Re: Chipone icn85xx support in x86 linux kernel

[Gregor Riepl <onitake@gmail.com>]

> At last I am releasing my icn85xx (tested on icn8528 on Chuwi Vi10,
> Baytrail) kernel space driver.
> It is unfortunately polling mode, but at the moment, I am happy with
> it and have no enough time to dig in with irq mode.
> 
> https://gitlab.com/SergK/icn85xx/tree/master

Good work!

A few comments - but bear in mind that I'm not an experienced kernel
developer. Someone else may be able to give better advice.

First: Patches to the kernel should be submitted according to these guidelines
at https://www.kernel.org/doc/Documentation/SubmittingPatches
In particular, a new driver should be a patch against the respective kernel
branch, which means it should live in the appropriate directory in the kernel
tree and include an addendum to the Kconfig/Makefile there. Standalone
Makefiles are nice when building drivers out-of-tree, but they are not
suitable for kernel code. This doesn't apply to your development repo, of
course. Testing may be easier when a driver can be built standalone.

Second: The patch should be formatted according to the Linux kernel coding
style at https://www.kernel.org/doc/Documentation/CodingStyle
There's also the checkpatch.pl utility in the kernel tree that can help you
find and fix mistakes.

But don't be discouraged by the lenghty guides - it's not that hard and
checkpatch helps a lot.

One thing I noticed on a quick glance is that line 833 in your code is
redundant. kzalloc aready clears the memory (hence, k_z_alloc).

Also: I highly recommend using devm_ variants of constructors, where
available. If you use those, you can reduce the footprint of your _remove()
function or even remove it completely. devm_input_allocate_device() is a good
example. You should use it instead of input_allocate_device(), and it will
take care of unregistering and deallocation automatically.