ubuntu-x-swat team mailing list archive

[Bryce Harrington <bryce@xxxxxxxxxxxxx>]

On Tue, Mar 15, 2011 at 09:28:02AM -0000, Ilan wrote:
> Bryce,
> 
> Are there additional details I should collect in case of future crashes
> that would be helpful to you and upstream?

Specifically what upstream requires for these bugs is a copy of the
following file taken while the system is locked up:

  /sys/kernel/debug/dri/0/i915_error_state

For example, sshing into the system while it's frozen can let you access
it.

The output of 'dmesg' from while the system is frozen can also be of
value.

Both of the above are collected automatically when apport catches the
GPU lockup.  However, such as in this case, *sometimes* it mysteriously
fails to capture anything useful.

The other thing that can be quite useful is if you can find a specific
set of steps that tends to result in the bug more often than not.  When
GPU lockups can be linked to specific sets of steps to reproduce, often
this greatly enables upstream to figure the issue out.

Unfortunately a lot of time with these GPU lockups it seems completely
random from a user-perspective as to when they occur.

Bryce