CONFIG_SOFTWARE_SUSPEND: on resume machine reboots at "restarting tasks"

[Aleksey Nogin]

I've been using the Raw Hide kernels recompiled with 
CONFIG_SOFTWARE_SUSPEND enabled (and CONFIG_APM disabled) and using ACPI 
software suspend. For a while it worked flawlessly, but then it broke - 
now at resume time, it starts resuming OK, but shortly after it says 
"restarting tasks", machine suddenly reboots.

Worked OK:
2.6.3-1.116 2.6.3-1.118

No longer OK:
2.6.3-2.1.253 2.6.3-2.1.253.2.1 2.6.4-1.298

Any ideas what might be going on? Thanks a lot!

P.S. Somewhere around the time it stopped working, I started 
experimenting with enabling and using SELinux, so it might be 
SELinux-related.

-- 
Aleksey Nogin

Home Page: http://nogin.org/
E-Mail: nogin(a)cs.caltech.edu (office), aleksey(a)nogin.org (personal)
Office: Jorgensen 70, tel: [phone]

[Jeremy Katz]

On Fri, 2004-04-02 at 14:02 -0800, Aleksey Nogin wrote:
>  I've been using the Raw Hide kernels recompiled with 
>  CONFIG_SOFTWARE_SUSPEND enabled (and CONFIG_APM disabled) and using ACPI 
>  software suspend. For a while it worked flawlessly, but then it broke - 
>  now at resume time, it starts resuming OK, but shortly after it says 
>  "restarting tasks", machine suddenly reboots. 
[snip]
>  Any ideas what might be going on? Thanks a lot! 
The newer kernels have the 4G/4G VM split enabled.  When you resume with
this patch applied, the machine triple faults and then reboots.  It's
being looked into...

Jeremy

Impact of 4G/4G split [Re: CONFIG_SOFTWARE_SUSPEND...]

[Per Bjornsson]

On Fri, 2004-04-02 at 14:45, Jeremy Katz wrote:

>  The newer kernels have the 4G/4G VM split enabled.  When you resume
> with
>  this patch applied, the machine triple faults and then reboots.  It's
>  being looked into... 
What is the performance impact of the 4G/4G split? I thought that it was
actually somewhat severe (haven't measured it myself though), so that
enabling this would only be useful when you're actually using a lot of
memory. Is this the case, and if it is, would it be a good idea to have
separate kernels for desktop/"low-memory" configurations without the
4G/4G split enabled? (I seem to recall that there might have been
special "hugemem" kernels with compile options chosen to optimize for
handling loads of memory as opposed to the normal kernels.)

Best regards,
Per Bjornsson

Impact of 4G/4G split [Re: CONFIG_SOFTWARE_SUSPEND...]

[Alan Cox]

On Fri, Apr 02, 2004 at 03:11:28PM -0800, Per Bjornsson wrote:
>  What is the performance impact of the 4G/4G split? I thought that it
> was
>  actually somewhat severe (haven't measured it myself though), so that 
I did some measurements on my laptop and its about 10%. Some things like
high speed networking seem to hurt a lot more and a few fairly bogus
tests like 1000 thread cycling are way worse. 

Alan

[Aleksey Nogin]

On 02.04.2004 14:45, Jeremy Katz wrote:

>  The newer kernels have the 4G/4G VM split enabled.  When you resume
> with
>  this patch applied, the machine triple faults and then reboots.  It's
>  being looked into... 
So if I do not care for 4G, but do care about software suspend, are 
there some config options I can toglle to make it work for me? Thanks!

-- 
Aleksey Nogin

Home Page: http://nogin.org/
E-Mail: nogin(a)cs.caltech.edu (office), aleksey(a)nogin.org (personal)
Office: Jorgensen 70, tel: [phone]

[Alan Cox]

On Fri, Apr 02, 2004 at 05:45:06PM -0500, Jeremy Katz wrote:
>  The newer kernels have the 4G/4G VM split enabled.  When you resume
> with
>  this patch applied, the machine triple faults and then reboots.  It's
>  being looked into... 
Interestingly this affects both APM and ACPI btw

[Warren Togami]

Alan Cox wrote:
>  On Fri, Apr 02, 2004 at 05:45:06PM -0500, Jeremy Katz wrote:
>  
> >The newer kernels have the 4G/4G VM split enabled.  When you resume with
> >this patch applied, the machine triple faults and then reboots.  It's
> >being looked into...
>  
>  
>  Interestingly this affects both APM and ACPI btw 
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=117032
4G/4G and broken S3 sleep

You can workaround this problem by recompiling the kernel with 4G/4G 
disabled or use the i586 kernel.  If you use the later workaround you 
will run into NPTL problems like this:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=91933

Warren

[Alan Cox]

On Fri, Apr 02, 2004 at 05:13:39PM -1000, Warren Togami wrote:
>  https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=117032
>  4G/4G and broken S3 sleep
>  
>  You can workaround this problem by recompiling the kernel with 4G/4G 
>  disabled or use the i586 kernel.  If you use the later workaround you 
>  will run into NPTL problems like this:
>  
>  https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=91933 
If you rebuild the kernel and db4 it seems ok. db* is one of the things
I always rebuild anyway because the RH one doesnt work with some packages
on non i686 boxes (eg phpwiki)

Alan

[Russell Coker]

On Sat, 3 Apr 2004 08:02, Aleksey Nogin <aleksey(a)nogin.org&gt; wrote:
>  now at resume time, it starts resuming OK, but shortly after it says
>  "restarting tasks", machine suddenly reboots. 
[...]
>  P.S. Somewhere around the time it stopped working, I started
>  experimenting with enabling and using SELinux, so it might be
>  SELinux-related. 
If you suspect SE Linux of causing such problems then the thing to do is to 
enable a serial console and capture the kernel messages to another machine.  
I have never seen SE Linux cause a kernel problem without logging it to the 
console (including serial console if one is enabled).  I have on several 
occasions seen SE Linux trigger a sudden reboot (by preventing the kernel 
from doing what it wants) in a way that only a serial console could capture.

-- 
http://www.coker.com.au/selinux/   My NSA Security Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP benchmark
http://www.coker.com.au/~russell/  My home page