Re: [PATCH 09/21] libs/guest: allow fetching a specific CPUID leaf from a cpu policy

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 23/03/2021 09:58, Roger Pau Monne wrote:
> diff --git a/tools/libs/guest/xg_cpuid_x86.c b/tools/libs/guest/xg_cpuid_x86.c
> index 48351f1c4c6..a1e1bf10d5c 100644
> --- a/tools/libs/guest/xg_cpuid_x86.c
> +++ b/tools/libs/guest/xg_cpuid_x86.c
> @@ -883,3 +883,45 @@ int xc_cpu_policy_serialise(xc_interface *xch, const 
> xc_cpu_policy_t p,
>      errno = 0;
>      return 0;
>  }
> +
> +int xc_cpu_policy_get_cpuid(xc_interface *xch, const xc_cpu_policy_t policy,
> +                            uint32_t leaf, uint32_t subleaf,
> +                            xen_cpuid_leaf_t *out)
> +{
> +    unsigned int nr_leaves, nr_msrs, i;
> +    xen_cpuid_leaf_t *leaves;
> +    int rc = xc_cpu_policy_get_size(xch, &nr_leaves, &nr_msrs);
> +
> +    if ( rc )
> +    {
> +        PERROR("Failed to obtain policy info size");
> +        return -1;
> +    }
> +
> +    leaves = calloc(nr_leaves, sizeof(*leaves));
> +    if ( !leaves )
> +    {
> +        PERROR("Failed to allocate resources");
> +        errno = ENOMEM;
> +        return -1;
> +    }
> +
> +    rc = xc_cpu_policy_serialise(xch, policy, leaves, &nr_leaves, NULL, 0);
> +    if ( rc )
> +        goto out;
> +
> +    for ( i = 0; i < nr_leaves; i++ )
> +        if ( leaves[i].leaf == leaf && leaves[i].subleaf == subleaf )
> +        {
> +            *out = leaves[i];
> +            goto out;
> +        }

Please adapt find_leaf(), probably by dropping xc_xend_cpuid and passing
in leaf/subleaf parameters.

Serialised leaves are sorted and there are plenty of them, so a log
search is better.

How frequent is this call going to be for the same policy?  With the
arrays embedded in a policy, they're still around, and serialise is an
expensive operation.

I wonder if it makes sense to try and keep both forms in sync, so we can
avoid redundant calls like this?

~Andrew