diff mbox series

[v2] net/virtio: handle Tx checksums correctly for tunnel packets

Message ID 20210916184955.2755-1-ivan.malov@oktetlabs.ru (mailing list archive)
State Accepted, archived
Delegated to: Maxime Coquelin
Headers
Series [v2] net/virtio: handle Tx checksums correctly for tunnel packets |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/github-robot: build success github build: passed
ci/Intel-compilation success Compilation OK
ci/intel-Testing fail Testing issues
ci/iol-intel-Functional success Functional Testing PASS
ci/iol-x86_64-unit-testing success Testing PASS
ci/iol-aarch64-compile-testing success Testing PASS
ci/iol-x86_64-compile-testing success Testing PASS

Commit Message

Ivan Malov Sept. 16, 2021, 6:49 p.m. UTC 
  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(-)

Comments

Andrew Rybchenko Oct. 14, 2021, 6:45 a.m. UTC | #1 
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);
>
Chenbo Xia Oct. 14, 2021, 7:12 a.m. UTC | #2 
> -----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>
Olivier Matz Oct. 15, 2021, 8:32 a.m. UTC | #3 
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
Andrew Rybchenko Oct. 18, 2021, 7:04 a.m. UTC | #4 
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.
Olivier Matz Oct. 18, 2021, 7:54 a.m. UTC | #5 
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.
Andrew Rybchenko Oct. 18, 2021, 8:20 a.m. UTC | #6 
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.
Maxime Coquelin Oct. 21, 2021, 12:28 p.m. UTC | #7 
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
diff mbox series

Patch

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);