diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2010-07-19 22:33:02 -0400 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2010-07-19 22:33:02 -0400 |
commit | ee1039307a8a64b038f9b8cdc6f9120ecd9dfe9b (patch) | |
tree | 05d4bb91752fdb1652a93bcd2ad4514cc697c2df /fs/btrfs | |
parent | d0c6f6258478e1dba532bf7c28e2cd6e1047d3a4 (diff) | |
parent | 2ebc3464781ad24474abcbd2274e6254689853b5 (diff) |
Merge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable
* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
Btrfs: fix checks in BTRFS_IOC_CLONE_RANGE
Btrfs: fix CLONE ioctl destination file size expansion to block boundary
Btrfs: fix split_leaf double split corner case
Diffstat (limited to 'fs/btrfs')
-rw-r--r-- | fs/btrfs/ctree.c | 129 | ||||
-rw-r--r-- | fs/btrfs/ioctl.c | 20 |
2 files changed, 126 insertions, 23 deletions
diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c index 0d1d966b0fe4..c3df14ce2cc2 100644 --- a/fs/btrfs/ctree.c +++ b/fs/btrfs/ctree.c | |||
@@ -2304,12 +2304,17 @@ noinline int btrfs_leaf_free_space(struct btrfs_root *root, | |||
2304 | return ret; | 2304 | return ret; |
2305 | } | 2305 | } |
2306 | 2306 | ||
2307 | /* | ||
2308 | * min slot controls the lowest index we're willing to push to the | ||
2309 | * right. We'll push up to and including min_slot, but no lower | ||
2310 | */ | ||
2307 | static noinline int __push_leaf_right(struct btrfs_trans_handle *trans, | 2311 | static noinline int __push_leaf_right(struct btrfs_trans_handle *trans, |
2308 | struct btrfs_root *root, | 2312 | struct btrfs_root *root, |
2309 | struct btrfs_path *path, | 2313 | struct btrfs_path *path, |
2310 | int data_size, int empty, | 2314 | int data_size, int empty, |
2311 | struct extent_buffer *right, | 2315 | struct extent_buffer *right, |
2312 | int free_space, u32 left_nritems) | 2316 | int free_space, u32 left_nritems, |
2317 | u32 min_slot) | ||
2313 | { | 2318 | { |
2314 | struct extent_buffer *left = path->nodes[0]; | 2319 | struct extent_buffer *left = path->nodes[0]; |
2315 | struct extent_buffer *upper = path->nodes[1]; | 2320 | struct extent_buffer *upper = path->nodes[1]; |
@@ -2327,7 +2332,7 @@ static noinline int __push_leaf_right(struct btrfs_trans_handle *trans, | |||
2327 | if (empty) | 2332 | if (empty) |
2328 | nr = 0; | 2333 | nr = 0; |
2329 | else | 2334 | else |
2330 | nr = 1; | 2335 | nr = max_t(u32, 1, min_slot); |
2331 | 2336 | ||
2332 | if (path->slots[0] >= left_nritems) | 2337 | if (path->slots[0] >= left_nritems) |
2333 | push_space += data_size; | 2338 | push_space += data_size; |
@@ -2469,10 +2474,14 @@ out_unlock: | |||
2469 | * | 2474 | * |
2470 | * returns 1 if the push failed because the other node didn't have enough | 2475 | * returns 1 if the push failed because the other node didn't have enough |
2471 | * room, 0 if everything worked out and < 0 if there were major errors. | 2476 | * room, 0 if everything worked out and < 0 if there were major errors. |
2477 | * | ||
2478 | * this will push starting from min_slot to the end of the leaf. It won't | ||
2479 | * push any slot lower than min_slot | ||
2472 | */ | 2480 | */ |
2473 | static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root | 2481 | static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root |
2474 | *root, struct btrfs_path *path, int data_size, | 2482 | *root, struct btrfs_path *path, |
2475 | int empty) | 2483 | int min_data_size, int data_size, |
2484 | int empty, u32 min_slot) | ||
2476 | { | 2485 | { |
2477 | struct extent_buffer *left = path->nodes[0]; | 2486 | struct extent_buffer *left = path->nodes[0]; |
2478 | struct extent_buffer *right; | 2487 | struct extent_buffer *right; |
@@ -2514,8 +2523,8 @@ static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root | |||
2514 | if (left_nritems == 0) | 2523 | if (left_nritems == 0) |
2515 | goto out_unlock; | 2524 | goto out_unlock; |
2516 | 2525 | ||
2517 | return __push_leaf_right(trans, root, path, data_size, empty, | 2526 | return __push_leaf_right(trans, root, path, min_data_size, empty, |
2518 | right, free_space, left_nritems); | 2527 | right, free_space, left_nritems, min_slot); |
2519 | out_unlock: | 2528 | out_unlock: |
2520 | btrfs_tree_unlock(right); | 2529 | btrfs_tree_unlock(right); |
2521 | free_extent_buffer(right); | 2530 | free_extent_buffer(right); |
@@ -2525,12 +2534,17 @@ out_unlock: | |||
2525 | /* | 2534 | /* |
2526 | * push some data in the path leaf to the left, trying to free up at | 2535 | * push some data in the path leaf to the left, trying to free up at |
2527 | * least data_size bytes. returns zero if the push worked, nonzero otherwise | 2536 | * least data_size bytes. returns zero if the push worked, nonzero otherwise |
2537 | * | ||
2538 | * max_slot can put a limit on how far into the leaf we'll push items. The | ||
2539 | * item at 'max_slot' won't be touched. Use (u32)-1 to make us do all the | ||
2540 | * items | ||
2528 | */ | 2541 | */ |
2529 | static noinline int __push_leaf_left(struct btrfs_trans_handle *trans, | 2542 | static noinline int __push_leaf_left(struct btrfs_trans_handle *trans, |
2530 | struct btrfs_root *root, | 2543 | struct btrfs_root *root, |
2531 | struct btrfs_path *path, int data_size, | 2544 | struct btrfs_path *path, int data_size, |
2532 | int empty, struct extent_buffer *left, | 2545 | int empty, struct extent_buffer *left, |
2533 | int free_space, int right_nritems) | 2546 | int free_space, u32 right_nritems, |
2547 | u32 max_slot) | ||
2534 | { | 2548 | { |
2535 | struct btrfs_disk_key disk_key; | 2549 | struct btrfs_disk_key disk_key; |
2536 | struct extent_buffer *right = path->nodes[0]; | 2550 | struct extent_buffer *right = path->nodes[0]; |
@@ -2549,9 +2563,9 @@ static noinline int __push_leaf_left(struct btrfs_trans_handle *trans, | |||
2549 | slot = path->slots[1]; | 2563 | slot = path->slots[1]; |
2550 | 2564 | ||
2551 | if (empty) | 2565 | if (empty) |
2552 | nr = right_nritems; | 2566 | nr = min(right_nritems, max_slot); |
2553 | else | 2567 | else |
2554 | nr = right_nritems - 1; | 2568 | nr = min(right_nritems - 1, max_slot); |
2555 | 2569 | ||
2556 | for (i = 0; i < nr; i++) { | 2570 | for (i = 0; i < nr; i++) { |
2557 | item = btrfs_item_nr(right, i); | 2571 | item = btrfs_item_nr(right, i); |
@@ -2712,10 +2726,14 @@ out: | |||
2712 | /* | 2726 | /* |
2713 | * push some data in the path leaf to the left, trying to free up at | 2727 | * push some data in the path leaf to the left, trying to free up at |
2714 | * least data_size bytes. returns zero if the push worked, nonzero otherwise | 2728 | * least data_size bytes. returns zero if the push worked, nonzero otherwise |
2729 | * | ||
2730 | * max_slot can put a limit on how far into the leaf we'll push items. The | ||
2731 | * item at 'max_slot' won't be touched. Use (u32)-1 to make us push all the | ||
2732 | * items | ||
2715 | */ | 2733 | */ |
2716 | static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root | 2734 | static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root |
2717 | *root, struct btrfs_path *path, int data_size, | 2735 | *root, struct btrfs_path *path, int min_data_size, |
2718 | int empty) | 2736 | int data_size, int empty, u32 max_slot) |
2719 | { | 2737 | { |
2720 | struct extent_buffer *right = path->nodes[0]; | 2738 | struct extent_buffer *right = path->nodes[0]; |
2721 | struct extent_buffer *left; | 2739 | struct extent_buffer *left; |
@@ -2761,8 +2779,9 @@ static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root | |||
2761 | goto out; | 2779 | goto out; |
2762 | } | 2780 | } |
2763 | 2781 | ||
2764 | return __push_leaf_left(trans, root, path, data_size, | 2782 | return __push_leaf_left(trans, root, path, min_data_size, |
2765 | empty, left, free_space, right_nritems); | 2783 | empty, left, free_space, right_nritems, |
2784 | max_slot); | ||
2766 | out: | 2785 | out: |
2767 | btrfs_tree_unlock(left); | 2786 | btrfs_tree_unlock(left); |
2768 | free_extent_buffer(left); | 2787 | free_extent_buffer(left); |
@@ -2855,6 +2874,64 @@ static noinline int copy_for_split(struct btrfs_trans_handle *trans, | |||
2855 | } | 2874 | } |
2856 | 2875 | ||
2857 | /* | 2876 | /* |
2877 | * double splits happen when we need to insert a big item in the middle | ||
2878 | * of a leaf. A double split can leave us with 3 mostly empty leaves: | ||
2879 | * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ] | ||
2880 | * A B C | ||
2881 | * | ||
2882 | * We avoid this by trying to push the items on either side of our target | ||
2883 | * into the adjacent leaves. If all goes well we can avoid the double split | ||
2884 | * completely. | ||
2885 | */ | ||
2886 | static noinline int push_for_double_split(struct btrfs_trans_handle *trans, | ||
2887 | struct btrfs_root *root, | ||
2888 | struct btrfs_path *path, | ||
2889 | int data_size) | ||
2890 | { | ||
2891 | int ret; | ||
2892 | int progress = 0; | ||
2893 | int slot; | ||
2894 | u32 nritems; | ||
2895 | |||
2896 | slot = path->slots[0]; | ||
2897 | |||
2898 | /* | ||
2899 | * try to push all the items after our slot into the | ||
2900 | * right leaf | ||
2901 | */ | ||
2902 | ret = push_leaf_right(trans, root, path, 1, data_size, 0, slot); | ||
2903 | if (ret < 0) | ||
2904 | return ret; | ||
2905 | |||
2906 | if (ret == 0) | ||
2907 | progress++; | ||
2908 | |||
2909 | nritems = btrfs_header_nritems(path->nodes[0]); | ||
2910 | /* | ||
2911 | * our goal is to get our slot at the start or end of a leaf. If | ||
2912 | * we've done so we're done | ||
2913 | */ | ||
2914 | if (path->slots[0] == 0 || path->slots[0] == nritems) | ||
2915 | return 0; | ||
2916 | |||
2917 | if (btrfs_leaf_free_space(root, path->nodes[0]) >= data_size) | ||
2918 | return 0; | ||
2919 | |||
2920 | /* try to push all the items before our slot into the next leaf */ | ||
2921 | slot = path->slots[0]; | ||
2922 | ret = push_leaf_left(trans, root, path, 1, data_size, 0, slot); | ||
2923 | if (ret < 0) | ||
2924 | return ret; | ||
2925 | |||
2926 | if (ret == 0) | ||
2927 | progress++; | ||
2928 | |||
2929 | if (progress) | ||
2930 | return 0; | ||
2931 | return 1; | ||
2932 | } | ||
2933 | |||
2934 | /* | ||
2858 | * split the path's leaf in two, making sure there is at least data_size | 2935 | * split the path's leaf in two, making sure there is at least data_size |
2859 | * available for the resulting leaf level of the path. | 2936 | * available for the resulting leaf level of the path. |
2860 | * | 2937 | * |
@@ -2876,6 +2953,7 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans, | |||
2876 | int wret; | 2953 | int wret; |
2877 | int split; | 2954 | int split; |
2878 | int num_doubles = 0; | 2955 | int num_doubles = 0; |
2956 | int tried_avoid_double = 0; | ||
2879 | 2957 | ||
2880 | l = path->nodes[0]; | 2958 | l = path->nodes[0]; |
2881 | slot = path->slots[0]; | 2959 | slot = path->slots[0]; |
@@ -2884,12 +2962,14 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans, | |||
2884 | return -EOVERFLOW; | 2962 | return -EOVERFLOW; |
2885 | 2963 | ||
2886 | /* first try to make some room by pushing left and right */ | 2964 | /* first try to make some room by pushing left and right */ |
2887 | if (data_size && ins_key->type != BTRFS_DIR_ITEM_KEY) { | 2965 | if (data_size) { |
2888 | wret = push_leaf_right(trans, root, path, data_size, 0); | 2966 | wret = push_leaf_right(trans, root, path, data_size, |
2967 | data_size, 0, 0); | ||
2889 | if (wret < 0) | 2968 | if (wret < 0) |
2890 | return wret; | 2969 | return wret; |
2891 | if (wret) { | 2970 | if (wret) { |
2892 | wret = push_leaf_left(trans, root, path, data_size, 0); | 2971 | wret = push_leaf_left(trans, root, path, data_size, |
2972 | data_size, 0, (u32)-1); | ||
2893 | if (wret < 0) | 2973 | if (wret < 0) |
2894 | return wret; | 2974 | return wret; |
2895 | } | 2975 | } |
@@ -2923,6 +3003,8 @@ again: | |||
2923 | if (mid != nritems && | 3003 | if (mid != nritems && |
2924 | leaf_space_used(l, mid, nritems - mid) + | 3004 | leaf_space_used(l, mid, nritems - mid) + |
2925 | data_size > BTRFS_LEAF_DATA_SIZE(root)) { | 3005 | data_size > BTRFS_LEAF_DATA_SIZE(root)) { |
3006 | if (data_size && !tried_avoid_double) | ||
3007 | goto push_for_double; | ||
2926 | split = 2; | 3008 | split = 2; |
2927 | } | 3009 | } |
2928 | } | 3010 | } |
@@ -2939,6 +3021,8 @@ again: | |||
2939 | if (mid != nritems && | 3021 | if (mid != nritems && |
2940 | leaf_space_used(l, mid, nritems - mid) + | 3022 | leaf_space_used(l, mid, nritems - mid) + |
2941 | data_size > BTRFS_LEAF_DATA_SIZE(root)) { | 3023 | data_size > BTRFS_LEAF_DATA_SIZE(root)) { |
3024 | if (data_size && !tried_avoid_double) | ||
3025 | goto push_for_double; | ||
2942 | split = 2 ; | 3026 | split = 2 ; |
2943 | } | 3027 | } |
2944 | } | 3028 | } |
@@ -3019,6 +3103,13 @@ again: | |||
3019 | } | 3103 | } |
3020 | 3104 | ||
3021 | return ret; | 3105 | return ret; |
3106 | |||
3107 | push_for_double: | ||
3108 | push_for_double_split(trans, root, path, data_size); | ||
3109 | tried_avoid_double = 1; | ||
3110 | if (btrfs_leaf_free_space(root, path->nodes[0]) >= data_size) | ||
3111 | return 0; | ||
3112 | goto again; | ||
3022 | } | 3113 | } |
3023 | 3114 | ||
3024 | static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans, | 3115 | static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans, |
@@ -3915,13 +4006,15 @@ int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root, | |||
3915 | extent_buffer_get(leaf); | 4006 | extent_buffer_get(leaf); |
3916 | 4007 | ||
3917 | btrfs_set_path_blocking(path); | 4008 | btrfs_set_path_blocking(path); |
3918 | wret = push_leaf_left(trans, root, path, 1, 1); | 4009 | wret = push_leaf_left(trans, root, path, 1, 1, |
4010 | 1, (u32)-1); | ||
3919 | if (wret < 0 && wret != -ENOSPC) | 4011 | if (wret < 0 && wret != -ENOSPC) |
3920 | ret = wret; | 4012 | ret = wret; |
3921 | 4013 | ||
3922 | if (path->nodes[0] == leaf && | 4014 | if (path->nodes[0] == leaf && |
3923 | btrfs_header_nritems(leaf)) { | 4015 | btrfs_header_nritems(leaf)) { |
3924 | wret = push_leaf_right(trans, root, path, 1, 1); | 4016 | wret = push_leaf_right(trans, root, path, 1, |
4017 | 1, 1, 0); | ||
3925 | if (wret < 0 && wret != -ENOSPC) | 4018 | if (wret < 0 && wret != -ENOSPC) |
3926 | ret = wret; | 4019 | ret = wret; |
3927 | } | 4020 | } |
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 4dbaf89b1337..9254b3d58dbe 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c | |||
@@ -1458,7 +1458,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |||
1458 | */ | 1458 | */ |
1459 | 1459 | ||
1460 | /* the destination must be opened for writing */ | 1460 | /* the destination must be opened for writing */ |
1461 | if (!(file->f_mode & FMODE_WRITE)) | 1461 | if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) |
1462 | return -EINVAL; | 1462 | return -EINVAL; |
1463 | 1463 | ||
1464 | ret = mnt_want_write(file->f_path.mnt); | 1464 | ret = mnt_want_write(file->f_path.mnt); |
@@ -1511,7 +1511,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |||
1511 | 1511 | ||
1512 | /* determine range to clone */ | 1512 | /* determine range to clone */ |
1513 | ret = -EINVAL; | 1513 | ret = -EINVAL; |
1514 | if (off >= src->i_size || off + len > src->i_size) | 1514 | if (off + len > src->i_size || off + len < off) |
1515 | goto out_unlock; | 1515 | goto out_unlock; |
1516 | if (len == 0) | 1516 | if (len == 0) |
1517 | olen = len = src->i_size - off; | 1517 | olen = len = src->i_size - off; |
@@ -1578,6 +1578,7 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |||
1578 | u64 disko = 0, diskl = 0; | 1578 | u64 disko = 0, diskl = 0; |
1579 | u64 datao = 0, datal = 0; | 1579 | u64 datao = 0, datal = 0; |
1580 | u8 comp; | 1580 | u8 comp; |
1581 | u64 endoff; | ||
1581 | 1582 | ||
1582 | size = btrfs_item_size_nr(leaf, slot); | 1583 | size = btrfs_item_size_nr(leaf, slot); |
1583 | read_extent_buffer(leaf, buf, | 1584 | read_extent_buffer(leaf, buf, |
@@ -1712,9 +1713,18 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd, | |||
1712 | btrfs_release_path(root, path); | 1713 | btrfs_release_path(root, path); |
1713 | 1714 | ||
1714 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; | 1715 | inode->i_mtime = inode->i_ctime = CURRENT_TIME; |
1715 | if (new_key.offset + datal > inode->i_size) | 1716 | |
1716 | btrfs_i_size_write(inode, | 1717 | /* |
1717 | new_key.offset + datal); | 1718 | * we round up to the block size at eof when |
1719 | * determining which extents to clone above, | ||
1720 | * but shouldn't round up the file size | ||
1721 | */ | ||
1722 | endoff = new_key.offset + datal; | ||
1723 | if (endoff > off+olen) | ||
1724 | endoff = off+olen; | ||
1725 | if (endoff > inode->i_size) | ||
1726 | btrfs_i_size_write(inode, endoff); | ||
1727 | |||
1718 | BTRFS_I(inode)->flags = BTRFS_I(src)->flags; | 1728 | BTRFS_I(inode)->flags = BTRFS_I(src)->flags; |
1719 | ret = btrfs_update_inode(trans, root, inode); | 1729 | ret = btrfs_update_inode(trans, root, inode); |
1720 | BUG_ON(ret); | 1730 | BUG_ON(ret); |