On Thu, 7 Apr 2016 17:11:19 +0000
John Guzik <john at shieldxnetworks.com> wrote:
> With the new jumbo frame patch a latent bug has appeared to have been
> uncovered: the VLAN offload is on the last rx descriptor, when rcd->eop is
> true, not the first rx descriptor, when rcd->sop is true and rcd->eop could
> be false.
> Most likely the reason this has not been seen before is that for non-jumbo
> frames rcd->eop has always been true when rcd->sop is true as well.
>
> The following patch should fix the problem:
>
> diff --git a/drivers/net/vmxnet3/vmxnet3_rxtx.c
> b/drivers/net/vmxnet3/vmxnet3_rxtx.c
> index 4ac0456..3d4a5eb 100644
> --- a/drivers/net/vmxnet3/vmxnet3_rxtx.c
> +++ b/drivers/net/vmxnet3/vmxnet3_rxtx.c
> @@ -587,12 +587,6 @@ vmxnet3_post_rx_bufs(vmxnet3_rx_queue_t *rxq, uint8_t
> ring_id)
> static void
> vmxnet3_rx_offload(const Vmxnet3_RxCompDesc *rcd, struct rte_mbuf *rxm)
> {
> - /* Check for hardware stripped VLAN tag */
> - if (rcd->ts) {
> - rxm->ol_flags |= PKT_RX_VLAN_PKT;
> - rxm->vlan_tci = rte_le_to_cpu_16((uint16_t)rcd->tci);
> - }
> -
> /* Check for RSS */
> if (rcd->rssType != VMXNET3_RCD_RSS_TYPE_NONE) {
> rxm->ol_flags |= PKT_RX_RSS_HASH;
> @@ -737,7 +731,15 @@ vmxnet3_recv_pkts(void *rx_queue, struct rte_mbuf
> **rx_pkts, uint16_t nb_pkts)
> rxq->last_seg = rxm;
>
> if (rcd->eop) {
> - rx_pkts[nb_rx++] = rxq->start_seg;
> + struct rte_mbuf *start = rxq->start_seg;
> +
> + /* Check for hardware stripped VLAN tag */
> + if (rcd->ts) {
> + start->ol_flags |= PKT_RX_VLAN_PKT;
> + start->vlan_tci =
> rte_le_to_cpu_16((uint16_t)rcd->tci);
> + }
> +
> + rx_pkts[nb_rx++] = start;
> rxq->start_seg = NULL;
> }
>
>
Yes, your analysis and bug fix looks correct. The Linux driver checks for VLAN
tag in the
last (eop) segment.
Acked-by: Stephen Hemminger <stephen at networkplumber.org>
