Andy Shevchenko writes:

On Mon, Nov 9, 2020 at 5:27 PM Alexandre Belloni
[off-list ref] wrote:

quoted

On 09/11/2020 17:16:49+0200, Andy Shevchenko wrote:

quoted

On Mon, Nov 9, 2020 at 4:32 PM Alexandre Belloni
[off-list ref] wrote:

quoted

On 09/11/2020 16:17:40+0200, Andy Shevchenko wrote:

...

quoted

quoted

quoted

quoted

quoted

+               dev_err(pctldev->dev, "Pin %d direction as %s is not possible\n",
+                       pin, input ? "input" : "output");

Do we need this noise? Isn't user space getting a proper error code as
per doc and can handle this?

Why would userspace get the error code?

Huh?! Why it shouldn't. How will users know if they are doing something wrong?

quoted

Userspace should never have to
handle gpios directly or you are doing something wrong.

This is true, but check how error codes are propagated to the user space.

your point is to remove an error message because the error may be
propagated to userspace. My point is that userspace should never use
gpios and the kernel has to be the consumer.

Tell this to plenty of users of old sysfs interface and to libgpiod ones.
If what you are saying had been true, we would have never had the new
ABI for GPIOs.

quoted

I don't see how your answer
is relevant here.

I have an opposite opinion.

quoted

Did you already check all the call sites from the
kernel too?

If you think we have to print a message on each possible error case
(but not always the one) we will get lost in the messages disaster and
dmesg overflow.
It is consumer who should decide if the setting is critical or not to
be printed to user.

I think the message is a valid one. I will change it to
dev_err_ratelimited() - that should prevent the dmesg flooding.

---Lars

--
Lars Povlsen,
Microchip