Al Viro <viro@xxxxxxxxxxxxxxxxxx> writes: > On Tue, Jan 03, 2017 at 01:51:36PM +1300, Eric W. Biederman wrote: > >> The only significant thing I see is that you have not taken the >> mount_lock on the path where new_mountpoint adds the new struct >> mountpoint into the mountpoint hash table. > > Umm... Point, but I really don't like that bouncing mount_lock up > and down there. It's not going to cause any serious overhead, > but it just looks ugly... ;-/ > > Let me think for a while... The other possibility is to grab namespace_sem in mntput_no_expire around the call of umount_mnt. That is the only path where put_mountpoint can be called where we are not holding namespace_sem. That works in the small but I haven't traced the callers of mntput and mntput_no_expire yet to see if it works in practice. Eric -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
