Re: [PATCH] powerpc: Remove device_type = "rtc" properties in .dts files

On Wed, Oct 22, 2008 at 02:09:01PM -0500, Matt Sealey wrote:

Anton Vorontsov wrote:

quoted

On Wed, Oct 22, 2008 at 01:40:50PM -0500, Matt Sealey wrote:

I think I got it. ;-) You think that I'm not aware of that we _can_
use the device_type for matching the nodes. Well, I'm aware of it,
sure we can. ;-)

I'm sure you are aware, I am just a little jumpy regarding this as the 
whole ePAPR-is-official thing and the direction Linux is taking with 
regards to redefining part of the device tree specs, that this could have 
been something a little more serious :)

quoted

But we don't use it for the rtc nodes, and we don't want to encourage
the usage for the flat trees. And that's the point of this patch.

Would it not be prudent to, while not actively encouraging it, at least 
mention device_type in any specifications as a legacy item (for real Open 
Firmware only) and for if a device should be in the tree as a generic, 
IEEE 1275-style device (i.e. there would be a set of well-defined client 
interface methods for it in a real OF)?

My basic concerns are for input/output as reported by /chosen - in case 
it is important exactly what is being used, there is at least one 
out-of-driver code snippet which checks if stdin and stdout are of type 
"serial" (or "failsafe") and auto-directs console to that - it would be 
nice to keep this clean and not dump a million serial-device-compatibles 
in another list here if someone wants to automatically choose between 
console output on the DIU or PSC for MPC5121e/MPC8610 for example, or 
wants to restrict the amount of fancy stuff it does on a terminal if it's 
a slow serial device, or perhaps even automatically invoke netconsole if 
it's set to "network"?

I know U-Boot doesn't have the intelligence to output to anything but a 
serial port these days on those devices, but as they say, there is no 
fate but what we make .. we should make sure it doesn't turn up that code 
is never suggested or attempted because supporting it in Linux would be 
too big a jump or too messy a patch :)

I don't feel competent to comment on "embedded-OF"/FDT design
decisions...

Let's Cc devicetree-discuss@ozlabs.org ?

-- 
Anton Vorontsov
email: cbouatmailru@gmail.com
irc://irc.freenode.net/bd2