diff mbox series

net/af_xdp: fix shared umem fill queue reserve

Message ID 20220311134513.25512-1-ciara.loftus@intel.com (mailing list archive)
State Accepted, archived
Delegated to: Ferruh Yigit
Headers
Series net/af_xdp: fix shared umem fill queue reserve |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK
ci/intel-Testing success Testing PASS
ci/github-robot: build success github build: passed
ci/iol-mellanox-Performance success Performance Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-aarch64-unit-testing success Testing PASS
ci/iol-aarch64-compile-testing success Testing PASS
ci/iol-x86_64-compile-testing success Testing PASS
ci/iol-x86_64-unit-testing success Testing PASS
ci/iol-abi-testing success Testing PASS
ci/iol-broadcom-Performance success Performance Testing PASS
ci/iol-broadcom-Functional success Functional Testing PASS

Commit Message

Loftus, Ciara March 11, 2022, 1:45 p.m. UTC 
  Commit 81fe6720f84f ("net/af_xdp: reserve fill queue before socket create")
moves the fill queue reserve logic to before the creation of the socket in
order to suppress kernel logs like:

XSK buffer pool does not provide enough addresses to fill 2047 buffers on
Rx ring 0

However, for queues that share umem, the fill queue reserve must occur
after the socket creation, because the fill queue is not valid until
that point.

This commit uses the umem refcnt value to determine whether the queue is
sharing a umem, and performs the fill queue reservation either before or
after the socket creation, depending on the refcnt value.

The kernel logs will still be seen for the shared umem queues.

Fixes: 81fe6720f84f ("net/af_xdp: reserve fill queue before socket create")

Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
---
 drivers/net/af_xdp/rte_eth_af_xdp.c | 22 ++++++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)

Comments

Thomas Monjalon March 14, 2022, 9:49 a.m. UTC | #1 
11/03/2022 14:45, Ciara Loftus:
> Commit 81fe6720f84f ("net/af_xdp: reserve fill queue before socket create")
> moves the fill queue reserve logic to before the creation of the socket in
> order to suppress kernel logs like:
> 
> XSK buffer pool does not provide enough addresses to fill 2047 buffers on
> Rx ring 0
> 
> However, for queues that share umem, the fill queue reserve must occur
> after the socket creation, because the fill queue is not valid until
> that point.
> 
> This commit uses the umem refcnt value to determine whether the queue is
> sharing a umem, and performs the fill queue reservation either before or
> after the socket creation, depending on the refcnt value.
> 
> The kernel logs will still be seen for the shared umem queues.
> 
> Fixes: 81fe6720f84f ("net/af_xdp: reserve fill queue before socket create")
> 
> Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>

Applied, thanks.
diff mbox series

Patch

diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c b/drivers/net/af_xdp/rte_eth_af_xdp.c
index 65479138d3..c7e5fb94df 100644
--- a/drivers/net/af_xdp/rte_eth_af_xdp.c
+++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
@@ -1277,11 +1277,13 @@  xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
 	int ret = 0;
 	int reserve_size = ETH_AF_XDP_DFLT_NUM_DESCS;
 	struct rte_mbuf *fq_bufs[reserve_size];
+	bool reserve_before;
 
 	rxq->umem = xdp_umem_configure(internals, rxq);
 	if (rxq->umem == NULL)
 		return -ENOMEM;
 	txq->umem = rxq->umem;
+	reserve_before = __atomic_load_n(&rxq->umem->refcnt, __ATOMIC_ACQUIRE) <= 1;
 
 #if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
 	ret = rte_pktmbuf_alloc_bulk(rxq->umem->mb_pool, fq_bufs, reserve_size);
@@ -1291,10 +1293,13 @@  xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
 	}
 #endif
 
-	ret = reserve_fill_queue(rxq->umem, reserve_size, fq_bufs, &rxq->fq);
-	if (ret) {
-		AF_XDP_LOG(ERR, "Failed to reserve fill queue.\n");
-		goto out_umem;
+	/* reserve fill queue of queues not (yet) sharing UMEM */
+	if (reserve_before) {
+		ret = reserve_fill_queue(rxq->umem, reserve_size, fq_bufs, &rxq->fq);
+		if (ret) {
+			AF_XDP_LOG(ERR, "Failed to reserve fill queue.\n");
+			goto out_umem;
+		}
 	}
 
 	cfg.rx_size = ring_size;
@@ -1335,6 +1340,15 @@  xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
 		goto out_umem;
 	}
 
+	if (!reserve_before) {
+		/* reserve fill queue of queues sharing UMEM */
+		ret = reserve_fill_queue(rxq->umem, reserve_size, fq_bufs, &rxq->fq);
+		if (ret) {
+			AF_XDP_LOG(ERR, "Failed to reserve fill queue.\n");
+			goto out_xsk;
+		}
+	}
+
 	/* insert the xsk into the xsks_map */
 	if (internals->custom_prog_configured) {
 		int err, fd;