Re: [bug report] ovl: make sure that real fid is 32bit aligned in memory

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

On Tue, May 05, 2020 at 07:13:20PM +0300, Amir Goldstein wrote:
> > The patch cbe7fba8edfc: "ovl: make sure that real fid is 32bit
> > aligned in memory" from Nov 15, 2019, leads to the following static
> > checker warning:
> >
> >         fs/overlayfs/export.c:791 ovl_fid_to_fh()
> >         warn: check that subtract can't underflow
> >
> > fs/overlayfs/export.c
> >    775  static struct ovl_fh *ovl_fid_to_fh(struct fid *fid, int buflen, int fh_type)
> >    776  {
> >    777          struct ovl_fh *fh;
> >    778
> >    779          /* If on-wire inner fid is aligned - nothing to do */
> >    780          if (fh_type == OVL_FILEID_V1)
> >    781                  return (struct ovl_fh *)fid;
> >    782
> >    783          if (fh_type != OVL_FILEID_V0)
> >    784                  return ERR_PTR(-EINVAL);
> >    785
> >    786          fh = kzalloc(buflen, GFP_KERNEL);
> 
> Doesn't Smatch warn on possible kmalloc(0)?
> That can't be good. Right?

No, no.  Allocating zero bytes is a useful feature.

	size = 0;
	buf = kzalloc(size, GFP_KERNEL);

	for (i = 0; i < size; i++)
		buf[i] = 42;
	memcpy(dest, buf, size);
	copy_to_user(p, dest, size);

That all works.  Neat, huh?

regards,
dan carpenter