aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-10-22 20:32:27 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-22 20:32:27 -0400
commit5fe3a5ae5c09d53b2b3c7a971e1d87ab3a747055 (patch)
tree1e0d3e10c83e456a1678c4e01acb5ff624129202
parent0fc0531e0a2174377a86fd6953ecaa00287d8f70 (diff)
parent39dc948c6921169e13224a97fa53188922acfde8 (diff)
Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs
* 'for-linus' of git://oss.sgi.com/xfs/xfs: (36 commits) xfs: semaphore cleanup xfs: Extend project quotas to support 32bit project ids xfs: remove xfs_buf wrappers xfs: remove xfs_cred.h xfs: remove xfs_globals.h xfs: remove xfs_version.h xfs: remove xfs_refcache.h xfs: fix the xfs_trans_committed xfs: remove unused t_callback field in struct xfs_trans xfs: fix bogus m_maxagi check in xfs_iget xfs: do not use xfs_mod_incore_sb_batch for per-cpu counters xfs: do not use xfs_mod_incore_sb for per-cpu counters xfs: remove XFS_MOUNT_NO_PERCPU_SB xfs: pack xfs_buf structure more tightly xfs: convert buffer cache hash to rbtree xfs: serialise inode reclaim within an AG xfs: batch inode reclaim lookup xfs: implement batched inode lookups for AG walking xfs: split out inode walk inode grabbing xfs: split inode AG walking into separate code for reclaim ...
Diffstat
-rw-r--r--fs/xfs/linux-2.6/xfs_buf.c219
-rw-r--r--fs/xfs/linux-2.6/xfs_buf.h70
-rw-r--r--fs/xfs/linux-2.6/xfs_cred.h28
-rw-r--r--fs/xfs/linux-2.6/xfs_fs_subr.c31
-rw-r--r--fs/xfs/linux-2.6/xfs_globals.c1
-rw-r--r--fs/xfs/linux-2.6/xfs_globals.h23
-rw-r--r--fs/xfs/linux-2.6/xfs_ioctl.c19
-rw-r--r--fs/xfs/linux-2.6/xfs_ioctl32.c5
-rw-r--r--fs/xfs/linux-2.6/xfs_ioctl32.h6
-rw-r--r--fs/xfs/linux-2.6/xfs_iops.c39
-rw-r--r--fs/xfs/linux-2.6/xfs_linux.h5
-rw-r--r--fs/xfs/linux-2.6/xfs_super.c24
-rw-r--r--fs/xfs/linux-2.6/xfs_super.h1
-rw-r--r--fs/xfs/linux-2.6/xfs_sync.c413
-rw-r--r--fs/xfs/linux-2.6/xfs_sync.h4
-rw-r--r--fs/xfs/linux-2.6/xfs_trace.h4
-rw-r--r--fs/xfs/linux-2.6/xfs_version.h29
-rw-r--r--fs/xfs/quota/xfs_dquot.c164
-rw-r--r--fs/xfs/quota/xfs_qm.c221
-rw-r--r--fs/xfs/quota/xfs_qm_bhv.c2
-rw-r--r--fs/xfs/quota/xfs_qm_syscalls.c16
-rw-r--r--fs/xfs/xfs_ag.h9
-rw-r--r--fs/xfs/xfs_alloc.c4
-rw-r--r--fs/xfs/xfs_alloc_btree.c33
-rw-r--r--fs/xfs/xfs_attr.c37
-rw-r--r--fs/xfs/xfs_bmap.c44
-rw-r--r--fs/xfs/xfs_bmap.h9
-rw-r--r--fs/xfs/xfs_btree.c56
-rw-r--r--fs/xfs/xfs_btree.h14
-rw-r--r--fs/xfs/xfs_buf_item.c7
-rw-r--r--fs/xfs/xfs_da_btree.c2
-rw-r--r--fs/xfs/xfs_dinode.h5
-rw-r--r--fs/xfs/xfs_dir2_leaf.c2
-rw-r--r--fs/xfs/xfs_fs.h7
-rw-r--r--fs/xfs/xfs_fsops.c14
-rw-r--r--fs/xfs/xfs_ialloc.c2
-rw-r--r--fs/xfs/xfs_ialloc_btree.c33
-rw-r--r--fs/xfs/xfs_iget.c4
-rw-r--r--fs/xfs/xfs_inode.c17
-rw-r--r--fs/xfs/xfs_inode.h30
-rw-r--r--fs/xfs/xfs_inode_item.c9
-rw-r--r--fs/xfs/xfs_itable.c3
-rw-r--r--fs/xfs/xfs_log.c5
-rw-r--r--fs/xfs/xfs_log_cil.c232
-rw-r--r--fs/xfs/xfs_log_recover.c25
-rw-r--r--fs/xfs/xfs_mount.c308
-rw-r--r--fs/xfs/xfs_mount.h9
-rw-r--r--fs/xfs/xfs_refcache.h52
-rw-r--r--fs/xfs/xfs_rename.c14
-rw-r--r--fs/xfs/xfs_rtalloc.c29
-rw-r--r--fs/xfs/xfs_sb.h10
-rw-r--r--fs/xfs/xfs_trans.c91
-rw-r--r--fs/xfs/xfs_trans.h3
-rw-r--r--fs/xfs/xfs_trans_buf.c2
-rw-r--r--fs/xfs/xfs_trans_inode.c30
-rw-r--r--fs/xfs/xfs_types.h2
-rw-r--r--fs/xfs/xfs_utils.c9
-rw-r--r--fs/xfs/xfs_utils.h3
-rw-r--r--fs/xfs/xfs_vnodeops.c65
-rw-r--r--fs/xfs/xfs_vnodeops.h6
60 files changed, 1185 insertions, 1375 deletions
diff --git a/fs/xfs/linux-2.6/xfs_buf.c b/fs/xfs/linux-2.6/xfs_buf.c
index f3ccaec5760a..ba5312802aa9 100644
--- a/fs/xfs/linux-2.6/xfs_buf.c
+++ b/fs/xfs/linux-2.6/xfs_buf.c
@@ -188,8 +188,8 @@ _xfs_buf_initialize(
188 atomic_set(&bp->b_hold, 1); 188 atomic_set(&bp->b_hold, 1);
189 init_completion(&bp->b_iowait); 189 init_completion(&bp->b_iowait);
190 INIT_LIST_HEAD(&bp->b_list); 190 INIT_LIST_HEAD(&bp->b_list);
191 INIT_LIST_HEAD(&bp->b_hash_list); 191 RB_CLEAR_NODE(&bp->b_rbnode);
192 init_MUTEX_LOCKED(&bp->b_sema); /* held, no waiters */ 192 sema_init(&bp->b_sema, 0); /* held, no waiters */
193 XB_SET_OWNER(bp); 193 XB_SET_OWNER(bp);
194 bp->b_target = target; 194 bp->b_target = target;
195 bp->b_file_offset = range_base; 195 bp->b_file_offset = range_base;
@@ -262,8 +262,6 @@ xfs_buf_free(
262{ 262{
263 trace_xfs_buf_free(bp, _RET_IP_); 263 trace_xfs_buf_free(bp, _RET_IP_);
264 264
265 ASSERT(list_empty(&bp->b_hash_list));
266
267 if (bp->b_flags & (_XBF_PAGE_CACHE|_XBF_PAGES)) { 265 if (bp->b_flags & (_XBF_PAGE_CACHE|_XBF_PAGES)) {
268 uint i; 266 uint i;
269 267
@@ -422,8 +420,10 @@ _xfs_buf_find(
422{ 420{
423 xfs_off_t range_base; 421 xfs_off_t range_base;
424 size_t range_length; 422 size_t range_length;
425 xfs_bufhash_t *hash; 423 struct xfs_perag *pag;
426 xfs_buf_t *bp, *n; 424 struct rb_node **rbp;
425 struct rb_node *parent;
426 xfs_buf_t *bp;
427 427
428 range_base = (ioff << BBSHIFT); 428 range_base = (ioff << BBSHIFT);
429 range_length = (isize << BBSHIFT); 429 range_length = (isize << BBSHIFT);
@@ -432,14 +432,37 @@ _xfs_buf_find(
432 ASSERT(!(range_length < (1 << btp->bt_sshift))); 432 ASSERT(!(range_length < (1 << btp->bt_sshift)));
433 ASSERT(!(range_base & (xfs_off_t)btp->bt_smask)); 433 ASSERT(!(range_base & (xfs_off_t)btp->bt_smask));
434 434
435 hash = &btp->bt_hash[hash_long((unsigned long)ioff, btp->bt_hashshift)]; 435 /* get tree root */
436 436 pag = xfs_perag_get(btp->bt_mount,
437 spin_lock(&hash->bh_lock); 437 xfs_daddr_to_agno(btp->bt_mount, ioff));
438 438
439 list_for_each_entry_safe(bp, n, &hash->bh_list, b_hash_list) { 439 /* walk tree */
440 ASSERT(btp == bp->b_target); 440 spin_lock(&pag->pag_buf_lock);
441 if (bp->b_file_offset == range_base && 441 rbp = &pag->pag_buf_tree.rb_node;
442 bp->b_buffer_length == range_length) { 442 parent = NULL;
443 bp = NULL;
444 while (*rbp) {
445 parent = *rbp;
446 bp = rb_entry(parent, struct xfs_buf, b_rbnode);
447
448 if (range_base < bp->b_file_offset)
449 rbp = &(*rbp)->rb_left;
450 else if (range_base > bp->b_file_offset)
451 rbp = &(*rbp)->rb_right;
452 else {
453 /*
454 * found a block offset match. If the range doesn't
455 * match, the only way this is allowed is if the buffer
456 * in the cache is stale and the transaction that made
457 * it stale has not yet committed. i.e. we are
458 * reallocating a busy extent. Skip this buffer and
459 * continue searching to the right for an exact match.
460 */
461 if (bp->b_buffer_length != range_length) {
462 ASSERT(bp->b_flags & XBF_STALE);
463 rbp = &(*rbp)->rb_right;
464 continue;
465 }
443 atomic_inc(&bp->b_hold); 466 atomic_inc(&bp->b_hold);
444 goto found; 467 goto found;
445 } 468 }
@@ -449,17 +472,21 @@ _xfs_buf_find(
449 if (new_bp) { 472 if (new_bp) {
450 _xfs_buf_initialize(new_bp, btp, range_base, 473 _xfs_buf_initialize(new_bp, btp, range_base,
451 range_length, flags); 474 range_length, flags);
452 new_bp->b_hash = hash; 475 rb_link_node(&new_bp->b_rbnode, parent, rbp);
453 list_add(&new_bp->b_hash_list, &hash->bh_list); 476 rb_insert_color(&new_bp->b_rbnode, &pag->pag_buf_tree);
477 /* the buffer keeps the perag reference until it is freed */
478 new_bp->b_pag = pag;
479 spin_unlock(&pag->pag_buf_lock);
454 } else { 480 } else {
455 XFS_STATS_INC(xb_miss_locked); 481 XFS_STATS_INC(xb_miss_locked);
482 spin_unlock(&pag->pag_buf_lock);
483 xfs_perag_put(pag);
456 } 484 }
457
458 spin_unlock(&hash->bh_lock);
459 return new_bp; 485 return new_bp;
460 486
461found: 487found:
462 spin_unlock(&hash->bh_lock); 488 spin_unlock(&pag->pag_buf_lock);
489 xfs_perag_put(pag);
463 490
464 /* Attempt to get the semaphore without sleeping, 491 /* Attempt to get the semaphore without sleeping,
465 * if this does not work then we need to drop the 492 * if this does not work then we need to drop the
@@ -625,8 +652,7 @@ void
625xfs_buf_readahead( 652xfs_buf_readahead(
626 xfs_buftarg_t *target, 653 xfs_buftarg_t *target,
627 xfs_off_t ioff, 654 xfs_off_t ioff,
628 size_t isize, 655 size_t isize)
629 xfs_buf_flags_t flags)
630{ 656{
631 struct backing_dev_info *bdi; 657 struct backing_dev_info *bdi;
632 658
@@ -634,8 +660,42 @@ xfs_buf_readahead(
634 if (bdi_read_congested(bdi)) 660 if (bdi_read_congested(bdi))
635 return; 661 return;
636 662
637 flags |= (XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD); 663 xfs_buf_read(target, ioff, isize,
638 xfs_buf_read(target, ioff, isize, flags); 664 XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD|XBF_DONT_BLOCK);
665}
666
667/*
668 * Read an uncached buffer from disk. Allocates and returns a locked
669 * buffer containing the disk contents or nothing.
670 */
671struct xfs_buf *
672xfs_buf_read_uncached(
673 struct xfs_mount *mp,
674 struct xfs_buftarg *target,
675 xfs_daddr_t daddr,
676 size_t length,
677 int flags)
678{
679 xfs_buf_t *bp;
680 int error;
681
682 bp = xfs_buf_get_uncached(target, length, flags);
683 if (!bp)
684 return NULL;
685
686 /* set up the buffer for a read IO */
687 xfs_buf_lock(bp);
688 XFS_BUF_SET_ADDR(bp, daddr);
689 XFS_BUF_READ(bp);
690 XFS_BUF_BUSY(bp);
691
692 xfsbdstrat(mp, bp);
693 error = xfs_buf_iowait(bp);
694 if (error || bp->b_error) {
695 xfs_buf_relse(bp);
696 return NULL;
697 }
698 return bp;
639} 699}
640 700
641xfs_buf_t * 701xfs_buf_t *
@@ -707,9 +767,10 @@ xfs_buf_associate_memory(
707} 767}
708 768
709xfs_buf_t * 769xfs_buf_t *
710xfs_buf_get_noaddr( 770xfs_buf_get_uncached(
771 struct xfs_buftarg *target,
711 size_t len, 772 size_t len,
712 xfs_buftarg_t *target) 773 int flags)
713{ 774{
714 unsigned long page_count = PAGE_ALIGN(len) >> PAGE_SHIFT; 775 unsigned long page_count = PAGE_ALIGN(len) >> PAGE_SHIFT;
715 int error, i; 776 int error, i;
@@ -725,7 +786,7 @@ xfs_buf_get_noaddr(
725 goto fail_free_buf; 786 goto fail_free_buf;
726 787
727 for (i = 0; i < page_count; i++) { 788 for (i = 0; i < page_count; i++) {
728 bp->b_pages[i] = alloc_page(GFP_KERNEL); 789 bp->b_pages[i] = alloc_page(xb_to_gfp(flags));
729 if (!bp->b_pages[i]) 790 if (!bp->b_pages[i])
730 goto fail_free_mem; 791 goto fail_free_mem;
731 } 792 }
@@ -740,7 +801,7 @@ xfs_buf_get_noaddr(
740 801
741 xfs_buf_unlock(bp); 802 xfs_buf_unlock(bp);
742 803
743 trace_xfs_buf_get_noaddr(bp, _RET_IP_); 804 trace_xfs_buf_get_uncached(bp, _RET_IP_);
744 return bp; 805 return bp;
745 806
746 fail_free_mem: 807 fail_free_mem:
@@ -774,29 +835,30 @@ void
774xfs_buf_rele( 835xfs_buf_rele(
775 xfs_buf_t *bp) 836 xfs_buf_t *bp)
776{ 837{
777 xfs_bufhash_t *hash = bp->b_hash; 838 struct xfs_perag *pag = bp->b_pag;
778 839
779 trace_xfs_buf_rele(bp, _RET_IP_); 840 trace_xfs_buf_rele(bp, _RET_IP_);
780 841
781 if (unlikely(!hash)) { 842 if (!pag) {
782 ASSERT(!bp->b_relse); 843 ASSERT(!bp->b_relse);
844 ASSERT(RB_EMPTY_NODE(&bp->b_rbnode));
783 if (atomic_dec_and_test(&bp->b_hold)) 845 if (atomic_dec_and_test(&bp->b_hold))
784 xfs_buf_free(bp); 846 xfs_buf_free(bp);
785 return; 847 return;
786 } 848 }
787 849
850 ASSERT(!RB_EMPTY_NODE(&bp->b_rbnode));
788 ASSERT(atomic_read(&bp->b_hold) > 0); 851 ASSERT(atomic_read(&bp->b_hold) > 0);
789 if (atomic_dec_and_lock(&bp->b_hold, &hash->bh_lock)) { 852 if (atomic_dec_and_lock(&bp->b_hold, &pag->pag_buf_lock)) {
790 if (bp->b_relse) { 853 if (bp->b_relse) {
791 atomic_inc(&bp->b_hold); 854 atomic_inc(&bp->b_hold);
792 spin_unlock(&hash->bh_lock); 855 spin_unlock(&pag->pag_buf_lock);
793 (*(bp->b_relse)) (bp); 856 bp->b_relse(bp);
794 } else if (bp->b_flags & XBF_FS_MANAGED) {
795 spin_unlock(&hash->bh_lock);
796 } else { 857 } else {
797 ASSERT(!(bp->b_flags & (XBF_DELWRI|_XBF_DELWRI_Q))); 858 ASSERT(!(bp->b_flags & (XBF_DELWRI|_XBF_DELWRI_Q)));
798 list_del_init(&bp->b_hash_list); 859 rb_erase(&bp->b_rbnode, &pag->pag_buf_tree);
799 spin_unlock(&hash->bh_lock); 860 spin_unlock(&pag->pag_buf_lock);
861 xfs_perag_put(pag);
800 xfs_buf_free(bp); 862 xfs_buf_free(bp);
801 } 863 }
802 } 864 }
@@ -859,7 +921,7 @@ xfs_buf_lock(
859 trace_xfs_buf_lock(bp, _RET_IP_); 921 trace_xfs_buf_lock(bp, _RET_IP_);
860 922
861 if (atomic_read(&bp->b_pin_count) && (bp->b_flags & XBF_STALE)) 923 if (atomic_read(&bp->b_pin_count) && (bp->b_flags & XBF_STALE))
862 xfs_log_force(bp->b_mount, 0); 924 xfs_log_force(bp->b_target->bt_mount, 0);
863 if (atomic_read(&bp->b_io_remaining)) 925 if (atomic_read(&bp->b_io_remaining))
864 blk_run_address_space(bp->b_target->bt_mapping); 926 blk_run_address_space(bp->b_target->bt_mapping);
865 down(&bp->b_sema); 927 down(&bp->b_sema);
@@ -970,7 +1032,6 @@ xfs_bwrite(
970{ 1032{
971 int error; 1033 int error;
972 1034
973 bp->b_mount = mp;
974 bp->b_flags |= XBF_WRITE; 1035 bp->b_flags |= XBF_WRITE;
975 bp->b_flags &= ~(XBF_ASYNC | XBF_READ); 1036 bp->b_flags &= ~(XBF_ASYNC | XBF_READ);
976 1037
@@ -991,8 +1052,6 @@ xfs_bdwrite(
991{ 1052{
992 trace_xfs_buf_bdwrite(bp, _RET_IP_); 1053 trace_xfs_buf_bdwrite(bp, _RET_IP_);
993 1054
994 bp->b_mount = mp;
995
996 bp->b_flags &= ~XBF_READ; 1055 bp->b_flags &= ~XBF_READ;
997 bp->b_flags |= (XBF_DELWRI | XBF_ASYNC); 1056 bp->b_flags |= (XBF_DELWRI | XBF_ASYNC);
998 1057
@@ -1001,7 +1060,7 @@ xfs_bdwrite(
1001 1060
1002/* 1061/*
1003 * Called when we want to stop a buffer from getting written or read. 1062 * Called when we want to stop a buffer from getting written or read.
1004 * We attach the EIO error, muck with its flags, and call biodone 1063 * We attach the EIO error, muck with its flags, and call xfs_buf_ioend
1005 * so that the proper iodone callbacks get called. 1064 * so that the proper iodone callbacks get called.
1006 */ 1065 */
1007STATIC int 1066STATIC int
@@ -1018,21 +1077,21 @@ xfs_bioerror(
1018 XFS_BUF_ERROR(bp, EIO); 1077 XFS_BUF_ERROR(bp, EIO);
1019 1078
1020 /* 1079 /*
1021 * We're calling biodone, so delete XBF_DONE flag. 1080 * We're calling xfs_buf_ioend, so delete XBF_DONE flag.
1022 */ 1081 */
1023 XFS_BUF_UNREAD(bp); 1082 XFS_BUF_UNREAD(bp);
1024 XFS_BUF_UNDELAYWRITE(bp); 1083 XFS_BUF_UNDELAYWRITE(bp);
1025 XFS_BUF_UNDONE(bp); 1084 XFS_BUF_UNDONE(bp);
1026 XFS_BUF_STALE(bp); 1085 XFS_BUF_STALE(bp);
1027 1086
1028 xfs_biodone(bp); 1087 xfs_buf_ioend(bp, 0);
1029 1088
1030 return EIO; 1089 return EIO;
1031} 1090}
1032 1091
1033/* 1092/*
1034 * Same as xfs_bioerror, except that we are releasing the buffer 1093 * Same as xfs_bioerror, except that we are releasing the buffer
1035 * here ourselves, and avoiding the biodone call. 1094 * here ourselves, and avoiding the xfs_buf_ioend call.
1036 * This is meant for userdata errors; metadata bufs come with 1095 * This is meant for userdata errors; metadata bufs come with
1037 * iodone functions attached, so that we can track down errors. 1096 * iodone functions attached, so that we can track down errors.
1038 */ 1097 */
@@ -1081,7 +1140,7 @@ int
1081xfs_bdstrat_cb( 1140xfs_bdstrat_cb(
1082 struct xfs_buf *bp) 1141 struct xfs_buf *bp)
1083{ 1142{
1084 if (XFS_FORCED_SHUTDOWN(bp->b_mount)) { 1143 if (XFS_FORCED_SHUTDOWN(bp->b_target->bt_mount)) {
1085 trace_xfs_bdstrat_shut(bp, _RET_IP_); 1144 trace_xfs_bdstrat_shut(bp, _RET_IP_);
1086 /* 1145 /*
1087 * Metadata write that didn't get logged but 1146 * Metadata write that didn't get logged but
@@ -1387,62 +1446,24 @@ xfs_buf_iomove(
1387 */ 1446 */
1388void 1447void
1389xfs_wait_buftarg( 1448xfs_wait_buftarg(
1390 xfs_buftarg_t *btp) 1449 struct xfs_buftarg *btp)
1391{
1392 xfs_buf_t *bp, *n;
1393 xfs_bufhash_t *hash;
1394 uint i;
1395
1396 for (i = 0; i < (1 << btp->bt_hashshift); i++) {
1397 hash = &btp->bt_hash[i];
1398again:
1399 spin_lock(&hash->bh_lock);
1400 list_for_each_entry_safe(bp, n, &hash->bh_list, b_hash_list) {
1401 ASSERT(btp == bp->b_target);
1402 if (!(bp->b_flags & XBF_FS_MANAGED)) {
1403 spin_unlock(&hash->bh_lock);
1404 /*
1405 * Catch superblock reference count leaks
1406 * immediately
1407 */
1408 BUG_ON(bp->b_bn == 0);
1409 delay(100);
1410 goto again;
1411 }
1412 }
1413 spin_unlock(&hash->bh_lock);
1414 }
1415}
1416
1417/*
1418 * Allocate buffer hash table for a given target.
1419 * For devices containing metadata (i.e. not the log/realtime devices)
1420 * we need to allocate a much larger hash table.
1421 */
1422STATIC void
1423xfs_alloc_bufhash(
1424 xfs_buftarg_t *btp,
1425 int external)
1426{ 1450{
1427 unsigned int i; 1451 struct xfs_perag *pag;
1452 uint i;
1428 1453
1429 btp->bt_hashshift = external ? 3 : 12; /* 8 or 4096 buckets */ 1454 for (i = 0; i < btp->bt_mount->m_sb.sb_agcount; i++) {
1430 btp->bt_hash = kmem_zalloc_large((1 << btp->bt_hashshift) * 1455 pag = xfs_perag_get(btp->bt_mount, i);
1431 sizeof(xfs_bufhash_t)); 1456 spin_lock(&pag->pag_buf_lock);
1432 for (i = 0; i < (1 << btp->bt_hashshift); i++) { 1457 while (rb_first(&pag->pag_buf_tree)) {
1433 spin_lock_init(&btp->bt_hash[i].bh_lock); 1458 spin_unlock(&pag->pag_buf_lock);
1434 INIT_LIST_HEAD(&btp->bt_hash[i].bh_list); 1459 delay(100);
1460 spin_lock(&pag->pag_buf_lock);
1461 }
1462 spin_unlock(&pag->pag_buf_lock);
1463 xfs_perag_put(pag);
1435 } 1464 }
1436} 1465}
1437 1466
1438STATIC void
1439xfs_free_bufhash(
1440 xfs_buftarg_t *btp)
1441{
1442 kmem_free_large(btp->bt_hash);
1443 btp->bt_hash = NULL;
1444}
1445
1446/* 1467/*
1447 * buftarg list for delwrite queue processing 1468 * buftarg list for delwrite queue processing
1448 */ 1469 */
@@ -1475,7 +1496,6 @@ xfs_free_buftarg(
1475 xfs_flush_buftarg(btp, 1); 1496 xfs_flush_buftarg(btp, 1);
1476 if (mp->m_flags & XFS_MOUNT_BARRIER) 1497 if (mp->m_flags & XFS_MOUNT_BARRIER)
1477 xfs_blkdev_issue_flush(btp); 1498 xfs_blkdev_issue_flush(btp);
1478 xfs_free_bufhash(btp);
1479 iput(btp->bt_mapping->host); 1499 iput(btp->bt_mapping->host);
1480 1500
1481 /* Unregister the buftarg first so that we don't get a 1501 /* Unregister the buftarg first so that we don't get a
@@ -1597,6 +1617,7 @@ out_error:
1597 1617
1598xfs_buftarg_t * 1618xfs_buftarg_t *
1599xfs_alloc_buftarg( 1619xfs_alloc_buftarg(
1620 struct xfs_mount *mp,
1600 struct block_device *bdev, 1621 struct block_device *bdev,
1601 int external, 1622 int external,
1602 const char *fsname) 1623 const char *fsname)
@@ -1605,6 +1626,7 @@ xfs_alloc_buftarg(
1605 1626
1606 btp = kmem_zalloc(sizeof(*btp), KM_SLEEP); 1627 btp = kmem_zalloc(sizeof(*btp), KM_SLEEP);
1607 1628
1629 btp->bt_mount = mp;
1608 btp->bt_dev = bdev->bd_dev; 1630 btp->bt_dev = bdev->bd_dev;
1609 btp->bt_bdev = bdev; 1631 btp->bt_bdev = bdev;
1610 if (xfs_setsize_buftarg_early(btp, bdev)) 1632 if (xfs_setsize_buftarg_early(btp, bdev))
@@ -1613,7 +1635,6 @@ xfs_alloc_buftarg(
1613 goto error; 1635 goto error;
1614 if (xfs_alloc_delwrite_queue(btp, fsname)) 1636 if (xfs_alloc_delwrite_queue(btp, fsname))
1615 goto error; 1637 goto error;
1616 xfs_alloc_bufhash(btp, external);
1617 return btp; 1638 return btp;
1618 1639
1619error: 1640error:
@@ -1904,7 +1925,7 @@ xfs_flush_buftarg(
1904 bp = list_first_entry(&wait_list, struct xfs_buf, b_list); 1925 bp = list_first_entry(&wait_list, struct xfs_buf, b_list);
1905 1926
1906 list_del_init(&bp->b_list); 1927 list_del_init(&bp->b_list);
1907 xfs_iowait(bp); 1928 xfs_buf_iowait(bp);
1908 xfs_buf_relse(bp); 1929 xfs_buf_relse(bp);
1909 } 1930 }
1910 } 1931 }
diff --git a/fs/xfs/linux-2.6/xfs_buf.h b/fs/xfs/linux-2.6/xfs_buf.h
index 9d021c73ea52..383a3f37cf98 100644
--- a/fs/xfs/linux-2.6/xfs_buf.h
+++ b/fs/xfs/linux-2.6/xfs_buf.h
@@ -51,7 +51,6 @@ typedef enum {
51#define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */ 51#define XBF_DONE (1 << 5) /* all pages in the buffer uptodate */
52#define XBF_DELWRI (1 << 6) /* buffer has dirty pages */ 52#define XBF_DELWRI (1 << 6) /* buffer has dirty pages */
53#define XBF_STALE (1 << 7) /* buffer has been staled, do not find it */ 53#define XBF_STALE (1 << 7) /* buffer has been staled, do not find it */
54#define XBF_FS_MANAGED (1 << 8) /* filesystem controls freeing memory */
55#define XBF_ORDERED (1 << 11)/* use ordered writes */ 54#define XBF_ORDERED (1 << 11)/* use ordered writes */
56#define XBF_READ_AHEAD (1 << 12)/* asynchronous read-ahead */ 55#define XBF_READ_AHEAD (1 << 12)/* asynchronous read-ahead */
57#define XBF_LOG_BUFFER (1 << 13)/* this is a buffer used for the log */ 56#define XBF_LOG_BUFFER (1 << 13)/* this is a buffer used for the log */
@@ -96,7 +95,6 @@ typedef unsigned int xfs_buf_flags_t;
96 { XBF_DONE, "DONE" }, \ 95 { XBF_DONE, "DONE" }, \
97 { XBF_DELWRI, "DELWRI" }, \ 96 { XBF_DELWRI, "DELWRI" }, \
98 { XBF_STALE, "STALE" }, \ 97 { XBF_STALE, "STALE" }, \
99 { XBF_FS_MANAGED, "FS_MANAGED" }, \
100 { XBF_ORDERED, "ORDERED" }, \ 98 { XBF_ORDERED, "ORDERED" }, \
101 { XBF_READ_AHEAD, "READ_AHEAD" }, \ 99 { XBF_READ_AHEAD, "READ_AHEAD" }, \
102 { XBF_LOCK, "LOCK" }, /* should never be set */\ 100 { XBF_LOCK, "LOCK" }, /* should never be set */\
@@ -123,14 +121,11 @@ typedef struct xfs_buftarg {
123 dev_t bt_dev; 121 dev_t bt_dev;
124 struct block_device *bt_bdev; 122 struct block_device *bt_bdev;
125 struct address_space *bt_mapping; 123 struct address_space *bt_mapping;
124 struct xfs_mount *bt_mount;
126 unsigned int bt_bsize; 125 unsigned int bt_bsize;
127 unsigned int bt_sshift; 126 unsigned int bt_sshift;
128 size_t bt_smask; 127 size_t bt_smask;
129 128
130 /* per device buffer hash table */
131 uint bt_hashshift;
132 xfs_bufhash_t *bt_hash;
133
134 /* per device delwri queue */ 129 /* per device delwri queue */
135 struct task_struct *bt_task; 130 struct task_struct *bt_task;
136 struct list_head bt_list; 131 struct list_head bt_list;
@@ -158,34 +153,41 @@ typedef int (*xfs_buf_bdstrat_t)(struct xfs_buf *);
158#define XB_PAGES 2 153#define XB_PAGES 2
159 154
160typedef struct xfs_buf { 155typedef struct xfs_buf {
156 /*
157 * first cacheline holds all the fields needed for an uncontended cache
158 * hit to be fully processed. The semaphore straddles the cacheline
159 * boundary, but the counter and lock sits on the first cacheline,
160 * which is the only bit that is touched if we hit the semaphore
161 * fast-path on locking.
162 */
163 struct rb_node b_rbnode; /* rbtree node */
164 xfs_off_t b_file_offset; /* offset in file */
165 size_t b_buffer_length;/* size of buffer in bytes */
166 atomic_t b_hold; /* reference count */
167 xfs_buf_flags_t b_flags; /* status flags */
161 struct semaphore b_sema; /* semaphore for lockables */ 168 struct semaphore b_sema; /* semaphore for lockables */
162 unsigned long b_queuetime; /* time buffer was queued */ 169
163 atomic_t b_pin_count; /* pin count */
164 wait_queue_head_t b_waiters; /* unpin waiters */ 170 wait_queue_head_t b_waiters; /* unpin waiters */
165 struct list_head b_list; 171 struct list_head b_list;
166 xfs_buf_flags_t b_flags; /* status flags */ 172 struct xfs_perag *b_pag; /* contains rbtree root */
167 struct list_head b_hash_list; /* hash table list */
168 xfs_bufhash_t *b_hash; /* hash table list start */
169 xfs_buftarg_t *b_target; /* buffer target (device) */ 173 xfs_buftarg_t *b_target; /* buffer target (device) */
170 atomic_t b_hold; /* reference count */
171 xfs_daddr_t b_bn; /* block number for I/O */ 174 xfs_daddr_t b_bn; /* block number for I/O */
172 xfs_off_t b_file_offset; /* offset in file */
173 size_t b_buffer_length;/* size of buffer in bytes */
174 size_t b_count_desired;/* desired transfer size */ 175 size_t b_count_desired;/* desired transfer size */
175 void *b_addr; /* virtual address of buffer */ 176 void *b_addr; /* virtual address of buffer */
176 struct work_struct b_iodone_work; 177 struct work_struct b_iodone_work;
177 atomic_t b_io_remaining; /* #outstanding I/O requests */
178 xfs_buf_iodone_t b_iodone; /* I/O completion function */ 178 xfs_buf_iodone_t b_iodone; /* I/O completion function */
179 xfs_buf_relse_t b_relse; /* releasing function */ 179 xfs_buf_relse_t b_relse; /* releasing function */
180 struct completion b_iowait; /* queue for I/O waiters */ 180 struct completion b_iowait; /* queue for I/O waiters */
181 void *b_fspriv; 181 void *b_fspriv;
182 void *b_fspriv2; 182 void *b_fspriv2;
183 struct xfs_mount *b_mount;
184 unsigned short b_error; /* error code on I/O */
185 unsigned int b_page_count; /* size of page array */
186 unsigned int b_offset; /* page offset in first page */
187 struct page **b_pages; /* array of page pointers */ 183 struct page **b_pages; /* array of page pointers */
188 struct page *b_page_array[XB_PAGES]; /* inline pages */ 184 struct page *b_page_array[XB_PAGES]; /* inline pages */
185 unsigned long b_queuetime; /* time buffer was queued */
186 atomic_t b_pin_count; /* pin count */
187 atomic_t b_io_remaining; /* #outstanding I/O requests */
188 unsigned int b_page_count; /* size of page array */
189 unsigned int b_offset; /* page offset in first page */
190 unsigned short b_error; /* error code on I/O */
189#ifdef XFS_BUF_LOCK_TRACKING 191#ifdef XFS_BUF_LOCK_TRACKING
190 int b_last_holder; 192 int b_last_holder;
191#endif 193#endif
@@ -204,11 +206,13 @@ extern xfs_buf_t *xfs_buf_read(xfs_buftarg_t *, xfs_off_t, size_t,
204 xfs_buf_flags_t); 206 xfs_buf_flags_t);
205 207
206extern xfs_buf_t *xfs_buf_get_empty(size_t, xfs_buftarg_t *); 208extern xfs_buf_t *xfs_buf_get_empty(size_t, xfs_buftarg_t *);
207extern xfs_buf_t *xfs_buf_get_noaddr(size_t, xfs_buftarg_t *); 209extern xfs_buf_t *xfs_buf_get_uncached(struct xfs_buftarg *, size_t, int);
208extern int xfs_buf_associate_memory(xfs_buf_t *, void *, size_t); 210extern int xfs_buf_associate_memory(xfs_buf_t *, void *, size_t);
209extern void xfs_buf_hold(xfs_buf_t *); 211extern void xfs_buf_hold(xfs_buf_t *);
210extern void xfs_buf_readahead(xfs_buftarg_t *, xfs_off_t, size_t, 212extern void xfs_buf_readahead(xfs_buftarg_t *, xfs_off_t, size_t);
211 xfs_buf_flags_t); 213struct xfs_buf *xfs_buf_read_uncached(struct xfs_mount *mp,
214 struct xfs_buftarg *target,
215 xfs_daddr_t daddr, size_t length, int flags);
212 216
213/* Releasing Buffers */ 217/* Releasing Buffers */
214extern void xfs_buf_free(xfs_buf_t *); 218extern void xfs_buf_free(xfs_buf_t *);
@@ -233,6 +237,8 @@ extern int xfs_buf_iorequest(xfs_buf_t *);
233extern int xfs_buf_iowait(xfs_buf_t *); 237extern int xfs_buf_iowait(xfs_buf_t *);
234extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *, 238extern void xfs_buf_iomove(xfs_buf_t *, size_t, size_t, void *,
235 xfs_buf_rw_t); 239 xfs_buf_rw_t);
240#define xfs_buf_zero(bp, off, len) \
241 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
236 242
237static inline int xfs_buf_geterror(xfs_buf_t *bp) 243static inline int xfs_buf_geterror(xfs_buf_t *bp)
238{ 244{
@@ -267,8 +273,6 @@ extern void xfs_buf_terminate(void);
267 XFS_BUF_DONE(bp); \ 273 XFS_BUF_DONE(bp); \
268 } while (0) 274 } while (0)
269 275
270#define XFS_BUF_UNMANAGE(bp) ((bp)->b_flags &= ~XBF_FS_MANAGED)
271
272#define XFS_BUF_DELAYWRITE(bp) ((bp)->b_flags |= XBF_DELWRI) 276#define XFS_BUF_DELAYWRITE(bp) ((bp)->b_flags |= XBF_DELWRI)
273#define XFS_BUF_UNDELAYWRITE(bp) xfs_buf_delwri_dequeue(bp) 277#define XFS_BUF_UNDELAYWRITE(bp) xfs_buf_delwri_dequeue(bp)
274#define XFS_BUF_ISDELAYWRITE(bp) ((bp)->b_flags & XBF_DELWRI) 278#define XFS_BUF_ISDELAYWRITE(bp) ((bp)->b_flags & XBF_DELWRI)
@@ -347,25 +351,11 @@ static inline void xfs_buf_relse(xfs_buf_t *bp)
347 xfs_buf_rele(bp); 351 xfs_buf_rele(bp);
348} 352}
349 353
350#define xfs_biodone(bp) xfs_buf_ioend(bp, 0)
351
352#define xfs_biomove(bp, off, len, data, rw) \
353 xfs_buf_iomove((bp), (off), (len), (data), \
354 ((rw) == XBF_WRITE) ? XBRW_WRITE : XBRW_READ)
355
356#define xfs_biozero(bp, off, len) \
357 xfs_buf_iomove((bp), (off), (len), NULL, XBRW_ZERO)
358
359#define xfs_iowait(bp) xfs_buf_iowait(bp)
360
361#define xfs_baread(target, rablkno, ralen) \
362 xfs_buf_readahead((target), (rablkno), (ralen), XBF_DONT_BLOCK)
363
364
365/* 354/*
366 * Handling of buftargs. 355 * Handling of buftargs.
367 */ 356 */
368extern xfs_buftarg_t *xfs_alloc_buftarg(struct block_device *, int, const char *); 357extern xfs_buftarg_t *xfs_alloc_buftarg(struct xfs_mount *,
358 struct block_device *, int, const char *);
369extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *); 359extern void xfs_free_buftarg(struct xfs_mount *, struct xfs_buftarg *);
370extern void xfs_wait_buftarg(xfs_buftarg_t *); 360extern void xfs_wait_buftarg(xfs_buftarg_t *);
371extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int, unsigned int); 361extern int xfs_setsize_buftarg(xfs_buftarg_t *, unsigned int, unsigned int);
diff --git a/fs/xfs/linux-2.6/xfs_cred.h b/fs/xfs/linux-2.6/xfs_cred.h
deleted file mode 100644
index 55bddf3b6091..000000000000
--- a/fs/xfs/linux-2.6/xfs_cred.h
+++ /dev/null
@@ -1,28 +0,0 @@
1/*
2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_CRED_H__
19#define __XFS_CRED_H__
20
21#include <linux/capability.h>
22
23/*
24 * Credentials
25 */
26typedef const struct cred cred_t;
27
28#endif /* __XFS_CRED_H__ */
diff --git a/fs/xfs/linux-2.6/xfs_fs_subr.c b/fs/xfs/linux-2.6/xfs_fs_subr.c
index 1f279b012f94..ed88ed16811c 100644
--- a/fs/xfs/linux-2.6/xfs_fs_subr.c
+++ b/fs/xfs/linux-2.6/xfs_fs_subr.c
@@ -32,10 +32,9 @@ xfs_tosspages(
32 xfs_off_t last, 32 xfs_off_t last,
33 int fiopt) 33 int fiopt)
34{ 34{
35 struct address_space *mapping = VFS_I(ip)->i_mapping; 35 /* can't toss partial tail pages, so mask them out */
36 36 last &= ~(PAGE_SIZE - 1);
37 if (mapping->nrpages) 37 truncate_inode_pages_range(VFS_I(ip)->i_mapping, first, last - 1);
38 truncate_inode_pages(mapping, first);
39} 38}
40 39
41int 40int
@@ -50,12 +49,11 @@ xfs_flushinval_pages(
50 49
51 trace_xfs_pagecache_inval(ip, first, last); 50 trace_xfs_pagecache_inval(ip, first, last);
52 51
53 if (mapping->nrpages) { 52 xfs_iflags_clear(ip, XFS_ITRUNCATED);
54 xfs_iflags_clear(ip, XFS_ITRUNCATED); 53 ret = filemap_write_and_wait_range(mapping, first,
55 ret = filemap_write_and_wait(mapping); 54 last == -1 ? LLONG_MAX : last);
56 if (!ret) 55 if (!ret)
57 truncate_inode_pages(mapping, first); 56 truncate_inode_pages_range(mapping, first, last);
58 }
59 return -ret; 57 return -ret;
60} 58}
61 59
@@ -71,10 +69,9 @@ xfs_flush_pages(
71 int ret = 0; 69 int ret = 0;
72 int ret2; 70 int ret2;
73 71
74 if (mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) { 72 xfs_iflags_clear(ip, XFS_ITRUNCATED);
75 xfs_iflags_clear(ip, XFS_ITRUNCATED); 73 ret = -filemap_fdatawrite_range(mapping, first,
76 ret = -filemap_fdatawrite(mapping); 74 last == -1 ? LLONG_MAX : last);
77 }
78 if (flags & XBF_ASYNC) 75 if (flags & XBF_ASYNC)
79 return ret; 76 return ret;
80 ret2 = xfs_wait_on_pages(ip, first, last); 77 ret2 = xfs_wait_on_pages(ip, first, last);
@@ -91,7 +88,9 @@ xfs_wait_on_pages(
91{ 88{
92 struct address_space *mapping = VFS_I(ip)->i_mapping; 89 struct address_space *mapping = VFS_I(ip)->i_mapping;
93 90
94 if (mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) 91 if (mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) {
95 return -filemap_fdatawait(mapping); 92 return -filemap_fdatawait_range(mapping, first,
93 last == -1 ? ip->i_size - 1 : last);
94 }
96 return 0; 95 return 0;
97} 96}
diff --git a/fs/xfs/linux-2.6/xfs_globals.c b/fs/xfs/linux-2.6/xfs_globals.c
index 2ae8b1ccb02e..76e81cff70b9 100644
--- a/fs/xfs/linux-2.6/xfs_globals.c
+++ b/fs/xfs/linux-2.6/xfs_globals.c
@@ -16,7 +16,6 @@
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */ 17 */
18#include "xfs.h" 18#include "xfs.h"
19#include "xfs_cred.h"
20#include "xfs_sysctl.h" 19#include "xfs_sysctl.h"
21 20
22/* 21/*
diff --git a/fs/xfs/linux-2.6/xfs_globals.h b/fs/xfs/linux-2.6/xfs_globals.h
deleted file mode 100644
index 69f71caf061c..000000000000
--- a/fs/xfs/linux-2.6/xfs_globals.h
+++ /dev/null
@@ -1,23 +0,0 @@
1/*
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_GLOBALS_H__
19#define __XFS_GLOBALS_H__
20
21extern uint64_t xfs_panic_mask; /* set to cause more panics */
22
23#endif /* __XFS_GLOBALS_H__ */
diff --git a/fs/xfs/linux-2.6/xfs_ioctl.c b/fs/xfs/linux-2.6/xfs_ioctl.c
index 3b9e626f7cd1..2ea238f6d38e 100644
--- a/fs/xfs/linux-2.6/xfs_ioctl.c
+++ b/fs/xfs/linux-2.6/xfs_ioctl.c
@@ -790,7 +790,7 @@ xfs_ioc_fsgetxattr(
790 xfs_ilock(ip, XFS_ILOCK_SHARED); 790 xfs_ilock(ip, XFS_ILOCK_SHARED);
791 fa.fsx_xflags = xfs_ip2xflags(ip); 791 fa.fsx_xflags = xfs_ip2xflags(ip);
792 fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog; 792 fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
793 fa.fsx_projid = ip->i_d.di_projid; 793 fa.fsx_projid = xfs_get_projid(ip);
794 794
795 if (attr) { 795 if (attr) {
796 if (ip->i_afp) { 796 if (ip->i_afp) {
@@ -909,10 +909,10 @@ xfs_ioctl_setattr(
909 return XFS_ERROR(EIO); 909 return XFS_ERROR(EIO);
910 910
911 /* 911 /*
912 * Disallow 32bit project ids because on-disk structure 912 * Disallow 32bit project ids when projid32bit feature is not enabled.
913 * is 16bit only.
914 */ 913 */
915 if ((mask & FSX_PROJID) && (fa->fsx_projid > (__uint16_t)-1)) 914 if ((mask & FSX_PROJID) && (fa->fsx_projid > (__uint16_t)-1) &&
915 !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
916 return XFS_ERROR(EINVAL); 916 return XFS_ERROR(EINVAL);
917 917
918 /* 918 /*
@@ -961,7 +961,7 @@ xfs_ioctl_setattr(
961 if (mask & FSX_PROJID) { 961 if (mask & FSX_PROJID) {
962 if (XFS_IS_QUOTA_RUNNING(mp) && 962 if (XFS_IS_QUOTA_RUNNING(mp) &&
963 XFS_IS_PQUOTA_ON(mp) && 963 XFS_IS_PQUOTA_ON(mp) &&
964 ip->i_d.di_projid != fa->fsx_projid) { 964 xfs_get_projid(ip) != fa->fsx_projid) {
965 ASSERT(tp); 965 ASSERT(tp);
966 code = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp, 966 code = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp,
967 capable(CAP_FOWNER) ? 967 capable(CAP_FOWNER) ?
@@ -1063,12 +1063,12 @@ xfs_ioctl_setattr(
1063 * Change the ownerships and register quota modifications 1063 * Change the ownerships and register quota modifications
1064 * in the transaction. 1064 * in the transaction.
1065 */ 1065 */
1066 if (ip->i_d.di_projid != fa->fsx_projid) { 1066 if (xfs_get_projid(ip) != fa->fsx_projid) {
1067 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) { 1067 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1068 olddquot = xfs_qm_vop_chown(tp, ip, 1068 olddquot = xfs_qm_vop_chown(tp, ip,
1069 &ip->i_gdquot, gdqp); 1069 &ip->i_gdquot, gdqp);
1070 } 1070 }
1071 ip->i_d.di_projid = fa->fsx_projid; 1071 xfs_set_projid(ip, fa->fsx_projid);
1072 1072
1073 /* 1073 /*
1074 * We may have to rev the inode as well as 1074 * We may have to rev the inode as well as
@@ -1088,8 +1088,8 @@ xfs_ioctl_setattr(
1088 xfs_diflags_to_linux(ip); 1088 xfs_diflags_to_linux(ip);
1089 } 1089 }
1090 1090
1091 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1091 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); 1092 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1092 xfs_ichgtime(ip, XFS_ICHGTIME_CHG);
1093 1093
1094 XFS_STATS_INC(xs_ig_attrchg); 1094 XFS_STATS_INC(xs_ig_attrchg);
1095 1095
@@ -1301,7 +1301,8 @@ xfs_file_ioctl(
1301 case XFS_IOC_ALLOCSP64: 1301 case XFS_IOC_ALLOCSP64:
1302 case XFS_IOC_FREESP64: 1302 case XFS_IOC_FREESP64:
1303 case XFS_IOC_RESVSP64: 1303 case XFS_IOC_RESVSP64:
1304 case XFS_IOC_UNRESVSP64: { 1304 case XFS_IOC_UNRESVSP64:
1305 case XFS_IOC_ZERO_RANGE: {
1305 xfs_flock64_t bf; 1306 xfs_flock64_t bf;
1306 1307
1307 if (copy_from_user(&bf, arg, sizeof(bf))) 1308 if (copy_from_user(&bf, arg, sizeof(bf)))
diff --git a/fs/xfs/linux-2.6/xfs_ioctl32.c b/fs/xfs/linux-2.6/xfs_ioctl32.c
index 6c83f7f62dc9..b3486dfa5520 100644
--- a/fs/xfs/linux-2.6/xfs_ioctl32.c
+++ b/fs/xfs/linux-2.6/xfs_ioctl32.c
@@ -164,7 +164,8 @@ xfs_ioctl32_bstat_copyin(
164 get_user(bstat->bs_extsize, &bstat32->bs_extsize) || 164 get_user(bstat->bs_extsize, &bstat32->bs_extsize) ||
165 get_user(bstat->bs_extents, &bstat32->bs_extents) || 165 get_user(bstat->bs_extents, &bstat32->bs_extents) ||
166 get_user(bstat->bs_gen, &bstat32->bs_gen) || 166 get_user(bstat->bs_gen, &bstat32->bs_gen) ||
167 get_user(bstat->bs_projid, &bstat32->bs_projid) || 167 get_user(bstat->bs_projid_lo, &bstat32->bs_projid_lo) ||
168 get_user(bstat->bs_projid_hi, &bstat32->bs_projid_hi) ||
168 get_user(bstat->bs_dmevmask, &bstat32->bs_dmevmask) || 169 get_user(bstat->bs_dmevmask, &bstat32->bs_dmevmask) ||
169 get_user(bstat->bs_dmstate, &bstat32->bs_dmstate) || 170 get_user(bstat->bs_dmstate, &bstat32->bs_dmstate) ||
170 get_user(bstat->bs_aextents, &bstat32->bs_aextents)) 171 get_user(bstat->bs_aextents, &bstat32->bs_aextents))
@@ -218,6 +219,7 @@ xfs_bulkstat_one_fmt_compat(
218 put_user(buffer->bs_extents, &p32->bs_extents) || 219 put_user(buffer->bs_extents, &p32->bs_extents) ||
219 put_user(buffer->bs_gen, &p32->bs_gen) || 220 put_user(buffer->bs_gen, &p32->bs_gen) ||
220 put_user(buffer->bs_projid, &p32->bs_projid) || 221 put_user(buffer->bs_projid, &p32->bs_projid) ||
222 put_user(buffer->bs_projid_hi, &p32->bs_projid_hi) ||
221 put_user(buffer->bs_dmevmask, &p32->bs_dmevmask) || 223 put_user(buffer->bs_dmevmask, &p32->bs_dmevmask) ||
222 put_user(buffer->bs_dmstate, &p32->bs_dmstate) || 224 put_user(buffer->bs_dmstate, &p32->bs_dmstate) ||
223 put_user(buffer->bs_aextents, &p32->bs_aextents)) 225 put_user(buffer->bs_aextents, &p32->bs_aextents))
@@ -574,6 +576,7 @@ xfs_file_compat_ioctl(
574 case XFS_IOC_FSGEOMETRY_V1: 576 case XFS_IOC_FSGEOMETRY_V1:
575 case XFS_IOC_FSGROWFSDATA: 577 case XFS_IOC_FSGROWFSDATA:
576 case XFS_IOC_FSGROWFSRT: 578 case XFS_IOC_FSGROWFSRT:
579 case XFS_IOC_ZERO_RANGE:
577 return xfs_file_ioctl(filp, cmd, p); 580 return xfs_file_ioctl(filp, cmd, p);
578#else 581#else
579 case XFS_IOC_ALLOCSP_32: 582 case XFS_IOC_ALLOCSP_32:
diff --git a/fs/xfs/linux-2.6/xfs_ioctl32.h b/fs/xfs/linux-2.6/xfs_ioctl32.h
index 1024c4f8ba0d..08b605792a99 100644
--- a/fs/xfs/linux-2.6/xfs_ioctl32.h
+++ b/fs/xfs/linux-2.6/xfs_ioctl32.h
@@ -65,8 +65,10 @@ typedef struct compat_xfs_bstat {
65 __s32 bs_extsize; /* extent size */ 65 __s32 bs_extsize; /* extent size */
66 __s32 bs_extents; /* number of extents */ 66 __s32 bs_extents; /* number of extents */
67 __u32 bs_gen; /* generation count */ 67 __u32 bs_gen; /* generation count */
68 __u16 bs_projid; /* project id */ 68 __u16 bs_projid_lo; /* lower part of project id */
69 unsigned char bs_pad[14]; /* pad space, unused */ 69#define bs_projid bs_projid_lo /* (previously just bs_projid) */
70 __u16 bs_projid_hi; /* high part of project id */
71 unsigned char bs_pad[12]; /* pad space, unused */
70 __u32 bs_dmevmask; /* DMIG event mask */ 72 __u32 bs_dmevmask; /* DMIG event mask */
71 __u16 bs_dmstate; /* DMIG state info */ 73 __u16 bs_dmstate; /* DMIG state info */
72 __u16 bs_aextents; /* attribute number of extents */ 74 __u16 bs_aextents; /* attribute number of extents */
diff --git a/fs/xfs/linux-2.6/xfs_iops.c b/fs/xfs/linux-2.6/xfs_iops.c
index b1fc2a6bfe83..ec858e09d546 100644
--- a/fs/xfs/linux-2.6/xfs_iops.c
+++ b/fs/xfs/linux-2.6/xfs_iops.c
@@ -95,41 +95,6 @@ xfs_mark_inode_dirty(
95} 95}
96 96
97/* 97/*
98 * Change the requested timestamp in the given inode.
99 * We don't lock across timestamp updates, and we don't log them but
100 * we do record the fact that there is dirty information in core.
101 */
102void
103xfs_ichgtime(
104 xfs_inode_t *ip,
105 int flags)
106{
107 struct inode *inode = VFS_I(ip);
108 timespec_t tv;
109 int sync_it = 0;
110
111 tv = current_fs_time(inode->i_sb);
112
113 if ((flags & XFS_ICHGTIME_MOD) &&
114 !timespec_equal(&inode->i_mtime, &tv)) {
115 inode->i_mtime = tv;
116 sync_it = 1;
117 }
118 if ((flags & XFS_ICHGTIME_CHG) &&
119 !timespec_equal(&inode->i_ctime, &tv)) {
120 inode->i_ctime = tv;
121 sync_it = 1;
122 }
123
124 /*
125 * Update complete - now make sure everyone knows that the inode
126 * is dirty.
127 */
128 if (sync_it)
129 xfs_mark_inode_dirty_sync(ip);
130}
131
132/*
133 * Hook in SELinux. This is not quite correct yet, what we really need 98 * Hook in SELinux. This is not quite correct yet, what we really need
134 * here (as we do for default ACLs) is a mechanism by which creation of 99 * here (as we do for default ACLs) is a mechanism by which creation of
135 * these attrs can be journalled at inode creation time (along with the 100 * these attrs can be journalled at inode creation time (along with the
@@ -224,7 +189,7 @@ xfs_vn_mknod(
224 } 189 }
225 190
226 xfs_dentry_to_name(&name, dentry); 191 xfs_dentry_to_name(&name, dentry);
227 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip, NULL); 192 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip);
228 if (unlikely(error)) 193 if (unlikely(error))
229 goto out_free_acl; 194 goto out_free_acl;
230 195
@@ -397,7 +362,7 @@ xfs_vn_symlink(
397 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO); 362 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
398 xfs_dentry_to_name(&name, dentry); 363 xfs_dentry_to_name(&name, dentry);
399 364
400 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip, NULL); 365 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip);
401 if (unlikely(error)) 366 if (unlikely(error))
402 goto out; 367 goto out;
403 368
diff --git a/fs/xfs/linux-2.6/xfs_linux.h b/fs/xfs/linux-2.6/xfs_linux.h
index 2fa0bd9ebc7f..214ddd71ff79 100644
--- a/fs/xfs/linux-2.6/xfs_linux.h
+++ b/fs/xfs/linux-2.6/xfs_linux.h
@@ -71,6 +71,7 @@
71#include <linux/random.h> 71#include <linux/random.h>
72#include <linux/ctype.h> 72#include <linux/ctype.h>
73#include <linux/writeback.h> 73#include <linux/writeback.h>
74#include <linux/capability.h>
74 75
75#include <asm/page.h> 76#include <asm/page.h>
76#include <asm/div64.h> 77#include <asm/div64.h>
@@ -79,14 +80,12 @@
79#include <asm/byteorder.h> 80#include <asm/byteorder.h>
80#include <asm/unaligned.h> 81#include <asm/unaligned.h>
81 82
82#include <xfs_cred.h>
83#include <xfs_vnode.h> 83#include <xfs_vnode.h>
84#include <xfs_stats.h> 84#include <xfs_stats.h>
85#include <xfs_sysctl.h> 85#include <xfs_sysctl.h>
86#include <xfs_iops.h> 86#include <xfs_iops.h>
87#include <xfs_aops.h> 87#include <xfs_aops.h>
88#include <xfs_super.h> 88#include <xfs_super.h>
89#include <xfs_globals.h>
90#include <xfs_buf.h> 89#include <xfs_buf.h>
91 90
92/* 91/*
@@ -144,7 +143,7 @@
144#define SYNCHRONIZE() barrier() 143#define SYNCHRONIZE() barrier()
145#define __return_address __builtin_return_address(0) 144#define __return_address __builtin_return_address(0)
146 145
147#define dfltprid 0 146#define XFS_PROJID_DEFAULT 0
148#define MAXPATHLEN 1024 147#define MAXPATHLEN 1024
149 148
150#define MIN(a,b) (min(a,b)) 149#define MIN(a,b) (min(a,b))
diff --git a/fs/xfs/linux-2.6/xfs_super.c b/fs/xfs/linux-2.6/xfs_super.c
index 08fd3102128c..ab31ce5aeaf9 100644
--- a/fs/xfs/linux-2.6/xfs_super.c
+++ b/fs/xfs/linux-2.6/xfs_super.c
@@ -44,7 +44,6 @@
44#include "xfs_buf_item.h" 44#include "xfs_buf_item.h"
45#include "xfs_utils.h" 45#include "xfs_utils.h"
46#include "xfs_vnodeops.h" 46#include "xfs_vnodeops.h"
47#include "xfs_version.h"
48#include "xfs_log_priv.h" 47#include "xfs_log_priv.h"
49#include "xfs_trans_priv.h" 48#include "xfs_trans_priv.h"
50#include "xfs_filestream.h" 49#include "xfs_filestream.h"
@@ -645,7 +644,7 @@ xfs_barrier_test(
645 XFS_BUF_ORDERED(sbp); 644 XFS_BUF_ORDERED(sbp);
646 645
647 xfsbdstrat(mp, sbp); 646 xfsbdstrat(mp, sbp);
648 error = xfs_iowait(sbp); 647 error = xfs_buf_iowait(sbp);
649 648
650 /* 649 /*
651 * Clear all the flags we set and possible error state in the 650 * Clear all the flags we set and possible error state in the
@@ -757,18 +756,20 @@ xfs_open_devices(
757 * Setup xfs_mount buffer target pointers 756 * Setup xfs_mount buffer target pointers
758 */ 757 */
759 error = ENOMEM; 758 error = ENOMEM;
760 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0, mp->m_fsname); 759 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
761 if (!mp->m_ddev_targp) 760 if (!mp->m_ddev_targp)
762 goto out_close_rtdev; 761 goto out_close_rtdev;
763 762
764 if (rtdev) { 763 if (rtdev) {
765 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1, mp->m_fsname); 764 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
765 mp->m_fsname);
766 if (!mp->m_rtdev_targp) 766 if (!mp->m_rtdev_targp)
767 goto out_free_ddev_targ; 767 goto out_free_ddev_targ;
768 } 768 }
769 769
770 if (logdev && logdev != ddev) { 770 if (logdev && logdev != ddev) {
771 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1, mp->m_fsname); 771 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
772 mp->m_fsname);
772 if (!mp->m_logdev_targp) 773 if (!mp->m_logdev_targp)
773 goto out_free_rtdev_targ; 774 goto out_free_rtdev_targ;
774 } else { 775 } else {
@@ -971,12 +972,7 @@ xfs_fs_inode_init_once(
971 972
972/* 973/*
973 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that 974 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
974 * we catch unlogged VFS level updates to the inode. Care must be taken 975 * we catch unlogged VFS level updates to the inode.
975 * here - the transaction code calls mark_inode_dirty_sync() to mark the
976 * VFS inode dirty in a transaction and clears the i_update_core field;
977 * it must clear the field after calling mark_inode_dirty_sync() to
978 * correctly indicate that the dirty state has been propagated into the
979 * inode log item.
980 * 976 *
981 * We need the barrier() to maintain correct ordering between unlogged 977 * We need the barrier() to maintain correct ordering between unlogged
982 * updates and the transaction commit code that clears the i_update_core 978 * updates and the transaction commit code that clears the i_update_core
@@ -1520,8 +1516,9 @@ xfs_fs_fill_super(
1520 if (error) 1516 if (error)
1521 goto out_free_fsname; 1517 goto out_free_fsname;
1522 1518
1523 if (xfs_icsb_init_counters(mp)) 1519 error = xfs_icsb_init_counters(mp);
1524 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB; 1520 if (error)
1521 goto out_close_devices;
1525 1522
1526 error = xfs_readsb(mp, flags); 1523 error = xfs_readsb(mp, flags);
1527 if (error) 1524 if (error)
@@ -1582,6 +1579,7 @@ xfs_fs_fill_super(
1582 xfs_freesb(mp); 1579 xfs_freesb(mp);
1583 out_destroy_counters: 1580 out_destroy_counters:
1584 xfs_icsb_destroy_counters(mp); 1581 xfs_icsb_destroy_counters(mp);
1582 out_close_devices:
1585 xfs_close_devices(mp); 1583 xfs_close_devices(mp);
1586 out_free_fsname: 1584 out_free_fsname:
1587 xfs_free_fsname(mp); 1585 xfs_free_fsname(mp);
diff --git a/fs/xfs/linux-2.6/xfs_super.h b/fs/xfs/linux-2.6/xfs_super.h
index 1ef4a4d2d997..50a3266c999e 100644
--- a/fs/xfs/linux-2.6/xfs_super.h
+++ b/fs/xfs/linux-2.6/xfs_super.h
@@ -62,6 +62,7 @@ extern void xfs_qm_exit(void);
62# define XFS_DBG_STRING "no debug" 62# define XFS_DBG_STRING "no debug"
63#endif 63#endif
64 64
65#define XFS_VERSION_STRING "SGI XFS"
65#define XFS_BUILD_OPTIONS XFS_ACL_STRING \ 66#define XFS_BUILD_OPTIONS XFS_ACL_STRING \
66 XFS_SECURITY_STRING \ 67 XFS_SECURITY_STRING \
67 XFS_REALTIME_STRING \ 68 XFS_REALTIME_STRING \
diff --git a/fs/xfs/linux-2.6/xfs_sync.c b/fs/xfs/linux-2.6/xfs_sync.c
index 81976ffed7d6..37d33254981d 100644
--- a/fs/xfs/linux-2.6/xfs_sync.c
+++ b/fs/xfs/linux-2.6/xfs_sync.c
@@ -39,42 +39,39 @@
39#include <linux/kthread.h> 39#include <linux/kthread.h>
40#include <linux/freezer.h> 40#include <linux/freezer.h>
41 41
42/*
43 * The inode lookup is done in batches to keep the amount of lock traffic and
44 * radix tree lookups to a minimum. The batch size is a trade off between
45 * lookup reduction and stack usage. This is in the reclaim path, so we can't
46 * be too greedy.
47 */
48#define XFS_LOOKUP_BATCH 32
42 49
43STATIC xfs_inode_t * 50STATIC int
44xfs_inode_ag_lookup( 51xfs_inode_ag_walk_grab(
45 struct xfs_mount *mp, 52 struct xfs_inode *ip)
46 struct xfs_perag *pag,
47 uint32_t *first_index,
48 int tag)
49{ 53{
50 int nr_found; 54 struct inode *inode = VFS_I(ip);
51 struct xfs_inode *ip;
52 55
53 /* 56 /* nothing to sync during shutdown */
54 * use a gang lookup to find the next inode in the tree 57 if (XFS_FORCED_SHUTDOWN(ip->i_mount))
55 * as the tree is sparse and a gang lookup walks to find 58 return EFSCORRUPTED;
56 * the number of objects requested. 59
57 */ 60 /* avoid new or reclaimable inodes. Leave for reclaim code to flush */
58 if (tag == XFS_ICI_NO_TAG) { 61 if (xfs_iflags_test(ip, XFS_INEW | XFS_IRECLAIMABLE | XFS_IRECLAIM))
59 nr_found = radix_tree_gang_lookup(&pag->pag_ici_root, 62 return ENOENT;
60 (void **)&ip, *first_index, 1); 63
61 } else { 64 /* If we can't grab the inode, it must on it's way to reclaim. */
62 nr_found = radix_tree_gang_lookup_tag(&pag->pag_ici_root, 65 if (!igrab(inode))
63 (void **)&ip, *first_index, 1, tag); 66 return ENOENT;
67
68 if (is_bad_inode(inode)) {
69 IRELE(ip);
70 return ENOENT;
64 } 71 }
65 if (!nr_found)
66 return NULL;
67 72
68 /* 73 /* inode is valid */
69 * Update the index for the next lookup. Catch overflows 74 return 0;
70 * into the next AG range which can occur if we have inodes
71 * in the last block of the AG and we are currently
72 * pointing to the last inode.
73 */
74 *first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
75 if (*first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
76 return NULL;
77 return ip;
78} 75}
79 76
80STATIC int 77STATIC int
@@ -83,49 +80,75 @@ xfs_inode_ag_walk(
83 struct xfs_perag *pag, 80 struct xfs_perag *pag,
84 int (*execute)(struct xfs_inode *ip, 81 int (*execute)(struct xfs_inode *ip,
85 struct xfs_perag *pag, int flags), 82 struct xfs_perag *pag, int flags),
86 int flags, 83 int flags)
87 int tag,
88 int exclusive,
89 int *nr_to_scan)
90{ 84{
91 uint32_t first_index; 85 uint32_t first_index;
92 int last_error = 0; 86 int last_error = 0;
93 int skipped; 87 int skipped;
88 int done;
89 int nr_found;
94 90
95restart: 91restart:
92 done = 0;
96 skipped = 0; 93 skipped = 0;
97 first_index = 0; 94 first_index = 0;
95 nr_found = 0;
98 do { 96 do {
97 struct xfs_inode *batch[XFS_LOOKUP_BATCH];
99 int error = 0; 98 int error = 0;
100 xfs_inode_t *ip; 99 int i;
101 100
102 if (exclusive) 101 read_lock(&pag->pag_ici_lock);
103 write_lock(&pag->pag_ici_lock); 102 nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
104 else 103 (void **)batch, first_index,
105 read_lock(&pag->pag_ici_lock); 104 XFS_LOOKUP_BATCH);
106 ip = xfs_inode_ag_lookup(mp, pag, &first_index, tag); 105 if (!nr_found) {
107 if (!ip) { 106 read_unlock(&pag->pag_ici_lock);
108 if (exclusive)
109 write_unlock(&pag->pag_ici_lock);
110 else
111 read_unlock(&pag->pag_ici_lock);
112 break; 107 break;
113 } 108 }
114 109
115 /* execute releases pag->pag_ici_lock */ 110 /*
116 error = execute(ip, pag, flags); 111 * Grab the inodes before we drop the lock. if we found
117 if (error == EAGAIN) { 112 * nothing, nr == 0 and the loop will be skipped.
118 skipped++; 113 */
119 continue; 114 for (i = 0; i < nr_found; i++) {
115 struct xfs_inode *ip = batch[i];
116
117 if (done || xfs_inode_ag_walk_grab(ip))
118 batch[i] = NULL;
119
120 /*
121 * Update the index for the next lookup. Catch overflows
122 * into the next AG range which can occur if we have inodes
123 * in the last block of the AG and we are currently
124 * pointing to the last inode.
125 */
126 first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
127 if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
128 done = 1;
129 }
130
131 /* unlock now we've grabbed the inodes. */
132 read_unlock(&pag->pag_ici_lock);
133
134 for (i = 0; i < nr_found; i++) {
135 if (!batch[i])
136 continue;
137 error = execute(batch[i], pag, flags);
138 IRELE(batch[i]);
139 if (error == EAGAIN) {
140 skipped++;
141 continue;
142 }
143 if (error && last_error != EFSCORRUPTED)
144 last_error = error;
120 } 145 }
121 if (error)
122 last_error = error;
123 146
124 /* bail out if the filesystem is corrupted. */ 147 /* bail out if the filesystem is corrupted. */
125 if (error == EFSCORRUPTED) 148 if (error == EFSCORRUPTED)
126 break; 149 break;
127 150
128 } while ((*nr_to_scan)--); 151 } while (nr_found && !done);
129 152
130 if (skipped) { 153 if (skipped) {
131 delay(1); 154 delay(1);
@@ -134,110 +157,32 @@ restart:
134 return last_error; 157 return last_error;
135} 158}
136 159
137/*
138 * Select the next per-ag structure to iterate during the walk. The reclaim
139 * walk is optimised only to walk AGs with reclaimable inodes in them.
140 */
141static struct xfs_perag *
142xfs_inode_ag_iter_next_pag(
143 struct xfs_mount *mp,
144 xfs_agnumber_t *first,
145 int tag)
146{
147 struct xfs_perag *pag = NULL;
148
149 if (tag == XFS_ICI_RECLAIM_TAG) {
150 int found;
151 int ref;
152
153 spin_lock(&mp->m_perag_lock);
154 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
155 (void **)&pag, *first, 1, tag);
156 if (found <= 0) {
157 spin_unlock(&mp->m_perag_lock);
158 return NULL;
159 }
160 *first = pag->pag_agno + 1;
161 /* open coded pag reference increment */
162 ref = atomic_inc_return(&pag->pag_ref);
163 spin_unlock(&mp->m_perag_lock);
164 trace_xfs_perag_get_reclaim(mp, pag->pag_agno, ref, _RET_IP_);
165 } else {
166 pag = xfs_perag_get(mp, *first);
167 (*first)++;
168 }
169 return pag;
170}
171
172int 160int
173xfs_inode_ag_iterator( 161xfs_inode_ag_iterator(
174 struct xfs_mount *mp, 162 struct xfs_mount *mp,
175 int (*execute)(struct xfs_inode *ip, 163 int (*execute)(struct xfs_inode *ip,
176 struct xfs_perag *pag, int flags), 164 struct xfs_perag *pag, int flags),
177 int flags, 165 int flags)
178 int tag,
179 int exclusive,
180 int *nr_to_scan)
181{ 166{
182 struct xfs_perag *pag; 167 struct xfs_perag *pag;
183 int error = 0; 168 int error = 0;
184 int last_error = 0; 169 int last_error = 0;
185 xfs_agnumber_t ag; 170 xfs_agnumber_t ag;
186 int nr;
187 171
188 nr = nr_to_scan ? *nr_to_scan : INT_MAX;
189 ag = 0; 172 ag = 0;
190 while ((pag = xfs_inode_ag_iter_next_pag(mp, &ag, tag))) { 173 while ((pag = xfs_perag_get(mp, ag))) {
191 error = xfs_inode_ag_walk(mp, pag, execute, flags, tag, 174 ag = pag->pag_agno + 1;
192 exclusive, &nr); 175 error = xfs_inode_ag_walk(mp, pag, execute, flags);
193 xfs_perag_put(pag); 176 xfs_perag_put(pag);
194 if (error) { 177 if (error) {
195 last_error = error; 178 last_error = error;
196 if (error == EFSCORRUPTED) 179 if (error == EFSCORRUPTED)
197 break; 180 break;
198 } 181 }
199 if (nr <= 0)
200 break;
201 } 182 }
202 if (nr_to_scan)
203 *nr_to_scan = nr;
204 return XFS_ERROR(last_error); 183 return XFS_ERROR(last_error);
205} 184}
206 185
207/* must be called with pag_ici_lock held and releases it */
208int
209xfs_sync_inode_valid(
210 struct xfs_inode *ip,
211 struct xfs_perag *pag)
212{
213 struct inode *inode = VFS_I(ip);
214 int error = EFSCORRUPTED;
215
216 /* nothing to sync during shutdown */
217 if (XFS_FORCED_SHUTDOWN(ip->i_mount))
218 goto out_unlock;
219
220 /* avoid new or reclaimable inodes. Leave for reclaim code to flush */
221 error = ENOENT;
222 if (xfs_iflags_test(ip, XFS_INEW | XFS_IRECLAIMABLE | XFS_IRECLAIM))
223 goto out_unlock;
224
225 /* If we can't grab the inode, it must on it's way to reclaim. */
226 if (!igrab(inode))
227 goto out_unlock;
228
229 if (is_bad_inode(inode)) {
230 IRELE(ip);
231 goto out_unlock;
232 }
233
234 /* inode is valid */
235 error = 0;
236out_unlock:
237 read_unlock(&pag->pag_ici_lock);
238 return error;
239}
240
241STATIC int 186STATIC int
242xfs_sync_inode_data( 187xfs_sync_inode_data(
243 struct xfs_inode *ip, 188 struct xfs_inode *ip,
@@ -248,10 +193,6 @@ xfs_sync_inode_data(
248 struct address_space *mapping = inode->i_mapping; 193 struct address_space *mapping = inode->i_mapping;
249 int error = 0; 194 int error = 0;
250 195
251 error = xfs_sync_inode_valid(ip, pag);
252 if (error)
253 return error;
254
255 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) 196 if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
256 goto out_wait; 197 goto out_wait;
257 198
@@ -268,7 +209,6 @@ xfs_sync_inode_data(
268 out_wait: 209 out_wait:
269 if (flags & SYNC_WAIT) 210 if (flags & SYNC_WAIT)
270 xfs_ioend_wait(ip); 211 xfs_ioend_wait(ip);
271 IRELE(ip);
272 return error; 212 return error;
273} 213}
274 214
@@ -280,10 +220,6 @@ xfs_sync_inode_attr(
280{ 220{
281 int error = 0; 221 int error = 0;
282 222
283 error = xfs_sync_inode_valid(ip, pag);
284 if (error)
285 return error;
286
287 xfs_ilock(ip, XFS_ILOCK_SHARED); 223 xfs_ilock(ip, XFS_ILOCK_SHARED);
288 if (xfs_inode_clean(ip)) 224 if (xfs_inode_clean(ip))
289 goto out_unlock; 225 goto out_unlock;
@@ -302,7 +238,6 @@ xfs_sync_inode_attr(
302 238
303 out_unlock: 239 out_unlock:
304 xfs_iunlock(ip, XFS_ILOCK_SHARED); 240 xfs_iunlock(ip, XFS_ILOCK_SHARED);
305 IRELE(ip);
306 return error; 241 return error;
307} 242}
308 243
@@ -318,8 +253,7 @@ xfs_sync_data(
318 253
319 ASSERT((flags & ~(SYNC_TRYLOCK|SYNC_WAIT)) == 0); 254 ASSERT((flags & ~(SYNC_TRYLOCK|SYNC_WAIT)) == 0);
320 255
321 error = xfs_inode_ag_iterator(mp, xfs_sync_inode_data, flags, 256 error = xfs_inode_ag_iterator(mp, xfs_sync_inode_data, flags);
322 XFS_ICI_NO_TAG, 0, NULL);
323 if (error) 257 if (error)
324 return XFS_ERROR(error); 258 return XFS_ERROR(error);
325 259
@@ -337,8 +271,7 @@ xfs_sync_attr(
337{ 271{
338 ASSERT((flags & ~SYNC_WAIT) == 0); 272 ASSERT((flags & ~SYNC_WAIT) == 0);
339 273
340 return xfs_inode_ag_iterator(mp, xfs_sync_inode_attr, flags, 274 return xfs_inode_ag_iterator(mp, xfs_sync_inode_attr, flags);
341 XFS_ICI_NO_TAG, 0, NULL);
342} 275}
343 276
344STATIC int 277STATIC int
@@ -698,6 +631,43 @@ __xfs_inode_clear_reclaim_tag(
698} 631}
699 632
700/* 633/*
634 * Grab the inode for reclaim exclusively.
635 * Return 0 if we grabbed it, non-zero otherwise.
636 */
637STATIC int
638xfs_reclaim_inode_grab(
639 struct xfs_inode *ip,
640 int flags)
641{
642
643 /*
644 * do some unlocked checks first to avoid unnecceary lock traffic.
645 * The first is a flush lock check, the second is a already in reclaim
646 * check. Only do these checks if we are not going to block on locks.
647 */
648 if ((flags & SYNC_TRYLOCK) &&
649 (!ip->i_flush.done || __xfs_iflags_test(ip, XFS_IRECLAIM))) {
650 return 1;
651 }
652
653 /*
654 * The radix tree lock here protects a thread in xfs_iget from racing
655 * with us starting reclaim on the inode. Once we have the
656 * XFS_IRECLAIM flag set it will not touch us.
657 */
658 spin_lock(&ip->i_flags_lock);
659 ASSERT_ALWAYS(__xfs_iflags_test(ip, XFS_IRECLAIMABLE));
660 if (__xfs_iflags_test(ip, XFS_IRECLAIM)) {
661 /* ignore as it is already under reclaim */
662 spin_unlock(&ip->i_flags_lock);
663 return 1;
664 }
665 __xfs_iflags_set(ip, XFS_IRECLAIM);
666 spin_unlock(&ip->i_flags_lock);
667 return 0;
668}
669
670/*
701 * Inodes in different states need to be treated differently, and the return 671 * Inodes in different states need to be treated differently, and the return
702 * value of xfs_iflush is not sufficient to get this right. The following table 672 * value of xfs_iflush is not sufficient to get this right. The following table
703 * lists the inode states and the reclaim actions necessary for non-blocking 673 * lists the inode states and the reclaim actions necessary for non-blocking
@@ -755,23 +725,6 @@ xfs_reclaim_inode(
755{ 725{
756 int error = 0; 726 int error = 0;
757 727
758 /*
759 * The radix tree lock here protects a thread in xfs_iget from racing
760 * with us starting reclaim on the inode. Once we have the
761 * XFS_IRECLAIM flag set it will not touch us.
762 */
763 spin_lock(&ip->i_flags_lock);
764 ASSERT_ALWAYS(__xfs_iflags_test(ip, XFS_IRECLAIMABLE));
765 if (__xfs_iflags_test(ip, XFS_IRECLAIM)) {
766 /* ignore as it is already under reclaim */
767 spin_unlock(&ip->i_flags_lock);
768 write_unlock(&pag->pag_ici_lock);
769 return 0;
770 }
771 __xfs_iflags_set(ip, XFS_IRECLAIM);
772 spin_unlock(&ip->i_flags_lock);
773 write_unlock(&pag->pag_ici_lock);
774
775 xfs_ilock(ip, XFS_ILOCK_EXCL); 728 xfs_ilock(ip, XFS_ILOCK_EXCL);
776 if (!xfs_iflock_nowait(ip)) { 729 if (!xfs_iflock_nowait(ip)) {
777 if (!(sync_mode & SYNC_WAIT)) 730 if (!(sync_mode & SYNC_WAIT))
@@ -868,13 +821,126 @@ reclaim:
868 821
869} 822}
870 823
824/*
825 * Walk the AGs and reclaim the inodes in them. Even if the filesystem is
826 * corrupted, we still want to try to reclaim all the inodes. If we don't,
827 * then a shut down during filesystem unmount reclaim walk leak all the
828 * unreclaimed inodes.
829 */
830int
831xfs_reclaim_inodes_ag(
832 struct xfs_mount *mp,
833 int flags,
834 int *nr_to_scan)
835{
836 struct xfs_perag *pag;
837 int error = 0;
838 int last_error = 0;
839 xfs_agnumber_t ag;
840 int trylock = flags & SYNC_TRYLOCK;
841 int skipped;
842
843restart:
844 ag = 0;
845 skipped = 0;
846 while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
847 unsigned long first_index = 0;
848 int done = 0;
849 int nr_found = 0;
850
851 ag = pag->pag_agno + 1;
852
853 if (trylock) {
854 if (!mutex_trylock(&pag->pag_ici_reclaim_lock)) {
855 skipped++;
856 continue;
857 }
858 first_index = pag->pag_ici_reclaim_cursor;
859 } else
860 mutex_lock(&pag->pag_ici_reclaim_lock);
861
862 do {
863 struct xfs_inode *batch[XFS_LOOKUP_BATCH];
864 int i;
865
866 write_lock(&pag->pag_ici_lock);
867 nr_found = radix_tree_gang_lookup_tag(
868 &pag->pag_ici_root,
869 (void **)batch, first_index,
870 XFS_LOOKUP_BATCH,
871 XFS_ICI_RECLAIM_TAG);
872 if (!nr_found) {
873 write_unlock(&pag->pag_ici_lock);
874 break;
875 }
876
877 /*
878 * Grab the inodes before we drop the lock. if we found
879 * nothing, nr == 0 and the loop will be skipped.
880 */
881 for (i = 0; i < nr_found; i++) {
882 struct xfs_inode *ip = batch[i];
883
884 if (done || xfs_reclaim_inode_grab(ip, flags))
885 batch[i] = NULL;
886
887 /*
888 * Update the index for the next lookup. Catch
889 * overflows into the next AG range which can
890 * occur if we have inodes in the last block of
891 * the AG and we are currently pointing to the
892 * last inode.
893 */
894 first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
895 if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino))
896 done = 1;
897 }
898
899 /* unlock now we've grabbed the inodes. */
900 write_unlock(&pag->pag_ici_lock);
901
902 for (i = 0; i < nr_found; i++) {
903 if (!batch[i])
904 continue;
905 error = xfs_reclaim_inode(batch[i], pag, flags);
906 if (error && last_error != EFSCORRUPTED)
907 last_error = error;
908 }
909
910 *nr_to_scan -= XFS_LOOKUP_BATCH;
911
912 } while (nr_found && !done && *nr_to_scan > 0);
913
914 if (trylock && !done)
915 pag->pag_ici_reclaim_cursor = first_index;
916 else
917 pag->pag_ici_reclaim_cursor = 0;
918 mutex_unlock(&pag->pag_ici_reclaim_lock);
919 xfs_perag_put(pag);
920 }
921
922 /*
923 * if we skipped any AG, and we still have scan count remaining, do
924 * another pass this time using blocking reclaim semantics (i.e
925 * waiting on the reclaim locks and ignoring the reclaim cursors). This
926 * ensure that when we get more reclaimers than AGs we block rather
927 * than spin trying to execute reclaim.
928 */
929 if (trylock && skipped && *nr_to_scan > 0) {
930 trylock = 0;
931 goto restart;
932 }
933 return XFS_ERROR(last_error);
934}
935
871int 936int
872xfs_reclaim_inodes( 937xfs_reclaim_inodes(
873 xfs_mount_t *mp, 938 xfs_mount_t *mp,
874 int mode) 939 int mode)
875{ 940{
876 return xfs_inode_ag_iterator(mp, xfs_reclaim_inode, mode, 941 int nr_to_scan = INT_MAX;
877 XFS_ICI_RECLAIM_TAG, 1, NULL); 942
943 return xfs_reclaim_inodes_ag(mp, mode, &nr_to_scan);
878} 944}
879 945
880/* 946/*
@@ -896,17 +962,16 @@ xfs_reclaim_inode_shrink(
896 if (!(gfp_mask & __GFP_FS)) 962 if (!(gfp_mask & __GFP_FS))
897 return -1; 963 return -1;
898 964
899 xfs_inode_ag_iterator(mp, xfs_reclaim_inode, 0, 965 xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK, &nr_to_scan);
900 XFS_ICI_RECLAIM_TAG, 1, &nr_to_scan); 966 /* terminate if we don't exhaust the scan */
901 /* if we don't exhaust the scan, don't bother coming back */
902 if (nr_to_scan > 0) 967 if (nr_to_scan > 0)
903 return -1; 968 return -1;
904 } 969 }
905 970
906 reclaimable = 0; 971 reclaimable = 0;
907 ag = 0; 972 ag = 0;
908 while ((pag = xfs_inode_ag_iter_next_pag(mp, &ag, 973 while ((pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
909 XFS_ICI_RECLAIM_TAG))) { 974 ag = pag->pag_agno + 1;
910 reclaimable += pag->pag_ici_reclaimable; 975 reclaimable += pag->pag_ici_reclaimable;
911 xfs_perag_put(pag); 976 xfs_perag_put(pag);
912 } 977 }
diff --git a/fs/xfs/linux-2.6/xfs_sync.h b/fs/xfs/linux-2.6/xfs_sync.h
index fe78726196f8..32ba6628290c 100644
--- a/fs/xfs/linux-2.6/xfs_sync.h
+++ b/fs/xfs/linux-2.6/xfs_sync.h
@@ -47,10 +47,10 @@ void __xfs_inode_set_reclaim_tag(struct xfs_perag *pag, struct xfs_inode *ip);
47void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag, 47void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag,
48 struct xfs_inode *ip); 48 struct xfs_inode *ip);
49 49
50int xfs_sync_inode_valid(struct xfs_inode *ip, struct xfs_perag *pag); 50int xfs_sync_inode_grab(struct xfs_inode *ip);
51int xfs_inode_ag_iterator(struct xfs_mount *mp, 51int xfs_inode_ag_iterator(struct xfs_mount *mp,
52 int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag, int flags), 52 int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag, int flags),
53 int flags, int tag, int write_lock, int *nr_to_scan); 53 int flags);
54 54
55void xfs_inode_shrinker_register(struct xfs_mount *mp); 55void xfs_inode_shrinker_register(struct xfs_mount *mp);
56void xfs_inode_shrinker_unregister(struct xfs_mount *mp); 56void xfs_inode_shrinker_unregister(struct xfs_mount *mp);
diff --git a/fs/xfs/linux-2.6/xfs_trace.h b/fs/xfs/linux-2.6/xfs_trace.h
index 8fe311a456e2..acef2e98c594 100644
--- a/fs/xfs/linux-2.6/xfs_trace.h
+++ b/fs/xfs/linux-2.6/xfs_trace.h
@@ -124,7 +124,7 @@ DEFINE_EVENT(xfs_perag_class, name, \
124 unsigned long caller_ip), \ 124 unsigned long caller_ip), \
125 TP_ARGS(mp, agno, refcount, caller_ip)) 125 TP_ARGS(mp, agno, refcount, caller_ip))
126DEFINE_PERAG_REF_EVENT(xfs_perag_get); 126DEFINE_PERAG_REF_EVENT(xfs_perag_get);
127DEFINE_PERAG_REF_EVENT(xfs_perag_get_reclaim); 127DEFINE_PERAG_REF_EVENT(xfs_perag_get_tag);
128DEFINE_PERAG_REF_EVENT(xfs_perag_put); 128DEFINE_PERAG_REF_EVENT(xfs_perag_put);
129DEFINE_PERAG_REF_EVENT(xfs_perag_set_reclaim); 129DEFINE_PERAG_REF_EVENT(xfs_perag_set_reclaim);
130DEFINE_PERAG_REF_EVENT(xfs_perag_clear_reclaim); 130DEFINE_PERAG_REF_EVENT(xfs_perag_clear_reclaim);
@@ -330,7 +330,7 @@ DEFINE_BUF_EVENT(xfs_buf_iowait_done);
330DEFINE_BUF_EVENT(xfs_buf_delwri_queue); 330DEFINE_BUF_EVENT(xfs_buf_delwri_queue);
331DEFINE_BUF_EVENT(xfs_buf_delwri_dequeue); 331DEFINE_BUF_EVENT(xfs_buf_delwri_dequeue);
332DEFINE_BUF_EVENT(xfs_buf_delwri_split); 332DEFINE_BUF_EVENT(xfs_buf_delwri_split);
333DEFINE_BUF_EVENT(xfs_buf_get_noaddr); 333DEFINE_BUF_EVENT(xfs_buf_get_uncached);
334DEFINE_BUF_EVENT(xfs_bdstrat_shut); 334DEFINE_BUF_EVENT(xfs_bdstrat_shut);
335DEFINE_BUF_EVENT(xfs_buf_item_relse); 335DEFINE_BUF_EVENT(xfs_buf_item_relse);
336DEFINE_BUF_EVENT(xfs_buf_item_iodone); 336DEFINE_BUF_EVENT(xfs_buf_item_iodone);
diff --git a/fs/xfs/linux-2.6/xfs_version.h b/fs/xfs/linux-2.6/xfs_version.h
deleted file mode 100644
index f8d279d7563a..000000000000
--- a/fs/xfs/linux-2.6/xfs_version.h
+++ /dev/null
@@ -1,29 +0,0 @@
1/*
2 * Copyright (c) 2001-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_VERSION_H__
19#define __XFS_VERSION_H__
20
21/*
22 * Dummy file that can contain a timestamp to put into the
23 * XFS init string, to help users keep track of what they're
24 * running
25 */
26
27#define XFS_VERSION_STRING "SGI XFS"
28
29#endif /* __XFS_VERSION_H__ */
diff --git a/fs/xfs/quota/xfs_dquot.c b/fs/xfs/quota/xfs_dquot.c
index e1a2f6800e01..faf8e1a83a12 100644
--- a/fs/xfs/quota/xfs_dquot.c
+++ b/fs/xfs/quota/xfs_dquot.c
@@ -463,87 +463,68 @@ xfs_qm_dqtobp(
463 uint flags) 463 uint flags)
464{ 464{
465 xfs_bmbt_irec_t map; 465 xfs_bmbt_irec_t map;
466 int nmaps, error; 466 int nmaps = 1, error;
467 xfs_buf_t *bp; 467 xfs_buf_t *bp;
468 xfs_inode_t *quotip; 468 xfs_inode_t *quotip = XFS_DQ_TO_QIP(dqp);
469 xfs_mount_t *mp; 469 xfs_mount_t *mp = dqp->q_mount;
470 xfs_disk_dquot_t *ddq; 470 xfs_disk_dquot_t *ddq;
471 xfs_dqid_t id; 471 xfs_dqid_t id = be32_to_cpu(dqp->q_core.d_id);
472 boolean_t newdquot;
473 xfs_trans_t *tp = (tpp ? *tpp : NULL); 472 xfs_trans_t *tp = (tpp ? *tpp : NULL);
474 473
475 mp = dqp->q_mount; 474 dqp->q_fileoffset = (xfs_fileoff_t)id / mp->m_quotainfo->qi_dqperchunk;
476 id = be32_to_cpu(dqp->q_core.d_id);
477 nmaps = 1;
478 newdquot = B_FALSE;
479 475
480 /* 476 xfs_ilock(quotip, XFS_ILOCK_SHARED);
481 * If we don't know where the dquot lives, find out. 477 if (XFS_IS_THIS_QUOTA_OFF(dqp)) {
482 */
483 if (dqp->q_blkno == (xfs_daddr_t) 0) {
484 /* We use the id as an index */
485 dqp->q_fileoffset = (xfs_fileoff_t)id /
486 mp->m_quotainfo->qi_dqperchunk;
487 nmaps = 1;
488 quotip = XFS_DQ_TO_QIP(dqp);
489 xfs_ilock(quotip, XFS_ILOCK_SHARED);
490 /* 478 /*
491 * Return if this type of quotas is turned off while we didn't 479 * Return if this type of quotas is turned off while we
492 * have an inode lock 480 * didn't have the quota inode lock.
493 */ 481 */
494 if (XFS_IS_THIS_QUOTA_OFF(dqp)) { 482 xfs_iunlock(quotip, XFS_ILOCK_SHARED);
495 xfs_iunlock(quotip, XFS_ILOCK_SHARED); 483 return ESRCH;
496 return (ESRCH); 484 }
497 } 485
486 /*
487 * Find the block map; no allocations yet
488 */
489 error = xfs_bmapi(NULL, quotip, dqp->q_fileoffset,
490 XFS_DQUOT_CLUSTER_SIZE_FSB, XFS_BMAPI_METADATA,
491 NULL, 0, &map, &nmaps, NULL);
492
493 xfs_iunlock(quotip, XFS_ILOCK_SHARED);
494 if (error)
495 return error;
496
497 ASSERT(nmaps == 1);
498 ASSERT(map.br_blockcount == 1);
499
500 /*
501 * Offset of dquot in the (fixed sized) dquot chunk.
502 */
503 dqp->q_bufoffset = (id % mp->m_quotainfo->qi_dqperchunk) *
504 sizeof(xfs_dqblk_t);
505
506 ASSERT(map.br_startblock != DELAYSTARTBLOCK);
507 if (map.br_startblock == HOLESTARTBLOCK) {
498 /* 508 /*
499 * Find the block map; no allocations yet 509 * We don't allocate unless we're asked to
500 */ 510 */
501 error = xfs_bmapi(NULL, quotip, dqp->q_fileoffset, 511 if (!(flags & XFS_QMOPT_DQALLOC))
502 XFS_DQUOT_CLUSTER_SIZE_FSB, 512 return ENOENT;
503 XFS_BMAPI_METADATA,
504 NULL, 0, &map, &nmaps, NULL);
505 513
506 xfs_iunlock(quotip, XFS_ILOCK_SHARED); 514 ASSERT(tp);
515 error = xfs_qm_dqalloc(tpp, mp, dqp, quotip,
516 dqp->q_fileoffset, &bp);
507 if (error) 517 if (error)
508 return (error); 518 return error;
509 ASSERT(nmaps == 1); 519 tp = *tpp;
510 ASSERT(map.br_blockcount == 1); 520 } else {
521 trace_xfs_dqtobp_read(dqp);
511 522
512 /* 523 /*
513 * offset of dquot in the (fixed sized) dquot chunk. 524 * store the blkno etc so that we don't have to do the
525 * mapping all the time
514 */ 526 */
515 dqp->q_bufoffset = (id % mp->m_quotainfo->qi_dqperchunk) * 527 dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
516 sizeof(xfs_dqblk_t);
517 if (map.br_startblock == HOLESTARTBLOCK) {
518 /*
519 * We don't allocate unless we're asked to
520 */
521 if (!(flags & XFS_QMOPT_DQALLOC))
522 return (ENOENT);
523
524 ASSERT(tp);
525 if ((error = xfs_qm_dqalloc(tpp, mp, dqp, quotip,
526 dqp->q_fileoffset, &bp)))
527 return (error);
528 tp = *tpp;
529 newdquot = B_TRUE;
530 } else {
531 /*
532 * store the blkno etc so that we don't have to do the
533 * mapping all the time
534 */
535 dqp->q_blkno = XFS_FSB_TO_DADDR(mp, map.br_startblock);
536 }
537 }
538 ASSERT(dqp->q_blkno != DELAYSTARTBLOCK);
539 ASSERT(dqp->q_blkno != HOLESTARTBLOCK);
540
541 /*
542 * Read in the buffer, unless we've just done the allocation
543 * (in which case we already have the buf).
544 */
545 if (!newdquot) {
546 trace_xfs_dqtobp_read(dqp);
547 528
548 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, 529 error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
549 dqp->q_blkno, 530 dqp->q_blkno,
@@ -552,13 +533,14 @@ xfs_qm_dqtobp(
552 if (error || !bp) 533 if (error || !bp)
553 return XFS_ERROR(error); 534 return XFS_ERROR(error);
554 } 535 }
536
555 ASSERT(XFS_BUF_ISBUSY(bp)); 537 ASSERT(XFS_BUF_ISBUSY(bp));
556 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0); 538 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
557 539
558 /* 540 /*
559 * calculate the location of the dquot inside the buffer. 541 * calculate the location of the dquot inside the buffer.
560 */ 542 */
561 ddq = (xfs_disk_dquot_t *)((char *)XFS_BUF_PTR(bp) + dqp->q_bufoffset); 543 ddq = (struct xfs_disk_dquot *)(XFS_BUF_PTR(bp) + dqp->q_bufoffset);
562 544
563 /* 545 /*
564 * A simple sanity check in case we got a corrupted dquot... 546 * A simple sanity check in case we got a corrupted dquot...
@@ -1176,18 +1158,18 @@ xfs_qm_dqflush(
1176 xfs_dquot_t *dqp, 1158 xfs_dquot_t *dqp,
1177 uint flags) 1159 uint flags)
1178{ 1160{
1179 xfs_mount_t *mp; 1161 struct xfs_mount *mp = dqp->q_mount;
1180 xfs_buf_t *bp; 1162 struct xfs_buf *bp;
1181 xfs_disk_dquot_t *ddqp; 1163 struct xfs_disk_dquot *ddqp;
1182 int error; 1164 int error;
1183 1165
1184 ASSERT(XFS_DQ_IS_LOCKED(dqp)); 1166 ASSERT(XFS_DQ_IS_LOCKED(dqp));
1185 ASSERT(!completion_done(&dqp->q_flush)); 1167 ASSERT(!completion_done(&dqp->q_flush));
1168
1186 trace_xfs_dqflush(dqp); 1169 trace_xfs_dqflush(dqp);
1187 1170
1188 /* 1171 /*
1189 * If not dirty, or it's pinned and we are not supposed to 1172 * If not dirty, or it's pinned and we are not supposed to block, nada.
1190 * block, nada.
1191 */ 1173 */
1192 if (!XFS_DQ_IS_DIRTY(dqp) || 1174 if (!XFS_DQ_IS_DIRTY(dqp) ||
1193 (!(flags & SYNC_WAIT) && atomic_read(&dqp->q_pincount) > 0)) { 1175 (!(flags & SYNC_WAIT) && atomic_read(&dqp->q_pincount) > 0)) {
@@ -1201,40 +1183,46 @@ xfs_qm_dqflush(
1201 * down forcibly. If that's the case we must not write this dquot 1183 * down forcibly. If that's the case we must not write this dquot
1202 * to disk, because the log record didn't make it to disk! 1184 * to disk, because the log record didn't make it to disk!
1203 */ 1185 */
1204 if (XFS_FORCED_SHUTDOWN(dqp->q_mount)) { 1186 if (XFS_FORCED_SHUTDOWN(mp)) {
1205 dqp->dq_flags &= ~(XFS_DQ_DIRTY); 1187 dqp->dq_flags &= ~XFS_DQ_DIRTY;
1206 xfs_dqfunlock(dqp); 1188 xfs_dqfunlock(dqp);
1207 return XFS_ERROR(EIO); 1189 return XFS_ERROR(EIO);
1208 } 1190 }
1209 1191
1210 /* 1192 /*
1211 * Get the buffer containing the on-disk dquot 1193 * Get the buffer containing the on-disk dquot
1212 * We don't need a transaction envelope because we know that the
1213 * the ondisk-dquot has already been allocated for.
1214 */ 1194 */
1215 if ((error = xfs_qm_dqtobp(NULL, dqp, &ddqp, &bp, XFS_QMOPT_DOWARN))) { 1195 error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
1196 mp->m_quotainfo->qi_dqchunklen, 0, &bp);
1197 if (error) {
1216 ASSERT(error != ENOENT); 1198 ASSERT(error != ENOENT);
1217 /*
1218 * Quotas could have gotten turned off (ESRCH)
1219 */
1220 xfs_dqfunlock(dqp); 1199 xfs_dqfunlock(dqp);
1221 return (error); 1200 return error;
1222 } 1201 }
1223 1202
1224 if (xfs_qm_dqcheck(&dqp->q_core, be32_to_cpu(ddqp->d_id), 1203 /*
1225 0, XFS_QMOPT_DOWARN, "dqflush (incore copy)")) { 1204 * Calculate the location of the dquot inside the buffer.
1226 xfs_force_shutdown(dqp->q_mount, SHUTDOWN_CORRUPT_INCORE); 1205 */
1206 ddqp = (struct xfs_disk_dquot *)(XFS_BUF_PTR(bp) + dqp->q_bufoffset);
1207
1208 /*
1209 * A simple sanity check in case we got a corrupted dquot..
1210 */
1211 if (xfs_qm_dqcheck(&dqp->q_core, be32_to_cpu(ddqp->d_id), 0,
1212 XFS_QMOPT_DOWARN, "dqflush (incore copy)")) {
1213 xfs_buf_relse(bp);
1214 xfs_dqfunlock(dqp);
1215 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1227 return XFS_ERROR(EIO); 1216 return XFS_ERROR(EIO);
1228 } 1217 }
1229 1218
1230 /* This is the only portion of data that needs to persist */ 1219 /* This is the only portion of data that needs to persist */
1231 memcpy(ddqp, &(dqp->q_core), sizeof(xfs_disk_dquot_t)); 1220 memcpy(ddqp, &dqp->q_core, sizeof(xfs_disk_dquot_t));
1232 1221
1233 /* 1222 /*
1234 * Clear the dirty field and remember the flush lsn for later use. 1223 * Clear the dirty field and remember the flush lsn for later use.
1235 */ 1224 */
1236 dqp->dq_flags &= ~(XFS_DQ_DIRTY); 1225 dqp->dq_flags &= ~XFS_DQ_DIRTY;
1237 mp = dqp->q_mount;
1238 1226
1239 xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn, 1227 xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
1240 &dqp->q_logitem.qli_item.li_lsn); 1228 &dqp->q_logitem.qli_item.li_lsn);
diff --git a/fs/xfs/quota/xfs_qm.c b/fs/xfs/quota/xfs_qm.c
index 9a92407109a1..f8e854b4fde8 100644
--- a/fs/xfs/quota/xfs_qm.c
+++ b/fs/xfs/quota/xfs_qm.c
@@ -55,8 +55,6 @@ uint ndquot;
55kmem_zone_t *qm_dqzone; 55kmem_zone_t *qm_dqzone;
56kmem_zone_t *qm_dqtrxzone; 56kmem_zone_t *qm_dqtrxzone;
57 57
58static cred_t xfs_zerocr;
59
60STATIC void xfs_qm_list_init(xfs_dqlist_t *, char *, int); 58STATIC void xfs_qm_list_init(xfs_dqlist_t *, char *, int);
61STATIC void xfs_qm_list_destroy(xfs_dqlist_t *); 59STATIC void xfs_qm_list_destroy(xfs_dqlist_t *);
62 60
@@ -837,7 +835,7 @@ xfs_qm_dqattach_locked(
837 xfs_qm_dqattach_one(ip, ip->i_d.di_gid, XFS_DQ_GROUP, 835 xfs_qm_dqattach_one(ip, ip->i_d.di_gid, XFS_DQ_GROUP,
838 flags & XFS_QMOPT_DQALLOC, 836 flags & XFS_QMOPT_DQALLOC,
839 ip->i_udquot, &ip->i_gdquot) : 837 ip->i_udquot, &ip->i_gdquot) :
840 xfs_qm_dqattach_one(ip, ip->i_d.di_projid, XFS_DQ_PROJ, 838 xfs_qm_dqattach_one(ip, xfs_get_projid(ip), XFS_DQ_PROJ,
841 flags & XFS_QMOPT_DQALLOC, 839 flags & XFS_QMOPT_DQALLOC,
842 ip->i_udquot, &ip->i_gdquot); 840 ip->i_udquot, &ip->i_gdquot);
843 /* 841 /*
@@ -1199,87 +1197,6 @@ xfs_qm_list_destroy(
1199 mutex_destroy(&(list->qh_lock)); 1197 mutex_destroy(&(list->qh_lock));
1200} 1198}
1201 1199
1202
1203/*
1204 * Stripped down version of dqattach. This doesn't attach, or even look at the
1205 * dquots attached to the inode. The rationale is that there won't be any
1206 * attached at the time this is called from quotacheck.
1207 */
1208STATIC int
1209xfs_qm_dqget_noattach(
1210 xfs_inode_t *ip,
1211 xfs_dquot_t **O_udqpp,
1212 xfs_dquot_t **O_gdqpp)
1213{
1214 int error;
1215 xfs_mount_t *mp;
1216 xfs_dquot_t *udqp, *gdqp;
1217
1218 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
1219 mp = ip->i_mount;
1220 udqp = NULL;
1221 gdqp = NULL;
1222
1223 if (XFS_IS_UQUOTA_ON(mp)) {
1224 ASSERT(ip->i_udquot == NULL);
1225 /*
1226 * We want the dquot allocated if it doesn't exist.
1227 */
1228 if ((error = xfs_qm_dqget(mp, ip, ip->i_d.di_uid, XFS_DQ_USER,
1229 XFS_QMOPT_DQALLOC | XFS_QMOPT_DOWARN,
1230 &udqp))) {
1231 /*
1232 * Shouldn't be able to turn off quotas here.
1233 */
1234 ASSERT(error != ESRCH);
1235 ASSERT(error != ENOENT);
1236 return error;
1237 }
1238 ASSERT(udqp);
1239 }
1240
1241 if (XFS_IS_OQUOTA_ON(mp)) {
1242 ASSERT(ip->i_gdquot == NULL);
1243 if (udqp)
1244 xfs_dqunlock(udqp);
1245 error = XFS_IS_GQUOTA_ON(mp) ?
1246 xfs_qm_dqget(mp, ip,
1247 ip->i_d.di_gid, XFS_DQ_GROUP,
1248 XFS_QMOPT_DQALLOC|XFS_QMOPT_DOWARN,
1249 &gdqp) :
1250 xfs_qm_dqget(mp, ip,
1251 ip->i_d.di_projid, XFS_DQ_PROJ,
1252 XFS_QMOPT_DQALLOC|XFS_QMOPT_DOWARN,
1253 &gdqp);
1254 if (error) {
1255 if (udqp)
1256 xfs_qm_dqrele(udqp);
1257 ASSERT(error != ESRCH);
1258 ASSERT(error != ENOENT);
1259 return error;
1260 }
1261 ASSERT(gdqp);
1262
1263 /* Reacquire the locks in the right order */
1264 if (udqp) {
1265 if (! xfs_qm_dqlock_nowait(udqp)) {
1266 xfs_dqunlock(gdqp);
1267 xfs_dqlock(udqp);
1268 xfs_dqlock(gdqp);
1269 }
1270 }
1271 }
1272
1273 *O_udqpp = udqp;
1274 *O_gdqpp = gdqp;
1275
1276#ifdef QUOTADEBUG
1277 if (udqp) ASSERT(XFS_DQ_IS_LOCKED(udqp));
1278 if (gdqp) ASSERT(XFS_DQ_IS_LOCKED(gdqp));
1279#endif
1280 return 0;
1281}
1282
1283/* 1200/*
1284 * Create an inode and return with a reference already taken, but unlocked 1201 * Create an inode and return with a reference already taken, but unlocked
1285 * This is how we create quota inodes 1202 * This is how we create quota inodes
@@ -1305,8 +1222,8 @@ xfs_qm_qino_alloc(
1305 return error; 1222 return error;
1306 } 1223 }
1307 1224
1308 if ((error = xfs_dir_ialloc(&tp, NULL, S_IFREG, 1, 0, 1225 error = xfs_dir_ialloc(&tp, NULL, S_IFREG, 1, 0, 0, 1, ip, &committed);
1309 &xfs_zerocr, 0, 1, ip, &committed))) { 1226 if (error) {
1310 xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | 1227 xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES |
1311 XFS_TRANS_ABORT); 1228 XFS_TRANS_ABORT);
1312 return error; 1229 return error;
@@ -1516,7 +1433,7 @@ xfs_qm_dqiterate(
1516 rablkcnt = map[i+1].br_blockcount; 1433 rablkcnt = map[i+1].br_blockcount;
1517 rablkno = map[i+1].br_startblock; 1434 rablkno = map[i+1].br_startblock;
1518 while (rablkcnt--) { 1435 while (rablkcnt--) {
1519 xfs_baread(mp->m_ddev_targp, 1436 xfs_buf_readahead(mp->m_ddev_targp,
1520 XFS_FSB_TO_DADDR(mp, rablkno), 1437 XFS_FSB_TO_DADDR(mp, rablkno),
1521 mp->m_quotainfo->qi_dqchunklen); 1438 mp->m_quotainfo->qi_dqchunklen);
1522 rablkno++; 1439 rablkno++;
@@ -1546,18 +1463,34 @@ xfs_qm_dqiterate(
1546 1463
1547/* 1464/*
1548 * Called by dqusage_adjust in doing a quotacheck. 1465 * Called by dqusage_adjust in doing a quotacheck.
1549 * Given the inode, and a dquot (either USR or GRP, doesn't matter), 1466 *
1550 * this updates its incore copy as well as the buffer copy. This is 1467 * Given the inode, and a dquot id this updates both the incore dqout as well
1551 * so that once the quotacheck is done, we can just log all the buffers, 1468 * as the buffer copy. This is so that once the quotacheck is done, we can
1552 * as opposed to logging numerous updates to individual dquots. 1469 * just log all the buffers, as opposed to logging numerous updates to
1470 * individual dquots.
1553 */ 1471 */
1554STATIC void 1472STATIC int
1555xfs_qm_quotacheck_dqadjust( 1473xfs_qm_quotacheck_dqadjust(
1556 xfs_dquot_t *dqp, 1474 struct xfs_inode *ip,
1475 xfs_dqid_t id,
1476 uint type,
1557 xfs_qcnt_t nblks, 1477 xfs_qcnt_t nblks,
1558 xfs_qcnt_t rtblks) 1478 xfs_qcnt_t rtblks)
1559{ 1479{
1560 ASSERT(XFS_DQ_IS_LOCKED(dqp)); 1480 struct xfs_mount *mp = ip->i_mount;
1481 struct xfs_dquot *dqp;
1482 int error;
1483
1484 error = xfs_qm_dqget(mp, ip, id, type,
1485 XFS_QMOPT_DQALLOC | XFS_QMOPT_DOWARN, &dqp);
1486 if (error) {
1487 /*
1488 * Shouldn't be able to turn off quotas here.
1489 */
1490 ASSERT(error != ESRCH);
1491 ASSERT(error != ENOENT);
1492 return error;
1493 }
1561 1494
1562 trace_xfs_dqadjust(dqp); 1495 trace_xfs_dqadjust(dqp);
1563 1496
@@ -1582,11 +1515,13 @@ xfs_qm_quotacheck_dqadjust(
1582 * There are no timers for the default values set in the root dquot. 1515 * There are no timers for the default values set in the root dquot.
1583 */ 1516 */
1584 if (dqp->q_core.d_id) { 1517 if (dqp->q_core.d_id) {
1585 xfs_qm_adjust_dqlimits(dqp->q_mount, &dqp->q_core); 1518 xfs_qm_adjust_dqlimits(mp, &dqp->q_core);
1586 xfs_qm_adjust_dqtimers(dqp->q_mount, &dqp->q_core); 1519 xfs_qm_adjust_dqtimers(mp, &dqp->q_core);
1587 } 1520 }
1588 1521
1589 dqp->dq_flags |= XFS_DQ_DIRTY; 1522 dqp->dq_flags |= XFS_DQ_DIRTY;
1523 xfs_qm_dqput(dqp);
1524 return 0;
1590} 1525}
1591 1526
1592STATIC int 1527STATIC int
@@ -1629,8 +1564,7 @@ xfs_qm_dqusage_adjust(
1629 int *res) /* result code value */ 1564 int *res) /* result code value */
1630{ 1565{
1631 xfs_inode_t *ip; 1566 xfs_inode_t *ip;
1632 xfs_dquot_t *udqp, *gdqp; 1567 xfs_qcnt_t nblks, rtblks = 0;
1633 xfs_qcnt_t nblks, rtblks;
1634 int error; 1568 int error;
1635 1569
1636 ASSERT(XFS_IS_QUOTA_RUNNING(mp)); 1570 ASSERT(XFS_IS_QUOTA_RUNNING(mp));
@@ -1650,51 +1584,24 @@ xfs_qm_dqusage_adjust(
1650 * the case in all other instances. It's OK that we do this because 1584 * the case in all other instances. It's OK that we do this because
1651 * quotacheck is done only at mount time. 1585 * quotacheck is done only at mount time.
1652 */ 1586 */
1653 if ((error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_EXCL, &ip))) { 1587 error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_EXCL, &ip);
1588 if (error) {
1654 *res = BULKSTAT_RV_NOTHING; 1589 *res = BULKSTAT_RV_NOTHING;
1655 return error; 1590 return error;
1656 } 1591 }
1657 1592
1658 /* 1593 ASSERT(ip->i_delayed_blks == 0);
1659 * Obtain the locked dquots. In case of an error (eg. allocation
1660 * fails for ENOSPC), we return the negative of the error number
1661 * to bulkstat, so that it can get propagated to quotacheck() and
1662 * making us disable quotas for the file system.
1663 */
1664 if ((error = xfs_qm_dqget_noattach(ip, &udqp, &gdqp))) {
1665 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1666 IRELE(ip);
1667 *res = BULKSTAT_RV_GIVEUP;
1668 return error;
1669 }
1670 1594
1671 rtblks = 0; 1595 if (XFS_IS_REALTIME_INODE(ip)) {
1672 if (! XFS_IS_REALTIME_INODE(ip)) {
1673 nblks = (xfs_qcnt_t)ip->i_d.di_nblocks;
1674 } else {
1675 /* 1596 /*
1676 * Walk thru the extent list and count the realtime blocks. 1597 * Walk thru the extent list and count the realtime blocks.
1677 */ 1598 */
1678 if ((error = xfs_qm_get_rtblks(ip, &rtblks))) { 1599 error = xfs_qm_get_rtblks(ip, &rtblks);
1679 xfs_iunlock(ip, XFS_ILOCK_EXCL); 1600 if (error)
1680 IRELE(ip); 1601 goto error0;
1681 if (udqp)
1682 xfs_qm_dqput(udqp);
1683 if (gdqp)
1684 xfs_qm_dqput(gdqp);
1685 *res = BULKSTAT_RV_GIVEUP;
1686 return error;
1687 }
1688 nblks = (xfs_qcnt_t)ip->i_d.di_nblocks - rtblks;
1689 } 1602 }
1690 ASSERT(ip->i_delayed_blks == 0);
1691 1603
1692 /* 1604 nblks = (xfs_qcnt_t)ip->i_d.di_nblocks - rtblks;
1693 * We can't release the inode while holding its dquot locks.
1694 * The inode can go into inactive and might try to acquire the dquotlocks.
1695 * So, just unlock here and do a vn_rele at the end.
1696 */
1697 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1698 1605
1699 /* 1606 /*
1700 * Add the (disk blocks and inode) resources occupied by this 1607 * Add the (disk blocks and inode) resources occupied by this
@@ -1709,26 +1616,36 @@ xfs_qm_dqusage_adjust(
1709 * and quotaoffs don't race. (Quotachecks happen at mount time only). 1616 * and quotaoffs don't race. (Quotachecks happen at mount time only).
1710 */ 1617 */
1711 if (XFS_IS_UQUOTA_ON(mp)) { 1618 if (XFS_IS_UQUOTA_ON(mp)) {
1712 ASSERT(udqp); 1619 error = xfs_qm_quotacheck_dqadjust(ip, ip->i_d.di_uid,
1713 xfs_qm_quotacheck_dqadjust(udqp, nblks, rtblks); 1620 XFS_DQ_USER, nblks, rtblks);
1714 xfs_qm_dqput(udqp); 1621 if (error)
1622 goto error0;
1715 } 1623 }
1716 if (XFS_IS_OQUOTA_ON(mp)) { 1624
1717 ASSERT(gdqp); 1625 if (XFS_IS_GQUOTA_ON(mp)) {
1718 xfs_qm_quotacheck_dqadjust(gdqp, nblks, rtblks); 1626 error = xfs_qm_quotacheck_dqadjust(ip, ip->i_d.di_gid,
1719 xfs_qm_dqput(gdqp); 1627 XFS_DQ_GROUP, nblks, rtblks);
1628 if (error)
1629 goto error0;
1720 } 1630 }
1721 /*
1722 * Now release the inode. This will send it to 'inactive', and
1723 * possibly even free blocks.
1724 */
1725 IRELE(ip);
1726 1631
1727 /* 1632 if (XFS_IS_PQUOTA_ON(mp)) {
1728 * Goto next inode. 1633 error = xfs_qm_quotacheck_dqadjust(ip, xfs_get_projid(ip),
1729 */ 1634 XFS_DQ_PROJ, nblks, rtblks);
1635 if (error)
1636 goto error0;
1637 }
1638
1639 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1640 IRELE(ip);
1730 *res = BULKSTAT_RV_DIDONE; 1641 *res = BULKSTAT_RV_DIDONE;
1731 return 0; 1642 return 0;
1643
1644error0:
1645 xfs_iunlock(ip, XFS_ILOCK_EXCL);
1646 IRELE(ip);
1647 *res = BULKSTAT_RV_GIVEUP;
1648 return error;
1732} 1649}
1733 1650
1734/* 1651/*
@@ -2224,7 +2141,7 @@ xfs_qm_write_sb_changes(
2224 2141
2225 2142
2226/* 2143/*
2227 * Given an inode, a uid and gid (from cred_t) make sure that we have 2144 * Given an inode, a uid, gid and prid make sure that we have
2228 * allocated relevant dquot(s) on disk, and that we won't exceed inode 2145 * allocated relevant dquot(s) on disk, and that we won't exceed inode
2229 * quotas by creating this file. 2146 * quotas by creating this file.
2230 * This also attaches dquot(s) to the given inode after locking it, 2147 * This also attaches dquot(s) to the given inode after locking it,
@@ -2332,7 +2249,7 @@ xfs_qm_vop_dqalloc(
2332 xfs_dqunlock(gq); 2249 xfs_dqunlock(gq);
2333 } 2250 }
2334 } else if ((flags & XFS_QMOPT_PQUOTA) && XFS_IS_PQUOTA_ON(mp)) { 2251 } else if ((flags & XFS_QMOPT_PQUOTA) && XFS_IS_PQUOTA_ON(mp)) {
2335 if (ip->i_d.di_projid != prid) { 2252 if (xfs_get_projid(ip) != prid) {
2336 xfs_iunlock(ip, lockflags); 2253 xfs_iunlock(ip, lockflags);
2337 if ((error = xfs_qm_dqget(mp, NULL, (xfs_dqid_t)prid, 2254 if ((error = xfs_qm_dqget(mp, NULL, (xfs_dqid_t)prid,
2338 XFS_DQ_PROJ, 2255 XFS_DQ_PROJ,
@@ -2454,7 +2371,7 @@ xfs_qm_vop_chown_reserve(
2454 } 2371 }
2455 if (XFS_IS_OQUOTA_ON(ip->i_mount) && gdqp) { 2372 if (XFS_IS_OQUOTA_ON(ip->i_mount) && gdqp) {
2456 if (XFS_IS_PQUOTA_ON(ip->i_mount) && 2373 if (XFS_IS_PQUOTA_ON(ip->i_mount) &&
2457 ip->i_d.di_projid != be32_to_cpu(gdqp->q_core.d_id)) 2374 xfs_get_projid(ip) != be32_to_cpu(gdqp->q_core.d_id))
2458 prjflags = XFS_QMOPT_ENOSPC; 2375 prjflags = XFS_QMOPT_ENOSPC;
2459 2376
2460 if (prjflags || 2377 if (prjflags ||
@@ -2558,7 +2475,7 @@ xfs_qm_vop_create_dqattach(
2558 ip->i_gdquot = gdqp; 2475 ip->i_gdquot = gdqp;
2559 ASSERT(XFS_IS_OQUOTA_ON(mp)); 2476 ASSERT(XFS_IS_OQUOTA_ON(mp));
2560 ASSERT((XFS_IS_GQUOTA_ON(mp) ? 2477 ASSERT((XFS_IS_GQUOTA_ON(mp) ?
2561 ip->i_d.di_gid : ip->i_d.di_projid) == 2478 ip->i_d.di_gid : xfs_get_projid(ip)) ==
2562 be32_to_cpu(gdqp->q_core.d_id)); 2479 be32_to_cpu(gdqp->q_core.d_id));
2563 xfs_trans_mod_dquot(tp, gdqp, XFS_TRANS_DQ_ICOUNT, 1); 2480 xfs_trans_mod_dquot(tp, gdqp, XFS_TRANS_DQ_ICOUNT, 1);
2564 } 2481 }
diff --git a/fs/xfs/quota/xfs_qm_bhv.c b/fs/xfs/quota/xfs_qm_bhv.c
index bea02d786c5d..45b5cb1788ab 100644
--- a/fs/xfs/quota/xfs_qm_bhv.c
+++ b/fs/xfs/quota/xfs_qm_bhv.c
@@ -81,7 +81,7 @@ xfs_qm_statvfs(
81 xfs_mount_t *mp = ip->i_mount; 81 xfs_mount_t *mp = ip->i_mount;
82 xfs_dquot_t *dqp; 82 xfs_dquot_t *dqp;
83 83
84 if (!xfs_qm_dqget(mp, NULL, ip->i_d.di_projid, XFS_DQ_PROJ, 0, &dqp)) { 84 if (!xfs_qm_dqget(mp, NULL, xfs_get_projid(ip), XFS_DQ_PROJ, 0, &dqp)) {
85 xfs_fill_statvfs_from_dquot(statp, &dqp->q_core); 85 xfs_fill_statvfs_from_dquot(statp, &dqp->q_core);
86 xfs_qm_dqput(dqp); 86 xfs_qm_dqput(dqp);
87 } 87 }
diff --git a/fs/xfs/quota/xfs_qm_syscalls.c b/fs/xfs/quota/xfs_qm_syscalls.c
index 45e5849df238..bdebc183223e 100644
--- a/fs/xfs/quota/xfs_qm_syscalls.c
+++ b/fs/xfs/quota/xfs_qm_syscalls.c
@@ -276,7 +276,7 @@ xfs_qm_scall_trunc_qfile(
276 goto out_unlock; 276 goto out_unlock;
277 } 277 }
278 278
279 xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 279 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
280 error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); 280 error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
281 281
282out_unlock: 282out_unlock:
@@ -875,21 +875,14 @@ xfs_dqrele_inode(
875 struct xfs_perag *pag, 875 struct xfs_perag *pag,
876 int flags) 876 int flags)
877{ 877{
878 int error;
879
880 /* skip quota inodes */ 878 /* skip quota inodes */
881 if (ip == ip->i_mount->m_quotainfo->qi_uquotaip || 879 if (ip == ip->i_mount->m_quotainfo->qi_uquotaip ||
882 ip == ip->i_mount->m_quotainfo->qi_gquotaip) { 880 ip == ip->i_mount->m_quotainfo->qi_gquotaip) {
883 ASSERT(ip->i_udquot == NULL); 881 ASSERT(ip->i_udquot == NULL);
884 ASSERT(ip->i_gdquot == NULL); 882 ASSERT(ip->i_gdquot == NULL);
885 read_unlock(&pag->pag_ici_lock);
886 return 0; 883 return 0;
887 } 884 }
888 885
889 error = xfs_sync_inode_valid(ip, pag);
890 if (error)
891 return error;
892
893 xfs_ilock(ip, XFS_ILOCK_EXCL); 886 xfs_ilock(ip, XFS_ILOCK_EXCL);
894 if ((flags & XFS_UQUOTA_ACCT) && ip->i_udquot) { 887 if ((flags & XFS_UQUOTA_ACCT) && ip->i_udquot) {
895 xfs_qm_dqrele(ip->i_udquot); 888 xfs_qm_dqrele(ip->i_udquot);
@@ -900,8 +893,6 @@ xfs_dqrele_inode(
900 ip->i_gdquot = NULL; 893 ip->i_gdquot = NULL;
901 } 894 }
902 xfs_iunlock(ip, XFS_ILOCK_EXCL); 895 xfs_iunlock(ip, XFS_ILOCK_EXCL);
903
904 IRELE(ip);
905 return 0; 896 return 0;
906} 897}
907 898
@@ -918,8 +909,7 @@ xfs_qm_dqrele_all_inodes(
918 uint flags) 909 uint flags)
919{ 910{
920 ASSERT(mp->m_quotainfo); 911 ASSERT(mp->m_quotainfo);
921 xfs_inode_ag_iterator(mp, xfs_dqrele_inode, flags, 912 xfs_inode_ag_iterator(mp, xfs_dqrele_inode, flags);
922 XFS_ICI_NO_TAG, 0, NULL);
923} 913}
924 914
925/*------------------------------------------------------------------------*/ 915/*------------------------------------------------------------------------*/
@@ -1175,7 +1165,7 @@ xfs_qm_internalqcheck_adjust(
1175 } 1165 }
1176 xfs_qm_internalqcheck_get_dquots(mp, 1166 xfs_qm_internalqcheck_get_dquots(mp,
1177 (xfs_dqid_t) ip->i_d.di_uid, 1167 (xfs_dqid_t) ip->i_d.di_uid,
1178 (xfs_dqid_t) ip->i_d.di_projid, 1168 (xfs_dqid_t) xfs_get_projid(ip),
1179 (xfs_dqid_t) ip->i_d.di_gid, 1169 (xfs_dqid_t) ip->i_d.di_gid,
1180 &ud, &gd); 1170 &ud, &gd);
1181 if (XFS_IS_UQUOTA_ON(mp)) { 1171 if (XFS_IS_UQUOTA_ON(mp)) {
diff --git a/fs/xfs/xfs_ag.h b/fs/xfs/xfs_ag.h
index 4917d4eed4ed..63c7a1a6c022 100644
--- a/fs/xfs/xfs_ag.h
+++ b/fs/xfs/xfs_ag.h
@@ -230,6 +230,15 @@ typedef struct xfs_perag {
230 rwlock_t pag_ici_lock; /* incore inode lock */ 230 rwlock_t pag_ici_lock; /* incore inode lock */
231 struct radix_tree_root pag_ici_root; /* incore inode cache root */ 231 struct radix_tree_root pag_ici_root; /* incore inode cache root */
232 int pag_ici_reclaimable; /* reclaimable inodes */ 232 int pag_ici_reclaimable; /* reclaimable inodes */
233 struct mutex pag_ici_reclaim_lock; /* serialisation point */
234 unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */
235
236 /* buffer cache index */
237 spinlock_t pag_buf_lock; /* lock for pag_buf_tree */
238 struct rb_root pag_buf_tree; /* ordered tree of active buffers */
239
240 /* for rcu-safe freeing */
241 struct rcu_head rcu_head;
233#endif 242#endif
234 int pagb_count; /* pagb slots in use */ 243 int pagb_count; /* pagb slots in use */
235} xfs_perag_t; 244} xfs_perag_t;
diff --git a/fs/xfs/xfs_alloc.c b/fs/xfs/xfs_alloc.c
index af168faccc7a..112abc439ca5 100644
--- a/fs/xfs/xfs_alloc.c
+++ b/fs/xfs/xfs_alloc.c
@@ -675,7 +675,7 @@ xfs_alloc_ag_vextent_near(
675 xfs_agblock_t gtbnoa; /* aligned ... */ 675 xfs_agblock_t gtbnoa; /* aligned ... */
676 xfs_extlen_t gtdiff; /* difference to right side entry */ 676 xfs_extlen_t gtdiff; /* difference to right side entry */
677 xfs_extlen_t gtlen; /* length of right side entry */ 677 xfs_extlen_t gtlen; /* length of right side entry */
678 xfs_extlen_t gtlena; /* aligned ... */ 678 xfs_extlen_t gtlena = 0; /* aligned ... */
679 xfs_agblock_t gtnew; /* useful start bno of right side */ 679 xfs_agblock_t gtnew; /* useful start bno of right side */
680 int error; /* error code */ 680 int error; /* error code */
681 int i; /* result code, temporary */ 681 int i; /* result code, temporary */
@@ -684,7 +684,7 @@ xfs_alloc_ag_vextent_near(
684 xfs_agblock_t ltbnoa; /* aligned ... */ 684 xfs_agblock_t ltbnoa; /* aligned ... */
685 xfs_extlen_t ltdiff; /* difference to left side entry */ 685 xfs_extlen_t ltdiff; /* difference to left side entry */
686 xfs_extlen_t ltlen; /* length of left side entry */ 686 xfs_extlen_t ltlen; /* length of left side entry */
687 xfs_extlen_t ltlena; /* aligned ... */ 687 xfs_extlen_t ltlena = 0; /* aligned ... */
688 xfs_agblock_t ltnew; /* useful start bno of left side */ 688 xfs_agblock_t ltnew; /* useful start bno of left side */
689 xfs_extlen_t rlen; /* length of returned extent */ 689 xfs_extlen_t rlen; /* length of returned extent */
690#if defined(DEBUG) && defined(__KERNEL__) 690#if defined(DEBUG) && defined(__KERNEL__)
diff --git a/fs/xfs/xfs_alloc_btree.c b/fs/xfs/xfs_alloc_btree.c
index 97f7328967fd..3916925e2584 100644
--- a/fs/xfs/xfs_alloc_btree.c
+++ b/fs/xfs/xfs_alloc_btree.c
@@ -280,38 +280,6 @@ xfs_allocbt_key_diff(
280 return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock; 280 return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
281} 281}
282 282
283STATIC int
284xfs_allocbt_kill_root(
285 struct xfs_btree_cur *cur,
286 struct xfs_buf *bp,
287 int level,
288 union xfs_btree_ptr *newroot)
289{
290 int error;
291
292 XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
293 XFS_BTREE_STATS_INC(cur, killroot);
294
295 /*
296 * Update the root pointer, decreasing the level by 1 and then
297 * free the old root.
298 */
299 xfs_allocbt_set_root(cur, newroot, -1);
300 error = xfs_allocbt_free_block(cur, bp);
301 if (error) {
302 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
303 return error;
304 }
305
306 XFS_BTREE_STATS_INC(cur, free);
307
308 xfs_btree_setbuf(cur, level, NULL);
309 cur->bc_nlevels--;
310
311 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
312 return 0;
313}
314
315#ifdef DEBUG 283#ifdef DEBUG
316STATIC int 284STATIC int
317xfs_allocbt_keys_inorder( 285xfs_allocbt_keys_inorder(
@@ -423,7 +391,6 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
423 391
424 .dup_cursor = xfs_allocbt_dup_cursor, 392 .dup_cursor = xfs_allocbt_dup_cursor,
425 .set_root = xfs_allocbt_set_root, 393 .set_root = xfs_allocbt_set_root,
426 .kill_root = xfs_allocbt_kill_root,
427 .alloc_block = xfs_allocbt_alloc_block, 394 .alloc_block = xfs_allocbt_alloc_block,
428 .free_block = xfs_allocbt_free_block, 395 .free_block = xfs_allocbt_free_block,
429 .update_lastrec = xfs_allocbt_update_lastrec, 396 .update_lastrec = xfs_allocbt_update_lastrec,
diff --git a/fs/xfs/xfs_attr.c b/fs/xfs/xfs_attr.c
index c2568242a901..c86375378810 100644
--- a/fs/xfs/xfs_attr.c
+++ b/fs/xfs/xfs_attr.c
@@ -355,16 +355,15 @@ xfs_attr_set_int(
355 if (mp->m_flags & XFS_MOUNT_WSYNC) { 355 if (mp->m_flags & XFS_MOUNT_WSYNC) {
356 xfs_trans_set_sync(args.trans); 356 xfs_trans_set_sync(args.trans);
357 } 357 }
358
359 if (!error && (flags & ATTR_KERNOTIME) == 0) {
360 xfs_trans_ichgtime(args.trans, dp,
361 XFS_ICHGTIME_CHG);
362 }
358 err2 = xfs_trans_commit(args.trans, 363 err2 = xfs_trans_commit(args.trans,
359 XFS_TRANS_RELEASE_LOG_RES); 364 XFS_TRANS_RELEASE_LOG_RES);
360 xfs_iunlock(dp, XFS_ILOCK_EXCL); 365 xfs_iunlock(dp, XFS_ILOCK_EXCL);
361 366
362 /*
363 * Hit the inode change time.
364 */
365 if (!error && (flags & ATTR_KERNOTIME) == 0) {
366 xfs_ichgtime(dp, XFS_ICHGTIME_CHG);
367 }
368 return(error == 0 ? err2 : error); 367 return(error == 0 ? err2 : error);
369 } 368 }
370 369
@@ -420,6 +419,9 @@ xfs_attr_set_int(
420 xfs_trans_set_sync(args.trans); 419 xfs_trans_set_sync(args.trans);
421 } 420 }
422 421
422 if ((flags & ATTR_KERNOTIME) == 0)
423 xfs_trans_ichgtime(args.trans, dp, XFS_ICHGTIME_CHG);
424
423 /* 425 /*
424 * Commit the last in the sequence of transactions. 426 * Commit the last in the sequence of transactions.
425 */ 427 */
@@ -427,13 +429,6 @@ xfs_attr_set_int(
427 error = xfs_trans_commit(args.trans, XFS_TRANS_RELEASE_LOG_RES); 429 error = xfs_trans_commit(args.trans, XFS_TRANS_RELEASE_LOG_RES);
428 xfs_iunlock(dp, XFS_ILOCK_EXCL); 430 xfs_iunlock(dp, XFS_ILOCK_EXCL);
429 431
430 /*
431 * Hit the inode change time.
432 */
433 if (!error && (flags & ATTR_KERNOTIME) == 0) {
434 xfs_ichgtime(dp, XFS_ICHGTIME_CHG);
435 }
436
437 return(error); 432 return(error);
438 433
439out: 434out:
@@ -567,6 +562,9 @@ xfs_attr_remove_int(xfs_inode_t *dp, struct xfs_name *name, int flags)
567 xfs_trans_set_sync(args.trans); 562 xfs_trans_set_sync(args.trans);
568 } 563 }
569 564
565 if ((flags & ATTR_KERNOTIME) == 0)
566 xfs_trans_ichgtime(args.trans, dp, XFS_ICHGTIME_CHG);
567
570 /* 568 /*
571 * Commit the last in the sequence of transactions. 569 * Commit the last in the sequence of transactions.
572 */ 570 */
@@ -574,13 +572,6 @@ xfs_attr_remove_int(xfs_inode_t *dp, struct xfs_name *name, int flags)
574 error = xfs_trans_commit(args.trans, XFS_TRANS_RELEASE_LOG_RES); 572 error = xfs_trans_commit(args.trans, XFS_TRANS_RELEASE_LOG_RES);
575 xfs_iunlock(dp, XFS_ILOCK_EXCL); 573 xfs_iunlock(dp, XFS_ILOCK_EXCL);
576 574
577 /*
578 * Hit the inode change time.
579 */
580 if (!error && (flags & ATTR_KERNOTIME) == 0) {
581 xfs_ichgtime(dp, XFS_ICHGTIME_CHG);
582 }
583
584 return(error); 575 return(error);
585 576
586out: 577out:
@@ -1995,7 +1986,7 @@ xfs_attr_rmtval_get(xfs_da_args_t *args)
1995 1986
1996 tmp = (valuelen < XFS_BUF_SIZE(bp)) 1987 tmp = (valuelen < XFS_BUF_SIZE(bp))
1997 ? valuelen : XFS_BUF_SIZE(bp); 1988 ? valuelen : XFS_BUF_SIZE(bp);
1998 xfs_biomove(bp, 0, tmp, dst, XBF_READ); 1989 xfs_buf_iomove(bp, 0, tmp, dst, XBRW_READ);
1999 xfs_buf_relse(bp); 1990 xfs_buf_relse(bp);
2000 dst += tmp; 1991 dst += tmp;
2001 valuelen -= tmp; 1992 valuelen -= tmp;
@@ -2125,9 +2116,9 @@ xfs_attr_rmtval_set(xfs_da_args_t *args)
2125 2116
2126 tmp = (valuelen < XFS_BUF_SIZE(bp)) ? valuelen : 2117 tmp = (valuelen < XFS_BUF_SIZE(bp)) ? valuelen :
2127 XFS_BUF_SIZE(bp); 2118 XFS_BUF_SIZE(bp);
2128 xfs_biomove(bp, 0, tmp, src, XBF_WRITE); 2119 xfs_buf_iomove(bp, 0, tmp, src, XBRW_WRITE);
2129 if (tmp < XFS_BUF_SIZE(bp)) 2120 if (tmp < XFS_BUF_SIZE(bp))
2130 xfs_biozero(bp, tmp, XFS_BUF_SIZE(bp) - tmp); 2121 xfs_buf_zero(bp, tmp, XFS_BUF_SIZE(bp) - tmp);
2131 if ((error = xfs_bwrite(mp, bp))) {/* GROT: NOTE: synchronous write */ 2122 if ((error = xfs_bwrite(mp, bp))) {/* GROT: NOTE: synchronous write */
2132 return (error); 2123 return (error);
2133 } 2124 }
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index f90dadd5a968..8abd12e32e13 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -614,7 +614,7 @@ xfs_bmap_add_extent(
614 nblks += cur->bc_private.b.allocated; 614 nblks += cur->bc_private.b.allocated;
615 ASSERT(nblks <= da_old); 615 ASSERT(nblks <= da_old);
616 if (nblks < da_old) 616 if (nblks < da_old)
617 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS, 617 xfs_icsb_modify_counters(ip->i_mount, XFS_SBS_FDBLOCKS,
618 (int64_t)(da_old - nblks), rsvd); 618 (int64_t)(da_old - nblks), rsvd);
619 } 619 }
620 /* 620 /*
@@ -1079,7 +1079,8 @@ xfs_bmap_add_extent_delay_real(
1079 diff = (int)(temp + temp2 - startblockval(PREV.br_startblock) - 1079 diff = (int)(temp + temp2 - startblockval(PREV.br_startblock) -
1080 (cur ? cur->bc_private.b.allocated : 0)); 1080 (cur ? cur->bc_private.b.allocated : 0));
1081 if (diff > 0 && 1081 if (diff > 0 &&
1082 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd)) { 1082 xfs_icsb_modify_counters(ip->i_mount, XFS_SBS_FDBLOCKS,
1083 -((int64_t)diff), rsvd)) {
1083 /* 1084 /*
1084 * Ick gross gag me with a spoon. 1085 * Ick gross gag me with a spoon.
1085 */ 1086 */
@@ -1089,16 +1090,18 @@ xfs_bmap_add_extent_delay_real(
1089 temp--; 1090 temp--;
1090 diff--; 1091 diff--;
1091 if (!diff || 1092 if (!diff ||
1092 !xfs_mod_incore_sb(ip->i_mount, 1093 !xfs_icsb_modify_counters(ip->i_mount,
1093 XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd)) 1094 XFS_SBS_FDBLOCKS,
1095 -((int64_t)diff), rsvd))
1094 break; 1096 break;
1095 } 1097 }
1096 if (temp2) { 1098 if (temp2) {
1097 temp2--; 1099 temp2--;
1098 diff--; 1100 diff--;
1099 if (!diff || 1101 if (!diff ||
1100 !xfs_mod_incore_sb(ip->i_mount, 1102 !xfs_icsb_modify_counters(ip->i_mount,
1101 XFS_SBS_FDBLOCKS, -((int64_t)diff), rsvd)) 1103 XFS_SBS_FDBLOCKS,
1104 -((int64_t)diff), rsvd))
1102 break; 1105 break;
1103 } 1106 }
1104 } 1107 }
@@ -1766,7 +1769,7 @@ xfs_bmap_add_extent_hole_delay(
1766 } 1769 }
1767 if (oldlen != newlen) { 1770 if (oldlen != newlen) {
1768 ASSERT(oldlen > newlen); 1771 ASSERT(oldlen > newlen);
1769 xfs_mod_incore_sb(ip->i_mount, XFS_SBS_FDBLOCKS, 1772 xfs_icsb_modify_counters(ip->i_mount, XFS_SBS_FDBLOCKS,
1770 (int64_t)(oldlen - newlen), rsvd); 1773 (int64_t)(oldlen - newlen), rsvd);
1771 /* 1774 /*
1772 * Nothing to do for disk quota accounting here. 1775 * Nothing to do for disk quota accounting here.
@@ -3111,9 +3114,10 @@ xfs_bmap_del_extent(
3111 * Nothing to do for disk quota accounting here. 3114 * Nothing to do for disk quota accounting here.
3112 */ 3115 */
3113 ASSERT(da_old >= da_new); 3116 ASSERT(da_old >= da_new);
3114 if (da_old > da_new) 3117 if (da_old > da_new) {
3115 xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, (int64_t)(da_old - da_new), 3118 xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
3116 rsvd); 3119 (int64_t)(da_old - da_new), rsvd);
3120 }
3117done: 3121done:
3118 *logflagsp = flags; 3122 *logflagsp = flags;
3119 return error; 3123 return error;
@@ -4526,13 +4530,13 @@ xfs_bmapi(
4526 -((int64_t)extsz), (flags & 4530 -((int64_t)extsz), (flags &
4527 XFS_BMAPI_RSVBLOCKS)); 4531 XFS_BMAPI_RSVBLOCKS));
4528 } else { 4532 } else {
4529 error = xfs_mod_incore_sb(mp, 4533 error = xfs_icsb_modify_counters(mp,
4530 XFS_SBS_FDBLOCKS, 4534 XFS_SBS_FDBLOCKS,
4531 -((int64_t)alen), (flags & 4535 -((int64_t)alen), (flags &
4532 XFS_BMAPI_RSVBLOCKS)); 4536 XFS_BMAPI_RSVBLOCKS));
4533 } 4537 }
4534 if (!error) { 4538 if (!error) {
4535 error = xfs_mod_incore_sb(mp, 4539 error = xfs_icsb_modify_counters(mp,
4536 XFS_SBS_FDBLOCKS, 4540 XFS_SBS_FDBLOCKS,
4537 -((int64_t)indlen), (flags & 4541 -((int64_t)indlen), (flags &
4538 XFS_BMAPI_RSVBLOCKS)); 4542 XFS_BMAPI_RSVBLOCKS));
@@ -4542,7 +4546,7 @@ xfs_bmapi(
4542 (int64_t)extsz, (flags & 4546 (int64_t)extsz, (flags &
4543 XFS_BMAPI_RSVBLOCKS)); 4547 XFS_BMAPI_RSVBLOCKS));
4544 else if (error) 4548 else if (error)
4545 xfs_mod_incore_sb(mp, 4549 xfs_icsb_modify_counters(mp,
4546 XFS_SBS_FDBLOCKS, 4550 XFS_SBS_FDBLOCKS,
4547 (int64_t)alen, (flags & 4551 (int64_t)alen, (flags &
4548 XFS_BMAPI_RSVBLOCKS)); 4552 XFS_BMAPI_RSVBLOCKS));
@@ -4744,8 +4748,12 @@ xfs_bmapi(
4744 * Check if writing previously allocated but 4748 * Check if writing previously allocated but
4745 * unwritten extents. 4749 * unwritten extents.
4746 */ 4750 */
4747 if (wr && mval->br_state == XFS_EXT_UNWRITTEN && 4751 if (wr &&
4748 ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_DELAY)) == 0)) { 4752 ((mval->br_state == XFS_EXT_UNWRITTEN &&
4753 ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_DELAY)) == 0)) ||
4754 (mval->br_state == XFS_EXT_NORM &&
4755 ((flags & (XFS_BMAPI_PREALLOC|XFS_BMAPI_CONVERT)) ==
4756 (XFS_BMAPI_PREALLOC|XFS_BMAPI_CONVERT))))) {
4749 /* 4757 /*
4750 * Modify (by adding) the state flag, if writing. 4758 * Modify (by adding) the state flag, if writing.
4751 */ 4759 */
@@ -4757,7 +4765,9 @@ xfs_bmapi(
4757 *firstblock; 4765 *firstblock;
4758 cur->bc_private.b.flist = flist; 4766 cur->bc_private.b.flist = flist;
4759 } 4767 }
4760 mval->br_state = XFS_EXT_NORM; 4768 mval->br_state = (mval->br_state == XFS_EXT_UNWRITTEN)
4769 ? XFS_EXT_NORM
4770 : XFS_EXT_UNWRITTEN;
4761 error = xfs_bmap_add_extent(ip, lastx, &cur, mval, 4771 error = xfs_bmap_add_extent(ip, lastx, &cur, mval,
4762 firstblock, flist, &tmp_logflags, 4772 firstblock, flist, &tmp_logflags,
4763 whichfork, (flags & XFS_BMAPI_RSVBLOCKS)); 4773 whichfork, (flags & XFS_BMAPI_RSVBLOCKS));
@@ -5200,7 +5210,7 @@ xfs_bunmapi(
5200 ip, -((long)del.br_blockcount), 0, 5210 ip, -((long)del.br_blockcount), 0,
5201 XFS_QMOPT_RES_RTBLKS); 5211 XFS_QMOPT_RES_RTBLKS);
5202 } else { 5212 } else {
5203 xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, 5213 xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
5204 (int64_t)del.br_blockcount, rsvd); 5214 (int64_t)del.br_blockcount, rsvd);
5205 (void)xfs_trans_reserve_quota_nblks(NULL, 5215 (void)xfs_trans_reserve_quota_nblks(NULL,
5206 ip, -((long)del.br_blockcount), 0, 5216 ip, -((long)del.br_blockcount), 0,
diff --git a/fs/xfs/xfs_bmap.h b/fs/xfs/xfs_bmap.h
index b13569a6179b..71ec9b6ecdfc 100644
--- a/fs/xfs/xfs_bmap.h
+++ b/fs/xfs/xfs_bmap.h
@@ -74,9 +74,12 @@ typedef struct xfs_bmap_free
74#define XFS_BMAPI_IGSTATE 0x080 /* Ignore state - */ 74#define XFS_BMAPI_IGSTATE 0x080 /* Ignore state - */
75 /* combine contig. space */ 75 /* combine contig. space */
76#define XFS_BMAPI_CONTIG 0x100 /* must allocate only one extent */ 76#define XFS_BMAPI_CONTIG 0x100 /* must allocate only one extent */
77#define XFS_BMAPI_CONVERT 0x200 /* unwritten extent conversion - */ 77/*
78 /* need write cache flushing and no */ 78 * unwritten extent conversion - this needs write cache flushing and no additional
79 /* additional allocation alignments */ 79 * allocation alignments. When specified with XFS_BMAPI_PREALLOC it converts
80 * from written to unwritten, otherwise convert from unwritten to written.
81 */
82#define XFS_BMAPI_CONVERT 0x200
80 83
81#define XFS_BMAPI_FLAGS \ 84#define XFS_BMAPI_FLAGS \
82 { XFS_BMAPI_WRITE, "WRITE" }, \ 85 { XFS_BMAPI_WRITE, "WRITE" }, \
diff --git a/fs/xfs/xfs_btree.c b/fs/xfs/xfs_btree.c
index 829af92f0fba..04f9cca8da7e 100644
--- a/fs/xfs/xfs_btree.c
+++ b/fs/xfs/xfs_btree.c
@@ -217,7 +217,7 @@ xfs_btree_del_cursor(
217 */ 217 */
218 for (i = 0; i < cur->bc_nlevels; i++) { 218 for (i = 0; i < cur->bc_nlevels; i++) {
219 if (cur->bc_bufs[i]) 219 if (cur->bc_bufs[i])
220 xfs_btree_setbuf(cur, i, NULL); 220 xfs_trans_brelse(cur->bc_tp, cur->bc_bufs[i]);
221 else if (!error) 221 else if (!error)
222 break; 222 break;
223 } 223 }
@@ -656,7 +656,7 @@ xfs_btree_reada_bufl(
656 656
657 ASSERT(fsbno != NULLFSBLOCK); 657 ASSERT(fsbno != NULLFSBLOCK);
658 d = XFS_FSB_TO_DADDR(mp, fsbno); 658 d = XFS_FSB_TO_DADDR(mp, fsbno);
659 xfs_baread(mp->m_ddev_targp, d, mp->m_bsize * count); 659 xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
660} 660}
661 661
662/* 662/*
@@ -676,7 +676,7 @@ xfs_btree_reada_bufs(
676 ASSERT(agno != NULLAGNUMBER); 676 ASSERT(agno != NULLAGNUMBER);
677 ASSERT(agbno != NULLAGBLOCK); 677 ASSERT(agbno != NULLAGBLOCK);
678 d = XFS_AGB_TO_DADDR(mp, agno, agbno); 678 d = XFS_AGB_TO_DADDR(mp, agno, agbno);
679 xfs_baread(mp->m_ddev_targp, d, mp->m_bsize * count); 679 xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
680} 680}
681 681
682STATIC int 682STATIC int
@@ -763,22 +763,19 @@ xfs_btree_readahead(
763 * Set the buffer for level "lev" in the cursor to bp, releasing 763 * Set the buffer for level "lev" in the cursor to bp, releasing
764 * any previous buffer. 764 * any previous buffer.
765 */ 765 */
766void 766STATIC void
767xfs_btree_setbuf( 767xfs_btree_setbuf(
768 xfs_btree_cur_t *cur, /* btree cursor */ 768 xfs_btree_cur_t *cur, /* btree cursor */
769 int lev, /* level in btree */ 769 int lev, /* level in btree */
770 xfs_buf_t *bp) /* new buffer to set */ 770 xfs_buf_t *bp) /* new buffer to set */
771{ 771{
772 struct xfs_btree_block *b; /* btree block */ 772 struct xfs_btree_block *b; /* btree block */
773 xfs_buf_t *obp; /* old buffer pointer */
774 773
775 obp = cur->bc_bufs[lev]; 774 if (cur->bc_bufs[lev])
776 if (obp) 775 xfs_trans_brelse(cur->bc_tp, cur->bc_bufs[lev]);
777 xfs_trans_brelse(cur->bc_tp, obp);
778 cur->bc_bufs[lev] = bp; 776 cur->bc_bufs[lev] = bp;
779 cur->bc_ra[lev] = 0; 777 cur->bc_ra[lev] = 0;
780 if (!bp) 778
781 return;
782 b = XFS_BUF_TO_BLOCK(bp); 779 b = XFS_BUF_TO_BLOCK(bp);
783 if (cur->bc_flags & XFS_BTREE_LONG_PTRS) { 780 if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
784 if (be64_to_cpu(b->bb_u.l.bb_leftsib) == NULLDFSBNO) 781 if (be64_to_cpu(b->bb_u.l.bb_leftsib) == NULLDFSBNO)
@@ -3011,6 +3008,43 @@ out0:
3011 return 0; 3008 return 0;
3012} 3009}
3013 3010
3011/*
3012 * Kill the current root node, and replace it with it's only child node.
3013 */
3014STATIC int
3015xfs_btree_kill_root(
3016 struct xfs_btree_cur *cur,
3017 struct xfs_buf *bp,
3018 int level,
3019 union xfs_btree_ptr *newroot)
3020{
3021 int error;
3022
3023 XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
3024 XFS_BTREE_STATS_INC(cur, killroot);
3025
3026 /*
3027 * Update the root pointer, decreasing the level by 1 and then
3028 * free the old root.
3029 */
3030 cur->bc_ops->set_root(cur, newroot, -1);
3031
3032 error = cur->bc_ops->free_block(cur, bp);
3033 if (error) {
3034 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
3035 return error;
3036 }
3037
3038 XFS_BTREE_STATS_INC(cur, free);
3039
3040 cur->bc_bufs[level] = NULL;
3041 cur->bc_ra[level] = 0;
3042 cur->bc_nlevels--;
3043
3044 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
3045 return 0;
3046}
3047
3014STATIC int 3048STATIC int
3015xfs_btree_dec_cursor( 3049xfs_btree_dec_cursor(
3016 struct xfs_btree_cur *cur, 3050 struct xfs_btree_cur *cur,
@@ -3195,7 +3229,7 @@ xfs_btree_delrec(
3195 * Make it the new root of the btree. 3229 * Make it the new root of the btree.
3196 */ 3230 */
3197 pp = xfs_btree_ptr_addr(cur, 1, block); 3231 pp = xfs_btree_ptr_addr(cur, 1, block);
3198 error = cur->bc_ops->kill_root(cur, bp, level, pp); 3232 error = xfs_btree_kill_root(cur, bp, level, pp);
3199 if (error) 3233 if (error)
3200 goto error0; 3234 goto error0;
3201 } else if (level > 0) { 3235 } else if (level > 0) {
diff --git a/fs/xfs/xfs_btree.h b/fs/xfs/xfs_btree.h
index 7fa07062bdda..82fafc66bd1f 100644
--- a/fs/xfs/xfs_btree.h
+++ b/fs/xfs/xfs_btree.h
@@ -152,9 +152,7 @@ struct xfs_btree_ops {
152 152
153 /* update btree root pointer */ 153 /* update btree root pointer */
154 void (*set_root)(struct xfs_btree_cur *cur, 154 void (*set_root)(struct xfs_btree_cur *cur,
155 union xfs_btree_ptr *nptr, int level_change); 155 union xfs_btree_ptr *nptr, int level_change);
156 int (*kill_root)(struct xfs_btree_cur *cur, struct xfs_buf *bp,
157 int level, union xfs_btree_ptr *newroot);
158 156
159 /* block allocation / freeing */ 157 /* block allocation / freeing */
160 int (*alloc_block)(struct xfs_btree_cur *cur, 158 int (*alloc_block)(struct xfs_btree_cur *cur,
@@ -399,16 +397,6 @@ xfs_btree_reada_bufs(
399 xfs_agblock_t agbno, /* allocation group block number */ 397 xfs_agblock_t agbno, /* allocation group block number */
400 xfs_extlen_t count); /* count of filesystem blocks */ 398 xfs_extlen_t count); /* count of filesystem blocks */
401 399
402/*
403 * Set the buffer for level "lev" in the cursor to bp, releasing
404 * any previous buffer.
405 */
406void
407xfs_btree_setbuf(
408 xfs_btree_cur_t *cur, /* btree cursor */
409 int lev, /* level in btree */
410 struct xfs_buf *bp); /* new buffer to set */
411
412 400
413/* 401/*
414 * Common btree core entry points. 402 * Common btree core entry points.
diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
index 1b09d7a280df..2686d0d54c5b 100644
--- a/fs/xfs/xfs_buf_item.c
+++ b/fs/xfs/xfs_buf_item.c
@@ -692,8 +692,7 @@ xfs_buf_item_init(
692 * the first. If we do already have one, there is 692 * the first. If we do already have one, there is
693 * nothing to do here so return. 693 * nothing to do here so return.
694 */ 694 */
695 if (bp->b_mount != mp) 695 ASSERT(bp->b_target->bt_mount == mp);
696 bp->b_mount = mp;
697 if (XFS_BUF_FSPRIVATE(bp, void *) != NULL) { 696 if (XFS_BUF_FSPRIVATE(bp, void *) != NULL) {
698 lip = XFS_BUF_FSPRIVATE(bp, xfs_log_item_t *); 697 lip = XFS_BUF_FSPRIVATE(bp, xfs_log_item_t *);
699 if (lip->li_type == XFS_LI_BUF) { 698 if (lip->li_type == XFS_LI_BUF) {
@@ -974,7 +973,7 @@ xfs_buf_iodone_callbacks(
974 xfs_buf_do_callbacks(bp, lip); 973 xfs_buf_do_callbacks(bp, lip);
975 XFS_BUF_SET_FSPRIVATE(bp, NULL); 974 XFS_BUF_SET_FSPRIVATE(bp, NULL);
976 XFS_BUF_CLR_IODONE_FUNC(bp); 975 XFS_BUF_CLR_IODONE_FUNC(bp);
977 xfs_biodone(bp); 976 xfs_buf_ioend(bp, 0);
978 return; 977 return;
979 } 978 }
980 979
@@ -1033,7 +1032,7 @@ xfs_buf_iodone_callbacks(
1033 xfs_buf_do_callbacks(bp, lip); 1032 xfs_buf_do_callbacks(bp, lip);
1034 XFS_BUF_SET_FSPRIVATE(bp, NULL); 1033 XFS_BUF_SET_FSPRIVATE(bp, NULL);
1035 XFS_BUF_CLR_IODONE_FUNC(bp); 1034 XFS_BUF_CLR_IODONE_FUNC(bp);
1036 xfs_biodone(bp); 1035 xfs_buf_ioend(bp, 0);
1037} 1036}
1038 1037
1039/* 1038/*
diff --git a/fs/xfs/xfs_da_btree.c b/fs/xfs/xfs_da_btree.c
index 30fa0e206fba..1c00bedb3175 100644
--- a/fs/xfs/xfs_da_btree.c
+++ b/fs/xfs/xfs_da_btree.c
@@ -2042,7 +2042,7 @@ xfs_da_do_buf(
2042 mappedbno, nmapped, 0, &bp); 2042 mappedbno, nmapped, 0, &bp);
2043 break; 2043 break;
2044 case 3: 2044 case 3:
2045 xfs_baread(mp->m_ddev_targp, mappedbno, nmapped); 2045 xfs_buf_readahead(mp->m_ddev_targp, mappedbno, nmapped);
2046 error = 0; 2046 error = 0;
2047 bp = NULL; 2047 bp = NULL;
2048 break; 2048 break;
diff --git a/fs/xfs/xfs_dinode.h b/fs/xfs/xfs_dinode.h
index e5b153b2e6a3..dffba9ba0db6 100644
--- a/fs/xfs/xfs_dinode.h
+++ b/fs/xfs/xfs_dinode.h
@@ -49,8 +49,9 @@ typedef struct xfs_dinode {
49 __be32 di_uid; /* owner's user id */ 49 __be32 di_uid; /* owner's user id */
50 __be32 di_gid; /* owner's group id */ 50 __be32 di_gid; /* owner's group id */
51 __be32 di_nlink; /* number of links to file */ 51 __be32 di_nlink; /* number of links to file */
52 __be16 di_projid; /* owner's project id */ 52 __be16 di_projid_lo; /* lower part of owner's project id */
53 __u8 di_pad[8]; /* unused, zeroed space */ 53 __be16 di_projid_hi; /* higher part owner's project id */
54 __u8 di_pad[6]; /* unused, zeroed space */
54 __be16 di_flushiter; /* incremented on flush */ 55 __be16 di_flushiter; /* incremented on flush */
55 xfs_timestamp_t di_atime; /* time last accessed */ 56 xfs_timestamp_t di_atime; /* time last accessed */
56 xfs_timestamp_t di_mtime; /* time last modified */ 57 xfs_timestamp_t di_mtime; /* time last modified */
diff --git a/fs/xfs/xfs_dir2_leaf.c b/fs/xfs/xfs_dir2_leaf.c
index 504be8640e91..ae891223be90 100644
--- a/fs/xfs/xfs_dir2_leaf.c
+++ b/fs/xfs/xfs_dir2_leaf.c
@@ -961,7 +961,7 @@ xfs_dir2_leaf_getdents(
961 if (i > ra_current && 961 if (i > ra_current &&
962 map[ra_index].br_blockcount >= 962 map[ra_index].br_blockcount >=
963 mp->m_dirblkfsbs) { 963 mp->m_dirblkfsbs) {
964 xfs_baread(mp->m_ddev_targp, 964 xfs_buf_readahead(mp->m_ddev_targp,
965 XFS_FSB_TO_DADDR(mp, 965 XFS_FSB_TO_DADDR(mp,
966 map[ra_index].br_startblock + 966 map[ra_index].br_startblock +
967 ra_offset), 967 ra_offset),
diff --git a/fs/xfs/xfs_fs.h b/fs/xfs/xfs_fs.h
index 87c2e9d02288..8f6fc1a96386 100644
--- a/fs/xfs/xfs_fs.h
+++ b/fs/xfs/xfs_fs.h
@@ -293,9 +293,11 @@ typedef struct xfs_bstat {
293 __s32 bs_extsize; /* extent size */ 293 __s32 bs_extsize; /* extent size */
294 __s32 bs_extents; /* number of extents */ 294 __s32 bs_extents; /* number of extents */
295 __u32 bs_gen; /* generation count */ 295 __u32 bs_gen; /* generation count */
296 __u16 bs_projid; /* project id */ 296 __u16 bs_projid_lo; /* lower part of project id */
297#define bs_projid bs_projid_lo /* (previously just bs_projid) */
297 __u16 bs_forkoff; /* inode fork offset in bytes */ 298 __u16 bs_forkoff; /* inode fork offset in bytes */
298 unsigned char bs_pad[12]; /* pad space, unused */ 299 __u16 bs_projid_hi; /* higher part of project id */
300 unsigned char bs_pad[10]; /* pad space, unused */
299 __u32 bs_dmevmask; /* DMIG event mask */ 301 __u32 bs_dmevmask; /* DMIG event mask */
300 __u16 bs_dmstate; /* DMIG state info */ 302 __u16 bs_dmstate; /* DMIG state info */
301 __u16 bs_aextents; /* attribute number of extents */ 303 __u16 bs_aextents; /* attribute number of extents */
@@ -448,6 +450,7 @@ typedef struct xfs_handle {
448/* XFS_IOC_SETBIOSIZE ---- deprecated 46 */ 450/* XFS_IOC_SETBIOSIZE ---- deprecated 46 */
449/* XFS_IOC_GETBIOSIZE ---- deprecated 47 */ 451/* XFS_IOC_GETBIOSIZE ---- deprecated 47 */
450#define XFS_IOC_GETBMAPX _IOWR('X', 56, struct getbmap) 452#define XFS_IOC_GETBMAPX _IOWR('X', 56, struct getbmap)
453#define XFS_IOC_ZERO_RANGE _IOW ('X', 57, struct xfs_flock64)
451 454
452/* 455/*
453 * ioctl commands that replace IRIX syssgi()'s 456 * ioctl commands that replace IRIX syssgi()'s
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index 43b1d5699335..a7c116e814af 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -144,12 +144,11 @@ xfs_growfs_data_private(
144 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb))) 144 if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
145 return error; 145 return error;
146 dpct = pct - mp->m_sb.sb_imax_pct; 146 dpct = pct - mp->m_sb.sb_imax_pct;
147 error = xfs_read_buf(mp, mp->m_ddev_targp, 147 bp = xfs_buf_read_uncached(mp, mp->m_ddev_targp,
148 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1), 148 XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
149 XFS_FSS_TO_BB(mp, 1), 0, &bp); 149 BBTOB(XFS_FSS_TO_BB(mp, 1)), 0);
150 if (error) 150 if (!bp)
151 return error; 151 return EIO;
152 ASSERT(bp);
153 xfs_buf_relse(bp); 152 xfs_buf_relse(bp);
154 153
155 new = nb; /* use new as a temporary here */ 154 new = nb; /* use new as a temporary here */
@@ -597,7 +596,8 @@ out:
597 * the extra reserve blocks from the reserve..... 596 * the extra reserve blocks from the reserve.....
598 */ 597 */
599 int error; 598 int error;
600 error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0); 599 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
600 fdblks_delta, 0);
601 if (error == ENOSPC) 601 if (error == ENOSPC)
602 goto retry; 602 goto retry;
603 } 603 }
diff --git a/fs/xfs/xfs_ialloc.c b/fs/xfs/xfs_ialloc.c
index 5371d2dc360e..0626a32c3447 100644
--- a/fs/xfs/xfs_ialloc.c
+++ b/fs/xfs/xfs_ialloc.c
@@ -212,7 +212,7 @@ xfs_ialloc_inode_init(
212 * to log a whole cluster of inodes instead of all the 212 * to log a whole cluster of inodes instead of all the
213 * individual transactions causing a lot of log traffic. 213 * individual transactions causing a lot of log traffic.
214 */ 214 */
215 xfs_biozero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog); 215 xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog);
216 for (i = 0; i < ninodes; i++) { 216 for (i = 0; i < ninodes; i++) {
217 int ioffset = i << mp->m_sb.sb_inodelog; 217 int ioffset = i << mp->m_sb.sb_inodelog;
218 uint isize = sizeof(struct xfs_dinode); 218 uint isize = sizeof(struct xfs_dinode);
diff --git a/fs/xfs/xfs_ialloc_btree.c b/fs/xfs/xfs_ialloc_btree.c
index d352862cefa0..16921f55c542 100644
--- a/fs/xfs/xfs_ialloc_btree.c
+++ b/fs/xfs/xfs_ialloc_btree.c
@@ -183,38 +183,6 @@ xfs_inobt_key_diff(
183 cur->bc_rec.i.ir_startino; 183 cur->bc_rec.i.ir_startino;
184} 184}
185 185
186STATIC int
187xfs_inobt_kill_root(
188 struct xfs_btree_cur *cur,
189 struct xfs_buf *bp,
190 int level,
191 union xfs_btree_ptr *newroot)
192{
193 int error;
194
195 XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
196 XFS_BTREE_STATS_INC(cur, killroot);
197
198 /*
199 * Update the root pointer, decreasing the level by 1 and then
200 * free the old root.
201 */
202 xfs_inobt_set_root(cur, newroot, -1);
203 error = xfs_inobt_free_block(cur, bp);
204 if (error) {
205 XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
206 return error;
207 }
208
209 XFS_BTREE_STATS_INC(cur, free);
210
211 cur->bc_bufs[level] = NULL;
212 cur->bc_nlevels--;
213
214 XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
215 return 0;
216}
217
218#ifdef DEBUG 186#ifdef DEBUG
219STATIC int 187STATIC int
220xfs_inobt_keys_inorder( 188xfs_inobt_keys_inorder(
@@ -309,7 +277,6 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
309 277
310 .dup_cursor = xfs_inobt_dup_cursor, 278 .dup_cursor = xfs_inobt_dup_cursor,
311 .set_root = xfs_inobt_set_root, 279 .set_root = xfs_inobt_set_root,
312 .kill_root = xfs_inobt_kill_root,
313 .alloc_block = xfs_inobt_alloc_block, 280 .alloc_block = xfs_inobt_alloc_block,
314 .free_block = xfs_inobt_free_block, 281 .free_block = xfs_inobt_free_block,
315 .get_minrecs = xfs_inobt_get_minrecs, 282 .get_minrecs = xfs_inobt_get_minrecs,
diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c
index b1ecc6f97ade..0cdd26932d8e 100644
--- a/fs/xfs/xfs_iget.c
+++ b/fs/xfs/xfs_iget.c
@@ -365,8 +365,8 @@ xfs_iget(
365 xfs_perag_t *pag; 365 xfs_perag_t *pag;
366 xfs_agino_t agino; 366 xfs_agino_t agino;
367 367
368 /* the radix tree exists only in inode capable AGs */ 368 /* reject inode numbers outside existing AGs */
369 if (XFS_INO_TO_AGNO(mp, ino) >= mp->m_maxagi) 369 if (XFS_INO_TO_AGNO(mp, ino) >= mp->m_sb.sb_agcount)
370 return EINVAL; 370 return EINVAL;
371 371
372 /* get the perag structure and ensure that it's inode capable */ 372 /* get the perag structure and ensure that it's inode capable */
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 34798f391c49..108c7a085f94 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -660,7 +660,8 @@ xfs_dinode_from_disk(
660 to->di_uid = be32_to_cpu(from->di_uid); 660 to->di_uid = be32_to_cpu(from->di_uid);
661 to->di_gid = be32_to_cpu(from->di_gid); 661 to->di_gid = be32_to_cpu(from->di_gid);
662 to->di_nlink = be32_to_cpu(from->di_nlink); 662 to->di_nlink = be32_to_cpu(from->di_nlink);
663 to->di_projid = be16_to_cpu(from->di_projid); 663 to->di_projid_lo = be16_to_cpu(from->di_projid_lo);
664 to->di_projid_hi = be16_to_cpu(from->di_projid_hi);
664 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); 665 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
665 to->di_flushiter = be16_to_cpu(from->di_flushiter); 666 to->di_flushiter = be16_to_cpu(from->di_flushiter);
666 to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec); 667 to->di_atime.t_sec = be32_to_cpu(from->di_atime.t_sec);
@@ -695,7 +696,8 @@ xfs_dinode_to_disk(
695 to->di_uid = cpu_to_be32(from->di_uid); 696 to->di_uid = cpu_to_be32(from->di_uid);
696 to->di_gid = cpu_to_be32(from->di_gid); 697 to->di_gid = cpu_to_be32(from->di_gid);
697 to->di_nlink = cpu_to_be32(from->di_nlink); 698 to->di_nlink = cpu_to_be32(from->di_nlink);
698 to->di_projid = cpu_to_be16(from->di_projid); 699 to->di_projid_lo = cpu_to_be16(from->di_projid_lo);
700 to->di_projid_hi = cpu_to_be16(from->di_projid_hi);
699 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad)); 701 memcpy(to->di_pad, from->di_pad, sizeof(to->di_pad));
700 to->di_flushiter = cpu_to_be16(from->di_flushiter); 702 to->di_flushiter = cpu_to_be16(from->di_flushiter);
701 to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec); 703 to->di_atime.t_sec = cpu_to_be32(from->di_atime.t_sec);
@@ -874,7 +876,7 @@ xfs_iread(
874 if (ip->i_d.di_version == 1) { 876 if (ip->i_d.di_version == 1) {
875 ip->i_d.di_nlink = ip->i_d.di_onlink; 877 ip->i_d.di_nlink = ip->i_d.di_onlink;
876 ip->i_d.di_onlink = 0; 878 ip->i_d.di_onlink = 0;
877 ip->i_d.di_projid = 0; 879 xfs_set_projid(ip, 0);
878 } 880 }
879 881
880 ip->i_delayed_blks = 0; 882 ip->i_delayed_blks = 0;
@@ -982,8 +984,7 @@ xfs_ialloc(
982 mode_t mode, 984 mode_t mode,
983 xfs_nlink_t nlink, 985 xfs_nlink_t nlink,
984 xfs_dev_t rdev, 986 xfs_dev_t rdev,
985 cred_t *cr, 987 prid_t prid,
986 xfs_prid_t prid,
987 int okalloc, 988 int okalloc,
988 xfs_buf_t **ialloc_context, 989 xfs_buf_t **ialloc_context,
989 boolean_t *call_again, 990 boolean_t *call_again,
@@ -1027,7 +1028,7 @@ xfs_ialloc(
1027 ASSERT(ip->i_d.di_nlink == nlink); 1028 ASSERT(ip->i_d.di_nlink == nlink);
1028 ip->i_d.di_uid = current_fsuid(); 1029 ip->i_d.di_uid = current_fsuid();
1029 ip->i_d.di_gid = current_fsgid(); 1030 ip->i_d.di_gid = current_fsgid();
1030 ip->i_d.di_projid = prid; 1031 xfs_set_projid(ip, prid);
1031 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad)); 1032 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
1032 1033
1033 /* 1034 /*
@@ -2725,7 +2726,7 @@ cluster_corrupt_out:
2725 XFS_BUF_UNDONE(bp); 2726 XFS_BUF_UNDONE(bp);
2726 XFS_BUF_STALE(bp); 2727 XFS_BUF_STALE(bp);
2727 XFS_BUF_ERROR(bp,EIO); 2728 XFS_BUF_ERROR(bp,EIO);
2728 xfs_biodone(bp); 2729 xfs_buf_ioend(bp, 0);
2729 } else { 2730 } else {
2730 XFS_BUF_STALE(bp); 2731 XFS_BUF_STALE(bp);
2731 xfs_buf_relse(bp); 2732 xfs_buf_relse(bp);
@@ -3008,7 +3009,7 @@ xfs_iflush_int(
3008 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad)); 3009 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
3009 memset(&(dip->di_pad[0]), 0, 3010 memset(&(dip->di_pad[0]), 0,
3010 sizeof(dip->di_pad)); 3011 sizeof(dip->di_pad));
3011 ASSERT(ip->i_d.di_projid == 0); 3012 ASSERT(xfs_get_projid(ip) == 0);
3012 } 3013 }
3013 } 3014 }
3014 3015
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 0898c5417d12..fac52290de90 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -134,8 +134,9 @@ typedef struct xfs_icdinode {
134 __uint32_t di_uid; /* owner's user id */ 134 __uint32_t di_uid; /* owner's user id */
135 __uint32_t di_gid; /* owner's group id */ 135 __uint32_t di_gid; /* owner's group id */
136 __uint32_t di_nlink; /* number of links to file */ 136 __uint32_t di_nlink; /* number of links to file */
137 __uint16_t di_projid; /* owner's project id */ 137 __uint16_t di_projid_lo; /* lower part of owner's project id */
138 __uint8_t di_pad[8]; /* unused, zeroed space */ 138 __uint16_t di_projid_hi; /* higher part of owner's project id */
139 __uint8_t di_pad[6]; /* unused, zeroed space */
139 __uint16_t di_flushiter; /* incremented on flush */ 140 __uint16_t di_flushiter; /* incremented on flush */
140 xfs_ictimestamp_t di_atime; /* time last accessed */ 141 xfs_ictimestamp_t di_atime; /* time last accessed */
141 xfs_ictimestamp_t di_mtime; /* time last modified */ 142 xfs_ictimestamp_t di_mtime; /* time last modified */
@@ -212,7 +213,6 @@ typedef struct xfs_icdinode {
212#ifdef __KERNEL__ 213#ifdef __KERNEL__
213 214
214struct bhv_desc; 215struct bhv_desc;
215struct cred;
216struct xfs_buf; 216struct xfs_buf;
217struct xfs_bmap_free; 217struct xfs_bmap_free;
218struct xfs_bmbt_irec; 218struct xfs_bmbt_irec;
@@ -335,6 +335,25 @@ xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
335} 335}
336 336
337/* 337/*
338 * Project quota id helpers (previously projid was 16bit only
339 * and using two 16bit values to hold new 32bit projid was choosen
340 * to retain compatibility with "old" filesystems).
341 */
342static inline prid_t
343xfs_get_projid(struct xfs_inode *ip)
344{
345 return (prid_t)ip->i_d.di_projid_hi << 16 | ip->i_d.di_projid_lo;
346}
347
348static inline void
349xfs_set_projid(struct xfs_inode *ip,
350 prid_t projid)
351{
352 ip->i_d.di_projid_hi = (__uint16_t) (projid >> 16);
353 ip->i_d.di_projid_lo = (__uint16_t) (projid & 0xffff);
354}
355
356/*
338 * Manage the i_flush queue embedded in the inode. This completion 357 * Manage the i_flush queue embedded in the inode. This completion
339 * queue synchronizes processes attempting to flush the in-core 358 * queue synchronizes processes attempting to flush the in-core
340 * inode back to disk. 359 * inode back to disk.
@@ -456,8 +475,8 @@ void xfs_inode_free(struct xfs_inode *ip);
456 * xfs_inode.c prototypes. 475 * xfs_inode.c prototypes.
457 */ 476 */
458int xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t, 477int xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t,
459 xfs_nlink_t, xfs_dev_t, cred_t *, xfs_prid_t, 478 xfs_nlink_t, xfs_dev_t, prid_t, int,
460 int, struct xfs_buf **, boolean_t *, xfs_inode_t **); 479 struct xfs_buf **, boolean_t *, xfs_inode_t **);
461 480
462uint xfs_ip2xflags(struct xfs_inode *); 481uint xfs_ip2xflags(struct xfs_inode *);
463uint xfs_dic2xflags(struct xfs_dinode *); 482uint xfs_dic2xflags(struct xfs_dinode *);
@@ -471,7 +490,6 @@ int xfs_iunlink(struct xfs_trans *, xfs_inode_t *);
471void xfs_iext_realloc(xfs_inode_t *, int, int); 490void xfs_iext_realloc(xfs_inode_t *, int, int);
472void xfs_iunpin_wait(xfs_inode_t *); 491void xfs_iunpin_wait(xfs_inode_t *);
473int xfs_iflush(xfs_inode_t *, uint); 492int xfs_iflush(xfs_inode_t *, uint);
474void xfs_ichgtime(xfs_inode_t *, int);
475void xfs_lock_inodes(xfs_inode_t **, int, uint); 493void xfs_lock_inodes(xfs_inode_t **, int, uint);
476void xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint); 494void xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint);
477 495
diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c
index fe00777e2796..c7ac020705df 100644
--- a/fs/xfs/xfs_inode_item.c
+++ b/fs/xfs/xfs_inode_item.c
@@ -223,15 +223,6 @@ xfs_inode_item_format(
223 nvecs = 1; 223 nvecs = 1;
224 224
225 /* 225 /*
226 * Make sure the linux inode is dirty. We do this before
227 * clearing i_update_core as the VFS will call back into
228 * XFS here and set i_update_core, so we need to dirty the
229 * inode first so that the ordering of i_update_core and
230 * unlogged modifications still works as described below.
231 */
232 xfs_mark_inode_dirty_sync(ip);
233
234 /*
235 * Clear i_update_core if the timestamps (or any other 226 * Clear i_update_core if the timestamps (or any other
236 * non-transactional modification) need flushing/logging 227 * non-transactional modification) need flushing/logging
237 * and we're about to log them with the rest of the core. 228 * and we're about to log them with the rest of the core.
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index 7e3626e5925c..dc1882adaf54 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -92,7 +92,8 @@ xfs_bulkstat_one_int(
92 * further change. 92 * further change.
93 */ 93 */
94 buf->bs_nlink = dic->di_nlink; 94 buf->bs_nlink = dic->di_nlink;
95 buf->bs_projid = dic->di_projid; 95 buf->bs_projid_lo = dic->di_projid_lo;
96 buf->bs_projid_hi = dic->di_projid_hi;
96 buf->bs_ino = ino; 97 buf->bs_ino = ino;
97 buf->bs_mode = dic->di_mode; 98 buf->bs_mode = dic->di_mode;
98 buf->bs_uid = dic->di_uid; 99 buf->bs_uid = dic->di_uid;
diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index ba8e36e0b4e7..cee4ab9f8a9e 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -1118,7 +1118,8 @@ xlog_alloc_log(xfs_mount_t *mp,
1118 iclog->ic_prev = prev_iclog; 1118 iclog->ic_prev = prev_iclog;
1119 prev_iclog = iclog; 1119 prev_iclog = iclog;
1120 1120
1121 bp = xfs_buf_get_noaddr(log->l_iclog_size, mp->m_logdev_targp); 1121 bp = xfs_buf_get_uncached(mp->m_logdev_targp,
1122 log->l_iclog_size, 0);
1122 if (!bp) 1123 if (!bp)
1123 goto out_free_iclog; 1124 goto out_free_iclog;
1124 if (!XFS_BUF_CPSEMA(bp)) 1125 if (!XFS_BUF_CPSEMA(bp))
@@ -1296,7 +1297,7 @@ xlog_bdstrat(
1296 if (iclog->ic_state & XLOG_STATE_IOERROR) { 1297 if (iclog->ic_state & XLOG_STATE_IOERROR) {
1297 XFS_BUF_ERROR(bp, EIO); 1298 XFS_BUF_ERROR(bp, EIO);
1298 XFS_BUF_STALE(bp); 1299 XFS_BUF_STALE(bp);
1299 xfs_biodone(bp); 1300 xfs_buf_ioend(bp, 0);
1300 /* 1301 /*
1301 * It would seem logical to return EIO here, but we rely on 1302 * It would seem logical to return EIO here, but we rely on
1302 * the log state machine to propagate I/O errors instead of 1303 * the log state machine to propagate I/O errors instead of
diff --git a/fs/xfs/xfs_log_cil.c b/fs/xfs/xfs_log_cil.c
index 7e206fc1fa36..23d6ceb5e97b 100644
--- a/fs/xfs/xfs_log_cil.c
+++ b/fs/xfs/xfs_log_cil.c
@@ -146,102 +146,6 @@ xlog_cil_init_post_recovery(
146} 146}
147 147
148/* 148/*
149 * Insert the log item into the CIL and calculate the difference in space
150 * consumed by the item. Add the space to the checkpoint ticket and calculate
151 * if the change requires additional log metadata. If it does, take that space
152 * as well. Remove the amount of space we addded to the checkpoint ticket from
153 * the current transaction ticket so that the accounting works out correctly.
154 *
155 * If this is the first time the item is being placed into the CIL in this
156 * context, pin it so it can't be written to disk until the CIL is flushed to
157 * the iclog and the iclog written to disk.
158 */
159static void
160xlog_cil_insert(
161 struct log *log,
162 struct xlog_ticket *ticket,
163 struct xfs_log_item *item,
164 struct xfs_log_vec *lv)
165{
166 struct xfs_cil *cil = log->l_cilp;
167 struct xfs_log_vec *old = lv->lv_item->li_lv;
168 struct xfs_cil_ctx *ctx = cil->xc_ctx;
169 int len;
170 int diff_iovecs;
171 int iclog_space;
172
173 if (old) {
174 /* existing lv on log item, space used is a delta */
175 ASSERT(!list_empty(&item->li_cil));
176 ASSERT(old->lv_buf && old->lv_buf_len && old->lv_niovecs);
177
178 len = lv->lv_buf_len - old->lv_buf_len;
179 diff_iovecs = lv->lv_niovecs - old->lv_niovecs;
180 kmem_free(old->lv_buf);
181 kmem_free(old);
182 } else {
183 /* new lv, must pin the log item */
184 ASSERT(!lv->lv_item->li_lv);
185 ASSERT(list_empty(&item->li_cil));
186
187 len = lv->lv_buf_len;
188 diff_iovecs = lv->lv_niovecs;
189 IOP_PIN(lv->lv_item);
190
191 }
192 len += diff_iovecs * sizeof(xlog_op_header_t);
193
194 /* attach new log vector to log item */
195 lv->lv_item->li_lv = lv;
196
197 spin_lock(&cil->xc_cil_lock);
198 list_move_tail(&item->li_cil, &cil->xc_cil);
199 ctx->nvecs += diff_iovecs;
200
201 /*
202 * If this is the first time the item is being committed to the CIL,
203 * store the sequence number on the log item so we can tell
204 * in future commits whether this is the first checkpoint the item is
205 * being committed into.
206 */
207 if (!item->li_seq)
208 item->li_seq = ctx->sequence;
209
210 /*
211 * Now transfer enough transaction reservation to the context ticket
212 * for the checkpoint. The context ticket is special - the unit
213 * reservation has to grow as well as the current reservation as we
214 * steal from tickets so we can correctly determine the space used
215 * during the transaction commit.
216 */
217 if (ctx->ticket->t_curr_res == 0) {
218 /* first commit in checkpoint, steal the header reservation */
219 ASSERT(ticket->t_curr_res >= ctx->ticket->t_unit_res + len);
220 ctx->ticket->t_curr_res = ctx->ticket->t_unit_res;
221 ticket->t_curr_res -= ctx->ticket->t_unit_res;
222 }
223
224 /* do we need space for more log record headers? */
225 iclog_space = log->l_iclog_size - log->l_iclog_hsize;
226 if (len > 0 && (ctx->space_used / iclog_space !=
227 (ctx->space_used + len) / iclog_space)) {
228 int hdrs;
229
230 hdrs = (len + iclog_space - 1) / iclog_space;
231 /* need to take into account split region headers, too */
232 hdrs *= log->l_iclog_hsize + sizeof(struct xlog_op_header);
233 ctx->ticket->t_unit_res += hdrs;
234 ctx->ticket->t_curr_res += hdrs;
235 ticket->t_curr_res -= hdrs;
236 ASSERT(ticket->t_curr_res >= len);
237 }
238 ticket->t_curr_res -= len;
239 ctx->space_used += len;
240
241 spin_unlock(&cil->xc_cil_lock);
242}
243
244/*
245 * Format log item into a flat buffers 149 * Format log item into a flat buffers
246 * 150 *
247 * For delayed logging, we need to hold a formatted buffer containing all the 151 * For delayed logging, we need to hold a formatted buffer containing all the
@@ -286,7 +190,7 @@ xlog_cil_format_items(
286 len += lv->lv_iovecp[index].i_len; 190 len += lv->lv_iovecp[index].i_len;
287 191
288 lv->lv_buf_len = len; 192 lv->lv_buf_len = len;
289 lv->lv_buf = kmem_zalloc(lv->lv_buf_len, KM_SLEEP|KM_NOFS); 193 lv->lv_buf = kmem_alloc(lv->lv_buf_len, KM_SLEEP|KM_NOFS);
290 ptr = lv->lv_buf; 194 ptr = lv->lv_buf;
291 195
292 for (index = 0; index < lv->lv_niovecs; index++) { 196 for (index = 0; index < lv->lv_niovecs; index++) {
@@ -300,21 +204,136 @@ xlog_cil_format_items(
300 } 204 }
301} 205}
302 206
207/*
208 * Prepare the log item for insertion into the CIL. Calculate the difference in
209 * log space and vectors it will consume, and if it is a new item pin it as
210 * well.
211 */
212STATIC void
213xfs_cil_prepare_item(
214 struct log *log,
215 struct xfs_log_vec *lv,
216 int *len,
217 int *diff_iovecs)
218{
219 struct xfs_log_vec *old = lv->lv_item->li_lv;
220
221 if (old) {
222 /* existing lv on log item, space used is a delta */
223 ASSERT(!list_empty(&lv->lv_item->li_cil));
224 ASSERT(old->lv_buf && old->lv_buf_len && old->lv_niovecs);
225
226 *len += lv->lv_buf_len - old->lv_buf_len;
227 *diff_iovecs += lv->lv_niovecs - old->lv_niovecs;
228 kmem_free(old->lv_buf);
229 kmem_free(old);
230 } else {
231 /* new lv, must pin the log item */
232 ASSERT(!lv->lv_item->li_lv);
233 ASSERT(list_empty(&lv->lv_item->li_cil));
234
235 *len += lv->lv_buf_len;
236 *diff_iovecs += lv->lv_niovecs;
237 IOP_PIN(lv->lv_item);
238
239 }
240
241 /* attach new log vector to log item */
242 lv->lv_item->li_lv = lv;
243
244 /*
245 * If this is the first time the item is being committed to the
246 * CIL, store the sequence number on the log item so we can
247 * tell in future commits whether this is the first checkpoint
248 * the item is being committed into.
249 */
250 if (!lv->lv_item->li_seq)
251 lv->lv_item->li_seq = log->l_cilp->xc_ctx->sequence;
252}
253
254/*
255 * Insert the log items into the CIL and calculate the difference in space
256 * consumed by the item. Add the space to the checkpoint ticket and calculate
257 * if the change requires additional log metadata. If it does, take that space
258 * as well. Remove the amount of space we addded to the checkpoint ticket from
259 * the current transaction ticket so that the accounting works out correctly.
260 */
303static void 261static void
304xlog_cil_insert_items( 262xlog_cil_insert_items(
305 struct log *log, 263 struct log *log,
306 struct xfs_log_vec *log_vector, 264 struct xfs_log_vec *log_vector,
307 struct xlog_ticket *ticket, 265 struct xlog_ticket *ticket)
308 xfs_lsn_t *start_lsn)
309{ 266{
310 struct xfs_log_vec *lv; 267 struct xfs_cil *cil = log->l_cilp;
311 268 struct xfs_cil_ctx *ctx = cil->xc_ctx;
312 if (start_lsn) 269 struct xfs_log_vec *lv;
313 *start_lsn = log->l_cilp->xc_ctx->sequence; 270 int len = 0;
271 int diff_iovecs = 0;
272 int iclog_space;
314 273
315 ASSERT(log_vector); 274 ASSERT(log_vector);
275
276 /*
277 * Do all the accounting aggregation and switching of log vectors
278 * around in a separate loop to the insertion of items into the CIL.
279 * Then we can do a separate loop to update the CIL within a single
280 * lock/unlock pair. This reduces the number of round trips on the CIL
281 * lock from O(nr_logvectors) to O(1) and greatly reduces the overall
282 * hold time for the transaction commit.
283 *
284 * If this is the first time the item is being placed into the CIL in
285 * this context, pin it so it can't be written to disk until the CIL is
286 * flushed to the iclog and the iclog written to disk.
287 *
288 * We can do this safely because the context can't checkpoint until we
289 * are done so it doesn't matter exactly how we update the CIL.
290 */
291 for (lv = log_vector; lv; lv = lv->lv_next)
292 xfs_cil_prepare_item(log, lv, &len, &diff_iovecs);
293
294 /* account for space used by new iovec headers */
295 len += diff_iovecs * sizeof(xlog_op_header_t);
296
297 spin_lock(&cil->xc_cil_lock);
298
299 /* move the items to the tail of the CIL */
316 for (lv = log_vector; lv; lv = lv->lv_next) 300 for (lv = log_vector; lv; lv = lv->lv_next)
317 xlog_cil_insert(log, ticket, lv->lv_item, lv); 301 list_move_tail(&lv->lv_item->li_cil, &cil->xc_cil);
302
303 ctx->nvecs += diff_iovecs;
304
305 /*
306 * Now transfer enough transaction reservation to the context ticket
307 * for the checkpoint. The context ticket is special - the unit
308 * reservation has to grow as well as the current reservation as we
309 * steal from tickets so we can correctly determine the space used
310 * during the transaction commit.
311 */
312 if (ctx->ticket->t_curr_res == 0) {
313 /* first commit in checkpoint, steal the header reservation */
314 ASSERT(ticket->t_curr_res >= ctx->ticket->t_unit_res + len);
315 ctx->ticket->t_curr_res = ctx->ticket->t_unit_res;
316 ticket->t_curr_res -= ctx->ticket->t_unit_res;
317 }
318
319 /* do we need space for more log record headers? */
320 iclog_space = log->l_iclog_size - log->l_iclog_hsize;
321 if (len > 0 && (ctx->space_used / iclog_space !=
322 (ctx->space_used + len) / iclog_space)) {
323 int hdrs;
324
325 hdrs = (len + iclog_space - 1) / iclog_space;
326 /* need to take into account split region headers, too */
327 hdrs *= log->l_iclog_hsize + sizeof(struct xlog_op_header);
328 ctx->ticket->t_unit_res += hdrs;
329 ctx->ticket->t_curr_res += hdrs;
330 ticket->t_curr_res -= hdrs;
331 ASSERT(ticket->t_curr_res >= len);
332 }
333 ticket->t_curr_res -= len;
334 ctx->space_used += len;
335
336 spin_unlock(&cil->xc_cil_lock);
318} 337}
319 338
320static void 339static void
@@ -638,7 +657,10 @@ xfs_log_commit_cil(
638 657
639 /* lock out background commit */ 658 /* lock out background commit */
640 down_read(&log->l_cilp->xc_ctx_lock); 659 down_read(&log->l_cilp->xc_ctx_lock);
641 xlog_cil_insert_items(log, log_vector, tp->t_ticket, commit_lsn); 660 if (commit_lsn)
661 *commit_lsn = log->l_cilp->xc_ctx->sequence;
662
663 xlog_cil_insert_items(log, log_vector, tp->t_ticket);
642 664
643 /* check we didn't blow the reservation */ 665 /* check we didn't blow the reservation */
644 if (tp->t_ticket->t_curr_res < 0) 666 if (tp->t_ticket->t_curr_res < 0)
diff --git a/fs/xfs/xfs_log_recover.c b/fs/xfs/xfs_log_recover.c
index 6f3f5fa37acf..966d3f97458c 100644
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -107,7 +107,8 @@ xlog_get_bp(
107 nbblks += log->l_sectBBsize; 107 nbblks += log->l_sectBBsize;
108 nbblks = round_up(nbblks, log->l_sectBBsize); 108 nbblks = round_up(nbblks, log->l_sectBBsize);
109 109
110 return xfs_buf_get_noaddr(BBTOB(nbblks), log->l_mp->m_logdev_targp); 110 return xfs_buf_get_uncached(log->l_mp->m_logdev_targp,
111 BBTOB(nbblks), 0);
111} 112}
112 113
113STATIC void 114STATIC void
@@ -167,7 +168,7 @@ xlog_bread_noalign(
167 XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp); 168 XFS_BUF_SET_TARGET(bp, log->l_mp->m_logdev_targp);
168 169
169 xfsbdstrat(log->l_mp, bp); 170 xfsbdstrat(log->l_mp, bp);
170 error = xfs_iowait(bp); 171 error = xfs_buf_iowait(bp);
171 if (error) 172 if (error)
172 xfs_ioerror_alert("xlog_bread", log->l_mp, 173 xfs_ioerror_alert("xlog_bread", log->l_mp,
173 bp, XFS_BUF_ADDR(bp)); 174 bp, XFS_BUF_ADDR(bp));
@@ -321,12 +322,13 @@ xlog_recover_iodone(
321 * this during recovery. One strike! 322 * this during recovery. One strike!
322 */ 323 */
323 xfs_ioerror_alert("xlog_recover_iodone", 324 xfs_ioerror_alert("xlog_recover_iodone",
324 bp->b_mount, bp, XFS_BUF_ADDR(bp)); 325 bp->b_target->bt_mount, bp,
325 xfs_force_shutdown(bp->b_mount, SHUTDOWN_META_IO_ERROR); 326 XFS_BUF_ADDR(bp));
327 xfs_force_shutdown(bp->b_target->bt_mount,
328 SHUTDOWN_META_IO_ERROR);
326 } 329 }
327 bp->b_mount = NULL;
328 XFS_BUF_CLR_IODONE_FUNC(bp); 330 XFS_BUF_CLR_IODONE_FUNC(bp);
329 xfs_biodone(bp); 331 xfs_buf_ioend(bp, 0);
330} 332}
331 333
332/* 334/*
@@ -2275,8 +2277,7 @@ xlog_recover_do_buffer_trans(
2275 XFS_BUF_STALE(bp); 2277 XFS_BUF_STALE(bp);
2276 error = xfs_bwrite(mp, bp); 2278 error = xfs_bwrite(mp, bp);
2277 } else { 2279 } else {
2278 ASSERT(bp->b_mount == NULL || bp->b_mount == mp); 2280 ASSERT(bp->b_target->bt_mount == mp);
2279 bp->b_mount = mp;
2280 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone); 2281 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2281 xfs_bdwrite(mp, bp); 2282 xfs_bdwrite(mp, bp);
2282 } 2283 }
@@ -2540,8 +2541,7 @@ xlog_recover_do_inode_trans(
2540 } 2541 }
2541 2542
2542write_inode_buffer: 2543write_inode_buffer:
2543 ASSERT(bp->b_mount == NULL || bp->b_mount == mp); 2544 ASSERT(bp->b_target->bt_mount == mp);
2544 bp->b_mount = mp;
2545 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone); 2545 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2546 xfs_bdwrite(mp, bp); 2546 xfs_bdwrite(mp, bp);
2547error: 2547error:
@@ -2678,8 +2678,7 @@ xlog_recover_do_dquot_trans(
2678 memcpy(ddq, recddq, item->ri_buf[1].i_len); 2678 memcpy(ddq, recddq, item->ri_buf[1].i_len);
2679 2679
2680 ASSERT(dq_f->qlf_size == 2); 2680 ASSERT(dq_f->qlf_size == 2);
2681 ASSERT(bp->b_mount == NULL || bp->b_mount == mp); 2681 ASSERT(bp->b_target->bt_mount == mp);
2682 bp->b_mount = mp;
2683 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone); 2682 XFS_BUF_SET_IODONE_FUNC(bp, xlog_recover_iodone);
2684 xfs_bdwrite(mp, bp); 2683 xfs_bdwrite(mp, bp);
2685 2684
@@ -3817,7 +3816,7 @@ xlog_do_recover(
3817 XFS_BUF_READ(bp); 3816 XFS_BUF_READ(bp);
3818 XFS_BUF_UNASYNC(bp); 3817 XFS_BUF_UNASYNC(bp);
3819 xfsbdstrat(log->l_mp, bp); 3818 xfsbdstrat(log->l_mp, bp);
3820 error = xfs_iowait(bp); 3819 error = xfs_buf_iowait(bp);
3821 if (error) { 3820 if (error) {
3822 xfs_ioerror_alert("xlog_do_recover", 3821 xfs_ioerror_alert("xlog_do_recover",
3823 log->l_mp, bp, XFS_BUF_ADDR(bp)); 3822 log->l_mp, bp, XFS_BUF_ADDR(bp));
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index aeb9d72ebf6e..b1498ab5a399 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -52,16 +52,11 @@ STATIC void xfs_icsb_balance_counter(xfs_mount_t *, xfs_sb_field_t,
52 int); 52 int);
53STATIC void xfs_icsb_balance_counter_locked(xfs_mount_t *, xfs_sb_field_t, 53STATIC void xfs_icsb_balance_counter_locked(xfs_mount_t *, xfs_sb_field_t,
54 int); 54 int);
55STATIC int xfs_icsb_modify_counters(xfs_mount_t *, xfs_sb_field_t,
56 int64_t, int);
57STATIC void xfs_icsb_disable_counter(xfs_mount_t *, xfs_sb_field_t); 55STATIC void xfs_icsb_disable_counter(xfs_mount_t *, xfs_sb_field_t);
58
59#else 56#else
60 57
61#define xfs_icsb_balance_counter(mp, a, b) do { } while (0) 58#define xfs_icsb_balance_counter(mp, a, b) do { } while (0)
62#define xfs_icsb_balance_counter_locked(mp, a, b) do { } while (0) 59#define xfs_icsb_balance_counter_locked(mp, a, b) do { } while (0)
63#define xfs_icsb_modify_counters(mp, a, b, c) do { } while (0)
64
65#endif 60#endif
66 61
67static const struct { 62static const struct {
@@ -199,6 +194,8 @@ xfs_uuid_unmount(
199 194
200/* 195/*
201 * Reference counting access wrappers to the perag structures. 196 * Reference counting access wrappers to the perag structures.
197 * Because we never free per-ag structures, the only thing we
198 * have to protect against changes is the tree structure itself.
202 */ 199 */
203struct xfs_perag * 200struct xfs_perag *
204xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno) 201xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno)
@@ -206,19 +203,43 @@ xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno)
206 struct xfs_perag *pag; 203 struct xfs_perag *pag;
207 int ref = 0; 204 int ref = 0;
208 205
209 spin_lock(&mp->m_perag_lock); 206 rcu_read_lock();
210 pag = radix_tree_lookup(&mp->m_perag_tree, agno); 207 pag = radix_tree_lookup(&mp->m_perag_tree, agno);
211 if (pag) { 208 if (pag) {
212 ASSERT(atomic_read(&pag->pag_ref) >= 0); 209 ASSERT(atomic_read(&pag->pag_ref) >= 0);
213 /* catch leaks in the positive direction during testing */
214 ASSERT(atomic_read(&pag->pag_ref) < 1000);
215 ref = atomic_inc_return(&pag->pag_ref); 210 ref = atomic_inc_return(&pag->pag_ref);
216 } 211 }
217 spin_unlock(&mp->m_perag_lock); 212 rcu_read_unlock();
218 trace_xfs_perag_get(mp, agno, ref, _RET_IP_); 213 trace_xfs_perag_get(mp, agno, ref, _RET_IP_);
219 return pag; 214 return pag;
220} 215}
221 216
217/*
218 * search from @first to find the next perag with the given tag set.
219 */
220struct xfs_perag *
221xfs_perag_get_tag(
222 struct xfs_mount *mp,
223 xfs_agnumber_t first,
224 int tag)
225{
226 struct xfs_perag *pag;
227 int found;
228 int ref;
229
230 rcu_read_lock();
231 found = radix_tree_gang_lookup_tag(&mp->m_perag_tree,
232 (void **)&pag, first, 1, tag);
233 if (found <= 0) {
234 rcu_read_unlock();
235 return NULL;
236 }
237 ref = atomic_inc_return(&pag->pag_ref);
238 rcu_read_unlock();
239 trace_xfs_perag_get_tag(mp, pag->pag_agno, ref, _RET_IP_);
240 return pag;
241}
242
222void 243void
223xfs_perag_put(struct xfs_perag *pag) 244xfs_perag_put(struct xfs_perag *pag)
224{ 245{
@@ -229,10 +250,18 @@ xfs_perag_put(struct xfs_perag *pag)
229 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_); 250 trace_xfs_perag_put(pag->pag_mount, pag->pag_agno, ref, _RET_IP_);
230} 251}
231 252
253STATIC void
254__xfs_free_perag(
255 struct rcu_head *head)
256{
257 struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head);
258
259 ASSERT(atomic_read(&pag->pag_ref) == 0);
260 kmem_free(pag);
261}
262
232/* 263/*
233 * Free up the resources associated with a mount structure. Assume that 264 * Free up the per-ag resources associated with the mount structure.
234 * the structure was initially zeroed, so we can tell which fields got
235 * initialized.
236 */ 265 */
237STATIC void 266STATIC void
238xfs_free_perag( 267xfs_free_perag(
@@ -244,10 +273,9 @@ xfs_free_perag(
244 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) { 273 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
245 spin_lock(&mp->m_perag_lock); 274 spin_lock(&mp->m_perag_lock);
246 pag = radix_tree_delete(&mp->m_perag_tree, agno); 275 pag = radix_tree_delete(&mp->m_perag_tree, agno);
247 ASSERT(pag);
248 ASSERT(atomic_read(&pag->pag_ref) == 0);
249 spin_unlock(&mp->m_perag_lock); 276 spin_unlock(&mp->m_perag_lock);
250 kmem_free(pag); 277 ASSERT(pag);
278 call_rcu(&pag->rcu_head, __xfs_free_perag);
251 } 279 }
252} 280}
253 281
@@ -444,7 +472,10 @@ xfs_initialize_perag(
444 pag->pag_agno = index; 472 pag->pag_agno = index;
445 pag->pag_mount = mp; 473 pag->pag_mount = mp;
446 rwlock_init(&pag->pag_ici_lock); 474 rwlock_init(&pag->pag_ici_lock);
475 mutex_init(&pag->pag_ici_reclaim_lock);
447 INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC); 476 INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
477 spin_lock_init(&pag->pag_buf_lock);
478 pag->pag_buf_tree = RB_ROOT;
448 479
449 if (radix_tree_preload(GFP_NOFS)) 480 if (radix_tree_preload(GFP_NOFS))
450 goto out_unwind; 481 goto out_unwind;
@@ -639,7 +670,6 @@ int
639xfs_readsb(xfs_mount_t *mp, int flags) 670xfs_readsb(xfs_mount_t *mp, int flags)
640{ 671{
641 unsigned int sector_size; 672 unsigned int sector_size;
642 unsigned int extra_flags;
643 xfs_buf_t *bp; 673 xfs_buf_t *bp;
644 int error; 674 int error;
645 675
@@ -652,28 +682,24 @@ xfs_readsb(xfs_mount_t *mp, int flags)
652 * access to the superblock. 682 * access to the superblock.
653 */ 683 */
654 sector_size = xfs_getsize_buftarg(mp->m_ddev_targp); 684 sector_size = xfs_getsize_buftarg(mp->m_ddev_targp);
655 extra_flags = XBF_LOCK | XBF_FS_MANAGED | XBF_MAPPED;
656 685
657 bp = xfs_buf_read(mp->m_ddev_targp, XFS_SB_DADDR, BTOBB(sector_size), 686reread:
658 extra_flags); 687 bp = xfs_buf_read_uncached(mp, mp->m_ddev_targp,
659 if (!bp || XFS_BUF_ISERROR(bp)) { 688 XFS_SB_DADDR, sector_size, 0);
660 xfs_fs_mount_cmn_err(flags, "SB read failed"); 689 if (!bp) {
661 error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM; 690 xfs_fs_mount_cmn_err(flags, "SB buffer read failed");
662 goto fail; 691 return EIO;
663 } 692 }
664 ASSERT(XFS_BUF_ISBUSY(bp));
665 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
666 693
667 /* 694 /*
668 * Initialize the mount structure from the superblock. 695 * Initialize the mount structure from the superblock.
669 * But first do some basic consistency checking. 696 * But first do some basic consistency checking.
670 */ 697 */
671 xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp)); 698 xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp));
672
673 error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags); 699 error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags);
674 if (error) { 700 if (error) {
675 xfs_fs_mount_cmn_err(flags, "SB validate failed"); 701 xfs_fs_mount_cmn_err(flags, "SB validate failed");
676 goto fail; 702 goto release_buf;
677 } 703 }
678 704
679 /* 705 /*
@@ -684,7 +710,7 @@ xfs_readsb(xfs_mount_t *mp, int flags)
684 "device supports only %u byte sectors (not %u)", 710 "device supports only %u byte sectors (not %u)",
685 sector_size, mp->m_sb.sb_sectsize); 711 sector_size, mp->m_sb.sb_sectsize);
686 error = ENOSYS; 712 error = ENOSYS;
687 goto fail; 713 goto release_buf;
688 } 714 }
689 715
690 /* 716 /*
@@ -692,33 +718,20 @@ xfs_readsb(xfs_mount_t *mp, int flags)
692 * re-read the superblock so the buffer is correctly sized. 718 * re-read the superblock so the buffer is correctly sized.
693 */ 719 */
694 if (sector_size < mp->m_sb.sb_sectsize) { 720 if (sector_size < mp->m_sb.sb_sectsize) {
695 XFS_BUF_UNMANAGE(bp);
696 xfs_buf_relse(bp); 721 xfs_buf_relse(bp);
697 sector_size = mp->m_sb.sb_sectsize; 722 sector_size = mp->m_sb.sb_sectsize;
698 bp = xfs_buf_read(mp->m_ddev_targp, XFS_SB_DADDR, 723 goto reread;
699 BTOBB(sector_size), extra_flags);
700 if (!bp || XFS_BUF_ISERROR(bp)) {
701 xfs_fs_mount_cmn_err(flags, "SB re-read failed");
702 error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
703 goto fail;
704 }
705 ASSERT(XFS_BUF_ISBUSY(bp));
706 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
707 } 724 }
708 725
709 /* Initialize per-cpu counters */ 726 /* Initialize per-cpu counters */
710 xfs_icsb_reinit_counters(mp); 727 xfs_icsb_reinit_counters(mp);
711 728
712 mp->m_sb_bp = bp; 729 mp->m_sb_bp = bp;
713 xfs_buf_relse(bp); 730 xfs_buf_unlock(bp);
714 ASSERT(XFS_BUF_VALUSEMA(bp) > 0);
715 return 0; 731 return 0;
716 732
717 fail: 733release_buf:
718 if (bp) { 734 xfs_buf_relse(bp);
719 XFS_BUF_UNMANAGE(bp);
720 xfs_buf_relse(bp);
721 }
722 return error; 735 return error;
723} 736}
724 737
@@ -991,42 +1004,35 @@ xfs_check_sizes(xfs_mount_t *mp)
991{ 1004{
992 xfs_buf_t *bp; 1005 xfs_buf_t *bp;
993 xfs_daddr_t d; 1006 xfs_daddr_t d;
994 int error;
995 1007
996 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks); 1008 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
997 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) { 1009 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) {
998 cmn_err(CE_WARN, "XFS: size check 1 failed"); 1010 cmn_err(CE_WARN, "XFS: filesystem size mismatch detected");
999 return XFS_ERROR(EFBIG); 1011 return XFS_ERROR(EFBIG);
1000 } 1012 }
1001 error = xfs_read_buf(mp, mp->m_ddev_targp, 1013 bp = xfs_buf_read_uncached(mp, mp->m_ddev_targp,
1002 d - XFS_FSS_TO_BB(mp, 1), 1014 d - XFS_FSS_TO_BB(mp, 1),
1003 XFS_FSS_TO_BB(mp, 1), 0, &bp); 1015 BBTOB(XFS_FSS_TO_BB(mp, 1)), 0);
1004 if (!error) { 1016 if (!bp) {
1005 xfs_buf_relse(bp); 1017 cmn_err(CE_WARN, "XFS: last sector read failed");
1006 } else { 1018 return EIO;
1007 cmn_err(CE_WARN, "XFS: size check 2 failed");
1008 if (error == ENOSPC)
1009 error = XFS_ERROR(EFBIG);
1010 return error;
1011 } 1019 }
1020 xfs_buf_relse(bp);
1012 1021
1013 if (mp->m_logdev_targp != mp->m_ddev_targp) { 1022 if (mp->m_logdev_targp != mp->m_ddev_targp) {
1014 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks); 1023 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks);
1015 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) { 1024 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) {
1016 cmn_err(CE_WARN, "XFS: size check 3 failed"); 1025 cmn_err(CE_WARN, "XFS: log size mismatch detected");
1017 return XFS_ERROR(EFBIG); 1026 return XFS_ERROR(EFBIG);
1018 } 1027 }
1019 error = xfs_read_buf(mp, mp->m_logdev_targp, 1028 bp = xfs_buf_read_uncached(mp, mp->m_logdev_targp,
1020 d - XFS_FSB_TO_BB(mp, 1), 1029 d - XFS_FSB_TO_BB(mp, 1),
1021 XFS_FSB_TO_BB(mp, 1), 0, &bp); 1030 XFS_FSB_TO_B(mp, 1), 0);
1022 if (!error) { 1031 if (!bp) {
1023 xfs_buf_relse(bp); 1032 cmn_err(CE_WARN, "XFS: log device read failed");
1024 } else { 1033 return EIO;
1025 cmn_err(CE_WARN, "XFS: size check 3 failed");
1026 if (error == ENOSPC)
1027 error = XFS_ERROR(EFBIG);
1028 return error;
1029 } 1034 }
1035 xfs_buf_relse(bp);
1030 } 1036 }
1031 return 0; 1037 return 0;
1032} 1038}
@@ -1601,7 +1607,7 @@ xfs_unmountfs_writesb(xfs_mount_t *mp)
1601 XFS_BUF_UNASYNC(sbp); 1607 XFS_BUF_UNASYNC(sbp);
1602 ASSERT(XFS_BUF_TARGET(sbp) == mp->m_ddev_targp); 1608 ASSERT(XFS_BUF_TARGET(sbp) == mp->m_ddev_targp);
1603 xfsbdstrat(mp, sbp); 1609 xfsbdstrat(mp, sbp);
1604 error = xfs_iowait(sbp); 1610 error = xfs_buf_iowait(sbp);
1605 if (error) 1611 if (error)
1606 xfs_ioerror_alert("xfs_unmountfs_writesb", 1612 xfs_ioerror_alert("xfs_unmountfs_writesb",
1607 mp, sbp, XFS_BUF_ADDR(sbp)); 1613 mp, sbp, XFS_BUF_ADDR(sbp));
@@ -1832,135 +1838,72 @@ xfs_mod_incore_sb_unlocked(
1832 */ 1838 */
1833int 1839int
1834xfs_mod_incore_sb( 1840xfs_mod_incore_sb(
1835 xfs_mount_t *mp, 1841 struct xfs_mount *mp,
1836 xfs_sb_field_t field, 1842 xfs_sb_field_t field,
1837 int64_t delta, 1843 int64_t delta,
1838 int rsvd) 1844 int rsvd)
1839{ 1845{
1840 int status; 1846 int status;
1841 1847
1842 /* check for per-cpu counters */
1843 switch (field) {
1844#ifdef HAVE_PERCPU_SB 1848#ifdef HAVE_PERCPU_SB
1845 case XFS_SBS_ICOUNT: 1849 ASSERT(field < XFS_SBS_ICOUNT || field > XFS_SBS_FDBLOCKS);
1846 case XFS_SBS_IFREE:
1847 case XFS_SBS_FDBLOCKS:
1848 if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
1849 status = xfs_icsb_modify_counters(mp, field,
1850 delta, rsvd);
1851 break;
1852 }
1853 /* FALLTHROUGH */
1854#endif 1850#endif
1855 default: 1851 spin_lock(&mp->m_sb_lock);
1856 spin_lock(&mp->m_sb_lock); 1852 status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd);
1857 status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd); 1853 spin_unlock(&mp->m_sb_lock);
1858 spin_unlock(&mp->m_sb_lock);
1859 break;
1860 }
1861 1854
1862 return status; 1855 return status;
1863} 1856}
1864 1857
1865/* 1858/*
1866 * xfs_mod_incore_sb_batch() is used to change more than one field 1859 * Change more than one field in the in-core superblock structure at a time.
1867 * in the in-core superblock structure at a time. This modification
1868 * is protected by a lock internal to this module. The fields and
1869 * changes to those fields are specified in the array of xfs_mod_sb
1870 * structures passed in.
1871 * 1860 *
1872 * Either all of the specified deltas will be applied or none of 1861 * The fields and changes to those fields are specified in the array of
1873 * them will. If any modified field dips below 0, then all modifications 1862 * xfs_mod_sb structures passed in. Either all of the specified deltas
1874 * will be backed out and EINVAL will be returned. 1863 * will be applied or none of them will. If any modified field dips below 0,
1864 * then all modifications will be backed out and EINVAL will be returned.
1865 *
1866 * Note that this function may not be used for the superblock values that
1867 * are tracked with the in-memory per-cpu counters - a direct call to
1868 * xfs_icsb_modify_counters is required for these.
1875 */ 1869 */
1876int 1870int
1877xfs_mod_incore_sb_batch(xfs_mount_t *mp, xfs_mod_sb_t *msb, uint nmsb, int rsvd) 1871xfs_mod_incore_sb_batch(
1872 struct xfs_mount *mp,
1873 xfs_mod_sb_t *msb,
1874 uint nmsb,
1875 int rsvd)
1878{ 1876{
1879 int status=0; 1877 xfs_mod_sb_t *msbp = &msb[0];
1880 xfs_mod_sb_t *msbp; 1878 int error = 0;
1881 1879
1882 /* 1880 /*
1883 * Loop through the array of mod structures and apply each 1881 * Loop through the array of mod structures and apply each individually.
1884 * individually. If any fail, then back out all those 1882 * If any fail, then back out all those which have already been applied.
1885 * which have already been applied. Do all of this within 1883 * Do all of this within the scope of the m_sb_lock so that all of the
1886 * the scope of the m_sb_lock so that all of the changes will 1884 * changes will be atomic.
1887 * be atomic.
1888 */ 1885 */
1889 spin_lock(&mp->m_sb_lock); 1886 spin_lock(&mp->m_sb_lock);
1890 msbp = &msb[0];
1891 for (msbp = &msbp[0]; msbp < (msb + nmsb); msbp++) { 1887 for (msbp = &msbp[0]; msbp < (msb + nmsb); msbp++) {
1892 /* 1888 ASSERT(msbp->msb_field < XFS_SBS_ICOUNT ||
1893 * Apply the delta at index n. If it fails, break 1889 msbp->msb_field > XFS_SBS_FDBLOCKS);
1894 * from the loop so we'll fall into the undo loop
1895 * below.
1896 */
1897 switch (msbp->msb_field) {
1898#ifdef HAVE_PERCPU_SB
1899 case XFS_SBS_ICOUNT:
1900 case XFS_SBS_IFREE:
1901 case XFS_SBS_FDBLOCKS:
1902 if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
1903 spin_unlock(&mp->m_sb_lock);
1904 status = xfs_icsb_modify_counters(mp,
1905 msbp->msb_field,
1906 msbp->msb_delta, rsvd);
1907 spin_lock(&mp->m_sb_lock);
1908 break;
1909 }
1910 /* FALLTHROUGH */
1911#endif
1912 default:
1913 status = xfs_mod_incore_sb_unlocked(mp,
1914 msbp->msb_field,
1915 msbp->msb_delta, rsvd);
1916 break;
1917 }
1918 1890
1919 if (status != 0) { 1891 error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field,
1920 break; 1892 msbp->msb_delta, rsvd);
1921 } 1893 if (error)
1894 goto unwind;
1922 } 1895 }
1896 spin_unlock(&mp->m_sb_lock);
1897 return 0;
1923 1898
1924 /* 1899unwind:
1925 * If we didn't complete the loop above, then back out 1900 while (--msbp >= msb) {
1926 * any changes made to the superblock. If you add code 1901 error = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field,
1927 * between the loop above and here, make sure that you 1902 -msbp->msb_delta, rsvd);
1928 * preserve the value of status. Loop back until 1903 ASSERT(error == 0);
1929 * we step below the beginning of the array. Make sure
1930 * we don't touch anything back there.
1931 */
1932 if (status != 0) {
1933 msbp--;
1934 while (msbp >= msb) {
1935 switch (msbp->msb_field) {
1936#ifdef HAVE_PERCPU_SB
1937 case XFS_SBS_ICOUNT:
1938 case XFS_SBS_IFREE:
1939 case XFS_SBS_FDBLOCKS:
1940 if (!(mp->m_flags & XFS_MOUNT_NO_PERCPU_SB)) {
1941 spin_unlock(&mp->m_sb_lock);
1942 status = xfs_icsb_modify_counters(mp,
1943 msbp->msb_field,
1944 -(msbp->msb_delta),
1945 rsvd);
1946 spin_lock(&mp->m_sb_lock);
1947 break;
1948 }
1949 /* FALLTHROUGH */
1950#endif
1951 default:
1952 status = xfs_mod_incore_sb_unlocked(mp,
1953 msbp->msb_field,
1954 -(msbp->msb_delta),
1955 rsvd);
1956 break;
1957 }
1958 ASSERT(status == 0);
1959 msbp--;
1960 }
1961 } 1904 }
1962 spin_unlock(&mp->m_sb_lock); 1905 spin_unlock(&mp->m_sb_lock);
1963 return status; 1906 return error;
1964} 1907}
1965 1908
1966/* 1909/*
@@ -1998,18 +1941,13 @@ xfs_getsb(
1998 */ 1941 */
1999void 1942void
2000xfs_freesb( 1943xfs_freesb(
2001 xfs_mount_t *mp) 1944 struct xfs_mount *mp)
2002{ 1945{
2003 xfs_buf_t *bp; 1946 struct xfs_buf *bp = mp->m_sb_bp;
2004 1947
2005 /* 1948 xfs_buf_lock(bp);
2006 * Use xfs_getsb() so that the buffer will be locked
2007 * when we call xfs_buf_relse().
2008 */
2009 bp = xfs_getsb(mp, 0);
2010 XFS_BUF_UNMANAGE(bp);
2011 xfs_buf_relse(bp);
2012 mp->m_sb_bp = NULL; 1949 mp->m_sb_bp = NULL;
1950 xfs_buf_relse(bp);
2013} 1951}
2014 1952
2015/* 1953/*
@@ -2496,7 +2434,7 @@ xfs_icsb_balance_counter(
2496 spin_unlock(&mp->m_sb_lock); 2434 spin_unlock(&mp->m_sb_lock);
2497} 2435}
2498 2436
2499STATIC int 2437int
2500xfs_icsb_modify_counters( 2438xfs_icsb_modify_counters(
2501 xfs_mount_t *mp, 2439 xfs_mount_t *mp,
2502 xfs_sb_field_t field, 2440 xfs_sb_field_t field,
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 622da2179a57..5861b4980740 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -53,7 +53,6 @@ typedef struct xfs_trans_reservations {
53 53
54#include "xfs_sync.h" 54#include "xfs_sync.h"
55 55
56struct cred;
57struct log; 56struct log;
58struct xfs_mount_args; 57struct xfs_mount_args;
59struct xfs_inode; 58struct xfs_inode;
@@ -91,6 +90,8 @@ extern void xfs_icsb_reinit_counters(struct xfs_mount *);
91extern void xfs_icsb_destroy_counters(struct xfs_mount *); 90extern void xfs_icsb_destroy_counters(struct xfs_mount *);
92extern void xfs_icsb_sync_counters(struct xfs_mount *, int); 91extern void xfs_icsb_sync_counters(struct xfs_mount *, int);
93extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int); 92extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
93extern int xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t,
94 int64_t, int);
94 95
95#else 96#else
96#define xfs_icsb_init_counters(mp) (0) 97#define xfs_icsb_init_counters(mp) (0)
@@ -98,6 +99,8 @@ extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
98#define xfs_icsb_reinit_counters(mp) do { } while (0) 99#define xfs_icsb_reinit_counters(mp) do { } while (0)
99#define xfs_icsb_sync_counters(mp, flags) do { } while (0) 100#define xfs_icsb_sync_counters(mp, flags) do { } while (0)
100#define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0) 101#define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
102#define xfs_icsb_modify_counters(mp, field, delta, rsvd) \
103 xfs_mod_incore_sb(mp, field, delta, rsvd)
101#endif 104#endif
102 105
103typedef struct xfs_mount { 106typedef struct xfs_mount {
@@ -232,8 +235,6 @@ typedef struct xfs_mount {
232#define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */ 235#define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */
233#define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred 236#define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred
234 * I/O size in stat() */ 237 * I/O size in stat() */
235#define XFS_MOUNT_NO_PERCPU_SB (1ULL << 23) /* don't use per-cpu superblock
236 counters */
237#define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams 238#define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams
238 allocator */ 239 allocator */
239#define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */ 240#define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
@@ -327,6 +328,8 @@ xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
327 * perag get/put wrappers for ref counting 328 * perag get/put wrappers for ref counting
328 */ 329 */
329struct xfs_perag *xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno); 330struct xfs_perag *xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno);
331struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *mp, xfs_agnumber_t agno,
332 int tag);
330void xfs_perag_put(struct xfs_perag *pag); 333void xfs_perag_put(struct xfs_perag *pag);
331 334
332/* 335/*
diff --git a/fs/xfs/xfs_refcache.h b/fs/xfs/xfs_refcache.h
deleted file mode 100644
index 2dec79edb510..000000000000
--- a/fs/xfs/xfs_refcache.h
+++ /dev/null
@@ -1,52 +0,0 @@
1/*
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_REFCACHE_H__
19#define __XFS_REFCACHE_H__
20
21#ifdef HAVE_REFCACHE
22/*
23 * Maximum size (in inodes) for the NFS reference cache
24 */
25#define XFS_REFCACHE_SIZE_MAX 512
26
27struct xfs_inode;
28struct xfs_mount;
29
30extern void xfs_refcache_insert(struct xfs_inode *);
31extern void xfs_refcache_purge_ip(struct xfs_inode *);
32extern void xfs_refcache_purge_mp(struct xfs_mount *);
33extern void xfs_refcache_purge_some(struct xfs_mount *);
34extern void xfs_refcache_resize(int);
35extern void xfs_refcache_destroy(void);
36
37extern void xfs_refcache_iunlock(struct xfs_inode *, uint);
38
39#else
40
41#define xfs_refcache_insert(ip) do { } while (0)
42#define xfs_refcache_purge_ip(ip) do { } while (0)
43#define xfs_refcache_purge_mp(mp) do { } while (0)
44#define xfs_refcache_purge_some(mp) do { } while (0)
45#define xfs_refcache_resize(size) do { } while (0)
46#define xfs_refcache_destroy() do { } while (0)
47
48#define xfs_refcache_iunlock(ip, flags) xfs_iunlock(ip, flags)
49
50#endif
51
52#endif /* __XFS_REFCACHE_H__ */
diff --git a/fs/xfs/xfs_rename.c b/fs/xfs/xfs_rename.c
index 8fca957200df..d2af0a8381a6 100644
--- a/fs/xfs/xfs_rename.c
+++ b/fs/xfs/xfs_rename.c
@@ -183,7 +183,7 @@ xfs_rename(
183 * tree quota mechanism would be circumvented. 183 * tree quota mechanism would be circumvented.
184 */ 184 */
185 if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && 185 if (unlikely((target_dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
186 (target_dp->i_d.di_projid != src_ip->i_d.di_projid))) { 186 (xfs_get_projid(target_dp) != xfs_get_projid(src_ip)))) {
187 error = XFS_ERROR(EXDEV); 187 error = XFS_ERROR(EXDEV);
188 goto error_return; 188 goto error_return;
189 } 189 }
@@ -211,7 +211,9 @@ xfs_rename(
211 goto error_return; 211 goto error_return;
212 if (error) 212 if (error)
213 goto abort_return; 213 goto abort_return;
214 xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 214
215 xfs_trans_ichgtime(tp, target_dp,
216 XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
215 217
216 if (new_parent && src_is_directory) { 218 if (new_parent && src_is_directory) {
217 error = xfs_bumplink(tp, target_dp); 219 error = xfs_bumplink(tp, target_dp);
@@ -249,7 +251,9 @@ xfs_rename(
249 &first_block, &free_list, spaceres); 251 &first_block, &free_list, spaceres);
250 if (error) 252 if (error)
251 goto abort_return; 253 goto abort_return;
252 xfs_ichgtime(target_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 254
255 xfs_trans_ichgtime(tp, target_dp,
256 XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
253 257
254 /* 258 /*
255 * Decrement the link count on the target since the target 259 * Decrement the link count on the target since the target
@@ -292,7 +296,7 @@ xfs_rename(
292 * inode isn't really being changed, but old unix file systems did 296 * inode isn't really being changed, but old unix file systems did
293 * it and some incremental backup programs won't work without it. 297 * it and some incremental backup programs won't work without it.
294 */ 298 */
295 xfs_ichgtime(src_ip, XFS_ICHGTIME_CHG); 299 xfs_trans_ichgtime(tp, src_ip, XFS_ICHGTIME_CHG);
296 300
297 /* 301 /*
298 * Adjust the link count on src_dp. This is necessary when 302 * Adjust the link count on src_dp. This is necessary when
@@ -315,7 +319,7 @@ xfs_rename(
315 if (error) 319 if (error)
316 goto abort_return; 320 goto abort_return;
317 321
318 xfs_ichgtime(src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 322 xfs_trans_ichgtime(tp, src_dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
319 xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE); 323 xfs_trans_log_inode(tp, src_dp, XFS_ILOG_CORE);
320 if (new_parent) 324 if (new_parent)
321 xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); 325 xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
index 891260fea11e..12a191385310 100644
--- a/fs/xfs/xfs_rtalloc.c
+++ b/fs/xfs/xfs_rtalloc.c
@@ -39,6 +39,7 @@
39#include "xfs_trans_space.h" 39#include "xfs_trans_space.h"
40#include "xfs_utils.h" 40#include "xfs_utils.h"
41#include "xfs_trace.h" 41#include "xfs_trace.h"
42#include "xfs_buf.h"
42 43
43 44
44/* 45/*
@@ -1883,13 +1884,13 @@ xfs_growfs_rt(
1883 /* 1884 /*
1884 * Read in the last block of the device, make sure it exists. 1885 * Read in the last block of the device, make sure it exists.
1885 */ 1886 */
1886 error = xfs_read_buf(mp, mp->m_rtdev_targp, 1887 bp = xfs_buf_read_uncached(mp, mp->m_rtdev_targp,
1887 XFS_FSB_TO_BB(mp, nrblocks - 1), 1888 XFS_FSB_TO_BB(mp, nrblocks - 1),
1888 XFS_FSB_TO_BB(mp, 1), 0, &bp); 1889 XFS_FSB_TO_B(mp, 1), 0);
1889 if (error) 1890 if (!bp)
1890 return error; 1891 return EIO;
1891 ASSERT(bp);
1892 xfs_buf_relse(bp); 1892 xfs_buf_relse(bp);
1893
1893 /* 1894 /*
1894 * Calculate new parameters. These are the final values to be reached. 1895 * Calculate new parameters. These are the final values to be reached.
1895 */ 1896 */
@@ -2215,7 +2216,6 @@ xfs_rtmount_init(
2215{ 2216{
2216 xfs_buf_t *bp; /* buffer for last block of subvolume */ 2217 xfs_buf_t *bp; /* buffer for last block of subvolume */
2217 xfs_daddr_t d; /* address of last block of subvolume */ 2218 xfs_daddr_t d; /* address of last block of subvolume */
2218 int error; /* error return value */
2219 xfs_sb_t *sbp; /* filesystem superblock copy in mount */ 2219 xfs_sb_t *sbp; /* filesystem superblock copy in mount */
2220 2220
2221 sbp = &mp->m_sb; 2221 sbp = &mp->m_sb;
@@ -2242,15 +2242,12 @@ xfs_rtmount_init(
2242 (unsigned long long) mp->m_sb.sb_rblocks); 2242 (unsigned long long) mp->m_sb.sb_rblocks);
2243 return XFS_ERROR(EFBIG); 2243 return XFS_ERROR(EFBIG);
2244 } 2244 }
2245 error = xfs_read_buf(mp, mp->m_rtdev_targp, 2245 bp = xfs_buf_read_uncached(mp, mp->m_rtdev_targp,
2246 d - XFS_FSB_TO_BB(mp, 1), 2246 d - XFS_FSB_TO_BB(mp, 1),
2247 XFS_FSB_TO_BB(mp, 1), 0, &bp); 2247 XFS_FSB_TO_B(mp, 1), 0);
2248 if (error) { 2248 if (!bp) {
2249 cmn_err(CE_WARN, 2249 cmn_err(CE_WARN, "XFS: realtime device size check failed");
2250 "XFS: realtime mount -- xfs_read_buf failed, returned %d", error); 2250 return EIO;
2251 if (error == ENOSPC)
2252 return XFS_ERROR(EFBIG);
2253 return error;
2254 } 2251 }
2255 xfs_buf_relse(bp); 2252 xfs_buf_relse(bp);
2256 return 0; 2253 return 0;
diff --git a/fs/xfs/xfs_sb.h b/fs/xfs/xfs_sb.h
index 1b017c657494..1eb2ba586814 100644
--- a/fs/xfs/xfs_sb.h
+++ b/fs/xfs/xfs_sb.h
@@ -80,10 +80,12 @@ struct xfs_mount;
80#define XFS_SB_VERSION2_RESERVED4BIT 0x00000004 80#define XFS_SB_VERSION2_RESERVED4BIT 0x00000004
81#define XFS_SB_VERSION2_ATTR2BIT 0x00000008 /* Inline attr rework */ 81#define XFS_SB_VERSION2_ATTR2BIT 0x00000008 /* Inline attr rework */
82#define XFS_SB_VERSION2_PARENTBIT 0x00000010 /* parent pointers */ 82#define XFS_SB_VERSION2_PARENTBIT 0x00000010 /* parent pointers */
83#define XFS_SB_VERSION2_PROJID32BIT 0x00000080 /* 32 bit project id */
83 84
84#define XFS_SB_VERSION2_OKREALFBITS \ 85#define XFS_SB_VERSION2_OKREALFBITS \
85 (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \ 86 (XFS_SB_VERSION2_LAZYSBCOUNTBIT | \
86 XFS_SB_VERSION2_ATTR2BIT) 87 XFS_SB_VERSION2_ATTR2BIT | \
88 XFS_SB_VERSION2_PROJID32BIT)
87#define XFS_SB_VERSION2_OKSASHFBITS \ 89#define XFS_SB_VERSION2_OKSASHFBITS \
88 (0) 90 (0)
89#define XFS_SB_VERSION2_OKREALBITS \ 91#define XFS_SB_VERSION2_OKREALBITS \
@@ -495,6 +497,12 @@ static inline void xfs_sb_version_removeattr2(xfs_sb_t *sbp)
495 sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT; 497 sbp->sb_versionnum &= ~XFS_SB_VERSION_MOREBITSBIT;
496} 498}
497 499
500static inline int xfs_sb_version_hasprojid32bit(xfs_sb_t *sbp)
501{
502 return xfs_sb_version_hasmorebits(sbp) &&
503 (sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT);
504}
505
498/* 506/*
499 * end of superblock version macros 507 * end of superblock version macros
500 */ 508 */
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index 1c47edaea0d2..f6d956b7711e 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -696,7 +696,7 @@ xfs_trans_reserve(
696 * fail if the count would go below zero. 696 * fail if the count would go below zero.
697 */ 697 */
698 if (blocks > 0) { 698 if (blocks > 0) {
699 error = xfs_mod_incore_sb(tp->t_mountp, XFS_SBS_FDBLOCKS, 699 error = xfs_icsb_modify_counters(tp->t_mountp, XFS_SBS_FDBLOCKS,
700 -((int64_t)blocks), rsvd); 700 -((int64_t)blocks), rsvd);
701 if (error != 0) { 701 if (error != 0) {
702 current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS); 702 current_restore_flags_nested(&tp->t_pflags, PF_FSTRANS);
@@ -767,7 +767,7 @@ undo_log:
767 767
768undo_blocks: 768undo_blocks:
769 if (blocks > 0) { 769 if (blocks > 0) {
770 (void) xfs_mod_incore_sb(tp->t_mountp, XFS_SBS_FDBLOCKS, 770 xfs_icsb_modify_counters(tp->t_mountp, XFS_SBS_FDBLOCKS,
771 (int64_t)blocks, rsvd); 771 (int64_t)blocks, rsvd);
772 tp->t_blk_res = 0; 772 tp->t_blk_res = 0;
773 } 773 }
@@ -1009,7 +1009,7 @@ void
1009xfs_trans_unreserve_and_mod_sb( 1009xfs_trans_unreserve_and_mod_sb(
1010 xfs_trans_t *tp) 1010 xfs_trans_t *tp)
1011{ 1011{
1012 xfs_mod_sb_t msb[14]; /* If you add cases, add entries */ 1012 xfs_mod_sb_t msb[9]; /* If you add cases, add entries */
1013 xfs_mod_sb_t *msbp; 1013 xfs_mod_sb_t *msbp;
1014 xfs_mount_t *mp = tp->t_mountp; 1014 xfs_mount_t *mp = tp->t_mountp;
1015 /* REFERENCED */ 1015 /* REFERENCED */
@@ -1017,55 +1017,61 @@ xfs_trans_unreserve_and_mod_sb(
1017 int rsvd; 1017 int rsvd;
1018 int64_t blkdelta = 0; 1018 int64_t blkdelta = 0;
1019 int64_t rtxdelta = 0; 1019 int64_t rtxdelta = 0;
1020 int64_t idelta = 0;
1021 int64_t ifreedelta = 0;
1020 1022
1021 msbp = msb; 1023 msbp = msb;
1022 rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0; 1024 rsvd = (tp->t_flags & XFS_TRANS_RESERVE) != 0;
1023 1025
1024 /* calculate free blocks delta */ 1026 /* calculate deltas */
1025 if (tp->t_blk_res > 0) 1027 if (tp->t_blk_res > 0)
1026 blkdelta = tp->t_blk_res; 1028 blkdelta = tp->t_blk_res;
1027
1028 if ((tp->t_fdblocks_delta != 0) && 1029 if ((tp->t_fdblocks_delta != 0) &&
1029 (xfs_sb_version_haslazysbcount(&mp->m_sb) || 1030 (xfs_sb_version_haslazysbcount(&mp->m_sb) ||
1030 (tp->t_flags & XFS_TRANS_SB_DIRTY))) 1031 (tp->t_flags & XFS_TRANS_SB_DIRTY)))
1031 blkdelta += tp->t_fdblocks_delta; 1032 blkdelta += tp->t_fdblocks_delta;
1032 1033
1033 if (blkdelta != 0) {
1034 msbp->msb_field = XFS_SBS_FDBLOCKS;
1035 msbp->msb_delta = blkdelta;
1036 msbp++;
1037 }
1038
1039 /* calculate free realtime extents delta */
1040 if (tp->t_rtx_res > 0) 1034 if (tp->t_rtx_res > 0)
1041 rtxdelta = tp->t_rtx_res; 1035 rtxdelta = tp->t_rtx_res;
1042
1043 if ((tp->t_frextents_delta != 0) && 1036 if ((tp->t_frextents_delta != 0) &&
1044 (tp->t_flags & XFS_TRANS_SB_DIRTY)) 1037 (tp->t_flags & XFS_TRANS_SB_DIRTY))
1045 rtxdelta += tp->t_frextents_delta; 1038 rtxdelta += tp->t_frextents_delta;
1046 1039
1040 if (xfs_sb_version_haslazysbcount(&mp->m_sb) ||
1041 (tp->t_flags & XFS_TRANS_SB_DIRTY)) {
1042 idelta = tp->t_icount_delta;
1043 ifreedelta = tp->t_ifree_delta;
1044 }
1045
1046 /* apply the per-cpu counters */
1047 if (blkdelta) {
1048 error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
1049 blkdelta, rsvd);
1050 if (error)
1051 goto out;
1052 }
1053
1054 if (idelta) {
1055 error = xfs_icsb_modify_counters(mp, XFS_SBS_ICOUNT,
1056 idelta, rsvd);
1057 if (error)
1058 goto out_undo_fdblocks;
1059 }
1060
1061 if (ifreedelta) {
1062 error = xfs_icsb_modify_counters(mp, XFS_SBS_IFREE,
1063 ifreedelta, rsvd);
1064 if (error)
1065 goto out_undo_icount;
1066 }
1067
1068 /* apply remaining deltas */
1047 if (rtxdelta != 0) { 1069 if (rtxdelta != 0) {
1048 msbp->msb_field = XFS_SBS_FREXTENTS; 1070 msbp->msb_field = XFS_SBS_FREXTENTS;
1049 msbp->msb_delta = rtxdelta; 1071 msbp->msb_delta = rtxdelta;
1050 msbp++; 1072 msbp++;
1051 } 1073 }
1052 1074
1053 /* apply remaining deltas */
1054
1055 if (xfs_sb_version_haslazysbcount(&mp->m_sb) ||
1056 (tp->t_flags & XFS_TRANS_SB_DIRTY)) {
1057 if (tp->t_icount_delta != 0) {
1058 msbp->msb_field = XFS_SBS_ICOUNT;
1059 msbp->msb_delta = tp->t_icount_delta;
1060 msbp++;
1061 }
1062 if (tp->t_ifree_delta != 0) {
1063 msbp->msb_field = XFS_SBS_IFREE;
1064 msbp->msb_delta = tp->t_ifree_delta;
1065 msbp++;
1066 }
1067 }
1068
1069 if (tp->t_flags & XFS_TRANS_SB_DIRTY) { 1075 if (tp->t_flags & XFS_TRANS_SB_DIRTY) {
1070 if (tp->t_dblocks_delta != 0) { 1076 if (tp->t_dblocks_delta != 0) {
1071 msbp->msb_field = XFS_SBS_DBLOCKS; 1077 msbp->msb_field = XFS_SBS_DBLOCKS;
@@ -1115,8 +1121,24 @@ xfs_trans_unreserve_and_mod_sb(
1115 if (msbp > msb) { 1121 if (msbp > msb) {
1116 error = xfs_mod_incore_sb_batch(tp->t_mountp, msb, 1122 error = xfs_mod_incore_sb_batch(tp->t_mountp, msb,
1117 (uint)(msbp - msb), rsvd); 1123 (uint)(msbp - msb), rsvd);
1118 ASSERT(error == 0); 1124 if (error)
1125 goto out_undo_ifreecount;
1119 } 1126 }
1127
1128 return;
1129
1130out_undo_ifreecount:
1131 if (ifreedelta)
1132 xfs_icsb_modify_counters(mp, XFS_SBS_IFREE, -ifreedelta, rsvd);
1133out_undo_icount:
1134 if (idelta)
1135 xfs_icsb_modify_counters(mp, XFS_SBS_ICOUNT, -idelta, rsvd);
1136out_undo_fdblocks:
1137 if (blkdelta)
1138 xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS, -blkdelta, rsvd);
1139out:
1140 ASSERT(error = 0);
1141 return;
1120} 1142}
1121 1143
1122/* 1144/*
@@ -1389,15 +1411,12 @@ xfs_trans_item_committed(
1389 */ 1411 */
1390STATIC void 1412STATIC void
1391xfs_trans_committed( 1413xfs_trans_committed(
1392 struct xfs_trans *tp, 1414 void *arg,
1393 int abortflag) 1415 int abortflag)
1394{ 1416{
1417 struct xfs_trans *tp = arg;
1395 struct xfs_log_item_desc *lidp, *next; 1418 struct xfs_log_item_desc *lidp, *next;
1396 1419
1397 /* Call the transaction's completion callback if there is one. */
1398 if (tp->t_callback != NULL)
1399 tp->t_callback(tp, tp->t_callarg);
1400
1401 list_for_each_entry_safe(lidp, next, &tp->t_items, lid_trans) { 1420 list_for_each_entry_safe(lidp, next, &tp->t_items, lid_trans) {
1402 xfs_trans_item_committed(lidp->lid_item, tp->t_lsn, abortflag); 1421 xfs_trans_item_committed(lidp->lid_item, tp->t_lsn, abortflag);
1403 xfs_trans_free_item_desc(lidp); 1422 xfs_trans_free_item_desc(lidp);
@@ -1525,7 +1544,7 @@ xfs_trans_commit_iclog(
1525 * running in simulation mode (the log is explicitly turned 1544 * running in simulation mode (the log is explicitly turned
1526 * off). 1545 * off).
1527 */ 1546 */
1528 tp->t_logcb.cb_func = (void(*)(void*, int))xfs_trans_committed; 1547 tp->t_logcb.cb_func = xfs_trans_committed;
1529 tp->t_logcb.cb_arg = tp; 1548 tp->t_logcb.cb_arg = tp;
1530 1549
1531 /* 1550 /*
diff --git a/fs/xfs/xfs_trans.h b/fs/xfs/xfs_trans.h
index c13c0f97b494..246286b77a86 100644
--- a/fs/xfs/xfs_trans.h
+++ b/fs/xfs/xfs_trans.h
@@ -399,8 +399,6 @@ typedef struct xfs_trans {
399 * transaction. */ 399 * transaction. */
400 struct xfs_mount *t_mountp; /* ptr to fs mount struct */ 400 struct xfs_mount *t_mountp; /* ptr to fs mount struct */
401 struct xfs_dquot_acct *t_dqinfo; /* acctg info for dquots */ 401 struct xfs_dquot_acct *t_dqinfo; /* acctg info for dquots */
402 xfs_trans_callback_t t_callback; /* transaction callback */
403 void *t_callarg; /* callback arg */
404 unsigned int t_flags; /* misc flags */ 402 unsigned int t_flags; /* misc flags */
405 int64_t t_icount_delta; /* superblock icount change */ 403 int64_t t_icount_delta; /* superblock icount change */
406 int64_t t_ifree_delta; /* superblock ifree change */ 404 int64_t t_ifree_delta; /* superblock ifree change */
@@ -473,6 +471,7 @@ void xfs_trans_dquot_buf(xfs_trans_t *, struct xfs_buf *, uint);
473void xfs_trans_inode_alloc_buf(xfs_trans_t *, struct xfs_buf *); 471void xfs_trans_inode_alloc_buf(xfs_trans_t *, struct xfs_buf *);
474int xfs_trans_iget(struct xfs_mount *, xfs_trans_t *, 472int xfs_trans_iget(struct xfs_mount *, xfs_trans_t *,
475 xfs_ino_t , uint, uint, struct xfs_inode **); 473 xfs_ino_t , uint, uint, struct xfs_inode **);
474void xfs_trans_ichgtime(struct xfs_trans *, struct xfs_inode *, int);
476void xfs_trans_ijoin_ref(struct xfs_trans *, struct xfs_inode *, uint); 475void xfs_trans_ijoin_ref(struct xfs_trans *, struct xfs_inode *, uint);
477void xfs_trans_ijoin(struct xfs_trans *, struct xfs_inode *); 476void xfs_trans_ijoin(struct xfs_trans *, struct xfs_inode *);
478void xfs_trans_log_buf(xfs_trans_t *, struct xfs_buf *, uint, uint); 477void xfs_trans_log_buf(xfs_trans_t *, struct xfs_buf *, uint, uint);
diff --git a/fs/xfs/xfs_trans_buf.c b/fs/xfs/xfs_trans_buf.c
index 90af025e6839..c47918c302a5 100644
--- a/fs/xfs/xfs_trans_buf.c
+++ b/fs/xfs/xfs_trans_buf.c
@@ -336,7 +336,7 @@ xfs_trans_read_buf(
336 ASSERT(!XFS_BUF_ISASYNC(bp)); 336 ASSERT(!XFS_BUF_ISASYNC(bp));
337 XFS_BUF_READ(bp); 337 XFS_BUF_READ(bp);
338 xfsbdstrat(tp->t_mountp, bp); 338 xfsbdstrat(tp->t_mountp, bp);
339 error = xfs_iowait(bp); 339 error = xfs_buf_iowait(bp);
340 if (error) { 340 if (error) {
341 xfs_ioerror_alert("xfs_trans_read_buf", mp, 341 xfs_ioerror_alert("xfs_trans_read_buf", mp,
342 bp, blkno); 342 bp, blkno);
diff --git a/fs/xfs/xfs_trans_inode.c b/fs/xfs/xfs_trans_inode.c
index cdc53a1050c5..ccb34532768b 100644
--- a/fs/xfs/xfs_trans_inode.c
+++ b/fs/xfs/xfs_trans_inode.c
@@ -118,6 +118,36 @@ xfs_trans_ijoin_ref(
118} 118}
119 119
120/* 120/*
121 * Transactional inode timestamp update. Requires the inode to be locked and
122 * joined to the transaction supplied. Relies on the transaction subsystem to
123 * track dirty state and update/writeback the inode accordingly.
124 */
125void
126xfs_trans_ichgtime(
127 struct xfs_trans *tp,
128 struct xfs_inode *ip,
129 int flags)
130{
131 struct inode *inode = VFS_I(ip);
132 timespec_t tv;
133
134 ASSERT(tp);
135 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
136 ASSERT(ip->i_transp == tp);
137
138 tv = current_fs_time(inode->i_sb);
139
140 if ((flags & XFS_ICHGTIME_MOD) &&
141 !timespec_equal(&inode->i_mtime, &tv)) {
142 inode->i_mtime = tv;
143 }
144 if ((flags & XFS_ICHGTIME_CHG) &&
145 !timespec_equal(&inode->i_ctime, &tv)) {
146 inode->i_ctime = tv;
147 }
148}
149
150/*
121 * This is called to mark the fields indicated in fieldmask as needing 151 * This is called to mark the fields indicated in fieldmask as needing
122 * to be logged when the transaction is committed. The inode must 152 * to be logged when the transaction is committed. The inode must
123 * already be associated with the given transaction. 153 * already be associated with the given transaction.
diff --git a/fs/xfs/xfs_types.h b/fs/xfs/xfs_types.h
index 320775295e32..26d1867d8156 100644
--- a/fs/xfs/xfs_types.h
+++ b/fs/xfs/xfs_types.h
@@ -73,8 +73,6 @@ typedef __int32_t xfs_tid_t; /* transaction identifier */
73typedef __uint32_t xfs_dablk_t; /* dir/attr block number (in file) */ 73typedef __uint32_t xfs_dablk_t; /* dir/attr block number (in file) */
74typedef __uint32_t xfs_dahash_t; /* dir/attr hash value */ 74typedef __uint32_t xfs_dahash_t; /* dir/attr hash value */
75 75
76typedef __uint16_t xfs_prid_t; /* prid_t truncated to 16bits in XFS */
77
78typedef __uint32_t xlog_tid_t; /* transaction ID type */ 76typedef __uint32_t xlog_tid_t; /* transaction ID type */
79 77
80/* 78/*
diff --git a/fs/xfs/xfs_utils.c b/fs/xfs/xfs_utils.c
index b7d5769d2df0..8b32d1a4c5a1 100644
--- a/fs/xfs/xfs_utils.c
+++ b/fs/xfs/xfs_utils.c
@@ -56,7 +56,6 @@ xfs_dir_ialloc(
56 mode_t mode, 56 mode_t mode,
57 xfs_nlink_t nlink, 57 xfs_nlink_t nlink,
58 xfs_dev_t rdev, 58 xfs_dev_t rdev,
59 cred_t *credp,
60 prid_t prid, /* project id */ 59 prid_t prid, /* project id */
61 int okalloc, /* ok to allocate new space */ 60 int okalloc, /* ok to allocate new space */
62 xfs_inode_t **ipp, /* pointer to inode; it will be 61 xfs_inode_t **ipp, /* pointer to inode; it will be
@@ -93,7 +92,7 @@ xfs_dir_ialloc(
93 * transaction commit so that no other process can steal 92 * transaction commit so that no other process can steal
94 * the inode(s) that we've just allocated. 93 * the inode(s) that we've just allocated.
95 */ 94 */
96 code = xfs_ialloc(tp, dp, mode, nlink, rdev, credp, prid, okalloc, 95 code = xfs_ialloc(tp, dp, mode, nlink, rdev, prid, okalloc,
97 &ialloc_context, &call_again, &ip); 96 &ialloc_context, &call_again, &ip);
98 97
99 /* 98 /*
@@ -197,7 +196,7 @@ xfs_dir_ialloc(
197 * other allocations in this allocation group, 196 * other allocations in this allocation group,
198 * this call should always succeed. 197 * this call should always succeed.
199 */ 198 */
200 code = xfs_ialloc(tp, dp, mode, nlink, rdev, credp, prid, 199 code = xfs_ialloc(tp, dp, mode, nlink, rdev, prid,
201 okalloc, &ialloc_context, &call_again, &ip); 200 okalloc, &ialloc_context, &call_again, &ip);
202 201
203 /* 202 /*
@@ -235,7 +234,7 @@ xfs_droplink(
235{ 234{
236 int error; 235 int error;
237 236
238 xfs_ichgtime(ip, XFS_ICHGTIME_CHG); 237 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
239 238
240 ASSERT (ip->i_d.di_nlink > 0); 239 ASSERT (ip->i_d.di_nlink > 0);
241 ip->i_d.di_nlink--; 240 ip->i_d.di_nlink--;
@@ -299,7 +298,7 @@ xfs_bumplink(
299{ 298{
300 if (ip->i_d.di_nlink >= XFS_MAXLINK) 299 if (ip->i_d.di_nlink >= XFS_MAXLINK)
301 return XFS_ERROR(EMLINK); 300 return XFS_ERROR(EMLINK);
302 xfs_ichgtime(ip, XFS_ICHGTIME_CHG); 301 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
303 302
304 ASSERT(ip->i_d.di_nlink > 0); 303 ASSERT(ip->i_d.di_nlink > 0);
305 ip->i_d.di_nlink++; 304 ip->i_d.di_nlink++;
diff --git a/fs/xfs/xfs_utils.h b/fs/xfs/xfs_utils.h
index f55b9678264f..456fca314933 100644
--- a/fs/xfs/xfs_utils.h
+++ b/fs/xfs/xfs_utils.h
@@ -19,8 +19,7 @@
19#define __XFS_UTILS_H__ 19#define __XFS_UTILS_H__
20 20
21extern int xfs_dir_ialloc(xfs_trans_t **, xfs_inode_t *, mode_t, xfs_nlink_t, 21extern int xfs_dir_ialloc(xfs_trans_t **, xfs_inode_t *, mode_t, xfs_nlink_t,
22 xfs_dev_t, cred_t *, prid_t, int, 22 xfs_dev_t, prid_t, int, xfs_inode_t **, int *);
23 xfs_inode_t **, int *);
24extern int xfs_droplink(xfs_trans_t *, xfs_inode_t *); 23extern int xfs_droplink(xfs_trans_t *, xfs_inode_t *);
25extern int xfs_bumplink(xfs_trans_t *, xfs_inode_t *); 24extern int xfs_bumplink(xfs_trans_t *, xfs_inode_t *);
26extern void xfs_bump_ino_vers2(xfs_trans_t *, xfs_inode_t *); 25extern void xfs_bump_ino_vers2(xfs_trans_t *, xfs_inode_t *);
diff --git a/fs/xfs/xfs_vnodeops.c b/fs/xfs/xfs_vnodeops.c
index 4c7c7bfb2b2f..8e4a63c4151a 100644
--- a/fs/xfs/xfs_vnodeops.c
+++ b/fs/xfs/xfs_vnodeops.c
@@ -114,7 +114,7 @@ xfs_setattr(
114 */ 114 */
115 ASSERT(udqp == NULL); 115 ASSERT(udqp == NULL);
116 ASSERT(gdqp == NULL); 116 ASSERT(gdqp == NULL);
117 code = xfs_qm_vop_dqalloc(ip, uid, gid, ip->i_d.di_projid, 117 code = xfs_qm_vop_dqalloc(ip, uid, gid, xfs_get_projid(ip),
118 qflags, &udqp, &gdqp); 118 qflags, &udqp, &gdqp);
119 if (code) 119 if (code)
120 return code; 120 return code;
@@ -184,8 +184,11 @@ xfs_setattr(
184 ip->i_size == 0 && ip->i_d.di_nextents == 0) { 184 ip->i_size == 0 && ip->i_d.di_nextents == 0) {
185 xfs_iunlock(ip, XFS_ILOCK_EXCL); 185 xfs_iunlock(ip, XFS_ILOCK_EXCL);
186 lock_flags &= ~XFS_ILOCK_EXCL; 186 lock_flags &= ~XFS_ILOCK_EXCL;
187 if (mask & ATTR_CTIME) 187 if (mask & ATTR_CTIME) {
188 xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 188 inode->i_mtime = inode->i_ctime =
189 current_fs_time(inode->i_sb);
190 xfs_mark_inode_dirty_sync(ip);
191 }
189 code = 0; 192 code = 0;
190 goto error_return; 193 goto error_return;
191 } 194 }
@@ -1253,8 +1256,7 @@ xfs_create(
1253 struct xfs_name *name, 1256 struct xfs_name *name,
1254 mode_t mode, 1257 mode_t mode,
1255 xfs_dev_t rdev, 1258 xfs_dev_t rdev,
1256 xfs_inode_t **ipp, 1259 xfs_inode_t **ipp)
1257 cred_t *credp)
1258{ 1260{
1259 int is_dir = S_ISDIR(mode); 1261 int is_dir = S_ISDIR(mode);
1260 struct xfs_mount *mp = dp->i_mount; 1262 struct xfs_mount *mp = dp->i_mount;
@@ -1266,7 +1268,7 @@ xfs_create(
1266 boolean_t unlock_dp_on_error = B_FALSE; 1268 boolean_t unlock_dp_on_error = B_FALSE;
1267 uint cancel_flags; 1269 uint cancel_flags;
1268 int committed; 1270 int committed;
1269 xfs_prid_t prid; 1271 prid_t prid;
1270 struct xfs_dquot *udqp = NULL; 1272 struct xfs_dquot *udqp = NULL;
1271 struct xfs_dquot *gdqp = NULL; 1273 struct xfs_dquot *gdqp = NULL;
1272 uint resblks; 1274 uint resblks;
@@ -1279,9 +1281,9 @@ xfs_create(
1279 return XFS_ERROR(EIO); 1281 return XFS_ERROR(EIO);
1280 1282
1281 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) 1283 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
1282 prid = dp->i_d.di_projid; 1284 prid = xfs_get_projid(dp);
1283 else 1285 else
1284 prid = dfltprid; 1286 prid = XFS_PROJID_DEFAULT;
1285 1287
1286 /* 1288 /*
1287 * Make sure that we have allocated dquot(s) on disk. 1289 * Make sure that we have allocated dquot(s) on disk.
@@ -1360,7 +1362,7 @@ xfs_create(
1360 * entry pointing to them, but a directory also the "." entry 1362 * entry pointing to them, but a directory also the "." entry
1361 * pointing to itself. 1363 * pointing to itself.
1362 */ 1364 */
1363 error = xfs_dir_ialloc(&tp, dp, mode, is_dir ? 2 : 1, rdev, credp, 1365 error = xfs_dir_ialloc(&tp, dp, mode, is_dir ? 2 : 1, rdev,
1364 prid, resblks > 0, &ip, &committed); 1366 prid, resblks > 0, &ip, &committed);
1365 if (error) { 1367 if (error) {
1366 if (error == ENOSPC) 1368 if (error == ENOSPC)
@@ -1391,7 +1393,7 @@ xfs_create(
1391 ASSERT(error != ENOSPC); 1393 ASSERT(error != ENOSPC);
1392 goto out_trans_abort; 1394 goto out_trans_abort;
1393 } 1395 }
1394 xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 1396 xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
1395 xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE); 1397 xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
1396 1398
1397 if (is_dir) { 1399 if (is_dir) {
@@ -1742,7 +1744,7 @@ xfs_remove(
1742 ASSERT(error != ENOENT); 1744 ASSERT(error != ENOENT);
1743 goto out_bmap_cancel; 1745 goto out_bmap_cancel;
1744 } 1746 }
1745 xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 1747 xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
1746 1748
1747 if (is_dir) { 1749 if (is_dir) {
1748 /* 1750 /*
@@ -1880,7 +1882,7 @@ xfs_link(
1880 * the tree quota mechanism could be circumvented. 1882 * the tree quota mechanism could be circumvented.
1881 */ 1883 */
1882 if (unlikely((tdp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && 1884 if (unlikely((tdp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1883 (tdp->i_d.di_projid != sip->i_d.di_projid))) { 1885 (xfs_get_projid(tdp) != xfs_get_projid(sip)))) {
1884 error = XFS_ERROR(EXDEV); 1886 error = XFS_ERROR(EXDEV);
1885 goto error_return; 1887 goto error_return;
1886 } 1888 }
@@ -1895,7 +1897,7 @@ xfs_link(
1895 &first_block, &free_list, resblks); 1897 &first_block, &free_list, resblks);
1896 if (error) 1898 if (error)
1897 goto abort_return; 1899 goto abort_return;
1898 xfs_ichgtime(tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 1900 xfs_trans_ichgtime(tp, tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
1899 xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE); 1901 xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE);
1900 1902
1901 error = xfs_bumplink(tp, sip); 1903 error = xfs_bumplink(tp, sip);
@@ -1933,8 +1935,7 @@ xfs_symlink(
1933 struct xfs_name *link_name, 1935 struct xfs_name *link_name,
1934 const char *target_path, 1936 const char *target_path,
1935 mode_t mode, 1937 mode_t mode,
1936 xfs_inode_t **ipp, 1938 xfs_inode_t **ipp)
1937 cred_t *credp)
1938{ 1939{
1939 xfs_mount_t *mp = dp->i_mount; 1940 xfs_mount_t *mp = dp->i_mount;
1940 xfs_trans_t *tp; 1941 xfs_trans_t *tp;
@@ -1955,7 +1956,7 @@ xfs_symlink(
1955 int byte_cnt; 1956 int byte_cnt;
1956 int n; 1957 int n;
1957 xfs_buf_t *bp; 1958 xfs_buf_t *bp;
1958 xfs_prid_t prid; 1959 prid_t prid;
1959 struct xfs_dquot *udqp, *gdqp; 1960 struct xfs_dquot *udqp, *gdqp;
1960 uint resblks; 1961 uint resblks;
1961 1962
@@ -1978,9 +1979,9 @@ xfs_symlink(
1978 1979
1979 udqp = gdqp = NULL; 1980 udqp = gdqp = NULL;
1980 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) 1981 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
1981 prid = dp->i_d.di_projid; 1982 prid = xfs_get_projid(dp);
1982 else 1983 else
1983 prid = (xfs_prid_t)dfltprid; 1984 prid = XFS_PROJID_DEFAULT;
1984 1985
1985 /* 1986 /*
1986 * Make sure that we have allocated dquot(s) on disk. 1987 * Make sure that we have allocated dquot(s) on disk.
@@ -2046,8 +2047,8 @@ xfs_symlink(
2046 /* 2047 /*
2047 * Allocate an inode for the symlink. 2048 * Allocate an inode for the symlink.
2048 */ 2049 */
2049 error = xfs_dir_ialloc(&tp, dp, S_IFLNK | (mode & ~S_IFMT), 2050 error = xfs_dir_ialloc(&tp, dp, S_IFLNK | (mode & ~S_IFMT), 1, 0,
2050 1, 0, credp, prid, resblks > 0, &ip, NULL); 2051 prid, resblks > 0, &ip, NULL);
2051 if (error) { 2052 if (error) {
2052 if (error == ENOSPC) 2053 if (error == ENOSPC)
2053 goto error_return; 2054 goto error_return;
@@ -2129,7 +2130,7 @@ xfs_symlink(
2129 &first_block, &free_list, resblks); 2130 &first_block, &free_list, resblks);
2130 if (error) 2131 if (error)
2131 goto error1; 2132 goto error1;
2132 xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 2133 xfs_trans_ichgtime(tp, dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
2133 xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE); 2134 xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);
2134 2135
2135 /* 2136 /*
@@ -2272,7 +2273,7 @@ xfs_alloc_file_space(
2272 count = len; 2273 count = len;
2273 imapp = &imaps[0]; 2274 imapp = &imaps[0];
2274 nimaps = 1; 2275 nimaps = 1;
2275 bmapi_flag = XFS_BMAPI_WRITE | (alloc_type ? XFS_BMAPI_PREALLOC : 0); 2276 bmapi_flag = XFS_BMAPI_WRITE | alloc_type;
2276 startoffset_fsb = XFS_B_TO_FSBT(mp, offset); 2277 startoffset_fsb = XFS_B_TO_FSBT(mp, offset);
2277 allocatesize_fsb = XFS_B_TO_FSB(mp, count); 2278 allocatesize_fsb = XFS_B_TO_FSB(mp, count);
2278 2279
@@ -2431,9 +2432,9 @@ xfs_zero_remaining_bytes(
2431 if (endoff > ip->i_size) 2432 if (endoff > ip->i_size)
2432 endoff = ip->i_size; 2433 endoff = ip->i_size;
2433 2434
2434 bp = xfs_buf_get_noaddr(mp->m_sb.sb_blocksize, 2435 bp = xfs_buf_get_uncached(XFS_IS_REALTIME_INODE(ip) ?
2435 XFS_IS_REALTIME_INODE(ip) ? 2436 mp->m_rtdev_targp : mp->m_ddev_targp,
2436 mp->m_rtdev_targp : mp->m_ddev_targp); 2437 mp->m_sb.sb_blocksize, XBF_DONT_BLOCK);
2437 if (!bp) 2438 if (!bp)
2438 return XFS_ERROR(ENOMEM); 2439 return XFS_ERROR(ENOMEM);
2439 2440
@@ -2459,7 +2460,7 @@ xfs_zero_remaining_bytes(
2459 XFS_BUF_READ(bp); 2460 XFS_BUF_READ(bp);
2460 XFS_BUF_SET_ADDR(bp, xfs_fsb_to_db(ip, imap.br_startblock)); 2461 XFS_BUF_SET_ADDR(bp, xfs_fsb_to_db(ip, imap.br_startblock));
2461 xfsbdstrat(mp, bp); 2462 xfsbdstrat(mp, bp);
2462 error = xfs_iowait(bp); 2463 error = xfs_buf_iowait(bp);
2463 if (error) { 2464 if (error) {
2464 xfs_ioerror_alert("xfs_zero_remaining_bytes(read)", 2465 xfs_ioerror_alert("xfs_zero_remaining_bytes(read)",
2465 mp, bp, XFS_BUF_ADDR(bp)); 2466 mp, bp, XFS_BUF_ADDR(bp));
@@ -2472,7 +2473,7 @@ xfs_zero_remaining_bytes(
2472 XFS_BUF_UNREAD(bp); 2473 XFS_BUF_UNREAD(bp);
2473 XFS_BUF_WRITE(bp); 2474 XFS_BUF_WRITE(bp);
2474 xfsbdstrat(mp, bp); 2475 xfsbdstrat(mp, bp);
2475 error = xfs_iowait(bp); 2476 error = xfs_buf_iowait(bp);
2476 if (error) { 2477 if (error) {
2477 xfs_ioerror_alert("xfs_zero_remaining_bytes(write)", 2478 xfs_ioerror_alert("xfs_zero_remaining_bytes(write)",
2478 mp, bp, XFS_BUF_ADDR(bp)); 2479 mp, bp, XFS_BUF_ADDR(bp));
@@ -2711,6 +2712,7 @@ xfs_change_file_space(
2711 xfs_off_t llen; 2712 xfs_off_t llen;
2712 xfs_trans_t *tp; 2713 xfs_trans_t *tp;
2713 struct iattr iattr; 2714 struct iattr iattr;
2715 int prealloc_type;
2714 2716
2715 if (!S_ISREG(ip->i_d.di_mode)) 2717 if (!S_ISREG(ip->i_d.di_mode))
2716 return XFS_ERROR(EINVAL); 2718 return XFS_ERROR(EINVAL);
@@ -2753,12 +2755,17 @@ xfs_change_file_space(
2753 * size to be changed. 2755 * size to be changed.
2754 */ 2756 */
2755 setprealloc = clrprealloc = 0; 2757 setprealloc = clrprealloc = 0;
2758 prealloc_type = XFS_BMAPI_PREALLOC;
2756 2759
2757 switch (cmd) { 2760 switch (cmd) {
2761 case XFS_IOC_ZERO_RANGE:
2762 prealloc_type |= XFS_BMAPI_CONVERT;
2763 xfs_tosspages(ip, startoffset, startoffset + bf->l_len, 0);
2764 /* FALLTHRU */
2758 case XFS_IOC_RESVSP: 2765 case XFS_IOC_RESVSP:
2759 case XFS_IOC_RESVSP64: 2766 case XFS_IOC_RESVSP64:
2760 error = xfs_alloc_file_space(ip, startoffset, bf->l_len, 2767 error = xfs_alloc_file_space(ip, startoffset, bf->l_len,
2761 1, attr_flags); 2768 prealloc_type, attr_flags);
2762 if (error) 2769 if (error)
2763 return error; 2770 return error;
2764 setprealloc = 1; 2771 setprealloc = 1;
@@ -2827,7 +2834,7 @@ xfs_change_file_space(
2827 if (ip->i_d.di_mode & S_IXGRP) 2834 if (ip->i_d.di_mode & S_IXGRP)
2828 ip->i_d.di_mode &= ~S_ISGID; 2835 ip->i_d.di_mode &= ~S_ISGID;
2829 2836
2830 xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG); 2837 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
2831 } 2838 }
2832 if (setprealloc) 2839 if (setprealloc)
2833 ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC; 2840 ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
diff --git a/fs/xfs/xfs_vnodeops.h b/fs/xfs/xfs_vnodeops.h
index d8dfa8d0dadd..f6702927eee4 100644
--- a/fs/xfs/xfs_vnodeops.h
+++ b/fs/xfs/xfs_vnodeops.h
@@ -2,7 +2,6 @@
2#define _XFS_VNODEOPS_H 1 2#define _XFS_VNODEOPS_H 1
3 3
4struct attrlist_cursor_kern; 4struct attrlist_cursor_kern;
5struct cred;
6struct file; 5struct file;
7struct iattr; 6struct iattr;
8struct inode; 7struct inode;
@@ -26,7 +25,7 @@ int xfs_inactive(struct xfs_inode *ip);
26int xfs_lookup(struct xfs_inode *dp, struct xfs_name *name, 25int xfs_lookup(struct xfs_inode *dp, struct xfs_name *name,
27 struct xfs_inode **ipp, struct xfs_name *ci_name); 26 struct xfs_inode **ipp, struct xfs_name *ci_name);
28int xfs_create(struct xfs_inode *dp, struct xfs_name *name, mode_t mode, 27int xfs_create(struct xfs_inode *dp, struct xfs_name *name, mode_t mode,
29 xfs_dev_t rdev, struct xfs_inode **ipp, cred_t *credp); 28 xfs_dev_t rdev, struct xfs_inode **ipp);
30int xfs_remove(struct xfs_inode *dp, struct xfs_name *name, 29int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
31 struct xfs_inode *ip); 30 struct xfs_inode *ip);
32int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip, 31int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
@@ -34,8 +33,7 @@ int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
34int xfs_readdir(struct xfs_inode *dp, void *dirent, size_t bufsize, 33int xfs_readdir(struct xfs_inode *dp, void *dirent, size_t bufsize,
35 xfs_off_t *offset, filldir_t filldir); 34 xfs_off_t *offset, filldir_t filldir);
36int xfs_symlink(struct xfs_inode *dp, struct xfs_name *link_name, 35int xfs_symlink(struct xfs_inode *dp, struct xfs_name *link_name,
37 const char *target_path, mode_t mode, struct xfs_inode **ipp, 36 const char *target_path, mode_t mode, struct xfs_inode **ipp);
38 cred_t *credp);
39int xfs_set_dmattrs(struct xfs_inode *ip, u_int evmask, u_int16_t state); 37int xfs_set_dmattrs(struct xfs_inode *ip, u_int evmask, u_int16_t state);
40int xfs_change_file_space(struct xfs_inode *ip, int cmd, 38int xfs_change_file_space(struct xfs_inode *ip, int cmd,
41 xfs_flock64_t *bf, xfs_off_t offset, int attr_flags); 39 xfs_flock64_t *bf, xfs_off_t offset, int attr_flags);
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-18 01:34:00 -0500