cifs: do not share tcons with DFS
authorPaulo Alcantara <pc@cjr.nz>
Mon, 20 Apr 2020 22:42:57 +0000 (19:42 -0300)
committerSteve French <stfrench@microsoft.com>
Thu, 23 Apr 2020 01:22:08 +0000 (20:22 -0500)
This disables tcon re-use for DFS shares.

tcon->dfs_path stores the path that the tcon should connect to when
doing failing over.

If that tcon is used multiple times e.g. 2 mounts using it with
different prefixpath, each will need a different dfs_path but there is
only one tcon. The other solution would be to split the tcon in 2
tcons during failover but that is much harder.

tcons could not be shared with DFS in cifs.ko because in a
DFS namespace like:

          //domain/dfsroot -> /serverA/dfsroot, /serverB/dfsroot

          //serverA/dfsroot/link -> /serverA/target1/aa/bb

          //serverA/dfsroot/link2 -> /serverA/target1/cc/dd

you can see that link and link2 are two DFS links that both resolve to
the same target share (/serverA/target1), so cifs.ko will only contain a
single tcon for both link and link2.

The problem with that is, if we (auto)mount "link" and "link2", cifs.ko
will only contain a single tcon for both DFS links so we couldn't
perform failover or refresh the DFS cache for both links because
tcon->dfs_path was set to either "link" or "link2", but not both --
which is wrong.

Signed-off-by: Paulo Alcantara (SUSE) <pc@cjr.nz>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Reviewed-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
fs/cifs/connect.c

index 63830f2..28268ed 100644 (file)
@@ -3375,6 +3375,10 @@ cifs_find_tcon(struct cifs_ses *ses, struct smb_vol *volume_info)
        spin_lock(&cifs_tcp_ses_lock);
        list_for_each(tmp, &ses->tcon_list) {
                tcon = list_entry(tmp, struct cifs_tcon, tcon_list);
+#ifdef CONFIG_CIFS_DFS_UPCALL
+               if (tcon->dfs_path)
+                       continue;
+#endif
                if (!match_tcon(tcon, volume_info))
                        continue;
                ++tcon->tc_count;