On Fri, 2009-04-03 at 02:52 +0100, Al Viro wrote: > On Fri, Apr 03, 2009 at 02:09:54AM +0100, Al Viro wrote: > > On Fri, Apr 03, 2009 at 12:18:59AM +0100, Al Viro wrote: > > > > > Eww... Let's not. First of all, the really useful part of nd->root for > > > normal case is that it allows to get ...->fs->root once. So it's better > > > to *cache* ->fs->root in there. At which point the flag disappears, > > > since it becomes simply nd->root.mnt != NULL. > > > > > > The interesting part is keeping refcounting happy. I'll see how to do > > > that - it's clearly useful on its own. Hopefully will post later tonight... > > > > ... and the ugly part is the check in follow_dotdot() - one about crossing > > out of chroot. Joy... > > OK, that's actually doable. > See git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6.git/ cache-root > for current variant; it's for comments *only* - not for merge. After that > one we should be able to do a modified variant of e.g. vfs_path_lookup() > that would do > get_path(root); > nd->root = root; > instead of > nd->root.mnt = NULL; > and have lookup with root set to given place without touching current->fs > at all. Warning: this is completely untested. Not even "does it boot?", > just a code dump. You have a potential problem in the case of do_filp_open(). It calls do_path_lookup(), but then expects to be able to reuse the same nameidata structure in subsequent calls to __do_follow_link(). Cheers Trond -- Trond Myklebust Linux NFS client maintainer NetApp Trond.Myklebust@xxxxxxxxxx www.netapp.com -- 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
