Re: [PATCH 08/12] ext4: hold i_rwsem until AIO completes

[Theodore Y. Ts'o]

On Tue, Jan 14, 2020 at 05:12:21PM +0100, Christoph Hellwig wrote:
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 0e8708b77da6..b6aa2d249b30 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -4777,9 +4777,6 @@ static long ext4_zero_range(struct file *file, loff_t offset,
>  	if (mode & FALLOC_FL_KEEP_SIZE)
>  		flags |= EXT4_GET_BLOCKS_KEEP_SIZE;
>  
> -	/* Wait all existing dio workers, newcomers will block on i_mutex */
> -	inode_dio_wait(inode);
> -
>  	/* Preallocate the range including the unaligned edges */
>  	if (partial_begin || partial_end) {
>  		ret = ext4_alloc_file_blocks(file,

I note that you've dropped the inode_dio_wait() in ext4's ZERO_RANGE,
COLLAPSE_RANGE, INSERT_RANGE, etc.  We had added these to avoid
problems when various fallocate operations which modify the inode's
logical->physical block mapping racing with direct I/O (both reads or
writes).

I don't see a replacement protection in this patch series.  How does
are file systems supported to protect against such races?

    	 	 	      	      	      - Ted