Re: [PATCH v2 1/2] x86: Add cpu_detect_cache_sizes() to init_intel(), add Quark to legacy_cache()

[Thomas Gleixner <tglx@linutronix.de>]

On Sat, 4 Oct 2014, Bryan O'Donoghue wrote:
> @@ -381,7 +381,18 @@ static void init_intel(struct cpuinfo_x86 *c)
>  #endif
>  	}
>  
> -	l2 = init_intel_cacheinfo(c);
> +	/* Detect legacy cache sizes */
> +	cpu_detect_cache_sizes(c);
> +
> +	/*
> +	 * If cache_size has not been initialized via legacy_cache()
> +	 * probe it via init_intel_cacheinfo().
> +	 */
> +	if (c->x86_cache_size == 0)
> +		l2 = init_intel_cacheinfo(c);
> +	else
> +		l2 = c->x86_cache_size;

Looking a bit deeper. This wont work because cpu_detect_cache_sizes()
will set c->x86_cache_size for all cpus with extended_cpuid_level >=
0x80000005, which is the case for everything modern. So this results
in init_intel_cacheinfo() not being called anymore.

So we might need the following conditional:

	l2 = init_intel_cacheinfo(c);
+	if (!c->x86_cache_size) {
+		cpu_detect_cache_sizes(c);
+		l2 = c->x86_cache_size;
+	}

If that does not work because c->x86_cache_size is not 0 after
init_intel_cacheinfo() we really need to make it family/model
dependent.

Thanks,

	tglx