From: Kumar Gala <hidden> Date: 2005-08-22 14:20:10
On Aug 22, 2005, at 9:12 AM, Olof Johansson wrote:
On Mon, Aug 22, 2005 at 04:10:30PM +1000, Stephen Rothwell wrote:
quoted
This moves all the very similar files - either the ppc64 file
included
the ppc file or they differed in simple comments.
[...]
quoted
include/asm-powerpc/pmac_feature.h | 379
++++++++++++++++++++++++++++++++++++
quoted
include/asm-powerpc/pmac_low_i2c.h | 43 ++++
Since the need is considered big enough to create a hierarchy for
the code under arch/powerpc/, wouldn't something corresponding be a
good idea for the include files? Or do we want to clutter them with
all platform-specific files in one directory?
I'm in agreement, but am wondering how many of the pmac specific
headers could end up in arch/powerpc/platforms/pmac in the future.
- kumar
From: Olof Johansson <hidden> Date: 2005-08-22 15:01:38
On Mon, Aug 22, 2005 at 09:20:10AM -0500, Kumar Gala wrote:
quoted
Since the need is considered big enough to create a hierarchy for
the code under arch/powerpc/, wouldn't something corresponding be a
good idea for the include files? Or do we want to clutter them with
all platform-specific files in one directory?
I'm in agreement, but am wondering how many of the pmac specific
headers could end up in arch/powerpc/platforms/pmac in the future.
Yeah, not sure -- I was thinking the same while I wrote that reply. There
are some pSeries and iSeries files that could maybe be moved there too.
-Olof
From: Benjamin Herrenschmidt <benh@kernel.crashing.org> Date: 2005-08-23 05:41:47
On Mon, 2005-08-22 at 09:20 -0500, Kumar Gala wrote:
On Aug 22, 2005, at 9:12 AM, Olof Johansson wrote:
quoted
On Mon, Aug 22, 2005 at 04:10:30PM +1000, Stephen Rothwell wrote:
quoted
This moves all the very similar files - either the ppc64 file
included
the ppc file or they differed in simple comments.
[...]
quoted
include/asm-powerpc/pmac_feature.h | 379
++++++++++++++++++++++++++++++++++++
quoted
include/asm-powerpc/pmac_low_i2c.h | 43 ++++
Since the need is considered big enough to create a hierarchy for
the code under arch/powerpc/, wouldn't something corresponding be a
good idea for the include files? Or do we want to clutter them with
all platform-specific files in one directory?
I'm in agreement, but am wondering how many of the pmac specific
headers could end up in arch/powerpc/platforms/pmac in the future.
Most of the ones that are in include/asm* are there because drivers
outside of arch/* use them.
Ben.
From: Kumar Gala <hidden> Date: 2005-08-23 17:05:36
On Aug 23, 2005, at 2:31 AM, Stephen Rothwell wrote:
On Tue, 23 Aug 2005 15:41:47 +1000 Benjamin Herrenschmidt
[off-list ref] wrote:
quoted
Most of the ones that are in include/asm* are there because drivers
outside of arch/* use them.
Indeed. Here is the result of grepping for all of the .h files in
patch 3 (minus the
obvious noncontentious ones like errno.h ...):
This seems reasonable then to have them in include/asm-powerpc. I
feel that there are a few files in include/asm-ppc that were only
used by code in arch/ppc/platforms but I'm probably just crazy.
- kumar
On Tue, Aug 23, 2005 at 12:05:36PM -0500, Kumar Gala wrote:
On Aug 23, 2005, at 2:31 AM, Stephen Rothwell wrote:
quoted
On Tue, 23 Aug 2005 15:41:47 +1000 Benjamin Herrenschmidt
[off-list ref] wrote:
quoted
Most of the ones that are in include/asm* are there because drivers
outside of arch/* use them.
Indeed. Here is the result of grepping for all of the .h files in
patch 3 (minus the
obvious noncontentious ones like errno.h ...):
This seems reasonable then to have them in include/asm-powerpc. I
No it doesn't. You can easily make the drivers get the includes from
<platforms/pmac/foo.h>, arch/$(ARCH) is already always in the search
path, I believe (if not, it's easy to add Makefile things to get it).
--
Tom Rini
http://gate.crashing.org/~trini/
On Wed, Aug 24, 2005 at 10:03:26PM +0200, Arnd Bergmann wrote:
On Middeweken 24 August 2005 20:20, Tom Rini wrote:
quoted
On Tue, Aug 23, 2005 at 12:05:36PM -0500, Kumar Gala wrote:
quoted
This seems reasonable then to have them in include/asm-powerpc. I
No it doesn't. You can easily make the drivers get the includes from
<platforms/pmac/foo.h>, arch/$(ARCH) is already always in the search
path, I believe (if not, it's easy to add Makefile things to get it).
I don't think any architecture except ppc32 has traditionally had
arch/$ARCH in its include path, and it would probably come as a surprise
to many kernel developers if this were used more.
True, but it's possible that the pmac drivers could also be converted to
use something else to pass around infos, register IO and such so the
headers can still live in arch/powerpc/platforms/pmac/
--
Tom Rini
http://gate.crashing.org/~trini/
On Middeweken 24 August 2005 20:20, Tom Rini wrote:
On Tue, Aug 23, 2005 at 12:05:36PM -0500, Kumar Gala wrote:
quoted
This seems reasonable then to have them in include/asm-powerpc. I
No it doesn't. You can easily make the drivers get the includes from
<platforms/pmac/foo.h>, arch/$(ARCH) is already always in the search
path, I believe (if not, it's easy to add Makefile things to get it).
I don't think any architecture except ppc32 has traditionally had
arch/$ARCH in its include path, and it would probably come as a surprise
to many kernel developers if this were used more.
Currently, there seem to be hardly any users of this "feature" in ppc32
that can't be trivially converted to including local files, so I'd
rather not see this moved over to arch/powerpc.
Arnd <><
On Wed, Aug 24, 2005 at 01:13:32PM -0700, Tom Rini wrote:
On Wed, Aug 24, 2005 at 10:03:26PM +0200, Arnd Bergmann wrote:
quoted
On Middeweken 24 August 2005 20:20, Tom Rini wrote:
quoted
No it doesn't. You can easily make the drivers get the includes from
<platforms/pmac/foo.h>, arch/$(ARCH) is already always in the search
path, I believe (if not, it's easy to add Makefile things to get it).
I don't think any architecture except ppc32 has traditionally had
arch/$ARCH in its include path, and it would probably come as a surprise
to many kernel developers if this were used more.
True, but it's possible that the pmac drivers could also be converted to
use something else to pass around infos, register IO and such so the
headers can still live in arch/powerpc/platforms/pmac/
These days, the macio bus is a real device-model bus. The code still
lives in drivers/macintosh, but it will be used by more than just
pmac machines whenever I get the 68k mac support finished. It seems
like we really should have an include/linux/macio.h, and make the
whole thing act more like a normal bus. I would like to abstract out
more of it anyway, particularly the DMA support. That way we wouldn't
have several different drivers for the same chip due to bus interface
differences (see mace.c and macmace.c for an example of current practice).
This would eliminate at least macio.h and dbdma.h. Hopefully it would
get most of the mac specific code organized in a more logical fashion.
Any comments/requirements/suggestions?
Brad Boyer
flar@allandria.com