Hello Sam,
On Mon, Jul 03, 2017 at 10:17:46PM +0200, Sam Ravnborg wrote:
> If there is no OFTREE support of_clk_get_by_name failed with
> -ENOENT, which caused clk_get to bail out.
> This had the effect that nothing was printed on the serial console
> with at91sam9263-ek.
>
> Fix this by modifying the logic to explicitly check for -EPROBE_DEFER
> like we do in the kernel.
>
> Fixes: 90f7eacb ("clk: let clk_get return errors from of_clk_get_by_name")
> Cc: Uwe Kleine-König <u.kleine-koenig@xxxxxxxxxxxxxx>
> Signed-off-by: Sam Ravnborg <sam@xxxxxxxxxxxx>
> ---
> drivers/clk/clkdev.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/clk/clkdev.c b/drivers/clk/clkdev.c
> index 6b1666355..939b067cc 100644
> --- a/drivers/clk/clkdev.c
> +++ b/drivers/clk/clkdev.c
> @@ -181,7 +181,7 @@ struct clk *clk_get(struct device_d *dev, const char *con_id)
>
> if (dev) {
> clk = of_clk_get_by_name(dev->device_node, con_id);
> - if (!IS_ERR(clk) || PTR_ERR(clk) != -ENODEV)
> + if (!IS_ERR(clk) || PTR_ERR(clk) == -EPROBE_DEFER)
Hmm, I don't remember which was the failure I saw that lead me to create
this patch. Probably I wanted to see EPROBE_DEFER.
Looking into the tree of functions that can be called from
of_clk_get_by_name I didn't find a function that returns ENODEV.
But consider a clock provider that tries to give you the clock and
triggers a EIO or ETIMEOUT. IMHO this should be given to the caller
instead of continuing with clk_get_sys. So I suggest
- if (!IS_ERR(clk) || PTR_ERR(clk) != -ENODEV)
+ if (!IS_ERR(clk) || PTR_ERR(clk) != -ENOENT)
instead of your patch.
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | http://www.pengutronix.de/ |
_______________________________________________
barebox mailing list
barebox@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/barebox
