diff mbox

[dpdk-dev] nfp: fix mbuf releasing when segments

Message ID 1493202427-25078-1-git-send-email-alejandro.lucero@netronome.com (mailing list archive)
State Accepted, archived
Delegated to: Ferruh Yigit
Headers

Checks

Context Check Description
ci/checkpatch warning coding style issues
ci/Intel-compilation success Compilation OK

Commit Message

Alejandro Lucero April 26, 2017, 10:27 a.m. UTC 
  If segments are used, just mbufs previously linked to head descriptor of
a mbuf chain are released. Other tx descriptor are used for the mbuf chain
but they keep their linked mbufs without releasing them. It is not a fatal
issue because sooner or later those descriptors will be head descriptors or
just used for a single mbuf packet, then those linked mbufs will be released.
However, this leads to apps needing bigger mbufs pools and some confusion
about memory requeriments. Indeed, because larget pools, some performance
impact could also be exptected due to cache misses.

With this patch all tx descriptors will release linked mbufs inside the
xmit function, and rte_pktmbuf_seg_free is used instead of rte_pktmbuf_free.

Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>
---
 drivers/net/nfp/nfp_net.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

Comments

Ferruh Yigit April 26, 2017, 8:07 p.m. UTC | #1 
On 4/26/2017 11:27 AM, Alejandro Lucero wrote:
> If segments are used, just mbufs previously linked to head descriptor of
> a mbuf chain are released. Other tx descriptor are used for the mbuf chain
> but they keep their linked mbufs without releasing them. It is not a fatal
> issue because sooner or later those descriptors will be head descriptors or
> just used for a single mbuf packet, then those linked mbufs will be released.
> However, this leads to apps needing bigger mbufs pools and some confusion
> about memory requeriments. Indeed, because larget pools, some performance
> impact could also be exptected due to cache misses.
> 
> With this patch all tx descriptors will release linked mbufs inside the
> xmit function, and rte_pktmbuf_seg_free is used instead of rte_pktmbuf_free.
> 
> Signed-off-by: Alejandro Lucero <alejandro.lucero@netronome.com>

Applied to dpdk-next-net/master, thanks.
diff mbox

Patch

diff --git a/drivers/net/nfp/nfp_net.c b/drivers/net/nfp/nfp_net.c
index eda87a5..5c5cba1 100644
--- a/drivers/net/nfp/nfp_net.c
+++ b/drivers/net/nfp/nfp_net.c
@@ -2096,18 +2096,20 @@  uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 		 */
 		pkt_size = pkt->pkt_len;
 
-		/* Releasing mbuf which was prefetched above */
-		if (*lmbuf)
-			rte_pktmbuf_free(*lmbuf);
-		/*
-		 * Linking mbuf with descriptor for being released
-		 * next time descriptor is used
-		 */
-		*lmbuf = pkt;
-
 		while (pkt_size) {
 			/* Copying TSO, VLAN and cksum info */
 			*txds = txd;
+
+			/* Releasing mbuf used by this descriptor previously*/
+			if (*lmbuf)
+				rte_pktmbuf_free_seg(*lmbuf);
+
+			/*
+			 * Linking mbuf with descriptor for being released
+			 * next time descriptor is used
+			 */
+			*lmbuf = pkt;
+
 			dma_size = pkt->data_len;
 			dma_addr = rte_mbuf_data_dma_addr(pkt);
 			PMD_TX_LOG(DEBUG, "Working with mbuf at dma address:"
@@ -2135,6 +2137,7 @@  uint32_t nfp_net_txq_full(struct nfp_net_txq *txq)
 			}
 			/* Referencing next free TX descriptor */
 			txds = &txq->txds[txq->wr_p];
+			lmbuf = &txq->txbufs[txq->wr_p].mbuf;
 			issued_descs++;
 		}
 		i++;