[ltp] ANNOUNCE: FC3 + swsusp kernel

[linux-thinkpad@linux-thinkpad.org]

On Mon, 15 Nov 2004, Alex Deucher wrote:

> On Sun, 14 Nov 2004 18:19:22 -0500, Volker Braun
> <volker.braun@physik.hu-berlin.de> wrote:
>> After a brief struggle to compile a kernel under Fedora Core 3, I got
>> everything working. I wrote an rpm spec file which does everything
>> necessary. You can get it and the finished kernel rpm at
>>
>> http://www.sas.upenn.edu/~vbraun/computing/T41/kernel.html
>>
>> This is a vanilla kernel with the tmpfs xattr patch, and of course the
>> software suspend 2 patch. Also, during installation it fixes the initrd
>> to call suspend2.
>
> Resume may work with the xorg radeon driver w/dri support if you
> switch to another VT before suspending.
> you may find these pages helpful as reference:
> http://visualisation.tudelft.nl/~cpbotha/dri_resume.html
> http://penguinppc.org/~daenzer/DRI/radeon-reinit.diff

Yep.  Resume works ok from X on a T40/ATI Radeon Mobility 9000 under
Fedora Core 3/Xorg 6.8.1 if you enable the following in swsusp2's
/etc/hibernate/hibernate.conf:

LeaveXBeforeSuspend yes
nVidiaHack yes

- nVidiaHack is misleadingly named, and launches a fake Xserver to
reinitialise the card, so works for an ATI too.  If you do this, at
least in the vanilla kernel, you don't need to lose dri.

This is with the vanilla kernel.org 2.6.9 kernel, with the initramfs
fix needed as identified by Volker above.  Neither of these steps
were necessary in FC2 and its Xorg version, but it's worth it as the
Mobility 9000 driver in Xorg 6.8.1 is much improved and I can finally
play Tuxracer without resorting to the appalling ATI drivers which
prevent hibernate completely :)

As for patching the Fedora kernel, which would be preferable,
I suspect you don't need to remove many Fedora patches Volker: a
swsusp2 developer working with me on this has managed to pre-patch the
current FC3 kernel with just two patches (about 100 lines in total)
which allow the current swsusp2 patches to patch cleanly.  There must
be a remaining problem, as the suspend works ok, the resume works ok,
and then the machine promptly reboots!  Hopefully this is a niggle to
be cleared up soon - the goal would be a very small patch set which
can be applied to each new Fedora kernel SRPM to allow the swsusp2
patches to work, so that people could do this independently...
and keeping the 60-odd Fedora kernel patches on board.

Volker - you've obviously done a fair bit on this: you have any
comments on what may be making this reboot happen?

Honey