Re: [PATCH v2 4/7] tty: serial: support device tree in pxa

[Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>]

On Fri, Jul 29, 2011 at 05:49:22PM +0100, Russell King - ARM Linux wrote:
> On Fri, Jul 29, 2011 at 10:45:39AM -0600, Grant Likely wrote:
> > On Thu, Jul 28, 2011 at 02:41:30PM +0800, Haojian Zhuang wrote:
> > > +#ifdef CONFIG_OF
> > > +	for (i = 0; i < PXA_SERIAL_NR; i++) {
> > > +		if (serial_pxa_ports[i] == NULL)
> > > +			break;
> > > +	}
> > > +	if (i >= PXA_SERIAL_NR) {
> > > +		pr_warn("can't find pxa serial port\n");
> > > +		return -ENODEV;
> > > +	}
> > > +
> > > +	if (of_property_read_u32(np, "clock-frequency", &clk)) {
> > > +		pr_warn("no clock-frequency property set\n");
> > > +		return -ENODEV;
> > > +	}
> > > +	if (of_property_read_u32(np, "current-speed", &spd) == 0)
> > > +		sport->port.custom_divisor = clk / (16 * spd);
> > > +
> > > +	sprintf(name, "pxa2xx-uart.%d", i);
> > > +	sport->clk = clk_get_sys(name, NULL);
> > > +	if (IS_ERR(sport->clk)) {
> > > +		ret = PTR_ERR(sport->clk);
> > > +		goto err_free;
> > > +	}
> > > +	sport->port.uartclk = clk;
> > > +#else
> > > +	i = dev->id;
> > >  	sport->clk = clk_get(&dev->dev, NULL);
> > >  	if (IS_ERR(sport->clk)) {
> > >  		ret = PTR_ERR(sport->clk);
> > >  		goto err_free;
> > >  	}
> > > +	sport->port.uartclk = clk_get_rate(sport->clk);
> > > +#endif
> > 
> > This means a kernel build can either support DT or non-DT, but not
> > both.  DT & non-DT booting are full supported with the same kernel
> > image, so don't do it this way.
> > 
> > Instead, check for the presence of an of_node.  If it is there, do the
> > DT parsing.  If now, still support the old method.
> 
> Oh ffs, why is this DT stuff causing all the clk stuff to have to change.
> And specifically why is stuff converting to use clk_get_sys().
> 
> clk_get_sys() is there to allow system devices to get their clocks.
> It's not for general drivers to use.
> 
> Please, stop this madness.

indeed.

g.