diff options
author | Mingming Cao <cmm@us.ibm.com> | 2006-06-25 08:48:06 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@g5.osdl.org> | 2006-06-25 13:01:10 -0400 |
commit | 1c2bf374a4b8c2e1a3e6ff3a64fb67272a8cd2e2 (patch) | |
tree | abb4e01a4bc146183d5cf1f0e18c1aa6e847aa5e /fs/ext3/balloc.c | |
parent | cedfb178ada245b6c52a654936b34d7996e26a1d (diff) |
[PATCH] ext3_fsblk_t: filesystem, group blocks and bug fixes
Some of the in-kernel ext3 block variable type are treated as signed 4 bytes
int type, thus limited ext3 filesystem to 8TB (4kblock size based). While
trying to fix them, it seems quite confusing in the ext3 code where some
blocks are filesystem-wide blocks, some are group relative offsets that need
to be signed value (as -1 has special meaning). So it seem saner to define
two types of physical blocks: one is filesystem wide blocks, another is
group-relative blocks. The following patches clarify these two types of
blocks in the ext3 code, and fix the type bugs which limit current 32 bit ext3
filesystem limit to 8TB.
With this series of patches and the percpu counter data type changes in the mm
tree, we are able to extend exts filesystem limit to 16TB.
This work is also a pre-request for the recent >32 bit ext3 work, and makes
the kernel to able to address 48 bit ext3 block a lot easier: Simply redefine
ext3_fsblk_t from unsigned long to sector_t and redefine the format string for
ext3 filesystem block corresponding.
Two RFC with a series patches have been posted to ext2-devel list and have
been reviewed and discussed:
http://marc.theaimsgroup.com/?l=ext2-devel&m=114722190816690&w=2
http://marc.theaimsgroup.com/?l=ext2-devel&m=114784919525942&w=2
Patches are tested on both 32 bit machine and 64 bit machine, <8TB ext3 and
>8TB ext3 filesystem(with the latest to be released e2fsprogs-1.39). Tests
includes overnight fsx, tiobench, dbench and fsstress.
This patch:
Defines ext3_fsblk_t and ext3_grpblk_t, and the printk format string for
filesystem wide blocks.
This patch classifies all block group relative blocks, and ext3_fsblk_t blocks
occurs in the same function where used to be confusing before. Also include
kernel bug fixes for filesystem wide in-kernel block variables. There are
some fileystem wide blocks are treated as int/unsigned int type in the kernel
currently, especially in ext3 block allocation and reservation code. This
patch fixed those bugs by converting those variables to ext3_fsblk_t(unsigned
long) type.
Signed-off-by: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'fs/ext3/balloc.c')
-rw-r--r-- | fs/ext3/balloc.c | 215 |
1 files changed, 112 insertions, 103 deletions
diff --git a/fs/ext3/balloc.c b/fs/ext3/balloc.c index 77927d6938f..b1633cd28ec 100644 --- a/fs/ext3/balloc.c +++ b/fs/ext3/balloc.c | |||
@@ -163,10 +163,10 @@ restart: | |||
163 | #endif | 163 | #endif |
164 | 164 | ||
165 | static int | 165 | static int |
166 | goal_in_my_reservation(struct ext3_reserve_window *rsv, int goal, | 166 | goal_in_my_reservation(struct ext3_reserve_window *rsv, ext3_grpblk_t grp_goal, |
167 | unsigned int group, struct super_block * sb) | 167 | unsigned int group, struct super_block * sb) |
168 | { | 168 | { |
169 | unsigned long group_first_block, group_last_block; | 169 | ext3_fsblk_t group_first_block, group_last_block; |
170 | 170 | ||
171 | group_first_block = le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block) + | 171 | group_first_block = le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block) + |
172 | group * EXT3_BLOCKS_PER_GROUP(sb); | 172 | group * EXT3_BLOCKS_PER_GROUP(sb); |
@@ -175,8 +175,8 @@ goal_in_my_reservation(struct ext3_reserve_window *rsv, int goal, | |||
175 | if ((rsv->_rsv_start > group_last_block) || | 175 | if ((rsv->_rsv_start > group_last_block) || |
176 | (rsv->_rsv_end < group_first_block)) | 176 | (rsv->_rsv_end < group_first_block)) |
177 | return 0; | 177 | return 0; |
178 | if ((goal >= 0) && ((goal + group_first_block < rsv->_rsv_start) | 178 | if ((grp_goal >= 0) && ((grp_goal + group_first_block < rsv->_rsv_start) |
179 | || (goal + group_first_block > rsv->_rsv_end))) | 179 | || (grp_goal + group_first_block > rsv->_rsv_end))) |
180 | return 0; | 180 | return 0; |
181 | return 1; | 181 | return 1; |
182 | } | 182 | } |
@@ -187,7 +187,7 @@ goal_in_my_reservation(struct ext3_reserve_window *rsv, int goal, | |||
187 | * Returns NULL if there are no windows or if all windows start after the goal. | 187 | * Returns NULL if there are no windows or if all windows start after the goal. |
188 | */ | 188 | */ |
189 | static struct ext3_reserve_window_node * | 189 | static struct ext3_reserve_window_node * |
190 | search_reserve_window(struct rb_root *root, unsigned long goal) | 190 | search_reserve_window(struct rb_root *root, ext3_fsblk_t goal) |
191 | { | 191 | { |
192 | struct rb_node *n = root->rb_node; | 192 | struct rb_node *n = root->rb_node; |
193 | struct ext3_reserve_window_node *rsv; | 193 | struct ext3_reserve_window_node *rsv; |
@@ -223,7 +223,7 @@ void ext3_rsv_window_add(struct super_block *sb, | |||
223 | { | 223 | { |
224 | struct rb_root *root = &EXT3_SB(sb)->s_rsv_window_root; | 224 | struct rb_root *root = &EXT3_SB(sb)->s_rsv_window_root; |
225 | struct rb_node *node = &rsv->rsv_node; | 225 | struct rb_node *node = &rsv->rsv_node; |
226 | unsigned int start = rsv->rsv_start; | 226 | ext3_fsblk_t start = rsv->rsv_start; |
227 | 227 | ||
228 | struct rb_node ** p = &root->rb_node; | 228 | struct rb_node ** p = &root->rb_node; |
229 | struct rb_node * parent = NULL; | 229 | struct rb_node * parent = NULL; |
@@ -310,20 +310,20 @@ void ext3_discard_reservation(struct inode *inode) | |||
310 | 310 | ||
311 | /* Free given blocks, update quota and i_blocks field */ | 311 | /* Free given blocks, update quota and i_blocks field */ |
312 | void ext3_free_blocks_sb(handle_t *handle, struct super_block *sb, | 312 | void ext3_free_blocks_sb(handle_t *handle, struct super_block *sb, |
313 | unsigned long block, unsigned long count, | 313 | ext3_fsblk_t block, unsigned long count, |
314 | int *pdquot_freed_blocks) | 314 | unsigned long *pdquot_freed_blocks) |
315 | { | 315 | { |
316 | struct buffer_head *bitmap_bh = NULL; | 316 | struct buffer_head *bitmap_bh = NULL; |
317 | struct buffer_head *gd_bh; | 317 | struct buffer_head *gd_bh; |
318 | unsigned long block_group; | 318 | unsigned long block_group; |
319 | unsigned long bit; | 319 | ext3_grpblk_t bit; |
320 | unsigned long i; | 320 | unsigned long i; |
321 | unsigned long overflow; | 321 | unsigned long overflow; |
322 | struct ext3_group_desc * desc; | 322 | struct ext3_group_desc * desc; |
323 | struct ext3_super_block * es; | 323 | struct ext3_super_block * es; |
324 | struct ext3_sb_info *sbi; | 324 | struct ext3_sb_info *sbi; |
325 | int err = 0, ret; | 325 | int err = 0, ret; |
326 | unsigned group_freed; | 326 | ext3_grpblk_t group_freed; |
327 | 327 | ||
328 | *pdquot_freed_blocks = 0; | 328 | *pdquot_freed_blocks = 0; |
329 | sbi = EXT3_SB(sb); | 329 | sbi = EXT3_SB(sb); |
@@ -333,7 +333,7 @@ void ext3_free_blocks_sb(handle_t *handle, struct super_block *sb, | |||
333 | block + count > le32_to_cpu(es->s_blocks_count)) { | 333 | block + count > le32_to_cpu(es->s_blocks_count)) { |
334 | ext3_error (sb, "ext3_free_blocks", | 334 | ext3_error (sb, "ext3_free_blocks", |
335 | "Freeing blocks not in datazone - " | 335 | "Freeing blocks not in datazone - " |
336 | "block = %lu, count = %lu", block, count); | 336 | "block = "E3FSBLK", count = %lu", block, count); |
337 | goto error_return; | 337 | goto error_return; |
338 | } | 338 | } |
339 | 339 | ||
@@ -369,7 +369,7 @@ do_more: | |||
369 | sbi->s_itb_per_group)) | 369 | sbi->s_itb_per_group)) |
370 | ext3_error (sb, "ext3_free_blocks", | 370 | ext3_error (sb, "ext3_free_blocks", |
371 | "Freeing blocks in system zones - " | 371 | "Freeing blocks in system zones - " |
372 | "Block = %lu, count = %lu", | 372 | "Block = "E3FSBLK", count = %lu", |
373 | block, count); | 373 | block, count); |
374 | 374 | ||
375 | /* | 375 | /* |
@@ -453,7 +453,8 @@ do_more: | |||
453 | bit + i, bitmap_bh->b_data)) { | 453 | bit + i, bitmap_bh->b_data)) { |
454 | jbd_unlock_bh_state(bitmap_bh); | 454 | jbd_unlock_bh_state(bitmap_bh); |
455 | ext3_error(sb, __FUNCTION__, | 455 | ext3_error(sb, __FUNCTION__, |
456 | "bit already cleared for block %lu", block + i); | 456 | "bit already cleared for block "E3FSBLK, |
457 | block + i); | ||
457 | jbd_lock_bh_state(bitmap_bh); | 458 | jbd_lock_bh_state(bitmap_bh); |
458 | BUFFER_TRACE(bitmap_bh, "bit already cleared"); | 459 | BUFFER_TRACE(bitmap_bh, "bit already cleared"); |
459 | } else { | 460 | } else { |
@@ -493,10 +494,10 @@ error_return: | |||
493 | 494 | ||
494 | /* Free given blocks, update quota and i_blocks field */ | 495 | /* Free given blocks, update quota and i_blocks field */ |
495 | void ext3_free_blocks(handle_t *handle, struct inode *inode, | 496 | void ext3_free_blocks(handle_t *handle, struct inode *inode, |
496 | unsigned long block, unsigned long count) | 497 | ext3_fsblk_t block, unsigned long count) |
497 | { | 498 | { |
498 | struct super_block * sb; | 499 | struct super_block * sb; |
499 | int dquot_freed_blocks; | 500 | unsigned long dquot_freed_blocks; |
500 | 501 | ||
501 | sb = inode->i_sb; | 502 | sb = inode->i_sb; |
502 | if (!sb) { | 503 | if (!sb) { |
@@ -525,7 +526,7 @@ void ext3_free_blocks(handle_t *handle, struct inode *inode, | |||
525 | * data-writes at some point, and disable it for metadata allocations or | 526 | * data-writes at some point, and disable it for metadata allocations or |
526 | * sync-data inodes. | 527 | * sync-data inodes. |
527 | */ | 528 | */ |
528 | static int ext3_test_allocatable(int nr, struct buffer_head *bh) | 529 | static int ext3_test_allocatable(ext3_grpblk_t nr, struct buffer_head *bh) |
529 | { | 530 | { |
530 | int ret; | 531 | int ret; |
531 | struct journal_head *jh = bh2jh(bh); | 532 | struct journal_head *jh = bh2jh(bh); |
@@ -542,11 +543,11 @@ static int ext3_test_allocatable(int nr, struct buffer_head *bh) | |||
542 | return ret; | 543 | return ret; |
543 | } | 544 | } |
544 | 545 | ||
545 | static int | 546 | static ext3_grpblk_t |
546 | bitmap_search_next_usable_block(int start, struct buffer_head *bh, | 547 | bitmap_search_next_usable_block(ext3_grpblk_t start, struct buffer_head *bh, |
547 | int maxblocks) | 548 | ext3_grpblk_t maxblocks) |
548 | { | 549 | { |
549 | int next; | 550 | ext3_grpblk_t next; |
550 | struct journal_head *jh = bh2jh(bh); | 551 | struct journal_head *jh = bh2jh(bh); |
551 | 552 | ||
552 | /* | 553 | /* |
@@ -576,10 +577,11 @@ bitmap_search_next_usable_block(int start, struct buffer_head *bh, | |||
576 | * the initial goal; then for a free byte somewhere in the bitmap; then | 577 | * the initial goal; then for a free byte somewhere in the bitmap; then |
577 | * for any free bit in the bitmap. | 578 | * for any free bit in the bitmap. |
578 | */ | 579 | */ |
579 | static int | 580 | static ext3_grpblk_t |
580 | find_next_usable_block(int start, struct buffer_head *bh, int maxblocks) | 581 | find_next_usable_block(ext3_grpblk_t start, struct buffer_head *bh, |
582 | ext3_grpblk_t maxblocks) | ||
581 | { | 583 | { |
582 | int here, next; | 584 | ext3_grpblk_t here, next; |
583 | char *p, *r; | 585 | char *p, *r; |
584 | 586 | ||
585 | if (start > 0) { | 587 | if (start > 0) { |
@@ -591,7 +593,7 @@ find_next_usable_block(int start, struct buffer_head *bh, int maxblocks) | |||
591 | * less than EXT3_BLOCKS_PER_GROUP. Aligning up to the | 593 | * less than EXT3_BLOCKS_PER_GROUP. Aligning up to the |
592 | * next 64-bit boundary is simple.. | 594 | * next 64-bit boundary is simple.. |
593 | */ | 595 | */ |
594 | int end_goal = (start + 63) & ~63; | 596 | ext3_grpblk_t end_goal = (start + 63) & ~63; |
595 | if (end_goal > maxblocks) | 597 | if (end_goal > maxblocks) |
596 | end_goal = maxblocks; | 598 | end_goal = maxblocks; |
597 | here = ext3_find_next_zero_bit(bh->b_data, end_goal, start); | 599 | here = ext3_find_next_zero_bit(bh->b_data, end_goal, start); |
@@ -628,7 +630,7 @@ find_next_usable_block(int start, struct buffer_head *bh, int maxblocks) | |||
628 | * zero (failure). | 630 | * zero (failure). |
629 | */ | 631 | */ |
630 | static inline int | 632 | static inline int |
631 | claim_block(spinlock_t *lock, int block, struct buffer_head *bh) | 633 | claim_block(spinlock_t *lock, ext3_grpblk_t block, struct buffer_head *bh) |
632 | { | 634 | { |
633 | struct journal_head *jh = bh2jh(bh); | 635 | struct journal_head *jh = bh2jh(bh); |
634 | int ret; | 636 | int ret; |
@@ -651,12 +653,13 @@ claim_block(spinlock_t *lock, int block, struct buffer_head *bh) | |||
651 | * new bitmap. In that case we must release write access to the old one via | 653 | * new bitmap. In that case we must release write access to the old one via |
652 | * ext3_journal_release_buffer(), else we'll run out of credits. | 654 | * ext3_journal_release_buffer(), else we'll run out of credits. |
653 | */ | 655 | */ |
654 | static int | 656 | static ext3_grpblk_t |
655 | ext3_try_to_allocate(struct super_block *sb, handle_t *handle, int group, | 657 | ext3_try_to_allocate(struct super_block *sb, handle_t *handle, int group, |
656 | struct buffer_head *bitmap_bh, int goal, | 658 | struct buffer_head *bitmap_bh, ext3_grpblk_t grp_goal, |
657 | unsigned long *count, struct ext3_reserve_window *my_rsv) | 659 | unsigned long *count, struct ext3_reserve_window *my_rsv) |
658 | { | 660 | { |
659 | int group_first_block, start, end; | 661 | ext3_fsblk_t group_first_block; |
662 | ext3_grpblk_t start, end; | ||
660 | unsigned long num = 0; | 663 | unsigned long num = 0; |
661 | 664 | ||
662 | /* we do allocation within the reservation window if we have a window */ | 665 | /* we do allocation within the reservation window if we have a window */ |
@@ -673,13 +676,13 @@ ext3_try_to_allocate(struct super_block *sb, handle_t *handle, int group, | |||
673 | if (end > EXT3_BLOCKS_PER_GROUP(sb)) | 676 | if (end > EXT3_BLOCKS_PER_GROUP(sb)) |
674 | /* reservation window crosses group boundary */ | 677 | /* reservation window crosses group boundary */ |
675 | end = EXT3_BLOCKS_PER_GROUP(sb); | 678 | end = EXT3_BLOCKS_PER_GROUP(sb); |
676 | if ((start <= goal) && (goal < end)) | 679 | if ((start <= grp_goal) && (grp_goal < end)) |
677 | start = goal; | 680 | start = grp_goal; |
678 | else | 681 | else |
679 | goal = -1; | 682 | grp_goal = -1; |
680 | } else { | 683 | } else { |
681 | if (goal > 0) | 684 | if (grp_goal > 0) |
682 | start = goal; | 685 | start = grp_goal; |
683 | else | 686 | else |
684 | start = 0; | 687 | start = 0; |
685 | end = EXT3_BLOCKS_PER_GROUP(sb); | 688 | end = EXT3_BLOCKS_PER_GROUP(sb); |
@@ -688,43 +691,43 @@ ext3_try_to_allocate(struct super_block *sb, handle_t *handle, int group, | |||
688 | BUG_ON(start > EXT3_BLOCKS_PER_GROUP(sb)); | 691 | BUG_ON(start > EXT3_BLOCKS_PER_GROUP(sb)); |
689 | 692 | ||
690 | repeat: | 693 | repeat: |
691 | if (goal < 0 || !ext3_test_allocatable(goal, bitmap_bh)) { | 694 | if (grp_goal < 0 || !ext3_test_allocatable(grp_goal, bitmap_bh)) { |
692 | goal = find_next_usable_block(start, bitmap_bh, end); | 695 | grp_goal = find_next_usable_block(start, bitmap_bh, end); |
693 | if (goal < 0) | 696 | if (grp_goal < 0) |
694 | goto fail_access; | 697 | goto fail_access; |
695 | if (!my_rsv) { | 698 | if (!my_rsv) { |
696 | int i; | 699 | int i; |
697 | 700 | ||
698 | for (i = 0; i < 7 && goal > start && | 701 | for (i = 0; i < 7 && grp_goal > start && |
699 | ext3_test_allocatable(goal - 1, | 702 | ext3_test_allocatable(grp_goal - 1, |
700 | bitmap_bh); | 703 | bitmap_bh); |
701 | i++, goal--) | 704 | i++, grp_goal--) |
702 | ; | 705 | ; |
703 | } | 706 | } |
704 | } | 707 | } |
705 | start = goal; | 708 | start = grp_goal; |
706 | 709 | ||
707 | if (!claim_block(sb_bgl_lock(EXT3_SB(sb), group), goal, bitmap_bh)) { | 710 | if (!claim_block(sb_bgl_lock(EXT3_SB(sb), group), grp_goal, bitmap_bh)) { |
708 | /* | 711 | /* |
709 | * The block was allocated by another thread, or it was | 712 | * The block was allocated by another thread, or it was |
710 | * allocated and then freed by another thread | 713 | * allocated and then freed by another thread |
711 | */ | 714 | */ |
712 | start++; | 715 | start++; |
713 | goal++; | 716 | grp_goal++; |
714 | if (start >= end) | 717 | if (start >= end) |
715 | goto fail_access; | 718 | goto fail_access; |
716 | goto repeat; | 719 | goto repeat; |
717 | } | 720 | } |
718 | num++; | 721 | num++; |
719 | goal++; | 722 | grp_goal++; |
720 | while (num < *count && goal < end | 723 | while (num < *count && grp_goal < end |
721 | && ext3_test_allocatable(goal, bitmap_bh) | 724 | && ext3_test_allocatable(grp_goal, bitmap_bh) |
722 | && claim_block(sb_bgl_lock(EXT3_SB(sb), group), goal, bitmap_bh)) { | 725 | && claim_block(sb_bgl_lock(EXT3_SB(sb), group), grp_goal, bitmap_bh)) { |
723 | num++; | 726 | num++; |
724 | goal++; | 727 | grp_goal++; |
725 | } | 728 | } |
726 | *count = num; | 729 | *count = num; |
727 | return goal - num; | 730 | return grp_goal - num; |
728 | fail_access: | 731 | fail_access: |
729 | *count = num; | 732 | *count = num; |
730 | return -1; | 733 | return -1; |
@@ -766,12 +769,13 @@ fail_access: | |||
766 | static int find_next_reservable_window( | 769 | static int find_next_reservable_window( |
767 | struct ext3_reserve_window_node *search_head, | 770 | struct ext3_reserve_window_node *search_head, |
768 | struct ext3_reserve_window_node *my_rsv, | 771 | struct ext3_reserve_window_node *my_rsv, |
769 | struct super_block * sb, int start_block, | 772 | struct super_block * sb, |
770 | int last_block) | 773 | ext3_fsblk_t start_block, |
774 | ext3_fsblk_t last_block) | ||
771 | { | 775 | { |
772 | struct rb_node *next; | 776 | struct rb_node *next; |
773 | struct ext3_reserve_window_node *rsv, *prev; | 777 | struct ext3_reserve_window_node *rsv, *prev; |
774 | int cur; | 778 | ext3_fsblk_t cur; |
775 | int size = my_rsv->rsv_goal_size; | 779 | int size = my_rsv->rsv_goal_size; |
776 | 780 | ||
777 | /* TODO: make the start of the reservation window byte-aligned */ | 781 | /* TODO: make the start of the reservation window byte-aligned */ |
@@ -873,10 +877,10 @@ static int find_next_reservable_window( | |||
873 | * | 877 | * |
874 | * @rsv: the reservation | 878 | * @rsv: the reservation |
875 | * | 879 | * |
876 | * @goal: The goal (group-relative). It is where the search for a | 880 | * @grp_goal: The goal (group-relative). It is where the search for a |
877 | * free reservable space should start from. | 881 | * free reservable space should start from. |
878 | * if we have a goal(goal >0 ), then start from there, | 882 | * if we have a grp_goal(grp_goal >0 ), then start from there, |
879 | * no goal(goal = -1), we start from the first block | 883 | * no grp_goal(grp_goal = -1), we start from the first block |
880 | * of the group. | 884 | * of the group. |
881 | * | 885 | * |
882 | * @sb: the super block | 886 | * @sb: the super block |
@@ -885,12 +889,12 @@ static int find_next_reservable_window( | |||
885 | * | 889 | * |
886 | */ | 890 | */ |
887 | static int alloc_new_reservation(struct ext3_reserve_window_node *my_rsv, | 891 | static int alloc_new_reservation(struct ext3_reserve_window_node *my_rsv, |
888 | int goal, struct super_block *sb, | 892 | ext3_grpblk_t grp_goal, struct super_block *sb, |
889 | unsigned int group, struct buffer_head *bitmap_bh) | 893 | unsigned int group, struct buffer_head *bitmap_bh) |
890 | { | 894 | { |
891 | struct ext3_reserve_window_node *search_head; | 895 | struct ext3_reserve_window_node *search_head; |
892 | int group_first_block, group_end_block, start_block; | 896 | ext3_fsblk_t group_first_block, group_end_block, start_block; |
893 | int first_free_block; | 897 | ext3_grpblk_t first_free_block; |
894 | struct rb_root *fs_rsv_root = &EXT3_SB(sb)->s_rsv_window_root; | 898 | struct rb_root *fs_rsv_root = &EXT3_SB(sb)->s_rsv_window_root; |
895 | unsigned long size; | 899 | unsigned long size; |
896 | int ret; | 900 | int ret; |
@@ -900,10 +904,10 @@ static int alloc_new_reservation(struct ext3_reserve_window_node *my_rsv, | |||
900 | group * EXT3_BLOCKS_PER_GROUP(sb); | 904 | group * EXT3_BLOCKS_PER_GROUP(sb); |
901 | group_end_block = group_first_block + EXT3_BLOCKS_PER_GROUP(sb) - 1; | 905 | group_end_block = group_first_block + EXT3_BLOCKS_PER_GROUP(sb) - 1; |
902 | 906 | ||
903 | if (goal < 0) | 907 | if (grp_goal < 0) |
904 | start_block = group_first_block; | 908 | start_block = group_first_block; |
905 | else | 909 | else |
906 | start_block = goal + group_first_block; | 910 | start_block = grp_goal + group_first_block; |
907 | 911 | ||
908 | size = my_rsv->rsv_goal_size; | 912 | size = my_rsv->rsv_goal_size; |
909 | 913 | ||
@@ -1057,14 +1061,15 @@ static void try_to_extend_reservation(struct ext3_reserve_window_node *my_rsv, | |||
1057 | * sorted double linked list should be fast. | 1061 | * sorted double linked list should be fast. |
1058 | * | 1062 | * |
1059 | */ | 1063 | */ |
1060 | static int | 1064 | static ext3_grpblk_t |
1061 | ext3_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle, | 1065 | ext3_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle, |
1062 | unsigned int group, struct buffer_head *bitmap_bh, | 1066 | unsigned int group, struct buffer_head *bitmap_bh, |
1063 | int goal, struct ext3_reserve_window_node * my_rsv, | 1067 | ext3_grpblk_t grp_goal, |
1068 | struct ext3_reserve_window_node * my_rsv, | ||
1064 | unsigned long *count, int *errp) | 1069 | unsigned long *count, int *errp) |
1065 | { | 1070 | { |
1066 | unsigned long group_first_block; | 1071 | ext3_fsblk_t group_first_block; |
1067 | int ret = 0; | 1072 | ext3_grpblk_t ret = 0; |
1068 | int fatal; | 1073 | int fatal; |
1069 | unsigned long num = *count; | 1074 | unsigned long num = *count; |
1070 | 1075 | ||
@@ -1090,12 +1095,12 @@ ext3_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle, | |||
1090 | */ | 1095 | */ |
1091 | if (my_rsv == NULL ) { | 1096 | if (my_rsv == NULL ) { |
1092 | ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh, | 1097 | ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh, |
1093 | goal, count, NULL); | 1098 | grp_goal, count, NULL); |
1094 | goto out; | 1099 | goto out; |
1095 | } | 1100 | } |
1096 | /* | 1101 | /* |
1097 | * goal is a group relative block number (if there is a goal) | 1102 | * grp_goal is a group relative block number (if there is a goal) |
1098 | * 0 < goal < EXT3_BLOCKS_PER_GROUP(sb) | 1103 | * 0 < grp_goal < EXT3_BLOCKS_PER_GROUP(sb) |
1099 | * first block is a filesystem wide block number | 1104 | * first block is a filesystem wide block number |
1100 | * first block is the block number of the first block in this group | 1105 | * first block is the block number of the first block in this group |
1101 | */ | 1106 | */ |
@@ -1119,24 +1124,24 @@ ext3_try_to_allocate_with_rsv(struct super_block *sb, handle_t *handle, | |||
1119 | */ | 1124 | */ |
1120 | while (1) { | 1125 | while (1) { |
1121 | if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) || | 1126 | if (rsv_is_empty(&my_rsv->rsv_window) || (ret < 0) || |
1122 | !goal_in_my_reservation(&my_rsv->rsv_window, goal, group, sb)) { | 1127 | !goal_in_my_reservation(&my_rsv->rsv_window, grp_goal, group, sb)) { |
1123 | if (my_rsv->rsv_goal_size < *count) | 1128 | if (my_rsv->rsv_goal_size < *count) |
1124 | my_rsv->rsv_goal_size = *count; | 1129 | my_rsv->rsv_goal_size = *count; |
1125 | ret = alloc_new_reservation(my_rsv, goal, sb, | 1130 | ret = alloc_new_reservation(my_rsv, grp_goal, sb, |
1126 | group, bitmap_bh); | 1131 | group, bitmap_bh); |
1127 | if (ret < 0) | 1132 | if (ret < 0) |
1128 | break; /* failed */ | 1133 | break; /* failed */ |
1129 | 1134 | ||
1130 | if (!goal_in_my_reservation(&my_rsv->rsv_window, goal, group, sb)) | 1135 | if (!goal_in_my_reservation(&my_rsv->rsv_window, grp_goal, group, sb)) |
1131 | goal = -1; | 1136 | grp_goal = -1; |
1132 | } else if (goal > 0 && (my_rsv->rsv_end-goal+1) < *count) | 1137 | } else if (grp_goal > 0 && (my_rsv->rsv_end-grp_goal+1) < *count) |
1133 | try_to_extend_reservation(my_rsv, sb, | 1138 | try_to_extend_reservation(my_rsv, sb, |
1134 | *count-my_rsv->rsv_end + goal - 1); | 1139 | *count-my_rsv->rsv_end + grp_goal - 1); |
1135 | 1140 | ||
1136 | if ((my_rsv->rsv_start >= group_first_block + EXT3_BLOCKS_PER_GROUP(sb)) | 1141 | if ((my_rsv->rsv_start >= group_first_block + EXT3_BLOCKS_PER_GROUP(sb)) |
1137 | || (my_rsv->rsv_end < group_first_block)) | 1142 | || (my_rsv->rsv_end < group_first_block)) |
1138 | BUG(); | 1143 | BUG(); |
1139 | ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh, goal, | 1144 | ret = ext3_try_to_allocate(sb, handle, group, bitmap_bh, grp_goal, |
1140 | &num, &my_rsv->rsv_window); | 1145 | &num, &my_rsv->rsv_window); |
1141 | if (ret >= 0) { | 1146 | if (ret >= 0) { |
1142 | my_rsv->rsv_alloc_hit += num; | 1147 | my_rsv->rsv_alloc_hit += num; |
@@ -1164,7 +1169,7 @@ out: | |||
1164 | 1169 | ||
1165 | static int ext3_has_free_blocks(struct ext3_sb_info *sbi) | 1170 | static int ext3_has_free_blocks(struct ext3_sb_info *sbi) |
1166 | { | 1171 | { |
1167 | int free_blocks, root_blocks; | 1172 | ext3_fsblk_t free_blocks, root_blocks; |
1168 | 1173 | ||
1169 | free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); | 1174 | free_blocks = percpu_counter_read_positive(&sbi->s_freeblocks_counter); |
1170 | root_blocks = le32_to_cpu(sbi->s_es->s_r_blocks_count); | 1175 | root_blocks = le32_to_cpu(sbi->s_es->s_r_blocks_count); |
@@ -1200,19 +1205,20 @@ int ext3_should_retry_alloc(struct super_block *sb, int *retries) | |||
1200 | * bitmap, and then for any free bit if that fails. | 1205 | * bitmap, and then for any free bit if that fails. |
1201 | * This function also updates quota and i_blocks field. | 1206 | * This function also updates quota and i_blocks field. |
1202 | */ | 1207 | */ |
1203 | int ext3_new_blocks(handle_t *handle, struct inode *inode, | 1208 | ext3_fsblk_t ext3_new_blocks(handle_t *handle, struct inode *inode, |
1204 | unsigned long goal, unsigned long *count, int *errp) | 1209 | ext3_fsblk_t goal, unsigned long *count, int *errp) |
1205 | { | 1210 | { |
1206 | struct buffer_head *bitmap_bh = NULL; | 1211 | struct buffer_head *bitmap_bh = NULL; |
1207 | struct buffer_head *gdp_bh; | 1212 | struct buffer_head *gdp_bh; |
1208 | int group_no; | 1213 | int group_no; |
1209 | int goal_group; | 1214 | int goal_group; |
1210 | int ret_block; | 1215 | ext3_grpblk_t grp_target_blk; /* blockgroup relative goal block */ |
1216 | ext3_grpblk_t grp_alloc_blk; /* blockgroup-relative allocated block*/ | ||
1217 | ext3_fsblk_t ret_block; /* filesyetem-wide allocated block */ | ||
1211 | int bgi; /* blockgroup iteration index */ | 1218 | int bgi; /* blockgroup iteration index */ |
1212 | int target_block; | ||
1213 | int fatal = 0, err; | 1219 | int fatal = 0, err; |
1214 | int performed_allocation = 0; | 1220 | int performed_allocation = 0; |
1215 | int free_blocks; | 1221 | ext3_grpblk_t free_blocks; /* number of free blocks in a group */ |
1216 | struct super_block *sb; | 1222 | struct super_block *sb; |
1217 | struct ext3_group_desc *gdp; | 1223 | struct ext3_group_desc *gdp; |
1218 | struct ext3_super_block *es; | 1224 | struct ext3_super_block *es; |
@@ -1285,16 +1291,17 @@ retry: | |||
1285 | my_rsv = NULL; | 1291 | my_rsv = NULL; |
1286 | 1292 | ||
1287 | if (free_blocks > 0) { | 1293 | if (free_blocks > 0) { |
1288 | ret_block = ((goal - le32_to_cpu(es->s_first_data_block)) % | 1294 | grp_target_blk = ((goal - le32_to_cpu(es->s_first_data_block)) % |
1289 | EXT3_BLOCKS_PER_GROUP(sb)); | 1295 | EXT3_BLOCKS_PER_GROUP(sb)); |
1290 | bitmap_bh = read_block_bitmap(sb, group_no); | 1296 | bitmap_bh = read_block_bitmap(sb, group_no); |
1291 | if (!bitmap_bh) | 1297 | if (!bitmap_bh) |
1292 | goto io_error; | 1298 | goto io_error; |
1293 | ret_block = ext3_try_to_allocate_with_rsv(sb, handle, group_no, | 1299 | grp_alloc_blk = ext3_try_to_allocate_with_rsv(sb, handle, |
1294 | bitmap_bh, ret_block, my_rsv, &num, &fatal); | 1300 | group_no, bitmap_bh, grp_target_blk, |
1301 | my_rsv, &num, &fatal); | ||
1295 | if (fatal) | 1302 | if (fatal) |
1296 | goto out; | 1303 | goto out; |
1297 | if (ret_block >= 0) | 1304 | if (grp_alloc_blk >= 0) |
1298 | goto allocated; | 1305 | goto allocated; |
1299 | } | 1306 | } |
1300 | 1307 | ||
@@ -1327,11 +1334,15 @@ retry: | |||
1327 | bitmap_bh = read_block_bitmap(sb, group_no); | 1334 | bitmap_bh = read_block_bitmap(sb, group_no); |
1328 | if (!bitmap_bh) | 1335 | if (!bitmap_bh) |
1329 | goto io_error; | 1336 | goto io_error; |
1330 | ret_block = ext3_try_to_allocate_with_rsv(sb, handle, group_no, | 1337 | /* |
1331 | bitmap_bh, -1, my_rsv, &num, &fatal); | 1338 | * try to allocate block(s) from this group, without a goal(-1). |
1339 | */ | ||
1340 | grp_alloc_blk = ext3_try_to_allocate_with_rsv(sb, handle, | ||
1341 | group_no, bitmap_bh, -1, my_rsv, | ||
1342 | &num, &fatal); | ||
1332 | if (fatal) | 1343 | if (fatal) |
1333 | goto out; | 1344 | goto out; |
1334 | if (ret_block >= 0) | 1345 | if (grp_alloc_blk >= 0) |
1335 | goto allocated; | 1346 | goto allocated; |
1336 | } | 1347 | } |
1337 | /* | 1348 | /* |
@@ -1360,18 +1371,19 @@ allocated: | |||
1360 | if (fatal) | 1371 | if (fatal) |
1361 | goto out; | 1372 | goto out; |
1362 | 1373 | ||
1363 | target_block = ret_block + group_no * EXT3_BLOCKS_PER_GROUP(sb) | 1374 | ret_block = grp_alloc_blk + group_no * EXT3_BLOCKS_PER_GROUP(sb) |
1364 | + le32_to_cpu(es->s_first_data_block); | 1375 | + le32_to_cpu(es->s_first_data_block); |
1365 | 1376 | ||
1366 | if (in_range(le32_to_cpu(gdp->bg_block_bitmap), target_block, num) || | 1377 | if (in_range(le32_to_cpu(gdp->bg_block_bitmap), ret_block, num) || |
1367 | in_range(le32_to_cpu(gdp->bg_inode_bitmap), target_block, num) || | 1378 | in_range(le32_to_cpu(gdp->bg_inode_bitmap), ret_block, num) || |
1368 | in_range(target_block, le32_to_cpu(gdp->bg_inode_table), | 1379 | in_range(ret_block, le32_to_cpu(gdp->bg_inode_table), |
1369 | EXT3_SB(sb)->s_itb_per_group) || | 1380 | EXT3_SB(sb)->s_itb_per_group) || |
1370 | in_range(target_block + num - 1, le32_to_cpu(gdp->bg_inode_table), | 1381 | in_range(ret_block + num - 1, le32_to_cpu(gdp->bg_inode_table), |
1371 | EXT3_SB(sb)->s_itb_per_group)) | 1382 | EXT3_SB(sb)->s_itb_per_group)) |
1372 | ext3_error(sb, "ext3_new_block", | 1383 | ext3_error(sb, "ext3_new_block", |
1373 | "Allocating block in system zone - " | 1384 | "Allocating block in system zone - " |
1374 | "blocks from %u, length %lu", target_block, num); | 1385 | "blocks from "E3FSBLK", length %lu", |
1386 | ret_block, num); | ||
1375 | 1387 | ||
1376 | performed_allocation = 1; | 1388 | performed_allocation = 1; |
1377 | 1389 | ||
@@ -1380,7 +1392,7 @@ allocated: | |||
1380 | struct buffer_head *debug_bh; | 1392 | struct buffer_head *debug_bh; |
1381 | 1393 | ||
1382 | /* Record bitmap buffer state in the newly allocated block */ | 1394 | /* Record bitmap buffer state in the newly allocated block */ |
1383 | debug_bh = sb_find_get_block(sb, target_block); | 1395 | debug_bh = sb_find_get_block(sb, ret_block); |
1384 | if (debug_bh) { | 1396 | if (debug_bh) { |
1385 | BUFFER_TRACE(debug_bh, "state when allocated"); | 1397 | BUFFER_TRACE(debug_bh, "state when allocated"); |
1386 | BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap state"); | 1398 | BUFFER_TRACE2(debug_bh, bitmap_bh, "bitmap state"); |
@@ -1393,24 +1405,21 @@ allocated: | |||
1393 | int i; | 1405 | int i; |
1394 | 1406 | ||
1395 | for (i = 0; i < num; i++) { | 1407 | for (i = 0; i < num; i++) { |
1396 | if (ext3_test_bit(ret_block, | 1408 | if (ext3_test_bit(grp_alloc_blk+i, |
1397 | bh2jh(bitmap_bh)->b_committed_data)) { | 1409 | bh2jh(bitmap_bh)->b_committed_data)) { |
1398 | printk("%s: block was unexpectedly set in " | 1410 | printk("%s: block was unexpectedly set in " |
1399 | "b_committed_data\n", __FUNCTION__); | 1411 | "b_committed_data\n", __FUNCTION__); |
1400 | } | 1412 | } |
1401 | } | 1413 | } |
1402 | } | 1414 | } |
1403 | ext3_debug("found bit %d\n", ret_block); | 1415 | ext3_debug("found bit %d\n", grp_alloc_blk); |
1404 | spin_unlock(sb_bgl_lock(sbi, group_no)); | 1416 | spin_unlock(sb_bgl_lock(sbi, group_no)); |
1405 | jbd_unlock_bh_state(bitmap_bh); | 1417 | jbd_unlock_bh_state(bitmap_bh); |
1406 | #endif | 1418 | #endif |
1407 | 1419 | ||
1408 | /* ret_block was blockgroup-relative. Now it becomes fs-relative */ | ||
1409 | ret_block = target_block; | ||
1410 | |||
1411 | if (ret_block + num - 1 >= le32_to_cpu(es->s_blocks_count)) { | 1420 | if (ret_block + num - 1 >= le32_to_cpu(es->s_blocks_count)) { |
1412 | ext3_error(sb, "ext3_new_block", | 1421 | ext3_error(sb, "ext3_new_block", |
1413 | "block(%d) >= blocks count(%d) - " | 1422 | "block("E3FSBLK") >= blocks count(%d) - " |
1414 | "block_group = %d, es == %p ", ret_block, | 1423 | "block_group = %d, es == %p ", ret_block, |
1415 | le32_to_cpu(es->s_blocks_count), group_no, es); | 1424 | le32_to_cpu(es->s_blocks_count), group_no, es); |
1416 | goto out; | 1425 | goto out; |
@@ -1421,7 +1430,7 @@ allocated: | |||
1421 | * list of some description. We don't know in advance whether | 1430 | * list of some description. We don't know in advance whether |
1422 | * the caller wants to use it as metadata or data. | 1431 | * the caller wants to use it as metadata or data. |
1423 | */ | 1432 | */ |
1424 | ext3_debug("allocating block %d. Goal hits %d of %d.\n", | 1433 | ext3_debug("allocating block %lu. Goal hits %d of %d.\n", |
1425 | ret_block, goal_hits, goal_attempts); | 1434 | ret_block, goal_hits, goal_attempts); |
1426 | 1435 | ||
1427 | spin_lock(sb_bgl_lock(sbi, group_no)); | 1436 | spin_lock(sb_bgl_lock(sbi, group_no)); |
@@ -1461,8 +1470,8 @@ out: | |||
1461 | return 0; | 1470 | return 0; |
1462 | } | 1471 | } |
1463 | 1472 | ||
1464 | int ext3_new_block(handle_t *handle, struct inode *inode, | 1473 | ext3_fsblk_t ext3_new_block(handle_t *handle, struct inode *inode, |
1465 | unsigned long goal, int *errp) | 1474 | ext3_fsblk_t goal, int *errp) |
1466 | { | 1475 | { |
1467 | unsigned long count = 1; | 1476 | unsigned long count = 1; |
1468 | 1477 | ||
@@ -1520,7 +1529,7 @@ unsigned long ext3_count_free_blocks(struct super_block *sb) | |||
1520 | } | 1529 | } |
1521 | 1530 | ||
1522 | static inline int | 1531 | static inline int |
1523 | block_in_use(unsigned long block, struct super_block *sb, unsigned char *map) | 1532 | block_in_use(ext3_fsblk_t block, struct super_block *sb, unsigned char *map) |
1524 | { | 1533 | { |
1525 | return ext3_test_bit ((block - | 1534 | return ext3_test_bit ((block - |
1526 | le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block)) % | 1535 | le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block)) % |