[PATCH v2 2/3] net/gve: fix GQI RDA multi-seg descriptor length and clean

Ryan Wolfe rnwolfe at google.com
Tue Nov 17 15:11:43 CET 2026


gve_tx_burst_ra() has three bugs on multi-segment bursts and ring wrap:

1. For non-TSO packets, hlen was set to tx_pkt->pkt_len instead of
   tx_pkt->data_len and passed as the first descriptor's seg_len,
   causing the device to overrun the head mbuf buffer on multi-segment
   frames while also transmitting the trailing segment descriptors.
2. When txq->nb_free dropped below nb_used mid-burst, gve_tx_burst_ra()
   aborted immediately without calling gve_tx_clean(txq) to reclaim
   completed descriptors.
3. Unsent packets left for the caller on ring-full backpressure were
   added to txq->stats.errors.

Use tx_pkt->data_len for the head segment of non-TSO frames, attempt
gve_tx_clean() before stopping when descriptors run low mid-burst, and
stop counting normal backpressure as Tx errors.

Fixes: a46583cf43c8 ("net/gve: support Rx/Tx")
Cc: stable at dpdk.org

Signed-off-by: Ryan Wolfe <rnwolfe at google.com>
---
 drivers/net/gve/gve_tx.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/net/gve/gve_tx.c b/drivers/net/gve/gve_tx.c
index d344648d0f..14495e3851 100644
--- a/drivers/net/gve/gve_tx.c
+++ b/drivers/net/gve/gve_tx.c
@@ -412,8 +412,11 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		if (ol_flags & RTE_MBUF_F_TX_TCP_SEG)
 			nb_used++;
 
-		if (txq->nb_free < nb_used)
-			goto end_of_tx;
+		if (txq->nb_free < nb_used) {
+			gve_tx_clean(txq);
+			if (txq->nb_free < nb_used)
+				goto end_of_tx;
+		}
 
 		tx_offload.l2_len = tx_pkt->l2_len;
 		tx_offload.l3_len = tx_pkt->l3_len;
@@ -425,7 +428,7 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 
 		hlen = ol_flags & RTE_MBUF_F_TX_TCP_SEG ?
 			(uint32_t)(tx_offload.l2_len + tx_offload.l3_len + tx_offload.l4_len) :
-			tx_pkt->pkt_len;
+			tx_pkt->data_len;
 		/*
 		 * if tso, the driver needs to fill 2 descs for 1 mbuf
 		 * so only put this mbuf into the 1st tx entry in sw ring
@@ -466,7 +469,6 @@ gve_tx_burst_ra(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 
 		txq->stats.packets += nb_tx;
 		txq->stats.bytes += bytes;
-		txq->stats.errors += nb_pkts - nb_tx;
 	}
 
 	return nb_tx;


More information about the dev mailing list