Re: Recipe for cooking 2.1.72's mm

[Pavel Machek <pavel@Elf.mj.gts.cz>]

Hi!

> > Sorry. There is a problem. It needs to be solved, not worked
> > around. (Notice, that same process does nothing bad to 2.0.28).
> 
> On my system, it just gives one or two out-of-memory kills
> of random processes. I'd really like it if those processes
> would be a little less random... Killing kerneld or crond
> (or X... remember those poor stateless-vga-card users) is
> IMHO worse than killing a program from some USER. Finding
> the most hoggy non-root process group and killing some of
> it's programs shouldn't be too difficult.

Aha. So you were unsuccessfull while reproducing. On my system no
process dies, but whole system is dead.

> btw: I'm using 2.1.66 with my mmap-age patch...
> 
> > And: Work around is bad. Imagine your machine with such behaviour on
> > 100MBit ethernet. Imagine me around (ping -f)ing your machine. That
> > can keep your pages low for as long as I want. You do not your machine
> > to go yo-yo (up and down and up and down ...).
> 
> Ok, so we should limit the amount of memory the kernel can grab
> for internal usage... Sysctl-wise of course, because some people
> have special purpose routing machines.

It might be nice idea. I'm just afraid that for every limit, you find
situation in which limit _must_ be exceeded or action is impossible.

								Pavel

-- 
I'm really pavel@atrey.karlin.mff.cuni.cz. 	   Pavel
Look at http://atrey.karlin.mff.cuni.cz/~pavel/ ;-).