[Intel-gfx] resume lagginess and other problems

[Andrew Lutomirski]

On Tue, Jan 26, 2010 at 4:47 PM, Jesse Barnes <jbarnes at virtuousgeek.org> wrote:
> On Tue, 26 Jan 2010 16:37:28 -0500
> Andrew Lutomirski <luto at mit.edu> wrote:
>> > Sounds like you're getting a bunch of hotplug notifications on resume,
>> > maybe due to lid events.  This may cause X to re-probe things, which
>> > can take some time and CPU in the kernel.
>>
>> As far as I can see, all of those "CHANGE" events have no effect (i.e.
>> nothing about the device node actually changed).  Why is the driver
>> generating them?
>
> Probably because ACPI is giving it LID events.
>
>> Is there any way to log ACPI events on resume to see if you're right?
>
> You could use acpi_listen maybe, or add some code to the kernel's ACPI
> button driver to trace it.

I 'cat'ted /proc/acpi/events and I'm pretty sure it wasn't due to a
flood of events.  But I was also running 2.6.33-rc2 not -rc5 (oops,
misconfigured grub) and I haven't seen it yet on -rc5, so I think it's
safe to ignore this bug report.

So far, -rc5 seems more stable than -rc2 as far as graphics go.

--Andy