<div class="gmail_quote">On Wed, Jan 4, 2012 at 14:52, Daniel Vetter <<a href="mailto:daniel.vetter@ffwll.ch">daniel.vetter@ffwll.ch</a>> wrote: <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Two things seem to do the trick on my ivb machine here: - prevent the gt from powering down while waiting for seqno notification interrupts by grabbing the force_wake in get_irq (and dropping it in put_irq again). - ordering writes from the ring's CS by reading a CS register, ACTHD seems to work. Only the blt&bsd ring on ivb seem to be massively affected by this, but for paranoia do this dance also on the render ring and on snb (i.e. all gpus with forcewake). Tested with Eric's glCopyPixels loop which without this patch scores a missed irq every few seconds. This patch needs my forcewake rework to use a spinlock instead of dev->struct_mutex. Cc: <a href="mailto:stable@kernel.org">stable@kernel.org</a> Cc: Eric Anholt <<a href="mailto:eric@anholt.net">eric@anholt.net</a>> Cc: Kenneth Graunke <<a href="mailto:kenneth@whitecape.org">kenneth@whitecape.org</a>> Cc: Eugeni Dodonov <<a href="mailto:eugeni.dodonov@intel.com">eugeni.dodonov@intel.com</a>> Signed-Off-by: Daniel Vetter <<a href="mailto:daniel.vetter@ffwll.ch">daniel.vetter@ffwll.ch</a>> </blockquote><div> This voodoo magic works here, no more missed irqs on my IVBs. Tested-by: Eugeni Dodonov <<a href="mailto:eugeni.dodonov@intel.com">eugeni.dodonov@intel.com</a>> Reviewed-by: Eugeni Dodonov <<a href="mailto:eugeni.dodonov@intel.com">eugeni.dodonov@intel.com</a>> </div><blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> static u32 +gen6_ring_get_seqno(struct intel_ring_buffer *ring) +{ + /* Workaround for force correct ordering between irq and seqno writes on + * ivb (and maybe also on snb). */ + intel_ring_get_active_head(ring); + return intel_read_status_page(ring, I915_GEM_HWS_INDEX); </blockquote><div> <bikeshedding> - the comment could be rewritten to "Workaround for force correct ordering between irq and seqno writes on ivb (and maybe also on snb) by reading a CS register, like ACTHD, prior reading status page". - we could do a 'return ring_get_seqno(ring);' instead of 'return intel_read_status_page(ring, I915_GEM_HWS_INDEX);' to reduce the Universe entropy :). </div></div></bikeshedding> -- Eugeni Dodonov<a href="http://eugeni.dodonov.net/" target="_blank"> </a>