Re: linux-next: build failure after merge of the bpf-next tree

From: Alexei Starovoitov
Date: Tue Sep 14 2021 - 21:27:32 EST


On Tue, Sep 14, 2021 at 4:59 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
>
> Hi Andrii,
>
> On Tue, 14 Sep 2021 16:40:37 -0700 Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote:
> >
> > On Tue, Sep 14, 2021 at 4:38 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> > >
> > > Hi Andrii,
> > >
> > > On Tue, 14 Sep 2021 16:25:55 -0700 Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx> wrote:
> > > >
> > > > On Mon, Sep 13, 2021 at 6:37 PM Stephen Rothwell <sfr@xxxxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > After merging the bpf-next tree, today's linux-next build (perf) failed
> > > > > like this:
> > > > >
> > > > > util/bpf-event.c: In function 'btf__load_from_kernel_by_id':
> > > > > util/bpf-event.c:27:8: error: 'btf__get_from_id' is deprecated: libbpf v0.6+: use btf__load_from_kernel_by_id instead [-Werror=deprecated-declarations]
> > > > > 27 | int err = btf__get_from_id(id, &btf);
> > > > > | ^~~
> > > > > In file included from util/bpf-event.c:5:
> > > > > /home/sfr/next/next/tools/lib/bpf/btf.h:54:16: note: declared here
> > > > > 54 | LIBBPF_API int btf__get_from_id(__u32 id, struct btf **btf);
> > > > > | ^~~~~~~~~~~~~~~~
> > > > > cc1: all warnings being treated as errors
> > > > >
> > > > > Caused by commit
> > > > >
> > > > > 0b46b7550560 ("libbpf: Add LIBBPF_DEPRECATED_SINCE macro for scheduling API deprecations")
> > > >
> > > > Should be fixed by [0], when applied to perf tree. Thanks for reporting!
> > > >
> > > > [0] https://patchwork.kernel.org/project/netdevbpf/patch/20210914170004.4185659-1-andrii@xxxxxxxxxx/
> > >
> > > That really needs to be applied to the bpf-next tree (presumably with
> > > the appropriate Acks).
> > >
> >
> > This is perf code that's not in bpf-next yet.
>
> Then you need to think of a solution for the bpf-next tree, as it will
> not build (allmodconfig) when combined with Linus' current tree. And
> when it is merged into the net-next tree, that tree will be broken as
> well.

It's not a problem of bpf-tree and it cannot be fixed in bpf-tree.
The bug is in perf tree.