[PATCH] x86, cpu, amd: Use 8-bit extractor for NodeId field

[PATCH] x86, cpu, amd: Use 8-bit extractor for NodeId field

[Aravind Gopalakrishnan]

The NodeId field in cpuid_ecx(0x8000001e) is a 8 bit field.
Although current extractor works fine, it will break if/when
the most significant 5 bits ever get used.

So, lets honor the field as architected in the BKDG

Reference: BKDG for AMD Fam15h Models 00h-0fh,
CPUID Fn8000_001E_ECX Node Identifiers

Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
---
 arch/x86/kernel/cpu/amd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
index a220239..9942b83 100644
--- a/arch/x86/kernel/cpu/amd.c
+++ b/arch/x86/kernel/cpu/amd.c
@@ -300,7 +300,7 @@ static void amd_get_topology(struct cpuinfo_x86 *c)
 
 		cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
 		nodes = ((ecx >> 8) & 7) + 1;
-		node_id = ecx & 7;
+		node_id = ecx & 0xff;
 
 		/* get compute unit information */
 		smp_num_siblings = ((ebx >> 8) & 3) + 1;
-- 
1.9.1

Re: [PATCH] x86, cpu, amd: Use 8-bit extractor for NodeId field

[Borislav Petkov]

On Mon, Jan 19, 2015 at 12:18:58PM -0600, Aravind Gopalakrishnan wrote:
> The NodeId field in cpuid_ecx(0x8000001e) is a 8 bit field.
> Although current extractor works fine, it will break if/when
> the most significant 5 bits ever get used.
> 
> So, lets honor the field as architected in the BKDG
> 
> Reference: BKDG for AMD Fam15h Models 00h-0fh,
> CPUID Fn8000_001E_ECX Node Identifiers
> 
> Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
> ---
>  arch/x86/kernel/cpu/amd.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
> index a220239..9942b83 100644
> --- a/arch/x86/kernel/cpu/amd.c
> +++ b/arch/x86/kernel/cpu/amd.c
> @@ -300,7 +300,7 @@ static void amd_get_topology(struct cpuinfo_x86 *c)
>  
>  		cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
>  		nodes = ((ecx >> 8) & 7) + 1;
> -		node_id = ecx & 7;
> +		node_id = ecx & 0xff;

That's {00000b,D18F0x60[NodeId]} and NodeId is 3 bits for enumerating
8 nodes so checking the 3 bits is actually correct as it says that the
other 5 are 0b.

Regardless, is this something hypothetical you've noticed from code
inspection or are we really going to have more than 8 nodes? If the
former, then patch is unnecessary churn.

-- 
Regards/Gruss,
    Boris.

ECO tip #101: Trim your mails when you reply.
--

Re: [PATCH] x86, cpu, amd: Use 8-bit extractor for NodeId field

[Aravind Gopalakrishnan]

On 1/19/2015 12:21 PM, Borislav Petkov wrote:
> On Mon, Jan 19, 2015 at 12:18:58PM -0600, Aravind Gopalakrishnan wrote:
>> The NodeId field in cpuid_ecx(0x8000001e) is a 8 bit field.
>> Although current extractor works fine, it will break if/when
>> the most significant 5 bits ever get used.
>>
>> So, lets honor the field as architected in the BKDG
>>
>> Reference: BKDG for AMD Fam15h Models 00h-0fh,
>> CPUID Fn8000_001E_ECX Node Identifiers
>>
>> Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@amd.com>
>> ---
>>   arch/x86/kernel/cpu/amd.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/x86/kernel/cpu/amd.c b/arch/x86/kernel/cpu/amd.c
>> index a220239..9942b83 100644
>> --- a/arch/x86/kernel/cpu/amd.c
>> +++ b/arch/x86/kernel/cpu/amd.c
>> @@ -300,7 +300,7 @@ static void amd_get_topology(struct cpuinfo_x86 *c)
>>   
>>   		cpuid(0x8000001e, &eax, &ebx, &ecx, &edx);
>>   		nodes = ((ecx >> 8) & 7) + 1;
>> -		node_id = ecx & 7;
>> +		node_id = ecx & 0xff;
> That's {00000b,D18F0x60[NodeId]} and NodeId is 3 bits for enumerating
> 8 nodes so checking the 3 bits is actually correct as it says that the
> other 5 are 0b.

Right. Just a concern that if the definition of those 5 bits changes 
then it would be a problem.

> Regardless, is this something hypothetical you've noticed from code
> inspection or are we really going to have more than 8 nodes? If the
> former, then patch is unnecessary churn.
>

Just something from code inspection. So not an immediate issue.

Thanks,
-Aravind.