From: David Daney <[email protected]>
When "numa=off" is passed on the command line, of_node_to_nid() still
returns the node number (which can be greater than zero). However, in
this case all the memory is associated with the dummy node zero. This
causes OOPS in kernel/irq/irqdomain.c:
domain = kzalloc_node(sizeof(*domain) + (sizeof(unsigned int) * size),
GFP_KERNEL, of_node_to_nid(of_node));
...
which in my case then caused the kernel to OOPS for the IRQ controller
on node 1:
[ 0.000000] [<fffffc00081bba84>] __alloc_pages_nodemask+0xa4/0xe68
[ 0.000000] [<fffffc00082163a8>] new_slab+0xd0/0x57c
[ 0.000000] [<fffffc000821879c>] ___slab_alloc+0x2e4/0x514
[ 0.000000] [<fffffc000823882c>] __slab_alloc+0x48/0x58
[ 0.000000] [<fffffc00082195a0>] __kmalloc_node+0xd0/0x2e0
[ 0.000000] [<fffffc00081119b8>] __irq_domain_add+0x7c/0x164
[ 0.000000] [<fffffc0008b75d30>] its_probe+0x784/0x81c
[ 0.000000] [<fffffc0008b75e10>] its_init+0x48/0x1b0
Fix by forcing of_node_to_nid() to return NUMA_NO_NODE when numa=off.
The kmalloc_node() family is perfectly happy when the node is
specified as NUMA_NO_NODE.
Reported-by: Gilbert Netzer <[email protected]>
Signed-off-by: David Daney <[email protected]>
---
arch/arm64/mm/numa.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/arch/arm64/mm/numa.c b/arch/arm64/mm/numa.c
index 778a985..6d34ebb 100644
--- a/arch/arm64/mm/numa.c
+++ b/arch/arm64/mm/numa.c
@@ -41,8 +41,10 @@ static __init int numa_parse_early_param(char *opt)
{
if (!opt)
return -EINVAL;
- if (!strncmp(opt, "off", 3))
+ if (!strncmp(opt, "off", 3)) {
+ __of_force_no_numa();
numa_off = true;
+ }
return 0;
}
@@ -432,6 +434,7 @@ static int __init dummy_numa_init(void)
return ret;
}
+ __of_force_no_numa();
numa_off = true;
return 0;
}
--
1.8.3.1
