Re: [PATCH 3/3] serial: Add kserial_rs485 to avoid wasted space due to .padding

["Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1644 bytes --]

On Fri, 26 Aug 2022, Andy Shevchenko wrote:

> On Fri, Aug 26, 2022 at 5:51 PM Ilpo Järvinen
> <ilpo.jarvinen@linux.intel.com> wrote:
> 
> > -static int user_rs485_to_kernel_serial_rs485(struct serial_rs485 *rs485,
> > +static int user_rs485_to_kernel_serial_rs485(struct kserial_rs485 *rs485,
> >                                              const struct serial_rs485 __user *rs485_user)
> >  {
> > -       if (copy_from_user(rs485, rs485_user, sizeof(*rs485)))
> > +       struct serial_rs485 rs485_uapi;
> > +
> > +       if (copy_from_user(&rs485_uapi, rs485_user, sizeof(*rs485)))
> >                 return -EFAULT;
> 
> > +       *rs485 = *((struct kserial_rs485 *)&rs485_uapi);
> 
> So with all assets we have we can be sure that on BE64 / BE32 machines
> this will be flawless. Is this assumption correct?

I think so. At worst it could trigger a build fail assuming the kernel 
would do some really odd struct layout reordering (which, according to 
the build bot, doesn't occur for any currently tested arch).

Now that you highlighted this line though, I started to wonder if it would 
be just better to use memcpy() instead as it would avoid those casts.

> > + * padding.
> > + */
> > +struct kserial_rs485 {
> > +       __u32   flags;                  /* RS485 feature flags */
> > +       __u32   delay_rts_before_send;  /* Delay before send (milliseconds) */
> > +       __u32   delay_rts_after_send;   /* Delay after send (milliseconds) */
> > +       struct {
> > +               __u8    addr_recv;
> > +               __u8    addr_dest;
> > +       };
> 
> Btw, can't we convert them to kernel doc?

Yes, why not.

-- 
 i.