[PATCH] Make new setting of panic_on_oom

[PATCH] Make new setting of panic_on_oom

[Yasunori Goto]

The current panic_on_oom may not work if there is a process using 
cpusets/mempolicy, because other nodes' memory may still free.
But some people want failover by panic ASAP even if they are used.
This patch makes new setting for its request.

This is not tested yet. But it would work.

Please apply.

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

---
 Documentation/sysctl/vm.txt |   23 +++++++++++++++++------
 mm/oom_kill.c               |    3 +++
 2 files changed, 20 insertions(+), 6 deletions(-)

Index: panic_on_oom2/Documentation/sysctl/vm.txt
===================================================================
--- panic_on_oom2.orig/Documentation/sysctl/vm.txt	2007-04-19 11:19:07.000000000 +0900
+++ panic_on_oom2/Documentation/sysctl/vm.txt	2007-04-19 14:11:24.000000000 +0900
@@ -197,11 +197,22 @@
 
 panic_on_oom
 
-This enables or disables panic on out-of-memory feature.  If this is set to 1,
-the kernel panics when out-of-memory happens.  If this is set to 0, the kernel
-will kill some rogue process, called oom_killer.  Usually, oom_killer can kill
-rogue processes and system will survive.  If you want to panic the system
-rather than killing rogue processes, set this to 1.
+This enables or disables panic on out-of-memory feature.
 
-The default value is 0.
+If this is set to 0, the kernel will kill some rogue process,
+called oom_killer.  Usually, oom_killer can kill rogue processes and
+system will survive.
+
+If this is set to 1, the kernel panics when out-of-memory happens.
+However, if a process limits using nodes by mempolicy/cpusets,
+and those nodes become memory exhaustion status, one process
+may be killed by oom-killer. No panic occurs in this case.
+Because other nodes' memory may be free. This means system total status
+may be not fatal yet.
 
+If this is set to 2, the kernel panics compulsorily even on the
+above-mentioned.
+
+The default value is 0.
+1 and 2 are for failover of clustering. Please select either
+according to your policy of failover.
Index: panic_on_oom2/mm/oom_kill.c
===================================================================
--- panic_on_oom2.orig/mm/oom_kill.c	2007-04-19 11:27:31.000000000 +0900
+++ panic_on_oom2/mm/oom_kill.c	2007-04-19 12:01:44.000000000 +0900
@@ -412,6 +412,9 @@
 	cpuset_lock();
 	read_lock(&tasklist_lock);
 
+	if (sysctl_panic_on_oom == 2)
+		panic("out of memory. Compulsory panic_on_oom is selected.\n");
+
 	/*
 	 * Check if there were limitations on the allocation (only relevant for
 	 * NUMA) that may require different handling.

-- 
Yasunori Goto 

Re: [PATCH] Make new setting of panic_on_oom

[Benjamin LaHaise]

On Fri, Apr 20, 2007 at 08:43:56PM +0900, Yasunori Goto wrote:
> 
> The current panic_on_oom may not work if there is a process using 
> cpusets/mempolicy, because other nodes' memory may still free.
> But some people want failover by panic ASAP even if they are used.
> This patch makes new setting for its request.
> 
> This is not tested yet. But it would work.

>  	read_lock(&tasklist_lock);
>  
> +	if (sysctl_panic_on_oom == 2)
> +		panic("out of memory. Compulsory panic_on_oom is selected.\n");
> +

Wouldn't it be safer to put the panic before the read_lock()?

		-ben
-- 
"Time is of no importance, Mr. President, only life is important."
Don't Email: <zyntrop@kvack.org>.

Re: [PATCH] Make new setting of panic_on_oom

[Christoph Lameter]

On Fri, 20 Apr 2007, Benjamin LaHaise wrote:

> On Fri, Apr 20, 2007 at 08:43:56PM +0900, Yasunori Goto wrote:
> > 
> > The current panic_on_oom may not work if there is a process using 
> > cpusets/mempolicy, because other nodes' memory may still free.
> > But some people want failover by panic ASAP even if they are used.
> > This patch makes new setting for its request.
> > 
> > This is not tested yet. But it would work.
> 
> >  	read_lock(&tasklist_lock);
> >  
> > +	if (sysctl_panic_on_oom == 2)
> > +		panic("out of memory. Compulsory panic_on_oom is selected.\n");
> > +
> 
> Wouldn't it be safer to put the panic before the read_lock()?

I agree. Otherwise the patch seem to be okay.

Re: [PATCH] Make new setting of panic_on_oom

[Yasunori Goto]

> > >  	read_lock(&tasklist_lock);
> > >  
> > > +	if (sysctl_panic_on_oom == 2)
> > > +		panic("out of memory. Compulsory panic_on_oom is selected.\n");
> > > +
> > 
> > Wouldn't it be safer to put the panic before the read_lock()?
> 
> I agree. Otherwise the patch seem to be okay.

Ok. This is take 2.
Thanks for your comment.

---------

The current panic_on_oom may not work if there is a process using 
cpusets/mempolicy, because other nodes' memory may remain.
But some people want failover by panic ASAP even if they are used.
This patch makes new setting for its request.

This is not tested yet. But it would work.

Please apply.

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

---
 Documentation/sysctl/vm.txt |   23 +++++++++++++++++------
 mm/oom_kill.c               |    3 +++
 2 files changed, 20 insertions(+), 6 deletions(-)

Index: panic_on_oom2/Documentation/sysctl/vm.txt
===================================================================
--- panic_on_oom2.orig/Documentation/sysctl/vm.txt	2007-04-21 12:39:09.000000000 +0900
+++ panic_on_oom2/Documentation/sysctl/vm.txt	2007-04-21 12:39:58.000000000 +0900
@@ -197,11 +197,22 @@
 
 panic_on_oom
 
-This enables or disables panic on out-of-memory feature.  If this is set to 1,
-the kernel panics when out-of-memory happens.  If this is set to 0, the kernel
-will kill some rogue process, called oom_killer.  Usually, oom_killer can kill
-rogue processes and system will survive.  If you want to panic the system
-rather than killing rogue processes, set this to 1.
+This enables or disables panic on out-of-memory feature.
 
-The default value is 0.
+If this is set to 0, the kernel will kill some rogue process,
+called oom_killer.  Usually, oom_killer can kill rogue processes and
+system will survive.
+
+If this is set to 1, the kernel panics when out-of-memory happens.
+However, if a process limits using nodes by mempolicy/cpusets,
+and those nodes become memory exhaustion status, one process
+may be killed by oom-killer. No panic occurs in this case.
+Because other nodes' memory may be free. This means system total status
+may be not fatal yet.
 
+If this is set to 2, the kernel panics compulsorily even on the
+above-mentioned.
+
+The default value is 0.
+1 and 2 are for failover of clustering. Please select either
+according to your policy of failover.
Index: panic_on_oom2/mm/oom_kill.c
===================================================================
--- panic_on_oom2.orig/mm/oom_kill.c	2007-04-21 12:39:09.000000000 +0900
+++ panic_on_oom2/mm/oom_kill.c	2007-04-21 12:40:31.000000000 +0900
@@ -409,6 +409,9 @@
 		show_mem();
 	}
 
+	if (sysctl_panic_on_oom == 2)
+		panic("out of memory. Compulsory panic_on_oom is selected.\n");
+
 	cpuset_lock();
 	read_lock(&tasklist_lock);
 


-- 
Yasunori Goto 

Re: [PATCH] Make new setting of panic_on_oom

[Benjamin LaHaise]

On Sat, Apr 21, 2007 at 01:28:16PM +0900, Yasunori Goto wrote:
> Ok. This is take 2.
> Thanks for your comment.
> 
> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>

Looks good.  Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>

		-ben
-- 
"Time is of no importance, Mr. President, only life is important."
Don't Email: <zyntrop@kvack.org>.

Re: [PATCH] Make new setting of panic_on_oom

[Yasunori Goto]

I tested this patch. It worked well.
So, I fixed its description.

Please apply.


----------

The current panic_on_oom may not work if there is a process using 
cpusets/mempolicy, because other nodes' memory may remain.
But some people want failover by panic ASAP even if they are used.
This patch makes new setting for its request.

This is tested on my ia64 box which has 3 nodes.

Please apply.

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>


---
 Documentation/sysctl/vm.txt |   23 +++++++++++++++++------
 mm/oom_kill.c               |    3 +++
 2 files changed, 20 insertions(+), 6 deletions(-)

Index: panic_on_oom2/Documentation/sysctl/vm.txt
===================================================================
--- panic_on_oom2.orig/Documentation/sysctl/vm.txt	2007-04-21 12:39:09.000000000 +0900
+++ panic_on_oom2/Documentation/sysctl/vm.txt	2007-04-21 12:39:58.000000000 +0900
@@ -197,11 +197,22 @@
 
 panic_on_oom
 
-This enables or disables panic on out-of-memory feature.  If this is set to 1,
-the kernel panics when out-of-memory happens.  If this is set to 0, the kernel
-will kill some rogue process, called oom_killer.  Usually, oom_killer can kill
-rogue processes and system will survive.  If you want to panic the system
-rather than killing rogue processes, set this to 1.
+This enables or disables panic on out-of-memory feature.
 
-The default value is 0.
+If this is set to 0, the kernel will kill some rogue process,
+called oom_killer.  Usually, oom_killer can kill rogue processes and
+system will survive.
+
+If this is set to 1, the kernel panics when out-of-memory happens.
+However, if a process limits using nodes by mempolicy/cpusets,
+and those nodes become memory exhaustion status, one process
+may be killed by oom-killer. No panic occurs in this case.
+Because other nodes' memory may be free. This means system total status
+may be not fatal yet.
 
+If this is set to 2, the kernel panics compulsorily even on the
+above-mentioned.
+
+The default value is 0.
+1 and 2 are for failover of clustering. Please select either
+according to your policy of failover.
Index: panic_on_oom2/mm/oom_kill.c
===================================================================
--- panic_on_oom2.orig/mm/oom_kill.c	2007-04-21 12:39:09.000000000 +0900
+++ panic_on_oom2/mm/oom_kill.c	2007-04-21 12:40:31.000000000 +0900
@@ -409,6 +409,9 @@
 		show_mem();
 	}
 
+	if (sysctl_panic_on_oom == 2)
+		panic("out of memory. Compulsory panic_on_oom is selected.\n");
+
 	cpuset_lock();
 	read_lock(&tasklist_lock);
 

-- 
Yasunori Goto