On Tue, Nov 27, 2007 at 06:49:13PM +0300, Sergei Shtylyov wrote:

Hello.

Hi Sergei,

Anton Vorontsov wrote:

quoted

This patch adds localbus and pata nodes to use CF IDE interface
on MPC8349E-mITX boards.

quoted

Patch also adds code to probe localbus.

quoted

Signed-off-by: Anton Vorontsov <redacted>
---
arch/powerpc/boot/dts/mpc8349emitx.dts    |   17 ++++++++++++++++-
arch/powerpc/platforms/83xx/mpc834x_itx.c |   17 +++++++++++++++++
2 files changed, 33 insertions(+), 1 deletions(-)

quoted

diff --git a/arch/powerpc/boot/dts/mpc8349emitx.dts 

b/arch/powerpc/boot/dts/mpc8349emitx.dts
index 5072f6d..7a97068 100644

--- a/arch/powerpc/boot/dts/mpc8349emitx.dts
+++ b/arch/powerpc/boot/dts/mpc8349emitx.dts

@@ -249,6 +249,21 @@

		device_type = "pci";
	};

+	localbus@e0005000 {
+		#address-cells = <2>;
+		#size-cells = <1>;
+		compatible = "fsl,mpc8349emitx-localbus",

   Board compatible bus?

This is what Documentation/powerpc/booting-without-of.txt suggests
for localbuses. I'm following.

quoted

+			     "fsl,mpc8349e-localbus",
+			     "fsl,pq2pro-localbus";
+		reg = <e0005000 d8>;
+		ranges = <3 0 f0000000 210>;

-
+		pata@3,0 {
+			compatible = "fsl,mpc8349emitx-pata", 
"pata-platform";
+			reg = <3 0 10 3 20c 4>;
+			ioport-shift = <1>;

   Bleh... that shift again. And this is surely not a good name for a 
property (where's I/O ports in your case?) -- why not call it "reg-shift" 
(well, I'd call it "reg-size" or "reg-stride" myself :-)?

1. "shift" because pata_platform using that name. I don't see any
   reason to contrive indirections. ioport-shift is what the whole
   Linux kernel using nowadays, and ioport-shift dts property
   anyway Linux-specific.

   I'm just following todays' conventions.

   If you feel really bad about that, I think better to fix that in
   the source of the badness -- pata_platform. It's easy, I can do
   that. Would you ack patch that converts whole pata_platform and
   users? Would Paul ack it?

   Still, is there any hardware that needs not power of 2 stride?

2. "ioport" because shift^Wstride ;-) applies only to the io range
   (yes, it's obvious, but worth open-wording, no?).


And btw, I can get rid of ioport-shift at all. And do fixups in
the pata_of_platform driver via .compatible matching. But I don't
want: it feels bad to list every needs-to-fixup board in the common
driver. It also feels not so great creating something like
pata-platform-stride-{1,2,4,...} compatible stuff. Heh.


Thanks,

-- 
Anton Vorontsov
email: cbou@mail.ru
backup email: ya-cbou@yandex.ru
irc://irc.freenode.net/bd2