Subject: Re: StrongARM K bug
To: None <steve@starswan.demon.co.uk>
From: Charles M. Hannum <root@ihack.net>
List: port-arm32
Date: 03/30/1999 04:59:06
I don't have much to say about the StrongARM bug -- except to say that
it does seem desirable to work around it transparently if possible --
but I want to take issue with this statement:
> after all, Linux manages to fix the 'F00F' bug
First of all, `Linux' didn't `fix' that. The first method for
`fixing' it (some klugery involving causing a page fault) came from
BSDI. It was my analysis that demonstrated that this method
introduced new bugs. The second method -- and the one we use -- came
(AFAIK) from Intel. So, saying that Linux `fixed' it is misleading;
they simply adopted a method devised by someone else.
Secondly, most operating systems include a workaround for it now.
Certainly NetBSD does; I wrote the code myself. It just doesn't put
up a big banner and yell at you about it, because there's no need to;
the bug is completely worked around in software and therefore, from a
user's perspective, doesn't exist any more. There isn't even a
measurable performance impact.