[ovs-dev] [threaded-learning 14/25] ofproto: Move rule_execute() out of ofopgroup_complete().

Ben Pfaff blp at nicira.com
Thu Sep 12 07:25:24 UTC 2013


On Wed, Sep 11, 2013 at 07:17:18PM -0700, Ethan Jackson wrote:
> I posted this review on the wrong patch, but here it is on the right one:
> 
> FWIW I think this is the right choice given the options.  May be worth
> adding a comment in the actual code as to why we can't just execute
> the rule and have to queue it instead.

Thanks.

I changed the comment to this.  I'm not sure it's entirely coherent (I'm
writing this past midnight, pretty late for me):

    /* Delayed rule executions.
     *
     * We delay calls to ->ofproto_class->rule_execute() past releasing
     * ofproto_mutex during a flow_mod, because otherwise a "learn" action
     * triggered by the executing the packet would try to recursively modify
     * the flow table and reacquire the global lock. */

> Acked-by: Ethan Jackson <ethan at nicira.com>



More information about the dev mailing list