On Thu, Sep 04, 2014 at 04:12:43PM +0200, Daniel Vetter wrote: > On Thu, Sep 04, 2014 at 02:33:44PM +0100, Chris Wilson wrote: > > On Thu, Sep 04, 2014 at 03:31:14PM +0200, Daniel Vetter wrote: > > > Didn't 3.17 just gain the mod_delayed_work interface? > > > > Yes. It was so exciting and doesn't appear to requeue. > > I guess we could do the mod_delayed_work in add_request and use the > schedule_delayed_work everywhere in i915_irq.c then? > > But when exactly does mod_delayed_work not requeue? Documentation at > least claims it does ... I think the first version I tested used relative jiffies in mod_delayed_work, but I'm not buying that we actually ever want to push the first hangcheck back simply because we are queueing more work... We can queue work right up until the we fill the ring behind a hung task in that case. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
