[PATCH dpdk v3 1/2] hash: avoid leaking entries on RCU defer queue failure

[Robin Jarry]

When rte_rcu_qsbr_dq_enqueue() fails in DQ mode, the deleted key slot
is never freed and becomes permanently leaked. Fall back to synchronous
reclamation instead of only logging an error.

Cc: stable at dpdk.org
Fixes: 769b2de7fb52 ("hash: implement RCU resources reclamation")

Signed-off-by: Robin Jarry <rjarry at redhat.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev at huawei.com>
---
 lib/hash/rte_cuckoo_hash.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/lib/hash/rte_cuckoo_hash.c b/lib/hash/rte_cuckoo_hash.c
index 3f382dd117ce..36b3b477a55b 100644
--- a/lib/hash/rte_cuckoo_hash.c
+++ b/lib/hash/rte_cuckoo_hash.c
@@ -1870,18 +1870,15 @@ __rte_hash_del_key_with_hash(const struct rte_hash *h, const void *key,
 		/* Key index where key is stored, adding the first dummy index */
 		rcu_dq_entry.key_idx = ret + 1;
 		rcu_dq_entry.ext_bkt_idx = index;
-		if (h->dq == NULL) {
+		if (h->dq == NULL || rte_rcu_qsbr_dq_enqueue(h->dq, &rcu_dq_entry) != 0) {
 			/* Wait for quiescent state change if using
-			 * RTE_HASH_QSBR_MODE_SYNC
+			 * RTE_HASH_QSBR_MODE_SYNC or if RCU enqueue failed.
 			 */
 			rte_rcu_qsbr_synchronize(h->hash_rcu_cfg->v,
 						 RTE_QSBR_THRID_INVALID);
 			__hash_rcu_qsbr_free_resource((void *)((uintptr_t)h),
 						      &rcu_dq_entry, 1);
-		} else if (h->dq)
-			/* Push into QSBR FIFO if using RTE_HASH_QSBR_MODE_DQ */
-			if (rte_rcu_qsbr_dq_enqueue(h->dq, &rcu_dq_entry) != 0)
-				HASH_LOG(ERR, "Failed to push QSBR FIFO");
+		}
 	}
 	__hash_rw_writer_unlock(h);
 	return ret;
-- 
2.53.0