Re: Dropped packets mapping IRQs for adjusted queue counts on i40e
Zvi Effron <[email protected]>
| Newsgroups | org.kernel.vger.xdp-newbies |
|---|---|
| Message-ID | <CAC1LvL2zmO1ntKeAoUMkJSarJBgxNhnTva3Di4047MTKqo8rPA@mail.gmail.com> |
On Sun, May 2, 2021 at 1:16 PM T K Sourabh <[email protected]> wrote: > > On Sat, May 1, 2021 at 11:26 PM Zvi Effron <[email protected]> wrote: > > > > I am wondering if not having the matching queues for the ingress > > interface and egress interface on the same cores might be a > > contributing factor? > I haven't tested but you could give it a try to see if packet drop happens. I have now tested matching the IRQs for queue N on both the ingress and egress interfaces so they are on the same core. (Queue 0 for ingress and egress are on the same core, the same for queue 1 on each, etc, but queue 0 and queue 1 are on different cores). This did not resolve the packet loss. I'm at a loss for what could be causing the packet loss. It's clearly something to do with the IRQ remapping, as all I need to do to remove the loss is comment out the write to the smp_affinity files. I'm suspecting it's something with how XDP_REDIRECT is implemented in the i40e driver, but I don't know if this is a) cross driver behavior, b) expected behavior, or c) a bug. --Zvi