Messages in this thread | | | Date | Tue, 19 May 2015 08:04:37 -0700 | From | "H. Peter Anvin" <> | Subject | Re: [RFC PATCH] x86, espfix: use spin_lock rather than mutex |
| |
On 05/18/2015 12:43 PM, Andy Lutomirski wrote: > On 05/15/2015 12:27 AM, H. Peter Anvin wrote: >> On 05/14/2015 11:54 PM, Ingo Molnar wrote: >>> >>> The only slightly subtle detail with that is to use alloc_pages_node() >>> with the secondary CPU's node, to make sure the espfix stack is >>> NUMA-local to the CPU that is going to use it. >>> >> >> It doesn't hurt, although it isn't super critical as each page will be >> shared among 64 CPUs. The whole espfix stack is only a single cacheline >> long. >> > > I don't think we actually need these pages allocated until we try to run > user code. Can we move this very late in initialization instead? >
Yes, we could, as long as it is run on each CPU before that CPU tries to run user code.
-hpa
| |