[PATCH 1/4] pNFS/NFSv4: Fix a layout segment leak in pnfs_layout_process()

[PATCH 1/4] pNFS/NFSv4: Fix a layout segment leak in pnfs_layout_process()

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

If the server returns a new stateid that does not match the one in our
cache, then pnfs_layout_process() will leak the layout segments returned
by pnfs_mark_layout_stateid_invalid().

Fixes: 9888d837f3cf ("pNFS: Force a retry of LAYOUTGET if the stateid doesn't match our cache")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 4f274f21c4ab..e68e6f8cb407 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -2417,6 +2417,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 	spin_unlock(&ino->i_lock);
 	lseg->pls_layout = lo;
 	NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg);
+	pnfs_free_lseg_list(&free_me);
 	return ERR_PTR(-EAGAIN);
 }
 
-- 
2.29.2

[PATCH 2/4] pNFS/NFSv4: Try to return invalid layout in pnfs_layout_process()

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

If the server returns a new stateid that does not match the one in our
cache, then try to return the one we hold instead of just invalidating
it on the client side. This ensures that both client and server will
agree that the stateid is invalid.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index e68e6f8cb407..d6262289cf4a 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -2398,7 +2398,13 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 		 * We got an entirely new state ID.  Mark all segments for the
 		 * inode invalid, and retry the layoutget
 		 */
-		pnfs_mark_layout_stateid_invalid(lo, &free_me);
+		struct pnfs_layout_range range = {
+			.iomode = IOMODE_ANY,
+			.length = NFS4_MAX_UINT64,
+		};
+		pnfs_set_plh_return_info(lo, IOMODE_ANY, 0);
+		pnfs_mark_matching_lsegs_return(lo, &lo->plh_return_segs,
+						&range, 0);
 		goto out_forget;
 	}
 
@@ -2417,7 +2423,6 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
 	spin_unlock(&ino->i_lock);
 	lseg->pls_layout = lo;
 	NFS_SERVER(ino)->pnfs_curr_ld->free_lseg(lseg);
-	pnfs_free_lseg_list(&free_me);
 	return ERR_PTR(-EAGAIN);
 }
 
-- 
2.29.2

[PATCH 3/4] pNFS/NFSv4: Update the layout barrier when we schedule a layoutreturn

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

When we're scheduling a layoutreturn, we need to ignore any further
incoming layouts with sequence ids that are going to be affected by the
layout return.

Fixes: 44ea8dfce021 ("NFS/pnfs: Reference the layout cred in pnfs_prepare_layoutreturn()")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 39 +++++++++++++++++++++------------------
 1 file changed, 21 insertions(+), 18 deletions(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index d6262289cf4a..acb63ec00053 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -324,6 +324,21 @@ pnfs_grab_inode_layout_hdr(struct pnfs_layout_hdr *lo)
 	return NULL;
 }
 
+/*
+ * Compare 2 layout stateid sequence ids, to see which is newer,
+ * taking into account wraparound issues.
+ */
+static bool pnfs_seqid_is_newer(u32 s1, u32 s2)
+{
+	return (s32)(s1 - s2) > 0;
+}
+
+static void pnfs_barrier_update(struct pnfs_layout_hdr *lo, u32 newseq)
+{
+	if (pnfs_seqid_is_newer(newseq, lo->plh_barrier))
+		lo->plh_barrier = newseq;
+}
+
 static void
 pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 			 u32 seq)
@@ -335,6 +350,7 @@ pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 	if (seq != 0) {
 		WARN_ON_ONCE(lo->plh_return_seq != 0 && lo->plh_return_seq != seq);
 		lo->plh_return_seq = seq;
+		pnfs_barrier_update(lo, seq);
 	}
 }
 
@@ -639,15 +655,6 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg,
 	return rv;
 }
 
-/*
- * Compare 2 layout stateid sequence ids, to see which is newer,
- * taking into account wraparound issues.
- */
-static bool pnfs_seqid_is_newer(u32 s1, u32 s2)
-{
-	return (s32)(s1 - s2) > 0;
-}
-
 static bool
 pnfs_should_free_range(const struct pnfs_layout_range *lseg_range,
 		 const struct pnfs_layout_range *recall_range)
@@ -984,8 +991,7 @@ pnfs_set_layout_stateid(struct pnfs_layout_hdr *lo, const nfs4_stateid *new,
 		new_barrier = be32_to_cpu(new->seqid);
 	else if (new_barrier == 0)
 		return;
-	if (pnfs_seqid_is_newer(new_barrier, lo->plh_barrier))
-		lo->plh_barrier = new_barrier;
+	pnfs_barrier_update(lo, new_barrier);
 }
 
 static bool
@@ -1183,20 +1189,17 @@ pnfs_prepare_layoutreturn(struct pnfs_layout_hdr *lo,
 		return false;
 	set_bit(NFS_LAYOUT_RETURN, &lo->plh_flags);
 	pnfs_get_layout_hdr(lo);
+	nfs4_stateid_copy(stateid, &lo->plh_stateid);
+	*cred = get_cred(lo->plh_lc_cred);
 	if (test_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags)) {
-		nfs4_stateid_copy(stateid, &lo->plh_stateid);
-		*cred = get_cred(lo->plh_lc_cred);
 		if (lo->plh_return_seq != 0)
 			stateid->seqid = cpu_to_be32(lo->plh_return_seq);
 		if (iomode != NULL)
 			*iomode = lo->plh_return_iomode;
 		pnfs_clear_layoutreturn_info(lo);
-		return true;
-	}
-	nfs4_stateid_copy(stateid, &lo->plh_stateid);
-	*cred = get_cred(lo->plh_lc_cred);
-	if (iomode != NULL)
+	} else if (iomode != NULL)
 		*iomode = IOMODE_ANY;
+	pnfs_barrier_update(lo, be32_to_cpu(stateid->seqid));
 	return true;
 }
 
-- 
2.29.2

[PATCH 4/4] pNFS/NFSv4: Improve rejection of out-of-order layouts

[trondmy]

From: Trond Myklebust <trond.myklebust@hammerspace.com>

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>
---
 fs/nfs/pnfs.c | 22 ++++++++++++++++------
 1 file 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);
-- 
2.29.2