On Tue, Sep 24, 2013 at 12:11:27PM +0100, Chris Wilson wrote: > On Tue, Sep 24, 2013 at 12:47:07PM +0200, Daniel Vetter wrote: > > The idea that was floated was to use the reg/unreg functions of > > DEFINE_TRACE_FN to do the irq get/put dance. The problem seems to be > > that doing fancy stuff from within the tracepoint itself isn't awesome > > for -rt locking. So we don't enable the interrupts when we don't need > > them, but only when the tracepoint is active. > > And besides the complaint was about trace_irq_seqno, was it not? It's the ->get_irq call hidden in i915_trace_irq_get within a TP_fast_assign clause. The racy business around accessing ring->trace_irq_seqno is a bit orthogonal I think. > > The slight problem is that I don't see any users of this stuff, so I > > think we can still go meh. > > Indeed, I do not know how to do what he wants under the confines of the > current tracepoint API. Hm, maybe I need to yell at him again. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
