Re: unkillable dpkg-query processes

2007-10-28 Thread Sébastien Bernard

Bernd Zeimetz a écrit :

Hi,

please note that the futex bug also happens on US II machines,
it is jsut almost impossible to reproduce it - it'll just hang
after random days of building.


Everyone who sees these UltraSPARC-III problems please send me PRECISE
and FULL description of how to install from scratch a machine and run
something that will trigger these errors.


Can you please check if the Kernel config I've attached to one of my
last mails is fine for you? The normal Debian installer doesn't
boot on the US III machines which use two CPUs in one board as the
installer's Kernel is a non-SMP Kernel, and the result is that the
machine throws a CPU exception and needs to be power-cycled

I've started to investigate there with the help of a contact from
Sun, but we both didn't have the time to finish this.
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=440720 if you want
to have a look, please ignore those troll postings from chealer
in between...

So to give you a recipe to install Debian on such a box, I need to
build an installer with a SMP Kernel for you. If the config is fine
for your needs, I could just use use it.


The other option is to use debootstrap, if you have some system
on the machine already - so if you want to use that instead of
messing with a network installer, please let me know.
Debootstrap should run on most systems, as long as they have
ar/tar/gunzip and a bash (probably sh is enough...).
Would be faster to use that, and faster to write a recipe for
that.

I'll mark all Qlogic firmware related points, so the recipe should
work on machines with (v440, v880, probably the Enterprise models,
too) and without FC (I guess the Blade 1000 and 2000).


If you don't have access to an US-III machine, I can find a way
to give you access to the RSC and serial console of our machine.


Cheers,

Bernd


Well, I got bitten twice with this bug.

First is on U60, unstable debian.
Since mono team decided that the mono is broken on Sparc (and despite
the fix provided by David Miller), I had to rebuild after enabling the sparc
arch in the source.

The hangs happens always at the end of the buid when invoking dh_shgenlibs in 
the build.

This is not 100% reproducable even in my env.

Second was sun blade 2000 SMP with Ubuntu gutsy, I wasn't able to update the 
xemacs21 package.
The machine hanged with invoking the post installation script.

This is not really reproducable now that I upgraded the packages.

The mono build is, in my humble opinion, the most interesting track to catch 
the bug.

Seb
-
To unsubscribe from this list: send the line unsubscribe sparclinux in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html


Stuck process on debian unstable bring the machine down

2007-05-30 Thread Sébastien Bernard

Hi, I'm running a smp kernel on an U60.
I was trying to compile the source package of mono 1.2.4-3 on my machine 
when I got something really strange.


The compilation process got stuck in the dh_shlibdeps invokation.

When searching for the shlib, I got a process dpkg query which suddendly 
went to 100% CPU usage.
After further examination, the top indicates that the proces uses 1700m 
of virtual mem, with 39m of rss.
Strangely, the process eats all cpu cycles of the machine bringing the 
load to a steady 70.
The process is unkillable and an strace on it shows absolutely nothing 
which is plain weird.


This problem shows up either with 2.6.21.3-smp kernel and 2.6.20.12-smp.
Those kernels are home compiled.

	Can someone help me explain this oddity. I'm rebooting the machine for 
the 3rd time in two whereas it was working flawlessly for days.


Seb
-
To unsubscribe from this list: send the line unsubscribe sparclinux in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html