Hi,
Ok, thanks for your reply.
I will continue to work on it.
B.R.
Peng
> -----Original Message-----
> From: Eli Britstein <[email protected]>
> Sent: Sunday, April 2, 2023 1:58 PM
> To: Nole Zhang <[email protected]>; Simon Horman
> <[email protected]>; [email protected]
> Cc: Ilya Maximets <[email protected]>; Chaoyong He
> <[email protected]>; Kevin Liu <[email protected]>; oss-drivers
> <[email protected]>
> Subject: RE: [PATCH dpdk-latest v3 0/6] Add support for DPDK meter HW
> offload
>
> Hi
>
> The kernel provides infrastructure that does not exist for DPDK. TC is
> different and cannot be imitated directly in all cases.
> Using a meter object created on one port (proxy) on another one (for flow) is
> illegal.
> Also, with this approach, even if fixed, offload will not work on if the port
> is
> added after the meter is created.
> Regarding modify, it should be on the fly. When a meter is updated, the
> related HW meters should be updated via their proxy ports (as the ones
> created them).
> If all flows using it are un-offloaded, they will not be re-offloaded.
>
> Thanks,
> Eli
>
> >-----Original Message-----
> >From: Nole Zhang <[email protected]>
> >Sent: Friday, 31 March 2023 6:20
> >To: Eli Britstein <[email protected]>; Simon Horman
> ><[email protected]>; [email protected]
> >Cc: Ilya Maximets <[email protected]>; Chaoyong He
> ><[email protected]>; Kevin Liu <[email protected]>;
> >oss-drivers <[email protected]>
> >Subject: RE: [PATCH dpdk-latest v3 0/6] Add support for DPDK meter HW
> >offload
> >
> >External email: Use caution opening links or attachments
> >
> >
> >Hi,
> > Thanks for your reply.
> > Our original idea is imitating OVS-TC, meter action and flow is
> >independence. First through the proxy port create the meter, then flow
> >use meter with meter id. So I only add the proxy port, doesn't revise
> >the flow with proxy port. So you think this idea cant be accepted?
> > As your said, when the flow first time use the meter/proxy
> >port, we offload the meter to the HW, when the last flow with the
> >meter/proxy port will be delete, we delete the meter from the HW.
> > About your idea, I have some issue about modify the flow,
> >because the flow cant get the meter status, so we can't through flow to
> >modify the meter in the HW. We have two way to solve it, one is modify
> >the meter, we delete all the flow with meter from the HW, the key is
> >through meter id lookup the all flow with the meter id from the HW.
> >Another is update the meter through the proxy port in the HW. Do you
> think which is better?
> >B.R.
> >Peng
> >
> >> -----Original Message-----
> >> From: Eli Britstein <[email protected]>
> >> Sent: Thursday, March 30, 2023 8:34 PM
> >> To: Simon Horman <[email protected]>;
> [email protected]
> >> Cc: Ilya Maximets <[email protected]>; Chaoyong He
> >> <[email protected]>; Nole Zhang <[email protected]>;
> >> Kevin Liu <[email protected]>; oss-drivers
> >> <[email protected]>
> >> Subject: RE: [PATCH dpdk-latest v3 0/6] Add support for DPDK meter HW
> >> offload
> >>
> >> Hi
> >>
> >> The usage of the proxy port is wrong.
> >> All transfer offload should be migrated to use it. There was [1-4],
> >> but it was not completed.
> >>
> >> Meters should be created only on the proxy ports. As all offloads are
> >> moved to use proxy ports, this is how they are shared.
> >> Since we don't know the proxy port when meter is added, it can be
> >> lazy created upon the first flow that uses this meter/proxy-port
> >> (referenced counted), and destroy upon the last flow destroy using it.
> >> As there might be multiple "proxy" ports, there might be multiple HW
> >> meter objects associated with the same SW one. This should be managed.
> >>
> >> [1]
> >>
> >https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatc
> h
> >>
> >work.ozlabs.org%2Fproject%2Fopenvswitch%2Flist%2F%3Fseries%3D30252
> 5
> >%26
> >>
> >state&data=05%7C01%7Celibr%40nvidia.com%7C0f1851b0a5e94982778108
> db
> >3196
> >>
> >c6d8%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C63815829588
> 4302
> >138%7
> >>
> >CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ
> B
> >TiI6Ik1
> >>
> >haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=%2B1fCFn4UeTJpbV
> Sr
> >m3Sh5KqA7%
> >> 2FG64u81oeBP9xLY0q8%3D&reserved=0
> >> =*
> >> [2]
> >>
> >https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatc
> h
> >>
> >work.ozlabs.org%2Fproject%2Fopenvswitch%2Flist%2F%3Fseries%3D31041
> 3
> >%26
> >>
> >state&data=05%7C01%7Celibr%40nvidia.com%7C0f1851b0a5e94982778108
> db
> >3196
> >>
> >c6d8%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C63815829588
> 4302
> >138%7
> >>
> >CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ
> B
> >TiI6Ik1
> >>
> >haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=nywcgXWj2VfgzYW5
> D
> >DgnJWm002r2
> >> Jv03VDShjkToMiA%3D&reserved=0
> >> =*
> >> [3]
> >>
> >https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatc
> h
> >>
> >work.ozlabs.org%2Fproject%2Fopenvswitch%2Flist%2F%3Fseries%3D31041
> 5
> >%26
> >>
> >state&data=05%7C01%7Celibr%40nvidia.com%7C0f1851b0a5e94982778108
> db
> >3196
> >>
> >c6d8%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C63815829588
> 4302
> >138%7
> >>
> >CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ
> B
> >TiI6Ik1
> >>
> >haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=9Ee5Qz9yCwx%2BHc
> F6
> >f%2BT3F8Oo
> >> NwZyYYcinT10wh7BG%2F0%3D&reserved=0
> >> =*
> >> [4]
> >>
> >https://nam11.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpatc
> h
> >>
> >work.ozlabs.org%2Fproject%2Fopenvswitch%2Flist%2F%3Fseries%3D34288
> 5
> >%26
> >>
> >state&data=05%7C01%7Celibr%40nvidia.com%7C0f1851b0a5e94982778108
> db
> >3196
> >>
> >c6d8%7C43083d15727340c1b7db39efd9ccc17a%7C0%7C0%7C63815829588
> 4302
> >138%7
> >>
> >CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJ
> B
> >TiI6Ik1
> >>
> >haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=rmW1%2FaUKHmj2
> Wd
> >nmHwcSIE7XZH
> >> 0pEjIOLvGoiCRF8KM%3D&reserved=0
> >> =*
> >>
> >> Thanks,
> >> Eli
> >>
> >> >-----Original Message-----
> >> >From: Simon Horman <[email protected]>
> >> >Sent: Thursday, 30 March 2023 14:21
> >> >To: [email protected]
> >> >Cc: Ilya Maximets <[email protected]>; Eli Britstein
> >> ><[email protected]>; Chaoyong He <[email protected]>; Peng
> >> Zhang
> >> ><[email protected]>; Jin Liu <[email protected]>; oss-
> >> >[email protected]
> >> >Subject: [PATCH dpdk-latest v3 0/6] Add support for DPDK meter HW
> >> >offload
> >> >
> >> >External email: Use caution opening links or attachments
> >> >
> >> >
> >> >Hi,
> >> >
> >> >this series adds support for DPDK meter HW offload.
> >> >
> >> >Changes between v2 and v3.
> >> >* Use common API for DPDK and non-DPDK meter offloads
> >> >* Make use of netdev_ports_traverse to offload the meter
> >> >* Add dpdk-latest to subject prefix
> >> >
> >> >Changes between v1 and v2:
> >> >* Add the prox mechanism: add the meter by proxy id
> >> >* Change the offload interface from netdev-dpdk to the
> >> >netdev-offload
> >> >* Changed base to dpdk-latest branch
> >> >
> >> >Peng Zhang (6):
> >> > netdev-offload-dpdk: use flow transfer proxy
> >> > netdev-offload: Let meter offload API can be used with DPDK
> >> > dpif-netdev: Offloading meter with DPDK
> >> > netdev-offload-dpdk: Implement meter offload API for DPDK
> >> > netdev-dpdk: add meter algorithms
> >> > netdev-dpdk-offload: Add support for meter action
> >> >
> >> > Documentation/howto/dpdk.rst | 5 +-
> >> > lib/dpif-netdev.c | 22 ++-
> >> > lib/netdev-dpdk.c | 306 +++++++++++++++++++++++++++++++++-
> >> > lib/netdev-dpdk.h | 43 +++++
> >> > lib/netdev-offload-dpdk.c | 97 +++++++++++
> >> > lib/netdev-offload-provider.h | 21 ++-
> >> > lib/netdev-offload-tc.c | 9 +-
> >> > lib/netdev-offload.c | 135 ++++++++++++++-
> >> > lib/netdev-offload.h | 9 +
> >> > 9 files changed, 633 insertions(+), 14 deletions(-)
> >> >
> >> >--
> >> >2.30.2
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
