--- Begin Message ---
On Tue, 27 Sep 2011 00:45:01 -0500
Jonathan Nieder <jrnieder@gmail.com> wrote:
> Hi John,
>
> John O'Hagan wrote:
>
> > I run daily-updated testing on an HP Mini 5102 with N10 graphics
> > card (uses i915) running XFCE4.8. When 3.0.0-1 kernel came in I
> > installed both RT and non- RT versions. On non-RT version, both
> > suspend-to-ram and resume work as expected. On RT kernel they work
> > with only XFCE desktop running, however if other programs are
> > running as well (I've tried Chromium and Sylpheed), when suspend is
> > attempted, the screen goes black, fan runs high, and keyboard
> > becomes unresponsive, requiring a hard shutdown every time. I've
> > reproduced this about a dozen times.
> >
> > Nothing unusual in pm-suspend log that I could see.
>
> Excellent detective work; thank you. It would be nice to pin this on
> a particular driver, if possible. So:
>
> 1. Can you reproduce it with suspend-to-disk, too? (That means
> "echo disk >/sys/power/state".)
Yes, behaviour is the same.
> 2. Can you reproduce it without the graphics driver loaded (e.g., if
> you can reproduce it by inducing some network activity with programs
> like "w3m" from a terminal in X, can you do the same by booting with
> "single" on the kernel command line and trying the same in the
> console)? I realize this might be hard to check.
This is confusing: I can reproduce it in single-user mode with nothing additional running with echo mem > /sys/power/state. Not every time but most times.
> 3. Since hey, one can be lucky sometimes: is it possible to catch the
> failure as it happens, for example by not suspending the console
> and suspending everything else? See
>
> https://raw.github.com/torvalds/linux/master/Documentation/power/basic-pm-debugging.txt
> https://raw.github.com/torvalds/linux/master/Documentation/networking/netconsole.txt
>
> (or the analagously named files in the linux-doc-3.0.0 package)
> for hints in that direction.
This is a little out of my comfort zone, but I did follow the procedure in the first link, running the sequential tests with /sys/power/pm_test. "freezer", "devices" and "platform" tests all work, the failure occurs with "processors", both for STR and STD. IFACT from the link, this means the problem is not with a driver but with processor states? But my laptop only has a single processor (even though in /sys/devices/system/cpu/ there is cpu0 and cpu1, only the latter has a file "online" mentioned in the link, is this relevant?).
> 4. Does Alt+Sysrq work in the broken state? If so, the following
> could be useful.
No, totally frozen.
[...]
>
> > ** Tainted: C (1024)
> > * Module from drivers/staging has been loaded.
> [...]
> > [ 8.960305] brcmutil: module is from the staging directory, the
> > quality is unknown, you have been warned. [ 9.019973] brcmsmac:
> > module is from the staging directory, the quality is unknown, you
> > have been warned.
>
> Is it possible to reproduce this without brcmsmac? E.g., does the b43
> driver support your card (I haven't checked)?
With this kernel AFAIK I can only use brcmsmac, but I can unload it and still reproduce the problem.
--- End Message ---