[PATCH 2/2] ARM: mach-imx: CAN clock fix for i.MX53

[stigge@antcom.de (Roland Stigge)]

On 10/26/2012 01:05 PM, Roland Stigge wrote:
> On 10/26/2012 12:30 PM, Marc Kleine-Budde wrote:
>>>>> +	clk[can1_serial_gate] = imx_clk_gate2("can1_serial_gate",
>>>>> "can_sel", MXC_CCM_CCGR4, 6); +	clk[can1_ipg_gate] =
>>>>> imx_clk_gate2("can1_ipg_gate", "ipg", MXC_CCM_CCGR4, 8); +
>>>>> clk[can2_serial_gate] = imx_clk_gate2("can2_serial_gate",
>>>>> "can_sel", MXC_CCM_CCGR4, 6); +	clk[can2_ipg_gate] =
>>>>> imx_clk_gate2("can2_ipg_gate", "ipg", MXC_CCM_CCGR4, 8);
>>>>
>>>> This doesn't.
>>>
>>> This may be right, but unfortunately, since the introduction of
>>> your can1 + can2 clocking change, the first block stopped working
>>> for me.
>>
>> You are effectively using can2's clock for can1. Are you sure you 
>> haven't mixed up can1 and can2?
> 
> Just using the above patch and patch 1/2 from this series (missing can
> pinmuxing), doing like this in custom .dts:
> 
> +                       can1: can at 53fc8000 {
> +                               pinctrl-names = "default";
> +                               pinctrl-0 = <&pinctrl_can1_1>;
> +                               clock-frequency = <66500000>;

That was the problem, since in flexcan.c probe():


	if (pdev->dev.of_node)
		of_property_read_u32(pdev->dev.of_node,
						"clock-frequency", &clock_freq);

	if (!clock_freq) {
		clk_ipg = devm_clk_get(&pdev->dev, "ipg");
		if (IS_ERR(clk_ipg)) {
			dev_err(&pdev->dev, "no ipg clock defined\n");
			err = PTR_ERR(clk_ipg);
			goto failed_clock;
		}
		clock_freq = clk_get_rate(clk_ipg);

		clk_per = devm_clk_get(&pdev->dev, "per");
		if (IS_ERR(clk_per)) {
			dev_err(&pdev->dev, "no per clock defined\n");
			err = PTR_ERR(clk_per);
			goto failed_clock;
		}
	}

Sorry for the noise.

The other patches are still current.

Thanks,

Roland