[Intel-gfx] [PATCH 2/2] drm/i915/mst: Use MST sideband message transaction for dpms

Wed Sep 13 07:32:40 UTC 2017

Op 12-09-17 om 22:11 schreef Pandiyan, Dhinakaran:
>
>
> On Tue, 2017-09-12 at 19:08 +0000, Pandiyan, Dhinakaran wrote:
>> On Tue, 2017-09-12 at 19:17 +0300, Ville Syrjälä wrote:
>>> On Tue, Sep 12, 2017 at 07:11:32PM +0300, Ville Syrjälä wrote:
>>>> On Tue, Sep 05, 2017 at 06:26:34PM -0700, Dhinakaran Pandiyan wrote:
>>>>> Use the POWER_DOWN_PHY and POWER_UP_PHY sideband message trasactions to
>>>>> set power states for downstream sinks. Apart from giving us the ability
>>>>> to set power state for individual sinks, this fixes the below test for
>>>>> me
>>>>>
>>>>> $ xrandr --display :0 --output DP-2-2-8 --off
>>>>> $ xrandr --display :0 --output DP-2-2-1 --off
>>>>> $ xrandr --display :0 --output DP-2-2-8 --auto #Black screen
>>>>> $ xrandr --display :0 --output DP-2-2-1 --auto
>>>>>
>>>>> Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
>>>>> Cc: Lyude <lyude at redhat.com>
>>>>> Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
>>>>> ---
>>>>>  drivers/gpu/drm/i915/intel_ddi.c    | 6 ++++--
>>>>>  drivers/gpu/drm/i915/intel_dp_mst.c | 8 ++++----
>>>>>  2 files changed, 8 insertions(+), 6 deletions(-)
>>>>>
>>>>> diff --git a/drivers/gpu/drm/i915/intel_ddi.c b/drivers/gpu/drm/i915/intel_ddi.c
>>>>> index 1da3bb2cc4b4..8aebacc0aa31 100644
>>>>> --- a/drivers/gpu/drm/i915/intel_ddi.c
>>>>> +++ b/drivers/gpu/drm/i915/intel_ddi.c
>>>>> @@ -2161,7 +2161,8 @@ static void intel_ddi_pre_enable_dp(struct intel_encoder *encoder,
>>>>>  		intel_prepare_dp_ddi_buffers(encoder);
>>>>>  
>>>>>  	intel_ddi_init_dp_buf_reg(encoder);
>>>>> -	intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON);
>>>>> +	if (!link_mst)
>>>>> +		intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_ON);
>>>>>  	intel_dp_start_link_train(intel_dp);
>>>>>  	if (port != PORT_A || INTEL_GEN(dev_priv) >= 9)
>>>>>  		intel_dp_stop_link_train(intel_dp);
>>>>> @@ -2240,7 +2241,8 @@ static void intel_ddi_post_disable(struct intel_encoder *intel_encoder,
>>>>>  	if (type == INTEL_OUTPUT_DP || type == INTEL_OUTPUT_EDP) {
>>>>>  		struct intel_dp *intel_dp = enc_to_intel_dp(encoder);
>>>>>  
>>>>> -		intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_OFF);
>>>>> +		if (old_crtc_state && old_conn_state)
>>>>> +			intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_OFF);
>>>> I would make that
>>>> !intel_crtc_has_type(DP_MST)
>>>>
> old_crtc_state, which intel_crtc_has_type() needs, is NULL for MST and
> the reason Maarten provided in his commit is
>
> " We also shouldn't pass crtc_state, because in that case the
>   disabling sequence may potentially be different depending on
>   which crtc is disabled last. Nice way to introduce bugs.
> "
> I am not 100% sure I understand the concern. But since that change was
> intentional I guess we can use the NULL values, like I've done, to
> identify the mst sinks. I am open to ideas.
I think checking for NULL crtc_state is enough, in case we ever decide to pass the real connector state.
For clarity, I would add something like bool is_dsp_mst = !crtc_state; /* When enabling the link for MST, this connector is never bound to a crtc */.
>
>
>>>>>  	}
>>>>>  
>>>>>  	val = I915_READ(DDI_BUF_CTL(port));
>>>>> diff --git a/drivers/gpu/drm/i915/intel_dp_mst.c b/drivers/gpu/drm/i915/intel_dp_mst.c
>>>>> index 8e3aad0ea60b..81e63724e24b 100644
>>>>> --- a/drivers/gpu/drm/i915/intel_dp_mst.c
>>>>> +++ b/drivers/gpu/drm/i915/intel_dp_mst.c
>>>>> @@ -167,12 +167,11 @@ static void intel_mst_post_disable_dp(struct intel_encoder *encoder,
>>>>>  	intel_dp->active_mst_links--;
>>>>>  
>>>>>  	intel_mst->connector = NULL;
>>>>> -	if (intel_dp->active_mst_links == 0) {
>>>>> +	if (intel_dp->active_mst_links == 0)
>>>>>  		intel_dig_port->base.post_disable(&intel_dig_port->base,
>>>>>  						  NULL, NULL);
>>>>> -
>>>>> -		intel_dp_sink_dpms(intel_dp, DRM_MODE_DPMS_OFF);
>>>>> -	}
>>>>> +	drm_dp_send_power_updown_phy(&intel_dp->mst_mgr, connector->port,
>>>>> +				     false);
>>>> One thing that bothers me here is whether we'll start getting bogus link
>>>> training requests during shutdown since we'll do the D3 only after the
>>>> stream has been stopped. If you'll look at commit 7618138d8b40
>>>> ("drm/i915/ddi: Avoid long delays during system suspend / eDP
>>>> disabling") Imre explicitly moved the D3 earlier for SST to avoid this.
>>>>
>> Interesting, that patch makes a lot of sense.
>>
>>>> Hmm. Actually by the time we call mst_post_disable() we should have
>>>> deallocated the VC etc. so the current stream should be already
>>>> terminated at that point. So I think we should be able to do the
>>>> PHY_POWER_DOWN before we shut down the link, just like we do for SST.
>>> Might also be nice to add comments near all the D3 calls to document
>>> that the order we use is there to prevent the sink from complaining
>>> about link loss when we shut down the link. Otherwise someone might
>>> get tempted to reorder things as a cleanup or something.
>>>
>> Sure.
>>
>>>>>  }
>>>>>  
>>>>>  static void intel_mst_pre_enable_dp(struct intel_encoder *encoder,
>>>>> @@ -197,6 +196,7 @@ static void intel_mst_pre_enable_dp(struct intel_encoder *encoder,
>>>>>  
>>>>>  	DRM_DEBUG_KMS("%d\n", intel_dp->active_mst_links);
>>>>>  
>>>>> +	drm_dp_send_power_updown_phy(&intel_dp->mst_mgr, connector->port, true);
>>>>>  	if (intel_dp->active_mst_links == 0)
>>>>>  		intel_dig_port->base.pre_enable(&intel_dig_port->base,
>>>>>  						pipe_config, NULL);
>>>>> -- 
>>>>> 2.11.0
>>>> -- 
>>>> Ville Syrjälä
>>>> Intel OTC
>>>> _______________________________________________
>>>> Intel-gfx mailing list
>>>> Intel-gfx at lists.freedesktop.org
>>>> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
>> _______________________________________________
>> Intel-gfx mailing list
>> Intel-gfx at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/intel-gfx