Re: [PATCH v11 06/20] fsverity: don't allow setting DAX file attribute on fsverity files

On 2026-07-10 19:46:43, Eric Biggers wrote:
> On Fri, Jul 10, 2026 at 10:52:34AM +0200, Andrey Albershteyn wrote:
> > When fsverity is enabled on the file, with FS_IOC_ENABLE_VERITY ioctl(),
> > it checks if file has DAX enabled and fails if that's true. However, the
> > opposite case is not checked.
> > 
> > Signed-off-by: Andrey Albershteyn <[email protected]>
> > ---
> >  fs/file_attr.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/fs/file_attr.c b/fs/file_attr.c
> > index bfb00d256dd5..5424ec4e3949 100644
> > --- a/fs/file_attr.c
> > +++ b/fs/file_attr.c
> > @@ -246,6 +246,11 @@ static int fileattr_set_prepare(struct inode *inode,
> >  	if (fa->fsx_cowextsize == 0)
> >  		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
> >  
> > +	/* Can not enable DAX on fsverity file */
> > +	if ((old_ma->fsx_xflags & FS_XFLAG_VERITY) &&
> > +			fa->fsx_xflags & FS_XFLAG_DAX)
> > +		return -EINVAL;
> > +
> 
> This sounds like a bug fix.  But actually ext4 already checks this in
> dax_compatible(), and the other filesystems that already supported
> fsverity (f2fs and btrfs) don't support DAX.  So I guess this is really
> lifting that to common code?  

yes

> But the existing check is still being left
> in place because it's still needed for ext4 specific flags anyway?

The flag mask seems to be used only in dax_compatible(), but I
suppose this's fine to leave verity flag if EXT4_DAX_MUT_EXCL will
be used elsewhere in future.

-- 
- Andrey