Messages in this thread | | | Date | Tue, 15 May 2018 14:35:42 +0200 | From | Peter Zijlstra <> | Subject | Re: [PATCH] mm: Add new vma flag VM_LOCAL_CPU |
| |
On Tue, May 15, 2018 at 01:07:51PM +0100, Mark Rutland wrote: > // speculatively allocates TLB
Ohh, right, I completely forgot about that, but that actually does happen. We had trouble with AMD doing just that only about a year ago or so IIRC.
CPUs are completely free to speculatively load TLB entries for pages they never actually end up touching (typically prefetcher based), as long as there's valid page-tables for them at the time.
So yes, you're right.
| |