Re: [PATCH] x86/topology: Fallback to SMT level only once

[Jirka Hladky]

Hi Peter,

yes, initially I have reported the issue to occur on Intel E5v3 CPU
(that CPU does not have CoD) but it has turned to be a fluctuation of
results. After repeating the test 10 times it has turned out that
Intel E5v3 CPU is not affected. I'm sorry for that.

I have then rerun the test on Opteron 6272 (same CPU as used by
authors of the paper) and there was performance degradation by factor
9x. Jirka has then provided the patch.

Thanks a lot
Jirka

On Mon, Aug 29, 2016 at 12:40 PM, Peter Zijlstra  wrote:
> On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
>> Jirka, Peter and Jean-Pierre reported performance drop on
>> some cpus after making cpu offline and online again.
>>
>> The reason is the kernel logic that falls back to SMT
>> level topology if more than one node is detected within
>> CPU package. During the system boot this logic cuts out
>> the DIE topology level and numa code adds NUMA level
>> on top of this.
>
> Its not SMT topology, and back when I asked if he had CoD enabled or
> such he said not.
>
>
> See also:
>
> http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com
>
> Arguably, that should have been split in two patches, but alas..

Re: [PATCH] x86/topology: Fallback to SMT level only once

[Jirka Hladky]

Hi Peter,

yes, initially I have reported the issue to occur on Intel E5v3 CPU
(that CPU does not have CoD) but it has turned to be a fluctuation of
results. After repeating the test 10 times it has turned out that
Intel E5v3 CPU is not affected. I'm sorry for that.

I have then rerun the test on Opteron 6272 (same CPU as used by
authors of the paper) and there was performance degradation by factor
9x. Jirka has then provided the patch.

Thanks a lot
Jirka

On Mon, Aug 29, 2016 at 12:40 PM, Peter Zijlstra  wrote:
> On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
>> Jirka, Peter and Jean-Pierre reported performance drop on
>> some cpus after making cpu offline and online again.
>>
>> The reason is the kernel logic that falls back to SMT
>> level topology if more than one node is detected within
>> CPU package. During the system boot this logic cuts out
>> the DIE topology level and numa code adds NUMA level
>> on top of this.
>
> Its not SMT topology, and back when I asked if he had CoD enabled or
> such he said not.
>
>
> See also:
>
> http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com
>
> Arguably, that should have been split in two patches, but alas..

Re: [PATCH] x86/topology: Fallback to SMT level only once

[Jiri Olsa]

On Mon, Aug 29, 2016 at 12:40:47PM +0200, Peter Zijlstra wrote:
> On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
> > Jirka, Peter and Jean-Pierre reported performance drop on
> > some cpus after making cpu offline and online again.
> > 
> > The reason is the kernel logic that falls back to SMT
> > level topology if more than one node is detected within
> > CPU package. During the system boot this logic cuts out
> > the DIE topology level and numa code adds NUMA level
> > on top of this.
> 
> Its not SMT topology, and back when I asked if he had CoD enabled or
> such he said not.
> 
> 
> See also:
> 
> http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com
> 
> Arguably, that should have been split in two patches, but alas..

ok, that change will work for us as well

are you going to pull it?

thanks,
jirka

Re: [PATCH] x86/topology: Fallback to SMT level only once

[Jiri Olsa]

On Mon, Aug 29, 2016 at 12:40:47PM +0200, Peter Zijlstra wrote:
> On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
> > Jirka, Peter and Jean-Pierre reported performance drop on
> > some cpus after making cpu offline and online again.
> > 
> > The reason is the kernel logic that falls back to SMT
> > level topology if more than one node is detected within
> > CPU package. During the system boot this logic cuts out
> > the DIE topology level and numa code adds NUMA level
> > on top of this.
> 
> Its not SMT topology, and back when I asked if he had CoD enabled or
> such he said not.
> 
> 
> See also:
> 
> http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com
> 
> Arguably, that should have been split in two patches, but alas..

ok, that change will work for us as well

are you going to pull it?

thanks,
jirka

Re: [PATCH] x86/topology: Fallback to SMT level only once

[Peter Zijlstra]

On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
> Jirka, Peter and Jean-Pierre reported performance drop on
> some cpus after making cpu offline and online again.
> 
> The reason is the kernel logic that falls back to SMT
> level topology if more than one node is detected within
> CPU package. During the system boot this logic cuts out
> the DIE topology level and numa code adds NUMA level
> on top of this.

Its not SMT topology, and back when I asked if he had CoD enabled or
such he said not.


See also:

http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com

Arguably, that should have been split in two patches, but alas..

Re: [PATCH] x86/topology: Fallback to SMT level only once

[Peter Zijlstra]

On Sun, Aug 28, 2016 at 08:19:46PM +0200, Jiri Olsa wrote:
> Jirka, Peter and Jean-Pierre reported performance drop on
> some cpus after making cpu offline and online again.
> 
> The reason is the kernel logic that falls back to SMT
> level topology if more than one node is detected within
> CPU package. During the system boot this logic cuts out
> the DIE topology level and numa code adds NUMA level
> on top of this.

Its not SMT topology, and back when I asked if he had CoD enabled or
such he said not.


See also:

http://lkml.kernel.org/r/1471559812-19967-3-git-send-email-srinivas.pandruv...@linux.intel.com

Arguably, that should have been split in two patches, but alas..

[PATCH] x86/topology: Fallback to SMT level only once

[Jiri Olsa]

Jirka, Peter and Jean-Pierre reported performance drop on
some cpus after making cpu offline and online again.

The reason is the kernel logic that falls back to SMT
level topology if more than one node is detected within
CPU package. During the system boot this logic cuts out
the DIE topology level and numa code adds NUMA level
on top of this.

After the boot if you reboot make the cpu offline and online
again, this logic resets the SMT level topology, removing
whole NUMA level stuff.

Ensuring the SMT topology fallback happens only once during
the boot so the NUMA topology level is kept once it's built.

This problem is one of the issues reported in the wastedcores
article [1]. There's similar patch to this issue attached
to the article [2].

[1] https://github.com/jplozi/wastedcores
[2] 
https://github.com/jplozi/wastedcores/blob/master/patches/missing_sched_domains_linux_4.1.patch

Cc: Jean-Pierre Lozi 
Cc: Jirka Hladky 
Cc: Petr Surý 
Signed-off-by: Jiri Olsa 
---
 arch/x86/kernel/smpboot.c | 12 +++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2a6e84a30a54..f2a769b2b3fe 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -487,7 +487,17 @@ static struct sched_domain_topology_level 
numa_inside_package_topology[] = {
  */
 static void primarily_use_numa_for_topology(void)
 {
-   set_sched_topology(numa_inside_package_topology);
+   static bool once;
+
+   /*
+* We need to run it only during boot, once we are
+* here due to getting cpu online again we have already
+* NUMA topology setup done.
+*/
+   if (!once) {
+   set_sched_topology(numa_inside_package_topology);
+   once = true;
+   }
 }
 
 void set_cpu_sibling_map(int cpu)
-- 
2.7.4

[PATCH] x86/topology: Fallback to SMT level only once

[Jiri Olsa]

Jirka, Peter and Jean-Pierre reported performance drop on
some cpus after making cpu offline and online again.

The reason is the kernel logic that falls back to SMT
level topology if more than one node is detected within
CPU package. During the system boot this logic cuts out
the DIE topology level and numa code adds NUMA level
on top of this.

After the boot if you reboot make the cpu offline and online
again, this logic resets the SMT level topology, removing
whole NUMA level stuff.

Ensuring the SMT topology fallback happens only once during
the boot so the NUMA topology level is kept once it's built.

This problem is one of the issues reported in the wastedcores
article [1]. There's similar patch to this issue attached
to the article [2].

[1] https://github.com/jplozi/wastedcores
[2] 
https://github.com/jplozi/wastedcores/blob/master/patches/missing_sched_domains_linux_4.1.patch

Cc: Jean-Pierre Lozi 
Cc: Jirka Hladky 
Cc: Petr Surý 
Signed-off-by: Jiri Olsa 
---
 arch/x86/kernel/smpboot.c | 12 +++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/smpboot.c b/arch/x86/kernel/smpboot.c
index 2a6e84a30a54..f2a769b2b3fe 100644
--- a/arch/x86/kernel/smpboot.c
+++ b/arch/x86/kernel/smpboot.c
@@ -487,7 +487,17 @@ static struct sched_domain_topology_level 
numa_inside_package_topology[] = {
  */
 static void primarily_use_numa_for_topology(void)
 {
-   set_sched_topology(numa_inside_package_topology);
+   static bool once;
+
+   /*
+* We need to run it only during boot, once we are
+* here due to getting cpu online again we have already
+* NUMA topology setup done.
+*/
+   if (!once) {
+   set_sched_topology(numa_inside_package_topology);
+   once = true;
+   }
 }
 
 void set_cpu_sibling_map(int cpu)
-- 
2.7.4