Btrfs: correctly set profile flags on seqlock retry
authorFilipe Manana <fdmanana@gmail.com>
Thu, 24 Apr 2014 14:15:29 +0000 (15:15 +0100)
committerChris Mason <clm@fb.com>
Thu, 24 Apr 2014 23:43:33 +0000 (16:43 -0700)
If we had to retry on the profiles seqlock (due to a concurrent write), we
would set bits on the input flags that corresponded both to the current
profile and to previous values of the profile.

Signed-off-by: Filipe David Borba Manana <fdmanana@gmail.com>
Signed-off-by: Chris Mason <clm@fb.com>
fs/btrfs/extent-tree.c

index 678cb352902c5195b12a7327e6cf96d5d6f08276..5590af92094bb67ea61c8ae397cc393b58b75ae6 100644 (file)
@@ -3543,11 +3543,13 @@ static u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
        return extended_to_chunk(flags | tmp);
 }
 
-static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
+static u64 get_alloc_profile(struct btrfs_root *root, u64 orig_flags)
 {
        unsigned seq;
+       u64 flags;
 
        do {
+               flags = orig_flags;
                seq = read_seqbegin(&root->fs_info->profiles_lock);
 
                if (flags & BTRFS_BLOCK_GROUP_DATA)