pNFS: Handle NFS4ERR_RECALLCONFLICT correctly in LAYOUTGET

Instead of giving up altogether and falling back to doing I/O
through the MDS, which may make the situation worse, wait for
2 lease periods for the callback to resolve itself, and then
try destroying the existing layout.

Only if this was an attempt at getting a first layout, do we
give up altogether, as the server is clearly crazy.

Fixes: 183d9e7b112aa ("pnfs: rework LAYOUTGET retry handling")
Cc: stable@vger.kernel.org # 4.7
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index c50d4ebab5c5b96bfca01161d7e1926cc0ce5b87..7d992362ff04dac8123818e08a5fed33976a72f3 100644 (file)
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -1505,7 +1505,7 @@ pnfs_update_layout(struct inode *ino,
        struct pnfs_layout_segment *lseg = NULL;
        nfs4_stateid stateid;
        long timeout = 0;
-       unsigned long giveup = jiffies + rpc_get_timeout(server->client);
+       unsigned long giveup = jiffies + (clp->cl_lease_time << 1);
        bool first;
 
        if (!pnfs_enabled_sb(NFS_SERVER(ino))) {
@@ -1649,9 +1649,18 @@ lookup_again:
        if (IS_ERR(lseg)) {
                switch(PTR_ERR(lseg)) {
                case -EBUSY:
-               case -ERECALLCONFLICT:
                        if (time_after(jiffies, giveup))
                                lseg = NULL;
+                       break;
+               case -ERECALLCONFLICT:
+                       /* Huh? We hold no layouts, how is there a recall? */
+                       if (first) {
+                               lseg = NULL;
+                               break;
+                       }
+                       /* Destroy the existing layout and start over */
+                       if (time_after(jiffies, giveup))
+                               pnfs_destroy_layout(NFS_I(ino));
                        /* Fallthrough */
                case -EAGAIN:
                        break;