Re: [BUG] set_mempolicy(MPOL_INTERLEAV) cause kernel panic

[KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>]

> On Wed, 2009-07-15 at 18:48 +0900, KOSAKI Motohiro wrote:
> > Hi
> > 
> > On 2.6.31-rc3, following test makes kernel panic immediately.
> > 
> >   numactl --interleave=all echo
> > 
> > Panic message is below. I don't think commit 58568d2a8 is correct patch.
> > 
> > old behavior:
> >   do_set_mempolicy
> >     mpol_new
> >       cpuset_update_task_memory_state
> >         guarantee_online_mems
> >           nodes_and(cs->mems_allowed, node_states[N_HIGH_MEMORY]);
> > 
> > but new code doesn't consider N_HIGH_MEMORY. Then, the userland program
> > passing non-online node bit makes crash, I guess.
> > 
> > Miao, What do you think?
> 
> This looks similar to the problem I tried to fix in:
> 
> 	http://marc.info/?l=linux-mm&m=124140637722309&w=4
> 
> Miao pointed out that the patch needs more work to track hot plug of
> nodes.  I've not had time to get back to this.
> 
> Interestingly, on ia64, the top cpuset mems_allowed gets set to all
> possible nodes, while on x86_64, it gets set to on-line nodes [or nodes
> with memory].  Maybe this is a to support hot-plug?

Maybe.

task->mems_allowed of the init process is initialized by node_possible_map.
if the system doesn't have memory hot-plug capability, node_possible_map
is equal to node_online_map.


-------------------------------------------------
@@ -867,6 +866,11 @@ static noinline int init_post(void)
 static int __init kernel_init(void * unused)
 {
        lock_kernel();
+
+       /*
+        * init can allocate pages on any node
+        */
+       set_mems_allowed(node_possible_map);




--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>