2009-10-29 16:27:21

by John Linn

[permalink] [raw]
Subject: virtual vs physical addresses to cache functions

Maybe this is a dumb question...

I believe that the kernel expects virtual addresses to be passed to the
defined cache functions across all architectures. Looking at PowerPC
and Blackfin confirm this I think.

Can someone confirm this or point to where it's documented as I have dug
for it without finding any specific details?

Thanks,
John

This email and any attachments are intended for the sole use of the named recipient(s) and contain(s) confidential information that may be proprietary, privileged or copyrighted under applicable law. If you are not the intended recipient, do not read, copy, or forward this email message or any attachments. Delete this email message and any attachments immediately.


2009-10-30 00:36:19

by Mike Frysinger

[permalink] [raw]
Subject: Re: virtual vs physical addresses to cache functions

On Thu, Oct 29, 2009 at 12:26, John Linn wrote:
> Maybe this is a dumb question...

you should see the stuff people put into hardware ...

> I believe that the kernel expects virtual addresses to be passed to the
> defined cache functions across all architectures.

i believe they're virtual since most (sane) hardware virtual memory
implementations are done with caches based on virtual addresses, but
what do i know (leading into next point ...)

> Looking at PowerPC and Blackfin confirm this I think.

the Blackfin arch (which is what i work on) is a bad example as it
lacks virtual memory support
-mike