[rfc][patch 0/5] 2.6.19-rc1: oom killer fixes

[Nick Piggin]

I've been prompted to take another look through the OOM killer because it
turns out it is killing tasks that have had their oom_adj set to -17 (which
is supposed to make them unkillable).

So there are a number of problems, firstly, the child and sibling thread
killing routines do not account for -17 children/siblings.

Secondly, most architecture specific pagefault handlers do a direct kill
of the current process if it takes a VM_FAULT_OOM. This is a pretty rare
thing to happen, because there isn't a lot of higher order allocations
happening, but it is not impossible. I think we can just call into the
OOM killer here, and return to userspace... but I'd like comments about
this.

Thanks,
Nick
--
SuSE Labs
 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/