ubuntu-x-swat team mailing list archive

[Bryce Harrington <bryce@xxxxxxxxxxxxx>]

On Fri, Mar 25, 2011 at 10:53:26AM -0000, Sergey Yanovich wrote:
> > It looks like your system ran out of memory for whatever reason. X
> just happens to be the first really critical thing to notice that
> there's no memory left. It tends to need memory for doing its stuff. ;-)
> The reason you can't restart X is because the GPU got left in an
> inconsistent state (which is probably a bug in itself, but not the
> primary issue here).
> 
> It is certainly not so:
> host2:~$ free
>              total       used       free     shared    buffers     cached
> Mem:       1533292     807152     726140          0      28352     374208
> -/+ buffers/cache:     404592    1128700
> Swap:      1888252          0    1888252

Well, let me rephrase.  The X errors shown in your logs are ones that
typically show up when it is unable to allocate memory for creating
graphics.  Whether that shows up as different numbers in free-memory
listings or not is immaterial.  Something in the system is causing X to
be unable to write to memory.  When I've seen this before it is because
something consumes all the memory; in those cases the memory consumption
was so sudden that we did not see it measurable in /proc/meminfo.

> > But, I can try to help you figure out what actually is causing this bug. (As it is, the info you've collected doesn't really do much to pinpoint what's going on.)
> 
> Thanks. I will collect necessary data.
> 
> > Can you tell me a little more about your system. Are you booted off a
> hard drive or usb key or cd?
> 
> The system boots from compact flash. CF is 100% IDE-compatible. I've no
> problems with this config since 9.04 or so.

Fwiw, when I last dealt with this class of issue it was something that
appeared only on USB flash drives.  In that case the issue was narrowed
down to an error in ubiquity that for whatever reason only showed up
when using it from USB flash (cd installs seemed to be unaffected),
though, thus isn't the same bug as you; however the symptoms are almost
exactly identical.

The issue was that the ubiquity client was requesting infinite screen
redraws for a widget on one of the dialogs; the system would keep going
for some variable period of time after that was displayed but inevitably
ended up locking up the system with X errors almost identical to what
your log shows.

> >  How long have you been running natty?
> 
> I've had natty kernel since 2.6.37-12 or earlier. The CPU on the system
> is jusy 1.1 MHz single core, so hardware acceleration in new kernels
> helps a lot. I dist-upgraded to full natty a week ago.
> 
> > When did you start noticing this bug?
> 
> The problem immediately appeared.
> 
> > Has it happened multiple times?
> 
> Is happens every boot, sooner or later.
> 
> > What is unique or unusual about your system?
> 
> It is configured to run from a flash, so most regular writes goes to
> memory (fstab attached). Other than that, it is a usual Vaio VGN TX-
> 770P.
> 
> > Does it always happen after about the same period of time?
> 
> No the period is different.
> 
> > Does running a certain app or combination of apps make it more likely
> to crash?
> 
> I haven't run natty long enough to gather data. So long my usual session
> was several gnome-terminal tabs, firefox and pidgin. I would say that
> crashes happen when I have pidgin talks, but it is not certain.

Ok, what I think you'll want to do is start noting down the episodes.
Each time it happens, note down approximately how long it had been up
and running and the list of applications you ran.

pidgin sounds like a good client to start examining since it is perhaps
less widely used than gnome-terminal and firefox so could have a bug
that simply hadn't shown up due to insufficient testing for this type of
system configuration.

Bryce