Re: [PATCH] usb: cdnsp: Useless condition has been removed

[Dan Carpenter <dan.carpenter@oracle.com>]

On Fri, Apr 30, 2021 at 03:11:45PM +0800, Peter Chen wrote:
> On 21-04-30 07:52:48, Dan Carpenter wrote:
> > On Fri, Apr 30, 2021 at 11:43:49AM +0800, Peter Chen wrote:
> > > On 21-04-29 10:49:14, Pawel Laszczak wrote:
> > > > From: Pawel Laszczak <pawell@cadence.com>
> > > > 
> > > > Patch removes the warning "variable dereferenced before
> > > > check 'pdev->dcbaa'" from cdnsp_mem_cleanup.
> > > 
> > > You may describe the real problem you fix, but not the warning
> > > message from some auto build system.
> > > 
> > 
> > I really feel people have become too lost in the weeds of what matters
> > and what does not.  For internships, people want picky feedback in case
> > they're forced to deal with the ultra drill sargent maintainers.  But
> > really the important thing in a commit message is if you can understand
> > the problem and the fix.  In this case everyone who is capable of
> > understanding the patch can understand the commit message.
> > 
> > Also if you're going to criticize someone's commit message then just
> > write it the way you want so they can copy and paste.  I had someone
> > yesterday say that my commit message where I deleted a NULL check and
> > related dead code was not clear enough that the behavior was
> > "intentional" and I took that to mean that they wanted me to say that
> > deleting the dead code did not change runtime behavior.  I'm still not
> > sure that's what they wanted me to add...  The point is no one can read
> > your mind, if you want a commit message to say something specific then
> > just say it instead of hinting around the bush and we have to send a v2
> > and v3 commit message.
> > 
> > regards,
> > dan carpenter
> 
> Hi Dan,
> 
> Thanks for your comments, it's the good suggestion. But everyone may have
> slight different criteria that what's really clear.
> 
> For this patch, I don't understand the commits, the pdev->dcbaa is set as
> NULL after dma_free_coherent, why it has this warning.
> 

Hm...  I guess it's a fair point that it's not clear to the reviewer
where the unchecked dereference happens.  Here is how I would write that
commit.

    This code generates a Smatch warning:

    drivers/usb/cdns3/cdnsp-mem.c:1085 cdnsp_mem_cleanup()
    warn: variable dereferenced before check 'pdev->dcbaa' (see line 1067)

    The unchecked dereference happens inside the function when we call:

    	cdnsp_free_priv_device(pdev);

    But fortunately, the "pdev->dcbaa" pointer can never be NULL so it
    does not lead to a runtime issue.  We can just remove the NULL check
    which silences the warning and makes the code consistent.

regards,
dan carpenter