Xen devel - Apr 2006 - [Xen-ia64-devel] PATCH: console.c: use ioremap instead of __va to map video memory

Hi,

on ia64 you need to map video are as uncached memory.
With this patch the vga console is now working on ia64.

Tested on ia64 only.

Tristan.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@lists.xensource.com
http://lists.xensource.com/xen-ia64-devel

On Fri, Apr 14, 2006 at 01:37:22PM +0100, Tristan Gingold
wrote:
> Hi,
> 
> on ia64 you need to map video are as uncached memory.
> With this patch the vga console is now working on ia64.
> 
> Tested on ia64 only.
It should work fine on x86 (though I haven''t tested!) since ioremap is
just defined to __va().

The comment in asm-x86/io.h says "we don''t need a real ioremap on
Xen/x86", but I think we do need it for the IOMMU work on x86-64,
since the hypervisor will need to map the IOMMU''s internal register
area for things like cache flushing. Is there a fundamental objection
to a real ioremap for x86?

Cheers,
Muli
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 14 Apr 2006, at 15:46, Muli Ben-Yehuda wrote:
> The comment in asm-x86/io.h says "we don''t need a real
ioremap on
> Xen/x86", but I think we do need it for the IOMMU work on x86-64,
> since the hypervisor will need to map the IOMMU''s internal
register
> area for things like cache flushing. Is there a fundamental objection
> to a real ioremap for x86?
No, it got removed as it wasn''t used, but it can easily be added back. 
It''s particularly easy since we probably don''t need iounmap().

  -- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, Apr 14, 2006 at 07:38:18PM +0100, Keir Fraser wrote:
> No, it got removed as it wasn''t used, but it can easily be added
back.
> It''s particularly easy since we probably don''t need
iounmap().
Excellent, thanks for the info.

Cheers,
Muli
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, Apr 14, 2006 at 07:38:18PM +0100, Keir Fraser
wrote:
> 
> On 14 Apr 2006, at 15:46, Muli Ben-Yehuda wrote:
> 
> >The comment in asm-x86/io.h says "we don''t need a real
ioremap on
> >Xen/x86", but I think we do need it for the IOMMU work on x86-64,
> >since the hypervisor will need to map the IOMMU''s internal
register
> >area for things like cache flushing. Is there a fundamental objection
> >to a real ioremap for x86?
> 
> No, it got removed as it wasn''t used, but it can easily be added
back.
> It''s particularly easy since we probably don''t need
iounmap().
Hmm, looking at the hg logs, it looks like ioremap() was added for
x86-64 with the current __va() implementation. Can you please point me
at the old version?

Thanks,
Muli
-- 
Muli Ben-Yehuda
http://www.mulix.org | http://mulix.livejournal.com/


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 24 Apr 2006, at 08:01, Muli Ben-Yehuda wrote:
>> No, it got removed as it wasn''t used, but it can easily be
added back.
>> It''s particularly easy since we probably don''t need
iounmap().
>
> Hmm, looking at the hg logs, it looks like ioremap() was added for
> x86-64 with the current __va() implementation. Can you please point me
> at the old version?
If you want the old ioremap(), see for example changeset 1452, and look 
for xen/arch/i386/ioremap.c. That implementation would still work even 
for x86/64 -- the bit that directly changes page tables should actually 
call map_pages_to_xen().

  -- Keir


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel