Re: [PATCH 3/3] x86: fix node_possible_map logic -v2

[Yinghai Lu <yhlu.kernel@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 441 bytes --]

On Tue, May 12, 2009 at 9:16 AM, Jack Steiner <steiner@sgi.com> wrote:
> On Tue, May 12, 2009 at 08:10:18AM -0700, Yinghai Lu wrote:
>> jack,
>>
>> can you confirm:
>> 1. without this patch, your system still have problem?
>> 2. does current tip/master work for you?
>
> The current tip/master fails with or without the patch. AFAICT, the
> failure is identical in both cases.

ok, can you try attach patch to check cpu_to_node mapping?

YH

[-- Attachment #2: debug_extra_numa_init_node.patch --]
[-- Type: text/x-patch, Size: 836 bytes --]

Subject: [PATCH] x86: extra debug for cpu_to_node mapping

print more info about cpu node mapping finally ?

Signed-off-by: Yinghai Lu <yinghai@kernel.org>

---
 arch/x86/mm/numa_64.c |   11 +++++++++++
 1 file changed, 11 insertions(+)

Index: linux-2.6/arch/x86/mm/numa_64.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/numa_64.c
+++ linux-2.6/arch/x86/mm/numa_64.c
@@ -633,6 +633,17 @@ void __init init_cpu_to_node(void)
 			continue;
 		numa_set_node(cpu, node);
 	}
+	printk(KERN_INFO "init_cpu_to_node:\n");
+	for_each_possible_cpu(cpu) {
+		int node;
+		u16 apicid;
+		node = early_cpu_to_node(cpu);
+		if (node == NUMA_NO_NODE)
+			continue;
+		apicid = cpu_to_apicid[cpu];
+		printk(KERN_INFO "cpu %d -> apicid %#x -> node %d\n",
+				 cpu, apicid, node);
+	}
 }
 #endif