Re: linux-next: build warning after merge of the landlock tree

[Mickaël Salaün]

On Tue, Oct 03, 2023 at 03:23:22PM +0200, Geert Uytterhoeven wrote:
> Hi Mickaël,
> 
> On Tue, Oct 3, 2023 at 3:15 PM Mickaël Salaün <mic@xxxxxxxxxxx> wrote:
> > On Tue, Oct 03, 2023 at 02:27:37PM +1100, Stephen Rothwell wrote:
> > > After merging the landlock tree, today's linux-next build (powerpc
> > > allyesconfig) produced this warning:
> > >
> > > samples/landlock/sandboxer.c: In function 'populate_ruleset_net':
> > > samples/landlock/sandboxer.c:170:78: warning: format '%llu' expects argument of type 'long long unsigned int', but argument 3 has type '__u64' {aka 'long unsigned int'} [-Wformat=]
> > >   170 |                                 "Failed to update the ruleset with port \"%llu\": %s\n",
> > >       |                                                                           ~~~^
> > >       |                                                                              |
> > >       |                                                                              long long unsigned int
> > >       |                                                                           %lu
> > >   171 |                                 net_port.port, strerror(errno));
> > >       |                                 ~~~~~~~~~~~~~
> > >       |                                         |
> > >       |                                         __u64 {aka long unsigned int}
> > >
> > > Introduced by commit
> > >
> > >   24889e7a2079 ("samples/landlock: Add network demo")
> >
> > PowerPC-64 follows the LP64 data model and then uses int-l64.h (instead of
> > int-ll64.h like most architectures) for user space code.
> >
> > Here is the same code with the (suggested) "%lu" token on x86_86:
> >
> >   samples/landlock/sandboxer.c: In function ‘populate_ruleset_net’:
> >   samples/landlock/sandboxer.c:170:77: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 3 has type ‘__u64’ {aka ‘long long unsigned int’} [-Werror=format=]
> >     170 |                                 "Failed to update the ruleset with port \"%lu\": %s\n",
> >         |                                                                           ~~^
> >         |                                                                             |
> >         |                                                                             long unsigned int
> >         |                                                                           %llu
> >     171 |                                 net_port.port, strerror(errno));
> >         |                                 ~~~~~~~~~~~~~
> >         |                                         |
> >         |                                         __u64 {aka long long unsigned int}
> >
> >
> > We would then need to cast __u64 to unsigned long long to avoid this warning,
> > which may look useless, of even buggy, for people taking a look at this sample.
> 
> In userspace code, you are supposed to #include <inttypes.h>
> and use PRIu64.

Thanks for these tips!

> 
> > Anyway, it makes more sense to cast it to __u16 because it is the
> > expected type for a TCP port. I'm updating the patch with that.
> > Konstantin, please take this fix for the next series:
> > https://git.kernel.org/mic/c/fc9de206a61a
> 
> Until someone passes a too large number, and it becomes truncated...

That should not happen because it is checked by the kernel (for this
specific case), but let's make it simple and print the 64-bit value.

> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 
> -- 
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
> 
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds