[Intel-gfx] Power saving using Display port HPD

[Jani Nikula]

On Thu, 10 Apr 2014, Arun Chandran <achandran at mvista.com> wrote:
> I have connected only a DP monitor on Bayley Bay platform and booting
> ubuntu 13.10 with the kernel
> http://cgit.freedesktop.org/drm-intel/log/?h=drm-intel-fixes.

Most of the time it's probably better to try either Linus' master or
drm-intel-nightly for latest stuff. Anything in drm-intel-fixes is
always included in -nightly, and merged to Linus' master with fairly
short intervals.

> I am experimenting with the HPD feature of display port with debug
> messages enabled. (drm.debug=0xe).
>
> When I plug in/out DP cable I am able to see the below kernel messages
> and ubuntu desktop is displayed on the DP monitor
>
> ##############################
>
>
> root at arun-valley-view:~# uname -a
> Linux arun-valley-view 3.14.0-00528-g67e9d82 #34 SMP PREEMPT Wed Apr 9
> 18:48:30 IST 2014 i686 i686 i686 GNU/Linux
>
> root at arun-valley-view:~# dmesg -c
>
> [ 1028.506188] [drm:intel_hpd_irq_handler] hotplug event received,
> stat 0x00100000
> [ 1028.506217] [drm:intel_hpd_irq_handler] Received HPD interrupt on
> PIN 5 - cnt: 0
> [ 1028.513858] [drm:i915_hotplug_work_func] running encoder hotplug functions
> [ 1028.513880] [drm:i915_hotplug_work_func] Connector eDP-1 (pin 5)
> received hotplug event.
> [ 1028.513896] [drm:i915_hotplug_work_func] Connector HDMI-A-2 (pin 5)
> received hotplug event.
> [ 1028.513913] [drm:_edp_panel_vdd_on] Turning eDP VDD on
> [ 1028.513931] [drm:_edp_panel_vdd_on] PP_STATUS: 0x80000008
> PP_CONTROL: 0xabcd000f
> [ 1028.515064] [drm:intel_dp_get_dpcd] DPCD: 11 0a 84 01 01 00 01 80
> 02 00 00 00 00 00 00
> [ 1028.516563] [drm:intel_dp_check_link_status] TMDS-22: channel EQ
> not ok, retraining
> [ 1028.530884] [drm:intel_dp_set_signal_levels] Using signal levels 00000000
> [ 1028.546113] [drm:intel_dp_set_signal_levels] Using signal levels 00000000
> [ 1028.547233] [drm:intel_dp_start_link_train] clock recovery OK
> [ 1028.548711] [drm:intel_dp_complete_link_train] Channel EQ done. DP
> Training successful
> [ 1028.549264] [drm:intel_dp_detect] [CONNECTOR:23:eDP-1]
> [ 1028.549761] [drm:intel_dp_probe_oui] Sink OUI: 00e04c
> [ 1028.550262] [drm:intel_dp_probe_oui] Branch OUI: 000000
> [ 1028.550343] [drm:intel_hdmi_detect] [CONNECTOR:21:HDMI-A-2]
> [ 1028.550595] [drm:gmbus_xfer] GMBUS [i915 gmbus dpc] NAK for addr: 0050 r(1)
> [ 1028.550611] [drm:drm_do_probe_ddc_edid] drm: skipping non-existent
> adapter i915 gmbus dpc
>
> ##################################
>
> After booting I am executing the below steps
>
> 1) unplug DP cable.
>
> 2) Mask the pin 18 (HPD) of DP cable using a paper strip then connect it
> back
>     I have checked the dmesg output and the there is no messages related to
> HPD. So kernel is
>     unaware that I have connected a DP monitor
>
> But still I see ubuntu desktop coming on the DP monitor.  This means kernel
> does not stop pumping display data after step 1.
>
> Is this the right behavior? . When kernel knows that the last existing
> display is removed (via the HPD feature) it can save power by stopping the
> data output to DP monitor (or shutting down the DP hardware related part.)
>
> Does anybody knows how to use the HPD feature to save power? Or anybody has
> done this before?

First things first, let's confirm what's actually happening. Do a fresh
boot, and do your steps above. Annotate the dmesg with 'echo "step N" >
/dev/kmsg' as root between steps, or split the dmesg at each step. Make
sure your final dmesg(s) cover unplugging, plugging, re-plugging.

I suspect the DP output does get switched off, and your re-plugging with
HDP masked goes wrong, actually triggering another hotplug. But let's
confirm that with the dmesgs.

BR,
Jani.

-- 
Jani Nikula, Intel Open Source Technology Center