On Tue, 1 Mar 2011 15:48:07 -0500 "J. Bruce Fields" <bfields@xxxxxxxxxxxx> wrote: > On Tue, Mar 01, 2011 at 02:38:27PM -0600, Steve French wrote: > > On Tue, Mar 1, 2011 at 2:34 PM, J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote: > > > On Tue, Mar 01, 2011 at 02:11:32PM -0600, Steve French wrote: > > >> I vaguely remember that the same problem can occur with nfsd on a > > >> local file system > > > > > > Not true; as I've said, exportable local filesystems know how to look up > > > a file on disk by filehandle. > > > > > >> and that nfs clients have to be able to recover from > > >> ESTALE (e.g. lookup/delete/create/lookup would fail with ESTALE > > >> otherwise). > > > > > > It is true, however, that there are some *other* cases when a server can > > > return ESTALE. In those cases (such as the one Peter Staubach discusses > > > in more detail at http://lwn.net/Articles/272684/) there may be some > > > recovery that a client might reasonably attempt. > > > > > > Alas, no client can be expected to recover from ESTALE errors on an > > > opened file--those will just be passed on to the application. > > > > If this is v4 (only) wouldn't we always have an open->dentry in cache? > > That's correct, but only as long as the server is running. We need > reboot recovery to work as well. > > Also: > > - NFSv4 uptake is still rather small as far as I know, so the v3 > behavior is important > - We don't currently have any way to export cifs as v4 only. > Steve, given that the SMB protocol has no lookup by filehandle, I don't see how this can ever reasonably work. It's really not going to be very useful if it's randomly going to fail if the server reboots or inodes get pushed out of the cache. I'd prefer to see this code dropped as I don't ever see it being something that is shippable for distros. Do you agree? -- Jeff Layton <jlayton@xxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-cifs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html