CML - LSPCON mode change timed out after multiple DPMS cycles

A number of employees at my company have been reporting an issue with i915 that appears to still be present on the latest upstream Linux kernel (as of writing this, v5.15-rc7), where after a number of DPMS on/off cycles with an external HDMI display the LSPCON appears to stop responding and die. This issue has been observed on a ThinkPad P1 3rd Gen, and if there's any questions about firmware issues here I should be able to get Lenovo involved.

I've attached a dmesg, note that the dmesg does appear to be truncated - however it still appears to show the LSPCON failure so I don't think it should be an issue. ounsal-dmesg. As well, the CPU here is a i7-10850H so the GPU should be CML.

If you need any more info let me know, and I'll be happy to help.

changed the description

added Community feature: display/LSPCON platform: CML labels

assigned to @raviteja

Is issue still reproducing? on latest drm-tip

I haven't heard anything back from the person who reported this to me originally so I would assume yes. I will double check though. Let me know if there's any info they can give you to help figure this out

I'm running into this issue. Let me know if you need any more information.

Tested against HEAD (ae085d7f9365de7da27ab5c0d16b12d51ea7fca9), will test on drm-tip and update.

The display I can trigger this bug with is 1440p native. If it's plugged in to a gnome-shell session and driven at 1080p things work fine. Switching to 1440p, switching to a TTY and booting with it plugged in all trigger the bug.

Am I right in assuming this is the same laptop as the one mentioned here?

Yes

I do have the same problem on an HP-ENVY 15-dr1977nz. It triggers when an external display is connected via HDMI and my laptop goes to sleeping mode.

Running on Ubuntu 22.04LTS, kernel:5.15.0-27-generic

If anyone has a tip/fix about this :) (and if more infos a needed just ask)

@aknautiyal Please have a look into this issue.

@Leautp Can you share the boot logs, along with the dmesg logs when the issue occur. There are two LSPCON chips with these platforms one from MCA other from Parade. Boot logs will help understand if there is any issue during initialization. Also request you to get the i915_display_info which will help to the F/W version that is flashed. cat /sys/kernel/debug/dri/0/i915_display_info

Hi, here are the requested files.

• boot_log.log
• dmesg_after_issue.log
• i915_display_info.txt

@Leautp For drm logs the drm.debug level should be change in the linux command line. Can you please add drm.debug=0x10e log_buf_len=1M to Linux commandline and then capture the boot and dmesg logs. This will help to get drm logs along with DPCD read and writes right from boot. Also request you to capture i915_display_info when you have connected the display through hdmi cable. Sorry to have missed to share this information earlier.

More about capturing logs and setting drm.debug : https://gitlab.freedesktop.org/drm/intel/-/wikis/How-to-file-i915-bugs More information on drm.debug log levels : https://01.org/linuxgraphics/gfx-docs/drm/gpu/drm-internals.html#c.drm_debug_category

Hi, this should be it. dmesg_after_issue_debug.logboot_log_debug.log i915_display_info_connected.txt

Thanks @Leautp I see from the logs : The LSPCON chip is Parade with FW version : 8.53 The issue is after resume, we are expecting the LSPCON to settle in PCON mode, but it seems to remain in LS mode. This seem to be like an earlier issue https://bugs.freedesktop.org/show_bug.cgi?id=107503. At that time this seem to be fixed by increasing the timeout to 400ms.

https://cgit.freedesktop.org/drm/drm-tip/commit/?id=59f1c8ab30d6f9042562949f42cbd3f3cf69de94

Can you try with increasing the timeout, just to see if the LSPCON chip is taking some time to settle?0001-drm-i915-lspcon-DEBUG-Increase-LsPCON-timeout-to-1-s.patch

@aknautiyal Applied on top of drm-tip and this fixes the issue for me!

@aknautiyal Maybe send the timeout increase for review?

@swick thanks for trying this out. Can you also share the dmesg logs and display info. Also wanted to see if this issue is specific to certain LsPCON and F/W version, is it the latest f/w revision?

@jani I was wondering, what should be right amount of timeout, I had just given an arbitrary value of 1000msec, increasing from 400msec, which itself was increased from 100msec. The patch also mentions that with 1000msec timeout, there are concerns about delaying valid timeout cases. https://patchwork.freedesktop.org/patch/245162/?series=48414&rev=1

Ran into some weirdness when I tried to get the logs and decided to test the patch again. With the patch when the display is plugged in when the shell has started everything works fine, unplugging and re-plugging multiple times all without any issue. dmesg_drm_tip_patch_display_added.log

When booting with the display plugged in the display is flickering slightly but then the shell starts with the right mode and without glitches. Re-plugging the display messes up the mode though. dmesg_drm_tip_patch_display_boot_replugged.log

Will re-test with an even bigger timeout.

Even bigger timeout doesn't help.

When booting and then plugging in the display every now and then even with the patch it stops working.

The patch definitely makes things better but the issue is not gone completely.

Maybe helped by cf37c0979e55 ("drm/i915/display/lspcon: Increase LSPCON mode settle timeout").

With the patch applied:

[   79.369619] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change timed out
[   79.369630] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change failed
[   79.369634] i915 0000:00:02.0: [drm] *ERROR* LSPCON resume failed
[   80.175993] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode hasn't settled
[   81.049591] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode hasn't settled
[   81.331750] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change timed out
[   81.331761] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change failed
[   81.331765] i915 0000:00:02.0: [drm] *ERROR* LSPCON resume failed
[   81.357312] Registered IR keymap rc-cec
[   81.357396] rc rc0: DP-1 as /devices/pci0000:00/0000:00:02.0/rc/rc0
[   81.357491] input: DP-1 as /devices/pci0000:00/0000:00:02.0/rc/rc0/input16
[   83.482572] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode hasn't settled
[   83.764610] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change timed out
[   83.764621] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode change failed
[   83.764625] i915 0000:00:02.0: [drm] *ERROR* LSPCON resume failed
[   84.571508] i915 0000:00:02.0: [drm] *ERROR* LSPCON mode hasn't settled
[   84.577316] i915 0000:00:02.0: [drm] *ERROR* Link Training Unsuccessful

@swick There has been some fixes in the AUX timeout when trying during AUX power down/up cycle. Can you please try on the latest drm-tip and share the logs?

unassigned @raviteja