Re: [PATCH] ext4: Free data blocks directly for ordered journal

["Theodore Ts'o" <tytso@mit.edu>]

On Sat, Jun 24, 2023 at 11:52:59PM +0800, Wang Jianjian wrote:
> ---
>  fs/ext4/mballoc.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index 7b2e36d103cb..41fdc2f8c061 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -6206,7 +6206,7 @@ static void ext4_mb_clear_bb(handle_t *handle, struct inode *inode,
>  	 * consistency guarantees.
>  	 */
>  	if (ext4_handle_valid(handle) &&
> -	    ((flags & EXT4_FREE_BLOCKS_METADATA) ||
> +	    ((ext4_should_order_data(inode) && (flags & EXT4_FREE_BLOCKS_METADATA)) ||
>  	     !ext4_should_writeback_data(inode))) {
>  		struct ext4_free_data *new_entry;
>  		/*

This is not a safe thing to do, so I have to reject this patch.  As
stated in the comment immediately above this code:

	/*
	 * We need to make sure we don't reuse the freed block until after the
	 * transaction is committed. We make an exception if the inode is to be
	 * written in writeback mode since writeback mode has weak data
	 * consistency guarantees.
	 */

In ordered mode, we *do* care that if a file is deleted, and then the
block gets reallocated and used for some new file --- and then we
crash before a transaction commits, then the not-actually-deleted file
would have its data blocks corrupted.

							- Ted