ppc64-sym2-fix.patch

ppc64-sym2-fix.patch

[Matthew Wilcox]

Hi Anton.  Looks like you forgot to cc me on this one:

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test7/2.6.0-test7-mm1/broken-out/ppc64-sym2-fix.patch

On first look, you're right.  We should do something more sensible if
pci_set_mwi() fails (eg, just report it to the user and carry on without
setting MWI).

But the comment indicates that some revisions of the 896 *require* MWI
to work around chip errata.  So I don't particularly want to do that.

The comment in include/asm-ppc64/pci.h seems to indicate that it actually
wouldn't do any harm to return success since the "hardware treats MWI
the same as memory write".

What do you think?

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: ppc64-sym2-fix.patch

[Anton Blanchard]

Hi,

> Hi Anton.  Looks like you forgot to cc me on this one:
> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.0-test7/2.6.0-test7-mm1/broken-out/ppc64-sym2-fix.patch

Yeah sorry about that, I did the quick fix and havent revisited it yet.

> The comment in include/asm-ppc64/pci.h seems to indicate that it actually
> wouldn't do any harm to return success since the "hardware treats MWI
> the same as memory write".

I didnt want to because we have certain firmware workarounds that
involve setting smaller cacheline sizes. I was trying to avoid undoing
that.

As Milton here just pointed out, the generic part of the MWI code only
sets the MWI bit (which we should ignore) so we should be good to do
this. I'll test and send the ppc64 specific fix to akpm.

Anton