diff options
author | Linus Torvalds <torvalds@woody.linux-foundation.org> | 2008-02-26 10:50:16 -0500 |
---|---|---|
committer | Linus Torvalds <torvalds@woody.linux-foundation.org> | 2008-02-26 10:50:16 -0500 |
commit | adefe11c5383face09068176af7dd4e3a80d8f6b (patch) | |
tree | 9a24b21e5b5ab6ae721c3b09f28bb3901e246705 /fs | |
parent | 3d6ce332537ad0a6f77caa469e09d7335187767b (diff) | |
parent | 5606bf5d0cbfbc3dfa78793a3793c43dd045fb1b (diff) |
Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
ext4: add missing ext4_journal_stop()
ext4: ext4_find_next_zero_bit needs an aligned address on some arch
ext4: set EXT4_EXTENTS_FL only for directory and regular files
ext4: Don't mark filesystem error if fallocate fails
ext4: Fix BUG when writing to an unitialized extent
ext4: Don't use ext4_dec_count() if not needed
ext4: modify block allocation algorithm for the last group
ext4: Don't claim block from group which has corrupt bitmap
ext4: Get journal write access before modifying the extent tree
ext4: Fix memory and buffer head leak in callers to ext4_ext_find_extent()
ext4: Don't leave behind a half-created inode if ext4_mkdir() fails
ext4: Fix kernel BUG at fs/ext4/mballoc.c:910!
ext4: Fix locking hierarchy violation in ext4_fallocate()
Remove incorrect BKL comments in ext4
Diffstat (limited to 'fs')
-rw-r--r-- | fs/ext4/dir.c | 2 | ||||
-rw-r--r-- | fs/ext4/extents.c | 59 | ||||
-rw-r--r-- | fs/ext4/ialloc.c | 22 | ||||
-rw-r--r-- | fs/ext4/inode.c | 56 | ||||
-rw-r--r-- | fs/ext4/mballoc.c | 80 | ||||
-rw-r--r-- | fs/ext4/migrate.c | 5 | ||||
-rw-r--r-- | fs/ext4/namei.c | 18 | ||||
-rw-r--r-- | fs/ext4/resize.c | 1 |
8 files changed, 176 insertions, 67 deletions
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c index 33888bb58144..2c23bade9aa6 100644 --- a/fs/ext4/dir.c +++ b/fs/ext4/dir.c | |||
@@ -46,7 +46,7 @@ const struct file_operations ext4_dir_operations = { | |||
46 | #ifdef CONFIG_COMPAT | 46 | #ifdef CONFIG_COMPAT |
47 | .compat_ioctl = ext4_compat_ioctl, | 47 | .compat_ioctl = ext4_compat_ioctl, |
48 | #endif | 48 | #endif |
49 | .fsync = ext4_sync_file, /* BKL held */ | 49 | .fsync = ext4_sync_file, |
50 | .release = ext4_release_dir, | 50 | .release = ext4_release_dir, |
51 | }; | 51 | }; |
52 | 52 | ||
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index bc7081f1fbe8..9ae6e67090cd 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c | |||
@@ -148,6 +148,7 @@ static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode, | |||
148 | { | 148 | { |
149 | struct ext4_inode_info *ei = EXT4_I(inode); | 149 | struct ext4_inode_info *ei = EXT4_I(inode); |
150 | ext4_fsblk_t bg_start; | 150 | ext4_fsblk_t bg_start; |
151 | ext4_fsblk_t last_block; | ||
151 | ext4_grpblk_t colour; | 152 | ext4_grpblk_t colour; |
152 | int depth; | 153 | int depth; |
153 | 154 | ||
@@ -169,8 +170,13 @@ static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode, | |||
169 | /* OK. use inode's group */ | 170 | /* OK. use inode's group */ |
170 | bg_start = (ei->i_block_group * EXT4_BLOCKS_PER_GROUP(inode->i_sb)) + | 171 | bg_start = (ei->i_block_group * EXT4_BLOCKS_PER_GROUP(inode->i_sb)) + |
171 | le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_first_data_block); | 172 | le32_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_first_data_block); |
172 | colour = (current->pid % 16) * | 173 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; |
174 | |||
175 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | ||
176 | colour = (current->pid % 16) * | ||
173 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | 177 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); |
178 | else | ||
179 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | ||
174 | return bg_start + colour + block; | 180 | return bg_start + colour + block; |
175 | } | 181 | } |
176 | 182 | ||
@@ -349,7 +355,7 @@ static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path) | |||
349 | #define ext4_ext_show_leaf(inode,path) | 355 | #define ext4_ext_show_leaf(inode,path) |
350 | #endif | 356 | #endif |
351 | 357 | ||
352 | static void ext4_ext_drop_refs(struct ext4_ext_path *path) | 358 | void ext4_ext_drop_refs(struct ext4_ext_path *path) |
353 | { | 359 | { |
354 | int depth = path->p_depth; | 360 | int depth = path->p_depth; |
355 | int i; | 361 | int i; |
@@ -2168,6 +2174,10 @@ static int ext4_ext_convert_to_initialized(handle_t *handle, | |||
2168 | newblock = iblock - ee_block + ext_pblock(ex); | 2174 | newblock = iblock - ee_block + ext_pblock(ex); |
2169 | ex2 = ex; | 2175 | ex2 = ex; |
2170 | 2176 | ||
2177 | err = ext4_ext_get_access(handle, inode, path + depth); | ||
2178 | if (err) | ||
2179 | goto out; | ||
2180 | |||
2171 | /* ex1: ee_block to iblock - 1 : uninitialized */ | 2181 | /* ex1: ee_block to iblock - 1 : uninitialized */ |
2172 | if (iblock > ee_block) { | 2182 | if (iblock > ee_block) { |
2173 | ex1 = ex; | 2183 | ex1 = ex; |
@@ -2200,16 +2210,20 @@ static int ext4_ext_convert_to_initialized(handle_t *handle, | |||
2200 | newdepth = ext_depth(inode); | 2210 | newdepth = ext_depth(inode); |
2201 | if (newdepth != depth) { | 2211 | if (newdepth != depth) { |
2202 | depth = newdepth; | 2212 | depth = newdepth; |
2203 | path = ext4_ext_find_extent(inode, iblock, NULL); | 2213 | ext4_ext_drop_refs(path); |
2214 | path = ext4_ext_find_extent(inode, iblock, path); | ||
2204 | if (IS_ERR(path)) { | 2215 | if (IS_ERR(path)) { |
2205 | err = PTR_ERR(path); | 2216 | err = PTR_ERR(path); |
2206 | path = NULL; | ||
2207 | goto out; | 2217 | goto out; |
2208 | } | 2218 | } |
2209 | eh = path[depth].p_hdr; | 2219 | eh = path[depth].p_hdr; |
2210 | ex = path[depth].p_ext; | 2220 | ex = path[depth].p_ext; |
2211 | if (ex2 != &newex) | 2221 | if (ex2 != &newex) |
2212 | ex2 = ex; | 2222 | ex2 = ex; |
2223 | |||
2224 | err = ext4_ext_get_access(handle, inode, path + depth); | ||
2225 | if (err) | ||
2226 | goto out; | ||
2213 | } | 2227 | } |
2214 | allocated = max_blocks; | 2228 | allocated = max_blocks; |
2215 | } | 2229 | } |
@@ -2230,9 +2244,6 @@ static int ext4_ext_convert_to_initialized(handle_t *handle, | |||
2230 | ex2->ee_len = cpu_to_le16(allocated); | 2244 | ex2->ee_len = cpu_to_le16(allocated); |
2231 | if (ex2 != ex) | 2245 | if (ex2 != ex) |
2232 | goto insert; | 2246 | goto insert; |
2233 | err = ext4_ext_get_access(handle, inode, path + depth); | ||
2234 | if (err) | ||
2235 | goto out; | ||
2236 | /* | 2247 | /* |
2237 | * New (initialized) extent starts from the first block | 2248 | * New (initialized) extent starts from the first block |
2238 | * in the current extent. i.e., ex2 == ex | 2249 | * in the current extent. i.e., ex2 == ex |
@@ -2276,9 +2287,22 @@ out: | |||
2276 | } | 2287 | } |
2277 | 2288 | ||
2278 | /* | 2289 | /* |
2290 | * Block allocation/map/preallocation routine for extents based files | ||
2291 | * | ||
2292 | * | ||
2279 | * Need to be called with | 2293 | * Need to be called with |
2280 | * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block | 2294 | * down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block |
2281 | * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem) | 2295 | * (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem) |
2296 | * | ||
2297 | * return > 0, number of of blocks already mapped/allocated | ||
2298 | * if create == 0 and these are pre-allocated blocks | ||
2299 | * buffer head is unmapped | ||
2300 | * otherwise blocks are mapped | ||
2301 | * | ||
2302 | * return = 0, if plain look up failed (blocks have not been allocated) | ||
2303 | * buffer head is unmapped | ||
2304 | * | ||
2305 | * return < 0, error case. | ||
2282 | */ | 2306 | */ |
2283 | int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, | 2307 | int ext4_ext_get_blocks(handle_t *handle, struct inode *inode, |
2284 | ext4_lblk_t iblock, | 2308 | ext4_lblk_t iblock, |
@@ -2623,7 +2647,7 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len) | |||
2623 | * modify 1 super block, 1 block bitmap and 1 group descriptor. | 2647 | * modify 1 super block, 1 block bitmap and 1 group descriptor. |
2624 | */ | 2648 | */ |
2625 | credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + 3; | 2649 | credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb) + 3; |
2626 | down_write((&EXT4_I(inode)->i_data_sem)); | 2650 | mutex_lock(&inode->i_mutex); |
2627 | retry: | 2651 | retry: |
2628 | while (ret >= 0 && ret < max_blocks) { | 2652 | while (ret >= 0 && ret < max_blocks) { |
2629 | block = block + ret; | 2653 | block = block + ret; |
@@ -2634,16 +2658,17 @@ retry: | |||
2634 | break; | 2658 | break; |
2635 | } | 2659 | } |
2636 | 2660 | ||
2637 | ret = ext4_ext_get_blocks(handle, inode, block, | 2661 | ret = ext4_get_blocks_wrap(handle, inode, block, |
2638 | max_blocks, &map_bh, | 2662 | max_blocks, &map_bh, |
2639 | EXT4_CREATE_UNINITIALIZED_EXT, 0); | 2663 | EXT4_CREATE_UNINITIALIZED_EXT, 0); |
2640 | WARN_ON(ret <= 0); | ||
2641 | if (ret <= 0) { | 2664 | if (ret <= 0) { |
2642 | ext4_error(inode->i_sb, "ext4_fallocate", | 2665 | #ifdef EXT4FS_DEBUG |
2643 | "ext4_ext_get_blocks returned error: " | 2666 | WARN_ON(ret <= 0); |
2644 | "inode#%lu, block=%u, max_blocks=%lu", | 2667 | printk(KERN_ERR "%s: ext4_ext_get_blocks " |
2668 | "returned error inode#%lu, block=%u, " | ||
2669 | "max_blocks=%lu", __func__, | ||
2645 | inode->i_ino, block, max_blocks); | 2670 | inode->i_ino, block, max_blocks); |
2646 | ret = -EIO; | 2671 | #endif |
2647 | ext4_mark_inode_dirty(handle, inode); | 2672 | ext4_mark_inode_dirty(handle, inode); |
2648 | ret2 = ext4_journal_stop(handle); | 2673 | ret2 = ext4_journal_stop(handle); |
2649 | break; | 2674 | break; |
@@ -2680,7 +2705,6 @@ retry: | |||
2680 | if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) | 2705 | if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) |
2681 | goto retry; | 2706 | goto retry; |
2682 | 2707 | ||
2683 | up_write((&EXT4_I(inode)->i_data_sem)); | ||
2684 | /* | 2708 | /* |
2685 | * Time to update the file size. | 2709 | * Time to update the file size. |
2686 | * Update only when preallocation was requested beyond the file size. | 2710 | * Update only when preallocation was requested beyond the file size. |
@@ -2692,21 +2716,18 @@ retry: | |||
2692 | * if no error, we assume preallocation succeeded | 2716 | * if no error, we assume preallocation succeeded |
2693 | * completely | 2717 | * completely |
2694 | */ | 2718 | */ |
2695 | mutex_lock(&inode->i_mutex); | ||
2696 | i_size_write(inode, offset + len); | 2719 | i_size_write(inode, offset + len); |
2697 | EXT4_I(inode)->i_disksize = i_size_read(inode); | 2720 | EXT4_I(inode)->i_disksize = i_size_read(inode); |
2698 | mutex_unlock(&inode->i_mutex); | ||
2699 | } else if (ret < 0 && nblocks) { | 2721 | } else if (ret < 0 && nblocks) { |
2700 | /* Handle partial allocation scenario */ | 2722 | /* Handle partial allocation scenario */ |
2701 | loff_t newsize; | 2723 | loff_t newsize; |
2702 | 2724 | ||
2703 | mutex_lock(&inode->i_mutex); | ||
2704 | newsize = (nblocks << blkbits) + i_size_read(inode); | 2725 | newsize = (nblocks << blkbits) + i_size_read(inode); |
2705 | i_size_write(inode, EXT4_BLOCK_ALIGN(newsize, blkbits)); | 2726 | i_size_write(inode, EXT4_BLOCK_ALIGN(newsize, blkbits)); |
2706 | EXT4_I(inode)->i_disksize = i_size_read(inode); | 2727 | EXT4_I(inode)->i_disksize = i_size_read(inode); |
2707 | mutex_unlock(&inode->i_mutex); | ||
2708 | } | 2728 | } |
2709 | } | 2729 | } |
2710 | 2730 | ||
2731 | mutex_unlock(&inode->i_mutex); | ||
2711 | return ret > 0 ? ret2 : ret; | 2732 | return ret > 0 ? ret2 : ret; |
2712 | } | 2733 | } |
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c index da18a74b966a..8036b9b5376b 100644 --- a/fs/ext4/ialloc.c +++ b/fs/ext4/ialloc.c | |||
@@ -702,7 +702,12 @@ got: | |||
702 | ei->i_dir_start_lookup = 0; | 702 | ei->i_dir_start_lookup = 0; |
703 | ei->i_disksize = 0; | 703 | ei->i_disksize = 0; |
704 | 704 | ||
705 | ei->i_flags = EXT4_I(dir)->i_flags & ~EXT4_INDEX_FL; | 705 | /* |
706 | * Don't inherit extent flag from directory. We set extent flag on | ||
707 | * newly created directory and file only if -o extent mount option is | ||
708 | * specified | ||
709 | */ | ||
710 | ei->i_flags = EXT4_I(dir)->i_flags & ~(EXT4_INDEX_FL|EXT4_EXTENTS_FL); | ||
706 | if (S_ISLNK(mode)) | 711 | if (S_ISLNK(mode)) |
707 | ei->i_flags &= ~(EXT4_IMMUTABLE_FL|EXT4_APPEND_FL); | 712 | ei->i_flags &= ~(EXT4_IMMUTABLE_FL|EXT4_APPEND_FL); |
708 | /* dirsync only applies to directories */ | 713 | /* dirsync only applies to directories */ |
@@ -745,12 +750,15 @@ got: | |||
745 | goto fail_free_drop; | 750 | goto fail_free_drop; |
746 | } | 751 | } |
747 | if (test_opt(sb, EXTENTS)) { | 752 | if (test_opt(sb, EXTENTS)) { |
748 | EXT4_I(inode)->i_flags |= EXT4_EXTENTS_FL; | 753 | /* set extent flag only for directory and file */ |
749 | ext4_ext_tree_init(handle, inode); | 754 | if (S_ISDIR(mode) || S_ISREG(mode)) { |
750 | err = ext4_update_incompat_feature(handle, sb, | 755 | EXT4_I(inode)->i_flags |= EXT4_EXTENTS_FL; |
751 | EXT4_FEATURE_INCOMPAT_EXTENTS); | 756 | ext4_ext_tree_init(handle, inode); |
752 | if (err) | 757 | err = ext4_update_incompat_feature(handle, sb, |
753 | goto fail; | 758 | EXT4_FEATURE_INCOMPAT_EXTENTS); |
759 | if (err) | ||
760 | goto fail; | ||
761 | } | ||
754 | } | 762 | } |
755 | 763 | ||
756 | ext4_debug("allocating inode %lu\n", inode->i_ino); | 764 | ext4_debug("allocating inode %lu\n", inode->i_ino); |
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 7dd9b50d5ebc..945cbf6cb1fc 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c | |||
@@ -403,6 +403,7 @@ static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind) | |||
403 | __le32 *start = ind->bh ? (__le32*) ind->bh->b_data : ei->i_data; | 403 | __le32 *start = ind->bh ? (__le32*) ind->bh->b_data : ei->i_data; |
404 | __le32 *p; | 404 | __le32 *p; |
405 | ext4_fsblk_t bg_start; | 405 | ext4_fsblk_t bg_start; |
406 | ext4_fsblk_t last_block; | ||
406 | ext4_grpblk_t colour; | 407 | ext4_grpblk_t colour; |
407 | 408 | ||
408 | /* Try to find previous block */ | 409 | /* Try to find previous block */ |
@@ -420,8 +421,13 @@ static ext4_fsblk_t ext4_find_near(struct inode *inode, Indirect *ind) | |||
420 | * into the same cylinder group then. | 421 | * into the same cylinder group then. |
421 | */ | 422 | */ |
422 | bg_start = ext4_group_first_block_no(inode->i_sb, ei->i_block_group); | 423 | bg_start = ext4_group_first_block_no(inode->i_sb, ei->i_block_group); |
423 | colour = (current->pid % 16) * | 424 | last_block = ext4_blocks_count(EXT4_SB(inode->i_sb)->s_es) - 1; |
425 | |||
426 | if (bg_start + EXT4_BLOCKS_PER_GROUP(inode->i_sb) <= last_block) | ||
427 | colour = (current->pid % 16) * | ||
424 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); | 428 | (EXT4_BLOCKS_PER_GROUP(inode->i_sb) / 16); |
429 | else | ||
430 | colour = (current->pid % 16) * ((last_block - bg_start) / 16); | ||
425 | return bg_start + colour; | 431 | return bg_start + colour; |
426 | } | 432 | } |
427 | 433 | ||
@@ -768,7 +774,6 @@ err_out: | |||
768 | * | 774 | * |
769 | * `handle' can be NULL if create == 0. | 775 | * `handle' can be NULL if create == 0. |
770 | * | 776 | * |
771 | * The BKL may not be held on entry here. Be sure to take it early. | ||
772 | * return > 0, # of blocks mapped or allocated. | 777 | * return > 0, # of blocks mapped or allocated. |
773 | * return = 0, if plain lookup failed. | 778 | * return = 0, if plain lookup failed. |
774 | * return < 0, error case. | 779 | * return < 0, error case. |
@@ -903,11 +908,38 @@ out: | |||
903 | */ | 908 | */ |
904 | #define DIO_CREDITS 25 | 909 | #define DIO_CREDITS 25 |
905 | 910 | ||
911 | |||
912 | /* | ||
913 | * | ||
914 | * | ||
915 | * ext4_ext4 get_block() wrapper function | ||
916 | * It will do a look up first, and returns if the blocks already mapped. | ||
917 | * Otherwise it takes the write lock of the i_data_sem and allocate blocks | ||
918 | * and store the allocated blocks in the result buffer head and mark it | ||
919 | * mapped. | ||
920 | * | ||
921 | * If file type is extents based, it will call ext4_ext_get_blocks(), | ||
922 | * Otherwise, call with ext4_get_blocks_handle() to handle indirect mapping | ||
923 | * based files | ||
924 | * | ||
925 | * On success, it returns the number of blocks being mapped or allocate. | ||
926 | * if create==0 and the blocks are pre-allocated and uninitialized block, | ||
927 | * the result buffer head is unmapped. If the create ==1, it will make sure | ||
928 | * the buffer head is mapped. | ||
929 | * | ||
930 | * It returns 0 if plain look up failed (blocks have not been allocated), in | ||
931 | * that casem, buffer head is unmapped | ||
932 | * | ||
933 | * It returns the error in case of allocation failure. | ||
934 | */ | ||
906 | int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block, | 935 | int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block, |
907 | unsigned long max_blocks, struct buffer_head *bh, | 936 | unsigned long max_blocks, struct buffer_head *bh, |
908 | int create, int extend_disksize) | 937 | int create, int extend_disksize) |
909 | { | 938 | { |
910 | int retval; | 939 | int retval; |
940 | |||
941 | clear_buffer_mapped(bh); | ||
942 | |||
911 | /* | 943 | /* |
912 | * Try to see if we can get the block without requesting | 944 | * Try to see if we can get the block without requesting |
913 | * for new file system block. | 945 | * for new file system block. |
@@ -921,12 +953,26 @@ int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block, | |||
921 | inode, block, max_blocks, bh, 0, 0); | 953 | inode, block, max_blocks, bh, 0, 0); |
922 | } | 954 | } |
923 | up_read((&EXT4_I(inode)->i_data_sem)); | 955 | up_read((&EXT4_I(inode)->i_data_sem)); |
924 | if (!create || (retval > 0)) | 956 | |
957 | /* If it is only a block(s) look up */ | ||
958 | if (!create) | ||
959 | return retval; | ||
960 | |||
961 | /* | ||
962 | * Returns if the blocks have already allocated | ||
963 | * | ||
964 | * Note that if blocks have been preallocated | ||
965 | * ext4_ext_get_block() returns th create = 0 | ||
966 | * with buffer head unmapped. | ||
967 | */ | ||
968 | if (retval > 0 && buffer_mapped(bh)) | ||
925 | return retval; | 969 | return retval; |
926 | 970 | ||
927 | /* | 971 | /* |
928 | * We need to allocate new blocks which will result | 972 | * New blocks allocate and/or writing to uninitialized extent |
929 | * in i_data update | 973 | * will possibly result in updating i_data, so we take |
974 | * the write lock of i_data_sem, and call get_blocks() | ||
975 | * with create == 1 flag. | ||
930 | */ | 976 | */ |
931 | down_write((&EXT4_I(inode)->i_data_sem)); | 977 | down_write((&EXT4_I(inode)->i_data_sem)); |
932 | /* | 978 | /* |
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c index dd0fcfcb35ce..ef97f19c2f9d 100644 --- a/fs/ext4/mballoc.c +++ b/fs/ext4/mballoc.c | |||
@@ -627,21 +627,19 @@ static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb, | |||
627 | return block; | 627 | return block; |
628 | } | 628 | } |
629 | 629 | ||
630 | static inline void *mb_correct_addr_and_bit(int *bit, void *addr) | ||
631 | { | ||
630 | #if BITS_PER_LONG == 64 | 632 | #if BITS_PER_LONG == 64 |
631 | #define mb_correct_addr_and_bit(bit, addr) \ | 633 | *bit += ((unsigned long) addr & 7UL) << 3; |
632 | { \ | 634 | addr = (void *) ((unsigned long) addr & ~7UL); |
633 | bit += ((unsigned long) addr & 7UL) << 3; \ | ||
634 | addr = (void *) ((unsigned long) addr & ~7UL); \ | ||
635 | } | ||
636 | #elif BITS_PER_LONG == 32 | 635 | #elif BITS_PER_LONG == 32 |
637 | #define mb_correct_addr_and_bit(bit, addr) \ | 636 | *bit += ((unsigned long) addr & 3UL) << 3; |
638 | { \ | 637 | addr = (void *) ((unsigned long) addr & ~3UL); |
639 | bit += ((unsigned long) addr & 3UL) << 3; \ | ||
640 | addr = (void *) ((unsigned long) addr & ~3UL); \ | ||
641 | } | ||
642 | #else | 638 | #else |
643 | #error "how many bits you are?!" | 639 | #error "how many bits you are?!" |
644 | #endif | 640 | #endif |
641 | return addr; | ||
642 | } | ||
645 | 643 | ||
646 | static inline int mb_test_bit(int bit, void *addr) | 644 | static inline int mb_test_bit(int bit, void *addr) |
647 | { | 645 | { |
@@ -649,34 +647,54 @@ static inline int mb_test_bit(int bit, void *addr) | |||
649 | * ext4_test_bit on architecture like powerpc | 647 | * ext4_test_bit on architecture like powerpc |
650 | * needs unsigned long aligned address | 648 | * needs unsigned long aligned address |
651 | */ | 649 | */ |
652 | mb_correct_addr_and_bit(bit, addr); | 650 | addr = mb_correct_addr_and_bit(&bit, addr); |
653 | return ext4_test_bit(bit, addr); | 651 | return ext4_test_bit(bit, addr); |
654 | } | 652 | } |
655 | 653 | ||
656 | static inline void mb_set_bit(int bit, void *addr) | 654 | static inline void mb_set_bit(int bit, void *addr) |
657 | { | 655 | { |
658 | mb_correct_addr_and_bit(bit, addr); | 656 | addr = mb_correct_addr_and_bit(&bit, addr); |
659 | ext4_set_bit(bit, addr); | 657 | ext4_set_bit(bit, addr); |
660 | } | 658 | } |
661 | 659 | ||
662 | static inline void mb_set_bit_atomic(spinlock_t *lock, int bit, void *addr) | 660 | static inline void mb_set_bit_atomic(spinlock_t *lock, int bit, void *addr) |
663 | { | 661 | { |
664 | mb_correct_addr_and_bit(bit, addr); | 662 | addr = mb_correct_addr_and_bit(&bit, addr); |
665 | ext4_set_bit_atomic(lock, bit, addr); | 663 | ext4_set_bit_atomic(lock, bit, addr); |
666 | } | 664 | } |
667 | 665 | ||
668 | static inline void mb_clear_bit(int bit, void *addr) | 666 | static inline void mb_clear_bit(int bit, void *addr) |
669 | { | 667 | { |
670 | mb_correct_addr_and_bit(bit, addr); | 668 | addr = mb_correct_addr_and_bit(&bit, addr); |
671 | ext4_clear_bit(bit, addr); | 669 | ext4_clear_bit(bit, addr); |
672 | } | 670 | } |
673 | 671 | ||
674 | static inline void mb_clear_bit_atomic(spinlock_t *lock, int bit, void *addr) | 672 | static inline void mb_clear_bit_atomic(spinlock_t *lock, int bit, void *addr) |
675 | { | 673 | { |
676 | mb_correct_addr_and_bit(bit, addr); | 674 | addr = mb_correct_addr_and_bit(&bit, addr); |
677 | ext4_clear_bit_atomic(lock, bit, addr); | 675 | ext4_clear_bit_atomic(lock, bit, addr); |
678 | } | 676 | } |
679 | 677 | ||
678 | static inline int mb_find_next_zero_bit(void *addr, int max, int start) | ||
679 | { | ||
680 | int fix = 0; | ||
681 | addr = mb_correct_addr_and_bit(&fix, addr); | ||
682 | max += fix; | ||
683 | start += fix; | ||
684 | |||
685 | return ext4_find_next_zero_bit(addr, max, start) - fix; | ||
686 | } | ||
687 | |||
688 | static inline int mb_find_next_bit(void *addr, int max, int start) | ||
689 | { | ||
690 | int fix = 0; | ||
691 | addr = mb_correct_addr_and_bit(&fix, addr); | ||
692 | max += fix; | ||
693 | start += fix; | ||
694 | |||
695 | return ext4_find_next_bit(addr, max, start) - fix; | ||
696 | } | ||
697 | |||
680 | static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max) | 698 | static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max) |
681 | { | 699 | { |
682 | char *bb; | 700 | char *bb; |
@@ -906,7 +924,7 @@ static void ext4_mb_mark_free_simple(struct super_block *sb, | |||
906 | unsigned short chunk; | 924 | unsigned short chunk; |
907 | unsigned short border; | 925 | unsigned short border; |
908 | 926 | ||
909 | BUG_ON(len >= EXT4_BLOCKS_PER_GROUP(sb)); | 927 | BUG_ON(len > EXT4_BLOCKS_PER_GROUP(sb)); |
910 | 928 | ||
911 | border = 2 << sb->s_blocksize_bits; | 929 | border = 2 << sb->s_blocksize_bits; |
912 | 930 | ||
@@ -946,12 +964,12 @@ static void ext4_mb_generate_buddy(struct super_block *sb, | |||
946 | 964 | ||
947 | /* initialize buddy from bitmap which is aggregation | 965 | /* initialize buddy from bitmap which is aggregation |
948 | * of on-disk bitmap and preallocations */ | 966 | * of on-disk bitmap and preallocations */ |
949 | i = ext4_find_next_zero_bit(bitmap, max, 0); | 967 | i = mb_find_next_zero_bit(bitmap, max, 0); |
950 | grp->bb_first_free = i; | 968 | grp->bb_first_free = i; |
951 | while (i < max) { | 969 | while (i < max) { |
952 | fragments++; | 970 | fragments++; |
953 | first = i; | 971 | first = i; |
954 | i = ext4_find_next_bit(bitmap, max, i); | 972 | i = mb_find_next_bit(bitmap, max, i); |
955 | len = i - first; | 973 | len = i - first; |
956 | free += len; | 974 | free += len; |
957 | if (len > 1) | 975 | if (len > 1) |
@@ -959,7 +977,7 @@ static void ext4_mb_generate_buddy(struct super_block *sb, | |||
959 | else | 977 | else |
960 | grp->bb_counters[0]++; | 978 | grp->bb_counters[0]++; |
961 | if (i < max) | 979 | if (i < max) |
962 | i = ext4_find_next_zero_bit(bitmap, max, i); | 980 | i = mb_find_next_zero_bit(bitmap, max, i); |
963 | } | 981 | } |
964 | grp->bb_fragments = fragments; | 982 | grp->bb_fragments = fragments; |
965 | 983 | ||
@@ -967,6 +985,10 @@ static void ext4_mb_generate_buddy(struct super_block *sb, | |||
967 | ext4_error(sb, __FUNCTION__, | 985 | ext4_error(sb, __FUNCTION__, |
968 | "EXT4-fs: group %lu: %u blocks in bitmap, %u in gd\n", | 986 | "EXT4-fs: group %lu: %u blocks in bitmap, %u in gd\n", |
969 | group, free, grp->bb_free); | 987 | group, free, grp->bb_free); |
988 | /* | ||
989 | * If we intent to continue, we consider group descritor | ||
990 | * corrupt and update bb_free using bitmap value | ||
991 | */ | ||
970 | grp->bb_free = free; | 992 | grp->bb_free = free; |
971 | } | 993 | } |
972 | 994 | ||
@@ -1778,7 +1800,7 @@ static void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac, | |||
1778 | buddy = mb_find_buddy(e4b, i, &max); | 1800 | buddy = mb_find_buddy(e4b, i, &max); |
1779 | BUG_ON(buddy == NULL); | 1801 | BUG_ON(buddy == NULL); |
1780 | 1802 | ||
1781 | k = ext4_find_next_zero_bit(buddy, max, 0); | 1803 | k = mb_find_next_zero_bit(buddy, max, 0); |
1782 | BUG_ON(k >= max); | 1804 | BUG_ON(k >= max); |
1783 | 1805 | ||
1784 | ac->ac_found++; | 1806 | ac->ac_found++; |
@@ -1818,11 +1840,11 @@ static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, | |||
1818 | i = e4b->bd_info->bb_first_free; | 1840 | i = e4b->bd_info->bb_first_free; |
1819 | 1841 | ||
1820 | while (free && ac->ac_status == AC_STATUS_CONTINUE) { | 1842 | while (free && ac->ac_status == AC_STATUS_CONTINUE) { |
1821 | i = ext4_find_next_zero_bit(bitmap, | 1843 | i = mb_find_next_zero_bit(bitmap, |
1822 | EXT4_BLOCKS_PER_GROUP(sb), i); | 1844 | EXT4_BLOCKS_PER_GROUP(sb), i); |
1823 | if (i >= EXT4_BLOCKS_PER_GROUP(sb)) { | 1845 | if (i >= EXT4_BLOCKS_PER_GROUP(sb)) { |
1824 | /* | 1846 | /* |
1825 | * IF we corrupt the bitmap we won't find any | 1847 | * IF we have corrupt bitmap, we won't find any |
1826 | * free blocks even though group info says we | 1848 | * free blocks even though group info says we |
1827 | * we have free blocks | 1849 | * we have free blocks |
1828 | */ | 1850 | */ |
@@ -1838,6 +1860,12 @@ static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac, | |||
1838 | ext4_error(sb, __FUNCTION__, "%d free blocks as per " | 1860 | ext4_error(sb, __FUNCTION__, "%d free blocks as per " |
1839 | "group info. But got %d blocks\n", | 1861 | "group info. But got %d blocks\n", |
1840 | free, ex.fe_len); | 1862 | free, ex.fe_len); |
1863 | /* | ||
1864 | * The number of free blocks differs. This mostly | ||
1865 | * indicate that the bitmap is corrupt. So exit | ||
1866 | * without claiming the space. | ||
1867 | */ | ||
1868 | break; | ||
1841 | } | 1869 | } |
1842 | 1870 | ||
1843 | ext4_mb_measure_extent(ac, &ex, e4b); | 1871 | ext4_mb_measure_extent(ac, &ex, e4b); |
@@ -3740,10 +3768,10 @@ static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b, | |||
3740 | } | 3768 | } |
3741 | 3769 | ||
3742 | while (bit < end) { | 3770 | while (bit < end) { |
3743 | bit = ext4_find_next_zero_bit(bitmap_bh->b_data, end, bit); | 3771 | bit = mb_find_next_zero_bit(bitmap_bh->b_data, end, bit); |
3744 | if (bit >= end) | 3772 | if (bit >= end) |
3745 | break; | 3773 | break; |
3746 | next = ext4_find_next_bit(bitmap_bh->b_data, end, bit); | 3774 | next = mb_find_next_bit(bitmap_bh->b_data, end, bit); |
3747 | if (next > end) | 3775 | if (next > end) |
3748 | next = end; | 3776 | next = end; |
3749 | start = group * EXT4_BLOCKS_PER_GROUP(sb) + bit + | 3777 | start = group * EXT4_BLOCKS_PER_GROUP(sb) + bit + |
@@ -3771,6 +3799,10 @@ static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b, | |||
3771 | (unsigned long) pa->pa_len); | 3799 | (unsigned long) pa->pa_len); |
3772 | ext4_error(sb, __FUNCTION__, "free %u, pa_free %u\n", | 3800 | ext4_error(sb, __FUNCTION__, "free %u, pa_free %u\n", |
3773 | free, pa->pa_free); | 3801 | free, pa->pa_free); |
3802 | /* | ||
3803 | * pa is already deleted so we use the value obtained | ||
3804 | * from the bitmap and continue. | ||
3805 | */ | ||
3774 | } | 3806 | } |
3775 | atomic_add(free, &sbi->s_mb_discarded); | 3807 | atomic_add(free, &sbi->s_mb_discarded); |
3776 | if (ac) | 3808 | if (ac) |
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c index 8c6c685b9d22..5c1e27de7755 100644 --- a/fs/ext4/migrate.c +++ b/fs/ext4/migrate.c | |||
@@ -43,6 +43,7 @@ static int finish_range(handle_t *handle, struct inode *inode, | |||
43 | 43 | ||
44 | if (IS_ERR(path)) { | 44 | if (IS_ERR(path)) { |
45 | retval = PTR_ERR(path); | 45 | retval = PTR_ERR(path); |
46 | path = NULL; | ||
46 | goto err_out; | 47 | goto err_out; |
47 | } | 48 | } |
48 | 49 | ||
@@ -74,6 +75,10 @@ static int finish_range(handle_t *handle, struct inode *inode, | |||
74 | } | 75 | } |
75 | retval = ext4_ext_insert_extent(handle, inode, path, &newext); | 76 | retval = ext4_ext_insert_extent(handle, inode, path, &newext); |
76 | err_out: | 77 | err_out: |
78 | if (path) { | ||
79 | ext4_ext_drop_refs(path); | ||
80 | kfree(path); | ||
81 | } | ||
77 | lb->first_pblock = 0; | 82 | lb->first_pblock = 0; |
78 | return retval; | 83 | return retval; |
79 | } | 84 | } |
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c index a9347fb43bcc..28aa2ed4297e 100644 --- a/fs/ext4/namei.c +++ b/fs/ext4/namei.c | |||
@@ -1804,12 +1804,8 @@ retry: | |||
1804 | inode->i_fop = &ext4_dir_operations; | 1804 | inode->i_fop = &ext4_dir_operations; |
1805 | inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize; | 1805 | inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize; |
1806 | dir_block = ext4_bread (handle, inode, 0, 1, &err); | 1806 | dir_block = ext4_bread (handle, inode, 0, 1, &err); |
1807 | if (!dir_block) { | 1807 | if (!dir_block) |
1808 | ext4_dec_count(handle, inode); /* is this nlink == 0? */ | 1808 | goto out_clear_inode; |
1809 | ext4_mark_inode_dirty(handle, inode); | ||
1810 | iput (inode); | ||
1811 | goto out_stop; | ||
1812 | } | ||
1813 | BUFFER_TRACE(dir_block, "get_write_access"); | 1809 | BUFFER_TRACE(dir_block, "get_write_access"); |
1814 | ext4_journal_get_write_access(handle, dir_block); | 1810 | ext4_journal_get_write_access(handle, dir_block); |
1815 | de = (struct ext4_dir_entry_2 *) dir_block->b_data; | 1811 | de = (struct ext4_dir_entry_2 *) dir_block->b_data; |
@@ -1832,7 +1828,8 @@ retry: | |||
1832 | ext4_mark_inode_dirty(handle, inode); | 1828 | ext4_mark_inode_dirty(handle, inode); |
1833 | err = ext4_add_entry (handle, dentry, inode); | 1829 | err = ext4_add_entry (handle, dentry, inode); |
1834 | if (err) { | 1830 | if (err) { |
1835 | inode->i_nlink = 0; | 1831 | out_clear_inode: |
1832 | clear_nlink(inode); | ||
1836 | ext4_mark_inode_dirty(handle, inode); | 1833 | ext4_mark_inode_dirty(handle, inode); |
1837 | iput (inode); | 1834 | iput (inode); |
1838 | goto out_stop; | 1835 | goto out_stop; |
@@ -2164,7 +2161,7 @@ static int ext4_unlink(struct inode * dir, struct dentry *dentry) | |||
2164 | dir->i_ctime = dir->i_mtime = ext4_current_time(dir); | 2161 | dir->i_ctime = dir->i_mtime = ext4_current_time(dir); |
2165 | ext4_update_dx_flag(dir); | 2162 | ext4_update_dx_flag(dir); |
2166 | ext4_mark_inode_dirty(handle, dir); | 2163 | ext4_mark_inode_dirty(handle, dir); |
2167 | ext4_dec_count(handle, inode); | 2164 | drop_nlink(inode); |
2168 | if (!inode->i_nlink) | 2165 | if (!inode->i_nlink) |
2169 | ext4_orphan_add(handle, inode); | 2166 | ext4_orphan_add(handle, inode); |
2170 | inode->i_ctime = ext4_current_time(inode); | 2167 | inode->i_ctime = ext4_current_time(inode); |
@@ -2214,7 +2211,7 @@ retry: | |||
2214 | err = __page_symlink(inode, symname, l, | 2211 | err = __page_symlink(inode, symname, l, |
2215 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); | 2212 | mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS); |
2216 | if (err) { | 2213 | if (err) { |
2217 | ext4_dec_count(handle, inode); | 2214 | clear_nlink(inode); |
2218 | ext4_mark_inode_dirty(handle, inode); | 2215 | ext4_mark_inode_dirty(handle, inode); |
2219 | iput (inode); | 2216 | iput (inode); |
2220 | goto out_stop; | 2217 | goto out_stop; |
@@ -2223,7 +2220,6 @@ retry: | |||
2223 | inode->i_op = &ext4_fast_symlink_inode_operations; | 2220 | inode->i_op = &ext4_fast_symlink_inode_operations; |
2224 | memcpy((char*)&EXT4_I(inode)->i_data,symname,l); | 2221 | memcpy((char*)&EXT4_I(inode)->i_data,symname,l); |
2225 | inode->i_size = l-1; | 2222 | inode->i_size = l-1; |
2226 | EXT4_I(inode)->i_flags &= ~EXT4_EXTENTS_FL; | ||
2227 | } | 2223 | } |
2228 | EXT4_I(inode)->i_disksize = inode->i_size; | 2224 | EXT4_I(inode)->i_disksize = inode->i_size; |
2229 | err = ext4_add_nondir(handle, dentry, inode); | 2225 | err = ext4_add_nondir(handle, dentry, inode); |
@@ -2407,7 +2403,7 @@ static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry, | |||
2407 | ext4_dec_count(handle, old_dir); | 2403 | ext4_dec_count(handle, old_dir); |
2408 | if (new_inode) { | 2404 | if (new_inode) { |
2409 | /* checked empty_dir above, can't have another parent, | 2405 | /* checked empty_dir above, can't have another parent, |
2410 | * ext3_dec_count() won't work for many-linked dirs */ | 2406 | * ext4_dec_count() won't work for many-linked dirs */ |
2411 | new_inode->i_nlink = 0; | 2407 | new_inode->i_nlink = 0; |
2412 | } else { | 2408 | } else { |
2413 | ext4_inc_count(handle, new_dir); | 2409 | ext4_inc_count(handle, new_dir); |
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c index 9477a2bd6ff2..e29efa0f9d62 100644 --- a/fs/ext4/resize.c +++ b/fs/ext4/resize.c | |||
@@ -1037,6 +1037,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es, | |||
1037 | ext4_warning(sb, __FUNCTION__, | 1037 | ext4_warning(sb, __FUNCTION__, |
1038 | "multiple resizers run on filesystem!"); | 1038 | "multiple resizers run on filesystem!"); |
1039 | unlock_super(sb); | 1039 | unlock_super(sb); |
1040 | ext4_journal_stop(handle); | ||
1040 | err = -EBUSY; | 1041 | err = -EBUSY; |
1041 | goto exit_put; | 1042 | goto exit_put; |
1042 | } | 1043 | } |