[SLAB] Shutdown cache_reaper when cpu goes down

[SLAB] Shutdown cache_reaper when cpu goes down

[Christoph Lameter]

Shutdown the cache_reaper if the cpu is brought down and set the 
cache_reap.func to NULL. Otherwise hotplug shuts down the reaper for good.

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.20-rc6-mm2/mm/slab.c
===================================================================
--- linux-2.6.20-rc6-mm2.orig/mm/slab.c	2007-01-29 14:27:34.199229828 -0600
+++ linux-2.6.20-rc6-mm2/mm/slab.c	2007-01-30 13:55:57.936838878 -0600
@@ -1271,6 +1271,18 @@ static int __cpuinit cpuup_callback(stru
 		start_cpu_timer(cpu);
 		break;
 #ifdef CONFIG_HOTPLUG_CPU
+  	case CPU_DOWN_PREPARE:
+		/*
+		 * Shutdown cache reaper. Note that the cache_chain_mutex is
+		 * held so that cache_reap() cannot modify reap_work
+		 * concurrently.
+		*/
+		cancel_rearming_delayed_work(&per_cpu(reap_work, cpu));
+		per_cpu(reap_work, cpu).work.func = NULL;
+  		break;
+  	case CPU_DOWN_FAILED:
+		start_cpu_timer(cpu);
+  		break;
 	case CPU_DEAD:
 		/*
 		 * Even if all the cpus of a node are down, we don't free the

Re: [SLAB] Shutdown cache_reaper when cpu goes down

[Christoph Lameter]

The comments do not explain correctly what is going on. Sorry Oleg but it 
seems that the protection of the assignment to reap_work is different that 
what we initially thought.

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: current/mm/slab.c
===================================================================
--- current.orig/mm/slab.c	2007-02-01 15:07:09.000000000 -0800
+++ current/mm/slab.c	2007-02-01 15:09:21.000000000 -0800
@@ -1274,10 +1274,12 @@ static int __cpuinit cpuup_callback(stru
   	case CPU_DOWN_PREPARE:
 		/*
 		 * Shutdown cache reaper. Note that the cache_chain_mutex is
-		 * held so that cache_reap() cannot modify reap_work
-		 * concurrently.
+		 * held so that if cache_reap() is invoked it cannot do
+		 * anything expensive but will only modify reap_work
+		 * and reschedule the timer.
 		*/
 		cancel_rearming_delayed_work(&per_cpu(reap_work, cpu));
+		/* Now the cache_reaper is guaranteed to be not running. */
 		per_cpu(reap_work, cpu).work.func = NULL;
   		break;
   	case CPU_DOWN_FAILED:

Re: [SLAB] Shutdown cache_reaper when cpu goes down

[Andrew Morton]

On Thu, 1 Feb 2007 11:13:29 -0800 (PST)
Christoph Lameter <clameter@sgi.com> wrote:

> Shutdown the cache_reaper if the cpu is brought down and set the 
> cache_reap.func to NULL. Otherwise hotplug shuts down the reaper for good.
> 
> Signed-off-by: Christoph Lameter <clameter@sgi.com>
> 
> Index: linux-2.6.20-rc6-mm2/mm/slab.c
> ===================================================================
> --- linux-2.6.20-rc6-mm2.orig/mm/slab.c	2007-01-29 14:27:34.199229828 -0600
> +++ linux-2.6.20-rc6-mm2/mm/slab.c	2007-01-30 13:55:57.936838878 -0600
> @@ -1271,6 +1271,18 @@ static int __cpuinit cpuup_callback(stru
>  		start_cpu_timer(cpu);
>  		break;
>  #ifdef CONFIG_HOTPLUG_CPU
> +  	case CPU_DOWN_PREPARE:
> +		/*
> +		 * Shutdown cache reaper. Note that the cache_chain_mutex is
> +		 * held so that cache_reap() cannot modify reap_work
> +		 * concurrently.
> +		*/
> +		cancel_rearming_delayed_work(&per_cpu(reap_work, cpu));
> +		per_cpu(reap_work, cpu).work.func = NULL;
> +  		break;
> +  	case CPU_DOWN_FAILED:
> +		start_cpu_timer(cpu);
> +  		break;
>  	case CPU_DEAD:
>  		/*
>  		 * Even if all the cpus of a node are down, we don't free the

Is this bug unique to the changes in -mm, or is it in mainline?

Re: [SLAB] Shutdown cache_reaper when cpu goes down

[Christoph Lameter]

On Thu, 1 Feb 2007, Andrew Morton wrote:

> Is this bug unique to the changes in -mm, or is it in mainline?

Mainline is also affected. However, mainline may need a different fix 
since the hotplug notifier scheme was changed.