pNFS/NFSv4: Improve rejection of out-of-order layouts

If a layoutget ends up being reordered w.r.t. a layoutreturn, e.g. due
to a layoutget-on-open not knowing a priori which file to lock, then we
must assume the layout is no longer being considered valid state by the
server.
Incrementally improve our ability to reject such states by using the
cached old stateid in conjunction with the plh_barrier to try to
identify them.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

1 files changed, 16 insertions, 6 deletions

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index acb63ec00053..af64b4e6fd1f 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -1000,7 +1000,7 @@ pnfs_layout_stateid_blocked(const struct pnfs_layout_hdr *lo,
 {
 	u32 seqid = be32_to_cpu(stateid->seqid);
 
-	return !pnfs_seqid_is_newer(seqid, lo->plh_barrier);
+	return !pnfs_seqid_is_newer(seqid, lo->plh_barrier) && lo->plh_barrier;
 }
 
 /* lget is set to 1 if called from inside send_layoutget call chain */
@@ -1912,6 +1912,11 @@ static void nfs_layoutget_end(struct pnfs_layout_hdr *lo)
 		wake_up_var(&lo->plh_outstanding);
 }
 
+static bool pnfs_is_first_layoutget(struct pnfs_layout_hdr *lo)
+{
+	return test_bit(NFS_LAYOUT_FIRST_LAYOUTGET, &lo->plh_flags);
+}
+
 static void pnfs_clear_first_layoutget(struct pnfs_layout_hdr *lo)
 {
 	unsigned long *bitlock = &lo->plh_flags;
@@ -2386,17 +2391,17 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		goto out_forget;
 	}
 
-	if (!pnfs_layout_is_valid(lo)) {
-		/* We have a completely new layout */
-		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, true);
-	} else if (nfs4_stateid_match_other(&lo->plh_stateid, &res->stateid)) {
+	if (nfs4_stateid_match_other(&lo->plh_stateid, &res->stateid)) {
 		/* existing state ID, make sure the sequence number matches. */
 		if (pnfs_layout_stateid_blocked(lo, &res->stateid)) {
+			if (!pnfs_layout_is_valid(lo) &&
+			    pnfs_is_first_layoutget(lo))
+				lo->plh_barrier = 0;
 			dprintk("%s forget reply due to sequence\n", __func__);
 			goto out_forget;
 		}
 		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, false);
-	} else {
+	} else if (pnfs_layout_is_valid(lo)) {
 		/*
 		 * We got an entirely new state ID.  Mark all segments for the
 		 * inode invalid, and retry the layoutget
@@ -2409,6 +2414,11 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs,
 						&range, 0);
 		goto out_forget;
+	} else {
+		/* We have a completely new layout */
+		if (!pnfs_is_first_layoutget(lo))
+			goto out_forget;
+		pnfs_set_layout_stateid(lo, &res->stateid, lgp->cred, true);
 	}
 
 	pnfs_get_lseg(lseg);