LPSP igt test on TGL

Above block diagram depicts lpsp on TGL, while earlier development of lpsp igt test it has been assumed that every non-edp panel is not a lpsp panel but it is not true on TGL, any HDMI/DP panel connected on pipe A and connected to PORT_{A,B,C} is lpsp panel, and DSI panel is also lpsp panel.

Currently i915_pm_lpsp igt has been categorized in edp-native and non-edp subtest. non-edp subtest require the fix to validate the lpsp panels.
My solution approach to check the port on which DP/HDMI connector is currently connected.
if (port <= PORT_C)
lpsp_should_be_enabled.
else
lpsp_should_be_disabled.

Currently there are two igt i915_pm_lpsp, i915_pm_rpm require to know the port of DP/HDMI connector, so we need a igt library function for that.

I am looking to parse the i915_display_info to get the port information, but at same time it seems a fragile solution to me as any change in i915_display_info will break the test, but i don't have any other idea to get the port information.

Please provide your opinion and feedback for the above solution approach.

Quoted 24 lines · Anshuman Gupta

Above block diagram depicts lpsp on TGL, while earlier development of lpsp igt test it has been assumed that every non-edp panel is not a lpsp panel but it is not true on TGL, any HDMI/DP panel connected on pipe A and connected to PORT_{A,B,C} is lpsp panel, and DSI panel is also lpsp panel.

Currently i915_pm_lpsp igt has been categorized in edp-native and non-edp subtest. non-edp subtest require the fix to validate the lpsp panels.
My solution approach to check the port on which DP/HDMI connector is currently connected.
if (port <= PORT_C)
lpsp_should_be_enabled.
else
lpsp_should_be_disabled.

Currently there are two igt i915_pm_lpsp, i915_pm_rpm require to know the port of DP/HDMI connector, so we need a igt library function for that.

I am looking to parse the i915_display_info to get the port information, but at same time it seems a fragile solution to me as any change in i915_display_info will break the test, but i don't have any other idea to get the port information.

Quoted 1 line

Please provide your opinion and feedback for the above solution approach.

I'd add a connector specific debugfs file that tells you whether LPSS (Low Power Single Pipe, for those who want to know what the acronym means) is possible and whether it's enabled. Alternatively add the info about LPSS to i915_display_info. But AFAICT there's no need for userspace to know about the port. This also avoids reading the HSW_PWR_WELL_CTL2 register directly.

BR,
Jani.

Signature · 1 line

Jani Nikula, Intel Open Source Graphics Center

Quoted 25 lines · Jani Nikula

Quoted 24 lines · Anshuman Gupta

Above block diagram depicts lpsp on TGL, while earlier development of lpsp igt test it has been assumed that every non-edp panel is not a lpsp panel but it is not true on TGL, any HDMI/DP panel connected on pipe A and connected to PORT_{A,B,C} is lpsp panel, and DSI panel is also lpsp panel.

Currently i915_pm_lpsp igt has been categorized in edp-native and non-edp subtest. non-edp subtest require the fix to validate the lpsp panels.
My solution approach to check the port on which DP/HDMI connector is currently connected.
if (port <= PORT_C)
lpsp_should_be_enabled.
else
lpsp_should_be_disabled.

Currently there are two igt i915_pm_lpsp, i915_pm_rpm require to know the port of DP/HDMI connector, so we need a igt library function for that.

I am looking to parse the i915_display_info to get the port information, but at same time it seems a fragile solution to me as any change in i915_display_info will break the test, but i don't have any other idea to get the port information.

Quoted 8 lines

Quoted 1 line

Please provide your opinion and feedback for the above solution approach.

I'd add a connector specific debugfs file that tells you whether LPSS (Low Power Single Pipe, for those who want to know what the acronym means) is possible and whether it's enabled. Alternatively add the info about LPSS to i915_display_info. But AFAICT there's no need for userspace to know about the port. This also avoids reading the HSW_PWR_WELL_CTL2 register directly.

Thanks Jani and Ville for your feedback and suggestion, based upon your feedback i have following solution approach.
Add a connector specific debugfs attributes i915_lpsp_cpable.
for ecah connector if it is lpsp capable, do modeset with pipe A crtc on that connector and validate if power wells above pg2 are turned off by using i915_power_domain_info.

Thanks ,
Anshuman

Quoted 7 lines

BR,
Jani.

Signature · 1 line

Jani Nikula, Intel Open Source Graphics Center

Quoted 34 lines · Anshuman Gupta

Quoted 25 lines · Jani Nikula

Quoted 24 lines · Anshuman Gupta

Above block diagram depicts lpsp on TGL, while earlier development of lpsp igt test it has been assumed that every non-edp panel is not a lpsp panel but it is not true on TGL, any HDMI/DP panel connected on pipe A and connected to PORT_{A,B,C} is lpsp panel, and DSI panel is also lpsp panel.

Currently i915_pm_lpsp igt has been categorized in edp-native and non-edp subtest. non-edp subtest require the fix to validate the lpsp panels.
My solution approach to check the port on which DP/HDMI connector is currently connected.
if (port <= PORT_C)
lpsp_should_be_enabled.
else
lpsp_should_be_disabled.

Currently there are two igt i915_pm_lpsp, i915_pm_rpm require to know the port of DP/HDMI connector, so we need a igt library function for that.

I am looking to parse the i915_display_info to get the port information, but at same time it seems a fragile solution to me as any change in i915_display_info will break the test, but i don't have any other idea to get the port information.

Quoted 8 lines

Quoted 1 line

Please provide your opinion and feedback for the above solution approach.

I'd add a connector specific debugfs file that tells you whether LPSS (Low Power Single Pipe, for those who want to know what the acronym means) is possible and whether it's enabled. Alternatively add the info about LPSS to i915_display_info. But AFAICT there's no need for userspace to know about the port. This also avoids reading the HSW_PWR_WELL_CTL2 register directly.

Quoted 6 lines

Thanks Jani and Ville for your feedback and suggestion, based upon your feedback i have following solution approach.
Add a connector specific debugfs attributes i915_lpsp_cpable.
for ecah connector if it is lpsp capable, do modeset with pipe A crtc on that connector and validate if power wells above pg2 are turned off by using i915_power_domain_info.

Why not add a debugfs file that will tell you both, instead of forcing the userspace to know about the connection between lpsp and pg2?

BR,
Jani.

Quoted 4 lines

Thanks ,
Anshuman

Quoted 1 line

Quoted 6 lines

Quoted 6 lines

BR,
Jani.

Signature · 1 line

Jani Nikula, Intel Open Source Graphics Center

Signature · 1 line

Jani Nikula, Intel Open Source Graphics Center

Quoted 3 lines · Jani Nikula

Quoted 2 lines · Anshuman Gupta

Quoted 1 line · Jani Nikula

On Thu, 27 Feb 2020, Anshuman Gupta <[email protected]> wrote:

Quoted 7 lines

Quoted 4 lines

Add a connector specific debugfs attributes i915_lpsp_cpable. for ecah connector if it is lpsp capable, do modeset with pipe A crtc on that connector and validate if power wells above pg2 are turned off by using i915_power_domain_info.

Why not add a debugfs file that will tell you both, instead of forcing the userspace to know about the connection between lpsp and pg2?

IMHO a DP/HDMI connector may not be connected to pipe A crtc in default mode.

So via igt we can have modeset on Pipe A with DP/HDMI connector and can validate lpsp for DP/HDMI connector.
Please correct me if am wrong.

Quoted 4 lines · Anshuman Gupta

Quoted 3 lines · Jani Nikula

Quoted 2 lines · Anshuman Gupta

Quoted 1 line · Jani Nikula

On Thu, 27 Feb 2020, Anshuman Gupta <[email protected]> wrote:

Quoted 8 lines

Quoted 7 lines

Quoted 4 lines

Add a connector specific debugfs attributes i915_lpsp_cpable. for ecah connector if it is lpsp capable, do modeset with pipe A crtc on that connector and validate if power wells above pg2 are turned off by using i915_power_domain_info.

Why not add a debugfs file that will tell you both, instead of forcing the userspace to know about the connection between lpsp and pg2?

IMHO a DP/HDMI connector may not be connected to pipe A crtc in default mode.

What is default mode?

Quoted 3 lines

So via igt we can have modeset on Pipe A with DP/HDMI connector and can validate lpsp for DP/HDMI connector.
Please correct me if am wrong.

Sorry, I am not sure what you're asking.

BR,
Jani.

Signature · 1 line

Jani Nikula, Intel Open Source Graphics Center

+-------------------------------------------------------------------------------+
|                                                                               |
|                                                                               |
|                                                                               |
|   +--------------------+        +-------------+         +------------+        |
|   |                    |        |             |         |            |        |
|   |    PIPE A          |        | DP/HDMI     |         |   PORT_A   |        |
|   |                    +--------+ TRANSCODER  +---------+            |        |
|   |                    |        |             |         |            |        |
|   +--------------------+        +------------->         +------------>        |
|                                 |             |         |            |        |
|                                 |             |         |   PORT_B   |        |
|                                 |             |         |            |        |
|                                 |             |         |            |        |
|                                 +------------->         +------------>        |
|                                 |             |         |            |        |
|                                 | DSI         |         |   PORT_C   |        |
|                                 | TRANSCODER  |         |            |        |
|                                 +------------->         +------------+        |
|                                 |             |                               |
|                                 |             |                               |
|                                 |             |                               |
|                                 |             |                               |
|                                 |             |                               |
|                                 |             |                               |
|                                 |             |                               |
|                                 +-------------+                               |
|                                                                               |
|                   LPSP on Tigerlake                                           |
|                                                                               |
+-------------------------------------------------------------------------------+

Above block diagram depicts lpsp on TGL, while earlier development of lpsp igt test it has been assumed that every non-edp panel is not a lpsp panel but it is not true on TGL,