Bug 105998 - [CI][SKL only] igt@* - dmesg-warn - *ERROR* failed to enable link training
Summary: [CI][SKL only] igt@* - dmesg-warn - *ERROR* failed to enable link training
Status: NEW
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: DRI git
Hardware: Other All
: high normal
Assignee: Manasi
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard: ReadyForDev
Keywords:
Depends on:
Blocks:
 
Reported: 2018-04-12 10:18 UTC by Marta Löfstedt
Modified: 2019-01-09 19:31 UTC (History)
3 users (show)

See Also:
i915 platform: SKL
i915 features: display/Other


Attachments
dmesg, drm.debug=0xe (12.67 MB, text/plain)
2018-09-15 15:15 UTC, Chris Murphy
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Marta Löfstedt 2018-04-12 10:18:38 UTC
[  183.898612] [drm:intel_dp_start_link_train [i915]] *ERROR* failed to enable link training
[  183.898698] [drm:intel_dp_start_link_train [i915]] *ERROR* [CONNECTOR:71:eDP-1] Link Training failed at link rate = 270000, lane count = 2

https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_17/fi-skl-6700hq/igt@kms_cursor_crc@cursor-64x64-dpms.html
https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_17/fi-skl-6700hq/igt@kms_atomic_transition@1x-modeset-transitions-nonblocking.html
Comment 2 Francesco Balestrieri 2018-06-01 07:27:53 UTC
Still reproducible.
Comment 3 Chris Murphy 2018-09-15 15:15:43 UTC
Created attachment 141573 [details]
dmesg, drm.debug=0xe

I'm seeing this on

00:02.0 VGA compatible controller [0300]: Intel Corporation Skylake GT2 [HD Graphics 520] [8086:1916] (rev 07) (prog-if 00 [VGA controller])
	Subsystem: Hewlett-Packard Company Device [103c:81a0]


[12446.213836] f29h.local kernel: [drm:intel_dp_start_link_train [i915]] *ERROR* failed to enable link training
[12446.213959] f29h.local kernel: [drm:intel_dp_start_link_train [i915]] [CONNECTOR:79:DP-1] Link Training failed at link rate = 162000, lane count = 4
[12446.214211] f29h.local kernel: [drm:intel_dp_modeset_retry_work_fn [i915]] [CONNECTOR:79:DP-1]


4.19.0-0.rc3.git2.1.fc30.x86_64 which is git 54eda9df17f3.


So far it's only happening on wake from S3, possibly the connected external display itself hasn't come out of power save at the initial link training? Subsequent link training appears to be OK.

On the primary display at about this time (hard to say), there is a flicker at GDM (GNOME login). I can't tell if it's related.
Comment 4 Martin Peres 2018-09-21 13:34:24 UTC
(In reply to Chris Murphy from comment #3)
> Created attachment 141573 [details]
> dmesg, drm.debug=0xe
> 
> I'm seeing this on
> 
> 00:02.0 VGA compatible controller [0300]: Intel Corporation Skylake GT2 [HD
> Graphics 520] [8086:1916] (rev 07) (prog-if 00 [VGA controller])
> 	Subsystem: Hewlett-Packard Company Device [103c:81a0]
> 
> 
> [12446.213836] f29h.local kernel: [drm:intel_dp_start_link_train [i915]]
> *ERROR* failed to enable link training
> [12446.213959] f29h.local kernel: [drm:intel_dp_start_link_train [i915]]
> [CONNECTOR:79:DP-1] Link Training failed at link rate = 162000, lane count =
> 4
> [12446.214211] f29h.local kernel: [drm:intel_dp_modeset_retry_work_fn
> [i915]] [CONNECTOR:79:DP-1]
> 
> 
> 4.19.0-0.rc3.git2.1.fc30.x86_64 which is git 54eda9df17f3.
> 
> 
> So far it's only happening on wake from S3, possibly the connected external
> display itself hasn't come out of power save at the initial link training?
> Subsequent link training appears to be OK.
> 
> On the primary display at about this time (hard to say), there is a flicker
> at GDM (GNOME login). I can't tell if it's related.

Thanks for this detailed report Chris, I am pulling in Manasi who has been developing this feature. She should be able to tell us for sure if your sensible explanation is sound, and how we could fix this.

However, I would recommend writing a new bug report, given that not all link training failures stem from the same root cause. Since our failing tests are not S3-related, you likely experience a separate issue that is best handled in a separate bug.

In any case, how much impact this issue has for you? This is important for us to prioritise bugs.
Comment 5 Manasi 2018-10-02 21:59:02 UTC
Looking at the logs, this link failure is happening because of the aux timeouts before the link training. Meaning that the AUX power domain is not on at that point.
What test case are you running to put trigger S3 state and waking it up from S3?
Would be interesting to try this with i915.disable_power_well=0 set for the kernel parameters or manually setting this in i915_params.c and rebuilding the kernel.
Let me know if that fixes it.

Manasi
Comment 6 Manasi 2019-01-09 19:31:44 UTC
Did you get a chance to try out what I had suggesting in my earlier comment?
Do you still see this issue on drm-tip?

Manasi


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.