On Tue, 2 Mar 2010 16:53:00 -0800 (PST)
David Rientjes <rientjes@xxxxxxxxxx> wrote:
> On Wed, 3 Mar 2010, KAMEZAWA Hiroyuki wrote:
>
> > memory_cgroup_out_of_memory() kills a task. and return VM_FAULT_OOM then,
> > page_fault_out_of_memory() kills another task.
> > and cause panic if panic_on_oom=1.
> >
>
> If mem_cgroup_out_of_memory() has returned, then it has already killed a
> task that will have TIF_MEMDIE set and therefore make the VM_FAULT_OOM oom
> a no-op. If the oom killed task subsequently returns VM_FAULT_OOM, we
> better panic because we've fully depleted memory reserves and no future
> memory freeing is guaranteed.
>
In patch 01-03, you don't modified panic_on_oom implementation.
And this patch, you don't modified the return code of memcg's charge code.
It still returns -ENOMEM.
Then, VM_FAULT_OOM is returned and page_fault_out_of_memory() calles this
and hit this.
case CONSTRAINT_NONE:
if (sysctl_panic_on_oom) {
dump_header(NULL, gfp_mask, order, NULL);
panic("out of memory. panic_on_oom is selected\n");
}
The system will panic. A hook, mem_cgroup_oom_called() is for avoiding this.
memcg's oom doesn't mean memory shortage, just means it his limit.
Thanks,
-Kame
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>