Re: wild gettimeofday on smp under 2.2.18

Peter T. Breuer (ptb@oboe.it.uc3m.es)
Wed, 10 Jan 2001 01:04:45 +0100 (MET)


"Alan Cox wrote:"
> Doesnt seem very wild to me, but something did go back 3uS which would imply
> the CPU tsc's are not synched. 2.2 doesnt like that, Boot with 'notsc' and
> repeat the experiment

No change with that, as you already know. But I just noticed that the
ekernel was compiled for i386. I've recompiled 2.2.18 for i686, and
this comes out of the boot messages:

checking TSC synchronization across CPUs:
BIOS BUG: CPU#0 improperly initialized, has 2045320 usecs TSC skew! FIXED.
BIOS BUG: CPU#1 improperly initialized, has -2045320 usecs TSC skew! FIXED.
PCI: PCI BIOS revision 2.10 entry at 0xf0730

This is the ASUS BX dual board with onboard scsi.

But the FIXED above doesn't seem to be true. I still get results like
those below from gettimeofday calls every five seconds. Clearly two
mixed sequences ..

978042124 !
978042135 *
978042134 !
978042139 !
978042144 !
978042149 !
978042158 *
978042159 !
978042164 !
978042169 !
978042176

Peter
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/