On Sat, Jan 15, 2011 at 3:24 AM, Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote: > On Fri, Jan 14, 2011 at 12:19:43PM +1100, Nick Piggin wrote: > >> Also, I don't like how you return a vfsmount * and use that in path walking. If >> you just returned success and allowed the path walk to continue looking up >> mounted directories, it would be cleaner code, IMO. > > No. Filesystem has no business touching mount tree; leave that to core VFS. > Analysis of what's going on is messy enough as it is, let's *NOT* complicate > it even further. > > I think we should take do_add_mount() calls into follow_automount(), making > the API > NULL => nothing to do here, just follow mount > ERR_PTR() => sod off > mnt => here's the thing I want mounted, do it and go there. If > something's allready got mounted here, just disregard this one and follow > mount as usual (== we'd lost the race) That would be fine. I just want the path walk to follow the mount in the usual way, rather than following something that the filesystem hands back directly. David explains that was a micro optimisation to avoid the lookup, and that the vfsmount should always be mounted at that point, which basically solved my conceptual problem with it. -- 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
