by Andrii Nakryiko

[permalink] [raw]

Subject: Re: [PATCH v1 02/13] libbpf: Make __printf define conditional

On Mon, Mar 11, 2024 at 11:54 AM Ian Rogers <[email protected]> wrote:
>
> On Mon, Mar 11, 2024 at 10:49 AM Andrii Nakryiko
> <[email protected]> wrote:
> >
> > On Sat, Mar 9, 2024 at 6:05 PM Ian Rogers <[email protected]> wrote:
> > >
> > > libbpf depends upon linux/err.h which has a linux/compiler.h
> > > dependency. In the kernel includes, as opposed to the tools version,
> > > linux/compiler.h includes linux/compiler_attributes.h which defines
> > > __printf. As the libbpf.c __printf definition isn't guarded by an
> > > ifndef, this leads to a duplicate definition compilation error when
> > > trying to update the tools/include/linux/compiler.h. Fix this by
> > > adding the missing ifndef.
> > >
> > > Signed-off-by: Ian Rogers <[email protected]>
> > > ---
> > > tools/lib/bpf/libbpf.c | 4 +++-
> > > 1 file changed, 3 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > > index afd09571c482..2152360b4b18 100644
> > > --- a/tools/lib/bpf/libbpf.c
> > > +++ b/tools/lib/bpf/libbpf.c
> > > @@ -66,7 +66,9 @@
> > > */
> > > #pragma GCC diagnostic ignored "-Wformat-nonliteral"
> > >
> > > -#define __printf(a, b) __attribute__((format(printf, a, b)))
> > > +#ifndef __printf
> > > +# define __printf(a, b) __attribute__((format(printf, a, b)))
> >
> > styling nit: don't add spaces between # and define, please
> >
> > overall LGTM
> >
> > Acked-by: Andrii Nakryiko <[email protected]>
> >
> > Two questions, though.
> >
> > 1. It seems like just dropping #define __printf in libbpf.c compiles
> > fine (I checked both building libbpf directly, and BPF selftest, and
> > perf, and bpftool directly, all of them built fine). So we can
> > probably just drop this. I'll need to add __printf on Github, but
> > that's fine.
> >
> > 2. Logistics. Which tree should this patch go through? Can I land it
> > in bpf-next or it's too much inconvenience for you?
>
> Thanks Andrii,
>
> dropping the #define (1) sgtm but the current compiler.h will fail to
> build libbpf.c without the later compiler.h update in this series.
> This causes another logistic issue for your point 2. Presumably if
> this patch goes through bpf-next, the first patch "tools bpf:
> Synchronize bpf.h with kernel uapi version" should also go through the
> bpf-next.
>

That's what I'm saying, it seems to work without your patches already.
At least on bpf-next/master. But it's ok, let's keep it and just add
#ifndef guard, that will make my life easier when syncing to Github
later one. Then the patch can go through other trees and eventually
make it into bpf-next and then Github. So please keep my ack for
#ifndef version, thanks.

> Thanks,
> Ian
>
>
> > > +#endif
> > >
> > > static struct bpf_map *bpf_object__add_map(struct bpf_object *obj);
> > > static bool prog_is_subprog(const struct bpf_object *obj, const struct bpf_program *prog);
> > > --
> > > 2.44.0.278.ge034bb2e1d-goog
> > >