Message ID | 20240215194325.1364466-1-dw@davidwei.uk (mailing list archive) |
---|---|
Headers | show |
Series | netdevsim: link and forward skbs between ports | expand |
Hi, On Thu, 2024-02-15 at 11:43 -0800, David Wei wrote: > This patchset adds the ability to link two netdevsim ports together and > forward skbs between them, similar to veth. The goal is to use netdevsim > for testing features e.g. zero copy Rx using io_uring. > > This feature was tested locally on QEMU, and a selftest is included. this apparently causes rtnetlink.sh self-tests failures: https://netdev.bots.linux.dev/flakes.html?tn-needle=rtnetlink-sh example failure: https://netdev-3.bots.linux.dev/vmksft-net/results/467721/18-rtnetlink-sh/stdout the ipsec_offload test (using netdevsim) fails. @Jakub: it looks like the rtnetlink.sh test is currently ignored by patchwork, skimming over the recent failures they are roughly correlated to this series submission: the test looks otherwise reasonably stable to me. Cheers, Paolo
On Fri, 16 Feb 2024 11:38:17 +0100 Paolo Abeni wrote: > On Thu, 2024-02-15 at 11:43 -0800, David Wei wrote: > > This patchset adds the ability to link two netdevsim ports together and > > forward skbs between them, similar to veth. The goal is to use netdevsim > > for testing features e.g. zero copy Rx using io_uring. > > > > This feature was tested locally on QEMU, and a selftest is included. > > this apparently causes rtnetlink.sh self-tests failures: > > https://netdev.bots.linux.dev/flakes.html?tn-needle=rtnetlink-sh > > example failure: > > https://netdev-3.bots.linux.dev/vmksft-net/results/467721/18-rtnetlink-sh/stdout > > the ipsec_offload test (using netdevsim) fails. > > @Jakub: it looks like the rtnetlink.sh test is currently ignored by > patchwork, skimming over the recent failures they are roughly > correlated to this series submission: the test looks otherwise > reasonably stable to me. Wow, great detective work! This is what the diff says: # 1c1 # < SA count=2 tx=0 # --- # > SA count=2 tx=3 I'm guessing that because we clear IFF_NOARP and ipsec code doesn't connect any peer it can't transmit? Any suggestions how to fix it? Insert a neigh entry manually?
On 2024-02-16 03:38, Paolo Abeni wrote: > Hi, > > On Thu, 2024-02-15 at 11:43 -0800, David Wei wrote: >> This patchset adds the ability to link two netdevsim ports together and >> forward skbs between them, similar to veth. The goal is to use netdevsim >> for testing features e.g. zero copy Rx using io_uring. >> >> This feature was tested locally on QEMU, and a selftest is included. > > this apparently causes rtnetlink.sh self-tests failures: > > https://netdev.bots.linux.dev/flakes.html?tn-needle=rtnetlink-sh > > example failure: > > https://netdev-3.bots.linux.dev/vmksft-net/results/467721/18-rtnetlink-sh/stdout > > the ipsec_offload test (using netdevsim) fails. Thanks for catching this Paulo. Can I sort this out in a separate patch, as the rtnetlink.sh test is disabled in CI right now? > > @Jakub: it looks like the rtnetlink.sh test is currently ignored by > patchwork, skimming over the recent failures they are roughly > correlated to this series submission: the test looks otherwise > reasonably stable to me. > > Cheers, > > Paolo >
On Fri, 16 Feb 2024 12:13:42 -0700 David Wei wrote: > > this apparently causes rtnetlink.sh self-tests failures: > > > > https://netdev.bots.linux.dev/flakes.html?tn-needle=rtnetlink-sh > > > > example failure: > > > > https://netdev-3.bots.linux.dev/vmksft-net/results/467721/18-rtnetlink-sh/stdout > > > > the ipsec_offload test (using netdevsim) fails. > > Thanks for catching this Paulo. Can I sort this out in a separate patch, > as the rtnetlink.sh test is disabled in CI right now? Looks like the code needs fixing, still, so please tack the rtnetlink fix onto the v12
Also looks like the new test managed to flake once while it was sitting in patchwork ? https://netdev-3.bots.linux.dev/vmksft-netdevsim-dbg/results/468440/13-peer-sh/stdout
On 2024-02-16 18:47, Jakub Kicinski wrote: > On Fri, 16 Feb 2024 12:13:42 -0700 David Wei wrote: >>> this apparently causes rtnetlink.sh self-tests failures: >>> >>> https://netdev.bots.linux.dev/flakes.html?tn-needle=rtnetlink-sh >>> >>> example failure: >>> >>> https://netdev-3.bots.linux.dev/vmksft-net/results/467721/18-rtnetlink-sh/stdout >>> >>> the ipsec_offload test (using netdevsim) fails. >> >> Thanks for catching this Paulo. Can I sort this out in a separate patch, >> as the rtnetlink.sh test is disabled in CI right now? > > Looks like the code needs fixing, still, so please tack the rtnetlink > fix onto the v12 Okay I'll add it on!
On 2024-02-16 18:48, Jakub Kicinski wrote: > Also looks like the new test managed to flake once while it was sitting > in patchwork ? > > https://netdev-3.bots.linux.dev/vmksft-netdevsim-dbg/results/468440/13-peer-sh/stdout I can't repro it locally in QEMU. Maybe it's due to the leaky ref bug? I'll send another revision and hopefully this doesn't happen again.
On Fri, 16 Feb 2024 20:27:09 -0700 David Wei wrote: > On 2024-02-16 18:48, Jakub Kicinski wrote: > > Also looks like the new test managed to flake once while it was sitting > > in patchwork ? > > > > https://netdev-3.bots.linux.dev/vmksft-netdevsim-dbg/results/468440/13-peer-sh/stdout > > I can't repro it locally in QEMU. Maybe it's due to the leaky ref bug? > > I'll send another revision and hopefully this doesn't happen again. Grep for wait_local_port_listen, maybe we should make sure listener had fully started?
On 2024-02-19 12:20, Jakub Kicinski wrote: > On Fri, 16 Feb 2024 20:27:09 -0700 David Wei wrote: >> On 2024-02-16 18:48, Jakub Kicinski wrote: >>> Also looks like the new test managed to flake once while it was sitting >>> in patchwork ? >>> >>> https://netdev-3.bots.linux.dev/vmksft-netdevsim-dbg/results/468440/13-peer-sh/stdout >> >> I can't repro it locally in QEMU. Maybe it's due to the leaky ref bug? >> >> I'll send another revision and hopefully this doesn't happen again. > > Grep for wait_local_port_listen, maybe we should make sure listener > had fully started? Ah, thanks. I'll add that into the selftest.