Re: PAT support for i386 and x86_64

["Cédric Augonnet" <cedric.augonnet@gmail.com>]

[Apologize for the double-post, messed up with my mailer... ]

2007/8/8, Andi Kleen <andi@firstfloor.org>:
> > I don't see why we have to worry about cache corruption in the case at
> > hand. Write-combining is needed to map io (typically pci-mem regions)
> > which are never mapped cachable anywhere, including in the linear map.
>
> If we WC them using PAT then there would be a UC<->WC conflict with
> the direct mapping and possibly others. That's already undefined
> and not allowed.
>
> After some very bad experiences in the past I'm not going to take
> chances on this.
>
> We really need to keep all possible mappings synchronized.
>
> -Andi
>
>

Hi,

First thanks for your reactions.

Andi, what i don't understand is, if we only put a minimal patch, like only
modifying the register, perhaps we may avoid having all vendors doing
their own recipe, possibly all messing the one the other. As you said ,
changing this register is absolutely not the actual issue with PAT, but don't
you think such a first step is needed to avoid conflicts since people _do_
already set that register from their driver.

I agree there is work for a comprehensive support of PAT, especially when
dealing with ioremap, but even though we only handle a smallish portion
of the problem yet, perhaps it's time to at least modify that register ?

Of course there is no problem for that being dependant on a CONFIG_PAT.

Kind regards,
Cédric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/