Re: PATCH 2.3.23.5: ACPI iomap fix
Johannes Erdfelt (jerdfelt@sventech.com)
Fri, 22 Oct 1999 17:05:05 -0400
On Fri, Oct 22, 1999, Jeff Garzik <jgarzik@mandrakesoft.com> wrote:
> Andy Henroid wrote:
> >
> > --- Jeff Garzik <jgarzik@mandrakesoft.com> wrote:
> > >
> > > Attached is a patch against 2.3.23 pre5 which fixes
> > > ACPI with regards to iomapping. The warning
> > > messages added in pre5 brought to light this
> > > bug. This patch works on my K6 w/ ACPI.
> >
> > OK, so I see your fix with some changes from
> > Linus in pre6. It's still wrong for all those
> > FACP accesses so I'll send out a fix for that.
> >
> > Thanks for the portability catch. Also, the
> > inline assembly is a pretty glaring problem.
> > Guess that should go in a header, but which?
>
> I think the big question is whether or not ACPI will exist on Alpha, or
> other non-x86 platforms. Since Athlon and AXP will be sharing
> motherboard technology it would not surprise me.
>
> If that is the case, all the direct mem references would need to
> eventually be purged from the driver. If the inline assembler is
> required, it should probably ifdef'd CONFIG_X86 in a header somewhere if
> nothing else.
>
> As an aside, USB is _already_ on AXP, so inline assembler and other
> x86-isms in the USB code needs to be examined and compartmentalized, or
> purged...
The code is a minimum and puring it may be difficult. I think we need to
try to minimize it as much as possible.
The reason the UHCI driver uses inline assembly is because UHCI uses a
concurrent DMA engine and could be in a r/m/w cycle when we change some
pointers. Our changes them become overwritten by the old value which is
bad.
I think that the code can be minimized to one place and reuse that. I
wish I had better understanding of alpha and powerpc assembly. I guess
this is the time to learn :)
JE
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/