On Mon, May 31, 2021 at 12:58:29PM +0300, Andy Shevchenko wrote:

On Mon, May 31, 2021 at 10:47:20AM +0200, Marc Kleine-Budde wrote:

quoted

On 26.05.2021 22:33:26, Andy Shevchenko wrote:

quoted

In some configurations, mainly ACPI-based, the clock frequency of the device
is supplied by very well established 'clock-frequency' property. Hence, try
to get it from the property at last if no other providers are available.

quoted

quoted

 		return dev_err_probe(&spi->dev, PTR_ERR(reg_xceiver),
 				     "Failed to get Transceiver regulator!\n");
 
-	clk = devm_clk_get(&spi->dev, NULL);
+	/* Always ask for fixed clock rate from a property. */
+	device_property_read_u32(&spi->dev, "clock-frequency", &rate);

what about error handling....?

Not needed, but rate should be assigned to 0, which is missed.

quoted

quoted

+	clk = devm_clk_get_optional(&spi->dev, NULL);
 	if (IS_ERR(clk))
 		return dev_err_probe(&spi->dev, PTR_ERR(clk),
 				     "Failed to get Oscillator (clock)!\n");
 	freq = clk_get_rate(clk);
+	if (freq == 0)
+		freq = rate;

... this means we don't fail if there is neither a clk nor a
clock-frequency property.

The following will check for it (which is already in the code)

  if (freq <= MCP251XFD_SYSCLOCK_HZ_MAX / MCP251XFD_OSC_PLL_MULTIPLIER) {

Even before, actually,

             /* Sanity check */
             if (freq < MCP251XFD_SYSCLOCK_HZ_MIN ||
		 freq > MCP251XFD_SYSCLOCK_HZ_MAX) {

quoted

I've send a v3 to fix this.

You mean I have to send v3?
Sure!

So, I am going to send a v3 with amended commit message and assigning rate
to 0.

-- 
With Best Regards,
Andy Shevchenko