On Tue, Oct 30, 2012 at 09:24:42AM -0700, Tony Lindgren wrote:

* Omar Ramirez Luna [off-list ref] [121030 05:20]:

quoted

Tony,

On 29 October 2012 12:52, Tony Lindgren [off-list ref] wrote:

quoted

quoted

--- /dev/null
+++ b/include/linux/platform_data/omap_mailbox.h

@@ -0,0 +1,105 @@

This file should only contain pure platform data needed
by the core omap code to pass to the mailbox driver.

Ok, looking at it closely, this header file is related to the API
itself, there is nothing that could be actually considered as pure
platform data, the structures are related with the mailbox framework
and even if I split this file into two, the additional header would
end up including the "platform_data" header unless I move
save/restore_ctx functions and then export them as symbols for the
API.

So, it might be better for the entire file to sit in
linux/include/mailbox/ then.

OK to me.
 

quoted

quoted

The mailbox API header should be somewhere else,
like include/linux/mailbox/mailbox-omap.h or similar.

Ok.

quoted

But shouldn't this all now be handled by using the
remoteproc framework?

Remoteproc doesn't handle the mailbox hardware directly, it still
relies in the mailbox framework for the low level communications.
E.g.: Proc1 has a message (virtqueue msg) queued to Proc2, uses
mailbox msg to generate an interrupt to Proc2, Proc2 queries the
message (virtqueue) based on the mailbox message received.

OK.

Greg, do these patches look OK to you to move to live under
drivers/mailbox?

Um, I don't know, I wasn't paying attention here, sorry.

greg k-h