ASSERT((in_f->ilf_size == 3) || (fields & XFS_ILOG_AFORK));
2347 ASSERT(!(fields & XFS_ILOG_DFORK) ||
2348 (len == in_f->ilf_dsize));
2349
2350 switch (fields & XFS_ILOG_DFORK) {
2351 case XFS_ILOG_DDATA:
2352 case XFS_ILOG_DEXT:
2353 memcpy(XFS_DFORK_DPTR(dip), src, len);
2354 break;
2355
2356 case XFS_ILOG_DBROOT:
2357 xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src, len,
2358 (xfs_bmdr_block_t *)XFS_DFORK_DPTR(dip),
2359 XFS_DFORK_DSIZE(dip, mp));
2360 break;
2361
2362 default:
2363
2364
2365
2366 ASSERT((fields & XFS_ILOG_DFORK) == 0);
2367 break;
2368 }
2369
2370
2371
2372
2373
2374
2375 if (in_f->ilf_fields & XFS_ILOG_AFORK) {
2376 if (in_f->ilf_fields & XFS_ILOG_DFORK) {
2377 attr_index = 3;
2378 } else {
2379 attr_index = 2;
2380 }
2381 len = item->ri_buf[attr_index].i_len;
2382 src = item->ri_buf[attr_index].i_addr;
2383 ASSERT(len == in_f->ilf_asize);
2384
2385 switch (in_f->ilf_fields & XFS_ILOG_AFORK) {
2386 case XFS_ILOG_ADATA:
2387 case XFS_ILOG_AEXT:
2388 dest = XFS_DFORK_APTR(dip);
2389 ASSERT(len <= XFS_DFORK_ASIZE(dip, mp));
2390 memcpy(dest, src, len);
2391 break;
2392
2393 case XFS_ILOG_ABROOT:
2394 dest = XFS_DFORK_APTR(dip);
2395 xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src,
2396 len, (xfs_bmdr_block_t*)dest,
2397 XFS_DFORK_ASIZE(dip, mp));
2398 break;
2399
2400 default:
2401 xlog_warn("XFS: xlog_recover_inode_pass2: Invalid flag");
2402 ASSERT(0);
2403 xfs_buf_relse(bp);
2404 error = EIO;
2405 goto error;
2406 }
2407 }
2408
2409write_inode_buffer:
2410 ASSERT(bp->b_target->bt_mount == mp);
2411 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2412 xfs_bdwrite(mp, bp);
2413error:
2414 if (need_free)
2415 kmem_free(in_f);
2416 return XFS_ERROR(error);
2417}
2418
2419
2420
2421
2422
2423
2424STATIC int
2425xlog_recover_quotaoff_pass1(
2426 xlog_t *log,
2427 xlog_recover_item_t *item)
2428{
2429 xfs_qoff_logformat_t *qoff_f = item->ri_buf[0].i_addr;
2430 ASSERT(qoff_f);
2431
2432
2433
2434
2435
2436 if (qoff_f->qf_flags & XFS_UQUOTA_ACCT)
2437 log->l_quotaoffs_flag |= XFS_DQ_USER;
2438 if (qoff_f->qf_flags & XFS_PQUOTA_ACCT)
2439 log->l_quotaoffs_flag |= XFS_DQ_PROJ;
2440 if (qoff_f->qf_flags & XFS_GQUOTA_ACCT)
2441 log->l_quotaoffs_flag |= XFS_DQ_GROUP;
2442
2443 return (0);
2444}
2445
2446
2447
2448
2449STATIC int
2450xlog_recover_dquot_pass2(
2451 xlog_t *log,
2452 xlog_recover_item_t *item)
2453{
2454 xfs_mount_t *mp = log->l_mp;
2455 xfs_buf_t *bp;
2456 struct xfs_disk_dquot *ddq, *recddq;
2457 int error;
2458 xfs_dq_logformat_t *dq_f;
2459 uint type;
2460
2461
2462
2463
2464
2465 if (mp->m_qflags == 0)
2466 return (0);
2467
2468 recddq = item->ri_buf[1].i_addr;
2469 if (recddq == NULL) {
2470 cmn_err(CE_ALERT,
2471 "XFS: NULL dquot in %s.", __func__);
2472 return XFS_ERROR(EIO);
2473 }
2474 if (item->ri_buf[1].i_len < sizeof(xfs_disk_dquot_t)) {
2475 cmn_err(CE_ALERT,
2476 "XFS: dquot too small (%d) in %s.",
2477 item->ri_buf[1].i_len, __func__);
2478 return XFS_ERROR(EIO);
2479 }
2480
2481
2482
2483
2484 type = recddq->d_flags & (XFS_DQ_USER | XFS_DQ_PROJ | XFS_DQ_GROUP);
2485 ASSERT(type);
2486 if (log->l_quotaoffs_flag & type)
2487 return (0);
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499 dq_f = item->ri_buf[0].i_addr;
2500 ASSERT(dq_f);
2501 if ((error = xfs_qm_dqcheck(recddq,
2502 dq_f->qlf_id,
2503 0, XFS_QMOPT_DOWARN,
2504 "xlog_recover_dquot_pass2 (log copy)"))) {
2505 return XFS_ERROR(EIO);
2506 }
2507 ASSERT(dq_f->qlf_len == 1);
2508
2509 error = xfs_read_buf(mp, mp->m_ddev_targp,
2510 dq_f->qlf_blkno,
2511 XFS_FSB_TO_BB(mp, dq_f->qlf_len),
2512 0, &bp);
2513 if (error) {
2514 xfs_ioerror_alert("xlog_recover_do..(read#3)", mp,
2515 bp, dq_f->qlf_blkno);
2516 return error;
2517 }
2518 ASSERT(bp);
2519 ddq = (xfs_disk_dquot_t *)xfs_buf_offset(bp, dq_f->qlf_boffset);
2520
2521
2522
2523
2524
2525
2526 if (xfs_qm_dqcheck(ddq, dq_f->qlf_id, 0, XFS_QMOPT_DOWARN,
2527 "xlog_recover_dquot_pass2")) {
2528 xfs_buf_relse(bp);
2529 return XFS_ERROR(EIO);
2530 }
2531
2532 memcpy(ddq, recddq, item->ri_buf[1].i_len);
2533
2534 ASSERT(dq_f->qlf_size == 2);
2535 ASSERT(bp->b_target->bt_mount == mp);
2536 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2537 xfs_bdwrite(mp, bp);
2538
2539 return (0);
2540}
2541
2542
2543
2544
2545
2546
2547
2548
2549STATIC int
2550xlog_recover_efi_pass2(
2551 xlog_t *log,
2552 xlog_recover_item_t *item,
2553 xfs_lsn_t lsn)
2554{
2555 int error;
2556 xfs_mount_t *mp = log->l_mp;
2557 xfs_efi_log_item_t *efip;
2558 xfs_efi_log_format_t *efi_formatp;
2559
2560 efi_formatp = item->ri_buf[0].i_addr;
2561
2562 efip = xfs_efi_init(mp, efi_formatp->efi_nextents);
2563 if ((error = xfs_efi_copy_format(&(item->ri_buf[0]),
2564 &(efip->efi_format)))) {
2565 xfs_efi_item_free(efip);
2566 return error;
2567 }
2568 atomic_set(&efip->efi_next_extent, efi_formatp->efi_nextents);
2569
2570 spin_lock(&log->l_ailp->xa_lock);
2571
2572
2573
2574 xfs_trans_ail_update(log->l_ailp, &efip->efi_item, lsn);
2575 return 0;
2576}
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587STATIC int
2588xlog_recover_efd_pass2(
2589 xlog_t *log,
2590 xlog_recover_item_t *item)
2591{
2592 xfs_efd_log_format_t *efd_formatp;
2593 xfs_efi_log_item_t *efip = NULL;
2594 xfs_log_item_t *lip;
2595 __uint64_t efi_id;
2596 struct xfs_ail_cursor cur;
2597 struct xfs_ail *ailp = log->l_ailp;
2598
2599 efd_formatp = item->ri_buf[0].i_addr;
2600 ASSERT((item->ri_buf[0].i_len == (sizeof(xfs_efd_log_format_32_t) +
2601 ((efd_formatp->efd_nextents - 1) * sizeof(xfs_extent_32_t)))) ||
2602 (item->ri_buf[0].i_len == (sizeof(xfs_efd_log_format_64_t) +
2603 ((efd_formatp->efd_nextents - 1) * sizeof(xfs_extent_64_t)))));
2604 efi_id = efd_formatp->efd_efi_id;
2605
2606
2607
2608
2609
2610 spin_lock(&ailp->xa_lock);
2611 lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
2612 while (lip != NULL) {
2613 if (lip->li_type == XFS_LI_EFI) {
2614 efip = (xfs_efi_log_item_t *)lip;
2615 if (efip->efi_format.efi_id == efi_id) {
2616
2617
2618
2619
2620 xfs_trans_ail_delete(ailp, lip);
2621 xfs_efi_item_free(efip);
2622 spin_lock(&ailp->xa_lock);
2623 break;
2624 }
2625 }
2626 lip = xfs_trans_ail_cursor_next(ailp, &cur);
2627 }
2628 xfs_trans_ail_cursor_done(ailp, &cur);
2629 spin_unlock(&ailp->xa_lock);
2630
2631 return 0;
2632}
2633
2634
2635
2636
2637
2638
2639STATIC void
2640xlog_recover_free_trans(
2641 struct xlog_recover *trans)
2642{
2643 xlog_recover_item_t *item, *n;
2644 int i;
2645
2646 list_for_each_entry_safe(item, n, &trans->r_itemq, ri_list) {
2647
2648 list_del(&item->ri_list);
2649 for (i = 0; i < item->ri_cnt; i++)
2650 kmem_free(item->ri_buf[i].i_addr);
2651
2652 kmem_free(item->ri_buf);
2653 kmem_free(item);
2654 }
2655
2656 kmem_free(trans);
2657}
2658
2659STATIC int
2660xlog_recover_commit_pass1(
2661 struct log *log,
2662 struct xlog_recover *trans,
2663 xlog_recover_item_t *item)
2664{
2665 trace_xfs_log_recover_item_recover(log, trans, item, XLOG_RECOVER_PASS1);
2666
2667 switch (ITEM_TYPE(item)) {
2668 case XFS_LI_BUF:
2669 return xlog_recover_buffer_pass1(log, item);
2670 case XFS_LI_QUOTAOFF:
2671 return xlog_recover_quotaoff_pass1(log, item);
2672 case XFS_LI_INODE:
2673 case XFS_LI_EFI:
2674 case XFS_LI_EFD:
2675 case XFS_LI_DQUOT:
2676
2677 return 0;
2678 default:
2679 xlog_warn(
2680 "XFS: invalid item type (%d) xlog_recover_commit_pass1",
2681 ITEM_TYPE(item));
2682 ASSERT(0);
2683 return XFS_ERROR(EIO);
2684 }
2685}
2686
2687STATIC int
2688xlog_recover_commit_pass2(
2689 struct log *log,
2690 struct xlog_recover *trans,
2691 xlog_recover_item_t *item)
2692{
2693 trace_xfs_log_recover_item_recover(log, trans, item, XLOG_RECOVER_PASS2);
2694
2695 switch (ITEM_TYPE(item)) {
2696 case XFS_LI_BUF:
2697 return xlog_recover_buffer_pass2(log, item);
2698 case XFS_LI_INODE:
2699 return xlog_recover_inode_pass2(log, item);
2700 case XFS_LI_EFI:
2701 return xlog_recover_efi_pass2(log, item, trans->r_lsn);
2702 case XFS_LI_EFD:
2703 return xlog_recover_efd_pass2(log, item);
2704 case XFS_LI_DQUOT:
2705 return xlog_recover_dquot_pass2(log, item);
2706 case XFS_LI_QUOTAOFF:
2707
2708 return 0;
2709 default:
2710 xlog_warn(
2711 "XFS: invalid item type (%d) xlog_recover_commit_pass2",
2712 ITEM_TYPE(item));
2713 ASSERT(0);
2714 return XFS_ERROR(EIO);
2715 }
2716}
2717
2718
2719
2720
2721
2722
2723
2724STATIC int
2725xlog_recover_commit_trans(
2726 struct log *log,
2727 struct xlog_recover *trans,
2728 int pass)
2729{
2730 int error = 0;
2731 xlog_recover_item_t *item;
2732
2733 hlist_del(&trans->r_list);
2734
2735 error = xlog_recover_reorder_trans(log, trans, pass);
2736 if (error)
2737 return error;
2738
2739 list_for_each_entry(item, &trans->r_itemq, ri_list) {
2740 if (pass == XLOG_RECOVER_PASS1)
2741 error = xlog_recover_commit_pass1(log, trans, item);
2742 else
2743 error = xlog_recover_commit_pass2(log, trans, item);
2744 if (error)
2745 return error;
2746 }
2747
2748 xlog_recover_free_trans(trans);
2749 return 0;
2750}
2751
2752STATIC int
2753xlog_recover_unmount_trans(
2754 xlog_recover_t *trans)
2755{
2756
2757 xlog_warn("XFS: xlog_recover_unmount_trans: Unmount LR");
2758 return 0;
2759}
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770STATIC int
2771xlog_recover_process_data(
2772 xlog_t *log,
2773 struct hlist_head rhash[],
2774 xlog_rec_header_t *rhead,
2775 xfs_caddr_t dp,
2776 int pass)
2777{
2778 xfs_caddr_t lp;
2779 int num_logops;
2780 xlog_op_header_t *ohead;
2781 xlog_recover_t *trans;
2782 xlog_tid_t tid;
2783 int error;
2784 unsigned long hash;
2785 uint flags;
2786
2787 lp = dp + be32_to_cpu(rhead->h_len);
2788 num_logops = be32_to_cpu(rhead->h_num_logops);
2789
2790
2791 if (xlog_header_check_recover(log->l_mp, rhead))
2792 return (XFS_ERROR(EIO));
2793
2794 while ((dp < lp) && num_logops) {
2795 ASSERT(dp + sizeof(xlog_op_header_t) <= lp);
2796 ohead = (xlog_op_header_t *)dp;
2797 dp += sizeof(xlog_op_header_t);
2798 if (ohead->oh_clientid != XFS_TRANSACTION &&
2799 ohead->oh_clientid != XFS_LOG) {
2800 xlog_warn(
2801 "XFS: xlog_recover_process_data: bad clientid");
2802 ASSERT(0);
2803 return (XFS_ERROR(EIO));
2804 }
2805 tid = be32_to_cpu(ohead->oh_tid);
2806 hash = XLOG_RHASH(tid);
2807 trans = xlog_recover_find_tid(&rhash[hash], tid);
2808 if (trans == NULL) {
2809 if (ohead->oh_flags & XLOG_START_TRANS)
2810 xlog_recover_new_tid(&rhash[hash], tid,
2811 be64_to_cpu(rhead->h_lsn));
2812 } else {
2813 if (dp + be32_to_cpu(ohead->oh_len) > lp) {
2814 xlog_warn(
2815 "XFS: xlog_recover_process_data: bad length");
2816 WARN_ON(1);
2817 return (XFS_ERROR(EIO));
2818 }
2819 flags = ohead->oh_flags & ~XLOG_END_TRANS;
2820 if (flags & XLOG_WAS_CONT_TRANS)
2821 flags &= ~XLOG_CONTINUE_TRANS;
2822 switch (flags) {
2823 case XLOG_COMMIT_TRANS:
2824 error = xlog_recover_commit_trans(log,
2825 trans, pass);
2826 break;
2827 case XLOG_UNMOUNT_TRANS:
2828 error = xlog_recover_unmount_trans(trans);
2829 break;
2830 case XLOG_WAS_CONT_TRANS:
2831 error = xlog_recover_add_to_cont_trans(log,
2832 trans, dp,
2833 be32_to_cpu(ohead->oh_len));
2834 break;
2835 case XLOG_START_TRANS:
2836 xlog_warn(
2837 "XFS: xlog_recover_process_data: bad transaction");
2838 ASSERT(0);
2839 error = XFS_ERROR(EIO);
2840 break;
2841 case 0:
2842 case XLOG_CONTINUE_TRANS:
2843 error = xlog_recover_add_to_trans(log, trans,
2844 dp, be32_to_cpu(ohead->oh_len));
2845 break;
2846 default:
2847 xlog_warn(
2848 "XFS: xlog_recover_process_data: bad flag");
2849 ASSERT(0);
2850 error = XFS_ERROR(EIO);
2851 break;
2852 }
2853 if (error)
2854 return error;
2855 }
2856 dp += be32_to_cpu(ohead->oh_len);
2857 num_logops--;
2858 }
2859 return 0;
2860}
2861
2862
2863
2864
2865
2866STATIC int
2867xlog_recover_process_efi(
2868 xfs_mount_t *mp,
2869 xfs_efi_log_item_t *efip)
2870{
2871 xfs_efd_log_item_t *efdp;
2872 xfs_trans_t *tp;
2873 int i;
2874 int error = 0;
2875 xfs_extent_t *extp;
2876 xfs_fsblock_t startblock_fsb;
2877
2878 ASSERT(!test_bit(XFS_EFI_RECOVERED, &efip->efi_flags));
2879
2880
2881
2882
2883
2884
2885 for (i = 0; i < efip->efi_format.efi_nextents; i++) {
2886 extp = &(efip->efi_format.efi_extents[i]);
2887 startblock_fsb = XFS_BB_TO_FSB(mp,
2888 XFS_FSB_TO_DADDR(mp, extp->ext_start));
2889 if ((startblock_fsb == 0) ||
2890 (extp->ext_len == 0) ||
2891 (startblock_fsb >= mp->m_sb.sb_dblocks) ||
2892 (extp->ext_len >= mp->m_sb.sb_agblocks)) {
2893
2894
2895
2896
2897 xfs_efi_release(efip, efip->efi_format.efi_nextents);
2898 return XFS_ERROR(EIO);
2899 }
2900 }
2901
2902 tp = xfs_trans_alloc(mp, 0);
2903 error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0, 0, 0);
2904 if (error)
2905 goto abort_error;
2906 efdp = xfs_trans_get_efd(tp, efip, efip->efi_format.efi_nextents);
2907
2908 for (i = 0; i < efip->efi_format.efi_nextents; i++) {
2909 extp = &(efip->efi_format.efi_extents[i]);
2910 error = xfs_free_extent(tp, extp->ext_start, extp->ext_len);
2911 if (error)
2912 goto abort_error;
2913 xfs_trans_log_efd_extent(tp, efdp, extp->ext_start,
2914 extp->ext_len);
2915 }
2916
2917 set_bit(XFS_EFI_RECOVERED, &efip->efi_flags);
2918 error = xfs_trans_commit(tp, 0);
2919 return error;
2920
2921abort_error:
2922 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
2923 return error;
2924}
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944STATIC int
2945xlog_recover_process_efis(
2946 xlog_t *log)
2947{
2948 xfs_log_item_t *lip;
2949 xfs_efi_log_item_t *efip;
2950 int error = 0;
2951 struct xfs_ail_cursor cur;
2952 struct xfs_ail *ailp;
2953
2954 ailp = log->l_ailp;
2955 spin_lock(&ailp->xa_lock);
2956 lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
2957 while (lip != NULL) {
2958
2959
2960
2961
2962 if (lip->li_type != XFS_LI_EFI) {
2963#ifdef DEBUG
2964 for (; lip; lip = xfs_trans_ail_cursor_next(ailp, &cur))
2965 ASSERT(lip->li_type != XFS_LI_EFI);
2966#endif
2967 break;
2968 }
2969
2970
2971
2972
2973 efip = (xfs_efi_log_item_t *)lip;
2974 if (test_bit(XFS_EFI_RECOVERED, &efip->efi_flags)) {
2975 lip = xfs_trans_ail_cursor_next(ailp, &cur);
2976 continue;
2977 }
2978
2979 spin_unlock(&ailp->xa_lock);
2980 error = xlog_recover_process_efi(log->l_mp, efip);
2981 spin_lock(&ailp->xa_lock);
2982 if (error)
2983 goto out;
2984 lip = xfs_trans_ail_cursor_next(ailp, &cur);
2985 }
2986out:
2987 xfs_trans_ail_cursor_done(ailp, &cur);
2988 spin_unlock(&ailp->xa_lock);
2989 return error;
2990}
2991
2992
2993
2994
2995
2996STATIC void
2997xlog_recover_clear_agi_bucket(
2998 xfs_mount_t *mp,
2999 xfs_agnumber_t agno,
3000 int bucket)
3001{
3002 xfs_trans_t *tp;
3003 xfs_agi_t *agi;
3004 xfs_buf_t *agibp;
3005 int offset;
3006 int error;
3007
3008 tp = xfs_trans_alloc(mp, XFS_TRANS_CLEAR_AGI_BUCKET);
3009 error = xfs_trans_reserve(tp, 0, XFS_CLEAR_AGI_BUCKET_LOG_RES(mp),
3010 0, 0, 0);
3011 if (error)
3012 goto out_abort;
3013
3014 error = xfs_read_agi(mp, tp, agno, &agibp);
3015 if (error)
3016 goto out_abort;
3017
3018 agi = XFS_BUF_TO_AGI(agibp);
3019 agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
3020 offset = offsetof(xfs_agi_t, agi_unlinked) +
3021 (sizeof(xfs_agino_t) * bucket);
3022 xfs_trans_log_buf(tp, agibp, offset,
3023 (offset + sizeof(xfs_agino_t) - 1));
3024
3025 error = xfs_trans_commit(tp, 0);
3026 if (error)
3027 goto out_error;
3028 return;
3029
3030out_abort:
3031 xfs_trans_cancel(tp, XFS_TRANS_ABORT);
3032out_error:
3033 xfs_fs_cmn_err(CE_WARN, mp, "xlog_recover_clear_agi_bucket: "
3034 "failed to clear agi %d. Continuing.", agno);
3035 return;
3036}
3037
3038STATIC xfs_agino_t
3039xlog_recover_process_one_iunlink(
3040 struct xfs_mount *mp,
3041 xfs_agnumber_t agno,
3042 xfs_agino_t agino,
3043 int bucket)
3044{
3045 struct xfs_buf *ibp;
3046 struct xfs_dinode *dip;
3047 struct xfs_inode *ip;
3048 xfs_ino_t ino;
3049 int error;
3050
3051 ino = XFS_AGINO_TO_INO(mp, agno, agino);
3052 error = xfs_iget(mp, NULL, ino, 0, 0, &ip);
3053 if (error)
3054 goto fail;
3055
3056
3057
3058
3059 error = xfs_itobp(mp, NULL, ip, &dip, &ibp, XBF_LOCK);
3060 if (error)
3061 goto fail_iput;
3062
3063 ASSERT(ip->i_d.di_nlink == 0);
3064 ASSERT(ip->i_d.di_mode != 0);
3065
3066
3067 agino = be32_to_cpu(dip->di_next_unlinked);
3068 xfs_buf_relse(ibp);
3069
3070
3071
3072
3073
3074 ip->i_d.di_dmevmask = 0;
3075
3076 IRELE(ip);
3077 return agino;
3078
3079 fail_iput:
3080 IRELE(ip);
3081 fail:
3082
3083
3084
3085
3086
3087
3088
3089
3090 xlog_recover_clear_agi_bucket(mp, agno, bucket);
3091 return NULLAGINO;
3092}
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106STATIC void
3107xlog_recover_process_iunlinks(
3108 xlog_t *log)
3109{
3110 xfs_mount_t *mp;
3111 xfs_agnumber_t agno;
3112 xfs_agi_t *agi;
3113 xfs_buf_t *agibp;
3114 xfs_agino_t agino;
3115 int bucket;
3116 int error;
3117 uint mp_dmevmask;
3118
3119 mp = log->l_mp;
3120
3121
3122
3123
3124 mp_dmevmask = mp->m_dmevmask;
3125 mp->m_dmevmask = 0;
3126
3127 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
3128
3129
3130
3131 error = xfs_read_agi(mp, NULL, agno, &agibp);
3132 if (error) {
3133
3134
3135
3136
3137
3138
3139 continue;
3140 }
3141 agi = XFS_BUF_TO_AGI(agibp);
3142
3143 for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++) {
3144 agino = be32_to_cpu(agi->agi_unlinked[bucket]);
3145 while (agino != NULLAGINO) {
3146
3147
3148
3149
3150
3151 xfs_buf_relse(agibp);
3152
3153 agino = xlog_recover_process_one_iunlink(mp,
3154 agno, agino, bucket);
3155
3156
3157
3158
3159
3160
3161 error = xfs_read_agi(mp, NULL, agno, &agibp);
3162 ASSERT(error == 0);
3163 agi = XFS_BUF_TO_AGI(agibp);
3164 }
3165 }
3166
3167
3168
3169
3170
3171 xfs_buf_relse(agibp);
3172 }
3173
3174 mp->m_dmevmask = mp_dmevmask;
3175}
3176
3177
3178#ifdef DEBUG
3179STATIC void
3180xlog_pack_data_checksum(
3181 xlog_t *log,
3182 xlog_in_core_t *iclog,
3183 int size)
3184{
3185 int i;
3186 __be32 *up;
3187 uint chksum = 0;
3188
3189 up = (__be32 *)iclog->ic_datap;
3190
3191 for (i = 0; i < (size >> 2); i++) {
3192 chksum ^= be32_to_cpu(*up);
3193 up++;
3194 }
3195 iclog->ic_header.h_chksum = cpu_to_be32(chksum);
3196}
3197#else
3198#define xlog_pack_data_checksum(log, iclog, size)
3199#endif
3200
3201
3202
3203
3204void
3205xlog_pack_data(
3206 xlog_t *log,
3207 xlog_in_core_t *iclog,
3208 int roundoff)
3209{
3210 int i, j, k;
3211 int size = iclog->ic_offset + roundoff;
3212 __be32 cycle_lsn;
3213 xfs_caddr_t dp;
3214
3215 xlog_pack_data_checksum(log, iclog, size);
3216
3217 cycle_lsn = CYCLE_LSN_DISK(iclog->ic_header.h_lsn);
3218
3219 dp = iclog->ic_datap;
3220 for (i = 0; i < BTOBB(size) &&
3221 i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
3222 iclog->ic_header.h_cycle_data[i] = *(__be32 *)dp;
3223 *(__be32 *)dp = cycle_lsn;
3224 dp += BBSIZE;
3225 }
3226
3227 if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
3228 xlog_in_core_2_t *xhdr = iclog->ic_data;
3229
3230 for ( ; i < BTOBB(size); i++) {
3231 j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
3232 k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
3233 xhdr[j].hic_xheader.xh_cycle_data[k] = *(__be32 *)dp;
3234 *(__be32 *)dp = cycle_lsn;
3235 dp += BBSIZE;
3236 }
3237
3238 for (i = 1; i < log->l_iclog_heads; i++) {
3239 xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
3240 }
3241 }
3242}
3243
3244STATIC void
3245xlog_unpack_data(
3246 xlog_rec_header_t *rhead,
3247 xfs_caddr_t dp,
3248 xlog_t *log)
3249{
3250 int i, j, k;
3251
3252 for (i = 0; i < BTOBB(be32_to_cpu(rhead->h_len)) &&
3253 i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
3254 *(__be32 *)dp = *(__be32 *)&rhead->h_cycle_data[i];
3255 dp += BBSIZE;
3256 }
3257
3258 if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
3259 xlog_in_core_2_t *xhdr = (xlog_in_core_2_t *)rhead;
3260 for ( ; i < BTOBB(be32_to_cpu(rhead->h_len)); i++) {
3261 j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
3262 k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
3263 *(__be32 *)dp = xhdr[j].hic_xheader.xh_cycle_data[k];
3264 dp += BBSIZE;
3265 }
3266 }
3267}
3268
3269STATIC int
3270xlog_valid_rec_header(
3271 xlog_t *log,
3272 xlog_rec_header_t *rhead,
3273 xfs_daddr_t blkno)
3274{
3275 int hlen;
3276
3277 if (unlikely(be32_to_cpu(rhead->h_magicno) != XLOG_HEADER_MAGIC_NUM)) {
3278 XFS_ERROR_REPORT("xlog_valid_rec_header(1)",
3279 XFS_ERRLEVEL_LOW, log->l_mp);
3280 return XFS_ERROR(EFSCORRUPTED);
3281 }
3282 if (unlikely(
3283 (!rhead->h_version ||
3284 (be32_to_cpu(rhead->h_version) & (~XLOG_VERSION_OKBITS))))) {
3285 xlog_warn("XFS: %s: unrecognised log version (%d).",
3286 __func__, be32_to_cpu(rhead->h_version));
3287 return XFS_ERROR(EIO);
3288 }
3289
3290
3291 hlen = be32_to_cpu(rhead->h_len);
3292 if (unlikely( hlen <= 0 || hlen > INT_MAX )) {
3293 XFS_ERROR_REPORT("xlog_valid_rec_header(2)",
3294 XFS_ERRLEVEL_LOW, log->l_mp);
3295 return XFS_ERROR(EFSCORRUPTED);
3296 }
3297 if (unlikely( blkno > log->l_logBBsize || blkno > INT_MAX )) {
3298 XFS_ERROR_REPORT("xlog_valid_rec_header(3)",
3299 XFS_ERRLEVEL_LOW, log->l_mp);
3300 return XFS_ERROR(EFSCORRUPTED);
3301 }
3302 return 0;
3303}
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313STATIC int
3314xlog_do_recovery_pass(
3315 xlog_t *log,
3316 xfs_daddr_t head_blk,
3317 xfs_daddr_t tail_blk,
3318 int pass)
3319{
3320 xlog_rec_header_t *rhead;
3321 xfs_daddr_t blk_no;
3322 xfs_caddr_t offset;
3323 xfs_buf_t *hbp, *dbp;
3324 int error = 0, h_size;
3325 int bblks, split_bblks;
3326 int hblks, split_hblks, wrapped_hblks;
3327 struct hlist_head rhash[XLOG_RHASH_SIZE];
3328
3329 ASSERT(head_blk != tail_blk);
3330
3331
3332
3333
3334
3335 if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
3336
3337
3338
3339
3340
3341 hbp = xlog_get_bp(log, 1);
3342 if (!hbp)
3343 return ENOMEM;
3344
3345 error = xlog_bread(log, tail_blk, 1, hbp, &offset);
3346 if (error)
3347 goto bread_err1;
3348
3349 rhead = (xlog_rec_header_t *)offset;
3350 error = xlog_valid_rec_header(log, rhead, tail_blk);
3351 if (error)
3352 goto bread_err1;
3353 h_size = be32_to_cpu(rhead->h_size);
3354 if ((be32_to_cpu(rhead->h_version) & XLOG_VERSION_2) &&
3355 (h_size > XLOG_HEADER_CYCLE_SIZE)) {
3356 hblks = h_size / XLOG_HEADER_CYCLE_SIZE;
3357 if (h_size % XLOG_HEADER_CYCLE_SIZE)
3358 hblks++;
3359 xlog_put_bp(hbp);
3360 hbp = xlog_get_bp(log, hblks);
3361 } else {
3362 hblks = 1;
3363 }
3364 } else {
3365 ASSERT(log->l_sectBBsize == 1);
3366 hblks = 1;
3367 hbp = xlog_get_bp(log, 1);
3368 h_size = XLOG_BIG_RECORD_BSIZE;
3369 }
3370
3371 if (!hbp)
3372 return ENOMEM;
3373 dbp = xlog_get_bp(log, BTOBB(h_size));
3374 if (!dbp) {
3375 xlog_put_bp(hbp);
3376 return ENOMEM;
3377 }
3378
3379 memset(rhash, 0, sizeof(rhash));
3380 if (tail_blk <= head_blk) {
3381 for (blk_no = tail_blk; blk_no < head_blk; ) {
3382 error = xlog_bread(log, blk_no, hblks, hbp, &offset);
3383 if (error)
3384 goto bread_err2;
3385
3386 rhead = (xlog_rec_header_t *)offset;
3387 error = xlog_valid_rec_header(log, rhead, blk_no);
3388 if (error)
3389 goto bread_err2;
3390
3391
3392 bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
3393 error = xlog_bread(log, blk_no + hblks, bblks, dbp,
3394 &offset);
3395 if (error)
3396 goto bread_err2;
3397
3398 xlog_unpack_data(rhead, offset, log);
3399 if ((error = xlog_recover_process_data(log,
3400 rhash, rhead, offset, pass)))
3401 goto bread_err2;
3402 blk_no += bblks + hblks;
3403 }
3404 } else {
3405
3406
3407
3408
3409
3410 blk_no = tail_blk;
3411 while (blk_no < log->l_logBBsize) {
3412
3413
3414
3415 offset = XFS_BUF_PTR(hbp);
3416 split_hblks = 0;
3417 wrapped_hblks = 0;
3418 if (blk_no + hblks <= log->l_logBBsize) {
3419
3420 error = xlog_bread(log, blk_no, hblks, hbp,
3421 &offset);
3422 if (error)
3423 goto bread_err2;
3424 } else {
3425
3426 if (blk_no != log->l_logBBsize) {
3427
3428 ASSERT(blk_no <= INT_MAX);
3429 split_hblks = log->l_logBBsize - (int)blk_no;
3430 ASSERT(split_hblks > 0);
3431 error = xlog_bread(log, blk_no,
3432 split_hblks, hbp,
3433 &offset);
3434 if (error)
3435 goto bread_err2;
3436 }
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450 wrapped_hblks = hblks - split_hblks;
3451 error = XFS_BUF_SET_PTR(hbp,
3452 offset + BBTOB(split_hblks),
3453 BBTOB(hblks - split_hblks));
3454 if (error)
3455 goto bread_err2;
3456
3457 error = xlog_bread_noalign(log, 0,
3458 wrapped_hblks, hbp);
3459 if (error)
3460 goto bread_err2;
3461
3462 error = XFS_BUF_SET_PTR(hbp, offset,
3463 BBTOB(hblks));
3464 if (error)
3465 goto bread_err2;
3466 }
3467 rhead = (xlog_rec_header_t *)offset;
3468 error = xlog_valid_rec_header(log, rhead,
3469 split_hblks ? blk_no : 0);
3470 if (error)
3471 goto bread_err2;
3472
3473 bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
3474 blk_no += hblks;
3475
3476
3477 if (blk_no + bblks <= log->l_logBBsize) {
3478 error = xlog_bread(log, blk_no, bblks, dbp,
3479 &offset);
3480 if (error)
3481 goto bread_err2;
3482 } else {
3483
3484
3485 offset = XFS_BUF_PTR(dbp);
3486 split_bblks = 0;
3487 if (blk_no != log->l_logBBsize) {
3488
3489
3490 ASSERT(!wrapped_hblks);
3491 ASSERT(blk_no <= INT_MAX);
3492 split_bblks =
3493 log->l_logBBsize - (int)blk_no;
3494 ASSERT(split_bblks > 0);
3495 error = xlog_bread(log, blk_no,
3496 split_bblks, dbp,
3497 &offset);
3498 if (error)
3499 goto bread_err2;
3500 }
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514 error = XFS_BUF_SET_PTR(dbp,
3515 offset + BBTOB(split_bblks),
3516 BBTOB(bblks - split_bblks));
3517 if (error)
3518 goto bread_err2;
3519
3520 error = xlog_bread_noalign(log, wrapped_hblks,
3521 bblks - split_bblks,
3522 dbp);
3523 if (error)
3524 goto bread_err2;
3525
3526 error = XFS_BUF_SET_PTR(dbp, offset, h_size);
3527 if (error)
3528 goto bread_err2;
3529 }
3530 xlog_unpack_data(rhead, offset, log);
3531 if ((error = xlog_recover_process_data(log, rhash,
3532 rhead, offset, pass)))
3533 goto bread_err2;
3534 blk_no += bblks;
3535 }
3536
3537 ASSERT(blk_no >= log->l_logBBsize);
3538 blk_no -= log->l_logBBsize;
3539
3540
3541 while (blk_no < head_blk) {
3542 error = xlog_bread(log, blk_no, hblks, hbp, &offset);
3543 if (error)
3544 goto bread_err2;
3545
3546 rhead = (xlog_rec_header_t *)offset;
3547 error = xlog_valid_rec_header(log, rhead, blk_no);
3548 if (error)
3549 goto bread_err2;
3550
3551 bblks = (int)BTOBB(be32_to_cpu(rhead->h_len));
3552 error = xlog_bread(log, blk_no+hblks, bblks, dbp,
3553 &offset);
3554 if (error)
3555 goto bread_err2;
3556
3557 xlog_unpack_data(rhead, offset, log);
3558 if ((error = xlog_recover_process_data(log, rhash,
3559 rhead, offset, pass)))
3560 goto bread_err2;
3561 blk_no += bblks + hblks;
3562 }
3563 }
3564
3565 bread_err2:
3566 xlog_put_bp(dbp);
3567 bread_err1:
3568 xlog_put_bp(hbp);
3569 return error;
3570}
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585STATIC int
3586xlog_do_log_recovery(
3587 xlog_t *log,
3588 xfs_daddr_t head_blk,
3589 xfs_daddr_t tail_blk)
3590{
3591 int error, i;
3592
3593 ASSERT(head_blk != tail_blk);
3594
3595
3596
3597
3598
3599 log->l_buf_cancel_table = kmem_zalloc(XLOG_BC_TABLE_SIZE *
3600 sizeof(struct list_head),
3601 KM_SLEEP);
3602 for (i = 0; i < XLOG_BC_TABLE_SIZE; i++)
3603 INIT_LIST_HEAD(&log->l_buf_cancel_table[i]);
3604
3605 error = xlog_do_recovery_pass(log, head_blk, tail_blk,
3606 XLOG_RECOVER_PASS1);
3607 if (error != 0) {
3608 kmem_free(log->l_buf_cancel_table);
3609 log->l_buf_cancel_table = NULL;
3610 return error;
3611 }
3612
3613
3614
3615
3616 error = xlog_do_recovery_pass(log, head_blk, tail_blk,
3617 XLOG_RECOVER_PASS2);
3618#ifdef DEBUG
3619 if (!error) {
3620 int i;
3621
3622 for (i = 0; i < XLOG_BC_TABLE_SIZE; i++)
3623 ASSERT(list_empty(&log->l_buf_cancel_table[i]));
3624 }
3625#endif
3626
3627 kmem_free(log->l_buf_cancel_table);
3628 log->l_buf_cancel_table = NULL;
3629
3630 return error;
3631}
3632
3633
3634
3635
3636STATIC int
3637xlog_do_recover(
3638 xlog_t *log,
3639 xfs_daddr_t head_blk,
3640 xfs_daddr_t tail_blk)
3641{
3642 int error;
3643 xfs_buf_t *bp;
3644 xfs_sb_t *sbp;
3645
3646
3647
3648
3649 error = xlog_do_log_recovery(log, head_blk, tail_blk);
3650 if (error) {
3651 return error;
3652 }
3653
3654 XFS_bflush(log->l_mp->m_ddev_targp);
3655
3656
3657
3658
3659 if (XFS_FORCED_SHUTDOWN(log->l_mp)) {
3660 return (EIO);
3661 }
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672 xlog_assign_tail_lsn(log->l_mp);
3673
3674
3675
3676
3677
3678 bp = xfs_getsb(log->l_mp, 0);
3679 XFS_BUF_UNDONE(bp);
3680 ASSERT(!(XFS_BUF_ISWRITE(bp)));
3681 ASSERT(!(XFS_BUF_ISDELAYWRITE(bp)));
3682 XFS_BUF_READ(bp);
3683 XFS_BUF_UNASYNC(bp);
3684 xfsbdstrat(log->l_mp, bp);
3685 error = xfs_buf_iowait(bp);
3686 if (error) {
3687 xfs_ioerror_alert("xlog_do_recover",
3688 log->l_mp, bp, XFS_BUF_ADDR(bp));
3689 ASSERT(0);
3690 xfs_buf_relse(bp);
3691 return error;
3692 }
3693
3694
3695 sbp = &log->l_mp->m_sb;
3696 xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
3697 ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
3698 ASSERT(xfs_sb_good_version(sbp));
3699 xfs_buf_relse(bp);
3700
3701
3702 xfs_icsb_reinit_counters(log->l_mp);
3703
3704 xlog_recover_check_summary(log);
3705
3706
3707 log->l_flags &= ~XLOG_ACTIVE_RECOVERY;
3708 return 0;
3709}
3710
3711
3712
3713
3714
3715
3716int
3717xlog_recover(
3718 xlog_t *log)
3719{
3720 xfs_daddr_t head_blk, tail_blk;
3721 int error;
3722
3723
3724 if ((error = xlog_find_tail(log, &head_blk, &tail_blk)))
3725 return error;
3726
3727 if (tail_blk != head_blk) {
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739 if ((error = xfs_dev_is_read_only(log->l_mp, "recovery"))) {
3740 return error;
3741 }
3742
3743 cmn_err(CE_NOTE,
3744 "Starting XFS recovery on filesystem: %s (logdev: %s)",
3745 log->l_mp->m_fsname, log->l_mp->m_logname ?
3746 log->l_mp->m_logname : "internal");
3747
3748 error = xlog_do_recover(log, head_blk, tail_blk);
3749 log->l_flags |= XLOG_RECOVERY_NEEDED;
3750 }
3751 return error;
3752}
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763int
3764xlog_recover_finish(
3765 xlog_t *log)
3766{
3767
3768
3769
3770
3771
3772
3773
3774
3775 if (log->l_flags & XLOG_RECOVERY_NEEDED) {
3776 int error;
3777 error = xlog_recover_process_efis(log);
3778 if (error) {
3779 cmn_err(CE_ALERT,
3780 "Failed to recover EFIs on filesystem: %s",
3781 log->l_mp->m_fsname);
3782 return error;
3783 }
3784
3785
3786
3787
3788
3789
3790 xfs_log_force(log->l_mp, XFS_LOG_SYNC);
3791
3792 xlog_recover_process_iunlinks(log);
3793
3794 xlog_recover_check_summary(log);
3795
3796 cmn_err(CE_NOTE,
3797 "Ending XFS recovery on filesystem: %s (logdev: %s)",
3798 log->l_mp->m_fsname, log->l_mp->m_logname ?
3799 log->l_mp->m_logname : "internal");
3800 log->l_flags &= ~XLOG_RECOVERY_NEEDED;
3801 } else {
3802 cmn_err(CE_DEBUG,
3803 "Ending clean XFS mount for filesystem: %s\n",
3804 log->l_mp->m_fsname);
3805 }
3806 return 0;
3807}
3808
3809
3810#if defined(DEBUG)
3811
3812
3813
3814
3815void
3816xlog_recover_check_summary(
3817 xlog_t *log)
3818{
3819 xfs_mount_t *mp;
3820 xfs_agf_t *agfp;
3821 xfs_buf_t *agfbp;
3822 xfs_buf_t *agibp;
3823 xfs_agnumber_t agno;
3824 __uint64_t freeblks;
3825 __uint64_t itotal;
3826 __uint64_t ifree;
3827 int error;
3828
3829 mp = log->l_mp;
3830
3831 freeblks = 0LL;
3832 itotal = 0LL;
3833 ifree = 0LL;
3834 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
3835 error = xfs_read_agf(mp, NULL, agno, 0, &agfbp);
3836 if (error) {
3837 xfs_fs_cmn_err(CE_ALERT, mp,
3838 "xlog_recover_check_summary(agf)"
3839 "agf read failed agno %d error %d",
3840 agno, error);
3841 } else {
3842 agfp = XFS_BUF_TO_AGF(agfbp);
3843 freeblks += be32_to_cpu(agfp->agf_freeblks) +
3844 be32_to_cpu(agfp->agf_flcount);
3845 xfs_buf_relse(agfbp);
3846 }
3847
3848 error = xfs_read_agi(mp, NULL, agno, &agibp);
3849 if (!error) {
3850 struct xfs_agi *agi = XFS_BUF_TO_AGI(agibp);
3851
3852 itotal += be32_to_cpu(agi->agi_count);
3853 ifree += be32_to_cpu(agi->agi_freecount);
3854 xfs_buf_relse(agibp);
3855 }
3856 }
3857}
3858#endif
3859