[PATCH] CPUSETS: remove SMP dependency

[PATCH] CPUSETS: remove SMP dependency

[Magnus Damm]

Remove the SMP dependency from CPUSETS.

Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
---

 include/linux/sched.h |    4 ++++
 init/Kconfig          |    1 -
 2 files changed, 4 insertions(+), 1 deletion(-)

--- from-0002/include/linux/sched.h
+++ to-work/include/linux/sched.h	2005-10-03 16:18:15.000000000 +0900
@@ -590,6 +590,10 @@ struct sched_domain {
 
 extern void partition_sched_domains(cpumask_t *partition1,
 				    cpumask_t *partition2);
+#else /* CONFIG_SMP */
+
+static inline void partition_sched_domains(cpumask_t *partition1,
+					   cpumask_t *partition2) {}
 #endif /* CONFIG_SMP */
 
 
--- from-0002/init/Kconfig
+++ to-work/init/Kconfig	2005-10-03 15:40:51.000000000 +0900
@@ -245,7 +245,6 @@ config IKCONFIG_PROC
 
 config CPUSETS
 	bool "Cpuset support"
-	depends on SMP
 	help
 	  This option will let you create and manage CPUSETs which
 	  allow dynamically partitioning a system into sets of CPUs and

Re: [PATCH] CPUSETS: remove SMP dependency

[Paul Jackson]

> Remove the SMP dependency from CPUSETS.

Acked-by: Paul Jackson <pj@sgi.com>

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

[PATCH] CPUSETS: remove SMP dependency

[Magnus Damm]

Remove the SMP dependency from CPUSETS.

Signed-off-by: Magnus Damm <magnus@valinux.co.jp>
---

 include/linux/sched.h |    4 ++++
 init/Kconfig          |    1 -
 2 files changed, 4 insertions(+), 1 deletion(-)

diff -urNp linux-2.6.14-rc5-mm1/include/linux/sched.h linux-2.6.14-rc5-mm1-cpusets_no_smp/include/linux/sched.h
--- linux-2.6.14-rc5-mm1/include/linux/sched.h	2005-10-24 18:18:17.000000000 +0900
+++ linux-2.6.14-rc5-mm1-cpusets_no_smp/include/linux/sched.h	2005-10-26 16:42:34.000000000 +0900
@@ -636,6 +636,10 @@ struct sched_domain {
 
 extern void partition_sched_domains(cpumask_t *partition1,
 				    cpumask_t *partition2);
+#else /* CONFIG_SMP */
+
+static inline void partition_sched_domains(cpumask_t *partition1,
+					   cpumask_t *partition2) {}
 #endif /* CONFIG_SMP */
 
 
diff -urNp linux-2.6.14-rc5-mm1/init/Kconfig linux-2.6.14-rc5-mm1-cpusets_no_smp/init/Kconfig
--- linux-2.6.14-rc5-mm1/init/Kconfig	2005-10-24 18:18:17.000000000 +0900
+++ linux-2.6.14-rc5-mm1-cpusets_no_smp/init/Kconfig	2005-10-26 16:42:34.000000000 +0900
@@ -266,7 +266,6 @@ config IKCONFIG_PROC
 
 config CPUSETS
 	bool "Cpuset support"
-	depends on SMP
 	help
 	  This option will let you create and manage CPUSETs which
 	  allow dynamically partitioning a system into sets of CPUs and

Re: [PATCH] CPUSETS: remove SMP dependency

[Andrew Morton]

Magnus Damm <magnus@valinux.co.jp> wrote:
>
> Remove the SMP dependency from CPUSETS.


Why?

Re: [PATCH] CPUSETS: remove SMP dependency

[Magnus Damm]

On 10/26/05, Andrew Morton <akpm@osdl.org> wrote:
> Magnus Damm <magnus@valinux.co.jp> wrote:
> >
> > Remove the SMP dependency from CPUSETS.
>
> Why?

http://www.ussg.iu.edu/hypermail/linux/kernel/0510.0/0279.html

/ magnus

Re: [PATCH] CPUSETS: remove SMP dependency

[Paul Jackson]

Andrew, responding to Magnus:
> > Remove the SMP dependency from CPUSETS.
> 
> Why?

As described in the posting that Magnus linked to, it seemed to me
like the cleaner approach - if there is no need to link two CONFIG
options, then don't.

Perhaps someone wants to build a uni-processor (UP) kernel for other
reasons, but still have it support running some stuff that depends
on cpusets being present.

But my vote on UP issues isn't worth much.  This patch is no biggie
to me either way.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

Re: [PATCH] CPUSETS: remove SMP dependency

[Magnus Damm]

On 10/27/05, Paul Jackson <pj@sgi.com> wrote:
> Andrew, responding to Magnus:
> > > Remove the SMP dependency from CPUSETS.
> >
> > Why?
>
> As described in the posting that Magnus linked to, it seemed to me
> like the cleaner approach - if there is no need to link two CONFIG
> options, then don't.
>
> Perhaps someone wants to build a uni-processor (UP) kernel for other
> reasons, but still have it support running some stuff that depends
> on cpusets being present.

Exactly, and the current Kconfig assumes that all NUMA systems are
configured with SMP. But that does not have to be the case, especially
with NUMA_EMU.

Depending on SMP only is wrong IMO because CPUSETS manages both sets
of cpus (SMP) and sets of mems (NUMA).

So CPUSETS should either depend on SMP || NUMA or nothing.

/ magnus