Skip to contentSkip to conversation

drm/i915: don't change DRM configuration when releasing load detect pipe

DISCUSSION Jesse Barnes 1 participant 2 messages Jul 9, 2010 3 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): 20100218145154.7aa7d557@jbarnes-piketon

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw
Quoted 25 lines · On Mon, 08 Mar 2010 21:59:29 +0100 Brice Goglin <[email protected]> wrote:
Quoted 24 lines · Eric Anholt
Quoted 9 lines · Jesse Barnes

When we get a CRTC to use for load detection, we restore its DPMS state if needed. We shouldn't, however, change the DRM configuration by calling drm_helper_disable_unused_functions when we release the CRTC. Doing so can cause problems with resume, since at suspend or lid close time, X may choose to probe outputs. If it doesn't re-probe them at open or resume time, LVDS won't be restored, since drm_helper_disable_unused_functions will have turned it off, preventing the mode set at lid open from restoring it.

It seems to me like the bug is that drm_helper_disable_unused_functions whacks the configuration. Why should it do that?

		if (connector->status == connector_status_disconnected)
			connector->encoder = NULL;

This function back in UMS used to be about getting the system from unknown bootup state to other junk being turned off after the initial modeset (which was of course broken, since having the other junk turned on during modeset could break things, so really it should have been intel_turn_everything_off_thanks() before doing a modeset).

Quoted 8 lines
Quoted 4 lines

Once that's fixed, it seems like it'd be a replacement for the turn-off-after-load-detect logic in the load detect pipe release function.

Did anything get applied to fix this issue in the end ? Jesse's initial patch fixed blank screens on my i 945 so I am still applying it to my kernels for now.

Arg, totally dropped this one.

Dave, please comment on the above.

Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center
#4 lore raw
Quoted 33 lines · On Fri, 9 Jul 2010 08:30:11 -0700 Jesse Barnes <[email protected]> wrote:
Quoted 30 lines · On Mon, 08 Mar 2010 21:59:29 +0100 Brice Goglin <[email protected]> wrote:
Quoted 28 lines · Eric Anholt
Quoted 9 lines · Jesse Barnes

When we get a CRTC to use for load detection, we restore its DPMS state if needed. We shouldn't, however, change the DRM configuration by calling drm_helper_disable_unused_functions when we release the CRTC. Doing so can cause problems with resume, since at suspend or lid close time, X may choose to probe outputs. If it doesn't re-probe them at open or resume time, LVDS won't be restored, since drm_helper_disable_unused_functions will have turned it off, preventing the mode set at lid open from restoring it.

It seems to me like the bug is that drm_helper_disable_unused_functions whacks the configuration. Why should it do that?

		if (connector->status == connector_status_disconnected)
			connector->encoder = NULL;

This function back in UMS used to be about getting the system from unknown bootup state to other junk being turned off after the initial modeset (which was of course broken, since having the other junk turned on during modeset could break things, so really it should have been intel_turn_everything_off_thanks() before doing a modeset).

Once that's fixed, it seems like it'd be a replacement for the turn-off-after-load-detect logic in the load detect pipe release function.

Quoted 7 lines
Quoted 3 lines

Did anything get applied to fix this issue in the end ? Jesse's initial patch fixed blank screens on my i 945 so I am still applying it to my kernels for now.

Arg, totally dropped this one.

Dave, please comment on the above.

I take that back; this shouldn't be required now that we unconditionally return connected from the LVDS detect hook.

Brice, are you seeing something different? I.e. is this patch required for you even on current kernels?

Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center