> On Jun 22, 2015, at 10:25 AM, Sorin Vinturis > <svintu...@cloudbasesolutions.com> wrote: > > Hi Eitan/Nithin, > > In normal execution flow all WFP objects are released after the extension is > disabled. The patch was meant to address the cases when the extension was not > properly disabled due to some unforeseen error, which does not deletes all > WFP objects like the OVS sublayer or callout. Many times I have seen an > improper activation of the OVS extension because the sublayer or callout > already existed. > > However I have not been able to reproduce the problem with the latest code > from the master branch. The extension is much more stable with the latest > patches and I cannot provide reproduction steps for this issue.
Sorin, Would it be possible to add some instrumentation code and force the EEXIST error? That would be one way to test things out, IMO. Eitan may have other ideas. thanks, -- Nithin _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
