Please note that I am not a subscriber to the linux-kernel mailing list but
*am* a subscriber to the smp list.
On Wed, 12 Apr 2000, you wrote:
> > I have not tried the 2.3.99 kernels but with 2.2.14 and earlier 2.3
> > kernels, my system had clock problems. These arose because the kernel
> > erroneously assumed that the TSC counts for both my processors would
> > increase at the same rate. My processors are different speeds.
> (well it's more like a BIOS bug but hey.)
Fair enough.
> > On 2.2.14, I tracked down and modified a patch which simply told the
> > kernel not
> > to use the TSC.
> >
> > Do you think this problem has been resolved with this kernel? If you
> > know for sure, great, let me know. Otherwise, I'll test it on my own
> > some time after I've been able to sleep.
>
> yep, i've written a patch for 2.2 to do this, check out:
>
> http://www.redhat.com/~mingo/smp-timersync-patches/
>
> i've integrated this into 2.3 around 2.3.21, so your box should work just
> fine under 2.3.99-pre5 + smp-2.3.99-5-B2.
It does not do so. I just tested it out on my system. The best way for me to
determine that my system does not work properly is to use ping. Under
2.3.99-pre5 + smp-2.3.99-5-B2, I get some ping times that are reasonable but I
also get some that are not. For example, about half the ping times are at
about 429494900 ms or greater. I get fewer ping times that are also incorrect
but are on the order of 5000 ms.
I will just point out that under 2.2.14 with my TSC patch, ping times all work
fine. This is not just a problem with ping, I see numerous other problems on
my system as a result of this, all of which go away with 2.2.14 and the TSC
patch. The TSC patch simply adds an option to configuration to DISABLE TSC
code in the kernel. It does nothing else and I only assembled it, I didn't
write it.
Here are some snippets from my syslog:
Apr 13 13:08:32 hatred kernel: calibrating APIC timer ...
Apr 13 13:08:32 hatred kernel: ..... CPU clock speed is 400.9489 MHz.
Apr 13 13:08:32 hatred kernel: ..... host bus clock speed is 66.8246 MHz.
Apr 13 13:08:32 hatred kernel: cpu: 0, clocks: 668246, slice: 222748
Apr 13 13:08:32 hatred kernel: CPU0<C0:668240,C:445488,D:4,S:222748,C:668246>
Apr 13 13:08:32 hatred kernel: cpu: 1, clocks: 668246, slice: 222748
Apr 13 13:08:32 hatred kernel: CPU1<C0:668240,C:222736,D:8,S:222748,C:668246>
Apr 13 13:08:32 hatred kernel: checking TSC synchronization across CPUs:
Apr 13 13:08:32 hatred kernel: BIOS BUG: CPU#0 improperly initialized, has -2492
273 usecs TSC skew! FIXED.
Apr 13 13:08:32 hatred kernel: BIOS BUG: CPU#1 improperly initialized, has 24922
73 usecs TSC skew! FIXED.
Apr 13 13:08:32 hatred kernel: Setting commenced=1, go go go
Apr 13 13:08:32 hatred kernel: mtrr: your CPUs had inconsistent fixed MTRR setti
ngs
Apr 13 13:08:32 hatred kernel: mtrr: probably your BIOS does not setup all CPUs
-- Christopher Thompson http://hypocrite.org/ One Ring to rule them all, One Ring to find them, One Ring to bring them all and in the darkness bind them -- Fellowship of the Rings, movie, Christmas 2001- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Apr 15 2000 - 21:00:22 EST