Messages in this thread | | | Date | Tue, 23 Aug 2011 22:56:16 +0200 | From | Borislav Petkov <> | Subject | Re: [uml-devel] SYSCALL, ptrace and syscall restart breakages (Re: [RFC] weird crap with vdso on uml/i386) |
| |
On Tue, Aug 23, 2011 at 03:15:58PM -0400, H. Peter Anvin wrote: > Right, but if you had said the difference had disappeared on current AMD > silicon it would be much less of an issue. That's why I wanted to find > that bit out from you.
Yeah, I dug this out in the APM, btw.
But no, I don't think the difference has disappeared - to the contrary, AFAICT, the intention is for SYSCALL to be the fastest way to do syscalls on x86 due to diminished number of segment checks etc. INT80 is legacy, slower, etc. I believe Andy measured a similar situation on Sandy Bridge with SYSCALL having latencies in the tens of nsecs range and INT80 being much slower. Ingo also measured a similar situation where the latency gap between the two on Intel is even bigger.
So, the only problem left now is what we're going to do with cases similar to what Al conjured up: http://marc.info/?l=linux-kernel&m=131412271112461&w=2
I don't know, maybe the most cowardly approach is to issue a warning and shrug with the shoulders, or do some asm magic... Also, do we care at all, how relevant is a case like that?
Hmmm.
-- Regards/Gruss, Boris.
Advanced Micro Devices GmbH Einsteinring 24, 85609 Dornach GM: Alberto Bozzo Reg: Dornach, Landkreis Muenchen HRB Nr. 43632 WEEE Registernr: 129 19551
| |