Re: "movb" for spin-unlock (was Re: namei() query)

From: Rogier Wolff (
Date: Mon Apr 24 2000 - 00:25:03 EST

Linus Torvalds wrote:
[Charset ISO-8859-1 unsupported, filtering to ASCII...]
> On Mon, 24 Apr 2000, Jakob _stergaard wrote:
> >
> > Dual PPro stepping 1 here. I was just finishing my ``success report'' as the
> > program locked up :(
> Good. Not because I like the result, but because this finally puts the
> thing to rest with not a shadow of a doubt. Andy Glew (and - modesty fails
> me again - I) was right. The only really universally safe unlock is indeed
> one that is atomic.
> However, I still think it would be fine to make this a config option,
> because obviously newer PPro (and all PII) cores will do the unlock
> correctly and much faster with the simple store.
> The question is really only how the config option should be presented to
> the user (a "generic" kernel would obviously have the CONFIG_SLOW_UNLOCK
> thing enabled, but should be for example just split the "PPro" config
> entry into "PPro" and "PII", and then for "PPro" have a furter "Slow but
> safe" option?)
> I think the "PPro" vs "PII/PIII" case would be nice - then people with
> newer CPU's would never even see the "is your CPU broken"? question.

Would it be an idea to just "go all the way" then? Make the menu split
PII and PIII. So far, this doesn't make a difference, but it might in
the future.

Also, this is something where Intel should be able to disclose which
processors actually have this problem. Testing shows: Stepping 9 and
above seems OK, stepping 1 is not OK.

And as, nowadays it is possible to upgrade the CPU in-circuit, even
under Linux, wouldn't it be an idea to just require the microcode
upgrade? This is EXACTLY what the microcode upgrade is for.


** ** ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
*       Common sense is the collection of                                *
******  prejudices acquired by age eighteen.   -- Albert Einstein ********

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to Please read the FAQ at

This archive was generated by hypermail 2b29 : Sun Apr 30 2000 - 21:00:07 EST