[PATCH 1/2] NFSv4/pNFS: Fix a layoutget livelock loop

[PATCH 1/2] NFSv4/pNFS: Fix a layoutget livelock loop

[trondmy]

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

If NFS_LAYOUT_RETURN_REQUESTED is set, but there is no value set for
the layout plh_return_seq, we can end up in a livelock loop in which
every layout segment retrieved by a new call to layoutget is immediately
invalidated by pnfs_layout_need_return().
To get around this, we should just set plh_return_seq to the current
value of the layout stateid's seqid.

Fixes: d474f96104bd ("NFS: Don't return layout segments that are in use")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 4ed4586bc1a2..51049499e98f 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -347,11 +347,15 @@ pnfs_set_plh_return_info(struct pnfs_layout_hdr *lo, enum pnfs_iomode iomode,
 		iomode = IOMODE_ANY;
 	lo->plh_return_iomode = iomode;
 	set_bit(NFS_LAYOUT_RETURN_REQUESTED, &lo->plh_flags);
-	if (seq != 0) {
-		WARN_ON_ONCE(lo->plh_return_seq != 0 && lo->plh_return_seq != seq);
+	/*
+	 * We must set lo->plh_return_seq to avoid livelocks with
+	 * pnfs_layout_need_return()
+	 */
+	if (seq == 0)
+		seq = be32_to_cpu(lo->plh_stateid.seqid);
+	if (!lo->plh_return_seq || pnfs_seqid_is_newer(seq, lo->plh_return_seq))
 		lo->plh_return_seq = seq;
-		pnfs_barrier_update(lo, seq);
-	}
+	pnfs_barrier_update(lo, seq);
 }
 
 static void
-- 
2.31.1

[PATCH 2/2] NFSv4/pNFS: Remove dead code

[trondmy]

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

Since commit 2b28a7bee453 ("fs, nfs: convert
pnfs_layout_hdr.plh_refcount from atomic_t to refcount_t") it has not
been legal to bump a zero refcount, so the code that tries to allow it
if the NFS_LSEG_VALID flag is still set would cause trouble. Luckily,
NFS_LSEG_VALID has its own refcount so we can never hit this bad code
snippet in practice. Remove it to avoid confusion.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 51049499e98f..7c9090a28e5c 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -596,10 +596,6 @@ pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 	inode = lo->plh_inode;
 
 	if (refcount_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) {
-		if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags)) {
-			spin_unlock(&inode->i_lock);
-			return;
-		}
 		pnfs_get_layout_hdr(lo);
 		pnfs_layout_remove_lseg(lo, lseg);
 		if (pnfs_cache_lseg_for_layoutreturn(lo, lseg))
-- 
2.31.1

[PATCH 2/2] NFSv4/pNFS: Remove dead code

[trondmy]

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

Since commit 2b28a7bee453 ("fs, nfs: convert
pnfs_layout_hdr.plh_refcount from atomic_t to refcount_t") it has not
been legal to bump a zero refcount, so the code that tries to allow it
if the NFS_LSEG_VALID flag is still set would cause trouble. Luckily,
NFS_LSEG_VALID has its own refcount so we can never hit this bad code
snippet in practice. Remove it to avoid confusion.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
---
 fs/nfs/pnfs.c | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 51049499e98f..7c9090a28e5c 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -596,10 +596,6 @@ pnfs_put_lseg(struct pnfs_layout_segment *lseg)
 	inode = lo->plh_inode;
 
 	if (refcount_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) {
-		if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags)) {
-			spin_unlock(&inode->i_lock);
-			return;
-		}
 		pnfs_get_layout_hdr(lo);
 		pnfs_layout_remove_lseg(lo, lseg);
 		if (pnfs_cache_lseg_for_layoutreturn(lo, lseg))
-- 
2.31.1