[Intel-gfx] [PATCH 1/2] drm/i915: Do a dummy DPCD read before the actual read

[Todd Previte]

On 10/17/2014 1:59 AM, Ville Syrjälä wrote:
> On Fri, Oct 17, 2014 at 11:43:21AM +0300, Jani Nikula wrote:
>> On Thu, 16 Oct 2014, ville.syrjala at linux.intel.com wrote:
>>> From: Ville Syrjälä <ville.syrjala at linux.intel.com>
>>>
>>> Sometimes we seem to get utter garbage from DPCD reads. The resulting
>>> buffer is filled with the same byte, and the operation completed without
>>> errors. My HP ZR24w monitor seems particularly susceptible to this
>>> problem once it's gone into a sleep mode.
>>>
>>> The issue seems to happen only for the first AUX message that wakes the
>>> sink up. But as the first AUX read we often do is the DPCD receiver
>>> cap it does wreak a bit of havoc with subsequent link training etc. when
>>> the receiver cap bw/lane/etc. information is garbage.
>> This makes me suspect our sink dpms and wake handling even more than I
>> already did. Someone(tm) should dig into the DP and hw specs again with
>> fresh eyes...
> Yeah when I last looked at the spec it was rather vague on the subject.
> On the other hand it seems to suggest that you're supposed to wake up
> the sink via DP_SET_POWER before any other AUX transfer, but on the
> other hand it seems to say AUX is fine as long as you remember to do
> the extended retry in case the AUX circuitry was asleep in the sink.

The sink is supposed to exit power-down mode within 1ms of detecting a 
differential signal voltage on the AUX lines, according to the spec. So 
in theory, any AUX transactions should be able to wake up the sink 
device. As you pointed out, the AUX retries will catch the case where 
the AUX channel is powered down. I know of one instance where the write 
to SET_POWER is required, and that's in one of the compliance tests. 
Outside of that though, I haven't seen anything where it's mandatory.

-T

>
> However DPCD 1.0 doesn't even support DP_SET_POWER so that does suggest
> that it's not really mandatory to wake things up first. Well, assuming
> DPCD 1.0 devices even exist.
>
> I was a bit suspicious of our AUX code as well, but IIRC didn't spot
> anything obviously incorrect there when I had a look. So might be this
> is just the sink being a bit buggy.
>
>>> A sufficient workaround seems to be to perform a single byte dummy read
>>> before reading the actual data. I suppose that just wakes up the sink
>>> sufficiently and we can just throw away the returned data in case it's
>>> crap. DP_DPCD_REV seems like a sufficiently safe location to read here.
>> Seems like a pretty harmless thing to do, and we already do loads of
>> dpcd reads anyway. We should throw this at some related bugs.
>>
>> So ack.
>>
>> BR,
>> Jani.
>>
>>
>>> Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
>>> ---
>>>   drivers/gpu/drm/i915/intel_dp.c | 7 +++++++
>>>   1 file changed, 7 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
>>> index 64c8e04..f07f02c 100644
>>> --- a/drivers/gpu/drm/i915/intel_dp.c
>>> +++ b/drivers/gpu/drm/i915/intel_dp.c
>>> @@ -2870,6 +2870,13 @@ intel_dp_dpcd_read_wake(struct drm_dp_aux *aux, unsigned int offset,
>>>   	ssize_t ret;
>>>   	int i;
>>>   
>>> +	/*
>>> +	 * Sometime we just get the same incorrect byte repeated
>>> +	 * over the entire buffer. Doing just one throw away read
>>> +	 * initially seems to "solve" it.
>>> +	 */
>>> +	drm_dp_dpcd_read(aux, DP_DPCD_REV, buffer, 1);
>>> +
>>>   	for (i = 0; i < 3; i++) {
>>>   		ret = drm_dp_dpcd_read(aux, offset, buffer, size);
>>>   		if (ret == size)
>>> -- 
>>> 2.0.4
>>>
>>> _______________________________________________
>>> Intel-gfx mailing list
>>> Intel-gfx at lists.freedesktop.org
>>> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>> -- 
>> Jani Nikula, Intel Open Source Technology Center