Re: oops in bk pull (oct 03)

[Linus Torvalds]

On Fri, 4 Oct 2002, Linus Torvalds wrote:
> 
> I think that is the real issue. We're mapping something - probably a host
> bridge - at address 0, and then accessing RAM (which is also is mapped at 
> PCI address 0) and the host bridge is unhappy.
> 
> So excluding the change is probably the right thing to do - it's just 
> fundamentally buggy to blindly put a base register at zero.

The more I think about this, the more convinced I am this is the case. We
just _mustn't_ set up a live PCI window at address 0, and expect it to not
cause confusion.

Also, we've seen before that we must not blindly disable a PCI window
either, since that will kill the system when the host bridge is disabled
and there is any pending DMA, for example (*). We saw that earlier in the
2.4.x tree - some host bridges will just ignore the disable (which means
that then we'd trigger the zero-base bug), and others will honour the
disable (which in turn will cause the DMA and other random problems).

This is all probably dependently on host bridge / MCH behaviour, so it
probably works fine on 90%+ of all machines, but clearly breaks enough to
not be a viable approach in general.

Ergo, the patch that looked so simple at first glance was really broken
for a number of really subtle reasons. 

		Linus

(*) And pending DMA is actually _normal_ on PC's at early bootup when we 
enumerate the PCI system - it's how USB keyboard and mouse emulation is 
done, together with SMI support in the BIOS.