Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-16 Thread Alison Schofield
On Tue, Feb 16, 2021 at 12:29:02PM +0100, Peter Zijlstra wrote: > On Wed, Feb 10, 2021 at 02:11:34PM -0800, Alison Schofield wrote: > > > This is equivalent to determining if x86_has_numa_in_package. > > Do you think there is an opportunity to set x86_has_numa_in_package > > earlier, and use it

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-16 Thread Peter Zijlstra
On Wed, Feb 10, 2021 at 02:11:34PM -0800, Alison Schofield wrote: > This is equivalent to determining if x86_has_numa_in_package. > Do you think there is an opportunity to set x86_has_numa_in_package > earlier, and use it here and in set_cpu_sibling_map()? Sure. Not sure that's actually clearer

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Alison Schofield
On Wed, Feb 10, 2021 at 08:38:42PM +0100, Peter Zijlstra wrote: > On Wed, Feb 10, 2021 at 07:22:03AM -0800, Dave Hansen wrote: > > On 2/10/21 12:05 AM, Peter Zijlstra wrote: > > >> +if (IS_ENABLED(CONFIG_NUMA)) > > >> +set_cpu_bug(c, X86_BUG_NUMA_SHARES_LLC); > > >> } > >

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Peter Zijlstra
On Wed, Feb 10, 2021 at 07:22:03AM -0800, Dave Hansen wrote: > On 2/10/21 12:05 AM, Peter Zijlstra wrote: > >> + if (IS_ENABLED(CONFIG_NUMA)) > >> + set_cpu_bug(c, X86_BUG_NUMA_SHARES_LLC); > >> } > > This seens wrong too, it shouldn't be allowed pre SKX. And ideally only > > be allowed

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Dave Hansen
On 2/10/21 12:10 AM, Peter Zijlstra wrote: > On Tue, Feb 09, 2021 at 11:09:27PM +, Luck, Tony wrote: >>> +#define X86_BUG_NUMA_SHARES_LLCX86_BUG(25) /* CPU may >>> enumerate an LLC shared by multiple NUMA nodes */ >> >> During internal review I wondered why this is a "BUG" rather

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Dave Hansen
On 2/10/21 12:05 AM, Peter Zijlstra wrote: >> +if (IS_ENABLED(CONFIG_NUMA)) >> +set_cpu_bug(c, X86_BUG_NUMA_SHARES_LLC); >> } > This seens wrong too, it shouldn't be allowed pre SKX. And ideally only > be allowed when SNC is enabled. Originally, this just added a few more models

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Peter Zijlstra
On Tue, Feb 09, 2021 at 11:09:27PM +, Luck, Tony wrote: > > +#define X86_BUG_NUMA_SHARES_LLCX86_BUG(25) /* CPU may > > enumerate an LLC shared by multiple NUMA nodes */ > > During internal review I wondered why this is a "BUG" rather than a "FEATURE" > bit. > > Apparently, the

Re: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-10 Thread Peter Zijlstra
On Tue, Feb 09, 2021 at 02:39:43PM -0800, Alison Schofield wrote: > Commit 1340ccfa9a9a ("x86,sched: Allow topologies where NUMA nodes > share an LLC") added a vendor and model specific check to skip the > topology_sane() check for Intel's Sky Lake Server CPUs where NUMA > nodes shared an LLC. >

RE: [PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-09 Thread Luck, Tony
> +#define X86_BUG_NUMA_SHARES_LLC X86_BUG(25) /* CPU may > enumerate an LLC shared by multiple NUMA nodes */ During internal review I wondered why this is a "BUG" rather than a "FEATURE" bit. Apparently, the suggestion for "BUG" came from earlier community discussions.

[PATCH] x86, sched: Allow NUMA nodes to share an LLC on Intel platforms

2021-02-09 Thread Alison Schofield
Commit 1340ccfa9a9a ("x86,sched: Allow topologies where NUMA nodes share an LLC") added a vendor and model specific check to skip the topology_sane() check for Intel's Sky Lake Server CPUs where NUMA nodes shared an LLC. This topology is no longer a quirk for Intel CPUs as Ice Lake and Sapphire