Re: Kernel OOPS when reloading i915 after resume from suspend

[Thomas Richter <thor@math.tu-berlin.de>]

Am 02.07.2014 23:22, schrieb Pavel Machek:
> Hi!
>
>> still experimenting with the resume from suspend on the Fujitsu
>> S6010. I can, however, still create a kernel oops. The kernel source
>> comes from alm_fixes5, kernel 3.15.0-rc7+. For that, do the
>> following:
>>
>> 1) Shut down X,
>> 2) Unbind the consoles:
>>
>> echo 0 > /sys/class/vtconsole/vtcon1/bind
>> echo 0 > /sys/class/vtconsole/vtcon0/bind
>>
>> 3) Remove the i915
>>
>> rmmod i915
>>
>> 4) Suspend the system
>>
>> pm-suspend
>
> Does it also break with plain echo mem > /sys/power/state?

No. If the modules stay resident, any user driven shutdown/resume events 
by writing into /sys/power/state work just fine.

> Does it break when you don't suspend at all?

No. Unloading and reloading i915 under normal conditions also works fine.

>> 5) Resume the system by pressing on the power-button.
>> 6) Reload the i915 module with
>>
>> modprobe i915.
>>
>> Result is a kernel-Oops:
>
> Sounds like problem in suspend/resume framework, actually :-(.

Good guess, but wrong. The BIOS leaves the 830GM in a state the i915 
does not seem to like.

Greetings,
	Thomas