RE: [PATCH net-next 0/8] Use max/min to simplify the code

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

From: Jakub Kicinski
> Sent: 27 August 2024 15:04
> 
> On Tue, 27 Aug 2024 07:45:02 +0300 Kalle Valo wrote:
> > > Do you mean some patches will go to other branches (such as mac80211)?
> >
> > Jakub means that your patchset had compilation errors, see the red on
> > patchwork:
> >
> > https://patchwork.kernel.org/project/netdevbpf/list/?series=882901&state=*&order=date
> 
> FWIW I prefer not to point noobs to the patchwork checks, lest they
> think it's a public CI and they can fling broken code at the list :(
> But yes, in case "code doesn't build" needs a further explanation:
> 
> net/core/pktgen.c: In function ‘pktgen_finalize_skb’:
> ./../include/linux/compiler_types.h:510:45: error: call to ‘__compiletime_assert_928’ declared with
> attribute error: min(datalen/frags, ((1UL) << 12)) signedness error
...
> ../net/core/pktgen.c:2796:28: note: in expansion of macro ‘min’
>  2796 |                 frag_len = min(datalen/frags, PAGE_SIZE);
>       |                            ^~~

I can't help feeling that a signed divide isn't intended here.
Which rather implies that both datalen and frags are signed types.
Whereas neither can be sensibly negative.
Perhaps that is the real bug?

	David

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