From: Paul Jackson <[email protected]>
Document the kernel boot parameter: relax_domain_level=.
Signed-off-by: Paul Jackson <[email protected]>
---
Documentation/cpusets.txt | 3 +++
Documentation/kernel-parameters.txt | 4 ++++
2 files changed, 7 insertions(+)
--- 2.6.26-rc5-mm3.orig/Documentation/cpusets.txt 2008-06-27 06:55:34.000000000 -0700
+++ 2.6.26-rc5-mm3/Documentation/cpusets.txt 2008-06-29 21:49:53.984674621 -0700
@@ -546,6 +546,9 @@ otherwise initial value -1 that indicate
( 4 : search nodes in a chunk of node [on NUMA system] )
( 5~ : search system wide [on NUMA system])
+The system default is 0 (no search). The system default can be
+changed using the relax_domain_level= boot parameter.
+
This file is per-cpuset and affect the sched domain where the cpuset
belongs to. Therefore if the flag 'sched_load_balance' of a cpuset
is disabled, then 'sched_relax_domain_level' have no effect since
--- 2.6.26-rc5-mm3.orig/Documentation/kernel-parameters.txt 2008-06-27 06:28:26.000000000 -0700
+++ 2.6.26-rc5-mm3/Documentation/kernel-parameters.txt 2008-06-29 21:51:10.477997208 -0700
@@ -1732,6 +1732,10 @@ and is between 256 and 4096 characters.
Format: <reboot_mode>[,<reboot_mode2>[,...]]
See arch/*/kernel/reboot.c or arch/*/kernel/process.c
+ relax_domain_level=
+ [KNL, SMP] Set schedulers default relax_domain_level.
+ See Documentation/cpusets.txt.
+
reserve= [KNL,BUGS] Force the kernel to ignore some iomem area
reservetop= [X86-32]
--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <[email protected]> 1.650.933.1373
Paul Jackson wrote:
> From: Paul Jackson <[email protected]>
>
> Document the kernel boot parameter: relax_domain_level=.
>
> Signed-off-by: Paul Jackson <[email protected]>
>
> ---
> Documentation/cpusets.txt | 3 +++
> Documentation/kernel-parameters.txt | 4 ++++
> 2 files changed, 7 insertions(+)
>
> --- 2.6.26-rc5-mm3.orig/Documentation/cpusets.txt 2008-06-27 06:55:34.000000000 -0700
> +++ 2.6.26-rc5-mm3/Documentation/cpusets.txt 2008-06-29 21:49:53.984674621 -0700
> @@ -546,6 +546,9 @@ otherwise initial value -1 that indicate
> ( 4 : search nodes in a chunk of node [on NUMA system] )
> ( 5~ : search system wide [on NUMA system])
>
> +The system default is 0 (no search). The system default can be
> +changed using the relax_domain_level= boot parameter.
> +
> This file is per-cpuset and affect the sched domain where the cpuset
> belongs to. Therefore if the flag 'sched_load_balance' of a cpuset
> is disabled, then 'sched_relax_domain_level' have no effect since
> --- 2.6.26-rc5-mm3.orig/Documentation/kernel-parameters.txt 2008-06-27 06:28:26.000000000 -0700
> +++ 2.6.26-rc5-mm3/Documentation/kernel-parameters.txt 2008-06-29 21:51:10.477997208 -0700
> @@ -1732,6 +1732,10 @@ and is between 256 and 4096 characters.
> Format: <reboot_mode>[,<reboot_mode2>[,...]]
> See arch/*/kernel/reboot.c or arch/*/kernel/process.c
>
> + relax_domain_level=
> + [KNL, SMP] Set schedulers default relax_domain_level.
> + See Documentation/cpusets.txt.
> +
is [KNL,SMP] more consistent ?
and s/schedulers/scheduler's/
but maybe I'm wrong due to my English skill.
> reserve= [KNL,BUGS] Force the kernel to ignore some iomem area
>
> reservetop= [X86-32]
>
Reviewed-by: Li Zefan <[email protected]>
Hi Paul,
Paul Jackson wrote:
> From: Paul Jackson <[email protected]>
>
> Document the kernel boot parameter: relax_domain_level=.
>
> Signed-off-by: Paul Jackson <[email protected]>
>
> ---
> Documentation/cpusets.txt | 3 +++
> Documentation/kernel-parameters.txt | 4 ++++
> 2 files changed, 7 insertions(+)
>
> --- 2.6.26-rc5-mm3.orig/Documentation/cpusets.txt 2008-06-27 06:55:34.000000000 -0700
> +++ 2.6.26-rc5-mm3/Documentation/cpusets.txt 2008-06-29 21:49:53.984674621 -0700
> @@ -546,6 +546,9 @@ otherwise initial value -1 that indicate
> ( 4 : search nodes in a chunk of node [on NUMA system] )
> ( 5~ : search system wide [on NUMA system])
>
> +The system default is 0 (no search). The system default can be
> +changed using the relax_domain_level= boot parameter.
The system default is architecture dependent. You can change your
system's default using the relax_domain_level= boot parameter.
Thanks,
H.Seto
On Mon, Jun 30, 2008 at 8:08 AM, Li Zefan <[email protected]> wrote:
> Paul Jackson wrote:
>> From: Paul Jackson <[email protected]>
>>
>> Document the kernel boot parameter: relax_domain_level=.
>>
>> Signed-off-by: Paul Jackson <[email protected]>
>>
>> ---
>> Documentation/cpusets.txt | 3 +++
>> Documentation/kernel-parameters.txt | 4 ++++
>> 2 files changed, 7 insertions(+)
>>
>> --- 2.6.26-rc5-mm3.orig/Documentation/cpusets.txt 2008-06-27 06:55:34.000000000 -0700
>> +++ 2.6.26-rc5-mm3/Documentation/cpusets.txt 2008-06-29 21:49:53.984674621 -0700
>> @@ -546,6 +546,9 @@ otherwise initial value -1 that indicate
>> ( 4 : search nodes in a chunk of node [on NUMA system] )
>> ( 5~ : search system wide [on NUMA system])
>>
>> +The system default is 0 (no search). The system default can be
>> +changed using the relax_domain_level= boot parameter.
>> +
>> This file is per-cpuset and affect the sched domain where the cpuset
>> belongs to. Therefore if the flag 'sched_load_balance' of a cpuset
>> is disabled, then 'sched_relax_domain_level' have no effect since
>> --- 2.6.26-rc5-mm3.orig/Documentation/kernel-parameters.txt 2008-06-27 06:28:26.000000000 -0700
>> +++ 2.6.26-rc5-mm3/Documentation/kernel-parameters.txt 2008-06-29 21:51:10.477997208 -0700
>> @@ -1732,6 +1732,10 @@ and is between 256 and 4096 characters.
>> Format: <reboot_mode>[,<reboot_mode2>[,...]]
>> See arch/*/kernel/reboot.c or arch/*/kernel/process.c
>>
>> + relax_domain_level=
>> + [KNL, SMP] Set schedulers default relax_domain_level.
>> + See Documentation/cpusets.txt.
>> +
>
> is [KNL,SMP] more consistent ?
>
> and s/schedulers/scheduler's/
>
> but maybe I'm wrong due to my English skill.
You are right. (Paul doesn't like apostrophes ;-).)
>> reserve= [KNL,BUGS] Force the kernel to ignore some iomem area
>>
>> reservetop= [X86-32]
>>
>
> Reviewed-by: Li Zefan <[email protected]>
>
>
--
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
man-pages online: http://www.kernel.org/doc/man-pages/online_pages.html
Found a bug? http://www.kernel.org/doc/man-pages/reporting_bugs.html
Hidetoshi Seto wrote:
> Hi Paul,
>
> Paul Jackson wrote:
>> From: Paul Jackson <[email protected]>
>>
>> Document the kernel boot parameter: relax_domain_level=.
>>
>> Signed-off-by: Paul Jackson <[email protected]>
>>
>> ---
>> Documentation/cpusets.txt | 3 +++
>> Documentation/kernel-parameters.txt | 4 ++++
>> 2 files changed, 7 insertions(+)
>>
>> --- 2.6.26-rc5-mm3.orig/Documentation/cpusets.txt 2008-06-27 06:55:34.000000000 -0700
>> +++ 2.6.26-rc5-mm3/Documentation/cpusets.txt 2008-06-29 21:49:53.984674621 -0700
>> @@ -546,6 +546,9 @@ otherwise initial value -1 that indicate
>> ( 4 : search nodes in a chunk of node [on NUMA system] )
>> ( 5~ : search system wide [on NUMA system])
>>
>> +The system default is 0 (no search). The system default can be
>> +changed using the relax_domain_level= boot parameter.
>
> The system default is architecture dependent. You can change your
> system's default using the relax_domain_level= boot parameter.
>
but the system default is 0 by default. and 0 means no search no matter
what architecture is. ;)
> Thanks,
> H.Seto
>
>
Li Zefan wrote:
> Hidetoshi Seto wrote:
>> The system default is architecture dependent. You can change your
>> system's default using the relax_domain_level= boot parameter.
>
> but the system default is 0 by default. and 0 means no search no matter
> what architecture is. ;)
If my git pull works correctly, it seems that the default is still -1...
[kernel/sched.c]
> static int default_relax_domain_level = -1;
AFAIK, for example, x86(SMT) search siblings by default.
Thanks,
H.Seto
Hidetoshi Seto wrote:
> Li Zefan wrote:
>> Hidetoshi Seto wrote:
>>> The system default is architecture dependent. You can change your
>>> system's default using the relax_domain_level= boot parameter.
>> but the system default is 0 by default. and 0 means no search no matter
>> what architecture is. ;)
>
> If my git pull works correctly, it seems that the default is still -1...
>
Ah, yes, it's my mistake. :(
> [kernel/sched.c]
>> static int default_relax_domain_level = -1;
>
> AFAIK, for example, x86(SMT) search siblings by default.
>
So the system default is architecture depentdent.
> Thanks,
> H.Seto
>
>
Michael wrote:
> Paul doesn't like apostrophes ;-)
You got that right -- I hate them ;).
--
I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <[email protected]> 1.940.382.4214