]> git.karo-electronics.de Git - linux-beck.git/commit
revalidate directories instiantiated via FIND_* in order to handle DFS referrals
authorJeff Layton <jlayton@redhat.com>
Tue, 25 Jun 2013 06:32:17 +0000 (01:32 -0500)
committerSteve French <smfrench@gmail.com>
Wed, 26 Jun 2013 22:31:41 +0000 (17:31 -0500)
commit2a2c41c07c710f2c1afe3748bdde40db9ea9d9e6
treeb3aefdf4ddae055c48e56f0f75a8fbdeb8847dcf
parent4a72dafa19ba77a2fb77ae676f8e3a0d6077c37c
revalidate directories instiantiated via FIND_* in order to handle DFS referrals

We've had a long-standing problem with DFS referral points. CIFS servers
generally try to make them look like directories in FIND_FIRST/NEXT
responses. When you go to try to do a FIND_FIRST on them though, the
server will then (correctly) return STATUS_PATH_NOT_COVERED. Mostly this
manifests as spurious EREMOTE errors back to userland.

This patch attempts to fix this by marking directories that are
discovered via FIND_FIRST/NEXT for revaldiation. When the lookup code
runs across them again, we'll reissue a QPathInfo against them and that
will make it chase the referral properly.

There is some performance penalty involved here and no I haven't
measured it -- it'll be highly dependent upon the workload and contents
of the mounted share. To try and mitigate that though, the code only
marks the inode for revalidation when it's possible to run across a DFS
referral. i.e.: when the kernel has DFS support built in and the share
is "in DFS"

[At the Microsoft plugfest we noted that usually the DFS links had
the REPARSE attribute tag enabled - DFS junctions are reparse points
after all - so I just added a check for that flag too so the
performance impact should be smaller - Steve]

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: Sachin Prabhu <sprabhu@redhat.com>
Signed-off-by: Steve French <smfrench@gmail.com>
fs/cifs/readdir.c