[PATCH 14/14] ARM: kirkwood: use devicetree to init sata_mv

[Jason]

On Wed, Mar 07, 2012 at 06:40:09PM +0000, Arnd Bergmann wrote:
> On Wednesday 07 March 2012, Jason Cooper wrote:
> > @@ -504,6 +504,16 @@ static int __init kirkwood_clock_gate(void)
> >                 kirkwood_clk_ctrl |= CGC_USB0;
> >                 of_node_put(np);
> >         }
> > +
> > +       np = of_find_compatible_node(NULL, NULL, "mrvl,orion-sata");
> > +       if (np && of_device_is_available(np)) {
> > +               int nr_ports;
> > +               kirkwood_clk_ctrl |= CGC_SATA0;
> > +               of_property_read_u32(np, "nr-ports", &nr_ports);
> > +               if (nr_ports > 1)
> > +                       kirkwood_clk_ctrl |= CGC_SATA1;
> > +               of_node_put(np);
> > +       }
> >  #endif
> >  
> >         /* For SATA: first shutdown the phy */
> 
> Is it guaranteed that it's always the first port that gets used if there
> is only one? If not, this might turn on the wrong one.

This is the same logic as mach-kirkwood/common.c:176

void __init kirkwood_sata_init(struct mv_sata_platform_data *sata_data)
{
        kirkwood_clk_ctrl |= CGC_SATA0;
        if (sata_data->n_ports > 1)
                kirkwood_clk_ctrl |= CGC_SATA1;

        orion_sata_init(sata_data, SATA_PHYS_BASE, IRQ_KIRKWOOD_SATA);
}

So, I don't think I'm introducing a regression.

thx,

Jason.