On Sat, 19 Apr 2025 at 12:07, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
> @@ -1987,12 +1988,27 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask,
> obj = inode;
> } else if (obj_type == FSNOTIFY_OBJ_TYPE_VFSMOUNT) {
> obj = path.mnt;
> + user_ns = real_mount(obj)->mnt_ns->user_ns;
> } else if (obj_type == FSNOTIFY_OBJ_TYPE_SB) {
> obj = path.mnt->mnt_sb;
> + user_ns = path.mnt->mnt_sb->s_user_ns;
The patch header notes that user_ns != &init_user_ns implies
FS_USERNS_MOUNT, but it'd be nice to document this with a WARN_ON() in
the code as well.
> } else if (obj_type == FSNOTIFY_OBJ_TYPE_MNTNS) {
> obj = mnt_ns_from_dentry(path.dentry);
> + user_ns = ((struct mnt_namespace *)obj)->user_ns;
It would be much more elegant if the type wasn't lost before this assignment.
Otherwise looks good:
Reviewed-by: Miklos Szeredi <mszeredi@xxxxxxxxxx>
Thanks,
Miklos
