[clug] nice.... but for memory?

Peter Barker pbarker at barker.dropbear.id.au
Wed Jan 21 07:16:59 GMT 2004


On Wed, 21 Jan 2004, Martin Pool wrote:

> Oh, possibly.  That does mean that the process can use up arbitrarily
> large amounts of memory just by forking, which kind of limits the
> usefulness.

If you're trying to stop malicious people, you've got a /lot/ of resource
starvation problems apart from memory. Now stopping an apache daemon
that's in a memory allocation death spiral screwing over your servers
(when you don't have time to debug other people's code) - that's where
ulimit is handy.

> I don't think you will necessarily be toast; when it reaches the limit
> things like brk() will just start to return ENOMEM, which libc should
> handle.  Of course most programs have probably not been tested in this
> domain and so will die.

Heh. And the /usual/ implementation of xmalloc is? :)

Now. Problem at hand... the only other thing I can think of which /may/
help is bsd process accounting, which I've never used. A short google
/does/ mention memory as something it will track. I've always thought of
that stuff as "run-it-as-a-batch-and-send-someone-a-bill" sort of thing,
though.

> Martin

Yours,
-- 
Peter Barker                          |   N    _--_|\ /---- Barham, Vic
Programmer,Sysadmin,Geek              | W + E /     /\
pbarker at barker.dropbear.id.au         |   S   \_,--?_*<-- Canberra
You need a bigger hammer.             |             v    [35S, 149E]
"They'll need a whole new Orwellian pseudo-crime-name for that... I
 suggest "digital molestation of kittens". -  Jeremi (14640) from Slashdot



More information about the linux mailing list