Status report for 2.6.12-rc2

[Benjamin Herrenschmidt]

On Sun, 2005-04-17 at 11:18 +0200, Wolfram Quester wrote:
> Hi,
> 
> I use debian on my Powerbook6,2 and recently upgraded from kernel 2.6.10
> with the patch Guido Günther provides at [1] to 2.6.12-rc2 with ben's
> tumbler/snapper patches applied.
> So far I've seen the following issues:
> 1. I tried the new nvidiafb - just because I'm curious and the help says 
>    "This driver supports graphics boards with the nVidia chips, TNT and
>    newer. For very old chipsets, such as the RIVA128, then use the
>    rivafb."
>    The kernel crashed during boot when it tried to switch to nvidiafb. I
>    could see the messages that appeared before this stage until I
>    switched the computer off. Booting with video=ofonly went well and
>    using the traditional rivafb works fine.
>    This situation reminds me to crashes I had sometime ago when I was
>    testing Guido's patches for rivafb when he tried to get proper
>    support for NV30 into rivafb

You should report that to the linux-fbdev mailing list so the maintainer
of this driver can try to figure out what's up.

> 2. Sometimes the machine crashes on suspend to disk. It only happens when I work
>    in X and it seems to freeze in the process of switching to a virtual
>    terminal. The screen shows a distorted image of the previous screen
>    (as if I had 8Bit colour depth with oversized pixels).
>    I never had such freezes with 2.6.10 and this is the issue which
>    annoys me most ATM. I don't have a way to clearly reproduce the
>    freeze. It happend in two out of perhaps 20 suspends. It only seems
>    to happen on the first suspend after boot.
>    The suspend itself is much much faster than with 2.6.10 however.

It dies at suspend time when switching to VC ? Hrm... looks like it
could be X and the framebuffer walking on eachother toes, not sure tho,
2.6.12-rc2 is supposed to have patches fixing such ...

> 3. I configured sound to use snd-powermac. Before I used
>    dmasound-{pmac,core} and it works mostly. After a
>    suspend-resume-cycle I have to kill xfce4-panel, which accesses
>    /dev/mixer0, rmmod snd-powermac, wait a second and reinsert the
>    module. 

Yes, the Alsa code isn't doing very well when sleep is triggered while
an application is playing. I may try to fix it some day, or you can
maybe report that to alsa-devel and have Takashi fix it :)

> 4. Something that is quite annoying but harmless is that after wakeup
>    the machine sometimes suspends again immediately. I had something
>    like this before. On resume the PMU seems to be some kind of
>    irritated and gives wrong information about remaining runtime.
>    If this remaining time is 0 pbbuttonsd suspends the machine again.

Possibly, I haven't really checked what's going on with suspend to disk.
I suspect we just haven't yet started polling again from the PMU and we
expose some crap to userland or something like htat ...

> Thanks,
> 
> Wolfi
> 
>   1. http://honk.physik.uni-konstanz.de/~agx/linux-ppc/kernel/2.6.11.6-agx0.diff
> 
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev at ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-dev
-- 
Benjamin Herrenschmidt <benh at kernel.crashing.org>