On 04/10/2024 13:35, Christoph Hellwig wrote:
On Fri, Oct 04, 2024 at 09:22:51AM +0000, John Garry wrote:
Add initial support for new flag FS_XFLAG_ATOMICWRITES.
This flag is a file attribute that mirrors an ondisk inode flag. Actual
support for untorn file writes (for now) depends on both the iflag and the
underlying storage devices, which we can only really check at statx and
pwritev2() time. This is the same story as FS_XFLAG_DAX, which signals to
the fs that we should try to enable the fsdax IO path on the file (instead
of the regular page cache), but applications have to query STAT_ATTR_DAX to
find out if they really got that IO path.
Current kernel support for atomic writes is based on HW support (for atomic
writes). Since for regular files XFS has no way to specify extent alignment
or granularity, atomic write size is limited to the FS block size.
I'm still confused why this flag is needed for the current version
of this patch set.
We should always be able to support atomic writes
<= block size if support by the block device.
Sure, that is true (about being able to atomically write 1x FS block if
the bdev support it).
But if we are going to add forcealign or similar later, then it would
make sense (to me) to have FS_XFLAG_ATOMICWRITES (and its other flags)
from the beginning. I mean, for example, if FS_XFLAG_FORCEALIGN were
enabled and we want atomic writes, setting FS_XFLAG_ATOMICWRITES would
be rejected if AG count is not aligned with extsize, or extsize is not a
power-of-2, or extsize exceeds bdev limits. So FS_XFLAG_ATOMICWRITES
could have some value there.
As such, it makes sense to have a consistent user experience and require
FS_XFLAG_ATOMICWRITES from the beginning.
Cheers,
John
