Re: [PATCH v2] x86,sched: allow topologies where NUMA nodes share an LLC

[Peter Zijlstra]

On Thu, Mar 22, 2018 at 01:49:22PM -0700, Alison Schofield wrote:
> +	/*
> +	 * Some Intel CPUs enumerate an LLC that is shared by
> +	 * multiple NUMA nodes.  The LLC on these systems is
> +	 * shared for off-package data access but private to the
> +	 * NUMA node (half of the package) for on-package access.
> +	 *
> +	 * CPUID can only enumerate the cache as being shared *or*
> +	 * unshared, but not this particular configuration.  The
> +	 * CPU in this case enumerates the cache to be shared
> +	 * across the entire package (spanning both NUMA nodes).
> +	 */
> +	if (!topology_same_node(c, o) &&
> +	    (c->x86_vendor == X86_VENDOR_INTEL &&
> +	     c->x86_model == INTEL_FAM6_SKYLAKE_X)) {
> +		/* Use NUMA instead of coregroups for scheduling: */
> +		x86_has_numa_in_package = true;
> +
> +		/*
> +		 * Return value doesn't actually matter because we
> +		 * are throwing away coregroups for scheduling anyway.
> +		 * Return false to bypass topology broken bug messages
> +		 * and fixups in sched_domain().
> +		 */
> +		return false;

IIRC that return value _does_ matter because the resulting mask still
ends up user visible in sysfs.

IIRC I went over with this dhansen a week or so ago, but I cannot now
recall what we settled on as being the right return value and for what
reason.