sysfs cpu?/topology is empty in 2.6.25 (32-bit Intel system)
Hi,
System topology on intel based system needs to be exported for non-numa case as well.
All parts of asm-i386/topology.h has come under #ifdef CONFIG_NUMA after the merge to asm-x86/topology.h
/sys/devices/system/cpu/cpu?/topology/* is populated based on ENABLE_TOPO_DEFINES
The sysfs cpu topology is not being populated on my dual socket dual core xeon 5160 processor based (x86 32 bit) system.
CONFIG_NUMA is not set in my case yet the topology is relevant and useful.
irqbalance daemon application depends on topology to build the cpus and package list and it fails on Fedora9 beta since the sysfs topology was not being populated in the 2.6.25 kernel.
I am not sure if it was intentional to not define ENABLE_TOPO_DEFINES for non-numa systems.
This fix has been tested on the above mentioned dual core, dual socket system.
Comments? Did I miss already posted fix?
--Vaidy
--- include/asm-x86/topology.h | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)
include/asm-x86/topology.h +10 -8
@@ -25,6 +25,16 @@#ifndef _ASM_X86_TOPOLOGY_H#define _ASM_X86_TOPOLOGY_H +#ifdef CONFIG_X86_32 +# ifdef CONFIG_X86_HT +# define ENABLE_TOPO_DEFINES +# endif +#else +# ifdef CONFIG_SMP +# define ENABLE_TOPO_DEFINES +# endif +#endif +#ifdef CONFIG_NUMA #include <linux/cpumask.h> #include <asm/mpspec.h>
@@ -112,10 +122,6 @@ extern unsigned long node_end_pfn[];extern unsigned long node_remap_size[];#define node_has_online_mem (nid) (node_start_pfn[nid] != node_end_pfn[nid]) -# ifdef CONFIG_X86_HT -# define ENABLE_TOPO_DEFINES -# endif -# define SD_CACHE_NICE_TRIES 1# define SD_IDLE_IDX 1# define SD_NEWIDLE_IDX 2
@@ -123,10 +129,6 @@ extern unsigned long node_remap_size[];#else -# ifdef CONFIG_SMP -# define ENABLE_TOPO_DEFINES -# endif -# define SD_CACHE_NICE_TRIES 2# define SD_IDLE_IDX 2# define SD_NEWIDLE_IDX 0
Hi,
I've tried this patch on an x86_64 system and it seems to be ok. One thing I've noticed is that turning on ENABLE_TOPO_DEFINES will turn on "arch_provides_topology_pointers". This indicates that the support functions can provide a pointer to various topology related cpu masks. I haven't checked the i386 arch code yet to determine if this is in fact true (or if the arch wants it to be true. It is a low impact area used for showing the cpu core and thread siblings.)
Thanks,
Mike
Quoted 78 lines · Vaidyanathan Srinivasan
Hi,
System topology on intel based system needs to be exported for non-numa case as well.
All parts of
asm-i386/topology.hhas come under #ifdefCONFIG_NUMAafter the merge toasm-x86/topology.h/sys/devices/system/cpu/cpu?/topology/* is populated based on
ENABLE_TOPO_DEFINESThe sysfs cpu topology is not being populated on my dual socket dual core xeon 5160 processor based (x86 32 bit) system.
CONFIG_NUMAis not set in my case yet the topology is relevant and useful.irqbalance daemon application depends on topology to build the cpus and package list and it fails on Fedora9 beta since the sysfs topology was not being populated in the 2.6.25 kernel.
I am not sure if it was intentional to not define
ENABLE_TOPO_DEFINESfor non-numa systems.This fix has been tested on the above mentioned dual core, dual socket system.
Comments? Did I miss already posted fix?
--Vaidy
Signed-off-by Vaidyanathan Srinivasan <[email protected]>--- include/asm-x86/topology.h | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-)include/asm-x86/topology.h +10
-8@@ -25,6 +25,16 @@#ifndef _ASM_X86_TOPOLOGY_H#define _ASM_X86_TOPOLOGY_H +#ifdef CONFIG_X86_32 +# ifdef CONFIG_X86_HT +# define ENABLE_TOPO_DEFINES +# endif +#else +# ifdef CONFIG_SMP +# define ENABLE_TOPO_DEFINES +# endif +#endif +#ifdef CONFIG_NUMA #include <linux/cpumask.h> #include <asm/mpspec.h> @@ -112,10 +122,6 @@ extern unsigned long node_end_pfn[];extern unsigned long node_remap_size[];#define node_has_online_mem (nid) (node_start_pfn[nid] != node_end_pfn[nid]) -# ifdef CONFIG_X86_HT -# define ENABLE_TOPO_DEFINES -# endif -# define SD_CACHE_NICE_TRIES 1# define SD_IDLE_IDX 1# define SD_NEWIDLE_IDX 2@@ -123,10 +129,6 @@ extern unsigned long node_remap_size[];#else -# ifdef CONFIG_SMP -# define ENABLE_TOPO_DEFINES -# endif -# define SD_CACHE_NICE_TRIES 2# define SD_IDLE_IDX 2# define SD_NEWIDLE_IDX 0
Ingo Molnar
AppliedQuoted 19 lines · Vaidyanathan Srinivasan
System topology on intel based system needs to be exported for non-numa case as well.
All parts of
asm-i386/topology.hhas come under #ifdefCONFIG_NUMAafter the merge toasm-x86/topology.h/sys/devices/system/cpu/cpu?/topology/* is populated based on
ENABLE_TOPO_DEFINESThe sysfs cpu topology is not being populated on my dual socket dual core xeon 5160 processor based (x86 32 bit) system.
CONFIG_NUMAis not set in my case yet the topology is relevant and useful.irqbalance daemon application depends on topology to build the cpus and package list and it fails on Fedora9 beta since the sysfs topology was not being populated in the 2.6.25 kernel.
Quoted 5 lines
I am not sure if it was intentional to not define
ENABLE_TOPO_DEFINESfor non-numa systems.This fix has been tested on the above mentioned dual core, dual socket system.
applied, thanks Vaidyanathan.
This was certainly not intentional - i suspect nobody noticed that irqbalanced was working without proper topology information?
Ingo
Quoted 25 lines · Ingo Molnar
Quoted 22 lines · Vaidyanathan Srinivasan
System topology on intel based system needs to be exported for non-numa case as well.
All parts of
asm-i386/topology.hhas come under #ifdefCONFIG_NUMAafter the merge toasm-x86/topology.h/sys/devices/system/cpu/cpu?/topology/* is populated based on
ENABLE_TOPO_DEFINESThe sysfs cpu topology is not being populated on my dual socket dual core xeon 5160 processor based (x86 32 bit) system.
CONFIG_NUMAis not set in my case yet the topology is relevant and useful.irqbalance daemon application depends on topology to build the cpus and package list and it fails on Fedora9 beta since the sysfs topology was not being populated in the 2.6.25 kernel.
I am not sure if it was intentional to not define
ENABLE_TOPO_DEFINESfor non-numa systems.
Quoted 7 lines
Quoted 2 lines
This fix has been tested on the above mentioned dual core, dual socket system.
applied, thanks Vaidyanathan.
This was certainly not intentional - i suspect nobody noticed that irqbalanced was working without proper topology information?
Hi Ingo,
Thank you for pushing the patch to 2.6.25-rc2. I have tested 2.6.26-rc2 for correct topology information and parsing by irqbalance daemon in the following system configurations:
32-bit Two socket dual core xeon 5160 processor system 32-bit Two socket quad core xeon E5450 processor system
--Vaidy