Re: [PATCH] drivers: net: ethernet: intel: e1000: e1000_ethtool.c coding style fixes

[Joe Perches <joe@perches.com>]

On Mon, 2014-08-18 at 08:45 -0700, Alexander Duyck wrote:
> On 08/18/2014 08:31 AM, Joe Perches wrote:
> > On Mon, 2014-08-18 at 08:29 -0700, Alexander Duyck wrote:
> >> Doing any kind of pointer math on a void pointer is generally unsafe as
> >> it is an incomplete type.  The only reason why it works in GCC is
> >> because GCC has a nonstandard extension that makes it report as having a
> >> size of 1.
> > 
> > I know.  It's used in quite a few places in kernel code
> > so I believe it's now a base assumption for the kernel.
> 
> Well that is something that should probably be fixed then.  I don't
> believe it is safe to be doing any kind of pointer math on a void pointer.
> 
> We really shouldn't be using any GCC specific bits unless we absolutely
> have to.

Good luck with that.

Here's a little coccinelle script to find them:

$ cat void_arithmetic.spatch
@@
void *p;
expression e;
@@

*	p + e
$

There are at least 23 uses just in lib/

$ spatch -sp_file void_arithmetic.spatch lib | \
  grep "^\-[^\-]" | wc -l
23

I'm fairly confident there are more than that,
but using spatch with --recursive-includes is
pretty slow on my machine.