* [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets [not found] <20210830142655.18373-1-ivan.malov@oktetlabs.ru> @ 2021-09-16 18:49 ` Ivan Malov 2021-10-14 6:45 ` Andrew Rybchenko ` (2 more replies) 0 siblings, 3 replies; 8+ messages in thread From: Ivan Malov @ 2021-09-16 18:49 UTC (permalink / raw) To: dev Cc: Maxime Coquelin, stable, Andrew Rybchenko, Chenbo Xia, Yuanhan Liu, Olivier Matz Tx prepare method calls rte_net_intel_cksum_prepare(), which handles tunnel packets correctly, but Tx burst path does not take tunnel presence into account when computing the offsets. Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") Cc: stable@dpdk.org Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> --- drivers/net/virtio/virtqueue.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h index 03957b2bd0..b83ff32efb 100644 --- a/drivers/net/virtio/virtqueue.h +++ b/drivers/net/virtio/virtqueue.h @@ -620,19 +620,21 @@ static inline void virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) { uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? + cookie->outer_l2_len + cookie->outer_l3_len : 0; if (cookie->ol_flags & PKT_TX_TCP_SEG) csum_l4 |= PKT_TX_TCP_CKSUM; switch (csum_l4) { case PKT_TX_UDP_CKSUM: - hdr->csum_start = cookie->l2_len + cookie->l3_len; + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; break; case PKT_TX_TCP_CKSUM: - hdr->csum_start = cookie->l2_len + cookie->l3_len; + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; break; @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) VIRTIO_NET_HDR_GSO_TCPV6 : VIRTIO_NET_HDR_GSO_TCPV4; hdr->gso_size = cookie->tso_segsz; - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + + cookie->l4_len; } else { ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); -- 2.20.1 ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-09-16 18:49 ` [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets Ivan Malov @ 2021-10-14 6:45 ` Andrew Rybchenko 2021-10-14 7:12 ` Xia, Chenbo 2021-10-21 12:28 ` Maxime Coquelin 2 siblings, 0 replies; 8+ messages in thread From: Andrew Rybchenko @ 2021-10-14 6:45 UTC (permalink / raw) To: Chenbo Xia, Maxime Coquelin Cc: stable, Yuanhan Liu, Olivier Matz, dev, Ivan Malov Hi Maxime, Chenbo, is the any problem with the patch below? Thanks, Andrew. On 9/16/21 9:49 PM, Ivan Malov wrote: > Tx prepare method calls rte_net_intel_cksum_prepare(), which > handles tunnel packets correctly, but Tx burst path does not > take tunnel presence into account when computing the offsets. > > Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") > Cc: stable@dpdk.org > > Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> > Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> > --- > drivers/net/virtio/virtqueue.h | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h > index 03957b2bd0..b83ff32efb 100644 > --- a/drivers/net/virtio/virtqueue.h > +++ b/drivers/net/virtio/virtqueue.h > @@ -620,19 +620,21 @@ static inline void > virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) > { > uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; > + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? > + cookie->outer_l2_len + cookie->outer_l3_len : 0; > > if (cookie->ol_flags & PKT_TX_TCP_SEG) > csum_l4 |= PKT_TX_TCP_CKSUM; > > switch (csum_l4) { > case PKT_TX_UDP_CKSUM: > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > break; > > case PKT_TX_TCP_CKSUM: > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > break; > @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) > VIRTIO_NET_HDR_GSO_TCPV6 : > VIRTIO_NET_HDR_GSO_TCPV4; > hdr->gso_size = cookie->tso_segsz; > - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; > + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + > + cookie->l4_len; > } else { > ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); > ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); > ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-09-16 18:49 ` [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets Ivan Malov 2021-10-14 6:45 ` Andrew Rybchenko @ 2021-10-14 7:12 ` Xia, Chenbo 2021-10-15 8:32 ` Olivier Matz 2021-10-21 12:28 ` Maxime Coquelin 2 siblings, 1 reply; 8+ messages in thread From: Xia, Chenbo @ 2021-10-14 7:12 UTC (permalink / raw) To: Ivan Malov, dev Cc: Maxime Coquelin, stable, Andrew Rybchenko, Yuanhan Liu, Olivier Matz > -----Original Message----- > From: Ivan Malov <ivan.malov@oktetlabs.ru> > Sent: Friday, September 17, 2021 2:50 AM > To: dev@dpdk.org > Cc: Maxime Coquelin <maxime.coquelin@redhat.com>; stable@dpdk.org; Andrew > Rybchenko <andrew.rybchenko@oktetlabs.ru>; Xia, Chenbo <chenbo.xia@intel.com>; > Yuanhan Liu <yuanhan.liu@linux.intel.com>; Olivier Matz > <olivier.matz@6wind.com> > Subject: [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel > packets > > Tx prepare method calls rte_net_intel_cksum_prepare(), which > handles tunnel packets correctly, but Tx burst path does not > take tunnel presence into account when computing the offsets. > > Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") > Cc: stable@dpdk.org > > Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> > Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> > --- > drivers/net/virtio/virtqueue.h | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > > diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h > index 03957b2bd0..b83ff32efb 100644 > --- a/drivers/net/virtio/virtqueue.h > +++ b/drivers/net/virtio/virtqueue.h > @@ -620,19 +620,21 @@ static inline void > virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) > { > uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; > + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? > + cookie->outer_l2_len + cookie->outer_l3_len : 0; > > if (cookie->ol_flags & PKT_TX_TCP_SEG) > csum_l4 |= PKT_TX_TCP_CKSUM; > > switch (csum_l4) { > case PKT_TX_UDP_CKSUM: > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > break; > > case PKT_TX_TCP_CKSUM: > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > break; > @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct > rte_mbuf *cookie) > VIRTIO_NET_HDR_GSO_TCPV6 : > VIRTIO_NET_HDR_GSO_TCPV4; > hdr->gso_size = cookie->tso_segsz; > - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; > + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + > + cookie->l4_len; > } else { > ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); > ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); > -- > 2.20.1 Reviewed-by: Chenbo Xia <chenbo.xia@intel.com> ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-10-14 7:12 ` Xia, Chenbo @ 2021-10-15 8:32 ` Olivier Matz 2021-10-18 7:04 ` Andrew Rybchenko 0 siblings, 1 reply; 8+ messages in thread From: Olivier Matz @ 2021-10-15 8:32 UTC (permalink / raw) To: Xia, Chenbo Cc: Ivan Malov, dev, Maxime Coquelin, stable, Andrew Rybchenko, Yuanhan Liu On Thu, Oct 14, 2021 at 07:12:29AM +0000, Xia, Chenbo wrote: > > -----Original Message----- > > From: Ivan Malov <ivan.malov@oktetlabs.ru> > > Sent: Friday, September 17, 2021 2:50 AM > > To: dev@dpdk.org > > Cc: Maxime Coquelin <maxime.coquelin@redhat.com>; stable@dpdk.org; Andrew > > Rybchenko <andrew.rybchenko@oktetlabs.ru>; Xia, Chenbo <chenbo.xia@intel.com>; > > Yuanhan Liu <yuanhan.liu@linux.intel.com>; Olivier Matz > > <olivier.matz@6wind.com> > > Subject: [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel > > packets > > > > Tx prepare method calls rte_net_intel_cksum_prepare(), which > > handles tunnel packets correctly, but Tx burst path does not > > take tunnel presence into account when computing the offsets. > > > > Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") > > Cc: stable@dpdk.org > > > > Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> > > Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> > > --- > > drivers/net/virtio/virtqueue.h | 9 ++++++--- > > 1 file changed, 6 insertions(+), 3 deletions(-) > > > > diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h > > index 03957b2bd0..b83ff32efb 100644 > > --- a/drivers/net/virtio/virtqueue.h > > +++ b/drivers/net/virtio/virtqueue.h > > @@ -620,19 +620,21 @@ static inline void > > virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) > > { > > uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; > > + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? > > + cookie->outer_l2_len + cookie->outer_l3_len : 0; > > > > if (cookie->ol_flags & PKT_TX_TCP_SEG) > > csum_l4 |= PKT_TX_TCP_CKSUM; > > > > switch (csum_l4) { > > case PKT_TX_UDP_CKSUM: > > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > > hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); > > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > > break; > > > > case PKT_TX_TCP_CKSUM: > > - hdr->csum_start = cookie->l2_len + cookie->l3_len; > > + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > > hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); > > hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > > break; > > @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct > > rte_mbuf *cookie) > > VIRTIO_NET_HDR_GSO_TCPV6 : > > VIRTIO_NET_HDR_GSO_TCPV4; > > hdr->gso_size = cookie->tso_segsz; > > - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; > > + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + > > + cookie->l4_len; > > } else { > > ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); > > ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); > > -- > > 2.20.1 > > Reviewed-by: Chenbo Xia <chenbo.xia@intel.com> > I have one comment to mention that from application perspective, it has to take care that the driver does not support outer tunnel offload (this matches the advertised capabilities). For instance, in case of a vxlan tunnel, if the outer checksum needs to be calculated, it has to be done by the application. In short, the application can ask to offload the inner part if no offload is required on the outer part. Also, since grep "PKT_TX_TUNNEL" in driver/net/ixgbe gives nothing, it seems the ixgbe driver does not support the same offload request than described in this patch: (m->ol_flags & PKT_TX_TUNNEL_MASK) == PKT_TX_TUNNEL_XXXXX m->outer_l2_len = outer l2 length m->outer_l3_len = outer l3 length m->l2_len = outer l4 length + tunnel len + inner l2 len m->l3_len = inner l3 len m->l4_len = inner l4 len An alternative for doing the same (that would work with ixgbe and current virtio) is to give: (m->ol_flags & PKT_TX_TUNNEL_MASK) == 0 m->l2_len = outer lengths + tunnel len + inner l2 len m->l3_len = inner l3 len m->l4_len = inner l4 len I think a capability may be missing to differentiate which drivers support which mode. Or, all drivers could be fixed to support both modes (and this would make this patch valid). Thanks, Olivier ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-10-15 8:32 ` Olivier Matz @ 2021-10-18 7:04 ` Andrew Rybchenko 2021-10-18 7:54 ` Olivier Matz 0 siblings, 1 reply; 8+ messages in thread From: Andrew Rybchenko @ 2021-10-18 7:04 UTC (permalink / raw) To: Olivier Matz, Xia, Chenbo Cc: Ivan Malov, dev, Maxime Coquelin, stable, Yuanhan Liu On 10/15/21 11:32 AM, Olivier Matz wrote: > On Thu, Oct 14, 2021 at 07:12:29AM +0000, Xia, Chenbo wrote: >>> -----Original Message----- >>> From: Ivan Malov <ivan.malov@oktetlabs.ru> >>> Sent: Friday, September 17, 2021 2:50 AM >>> To: dev@dpdk.org >>> Cc: Maxime Coquelin <maxime.coquelin@redhat.com>; stable@dpdk.org; Andrew >>> Rybchenko <andrew.rybchenko@oktetlabs.ru>; Xia, Chenbo <chenbo.xia@intel.com>; >>> Yuanhan Liu <yuanhan.liu@linux.intel.com>; Olivier Matz >>> <olivier.matz@6wind.com> >>> Subject: [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel >>> packets >>> >>> Tx prepare method calls rte_net_intel_cksum_prepare(), which >>> handles tunnel packets correctly, but Tx burst path does not >>> take tunnel presence into account when computing the offsets. >>> >>> Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") >>> Cc: stable@dpdk.org >>> >>> Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> >>> Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> >>> --- >>> drivers/net/virtio/virtqueue.h | 9 ++++++--- >>> 1 file changed, 6 insertions(+), 3 deletions(-) >>> >>> diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h >>> index 03957b2bd0..b83ff32efb 100644 >>> --- a/drivers/net/virtio/virtqueue.h >>> +++ b/drivers/net/virtio/virtqueue.h >>> @@ -620,19 +620,21 @@ static inline void >>> virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) >>> { >>> uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; >>> + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? >>> + cookie->outer_l2_len + cookie->outer_l3_len : 0; >>> >>> if (cookie->ol_flags & PKT_TX_TCP_SEG) >>> csum_l4 |= PKT_TX_TCP_CKSUM; >>> >>> switch (csum_l4) { >>> case PKT_TX_UDP_CKSUM: >>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; >>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; >>> hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); >>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; >>> break; >>> >>> case PKT_TX_TCP_CKSUM: >>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; >>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; >>> hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); >>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; >>> break; >>> @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct >>> rte_mbuf *cookie) >>> VIRTIO_NET_HDR_GSO_TCPV6 : >>> VIRTIO_NET_HDR_GSO_TCPV4; >>> hdr->gso_size = cookie->tso_segsz; >>> - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; >>> + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + >>> + cookie->l4_len; >>> } else { >>> ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); >>> ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); >>> -- >>> 2.20.1 >> >> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com> >> > > I have one comment to mention that from application perspective, it has > to take care that the driver does not support outer tunnel offload (this > matches the advertised capabilities). For instance, in case of a vxlan > tunnel, if the outer checksum needs to be calculated, it has to be done > by the application. In short, the application can ask to offload the > inner part if no offload is required on the outer part. Yes. It is really complicated. Let's name below case (A). > Also, since grep "PKT_TX_TUNNEL" in driver/net/ixgbe gives nothing, it > seems the ixgbe driver does not support the same offload request than > described in this patch: > (m->ol_flags & PKT_TX_TUNNEL_MASK) == PKT_TX_TUNNEL_XXXXX > m->outer_l2_len = outer l2 length > m->outer_l3_len = outer l3 length > m->l2_len = outer l4 length + tunnel len + inner l2 len > m->l3_len = inner l3 len > m->l4_len = inner l4 len True. Let's name below case (B). > An alternative for doing the same (that would work with ixgbe and > current virtio) is to give: > (m->ol_flags & PKT_TX_TUNNEL_MASK) == 0 > m->l2_len = outer lengths + tunnel len + inner l2 len > m->l3_len = inner l3 len > m->l4_len = inner l4 len Possible, but I think it is a bit less friendly. If application supports both cases with and without outer IPv4 checksum offload, it will require more extra steps from the application to remove tunnel mask and collapse outer lengths into l2_len (in addition outer IPv4 checksum calculation in SW). > I think a capability may be missing to differentiate which drivers > support which mode. Yes, something like TX_INNER_CKSUM. Some drivers already support it, some do not. > Or, all drivers could be fixed to support both modes > (and this would make this patch valid). I disagree with the point that something is required to make the patch valid. Before the patch the driver supports just (B). After the patch the driver supports both (A) and (B). It makes the driver to support more cases. Please, note that rte_net_intel_cksum_prepare() supports both (A) and (B) as well. ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-10-18 7:04 ` Andrew Rybchenko @ 2021-10-18 7:54 ` Olivier Matz 2021-10-18 8:20 ` Andrew Rybchenko 0 siblings, 1 reply; 8+ messages in thread From: Olivier Matz @ 2021-10-18 7:54 UTC (permalink / raw) To: Andrew Rybchenko Cc: Xia, Chenbo, Ivan Malov, dev, Maxime Coquelin, stable, Yuanhan Liu On Mon, Oct 18, 2021 at 10:04:46AM +0300, Andrew Rybchenko wrote: > On 10/15/21 11:32 AM, Olivier Matz wrote: > > On Thu, Oct 14, 2021 at 07:12:29AM +0000, Xia, Chenbo wrote: > >>> -----Original Message----- > >>> From: Ivan Malov <ivan.malov@oktetlabs.ru> > >>> Sent: Friday, September 17, 2021 2:50 AM > >>> To: dev@dpdk.org > >>> Cc: Maxime Coquelin <maxime.coquelin@redhat.com>; stable@dpdk.org; Andrew > >>> Rybchenko <andrew.rybchenko@oktetlabs.ru>; Xia, Chenbo <chenbo.xia@intel.com>; > >>> Yuanhan Liu <yuanhan.liu@linux.intel.com>; Olivier Matz > >>> <olivier.matz@6wind.com> > >>> Subject: [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel > >>> packets > >>> > >>> Tx prepare method calls rte_net_intel_cksum_prepare(), which > >>> handles tunnel packets correctly, but Tx burst path does not > >>> take tunnel presence into account when computing the offsets. > >>> > >>> Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") > >>> Cc: stable@dpdk.org > >>> > >>> Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> > >>> Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> > >>> --- > >>> drivers/net/virtio/virtqueue.h | 9 ++++++--- > >>> 1 file changed, 6 insertions(+), 3 deletions(-) > >>> > >>> diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h > >>> index 03957b2bd0..b83ff32efb 100644 > >>> --- a/drivers/net/virtio/virtqueue.h > >>> +++ b/drivers/net/virtio/virtqueue.h > >>> @@ -620,19 +620,21 @@ static inline void > >>> virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) > >>> { > >>> uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; > >>> + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? > >>> + cookie->outer_l2_len + cookie->outer_l3_len : 0; > >>> > >>> if (cookie->ol_flags & PKT_TX_TCP_SEG) > >>> csum_l4 |= PKT_TX_TCP_CKSUM; > >>> > >>> switch (csum_l4) { > >>> case PKT_TX_UDP_CKSUM: > >>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; > >>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > >>> hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); > >>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > >>> break; > >>> > >>> case PKT_TX_TCP_CKSUM: > >>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; > >>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; > >>> hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); > >>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; > >>> break; > >>> @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct > >>> rte_mbuf *cookie) > >>> VIRTIO_NET_HDR_GSO_TCPV6 : > >>> VIRTIO_NET_HDR_GSO_TCPV4; > >>> hdr->gso_size = cookie->tso_segsz; > >>> - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; > >>> + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + > >>> + cookie->l4_len; > >>> } else { > >>> ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); > >>> ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); > >>> -- > >>> 2.20.1 > >> > >> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com> > >> > > > > I have one comment to mention that from application perspective, it has > > to take care that the driver does not support outer tunnel offload (this > > matches the advertised capabilities). For instance, in case of a vxlan > > tunnel, if the outer checksum needs to be calculated, it has to be done > > by the application. In short, the application can ask to offload the > > inner part if no offload is required on the outer part. > > Yes. It is really complicated. > > Let's name below case (A). > > > Also, since grep "PKT_TX_TUNNEL" in driver/net/ixgbe gives nothing, it > > seems the ixgbe driver does not support the same offload request than > > described in this patch: > > (m->ol_flags & PKT_TX_TUNNEL_MASK) == PKT_TX_TUNNEL_XXXXX > > m->outer_l2_len = outer l2 length > > m->outer_l3_len = outer l3 length > > m->l2_len = outer l4 length + tunnel len + inner l2 len > > m->l3_len = inner l3 len > > m->l4_len = inner l4 len > > True. > > Let's name below case (B). > > > An alternative for doing the same (that would work with ixgbe and > > current virtio) is to give: > > (m->ol_flags & PKT_TX_TUNNEL_MASK) == 0 > > m->l2_len = outer lengths + tunnel len + inner l2 len > > m->l3_len = inner l3 len > > m->l4_len = inner l4 len > > Possible, but I think it is a bit less friendly. > If application supports both cases with and without > outer IPv4 checksum offload, it will require more extra > steps from the application to remove tunnel mask and > collapse outer lengths into l2_len (in addition outer > IPv4 checksum calculation in SW). > > > I think a capability may be missing to differentiate which drivers > > support which mode. > > Yes, something like TX_INNER_CKSUM. > Some drivers already support it, some do not. > > Or, all drivers could be fixed to support both modes > > (and this would make this patch valid). > > I disagree with the point that something is required to > make the patch valid. Before the patch the driver supports > just (B). After the patch the driver supports both (A) and (B). > It makes the driver to support more cases. Yes, I misspoke, I wanted to say that today it is difficult for an application to use this new mode because it does not know which driver supports it. Fixing all drivers would solve the issue without a capability. A drivers/hw supporting one mode can easily support the other one (this patch demonstrates it), so maybe it's not worth having a capability. But yes, this is not a good reason to reject the patch. So: Reviewed-by: Olivier Matz <olivier.matz@6wind.com> > Please, note that rte_net_intel_cksum_prepare() supports > both (A) and (B) as well. ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-10-18 7:54 ` Olivier Matz @ 2021-10-18 8:20 ` Andrew Rybchenko 0 siblings, 0 replies; 8+ messages in thread From: Andrew Rybchenko @ 2021-10-18 8:20 UTC (permalink / raw) To: Olivier Matz Cc: Xia, Chenbo, Ivan Malov, dev, Maxime Coquelin, stable, Yuanhan Liu On 10/18/21 10:54 AM, Olivier Matz wrote: > On Mon, Oct 18, 2021 at 10:04:46AM +0300, Andrew Rybchenko wrote: >> On 10/15/21 11:32 AM, Olivier Matz wrote: >>> On Thu, Oct 14, 2021 at 07:12:29AM +0000, Xia, Chenbo wrote: >>>>> -----Original Message----- >>>>> From: Ivan Malov <ivan.malov@oktetlabs.ru> >>>>> Sent: Friday, September 17, 2021 2:50 AM >>>>> To: dev@dpdk.org >>>>> Cc: Maxime Coquelin <maxime.coquelin@redhat.com>; stable@dpdk.org; Andrew >>>>> Rybchenko <andrew.rybchenko@oktetlabs.ru>; Xia, Chenbo <chenbo.xia@intel.com>; >>>>> Yuanhan Liu <yuanhan.liu@linux.intel.com>; Olivier Matz >>>>> <olivier.matz@6wind.com> >>>>> Subject: [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel >>>>> packets >>>>> >>>>> Tx prepare method calls rte_net_intel_cksum_prepare(), which >>>>> handles tunnel packets correctly, but Tx burst path does not >>>>> take tunnel presence into account when computing the offsets. >>>>> >>>>> Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") >>>>> Cc: stable@dpdk.org >>>>> >>>>> Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> >>>>> Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> >>>>> --- >>>>> drivers/net/virtio/virtqueue.h | 9 ++++++--- >>>>> 1 file changed, 6 insertions(+), 3 deletions(-) >>>>> >>>>> diff --git a/drivers/net/virtio/virtqueue.h b/drivers/net/virtio/virtqueue.h >>>>> index 03957b2bd0..b83ff32efb 100644 >>>>> --- a/drivers/net/virtio/virtqueue.h >>>>> +++ b/drivers/net/virtio/virtqueue.h >>>>> @@ -620,19 +620,21 @@ static inline void >>>>> virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *cookie) >>>>> { >>>>> uint64_t csum_l4 = cookie->ol_flags & PKT_TX_L4_MASK; >>>>> + uint16_t o_l23_len = (cookie->ol_flags & PKT_TX_TUNNEL_MASK) ? >>>>> + cookie->outer_l2_len + cookie->outer_l3_len : 0; >>>>> >>>>> if (cookie->ol_flags & PKT_TX_TCP_SEG) >>>>> csum_l4 |= PKT_TX_TCP_CKSUM; >>>>> >>>>> switch (csum_l4) { >>>>> case PKT_TX_UDP_CKSUM: >>>>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; >>>>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; >>>>> hdr->csum_offset = offsetof(struct rte_udp_hdr, dgram_cksum); >>>>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; >>>>> break; >>>>> >>>>> case PKT_TX_TCP_CKSUM: >>>>> - hdr->csum_start = cookie->l2_len + cookie->l3_len; >>>>> + hdr->csum_start = o_l23_len + cookie->l2_len + cookie->l3_len; >>>>> hdr->csum_offset = offsetof(struct rte_tcp_hdr, cksum); >>>>> hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM; >>>>> break; >>>>> @@ -650,7 +652,8 @@ virtqueue_xmit_offload(struct virtio_net_hdr *hdr, struct >>>>> rte_mbuf *cookie) >>>>> VIRTIO_NET_HDR_GSO_TCPV6 : >>>>> VIRTIO_NET_HDR_GSO_TCPV4; >>>>> hdr->gso_size = cookie->tso_segsz; >>>>> - hdr->hdr_len = cookie->l2_len + cookie->l3_len + cookie->l4_len; >>>>> + hdr->hdr_len = o_l23_len + cookie->l2_len + cookie->l3_len + >>>>> + cookie->l4_len; >>>>> } else { >>>>> ASSIGN_UNLESS_EQUAL(hdr->gso_type, 0); >>>>> ASSIGN_UNLESS_EQUAL(hdr->gso_size, 0); >>>>> -- >>>>> 2.20.1 >>>> >>>> Reviewed-by: Chenbo Xia <chenbo.xia@intel.com> >>>> >>> >>> I have one comment to mention that from application perspective, it has >>> to take care that the driver does not support outer tunnel offload (this >>> matches the advertised capabilities). For instance, in case of a vxlan >>> tunnel, if the outer checksum needs to be calculated, it has to be done >>> by the application. In short, the application can ask to offload the >>> inner part if no offload is required on the outer part. >> >> Yes. It is really complicated. >> >> Let's name below case (A). >> >>> Also, since grep "PKT_TX_TUNNEL" in driver/net/ixgbe gives nothing, it >>> seems the ixgbe driver does not support the same offload request than >>> described in this patch: >>> (m->ol_flags & PKT_TX_TUNNEL_MASK) == PKT_TX_TUNNEL_XXXXX >>> m->outer_l2_len = outer l2 length >>> m->outer_l3_len = outer l3 length >>> m->l2_len = outer l4 length + tunnel len + inner l2 len >>> m->l3_len = inner l3 len >>> m->l4_len = inner l4 len >> >> True. >> >> Let's name below case (B). >> >>> An alternative for doing the same (that would work with ixgbe and >>> current virtio) is to give: >>> (m->ol_flags & PKT_TX_TUNNEL_MASK) == 0 >>> m->l2_len = outer lengths + tunnel len + inner l2 len >>> m->l3_len = inner l3 len >>> m->l4_len = inner l4 len >> >> Possible, but I think it is a bit less friendly. >> If application supports both cases with and without >> outer IPv4 checksum offload, it will require more extra >> steps from the application to remove tunnel mask and >> collapse outer lengths into l2_len (in addition outer >> IPv4 checksum calculation in SW). >> >>> I think a capability may be missing to differentiate which drivers >>> support which mode. >> >> Yes, something like TX_INNER_CKSUM. >> Some drivers already support it, some do not. >> > Or, all drivers could be fixed to support both modes >>> (and this would make this patch valid). >> >> I disagree with the point that something is required to >> make the patch valid. Before the patch the driver supports >> just (B). After the patch the driver supports both (A) and (B). >> It makes the driver to support more cases. > > Yes, I misspoke, I wanted to say that today it is difficult for an > application to use this new mode because it does not know which driver > supports it. Fixing all drivers would solve the issue without a > capability. > > A drivers/hw supporting one mode can easily support the other one (this > patch demonstrates it), so maybe it's not worth having a capability. Good point. I'll keep it in mind and try to find time to take a look at other drivers as well. > > But yes, this is not a good reason to reject the patch. > > So: > Reviewed-by: Olivier Matz <olivier.matz@6wind.com> > > >> Please, note that rte_net_intel_cksum_prepare() supports >> both (A) and (B) as well. ^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets 2021-09-16 18:49 ` [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets Ivan Malov 2021-10-14 6:45 ` Andrew Rybchenko 2021-10-14 7:12 ` Xia, Chenbo @ 2021-10-21 12:28 ` Maxime Coquelin 2 siblings, 0 replies; 8+ messages in thread From: Maxime Coquelin @ 2021-10-21 12:28 UTC (permalink / raw) To: Ivan Malov, dev Cc: stable, Andrew Rybchenko, Chenbo Xia, Yuanhan Liu, Olivier Matz On 9/16/21 20:49, Ivan Malov wrote: > Tx prepare method calls rte_net_intel_cksum_prepare(), which > handles tunnel packets correctly, but Tx burst path does not > take tunnel presence into account when computing the offsets. > > Fixes: 58169a9c8153 ("net/virtio: support Tx checksum offload") > Cc: stable@dpdk.org > > Signed-off-by: Ivan Malov <ivan.malov@oktetlabs.ru> > Reviewed-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru> > --- > drivers/net/virtio/virtqueue.h | 9 ++++++--- > 1 file changed, 6 insertions(+), 3 deletions(-) > Applied to dpdk-next-virtio/main. Thanks, Maxime ^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2021-10-21 12:29 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <20210830142655.18373-1-ivan.malov@oktetlabs.ru>
2021-09-16 18:49 ` [dpdk-stable] [PATCH v2] net/virtio: handle Tx checksums correctly for tunnel packets Ivan Malov
2021-10-14 6:45 ` Andrew Rybchenko
2021-10-14 7:12 ` Xia, Chenbo
2021-10-15 8:32 ` Olivier Matz
2021-10-18 7:04 ` Andrew Rybchenko
2021-10-18 7:54 ` Olivier Matz
2021-10-18 8:20 ` Andrew Rybchenko
2021-10-21 12:28 ` Maxime Coquelin
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).