>>>>> "Paul" == Paul Jakma <paul@clubi.ie> writes:
> size issues aside, proper accounting that allows linux to better
> predict when to say no is surely better than after the fact
> killing. It seems cleaner to have code to predict, than to kill. Then
> we could gracefully say no, rather than ungracefully have to say
> "die" to quite possibly innocent programmes.
The usual problem with "no overcommit" is that it is grossly wasting
swap space. The typical example is system(3) which does a fork+exec(/bin/sh):
if you call it from a 500MB process, you temporarily need another 500MB
of swap even though it will never really be needed (since it's cow-shared).
Of course, swap space is cheap, so it's not necessarily such a big issue,
but on the other hand, the cases where a half-clever OOM-killer doesn't
cut the mustard are rather far and few apart.
Stefan
-
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/
This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:18 EST