CVE-2019-0154 caused major power problem, pls backport upstream's fix

[Tim Richardson]

https://gitlab.freedesktop.org/drm/intel/issues/614
the fix released as CVE-2019-0154 increases idle power consumption on intel i915 6th gen and later.
That's because the GPU does not enter RC6 any more.

For example, my Thinkpad P50 (6th gen) idles at 9w with current 5.3 kernels. Previously, it idled at 5W.
Effects on a T480 (8th gen) are less dramatic but still significant.

5.5RC1 has the maintainer's second pass at this, which restores RC6 idling

The maintainer's patch for 5.3 is here:
 https://gitlab.freedesktop.org/drm/intel/issues/614#note_366057

I have applied this to tag Ubuntu-hwe-5.3.0-25.27~18.04.2 and it builds. I am testing it now.

I think it would be very helpful to backport this if you can.

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

This bug is missing log files that will aid in diagnosing the problem. While running an Ubuntu kernel (not a mainline or third-party kernel) please enter the following command in a terminal window:

apport-collect 1856653

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Tim Richardson (tim-richardson)]

logs not applicable

[Tim Richardson (tim-richardson)]

If the power-saving patch is backported,will this be updated? I'm on own patched kernel so I won't notice.
It's an awful problem, by the way, it harms battery life on any laptop with intel graphics from 2016 or later.

[Tim Richardson (tim-richardson)]

PS I've been running on hwe 5.3 from Ubuntu's kernel git repository for a few weeks, with the patch linked above, and it is stable on my Thinkpad T480.

[Kai-Heng Feng (kaihengfeng)]

Is applying snippet [1] alone fixes the issue?

[1] https://gitlab.freedesktop.org/drm/intel/issues/614#note_366057

[Tim Richardson (tim-richardson)]

Yes, that little patch applies cleanly to every 5.3 hwe build I've done (about five I'd say)
To be clear:

diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 82b7ace62d97..b18f281d5d12 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -510,7 +510,11 @@ execlists_user_begin(struct intel_engine_execlists *execlists,
 inline void
 execlists_user_end(struct intel_engine_execlists *execlists)
 {
+ struct intel_engine_cs *engine =
+ container_of(execlists, typeof(*engine), execlists);
+
        execlists_clear_active(execlists, EXECLISTS_ACTIVE_USER);
+ mod_delayed_work(engine->i915->wq, &engine->i915->gem.retire_work, 0);
 }

 static inline void

[Tim Richardson (tim-richardson)]

Also, please note the upstream bug discussion also includes patches for 5.4
It seems that 5.4 is destined for 20.04. So far 5.4 upstream is still using the first release of this patch, which means 5.4 has the i915 RC6 bug. 5.4 is a LTS kernel bug so it strange to me, but please, I hope you can consider making sure that 20.04 does not launch with this nasty power regression.

[Tim Richardson (tim-richardson)]

Confirming that a fix has been backported to 20.04, currently in the proposed kernel. I'm testing it, the power usage is great.

[Tim Richardson (tim-richardson)]

not 'a fix', the intel maintainer's 5.4 patch has been applied by the ubuntu devs.