[Debian-eeepc-devel] Some experiences with kernel 2.6.31

Luca Niccoli lultimouomo at gmail.com
Fri Oct 9 21:00:35 UTC 2009


2009/10/9 Darren Salt <linux at youmustbejoking.demon.co.uk>:

> You should file a bug report against the kernel, then (not against
> linux-source-*, since this will probably be the same in linux-image-*).

I'll take a day to see if I can fix it myself and send a patch, I
doubt anyone will be happy to mess with a staging driver that is
supposed to die soon (The fact that the kernel ships an outdated
version of the driver makes me think so)


> http://bugzilla.kernel.org/show_bug.cgi?id=13390

Seems fixeds now (I used the 2.6.32 patch, it applies without fuzz or errors)

> It should wake up after 30s or so.

I didn't try, I prefer dropping a script in /etc/pm/sleep.d anyway :)


> /var/lib/eeepc-acpi-scripts/cpufv contains the current setting. (It has to be
> there, not in /etc, so that the script can safely modify it without breaking
> Policy.)

Oh, maybe it would be nice to put a comment in /etc/default/eeepc-acpi-scripts?
I'll prepare a patch...


> There is one known cause of this: i2c_i801. That module is blacklisted (if
> you have eeepc-acpi-scripts 1.1.2 or a git snapshot); if it's present in your
> kernel, REMOVE IT and the delay will go away.

It wasn't present, I looked better and found that is probably a BIOS
bug for the 901, loading the module takes ~3s...
http://bugzilla.kernel.org/show_bug.cgi?id=12243
I updated to the latest BIOS but it isn't fixed.
Unfortunately I don't see a way to behave differently according to the
eee model...

> Have you timed it with those built in?

My experience with past kernels has been that building things in the
kernel doesn't speed up the boot process, at least since
module-init-tools 3.5 - the exception being the drivers needed to
mount the root filesystem.
Anyway I'm building a kernel with everything needed built in but
eeepc_laptop, for obvious reasons.

> Nice to know that something other than GNOME benefits from that patch :-)

Now there is just one button on the 901 that doesn't send an XF86
keycode, SOFTBTN2.
SOFTBTN1 sends XF86Screensaver , SOFTBN3/4 send XF86Launch2/3
respectively, and Fn/F6 sends XF86Launch1 (the other Fn keys all send
the natural XF86 keycode)
I don't to what this is due, though.
It's nice to have them reported to X, it allows better integration
with the power/hotkey manager


> Maybe it's worth forcibly setting this and seeing if anybody complains? ;-)

I'm all for it, I always felt the other behaviour was unnatural :)



More information about the Debian-eeepc-devel mailing list