Re: [PATCH 5.15 000/846] 5.15.17-rc1 review

[Andrii Nakryiko]

On Mon, Jan 24, 2022 at 12:36 PM Holger Hoffstätte
<holger@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>
> On 2022-01-24 19:31, Greg Kroah-Hartman wrote:
> > This is the start of the stable review cycle for the 5.15.17 release.
>
> Oh noes :(
>
>    DESCEND bpf/resolve_btfids
>    MKDIR     /tmp/linux-5.15.17/tools/bpf/resolve_btfids//libbpf
>    GEN     /tmp/linux-5.15.17/tools/bpf/resolve_btfids/libbpf/bpf_helper_defs.h
>    MKDIR   /tmp/linux-5.15.17/tools/bpf/resolve_btfids/libbpf/staticobjs/
>    CC      /tmp/linux-5.15.17/tools/bpf/resolve_btfids/libbpf/staticobjs/libbpf.o
> libbpf.c: In function 'bpf_object__elf_collect':
> libbpf.c:3038:31: error: invalid type argument of '->' (have 'GElf_Shdr' {aka 'Elf64_Shdr'})
>   3038 |                         if (sh->sh_type != SHT_PROGBITS)
>        |                               ^~
> libbpf.c:3042:31: error: invalid type argument of '->' (have 'GElf_Shdr' {aka 'Elf64_Shdr'})
>   3042 |                         if (sh->sh_type != SHT_PROGBITS)
>        |                               ^~
> make[4]: *** [/tmp/linux-5.15.17/tools/build/Makefile.build:97: /tmp/linux-5.15.17/tools/bpf/resolve_btfids/libbpf/staticobjs/libbpf.o] Error 1
> make[3]: *** [Makefile:158: /tmp/linux-5.15.17/tools/bpf/resolve_btfids/libbpf/staticobjs/libbpf-in.o] Error 2
> make[2]: *** [Makefile:44: /tmp/linux-5.15.17/tools/bpf/resolve_btfids//libbpf/libbpf.a] Error 2
> make[1]: *** [Makefile:72: bpf/resolve_btfids] Error 2
> make: *** [Makefile:1371: tools/bpf/resolve_btfids] Error 2
>
> Reverting "libbpf-validate-that-.btf-and-.btf.ext-sections-cont.patcht" aka
> this one:
>
> > Andrii Nakryiko <andrii@xxxxxxxxxx>
> >      libbpf: Validate that .BTF and .BTF.ext sections contain data
>
> makes it build & run fine. I looked for followups but couldn't find anything that
> stood out, maybe the BPF folks (cc'ed) know what's missing/wrong.
>

That small fix depends on much bigger refactoring in ad23b7238474
("libbpf: Use Elf64-specific types explicitly for dealing with ELF").
I think this small fix can be dropped.

That's sort of a general rule with libbpf-related fixes, they are
usually not that critical to backport to stable, because most users
use/build libbpf from its Github mirror, which is always taken from
latest bpf-next. Libbpf is also not supposed to be used with untrusted
inputs (i.e., BPF object files) as BPF programs are loaded into the
kernel under root.


> cheers
> Holger