Re: [PATCH] thunderbolt: handle possible NULL pointer from get_device()

[Greg KH <gregkh@linuxfoundation.org>]

On Fri, Jun 09, 2023 at 11:05:59AM +0300, Dmitry Antipov wrote:
> On 6/9/23 10:46, Greg KH wrote:
> 
> > Again, how did you test this?
> 
> Did you look at the patch header? For that particular case, the static
> analysis tool complains that the value returned by get_device() is most
> likely should be checked just because it is checked on a lot of other
> code paths. Usually it may be a good precaution to handle the very rare
> and unexpected errors; again, if you're sure that this is not the case,
> just disregard it.

Just because a static tool said "this might be wrong" does not mean you
do not need to actually test your change or do some work to verify that
it is a sane change at all.

So far I have seen more and more false-positives from this "tool" of
your group that I am very inclined to just tell all kernel maintainers
to ignore them for a very long time as you are not following the
documented rules for such patches as outlined in
Documentation/process/researcher-guidelines.rst

Please read that and fix your tool, and your submission process, I've
said this many times already.

greg k-h