Re: The BogoMIPS value sometimes too low on Intel Mobile P3
On Thu, Sep 08, 2005 at 12:28:06PM +0100, Martin Vlk wrote:
> Hi folks,
> I am running a custom-built kernel 2.6.10 on an Intel Mobile P3
> processor. (Acer TravelMate 620, Debian 3.1)
> >From time to time it happens to me that on boot-up the USB mouse
> >doesn't work.
> When I try a USB camera in this situation it doesn't work either.
> I discovered that when the USB devices don't work the BogoMIPS value
> calculated is far lower than it should be. For example it is 32
> whereas normally it should be around 1450.
> The low BogoMIPS value causes the USB init delays to be calculated too
> short and the devices are not given enough time to initialize.
> It happens in approximately 45% of boot-ups and the only way I know of
> to make it work again is restart. Very often restart is needed
> multiple times before it starts working again.
> I tried to switch processor power management settings in BIOS, but
> with no success so far. Also today I discovered another BIOS setting
> related to Intel SpeedStep technology and I tried to change the value
> to try if I get any better behaviour. Will see in few days.
> Attached are log examples from both successful and unsuccessful
> boot-ups. I noticed the detected processor speed varies and also
> different hi-res timesources are used (tsc, pmtmr). Is that of any
> Any idea what the problem and solution is?
Not off hand. Though it is almost certainly a kernel problem, and quite
likely an ACPI issue. Before investigating further, would it be possible
for you to test a newer kernel, like 2.6.13. As you are doing a custom
build, it shouldn't be too much bother to just get it from kernel.org.
Or if you are relying on debian packages, get linux-source-2.6.12 from
the Debian archive - we are currently working on getting 2.6.13 ready.
The next thing that I would try is booting with pci=noacpi, or trying
out the various acpi options listed in
Documentation/kernel-parameters.txt in the kernel tree. This may end up
making other things work in a less than optimal way, but it is a good
way to explore if the problem is acpi related or not.