Message ID | 20200212145447.2136-1-tomaszx.konieczny@intel.com |
---|---|
State | Rejected |
Headers | show |
Series | [ovs-dev,1/1] docs: Add HWOL validated NIC | expand |
On 2/12/20 3:54 PM, Tomasz Konieczny wrote: > Signed-off-by: Tomasz Konieczny <tomaszx.konieczny@intel.com> > --- > Documentation/howto/dpdk.rst | 19 +++++++++++++++---- > Documentation/intro/install/dpdk.rst | 4 ++++ > 2 files changed, 19 insertions(+), 4 deletions(-) > > diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst > index be950d7..e890323 100644 > --- a/Documentation/howto/dpdk.rst > +++ b/Documentation/howto/dpdk.rst > @@ -377,10 +377,21 @@ drivers that support the configured rte_flow actions. > Partial flow offload requires support of "MARK + RSS" actions. Full > hardware offload requires support of the actions listed below. > > -The validated NICs are: > - > -- Mellanox (ConnectX-4, ConnectX-4 Lx, ConnectX-5) > -- Napatech (NT200B01) > +.. table:: Validated NICs > + > + +----------+---------------+--------------+-----------------+ > + | Vendor | Model | Full offload | Partial offload | > + +==========+===============+==============+=================+ > + | Intel | 700 Series | no | yes | > + +----------+---------------+--------------+-----------------+ > + | Mellanox | ConnectX-4 | | yes | > + | +---------------+--------------+-----------------+ > + | | ConnectX-4 Lx | | yes | > + | +---------------+--------------+-----------------+ > + | | ConnectX-5 | | yes | > + +----------+---------------+--------------+-----------------+ > + | Napatech | NT200B01 | | yes | > + +----------+---------------+--------------+-----------------+ Do we need this at all? Maybe it's better to just remove "validated NICs" section? Rationale: We can't test all NICs and it's literally not an OVS responsibility to test every single NIC. OVS just uses DPDK API provided by the DPDK library and we're clearly stating which functionality we're requiring from NIC. Users should check the DPDK documentation if their driver supports required offloading features. Above list of validated NICs forces us to test these NICs for each release to be sure that they are still working. I doubt someone does that. Thoughts? Best regards, Ilya Maximets. > > Supported protocols for hardware offload matches are: > > diff --git a/Documentation/intro/install/dpdk.rst b/Documentation/intro/install/dpdk.rst > index dbf88ec..db5910c 100644 > --- a/Documentation/intro/install/dpdk.rst > +++ b/Documentation/intro/install/dpdk.rst > @@ -687,6 +687,10 @@ Limitations > around is temporary and is expected to be removed once a method is provided > by DPDK to query the upper bound MTU value for a given device. > > +- Flow Hardware Offload: i40e devices (Intel(R) 700 Series) do not support > + L2: Ethernet and L4: ICMP for offload matches. Flow rule needs to specify > + highest supported layer for traffic match. We could document known issues though. Best regards, Ilya Maximets.
On Wed, Feb 12, 2020 at 04:10:28PM +0100, Ilya Maximets wrote: > On 2/12/20 3:54 PM, Tomasz Konieczny wrote: > > Signed-off-by: Tomasz Konieczny <tomaszx.konieczny@intel.com> > > --- > > Documentation/howto/dpdk.rst | 19 +++++++++++++++---- > > Documentation/intro/install/dpdk.rst | 4 ++++ > > 2 files changed, 19 insertions(+), 4 deletions(-) > > > > diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst > > index be950d7..e890323 100644 > > --- a/Documentation/howto/dpdk.rst > > +++ b/Documentation/howto/dpdk.rst > > @@ -377,10 +377,21 @@ drivers that support the configured rte_flow actions. > > Partial flow offload requires support of "MARK + RSS" actions. Full > > hardware offload requires support of the actions listed below. > > > > -The validated NICs are: > > - > > -- Mellanox (ConnectX-4, ConnectX-4 Lx, ConnectX-5) > > -- Napatech (NT200B01) > > +.. table:: Validated NICs > > + > > + +----------+---------------+--------------+-----------------+ > > + | Vendor | Model | Full offload | Partial offload | > > + +==========+===============+==============+=================+ > > + | Intel | 700 Series | no | yes | > > + +----------+---------------+--------------+-----------------+ > > + | Mellanox | ConnectX-4 | | yes | > > + | +---------------+--------------+-----------------+ > > + | | ConnectX-4 Lx | | yes | > > + | +---------------+--------------+-----------------+ > > + | | ConnectX-5 | | yes | > > + +----------+---------------+--------------+-----------------+ > > + | Napatech | NT200B01 | | yes | > > + +----------+---------------+--------------+-----------------+ > > > Do we need this at all? Maybe it's better to just remove > "validated NICs" section? > > Rationale: We can't test all NICs and it's literally not an OVS > responsibility to test every single NIC. OVS just uses DPDK API > provided by the DPDK library and we're clearly stating which > functionality we're requiring from NIC. Users should check the > DPDK documentation if their driver supports required offloading > features. Above list of validated NICs forces us to test these > NICs for each release to be sure that they are still working. > I doubt someone does that. > > Thoughts? +1 here. > > > > Supported protocols for hardware offload matches are: > > > > diff --git a/Documentation/intro/install/dpdk.rst b/Documentation/intro/install/dpdk.rst > > index dbf88ec..db5910c 100644 > > --- a/Documentation/intro/install/dpdk.rst > > +++ b/Documentation/intro/install/dpdk.rst > > @@ -687,6 +687,10 @@ Limitations > > around is temporary and is expected to be removed once a method is provided > > by DPDK to query the upper bound MTU value for a given device. > > > > +- Flow Hardware Offload: i40e devices (Intel(R) 700 Series) do not support > > + L2: Ethernet and L4: ICMP for offload matches. Flow rule needs to specify > > + highest supported layer for traffic match. > > We could document known issues though. +1 here as well. fbl
On 12/02/2020 19:53, Flavio Leitner wrote: > On Wed, Feb 12, 2020 at 04:10:28PM +0100, Ilya Maximets wrote: >> On 2/12/20 3:54 PM, Tomasz Konieczny wrote: >>> Signed-off-by: Tomasz Konieczny <tomaszx.konieczny@intel.com> >>> --- >>> Documentation/howto/dpdk.rst | 19 +++++++++++++++---- >>> Documentation/intro/install/dpdk.rst | 4 ++++ >>> 2 files changed, 19 insertions(+), 4 deletions(-) >>> >>> diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst >>> index be950d7..e890323 100644 >>> --- a/Documentation/howto/dpdk.rst >>> +++ b/Documentation/howto/dpdk.rst >>> @@ -377,10 +377,21 @@ drivers that support the configured rte_flow actions. >>> Partial flow offload requires support of "MARK + RSS" actions. Full >>> hardware offload requires support of the actions listed below. >>> >>> -The validated NICs are: >>> - >>> -- Mellanox (ConnectX-4, ConnectX-4 Lx, ConnectX-5) >>> -- Napatech (NT200B01) >>> +.. table:: Validated NICs >>> + >>> + +----------+---------------+--------------+-----------------+ >>> + | Vendor | Model | Full offload | Partial offload | >>> + +==========+===============+==============+=================+ >>> + | Intel | 700 Series | no | yes | >>> + +----------+---------------+--------------+-----------------+ >>> + | Mellanox | ConnectX-4 | | yes | >>> + | +---------------+--------------+-----------------+ >>> + | | ConnectX-4 Lx | | yes | >>> + | +---------------+--------------+-----------------+ >>> + | | ConnectX-5 | | yes | >>> + +----------+---------------+--------------+-----------------+ >>> + | Napatech | NT200B01 | | yes | >>> + +----------+---------------+--------------+-----------------+ >> >> >> Do we need this at all? Maybe it's better to just remove >> "validated NICs" section? >> >> Rationale: We can't test all NICs and it's literally not an OVS >> responsibility to test every single NIC. OVS just uses DPDK API >> provided by the DPDK library and we're clearly stating which >> functionality we're requiring from NIC. Users should check the >> DPDK documentation if their driver supports required offloading >> features. Above list of validated NICs forces us to test these >> NICs for each release to be sure that they are still working. >> I doubt someone does that. >> >> Thoughts? > > +1 here. > +1. Could add a pointer like, "For more information about which NICs support rte_flow actions, see 'Flow API' row in http://doc.dpdk.org/guides-19.11/nics/overview.html" >>> >>> Supported protocols for hardware offload matches are: >>> >>> diff --git a/Documentation/intro/install/dpdk.rst b/Documentation/intro/install/dpdk.rst >>> index dbf88ec..db5910c 100644 >>> --- a/Documentation/intro/install/dpdk.rst >>> +++ b/Documentation/intro/install/dpdk.rst >>> @@ -687,6 +687,10 @@ Limitations >>> around is temporary and is expected to be removed once a method is provided >>> by DPDK to query the upper bound MTU value for a given device. >>> >>> +- Flow Hardware Offload: i40e devices (Intel(R) 700 Series) do not support >>> + L2: Ethernet and L4: ICMP for offload matches. Flow rule needs to specify >>> + highest supported layer for traffic match. >> >> We could document known issues though. > > +1 here as well. > fbl > _______________________________________________ > dev mailing list > dev@openvswitch.org > https://mail.openvswitch.org/mailman/listinfo/ovs-dev >
>-----Original Message----- >From: dev <ovs-dev-bounces@openvswitch.org> On Behalf Of Kevin Traynor >Sent: 13 February 2020 11:37 >To: Flavio Leitner <fbl@sysclose.org>; Ilya Maximets <i.maximets@ovn.org> >Cc: dev@openvswitch.org; David Marchand <david.marchand@redhat.com> >Subject: Re: [ovs-dev] [PATCH 1/1] docs: Add HWOL validated NIC > >On 12/02/2020 19:53, Flavio Leitner wrote: >> On Wed, Feb 12, 2020 at 04:10:28PM +0100, Ilya Maximets wrote: >>> On 2/12/20 3:54 PM, Tomasz Konieczny wrote: >>>> Signed-off-by: Tomasz Konieczny <tomaszx.konieczny@intel.com> >>>> --- >>>> Documentation/howto/dpdk.rst | 19 +++++++++++++++---- >>>> Documentation/intro/install/dpdk.rst | 4 ++++ >>>> 2 files changed, 19 insertions(+), 4 deletions(-) >>>> >>>> diff --git a/Documentation/howto/dpdk.rst >>>> b/Documentation/howto/dpdk.rst index be950d7..e890323 100644 >>>> --- a/Documentation/howto/dpdk.rst >>>> +++ b/Documentation/howto/dpdk.rst >>>> @@ -377,10 +377,21 @@ drivers that support the configured rte_flow >actions. >>>> Partial flow offload requires support of "MARK + RSS" actions. Full >>>> hardware offload requires support of the actions listed below. >>>> >>>> -The validated NICs are: >>>> - >>>> -- Mellanox (ConnectX-4, ConnectX-4 Lx, ConnectX-5) >>>> -- Napatech (NT200B01) >>>> +.. table:: Validated NICs >>>> + >>>> + +----------+---------------+--------------+-----------------+ >>>> + | Vendor | Model | Full offload | Partial offload | >>>> + >+==========+===============+==============+=================+ >>>> + | Intel | 700 Series | no | yes | >>>> + +----------+---------------+--------------+-----------------+ >>>> + | Mellanox | ConnectX-4 | | yes | >>>> + | +---------------+--------------+-----------------+ >>>> + | | ConnectX-4 Lx | | yes | >>>> + | +---------------+--------------+-----------------+ >>>> + | | ConnectX-5 | | yes | >>>> + +----------+---------------+--------------+-----------------+ >>>> + | Napatech | NT200B01 | | yes | >>>> + +----------+---------------+--------------+-----------------+ >>> >>> >>> Do we need this at all? Maybe it's better to just remove "validated >>> NICs" section? >>> >>> Rationale: We can't test all NICs and it's literally not an OVS >>> responsibility to test every single NIC. OVS just uses DPDK API >>> provided by the DPDK library and we're clearly stating which >>> functionality we're requiring from NIC. Users should check the DPDK >>> documentation if their driver supports required offloading features. >>> Above list of validated NICs forces us to test these NICs for each >>> release to be sure that they are still working. >>> I doubt someone does that. >>> >>> Thoughts? >> >> +1 here. >> > >+1. Could add a pointer like, "For more information about which NICs >support rte_flow actions, see 'Flow API' row in http://doc.dpdk.org/guides- >19.11/nics/overview.html" > Actually it's not enough. Even though table says "Flow API Y", it doesn't mean all rte_flow actions are supported. That said I also think DPDK documentation would be a better place for this, but it needs to be expanded. >>>> >>>> Supported protocols for hardware offload matches are: >>>> >>>> diff --git a/Documentation/intro/install/dpdk.rst >>>> b/Documentation/intro/install/dpdk.rst >>>> index dbf88ec..db5910c 100644 >>>> --- a/Documentation/intro/install/dpdk.rst >>>> +++ b/Documentation/intro/install/dpdk.rst >>>> @@ -687,6 +687,10 @@ Limitations >>>> around is temporary and is expected to be removed once a method is >provided >>>> by DPDK to query the upper bound MTU value for a given device. >>>> >>>> +- Flow Hardware Offload: i40e devices (Intel(R) 700 Series) do not >>>> +support >>>> + L2: Ethernet and L4: ICMP for offload matches. Flow rule needs to >>>> +specify >>>> + highest supported layer for traffic match. >>> >>> We could document known issues though. >> >> +1 here as well. >> fbl >> _______________________________________________ >> dev mailing list >> dev@openvswitch.org >> https://mail.openvswitch.org/mailman/listinfo/ovs-dev >> > >_______________________________________________ >dev mailing list >dev@openvswitch.org >https://mail.openvswitch.org/mailman/listinfo/ovs-dev Regards Tomasz Konieczny --------------------------------------------------------------------- Intel Corporation (UK) Ltd. Co. Reg. #1134945 Pipers Way, Swindon SN3 1RJ
diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst index be950d7..e890323 100644 --- a/Documentation/howto/dpdk.rst +++ b/Documentation/howto/dpdk.rst @@ -377,10 +377,21 @@ drivers that support the configured rte_flow actions. Partial flow offload requires support of "MARK + RSS" actions. Full hardware offload requires support of the actions listed below. -The validated NICs are: - -- Mellanox (ConnectX-4, ConnectX-4 Lx, ConnectX-5) -- Napatech (NT200B01) +.. table:: Validated NICs + + +----------+---------------+--------------+-----------------+ + | Vendor | Model | Full offload | Partial offload | + +==========+===============+==============+=================+ + | Intel | 700 Series | no | yes | + +----------+---------------+--------------+-----------------+ + | Mellanox | ConnectX-4 | | yes | + | +---------------+--------------+-----------------+ + | | ConnectX-4 Lx | | yes | + | +---------------+--------------+-----------------+ + | | ConnectX-5 | | yes | + +----------+---------------+--------------+-----------------+ + | Napatech | NT200B01 | | yes | + +----------+---------------+--------------+-----------------+ Supported protocols for hardware offload matches are: diff --git a/Documentation/intro/install/dpdk.rst b/Documentation/intro/install/dpdk.rst index dbf88ec..db5910c 100644 --- a/Documentation/intro/install/dpdk.rst +++ b/Documentation/intro/install/dpdk.rst @@ -687,6 +687,10 @@ Limitations around is temporary and is expected to be removed once a method is provided by DPDK to query the upper bound MTU value for a given device. +- Flow Hardware Offload: i40e devices (Intel(R) 700 Series) do not support + L2: Ethernet and L4: ICMP for offload matches. Flow rule needs to specify + highest supported layer for traffic match. + Reporting Bugs --------------
Signed-off-by: Tomasz Konieczny <tomaszx.konieczny@intel.com> --- Documentation/howto/dpdk.rst | 19 +++++++++++++++---- Documentation/intro/install/dpdk.rst | 4 ++++ 2 files changed, 19 insertions(+), 4 deletions(-)