I've had a few instances in which a machine ran out of virtual memory, and
became dependent on the kernel killing processes. I'm used to
the kernel having some form of methodology to select an application based
on memory growth and possibly other factors before killing it.
A specific instance that stands out very clearly as a problem is as such:
My machine had roughly 128MB of memory allocated, leaving 128M physical ram
and 128M swap free... I started up ggv 0.61 and started reading through a
pdf file. At some point I must have run across a bug, because ggv began
consuming all the memory it could very rapidly... Behaving as if it
was executing "while (1) { foo = malloc (size); }"...
With ggv allocating roughly a 1/4 gig of memory in 2 minutes or less, one
would assume that it would be a prime candidate for killing. Instead
I found that the kernel (or whomever responsible) had killed sendmail, named,
cron, and one of a few processes that could log me out (X/gdm/...).
There have been a few cases of this occurring to me, but I've always dismissed
them because I've never had a particularly clear case.
Obviously a work around is to setup process memory limits to prevent a single
runaway program from causes semi-important processes from being killed.
I am, however, concerned about what I've noticed. Has anyone else seen
this behavior, any thoughts?
Kernel version is 2.2.12+memleak fix.
Hardware is:
Asus p2b-ls (rev 1.3)
256Meg PC100 sdram (2x 128meg dimms)
Celeron 400
Spattering of disks and other things that don't seem to relevant with memory
I have seen this occur on two different machines, but the particularly clear
case I described is on the machine described above.
I am on the l-k mailing list, but am really behind on reading it. Please
cc: responses to me, or I fear I may lose them among the 5000+ messages in
my l-k box. :)
Thanks
--- Joe Diehl <joed@ptn.net> Information and Communication Systems Paragon Technology, Inc.- 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.tux.org/lkml/