[6/6] net/vhost: perform SW checksum in Tx path
Checks
Commit Message
Virtio specification supports guest checksum offloading
for L4, which is enabled with VIRTIO_NET_F_GUEST_CSUM
feature negotiation. However, the Vhost PMD does not
advertise Tx checksum offload capabilities.
Advertising these offload capabilities at the ethdev level
is not enough, because we could still end-up with the
application enabling these offloads while the guest not
negotiating it.
This patch advertizes the Tx checksum offload capabilities,
and introduces a compatibility layer to cover the case
VIRTIO_NET_F_GUEST_CSUM has not been negotiated but the
application does configure the Tx checksum offloads. This
function performs the L4 Tx checksum in SW for UDP and TCP.
Compared to Rx SW checksum, the Tx SW checksum function
needs to compute the pseudo-header checksum, as we cannot
knwo whether it was done before.
This patch does not advertize SCTP checksum offloading
capability for now, but it could be handled later if the
need arises.
Reported-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
---
drivers/net/vhost/rte_eth_vhost.c | 62 +++++++++++++++++++++++++++++++
1 file changed, 62 insertions(+)
Comments
> -----Original Message-----
> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> Sent: 2022年5月5日 18:27
> To: dev@dpdk.org; jasowang@redhat.com; Xia, Chenbo
> <chenbo.xia@intel.com>; david.marchand@redhat.com;
> olivier.matz@6wind.com
> Cc: stable@dpdk.org; Maxime Coquelin <maxime.coquelin@redhat.com>
> Subject: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>
> Virtio specification supports guest checksum offloading for L4, which is
> enabled with VIRTIO_NET_F_GUEST_CSUM feature negotiation. However,
> the Vhost PMD does not advertise Tx checksum offload capabilities.
>
> Advertising these offload capabilities at the ethdev level is not enough,
> because we could still end-up with the application enabling these offloads
> while the guest not negotiating it.
>
> This patch advertizes the Tx checksum offload capabilities, and introduces a
> compatibility layer to cover the case VIRTIO_NET_F_GUEST_CSUM has not
> been negotiated but the application does configure the Tx checksum
> offloads. This function performs the L4 Tx checksum in SW for UDP and TCP.
> Compared to Rx SW checksum, the Tx SW checksum function needs to
> compute the pseudo-header checksum, as we cannot knwo whether it was
> done before.
>
> This patch does not advertize SCTP checksum offloading capability for now,
> but it could be handled later if the need arises.
In virtio_enqueue_offload(), if RTE_MBUF_F_TX_IP_CKSUM is set, we will
performs the L3 Tx checksum, why do not we advertise IPV4 checksum offloading capability?
Will we advertise it later?
>
> Reported-by: Jason Wang <jasowang@redhat.com>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
> ---
> drivers/net/vhost/rte_eth_vhost.c | 62
> +++++++++++++++++++++++++++++++
> 1 file changed, 62 insertions(+)
>
> diff --git a/drivers/net/vhost/rte_eth_vhost.c
> b/drivers/net/vhost/rte_eth_vhost.c
> index d5303f7368..52a802de05 100644
> --- a/drivers/net/vhost/rte_eth_vhost.c
> +++ b/drivers/net/vhost/rte_eth_vhost.c
> @@ -114,6 +114,7 @@ struct pmd_internal {
> rte_atomic32_t started;
> bool vlan_strip;
> bool rx_sw_csum;
> + bool tx_sw_csum;
> };
>
> struct internal_list {
> @@ -370,8 +371,10 @@ vhost_dev_csum_configure(struct rte_eth_dev
> *eth_dev) {
> struct pmd_internal *internal = eth_dev->data->dev_private;
> const struct rte_eth_rxmode *rxmode = ð_dev->data-
> >dev_conf.rxmode;
> + const struct rte_eth_txmode *txmode = ð_dev->data-
> >dev_conf.txmode;
>
> internal->rx_sw_csum = false;
> + internal->tx_sw_csum = false;
>
> /* SW checksum is not compatible with legacy mode */
> if (!(internal->flags &
> RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS))
> @@ -384,6 +387,56 @@ vhost_dev_csum_configure(struct rte_eth_dev
> *eth_dev)
> internal->rx_sw_csum = true;
> }
> }
> +
> + if (!(internal->features & (1ULL << VIRTIO_NET_F_GUEST_CSUM))) {
> + if (txmode->offloads &
> + (RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
> RTE_ETH_TX_OFFLOAD_TCP_CKSUM)) {
> + VHOST_LOG(NOTICE, "Tx csum will be done in SW,
> may impact performance.");
> + internal->tx_sw_csum = true;
> + }
> + }
> +}
> +
> +static void
> +vhost_dev_tx_sw_csum(struct rte_mbuf *mbuf) {
> + uint32_t hdr_len;
> + uint16_t csum = 0, csum_offset;
> +
> + switch (mbuf->ol_flags & RTE_MBUF_F_TX_L4_MASK) {
> + case RTE_MBUF_F_TX_L4_NO_CKSUM:
> + return;
> + case RTE_MBUF_F_TX_TCP_CKSUM:
> + csum_offset = offsetof(struct rte_tcp_hdr, cksum);
> + break;
> + case RTE_MBUF_F_TX_UDP_CKSUM:
> + csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum);
> + break;
> + default:
> + /* Unsupported packet type. */
> + return;
> + }
> +
> + hdr_len = mbuf->l2_len + mbuf->l3_len;
> + csum_offset += hdr_len;
> +
> + /* Prepare the pseudo-header checksum */
> + if (rte_net_intel_cksum_prepare(mbuf) < 0)
> + return;
> +
> + if (rte_raw_cksum_mbuf(mbuf, hdr_len,
> rte_pktmbuf_pkt_len(mbuf) - hdr_len, &csum) < 0)
> + return;
> +
> + csum = ~csum;
> + /* See RFC768 */
> + if (unlikely((mbuf->packet_type & RTE_PTYPE_L4_UDP) && csum ==
> 0))
> + csum = 0xffff;
> +
> + if (rte_pktmbuf_data_len(mbuf) >= csum_offset + 1)
> + *rte_pktmbuf_mtod_offset(mbuf, uint16_t *, csum_offset)
> = csum;
> +
> + mbuf->ol_flags &= ~RTE_MBUF_F_TX_L4_MASK;
> + mbuf->ol_flags |= RTE_MBUF_F_TX_L4_NO_CKSUM;
> }
>
> static void
> @@ -513,6 +566,10 @@ eth_vhost_tx(void *q, struct rte_mbuf **bufs,
> uint16_t nb_bufs)
> }
> }
>
> + if (r->internal->tx_sw_csum)
> + vhost_dev_tx_sw_csum(m);
> +
> +
> bufs[nb_send] = m;
> ++nb_send;
> }
> @@ -1359,6 +1416,11 @@ eth_dev_info(struct rte_eth_dev *dev,
>
> dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
> RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
> + if (internal->flags & RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS)
> {
> + dev_info->tx_offload_capa |=
> RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
> + RTE_ETH_TX_OFFLOAD_TCP_CKSUM;
> + }
> +
> dev_info->rx_offload_capa = RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
> if (internal->flags & RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS)
> {
> dev_info->rx_offload_capa |=
> RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
> --
> 2.35.1
Hi Wenwu,
Sorry, I missed your review.
On 5/7/22 05:20, Ma, WenwuX wrote:
>
>> -----Original Message-----
>> From: Maxime Coquelin <maxime.coquelin@redhat.com>
>> Sent: 2022年5月5日 18:27
>> To: dev@dpdk.org; jasowang@redhat.com; Xia, Chenbo
>> <chenbo.xia@intel.com>; david.marchand@redhat.com;
>> olivier.matz@6wind.com
>> Cc: stable@dpdk.org; Maxime Coquelin <maxime.coquelin@redhat.com>
>> Subject: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>>
>> Virtio specification supports guest checksum offloading for L4, which is
>> enabled with VIRTIO_NET_F_GUEST_CSUM feature negotiation. However,
>> the Vhost PMD does not advertise Tx checksum offload capabilities.
>>
>> Advertising these offload capabilities at the ethdev level is not enough,
>> because we could still end-up with the application enabling these offloads
>> while the guest not negotiating it.
>>
>> This patch advertizes the Tx checksum offload capabilities, and introduces a
>> compatibility layer to cover the case VIRTIO_NET_F_GUEST_CSUM has not
>> been negotiated but the application does configure the Tx checksum
>> offloads. This function performs the L4 Tx checksum in SW for UDP and TCP.
>> Compared to Rx SW checksum, the Tx SW checksum function needs to
>> compute the pseudo-header checksum, as we cannot knwo whether it was
>> done before.
>>
>> This patch does not advertize SCTP checksum offloading capability for now,
>> but it could be handled later if the need arises.
>
> In virtio_enqueue_offload(), if RTE_MBUF_F_TX_IP_CKSUM is set, we will
> performs the L3 Tx checksum, why do not we advertise IPV4 checksum offloading capability?
> Will we advertise it later?
>
Indeed, we have an IPv4 SW checksum fallback in Vhost library.
We could think about adding the capability, but that's not urgent I
think. Do you have a use-case where it is needed?
Regards,
Maxime
> -----Original Message-----
> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> Sent: 2022年6月2日 17:07
> To: Ma, WenwuX <wenwux.ma@intel.com>; dev@dpdk.org;
> jasowang@redhat.com; Xia, Chenbo <chenbo.xia@intel.com>;
> david.marchand@redhat.com; Matz, Olivier <olivier.matz@6wind.com>
> Cc: stable@dpdk.org
> Subject: Re: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>
> Hi Wenwu,
>
> Sorry, I missed your review.
>
> On 5/7/22 05:20, Ma, WenwuX wrote:
> >
> >> -----Original Message-----
> >> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> >> Sent: 2022年5月5日 18:27
> >> To: dev@dpdk.org; jasowang@redhat.com; Xia, Chenbo
> >> <chenbo.xia@intel.com>; david.marchand@redhat.com;
> >> olivier.matz@6wind.com
> >> Cc: stable@dpdk.org; Maxime Coquelin <maxime.coquelin@redhat.com>
> >> Subject: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
> >>
> >> Virtio specification supports guest checksum offloading for L4, which
> >> is enabled with VIRTIO_NET_F_GUEST_CSUM feature negotiation.
> However,
> >> the Vhost PMD does not advertise Tx checksum offload capabilities.
> >>
> >> Advertising these offload capabilities at the ethdev level is not
> >> enough, because we could still end-up with the application enabling
> >> these offloads while the guest not negotiating it.
> >>
> >> This patch advertizes the Tx checksum offload capabilities, and
> >> introduces a compatibility layer to cover the case
> >> VIRTIO_NET_F_GUEST_CSUM has not been negotiated but the
> application
> >> does configure the Tx checksum offloads. This function performs the L4 Tx
> checksum in SW for UDP and TCP.
> >> Compared to Rx SW checksum, the Tx SW checksum function needs to
> >> compute the pseudo-header checksum, as we cannot knwo whether it
> was
> >> done before.
> >>
> >> This patch does not advertize SCTP checksum offloading capability for
> >> now, but it could be handled later if the need arises.
> >
> > In virtio_enqueue_offload(), if RTE_MBUF_F_TX_IP_CKSUM is set, we will
> > performs the L3 Tx checksum, why do not we advertise IPV4 checksum
> offloading capability?
> > Will we advertise it later?
> >
>
> Indeed, we have an IPv4 SW checksum fallback in Vhost library.
> We could think about adding the capability, but that's not urgent I think. Do
> you have a use-case where it is needed?
>
The GRO/GSO library doesn't re-calculate IPv4 checksums for merged/fragmented packets, it will cause iperf in the vm to fail.
> Regards,
> Maxime
Hi Wenwu,
On 6/6/22 11:44, Ma, WenwuX wrote:
>
>
>> -----Original Message-----
>> From: Maxime Coquelin <maxime.coquelin@redhat.com>
>> Sent: 2022年6月2日 17:07
>> To: Ma, WenwuX <wenwux.ma@intel.com>; dev@dpdk.org;
>> jasowang@redhat.com; Xia, Chenbo <chenbo.xia@intel.com>;
>> david.marchand@redhat.com; Matz, Olivier <olivier.matz@6wind.com>
>> Cc: stable@dpdk.org
>> Subject: Re: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>>
>> Hi Wenwu,
>>
>> Sorry, I missed your review.
>>
>> On 5/7/22 05:20, Ma, WenwuX wrote:
>>>
>>>> -----Original Message-----
>>>> From: Maxime Coquelin <maxime.coquelin@redhat.com>
>>>> Sent: 2022年5月5日 18:27
>>>> To: dev@dpdk.org; jasowang@redhat.com; Xia, Chenbo
>>>> <chenbo.xia@intel.com>; david.marchand@redhat.com;
>>>> olivier.matz@6wind.com
>>>> Cc: stable@dpdk.org; Maxime Coquelin <maxime.coquelin@redhat.com>
>>>> Subject: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>>>>
>>>> Virtio specification supports guest checksum offloading for L4, which
>>>> is enabled with VIRTIO_NET_F_GUEST_CSUM feature negotiation.
>> However,
>>>> the Vhost PMD does not advertise Tx checksum offload capabilities.
>>>>
>>>> Advertising these offload capabilities at the ethdev level is not
>>>> enough, because we could still end-up with the application enabling
>>>> these offloads while the guest not negotiating it.
>>>>
>>>> This patch advertizes the Tx checksum offload capabilities, and
>>>> introduces a compatibility layer to cover the case
>>>> VIRTIO_NET_F_GUEST_CSUM has not been negotiated but the
>> application
>>>> does configure the Tx checksum offloads. This function performs the L4 Tx
>> checksum in SW for UDP and TCP.
>>>> Compared to Rx SW checksum, the Tx SW checksum function needs to
>>>> compute the pseudo-header checksum, as we cannot knwo whether it
>> was
>>>> done before.
>>>>
>>>> This patch does not advertize SCTP checksum offloading capability for
>>>> now, but it could be handled later if the need arises.
>>>
>>> In virtio_enqueue_offload(), if RTE_MBUF_F_TX_IP_CKSUM is set, we will
>>> performs the L3 Tx checksum, why do not we advertise IPV4 checksum
>> offloading capability?
>>> Will we advertise it later?
>>>
>>
>> Indeed, we have an IPv4 SW checksum fallback in Vhost library.
>> We could think about adding the capability, but that's not urgent I think. Do
>> you have a use-case where it is needed?
>>
> The GRO/GSO library doesn't re-calculate IPv4 checksums for merged/fragmented packets, it will cause iperf in the vm to fail.
Can you please elaborate?
If we don't expose the IPv4 checksum availability, it would be done by
the application using the Vhost PMD, so the result will be the same from
the VM point of view. Am I missing something?
Thanks,
Maxime
>> Regards,
>> Maxime
>
> -----Original Message-----
> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> Sent: 2022年6月8日 16:14
> To: Ma, WenwuX <wenwux.ma@intel.com>; dev@dpdk.org;
> jasowang@redhat.com; Xia, Chenbo <chenbo.xia@intel.com>;
> david.marchand@redhat.com; Matz, Olivier <olivier.matz@6wind.com>
> Cc: stable@dpdk.org
> Subject: Re: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
>
> Hi Wenwu,
>
> On 6/6/22 11:44, Ma, WenwuX wrote:
> >
> >
> >> -----Original Message-----
> >> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> >> Sent: 2022年6月2日 17:07
> >> To: Ma, WenwuX <wenwux.ma@intel.com>; dev@dpdk.org;
> >> jasowang@redhat.com; Xia, Chenbo <chenbo.xia@intel.com>;
> >> david.marchand@redhat.com; Matz, Olivier <olivier.matz@6wind.com>
> >> Cc: stable@dpdk.org
> >> Subject: Re: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
> >>
> >> Hi Wenwu,
> >>
> >> Sorry, I missed your review.
> >>
> >> On 5/7/22 05:20, Ma, WenwuX wrote:
> >>>
> >>>> -----Original Message-----
> >>>> From: Maxime Coquelin <maxime.coquelin@redhat.com>
> >>>> Sent: 2022年5月5日 18:27
> >>>> To: dev@dpdk.org; jasowang@redhat.com; Xia, Chenbo
> >>>> <chenbo.xia@intel.com>; david.marchand@redhat.com;
> >>>> olivier.matz@6wind.com
> >>>> Cc: stable@dpdk.org; Maxime Coquelin
> <maxime.coquelin@redhat.com>
> >>>> Subject: [PATCH 6/6] net/vhost: perform SW checksum in Tx path
> >>>>
> >>>> Virtio specification supports guest checksum offloading for L4,
> >>>> which is enabled with VIRTIO_NET_F_GUEST_CSUM feature
> negotiation.
> >> However,
> >>>> the Vhost PMD does not advertise Tx checksum offload capabilities.
> >>>>
> >>>> Advertising these offload capabilities at the ethdev level is not
> >>>> enough, because we could still end-up with the application enabling
> >>>> these offloads while the guest not negotiating it.
> >>>>
> >>>> This patch advertizes the Tx checksum offload capabilities, and
> >>>> introduces a compatibility layer to cover the case
> >>>> VIRTIO_NET_F_GUEST_CSUM has not been negotiated but the
> >> application
> >>>> does configure the Tx checksum offloads. This function performs the
> >>>> L4 Tx
> >> checksum in SW for UDP and TCP.
> >>>> Compared to Rx SW checksum, the Tx SW checksum function needs to
> >>>> compute the pseudo-header checksum, as we cannot knwo whether it
> >> was
> >>>> done before.
> >>>>
> >>>> This patch does not advertize SCTP checksum offloading capability
> >>>> for now, but it could be handled later if the need arises.
> >>>
> >>> In virtio_enqueue_offload(), if RTE_MBUF_F_TX_IP_CKSUM is set, we
> >>> will performs the L3 Tx checksum, why do not we advertise IPV4
> >>> checksum
> >> offloading capability?
> >>> Will we advertise it later?
> >>>
> >>
> >> Indeed, we have an IPv4 SW checksum fallback in Vhost library.
> >> We could think about adding the capability, but that's not urgent I
> >> think. Do you have a use-case where it is needed?
> >>
> > The GRO/GSO library doesn't re-calculate IPv4 checksums for
> merged/fragmented packets, it will cause iperf in the vm to fail.
>
> Can you please elaborate?
>
> If we don't expose the IPv4 checksum availability, it would be done by the
> application using the Vhost PMD, so the result will be the same from the VM
> point of view. Am I missing something?
>
Ok, it is not urgent.
> Thanks,
> Maxime
>
> >> Regards,
> >> Maxime
> >
@@ -114,6 +114,7 @@ struct pmd_internal {
rte_atomic32_t started;
bool vlan_strip;
bool rx_sw_csum;
+ bool tx_sw_csum;
};
struct internal_list {
@@ -370,8 +371,10 @@ vhost_dev_csum_configure(struct rte_eth_dev *eth_dev)
{
struct pmd_internal *internal = eth_dev->data->dev_private;
const struct rte_eth_rxmode *rxmode = ð_dev->data->dev_conf.rxmode;
+ const struct rte_eth_txmode *txmode = ð_dev->data->dev_conf.txmode;
internal->rx_sw_csum = false;
+ internal->tx_sw_csum = false;
/* SW checksum is not compatible with legacy mode */
if (!(internal->flags & RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS))
@@ -384,6 +387,56 @@ vhost_dev_csum_configure(struct rte_eth_dev *eth_dev)
internal->rx_sw_csum = true;
}
}
+
+ if (!(internal->features & (1ULL << VIRTIO_NET_F_GUEST_CSUM))) {
+ if (txmode->offloads &
+ (RTE_ETH_TX_OFFLOAD_UDP_CKSUM | RTE_ETH_TX_OFFLOAD_TCP_CKSUM)) {
+ VHOST_LOG(NOTICE, "Tx csum will be done in SW, may impact performance.");
+ internal->tx_sw_csum = true;
+ }
+ }
+}
+
+static void
+vhost_dev_tx_sw_csum(struct rte_mbuf *mbuf)
+{
+ uint32_t hdr_len;
+ uint16_t csum = 0, csum_offset;
+
+ switch (mbuf->ol_flags & RTE_MBUF_F_TX_L4_MASK) {
+ case RTE_MBUF_F_TX_L4_NO_CKSUM:
+ return;
+ case RTE_MBUF_F_TX_TCP_CKSUM:
+ csum_offset = offsetof(struct rte_tcp_hdr, cksum);
+ break;
+ case RTE_MBUF_F_TX_UDP_CKSUM:
+ csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum);
+ break;
+ default:
+ /* Unsupported packet type. */
+ return;
+ }
+
+ hdr_len = mbuf->l2_len + mbuf->l3_len;
+ csum_offset += hdr_len;
+
+ /* Prepare the pseudo-header checksum */
+ if (rte_net_intel_cksum_prepare(mbuf) < 0)
+ return;
+
+ if (rte_raw_cksum_mbuf(mbuf, hdr_len, rte_pktmbuf_pkt_len(mbuf) - hdr_len, &csum) < 0)
+ return;
+
+ csum = ~csum;
+ /* See RFC768 */
+ if (unlikely((mbuf->packet_type & RTE_PTYPE_L4_UDP) && csum == 0))
+ csum = 0xffff;
+
+ if (rte_pktmbuf_data_len(mbuf) >= csum_offset + 1)
+ *rte_pktmbuf_mtod_offset(mbuf, uint16_t *, csum_offset) = csum;
+
+ mbuf->ol_flags &= ~RTE_MBUF_F_TX_L4_MASK;
+ mbuf->ol_flags |= RTE_MBUF_F_TX_L4_NO_CKSUM;
}
static void
@@ -513,6 +566,10 @@ eth_vhost_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
}
}
+ if (r->internal->tx_sw_csum)
+ vhost_dev_tx_sw_csum(m);
+
+
bufs[nb_send] = m;
++nb_send;
}
@@ -1359,6 +1416,11 @@ eth_dev_info(struct rte_eth_dev *dev,
dev_info->tx_offload_capa = RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
+ if (internal->flags & RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS) {
+ dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
+ RTE_ETH_TX_OFFLOAD_TCP_CKSUM;
+ }
+
dev_info->rx_offload_capa = RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
if (internal->flags & RTE_VHOST_USER_NET_COMPLIANT_OL_FLAGS) {
dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_UDP_CKSUM |