On Tue, Mar 17, 2020 at 02:31:02PM +0800, robbieko wrote:
> From: Robbie Ko <robbieko@xxxxxxxxxxxx>
This is not a trivial patch that could go without a changelog.
> Fixes: aab15e8ec2576 ("Btrfs: fix rare chances for data loss when doing a fast fsync")
> Signed-off-by: Robbie Ko <robbieko@xxxxxxxxxxxx>
> ---
> fs/btrfs/file.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
> index a16da274c9aa..ae903da21588 100644
> --- a/fs/btrfs/file.c
> +++ b/fs/btrfs/file.c
> @@ -2124,6 +2124,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
> */
> ret = start_ordered_ops(inode, start, end);
> if (ret) {
> + up_write(&BTRFS_I(inode)->dio_sem);
I did not spot on first sight that there's was missing semaphore unlock
and a few lines below there's down_write(dio_sem). Turns out there are
two calls to start_ordered_ops, one before dio_sem and one inside the
locked section. So I solved the puzzle but I'd prefer not having to and
get a patch with instructions.
