On Wed, Mar 06, 2024 at 08:08:36AM +0100, Stefano Brivio wrote:
> instead of htons_constant(), which is for... constants.
> 
> Fixes: 5bf200ae8a1a ("tcp, udp: Don't include destination address in partially precomputed csums")
> Signed-off-by: Stefano Brivio <sbrivio@redhat.com>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

It seems to get the job done, at what's probably negligible practical
cost.  My perfectionist side has some misgivings:

AIUI, the point of htons_constant() isn't so much that it *requires* a
constant, but that because it's open-coded in plain arithmetic
operations the compiler will be able to constant fold it, if it is
invoked with a constant parameter.  Since htons() is a library
function, it probably can't be elided in that way.  The cost of
htons_constant() is that it may be a less optimal implementation when
the calculation really does need to be done at runtime.

I'm still a bit baffled at that Coverity warning: I can't see why it
doesn't preclude any situation where you want to write out
calculations for clarity, even though you know the result will be a
constant (and you expect the compiler to fold it).

> ---
>  ip.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/ip.h b/ip.h
> index 9be4778..b9aedf6 100644
> --- a/ip.h
> +++ b/ip.h
> @@ -38,7 +38,7 @@
>  		.daddr		= 0,					\
>  	}
>  #define L2_BUF_IP4_PSUM(proto)	((uint32_t)htons_constant(0x4500) +	\
> -				 (uint32_t)htons_constant(0xff00 | (proto)))
> +				 (uint32_t)htons(0xff00 | (proto)))
>  
>  #define L2_BUF_IP6_INIT(proto)						\
>  	{								\

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson