High Speed Timer

[bobcal2008]

I'm running openSUSE 11.1 with the x86-64 bit 2.6.27.29.01 kernel as my host

VMware 6.5.3 build 185404

mostly Windows XP Pro guests

I sometimes receive a high speed timer error in my guest Windows XP Pro that says the high speed timer is not available and sometimes my guests slow down. The error says to read . The guest OS symptoms are exactly like in the knowledge base document, but the knowledge base says this shouldn't be a problem in the 2.6 kernel. The document also says to add "host.useFastclock = FALSE" to the configuration file. Should I do this? Is this document just old and needs to be up dated, or do I have a different problem?

[AWo]

Welcome to the forums!

The guest OS symptoms are exactly like in the knowledge base document, but the knowledge base says this shouldn't be a problem in the 2.6 kernel.

But it didn't say that this shouldn't be a problem with a SuSE 2.6 Kernel. Many distribution change the configuration of the standard kernel so it is possible that your kernel was compiled with different values.

Check the file mentioned in the KB article (Locate the following line in /usr/src/linux-2.4/include/asm-i386/param.h) for the parameter "#define HZ 100" and you know which value is used in your installation.

BTW, have you installed the VMware Tools inside your guests?

AWo

VCP / VMware vEXPERT 2009

• I rent firewood out ***

[bobcal2008]

Thanks!

I will try this tonight. Yes I am installing VMware tools.

[bobcal2008]

It was 100, I changed it. Now I just need to figured out how to compile the kernel. No small job for a newbie. But thanks. I will also change the guest config file and add the host.useFastclock = FALSE.

One last question. If changing the clock to 1000 fixed the Host clock, why tell the guest not to use it?

[AWo]

Whithout knowing it exactly I can only guess a little bit :

The frequency set in this file is (at least) responsible for the amount of processor interrupts fired per second to read the PIT hardware clock (n+1 where n is the amount of processors, so if this is set to 1000 and you have 2 processors 3.000 interrupts will be issued per second). That leads to an unreliable guest clock in earlier Linux kernels as only the system actually owning the CPU can do this (so many requests are lost). Read this for background information: http://www.vmware.com/pdf/vmware_timekeeping.pdf

If the frequency set there is to small (which helps to avoid the problem described above) the guest OS may need to ask the RTC clock for higher timer rates. But that could lead to the problems described in the KB article you pointed to. So if you set the timer frequenzy high enough (let's say 1000) there's no need to ask the RTC device. As the newer Linux kernels doesn't use the PIT timer anymore a higher frequency doesn't lead to the clock problems mentioned above. So using both, the switch not to use the RTC device and providing a frequency high enough in that file should address both issues.

AWo

VCP / VMware vEXPERT 2009

\*** I rent firewood out ***