On Mon, 2009-06-01 at 01:57 +0900, Atsushi Nemoto wrote:
> : This is a reply for comments from Matt Mackall via mm-commit mail.
>
> On Fri, 29 May 2009 19:21:22 -0500, Matt Mackall <mpm@xxxxxxxxxxx> wrote:
> > > +++ a/arch/mips/txx9/generic/setup_tx4939.c
> > > @@ -494,6 +494,24 @@ void __init tx4939_aclc_init(void)
> > > TXX9_IRQ_BASE + TX4939_IR_ACLC, 1, 0, 1);
> > > }
> > >
> > > +#define TX4939_RNG_REG ((TX4939_CRYPTO_REG & 0xfffffffffULL) + 0xb0)
> >
> > This isn't the best place for this define?
>
> Well, I can move it to include/asm/txx9/tx4939.h.
>
> > > +++ a/arch/mips/txx9/rbtx4939/setup.c
> > > @@ -501,6 +501,7 @@ static void __init rbtx4939_device_init(
> > > tx4939_dmac_init(0, 2);
> > > tx4939_aclc_init();
> > > platform_device_register_simple("txx9aclc-generic", -1, NULL, 0);
> > > + tx4939_rng_init();
> > > }
> > >
> > > static void __init rbtx4939_setup(void)
> >
> > Not clear to me how this works when this is a module?
>
> This patch add a registration of a platform device for RNG to vmlinux.
> And the other patch add a driver module for the RNG. This strategy is
> fairly common for SoCs or embedded platforms.
If your driver is built as a module (which your patch allows), the above
won't work, right?
--
http://selenic.com : development and support for Mercurial and Linux
