Re: [PATCH v2 2/2] add check for pointers with __free attribute initialized to NULL

[Joe Perches <joe@perches.com>]

On Fri, 2025-10-24 at 11:14 -0700, dan.j.williams@intel.com wrote:
> Joe Perches wrote:
> > On Fri, 2025-10-24 at 22:59 +0530, Ally Heev wrote:
> > > pointers with __free attribute initialized to NULL
> > > pose potential cleanup issues [1] when a function uses
> > > interdependent variables with cleanup attributes
> > > 
> > > Link: https://docs.kernel.org/core-api/cleanup.html [1]
> > > Link: https://lore.kernel.org/all/68f7b830ec21a_10e910070@dwillia2-mobl4.notmuch/
> > > Suggested-by: Dan Williams <dan.j.williams@intel.com>
> > > Signed-off-by: Ally Heev <allyheev@gmail.com>
> > []
> > > diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
> > []
> > > @@ -7728,6 +7728,12 @@ sub process {
> > >  			ERROR("UNINITIALIZED_PTR_WITH_FREE",
> > >  			      "pointer '$1' with __free attribute should be initialized\n" . $herecurr);
> > >  		}
> > > +
> > > +# check for pointers with __free attribute initialized to NULL
> > > +		while ($line =~ /\*\s*($Ident)\s+$FreeAttribute\s*=\s*NULL\b/g) {
> > > +			WARN("NULL_INITIALIZED_PTR_WITH_FREE",
> > > +			      "pointer '$1' with __free attribute should be initialized to a non-NULL address\n" . $herecurr);
> > > +		}
> > >  	}
> > 
> > I think this a poor idea as almost all the instances where this
> > initialization is done are fine.
> > 
> > And there are a lot of them.
> > 
> > $ git grep -P '\b__free\b.*=\s*NULL\s*;' | wc -l
> > 490
> 
> That is significant. ...but you did say "almost" above. What about
> moving this from WARN level to CHK level?

I have no idea how many instances in the tree are inappropriate.
Do you? I believe it to be a difficult analysis problem.

But given the number is likely to be extremely low, I think it should
not be added to checkpatch even as a CHK.

If you can show that the reporting rate of defects is significant,
say >10%, then OK, but I rather doubt it's that high.