On Fri, Apr 05, 2024 at 09:40:29PM +0000, Günther Noack wrote: > These IOCTL commands should be implemented by setting attributes on the > superblock, rather than in the IOCTL hooks in struct file_operations. > > By returning -ENOTTY instead of -ENOIOCTLCMD, we instruct the fs/ioctl.c > logic to return -ENOTTY immediately, rather than attempting to call > f_op->unlocked_ioctl() or f_op->compat_ioctl() as a fallback. > > Why this is safe: > > Before this change, fs/ioctl.c would unsuccessfully attempt calling the > IOCTL hooks, and then return -ENOTTY. By returning -ENOTTY directly, we > return the same error code immediately, but save ourselves the fallback > attempt. > > Motivation: > > This simplifies the logic for these IOCTL commands and lets us reason about > the side effects of these IOCTLs more easily. It will be possible to > permit these IOCTLs under LSM IOCTL policies, without having to worry about > them getting dispatched to problematic device drivers (which sometimes do > work before looking at the IOCTL command number). > > Link: https://lore.kernel.org/all/cnwpkeovzbumhprco7q2c2y6zxzmxfpwpwe3tyy6c3gg2szgqd@vfzjaw5v5imr/ > Cc: Kent Overstreet <kent.overstreet@xxxxxxxxx> > Cc: Christian Brauner <brauner@xxxxxxxxxx> > Cc: Jan Kara <jack@xxxxxxx> > Cc: Dave Chinner <dchinner@xxxxxxxxxx> > Cc: Darrick J. Wong <djwong@xxxxxxxxxx> > Cc: Theodore Ts'o <tytso@xxxxxxx> > Cc: Josef Bacik <josef@xxxxxxxxxxxxxx> > Signed-off-by: Günther Noack <gnoack@xxxxxxxxxx> Acked-by: Kent Overstreet <kent.overstreet@xxxxxxxxx> > --- > fs/ioctl.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/fs/ioctl.c b/fs/ioctl.c > index 1d5abfdf0f22..fb0628e680c4 100644 > --- a/fs/ioctl.c > +++ b/fs/ioctl.c > @@ -769,7 +769,7 @@ static int ioctl_getfsuuid(struct file *file, void __user *argp) > struct fsuuid2 u = { .len = sb->s_uuid_len, }; > > if (!sb->s_uuid_len) > - return -ENOIOCTLCMD; > + return -ENOTTY; > > memcpy(&u.uuid[0], &sb->s_uuid, sb->s_uuid_len); > > @@ -781,7 +781,7 @@ static int ioctl_get_fs_sysfs_path(struct file *file, void __user *argp) > struct super_block *sb = file_inode(file)->i_sb; > > if (!strlen(sb->s_sysfs_name)) > - return -ENOIOCTLCMD; > + return -ENOTTY; > > struct fs_sysfs_path u = {}; > > -- > 2.44.0.478.gd926399ef9-goog >
