On Tue, 31 Dec 2013, Ben Greear wrote:
> On 12/31/2013 08:09 AM, Julia Lawall wrote:
> >
> >
> > On Tue, 31 Dec 2013, Ben Greear wrote:
> >
> > > On 12/30/2013 10:32 PM, Julia Lawall wrote:
> > > > > > > > I'm just thinking of a programmer, e.g. changing a struct like
> > > > > > > > this:
> > > > > > > >
> > > > > > > > struct foo {
> > > > > > > > u8 addr[ETH_ALEN];
> > > > > > > > - u16 dummy;
> > > > > > > > };
> > > > >
> > > > > I don't know of a way to catch that.
> > > > > Anyone else?
> > > >
> > > > Well, one could have a semantic patch that checks for that. But the
> > > > problem is that it is very slow, and it only covers the cases that I can
> > > > transform automatically, which currently means no pointers, only
> > > > explicit
> > > > arrays.
> > > >
> > > > On the other hand, I am finding the structure definition, so I can
> > > > easily
> > > > update the structure definition with an appropriate comment.
> > > >
> > > > struct foo {
> > > > u8 addr[ETH_ALEN]; /* must be followed by two bytes in the
> > > > structure */
> > > > u16 dummy;
> > > > };
> > > >
> > > > Unfortunately it is kind of verbose. Could there be an attribute? That
> > > > could even easily be checked.
> > >
> > > Can you not just add a build-time macro to check that sizeof(foo) >= 8
> > > for each of these struct foos? Or, is it required that the dummy field
> > > be there and be not used by anything else?
> >
> > It doesn't matter what the field is used for. The problem is that is it
> > necessary to ensure a property of the position of addr within the
> > structure. It has to have at least 16 bytes after it.
>
> You mean 16 bits?
Oops, yes. 16 bits.
> >
> > But maybe something with sizeof(foo) and offset_of would do?
> >
> > Could the macro be put near the declaration of the structure somehow?
>
> I think that would work, but do not know all of the details of such
> macros, so it's possible there is some catch.
>
> If nothing else, then some run-time code that calculates the offset off
> and asserts if it is broken in module initialization or similar might
> be good enough.
Could be OK. Something right in or after the structure declaration would
be nicest.
julia
>
> Thanks,
> Ben
>
> --
> Ben Greear <greearb@xxxxxxxxxxxxxxx>
> Candela Technologies Inc http://www.candelatech.com
>
> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
