Re: [PATCH v11 2/2] xfs: add FALLOC_FL_WRITE_ZEROES to XFS code base

["Darrick J. Wong" <djwong@kernel.org>]

On Tue, Mar 10, 2026 at 12:22:54PM +0100, Lukas Herbolt wrote:
> On 2026-03-10 11:10, Pankaj Raghav (Samsung) wrote:
> > > 	if (xfs_is_always_cow_inode(ip) ||
> > 
> > Why should we check for this here but we don't do this check in
> > xfs_falloc_zero_range()?
> > 
> > There is a comment in xfs_falloc_allocate_range():
> > 
> > 	/*
> > 	 * If always_cow mode we can't use preallocations and thus should not
> > 	 * create them.
> > 	 */
> > 	if (xfs_is_always_cow_inode(XFS_I(inode)))
> > 		return -EOPNOTSUPP;
> > 
> > Don't we also use preallocations in xfs_falloc_zero_range()?
> > 
> 
> For the xfs_falloc_zero_range we return 0 on COW from xfs_alloc_file_space
> and that
> results in creating a hole in the file instead of having preallocated
> extents as shown
> bellow.
> 
> But for the xfs_falloc_write_zero_range(), we want fail as soon as possible
> as possible.
> 
> Feel free to correct me if I am wrong.

You're correct, write-zero-range wants to actually write zeroes to
storage to avoid remapping overhead during a write.  However, zoned /
alwayscow files always remap so this mode is not supported for them.

--D

> int
> xfs_alloc_file_space(
>         struct xfs_inode        *ip,
>         xfs_off_t               offset,
>         xfs_off_t               len)
> {
>         xfs_mount_t             *mp = ip->i_mount;
>         xfs_off_t               count;
>         xfs_filblks_t           allocatesize_fsb;
>         xfs_extlen_t            extsz, temp;
>         xfs_fileoff_t           startoffset_fsb;
>         xfs_fileoff_t           endoffset_fsb;
>         int                     rt;
>         xfs_trans_t             *tp;
>         xfs_bmbt_irec_t         imaps[1], *imapp;
>         int                     error;
> 
>         if (xfs_is_always_cow_inode(ip))
>                 return 0;
> 
> 
> root@build-00:/mnt# dd if=/dev/urandom of=/mnt/test.bin bs=10M oflag=direct
> count=1
> 1+0 records in
> 1+0 records out
> 10485760 bytes (10 MB, 10 MiB) copied, 0.167452 s, 62.6 MB/s
> root@build-00:/mnt# xfs_bmap -vp  test.bin
> test.bin:
>  EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL FLAGS
>    0: [0..20479]:      192..20671        0 (192..20671)     20480 000000
> root@build-00:/mnt# strace -e fallocate fallocate -zl 2M  test.bin
> fallocate(3, FALLOC_FL_ZERO_RANGE, 0, 2097152) = 0
> +++ exited with 0 +++
> root@build-00:/mnt# xfs_bmap -vp  test.bin
> test.bin:
>  EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL FLAGS
>    0: [0..4095]:       hole                                  4096
>    1: [4096..20479]:   4288..20671       0 (4288..20671)    16384 000000
> root@build-00:/mnt# dd if=/dev/urandom of=/mnt/test.bin bs=1M oflag=direct
> count=1 conv=notrunc
> 1+0 records in
> 1+0 records out
> 1048576 bytes (1.0 MB, 1.0 MiB) copied, 0.0172151 s, 60.9 MB/s
> root@build-00:/mnt# xfs_bmap -vp  test.bin
> test.bin:
>  EXT: FILE-OFFSET      BLOCK-RANGE      AG AG-OFFSET        TOTAL FLAGS
>    0: [0..2047]:       192..2239         0 (192..2239)       2048 000000
>    1: [2048..4095]:    hole                                  2048
>    2: [4096..20479]:   4288..20671       0 (4288..20671)    16384 000000
> root@build-00:/mnt# cat /sys/fs/xfs/debug/
> always_cow          bload_node_slack    larp                mount_delay
> bload_leaf_slack    bug_on_assert       log_recovery_delay  pwork_threads
> root@build-00:/mnt# cat /sys/fs/xfs/debug/always_cow
> 1
> 
> 
> -- 
> -lhe
>