Re: Trouble with 116 (was Re: Notebooks)

Linus Torvalds (torvalds@transmeta.com)
Wed, 19 Aug 1998 10:03:57 -0700 (PDT)

On Wed, 19 Aug 1998, Craig Schlenter wrote:
>
> I had my box 'lock-up' on me with a slightly modified 2.1.116. I'm running
> with the above line fixed and with the page ageing #ifdef set to 1 instead
> of 0. I have 32M RAM, 128M swap and had KDE, netscape, pine, winamp, kpm
> and a couple of terminals running. I tried to switch desktops and it froze
> solid from an X perspective. I managed to ssh into it though from another
> machine to shut it down. dmesg said something like 'out of memory for X',
> 'out of memory for klogd'. The X process (S3 video) didn't seem to be
> around anymore and the console was left displaying the X screen and was
> essentially useless.

Ok. This is a "good" lockup as far as the kernel is concerned.

It probably means that the kernel is a bit too eager to return memory
allocation failures, which is something I didn't expect from the changes,
but should be a matter of turning a knob or two.

So what happened was that the kernel decided (much too early, for some as
of yet specified reason) that you didn't have enough memory free at some
allocation, and as a result it decided to refuse the allocation. It so
happened that that refusal led to the killing of the process that was
refused, namely X in this case.

I suspect that the changes in 2.1.116 aren't per se what showed it, but
that the different scheduling patterns of the low memory case is showing
cases that could have happened before but were harder to trigger. It
shouldn't be too hard to find.

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html