在 2025/2/7 00:43, hch@xxxxxxxxxxxxx 写道:
On Tue, Feb 04, 2025 at 02:00:23PM +1030, Qu Wenruo wrote:
[BUG]
It is a long known bug that VM image on btrfs can lead to data csum
mismatch, if the qemu is using direct-io for the image (this is commonly
known as cache mode none).
[CAUSE]
Inside the VM, if the fs is EXT4 or XFS, or even NTFS from Windows, the
fs is allowed to dirty/modify the folio even the folio is under
writeback (as long as the address space doesn't have AS_STABLE_WRITES
flag inherited from the block device).
Btw, can you add an xfstests that reproduces this by modifying pages
under direct I/O? That would be really helpful to verify the code when
we want to turn back on real direct I/O eventually when the VM is fixed
to prevent the modifications.
Sure, although I'm afraid it will need to be a C program instead.
Thanks,
Qu
