Baslinux mailing list

[baslinux AT lists.ibiblio.org]

I'm not sure if this discussion is on-topic enough, but the list
traffic has been minimal lately. Please let us know if we should be
discussing this somewhere else.

Sindi wrote:
> On Sat, 24 Mar 2007 baslinux AT lists.ibiblio.org [Sam] wrote:
>
> > some AMD cpus (the k6 and k7 in particular) suffer from the "hlt bug"
> > when used with ceartain chipsets (youll never know which they wont get
> > along with till you try -- it even happenes to some OEM boxen) and the
> > bug only shows when using NT/2k/xp versions of the 'doze and of course
> > linux... dos and 9x arent affected because they dont issue the 'hlt'
> > instruction...

The machine I am writing this on will spontaneously reset if 'no-hlt'
is used on power up, because of the sudden change in temperature. If I
let it reset a few times, it will stabilize and then work perfectly,
unless I reboot without no-hlt, in which case it will reset if the CPU
is idle (temperature drop). And if I cold boot with 'no-hlt', it also
works fine unless the CPU is used (fewer HLTs issued). Strange, eh? I
don't know anyone else with this problem, just thought it might be
interesting for those of us who use junk computers.

> > AFAIK there is no workaround for windows

There is a device driver and registry patch that forces Win9x to use
HLT. It's designed for qemu but also works on real hardware. XP might
have a similar hack available.

> > but in linux
> > just add append=no-hlt to lilo.conf or to the loadlin line and linux
> > will run fine...
>
> Thanks.  How do we determine if this is such a cpu?

If it seems to be "less broken" with either no-hlt or without it, the
computer likely has this problem.

See also the kernel documentation (search Config.in in kernel source
for no-hlt) for some other things to try when dealing with broken
hardware.

David