From mboxrd@z Thu Jan 1 00:00:00 1970 From: Suravee Suthikulanit Subject: Re: Time Skewing on Windows XP Date: Tue, 12 Mar 2013 13:25:16 -0500 Message-ID: <513F730C.9030805@amd.com> References: <513EB1BC.9050803@amd.com> <291EDFCB1E9E224A99088639C4762022013F7D2D2E18@LONPMAILBOX01.citrite.net> <513F4FD7.8020305@amd.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <513F4FD7.8020305@amd.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Paul Durrant Cc: George Dunlap , Jan Beulich , "xen-devel@lists.xen.org" List-Id: xen-devel@lists.xenproject.org On 3/12/2013 10:55 AM, Suravee Suthikulanit wrote: > On 3/12/2013 4:13 AM, Paul Durrant wrote: >>> -----Original Message----- >>> From: xen-devel-bounces@lists.xen.org [mailto:xen-devel- >>> bounces@lists.xen.org] On Behalf Of Suravee Suthikulpanit >>> Sent: 12 March 2013 04:40 >>> To: George Dunlap; Jan Beulich >>> Cc: xen-devel@lists.xen.org >>> Subject: [Xen-devel] Time Skewing on Windows XP >>> >>> Hi, >>> >>> While I was investigating the following issue on Windows XP (both >>> 32-bit and >>> 64-bit): >>> >>> * AMD NPT performance regression after c/s 24770:7f79475d3de7 >>> Reference:http://marc.info/?l=xen-devel&m=135075376805215 >>> >>> >>> On the latest source form xen-unstable, I ran into an issue where >>> the timing >>> on the HVM guests skewing about 2x slower than the actual wall clock >>> time. >>> This results in the system time slowing down. This is regardless of the >>> cpufreq governor scaling. (I tried with both ondemand and performance). >>> >>> However, I don't see the same behavior on the Win7 HVM guests. Is >>> this a >>> known issue. I assume that XP and Win7 uses different mechanism for >>> keeping time (e.g. rdtsc vs. HPET)? >>> >> I think XP uses TSC and more recent versions of Windows use PM TIMER. >> Try booting XP with the /USEPMTIMER switch in boot.ini. >> >> Paul > Paul, > > That still doesn't help in this case. I still see the WindowsXP guest > time skewing. Also, I would not think the normal RDTSC skewing should > be this severe (0.5x comparing to the wall-clock time). > > Suravee Ok, I found that the WinXP is actually using PMTIMER. After inserting the debug messages, I can see that the function pmt_update_timer() is continuously being called by the handle_pmt_io(). Suravee > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel >