On Thu, Feb 12, 2015 at 11:30:10PM -0600, Mike Christie wrote: > On 02/12/2015 06:33 PM, Chris Leech wrote: > > It looks like the communication with iscsiuio has a similar case where > > a polling function reschedules itself, following up with a patch to fix > > the delay there. > > Have you been able to hit that iscsiuio reschedule code path? There was > a bug where we could call actor_timer multiple times on undeleted timer. > I made the attached patch (made over github tree).
Sorry for the delay on this. I hadn't hit that case when I sent the patch, it was purely looking for the same pattern. I resorted to a forced scenerio, disabling iscsiuio and modifying uip_broadcast to return ISCSI_ERR_AGAIN on connection failure, to test the EGAIN handling. That did reproduce the high CPU load immediate polling, after my patch the retry was in 1 second intervals as expected. I'll take a look at your update to address the double actor_timer call to make sure it at least behaves well in the same case. - Chris -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/open-iscsi. For more options, visit https://groups.google.com/d/optout.
