Messages in this thread | | | Subject | Re: [PATCH] x86-64: espfix for 64-bit mode *PROTOTYPE* | From | "H. Peter Anvin" <> | Date | Mon, 21 Apr 2014 18:14:54 -0700 |
| |
I wanted to avoid the "another cpu made this allocation, now I have to free" crap, but I also didn't want to grab the lock if there was no work needed.
On April 21, 2014 6:06:19 PM PDT, Andrew Lutomirski <amluto@gmail.com> wrote: >On Mon, Apr 21, 2014 at 5:53 PM, H. Peter Anvin <hpa@zytor.com> wrote: >> Well, if 2^17 CPUs are allocated we might 2K pages allocated. We >could easily do a bitmap here, of course. NR_CPUS/64 is a small >number, and would reduce the code complexity. >> > >Even simpler: just get rid of the check entirely. That is, break out >of the higher level loops once one of them is set (this should be a >big speedup regardless) and don't allocate the page if the first PTE >is already pointing at something. > >After all, espfix_already_there is mostly a duplicate of >init_espfix_cpu. > >--Andy
-- Sent from my mobile phone. Please pardon brevity and lack of formatting.
| |