Random lockups with i915 driver enabled

Random lockups with i915 driver enabled

[Michael Schanne]

Hi,

I am troubleshooting random lockups using RT patch 6.12.28-rt10.  The
lockups seem to be completely random, sometimes occurring within 10
minutes of bootup, sometimes several hours, up to several days.  When
the issue occurs, the system is completely locked up.  Keyboard and
mouse are unresponsive, serial console unresponsive, network
unresponsive.  There is no clear pattern of what triggers the lockups,
it can happen while the system is sitting idle.  I do not get any
output on the serial port when this happens.  At the moment, I am
suspicious of the i915 driver.  Things I have tried:

Revert RT patch - no lockups over 4 days
Install RT patch but disable DRM_I915 in kernel config - no lockups
over 24 hours

I have tried enabling hardlockup_all_cpu_backtrace=1 /
softlockup_all_cpu_backtrace=1 in the boot parameters, and had several
lockups in this configuration, but I still get no output.

Due to the random nature of this, it is very difficult to know whether
the issue is gone or just not enough time has passed.  Also, I have
built 4x systems in identical configuration, only 2x show lockups and
the other 2x don't.  Software is exactly the same on all 4x machines.

Can anyone suggest any debug techniques to nail down root cause when
the next lockup occurs?  Or are there any potential i915 fixes I could
try out?

Hardware: Intel i5-6500TE quad core CPU (Skylake), 16 GB RAM

Thanks,
Mike

Re: Random lockups with i915 driver enabled

[Sebastian Andrzej Siewior]

On 2025-12-03 16:23:01 [-0500], Michael Schanne wrote:
> Hi,
Hi,

> Can anyone suggest any debug techniques to nail down root cause when
> the next lockup occurs?  Or are there any potential i915 fixes I could
> try out?

An 8250-UART would be probably help.
Does lockdep (CONFIG_PROVE_LOCKING) and/ or CONFIG_DEBUG_ATOMIC_SLEEP
complain?
Can you setup pstore and use it for locking?
Does your BIOS reset your memory on reset? If not, and you can force a
reset in the lockup case, you could try to stuff some information into
the tracing buffer. See 
	https://docs.kernel.org/trace/debugging.html#persistent-buffers-across-boots


Can you try to use the XE driver instead of i915?

> Hardware: Intel i5-6500TE quad core CPU (Skylake), 16 GB RAM
> 
> Thanks,
> Mike

Sebastian