Can you please submit the exact same change for drm/i915/intel_sdvoc.c too? The sdvo encoders can be the same ones for psb and i915 so having the same code would be great. Separate patch for drm/i915 ofc.
Are you absolutely sure that is a hardware error and not a software SDVO protocol goof? TARGET_NOT_SPECIFIED seems pretty specific.
No idea, but since the code still retries a limited amount and then still fails I can imagine that a few sdvo encoders just have a hard time waking up and responding with something useful. In all cases we log the actual return value and everything interesting.
I also have no idea if we're actually fixing a symptom instead of a real unrelated problem. What I know is that when I first reported this problem last year, Alan Cox told me he had received reports of such problems before, so this affects other Poulsbo users at least.
But I can't know for sure that it's related to hardware or the driver.
Either way, this patch can't make harm as far as I can see, but I cannot test it as I don't have the required hardware.
I'll be Submitting the updated patch set.
- Guillaume