[PATCH v2] platform/x86: ISST: Fix optimization with use of numa

[PATCH v2] platform/x86: ISST: Fix optimization with use of numa

[Srinivas Pandruvada]

When numa is used to map CPU to PCI device, the optimized path to read
from cached data is not working and still calls _isst_if_get_pci_dev().

The reason is that when caching the mapping, numa information is not
available as it is read later. So move the assignment of
isst_cpu_info[cpu].numa_node before calling _isst_if_get_pci_dev().

Fixes: aa2ddd242572 ("platform/x86: ISST: Use numa node id for cpu pci dev mapping")
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
---
v2:
Fixed "Fixes tag" as suggested by Stephen Rothwell

Update:
	The fixes tag was wrong
This patch can wait for next merge window as these systems are not
shipped.

 drivers/platform/x86/intel_speed_select_if/isst_if_common.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/platform/x86/intel_speed_select_if/isst_if_common.c b/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
index 6f0cc679c8e5..8a4d52a9028d 100644
--- a/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
+++ b/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
@@ -379,6 +379,8 @@ static int isst_if_cpu_online(unsigned int cpu)
 	u64 data;
 	int ret;
 
+	isst_cpu_info[cpu].numa_node = cpu_to_node(cpu);
+
 	ret = rdmsrl_safe(MSR_CPU_BUS_NUMBER, &data);
 	if (ret) {
 		/* This is not a fatal error on MSR mailbox only I/F */
@@ -397,7 +399,6 @@ static int isst_if_cpu_online(unsigned int cpu)
 		return ret;
 	}
 	isst_cpu_info[cpu].punit_cpu_id = data;
-	isst_cpu_info[cpu].numa_node = cpu_to_node(cpu);
 
 	isst_restore_msr_local(cpu);
 
-- 
2.31.1

Re: [PATCH v2] platform/x86: ISST: Fix optimization with use of numa

[Hans de Goede]

Hi,

On 8/11/21 5:36 AM, Srinivas Pandruvada wrote:
> When numa is used to map CPU to PCI device, the optimized path to read
> from cached data is not working and still calls _isst_if_get_pci_dev().
> 
> The reason is that when caching the mapping, numa information is not
> available as it is read later. So move the assignment of
> isst_cpu_info[cpu].numa_node before calling _isst_if_get_pci_dev().
> 
> Fixes: aa2ddd242572 ("platform/x86: ISST: Use numa node id for cpu pci dev mapping")
> Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
> ---
> v2:
> Fixed "Fixes tag" as suggested by Stephen Rothwell

Thanks, but I just fixed this myself in the pdx86/for-next tree and I've
done a forced-push of for-next with the fixed Fixes tag.

Either way this is resolved now :)

Regards,

Hans




> 
> Update:
> 	The fixes tag was wrong
> This patch can wait for next merge window as these systems are not
> shipped.
> 
>  drivers/platform/x86/intel_speed_select_if/isst_if_common.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/x86/intel_speed_select_if/isst_if_common.c b/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
> index 6f0cc679c8e5..8a4d52a9028d 100644
> --- a/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
> +++ b/drivers/platform/x86/intel_speed_select_if/isst_if_common.c
> @@ -379,6 +379,8 @@ static int isst_if_cpu_online(unsigned int cpu)
>  	u64 data;
>  	int ret;
>  
> +	isst_cpu_info[cpu].numa_node = cpu_to_node(cpu);
> +
>  	ret = rdmsrl_safe(MSR_CPU_BUS_NUMBER, &data);
>  	if (ret) {
>  		/* This is not a fatal error on MSR mailbox only I/F */
> @@ -397,7 +399,6 @@ static int isst_if_cpu_online(unsigned int cpu)
>  		return ret;
>  	}
>  	isst_cpu_info[cpu].punit_cpu_id = data;
> -	isst_cpu_info[cpu].numa_node = cpu_to_node(cpu);
>  
>  	isst_restore_msr_local(cpu);
>  
>