Re: [PATCH v8 5/6] serial: Support for RS-485 multipoint addresses

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Mon, Jun 20, 2022 at 02:26:17PM +0300, Ilpo Järvinen wrote:
> On Mon, 20 Jun 2022, Andy Shevchenko wrote:
> > On Mon, Jun 20, 2022 at 09:40:29AM +0300, Ilpo Järvinen wrote:

...

> > > The changes to serial_rs485 struct were test built with a few traps to
> > > detect mislayouting on archs lkp/0day builts for (all went fine):
> > >   BUILD_BUG_ON(((&rs485.delay_rts_after_send) + 1) != &rs485.padding[0]);
> > >   BUILD_BUG_ON(&rs485.padding[1] != &rs485.padding1[0]);
> > >   BUILD_BUG_ON(sizeof(rs485) != ((u8 *)(&rs485.padding[4]) -
> > > 				 ((u8 *)&rs485.flags) + sizeof(__u32)));
> > 
> > You may add static_asserts() for the above mentioned cases.
> 
> I'll add into the end of serial_core.h but in a cleaned up form
> using offsetof(). Those above look rather ugly :-).

Agree!

...

> > > -	__u32	padding[5];		/* Memory is cheap, new structs
> > > -					   are a royal PITA .. */
> > > +	union {
> > > +		/* v1 */
> > > +		__u32	padding[5];		/* Memory is cheap, new structs are a pain */
> > > +
> > > +		/* v2 (adds addressing mode fields) */
> > 
> > How user space will inform a kernel that it's trying v2?
> >
> > Usually when we have a union, it should be accompanied with the enum or version
> > or something to tell which part of it is in use. I can imagine that in this case
> > it's implied by the IOCTL parameters that never should be used on a garbage.
> > 
> > Either add a commit message / UAPI comment or add a version field or ...?
> > 
> > > +		struct {
> > > +			__u8	addr_recv;
> > > +			__u8	addr_dest;
> 
> The flags in .flags indicate when these two new fields are in use. Do you 
> think I need something beyond that. Maybe I should remove those comments 
> so they don't mislead you to think it's a "version" for real?

Yes, either drop this versioning, or replace with a comment on top of a union
like:

	/* The fields are defined by flags */

-- 
With Best Regards,
Andy Shevchenko