[Cbe-oss-dev] [patch 08/16] powerpc and x86_64: Initial port of PCIe endpoint device driver for Axon

[Arnd Bergmann]

On Thursday 13 March 2008, Murali Iyer wrote:
> On Thu, 2008-03-13 at 05:55 +0100, Arnd Bergmann wrote:

> > Do I understand this correctly that this is unidirectional? What do you use
> > for the opposite way?
> Yep. this is from host to cell. For opposite the UTL stuff to generate
> PCIe interrupt.. we are using shared one and need to start using MSI
> but having lack of some support host systems...
> and, this driver is extracting mails what host has sent _not_ the
> sending part itself.

Ok, is this abstracted in a higher-level interface so that the user
level driver and the network driver don't need to care whether they
use MSI or mailbox? If not, would it be reasonable to change?

> > > +#include <linux/platform_device.h>
> > > +
> > > +/*  function prototypes */
> > > +int mbx_init(void);	/* mailbox driver initialization */
> > > +void mbx_exit(void);	/* mailbox driver exit */
> > > +
> > 
> > Nor these prototypes. Just make the mailbox driver a standalone module
> > and mark these functions module_init/module_exit.
> yep. but sync between probes needed prior to that.

Right. The way to achieve that is to keep the dependencies between the
drivers as a directed acyclic graph, which they most likely are already.
If you use loadable modules, this will automatically result in the right
order.

When you have all drivers builtin, you can rely on link order (from Makefile),
or use explicit initcall levels (subsys_initcall, device_initcall,
late_initcall, ...).

	Arnd <><