Ticket #11398 (new defect)

Opened 2 years ago

Last modified 2 years ago

On XO-1.75, flash of random data appears on screen at reboot or power-down

Reported by: wad Owned by:
Priority: normal Milestone: 1.75-software
Component: kernel Version: 1.75-B1
Keywords: XO-1.75, reboot, DCON Cc:
Action Needed: reproduce Verified: no
Deployments affected: Blocked By:
Blocking:

Description

On an XO-1.75 B1 or C1 laptop, running recent firmware and OS builds (like Q4C03 and OS883), a flash of random data frequently appears on the screen when rebooting or powering down.

The random data appears as colored static. The timing is just as the UL warning is being put up -- the static usually changes to become the UL warning. On at least one occasion, the static remained displayed until the laptop rebooted.

My hypothesis is that this is a race condition in the code. The GPIO which tells the DCON to load its frame memory from the video stream is asserted, but before it has time to load the memory (it takes at least two frame interrupts after GPIO assertion), the DCON is told via the I2C bus to switch the display to "DCON master" mode.

In some cases, the SoC goes so far as to disable the video output before the DCON has a chance to grab the video, and the display memory remains uninitialized until OFW starts up again.

Change History

Changed 2 years ago by pgf

your description of the symptom ("colored static") is very similar to the symptom we see that was just reported in #11397 (where the DCON fails to load under X). i suspect a dup. DCON freeze is reliably broken in under X, and the DCON is frozen between each of the splash screens one sees during shutdown.

Changed 2 years ago by wad

Possibly, but not necessarily the same piece of code. What I am talking about is a transient phenomenon, where the DCON load actually works, but the switch of video output control happens too soon.

If the code allows a suspend to happen while the DCON is still waiting to load, it would result in a colored static screen during the suspend.

Changed 2 years ago by Quozl

  • milestone changed from Not Triaged to 1.75-software

Changed 2 years ago by Quozl

  • next_action changed from diagnose to reproduce

interesting to know if this still occurs in os12, which fixed #11397.

Note: See TracTickets for help on using tickets.