mapping back to 36 bit physical

mapping back to 36 bit physical

[Claudia Salzberg]

Greetings.
How does one go about aquiring the 36-bit physical address on a 440XX if
you have the corresponding va from an ioremap (when your driver is not
the one that did the ioremap so that you don't have the pa saved)?  I see
that ioremap does support 36-bit I/O mapping but _va and _pa only play
with unsigned long's.  So does iopa.  So do the virt_to_* 's I was able
to find.  Also, does anyone have plans for incorporating memory access
hooks above 4GB address space?

Thanks,

Claudia Salzberg


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: mapping back to 36 bit physical

[Tom Rini]

On Wed, Dec 18, 2002 at 02:07:45PM +0600, Claudia Salzberg wrote:

> Greetings.
> How does one go about aquiring the 36-bit physical address on a 440XX if
> you have the corresponding va from an ioremap (when your driver is not
> the one that did the ioremap so that you don't have the pa saved)? ?I see
> that ioremap does support 36-bit I/O mapping but _va and _pa only play
> with unsigned long's. ?So does iopa. ?So do the virt_to_* 's I was able
> to find. ?Also, does anyone have plans for incorporating memory access
> hooks above 4GB address space?

Erm.  Have you looked at the code in the linuxppc_2_4_devel tree?  I
believe all of these issues have already been addressed.

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: mapping back to 36 bit physical

[Matt Porter]

On Wed, Dec 18, 2002 at 02:07:45PM +0600, Claudia Salzberg wrote:
>
> Greetings.
> How does one go about aquiring the 36-bit physical address on a 440XX if
> you have the corresponding va from an ioremap (when your driver is not
> the one that did the ioremap so that you don't have the pa saved)?  I see
> that ioremap does support 36-bit I/O mapping but _va and _pa only play
> with unsigned long's.  So does iopa.  So do the virt_to_* 's I was able
> to find.  Also, does anyone have plans for incorporating memory access
> hooks above 4GB address space?

Everything is in the linuxppc-2.5 and linuxppc_2_4_devel to support all
normal kernel functionality with the exception of remap_page_range().
As part of a recent checkin, the commit notice gave a reference to an
ftp site to get a remap_page_range() 64-bit patch.  Check your changesets
or linuxppc-commit list archives.

You need to save the pa.  virt_to_* and friends are only valid for
kernel lowmem addresses as per Documentation/IO-mapping.txt.  Regardless,
all new drivers should be using the dma mapping api and saving the pa
as required.

Regards,
--
Matt Porter
porter@cox.net
This is Linux Country. On a quiet night, you can hear Windows reboot.

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/