[arch-general] Application stability problems (kernel 3.1, Intel hardware)

[David C. Rankin]

On 11/18/2011 09:41 AM, Ciprian Dorin Craciun wrote:
>      Hello all!
>
>      I'm encountering some very strange problems with my ArchLinux
> lately... And I was wondering if someone has similar problems?
>
>      Context: switched to 3.1.1 kernel, all packages are updated. I
> have a Lenovo X1 Thinkpad -- mainly everything is Intel based.
>
>      Strange behaviour:
>      * Firefox crashes randomly (or takes the CPU to 100%); (no error on stderr;)
>      * All OpenSSL based programs complain from time to time of invalid
> received data;
>      * Flash makes strange noises, strange display artefacts, usually
> just crashes;
>      * mouse jumps in the upper left corner (or sometimes just stays at
> top or left);
>      * nothing strange in dmesg;
>
>      Observations:
>      * it doesn't happen with a custom built kernel 2.6.38.x;
>      * it does happen with either custom 3.1.1 kernel or ArchLinux built one;
>      * it seems that when I'm at the office (with a LAN connection) it
> doesn't happen as often;
>      * it seems that at home (with a Wifi WLAN) it does happen a lot;
>
>      Thanks,
>      Ciprian.
>
>      P.S.: At the moment I don't have time to diagnose better the problem...
>

Ciprian,

   I have just experienced a complete crash of my i686 server running 3.1.1 - 
this server never crashes, and I can't explain what caused it to crash. This 
box is an older Gigabyte GA-7400SL with Athlon XP 3200 (boots to runlevel 3) 
(nothing in the logs points to any specific problem):

Last entry before crash (nothing remarkable earlier):

Nov 17 16:53:00 phoenix dovecot: imap-login: Login: user=<david>, 
method=PLAIN, rip=66.76.64.12, lip=192.168.7.15, mpid=26105, TLS

First entry after restart:

Nov 18 09:57:21 phoenix kernel: [    0.000000] Initializing cgroup subsys cpuset

??

   So something has happened. I've increased the log debug level and will keep 
watching. I am also interested in whether others have experienced this.

   On the plus side, the hwclock/sysclock drift I experienced with 3.0.x 
kernels is gone in 3.1:

[10:22 providence:/etc] # date; hwclock -r
Fri Nov 18 10:22:47 CST 2011
Fri 18 Nov 2011 10:22:47 AM CST  -0.031797 seconds

   If I catch anything useful in the logs, I'll pass it along...

   For your specific issues:

 >      Strange behaviour:
 >      * Firefox crashes randomly (or takes the CPU to 100%); (no error on 
stderr;)
 >      * All OpenSSL based programs complain from time to time of invalid
 > received data;
 >      * Flash makes strange noises, strange display artefacts, usually
 > just crashes;
 >      * mouse jumps in the upper left corner (or sometimes just stays at
 > top or left);
 >      * nothing strange in dmesg;


   This all looks "Desktop" related (except the SSL invalid received data). 
Have you tried another desktop? (i.e. fluxbox, openbox, gnome, kde, etc..) If 
you are using either kde4 or gnome shell, they are both under heavy 
development and I wouldn't put it past being a desktop issue.

-- 
David C. Rankin, J.D.,P.E.