A race condition has been identified during datapath destruction, along with the port offload flushes issued.
This series addresses these race conditions, cleaning up the port deletion process. The last patch also cleans up the offload structure. It is not strictly necessary like the first two fixes, so I put it last. It can wait until after the code-freeze to be integrated. I tested for a few hours without ASAN enabled without seeing issues. ASAN has been executed as part of the github CI: https://github.com/grivet/ovs/actions/runs/1795624401 It is however not too relevant, as no offloads are inserted during CI. The following patch was used to fix an unrelated CI issue: https://patchwork.ozlabs.org/project/openvswitch/patch/20220204150445.1481457-1-i.maxim...@ovn.org/ I also ran datapath creation + deletion loop with ASAN on an offload test setup, but the execution was excruciatingly slow and could not progress much. It reached datapath deletion without panicking and no crash was seen, even though I had to interrupt the test after a few hours. Gaetan Rivet (2): dpif-netdev: Move port flush after datapath reconfiguration dpif-netdev: Use dp_netdev reference in offload threads Sriharsha Basavapatna (1): dpif-netdev: Fix a race condition in deletion of offloaded flows lib/dpif-netdev.c | 71 ++++++++++++++++++++++++++--------------------- 1 file changed, 40 insertions(+), 31 deletions(-) -- 2.31.1 _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
