Re: [PATCH 1/6] serial: 8250_dw: add support for clk api

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Fri, Mar 15, 2013 at 09:15:11PM -0300, Emilio López wrote:
> Hello Russell,
> 
> El 15/03/13 19:39, Russell King - ARM Linux escribió:
> > On Fri, Mar 15, 2013 at 09:06:23PM +0100, Maxime Ripard wrote:
> >> +	/* clock got configured through clk api, all done */
> >> +	if (p->uartclk)
> > 
> > 	if (IS_ERR(p->uartclk))
> > 
> 
> Isn't IS_ERR for pointers? p->uartclk is an unsigned int

Right, sorry, ignore that.

> >> +		return 0;
> >> +
> >> +	/* try to find out clock frequency from DT as fallback */
> >>  	if (of_property_read_u32(np, "clock-frequency", &val)) {
> >> -		dev_err(p->dev, "no clock-frequency property set\n");
> >> +		dev_err(p->dev, "clk or clock-frequency not defined\n");
> >>  		return -EINVAL;
> >>  	}
> >>  	p->uartclk = val;
> >> @@ -294,9 +301,21 @@ static int dw8250_probe(struct platform_device *pdev)
> >>  	if (!uart.port.membase)
> >>  		return -ENOMEM;
> >>  
> >> +	data = devm_kzalloc(&pdev->dev, sizeof(*data), GFP_KERNEL);
> >> +	if (!data)
> >> +		return -ENOMEM;
> >> +
> >> +	data->clk = devm_clk_get(&pdev->dev, NULL);
> >> +	if (IS_ERR(data->clk))
> >> +		data->clk = NULL;
> >> +	else
> >> +		clk_prepare_enable(data->clk);
> > 
> > 	if (!IS_ERR(data->clk))
> > 		clk_prepare_enable(data->clk);
> > 
> 
> See below
> 
> >> +
> >>  	uart.port.iotype = UPIO_MEM;
> >>  	uart.port.serial_in = dw8250_serial_in;
> >>  	uart.port.serial_out = dw8250_serial_out;
> >> +	uart.port.private_data = data;
> >> +	uart.port.uartclk = clk_get_rate(data->clk);
> > 
> > What if data->clk is invalid?
> > 
> > 	if (!IS_ERR(data->clk)
> > 		uart.port.uartclk = clk_get_rate(data->clk);
> > 
> 
> I'm not sure if it is coincidental or the way it is supposed to be, but
> when using the common clock framework, if you pass a NULL to
> clk_get_rate, the function explicitly checks for it and returns 0. I
> relied on that behaviour when implementing this; see the if..else block
> above. Is this not always the case on other clock drivers?

That's something that the common clock framework decided to do.  It's
not a defined part of the API though, so drivers shouldn't rely on
this behaviour meaning anything special.