[PATCH v2] cpu: Fix node state for whether it contains CPU

[PATCH v2] cpu: Fix node state for whether it contains CPU

[Tim Chen]

In current kernel code, we only call node_set_state(cpu_to_node(cpu),
N_CPU) when a cpu is hot plugged.  But we do not set the node state for
N_CPU when the cpus are brought online during boot.

So this could lead to failure when we check to see
if a node contains cpu with node_state(node_id, N_CPU).

One use case is in the node_reclaime function:

        /*
         * Only run node reclaim on the local node or on nodes that do
         * not
         * have associated processors. This will favor the local
         * processor
         * over remote processors and spread off node memory allocations
         * as wide as possible.
         */
        if (node_state(pgdat->node_id, N_CPU) && pgdat->node_id !=
                numa_node_id())
                return NODE_RECLAIM_NOSCAN;

I instrumented the kernel to call this function after boot and it
always returns 0 on a x86 desktop machine until I apply
the attached patch.

int num_cpu_node(void)
{
       int i, nr_cpu_nodes = 0;

       for_each_node(i) {
               if (node_state(i, N_CPU))
                       ++ nr_cpu_nodes;
       }

       return nr_cpu_nodes;
}

Fix this by checking each node for online CPU when we initialize
vmstat that's responsible for maintaining node state.

v2:
1. Fix the problem for all architectures in the generic path
inside vmstat, not just for x86.

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
---
 mm/vmstat.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/mm/vmstat.c b/mm/vmstat.c
index 89cec42..d83f953 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1794,6 +1794,16 @@ static void __init start_shepherd_timer(void)
 		round_jiffies_relative(sysctl_stat_interval));
 }
 
+static void __init init_cpu_node_state(void)
+{
+	int cpu;
+
+	get_online_cpus();
+	for_each_online_cpu(cpu)
+		node_set_state(cpu_to_node(cpu), N_CPU);
+	put_online_cpus();
+}
+
 static void vmstat_cpu_dead(int node)
 {
 	int cpu;
@@ -1851,6 +1861,7 @@ static int __init setup_vmstat(void)
 #ifdef CONFIG_SMP
 	cpu_notifier_register_begin();
 	__register_cpu_notifier(&vmstat_notifier);
+	init_cpu_node_state();
 
 	start_shepherd_timer();
 	cpu_notifier_register_done();
-- 
2.5.5

Re: [PATCH v2] cpu: Fix node state for whether it contains CPU

[David Rientjes]

On Mon, 29 Aug 2016, Tim Chen wrote:

> In current kernel code, we only call node_set_state(cpu_to_node(cpu),
> N_CPU) when a cpu is hot plugged.  But we do not set the node state for
> N_CPU when the cpus are brought online during boot.
> 
> So this could lead to failure when we check to see
> if a node contains cpu with node_state(node_id, N_CPU).
> 
> One use case is in the node_reclaime function:
> 
>         /*
>          * Only run node reclaim on the local node or on nodes that do
>          * not
>          * have associated processors. This will favor the local
>          * processor
>          * over remote processors and spread off node memory allocations
>          * as wide as possible.
>          */
>         if (node_state(pgdat->node_id, N_CPU) && pgdat->node_id !=
>                 numa_node_id())
>                 return NODE_RECLAIM_NOSCAN;
> 
> I instrumented the kernel to call this function after boot and it
> always returns 0 on a x86 desktop machine until I apply
> the attached patch.
> 
> int num_cpu_node(void)
> {
>        int i, nr_cpu_nodes = 0;
> 
>        for_each_node(i) {
>                if (node_state(i, N_CPU))
>                        ++ nr_cpu_nodes;
>        }
> 
>        return nr_cpu_nodes;
> }
> 
> Fix this by checking each node for online CPU when we initialize
> vmstat that's responsible for maintaining node state.
> 

That would mean that when node_reclaim_mode is enabled that we weren't 
properly returning NODE_RECLAIM_NOSCAN if a remote node had its own cpus 
and PGDAT_RECLAIM_LOCKED wasn't already set, so this seems like it could 
result in a performance improvement.

> v2:
> 1. Fix the problem for all architectures in the generic path
> inside vmstat, not just for x86.
> 
> Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>

Acked-by: David Rientjes <rientjes@google.com>