On Tue, Sep 27, 2022 at 1:03 AM Ilya Maximets <[email protected]> wrote: > On 9/23/22 15:46, David Marchand wrote: > > On Fri, Sep 23, 2022 at 2:43 PM Kevin Traynor <[email protected]> wrote: > >> > >> On 22/09/2022 13:40, Michael Phelan wrote: > >>> Update OVS CLI and relevant documentation to use DPDK 20.11.6. > >>> > >>> A bug was introduced in DPDK 20.11.5 by the commit 33f2e3756186 ("vhost: > >>> fix unsafe vring addresses modifications"). > >>> This bug can cause a deadlock when vIOMMU is enabled and NUMA > >>> reallocation of the virtqueues happen. > >>> A fix [1] has been posted and pushed to the DPDK 20.11 branch. > >>> If a user wishes to avoid the issue then it is recommended to use DPDK > >>> 20.11.4 until the release of DPDK 20.11.7. > >>> It should be noted that DPDK 20.11.4 does not benefit from the numerous > >>> bug fixes addressed since its release. > >>> If a user wishes to benefit from these fixes it is recommended to use > >>> DPDK 20.11.6. > >>> > >>> [1] > >>> https://patches.dpdk.org/project/dpdk/patch/[email protected]/ > >>> Signed-off-by: Michael Phelan <[email protected]> > >>> > >> > >> For branches 2.15 [0] and 2.16 [1] I ran github actions and it failed. > >> For 2.16 branch I removed the patch and it passed [2]. It seems like > >> that the meson used (0.47.1 - which is min version for 20.11) does not > >> like the 20.11.5/6 package, or there is some other github effect. It is > >> working fine with 20.11.4. > >> > >> Afterwards, checking the ovs-build mailing [4] list I also see failures > >> here and an additional failure for 2.17 branch. So all these failures > >> need to checked. > >> > >> [1] https://github.com/kevintraynor/ovs/actions/runs/3111862351 > >> [2] https://github.com/kevintraynor/ovs/actions/runs/3111865180 > >> [3] https://github.com/kevintraynor/ovs/actions/runs/3112089634 > >> [4] > >> https://mail.openvswitch.org/pipermail/ovs-build/2022-September/date.html > > > > This looks like a regression in 20.11 LTS with older meson. > > Adding 20.11 LTS maintainers to the thread. > > I'm guessing that this regression will not be fixed until the next > series of DPDK stable releases. And since we're testing OVS here, > not DPDK, we may just choose a better version of meson. Would be > also nice to have one that works with python 3.10+, so we can actually > use more recent versions of python without capping it at 3.9. > > Any suggestions for a version to use?
I agree that as consumers of DPDK, OVS could select a more recent version of meson that fits its needs. Just reinstating that we should stick to a fixed version. Regressions in meson are common. DPDK 22.11 just bumped the minimum version to 0.53 (.2 because of a regression in x86 cross builds). As for your ask here, I tried meson 0.53.2 + python 3.10 against my dpdk-latest branch, and it seems to work (waiting for all jobs to finish..). https://github.com/david-marchand/ovs/actions/runs/3234859080/jobs/5298498558 -- David marchand _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
