On Mon 07-08-17 16:12:51, Mimi Zohar wrote: > On Mon, 2017-08-07 at 12:04 +0200, Jan Kara wrote: > > > For DAX, unlike do_blockdev_direct_IO() which takes the lock, reading > > > the file with O_DIRECT is fine, as dax_iomap_rw() only checks that the > > > lock has been taken. Assuming the file system is mounted with > > > i_version, the file hash is updated properly. > > > > Yes, for DAX direct IO is basically no different but frankly I would just > > refuse O_DIRECT on DAX inodes as well just for the consistency sake. > > Ok. So I shouldn't revert the original commit, which fails the > O_DIRECT open for either the buffered read or DAX. I'll just move the > code to a bit later, so that the failure is added to the measurement > list. > > The original commit returned -EACCES. On xfs, the open for direct IO > buffer read fails with -EINVAL. Do you have a preference IMA should > return? Not really. -EINVAL is more traditional when direct IO is not supported but since IMA denies access to the file, -EACCES makes sense as well. Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
