RE: [PATCH net-next 3/3] tcp: derive delack_max from rto_min

[David Laight <David.Laight@ACULAB.COM>]

From: Eric Dumazet
> Sent: 22 September 2023 11:53
> 
> On Fri, Sep 22, 2023 at 11:59 AM David Laight <David.Laight@aculab.com> wrote:
> >
> > From: Eric Dumazet
> > > Sent: 21 September 2023 13:58
> > >
> > > On Thu, Sep 21, 2023 at 2:37 PM David Ahern <dsahern@kernel.org> wrote:
> > > >
> > >
> > > > My comment is solely about mismatch on data types. I am surprised use of
> > > > max_t with mixed data types does not throw a compiler warning.
> > >
> > > This was intentional.
> > >
> > > This is max_t() purpose really.
> >
> > Apart from when it gets used to accidentally mask high bits :-)
> > (Although hat is usually consigned to min_t()).
> 
> As explained, this is not an accident, but a conscious decision I made.
> 
> >
> > Here
> >         u32 delack_from_rto_min = max(rto_min, 2u) - 1;
> > would probably be safer (as in have no casts that might have
> > unwanted side effects).
> >
> 
> I find my solution more readable.

It has to be said I didn't really like mine either :-)
A better alternative would be:
	max((int)rto_min - 1, 1)
to make it absolutely clear what is going on.

Far too many of the min_t() and max_t() are just used to silence
the over-enthusiastic type checking of min() and max().
So code doing something different might be best making it more obvious.

Does 'ip route' stop very large values for rto_min?
Unsigned values with the top bit set might cause 'interesting' behaviour!

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)