Re: [PATCH bpf-next 1/2] libbpf: add option to force-anonymize nested structs for BTF dump

[Eduard Zingerman <eddyz87@gmail.com>]

On Wed, 2025-12-17 at 15:34 -0800, Alexei Starovoitov wrote:
> On Wed, Dec 17, 2025 at 2:47 PM Eduard Zingerman <eddyz87@gmail.com> wrote:
> >
> > >   $ cat ms-ext-test2.c
> > >   struct foo {
> > >     int a;
> > >   } __attribute__((preserve_access_index));
> > >
> > >   struct bar {
> > >     struct foo;
> > >   } __attribute__((preserve_access_index));
> > >
> > >   int buz(struct bar *bar) {
> > >     return bar->a;
> > >   }
> > >
> > >   $ clang -O2 -g -fms-extensions --target=bpf -c ms-ext-test2.c
> > >   ms-ext-test2.c:6:3: warning: anonymous structs are a Microsoft extension [-Wmicrosoft-anon-tag]
> > >       6 |   struct foo;
> > >         |   ^~~~~~~~~~
> > >   1 warning generated.
> > >
> > >   $ llvm-objdump -Sdr ms-ext-test2.o
> > >
> > >   ms-ext-test2.o: file format elf64-bpf
> > >
> > >   Disassembly of section .text:
> > >
> > >   0000000000000000 <buz>:
> > >   ;   return bar->a;
> > >          0:       61 10 00 00 00 00 00 00 w0 = *(u32 *)(r1 + 0x0)
> > >                   0000000000000000:  CO-RE <byte_off> [2] struct bar::<anon 0>.a (0:0:0)
> > >          1:       95 00 00 00 00 00 00 00 exit
> > >
> > > Note the "<anon 0>" in the relocation.
> > > It appears that we loose no information if structures are unrolled.
>
> Forgot to mention the CORE concern earlier...
> Does the above work with current logic in relo_core.c ?
> If not, we should definitely unconditionally unroll
> to avoid fixing CORE.

I think there might be an issue with CO-RE.
Here is an example:

  struct foo {
    int a;
  } __attribute__((preserve_access_index));

  struct bar {
  #ifdef USE_MS
    struct foo;
  #else
    struct { int a; };
  #endif
  } __attribute__((preserve_access_index));

  int buz(struct bar *bar) {
    return bar->a;
  }

Here is what I get with USE_MS:

  $ llvm-objdump -Sdr ms-ext-test2.o

  ms-ext-test2.o: file format elf64-bpf

  Disassembly of section .text:

  0000000000000000 <buz>:
  ;   return bar->a;
         0:       61 10 00 00 00 00 00 00 w0 = *(u32 *)(r1 + 0x0)
                  0000000000000000:  CO-RE <byte_off> [2] struct bar::<anon 0>.a (0:0:0)
         1:       95 00 00 00 00 00 00 00 exit

  $ bpftool btf dump file ms-ext-test2.o
  [1] PTR '(anon)' type_id=2
  [2] STRUCT 'bar' size=4 vlen=1
          '(anon)' type_id=3 bits_offset=0
  [3] STRUCT 'foo' size=4 vlen=1
          'a' type_id=4 bits_offset=0
  [4] INT 'int' size=4 bits_offset=0 nr_bits=32 encoding=SIGNED
  [5] FUNC_PROTO '(anon)' ret_type_id=4 vlen=1
          'bar' type_id=1
  [6] FUNC 'buz' type_id=5 linkage=global

And here is without USE_MS:

  $ llvm-objdump -Sdr ms-ext-test2.o

  ms-ext-test2.o: file format elf64-bpf

  Disassembly of section .text:

  0000000000000000 <buz>:
  ;   return bar->a;
         0:       61 10 00 00 00 00 00 00 w0 = *(u32 *)(r1 + 0x0)
                  0000000000000000:  CO-RE <byte_off> [2] struct bar::<anon 0>.a (0:0:0)
         1:       95 00 00 00 00 00 00 00 exit

  $ bpftool btf dump file ms-ext-test2.o
  [1] PTR '(anon)' type_id=2
  [2] STRUCT 'bar' size=4 vlen=1
          '(anon)' type_id=3 bits_offset=0
  [3] STRUCT '(anon)' size=4 vlen=1
          'a' type_id=4 bits_offset=0
  [4] INT 'int' size=4 bits_offset=0 nr_bits=32 encoding=SIGNED
  [5] FUNC_PROTO '(anon)' ret_type_id=4 vlen=1
          'bar' type_id=1
  [6] FUNC 'buz' type_id=5 linkage=global

So, with USE_MS the relocation captures the offset inside 'struct foo'.
And this is important for CO-RE offsets resolution.
So unrolling structures is actually a problem.