Skip to content

Commit

Permalink
sched/numa: Fix initialization of sched_domain_topology for NUMA
Browse files Browse the repository at this point in the history
Jet Chen has reported a kernel panics when booting qemu-system-x86_64 with
kvm64 cpu. A panic occured while building the sched_domain.

In sched_init_numa, we create a new topology table in which both default
levels and numa levels are copied. The last row of the table must have a null
pointer in the mask field.

The current implementation doesn't add this last row in the computation of the
table size. So we add 1 row in the allocation size that will be used as the
last row of the table. The kzalloc will ensure that the mask field is NULL.

Reported-by: Jet Chen <jet.chen@intel.com>
Tested-by: Jet Chen <jet.chen@intel.com>
Signed-off-by: Vincent Guittot <vincent.guittot@linaro.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: fengguang.wu@intel.com
Link: http://lkml.kernel.org/r/1399972261-25693-1-git-send-email-vincent.guittot@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
  • Loading branch information
Vincent Guittot authored and Ingo Molnar committed May 22, 2014
1 parent 8bf2143 commit c515db8
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion kernel/sched/core.c
Original file line number Diff line number Diff line change
Expand Up @@ -6231,7 +6231,7 @@ static void sched_init_numa(void)
/* Compute default topology size */
for (i = 0; sched_domain_topology[i].mask; i++);

tl = kzalloc((i + level) *
tl = kzalloc((i + level + 1) *
sizeof(struct sched_domain_topology_level), GFP_KERNEL);
if (!tl)
return;
Expand Down

0 comments on commit c515db8

Please sign in to comment.