Tested-by: Michael Qiu <michael.qiu at intel.com>
- OS: Fedora20 3.11.10-301
- GCC: gcc version 4.8.3 2014911
- CPU: Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
- NIC: Ethernet controller: Intel Corporation Device 15a4 (rev 01)
- Default x86_64-native-linuxapp-gcc configuration
- Total 5 cases, 5 passed, 0 failed
- Case: Normal frames with no jumbo frame support
- Case: Jumbo frames with no jumbo frame support
- Case: Normal frames with jumbo frame support
- Case: Jumbo frames with jumbo frame support
- Case: Frames bigger than jumbo frames, wwith jumbo frame support
On 5/29/2015 4:11 PM, Chen, Jing D wrote:
> From: "Chen Jing D(Mark)" <jing.d.chen at intel.com>
>
> In TX side, bit FM10K_TXD_FLAG_LAST in TX descriptor only is set
> in the last descriptor for multi-segment packets. But current
> implementation didn't set all the fields of TX descriptor, which
> will cause descriptors processed now to re-use fields set in last
> scroll. If FM10K_TXD_FLAG_LAST bit was set in the last round and
> it happened this is not the last descriptor of a multi-segnment
> packet, HW will send out the incomplete packet out and leads to
> data intergrity issue.
>
> Signed-off-by: Chen Jing D(Mark) <jing.d.chen at intel.com>
> ---
> drivers/net/fm10k/fm10k_rxtx.c | 15 +++++++++++++--
> 1 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/fm10k/fm10k_rxtx.c b/drivers/net/fm10k/fm10k_rxtx.c
> index 56df6cd..f5d1ad0 100644
> --- a/drivers/net/fm10k/fm10k_rxtx.c
> +++ b/drivers/net/fm10k/fm10k_rxtx.c
> @@ -402,9 +402,9 @@ static inline void tx_xmit_pkt(struct fm10k_tx_queue *q,
> struct rte_mbuf *mb)
> q->nb_used = q->nb_used + mb->nb_segs;
> }
>
> - q->hw_ring[last_id].flags = flags;
> q->nb_free -= mb->nb_segs;
>
> + q->hw_ring[q->next_free].flags = 0;
> /* set checksum flags on first descriptor of packet. SCTP checksum
> * offload is not supported, but we do not explicitly check for this
> * case in favor of greatly simplified processing. */
> @@ -415,16 +415,27 @@ static inline void tx_xmit_pkt(struct fm10k_tx_queue
> *q, struct rte_mbuf *mb)
> if (mb->ol_flags & PKT_TX_VLAN_PKT)
> q->hw_ring[q->next_free].vlan = mb->vlan_tci;
>
> + q->sw_ring[q->next_free] = mb;
> + q->hw_ring[q->next_free].buffer_addr =
> + rte_cpu_to_le_64(MBUF_DMA_ADDR(mb));
> + q->hw_ring[q->next_free].buflen =
> + rte_cpu_to_le_16(rte_pktmbuf_data_len(mb));
> + if (++q->next_free == q->nb_desc)
> + q->next_free = 0;
> +
> /* fill up the rings */
> - for (; mb != NULL; mb = mb->next) {
> + for (mb = mb->next; mb != NULL; mb = mb->next) {
> q->sw_ring[q->next_free] = mb;
> q->hw_ring[q->next_free].buffer_addr =
> rte_cpu_to_le_64(MBUF_DMA_ADDR(mb));
> q->hw_ring[q->next_free].buflen =
> rte_cpu_to_le_16(rte_pktmbuf_data_len(mb));
> + q->hw_ring[q->next_free].flags = 0;
> if (++q->next_free == q->nb_desc)
> q->next_free = 0;
> }
> +
> + q->hw_ring[last_id].flags = flags;
> }
>
> uint16_t
