Re: [PATCH v5 0/6] Add support for FORM2 associativity

["Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>]

On 7/13/21 7:57 PM, Daniel Henrique Barboza wrote:
> Aneesh,
> 
> This series compiles with a configuration made with "pseries_le_defconfig"
> but fails with a config based on an existing RHEL8 config.
> 
> The reason, which is hinted in the robot replies in patch 4, is that you 
> defined
> a "__vphn_get_associativity" inside a #ifdef CONFIG_PPC_SPLPAR guard but 
> didn't
> define how the function would behave without the config, and you ended up
> using the function elsewhere.
> 
> This fixes the compilation but I'm not sure if this is what you intended
> for this function:
> 
> 
> diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
> index c68846fc9550..6e8551d16b7a 100644
> --- a/arch/powerpc/mm/numa.c
> +++ b/arch/powerpc/mm/numa.c
> @@ -680,6 +680,11 @@ static int vphn_get_nid(long lcpu)
> 
>   }
>   #else
> +static int __vphn_get_associativity(long lcpu, __be32 *associativity)
> +{
> +       return -1;
> +}
> +
>   static int vphn_get_nid(long unused)
>   {
>          return NUMA_NO_NODE;
> 
> 
> I'll post a new version of the QEMU FORM2 changes using these patches as 
> is (with
> the above fixup), but I guess you'll want to post a v6.
> 

kernel test robot did report that earlier and I have that fixed in my 
local tree. I haven't posted v6 yet because I want to close the review 
on the approach with v5 patchset.

-aneesh