[v2,bpf-next,0/5] CO-RE relocation selftests fixes

Lorenz Bauer noticed that core_reloc selftest has two inverted CHECK()
conditions, allowing failing tests to pass unnoticed. Fixing that opened up
few long-standing (field existence and direct memory bitfields) and one recent
failures (BTF_KIND_FLOAT relos).

This patch set fixes core_reloc selftest to capture such failures reliably in
the future. It also fixes all the newly failing tests. See individual patches
for details.

This patch set also completes a set of ASSERT_xxx() macros, so now there
should be a very little reason to use verbose and error-prone generic CHECK()
macro.

v1->v2:
  - updated bpf_core_fields_are_compat() comment to mention FLOAT (Lorenz).

Cc: Lorenz Bauer <lmb@cloudflare.com>

Andrii Nakryiko (5):
  selftests/bpf: add remaining ASSERT_xxx() variants
  libbpf: support BTF_KIND_FLOAT during type compatibility checks in
    CO-RE
  selftests/bpf: fix BPF_CORE_READ_BITFIELD() macro
  selftests/bpf: fix field existence CO-RE reloc tests
  selftests/bpf: fix core_reloc test runner

 tools/lib/bpf/bpf_core_read.h                 | 16 ++++--
 tools/lib/bpf/libbpf.c                        |  6 ++-
 .../selftests/bpf/prog_tests/btf_dump.c       |  2 +-
 .../selftests/bpf/prog_tests/btf_endian.c     |  4 +-
 .../selftests/bpf/prog_tests/cgroup_link.c    |  2 +-
 .../selftests/bpf/prog_tests/core_reloc.c     | 51 +++++++++++--------
 .../selftests/bpf/prog_tests/kfree_skb.c      |  2 +-
 .../selftests/bpf/prog_tests/resolve_btfids.c |  7 +--
 .../selftests/bpf/prog_tests/snprintf_btf.c   |  4 +-
 ...ore_reloc_existence___err_wrong_arr_kind.c |  3 --
 ...loc_existence___err_wrong_arr_value_type.c |  3 --
 ...ore_reloc_existence___err_wrong_int_kind.c |  3 --
 ..._core_reloc_existence___err_wrong_int_sz.c |  3 --
 ...ore_reloc_existence___err_wrong_int_type.c |  3 --
 ..._reloc_existence___err_wrong_struct_type.c |  3 --
 ..._core_reloc_existence___wrong_field_defs.c |  3 ++
 .../selftests/bpf/progs/core_reloc_types.h    | 20 +-------
 tools/testing/selftests/bpf/test_progs.h      | 50 +++++++++++++++++-
 18 files changed, 108 insertions(+), 77 deletions(-)
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_arr_kind.c
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_arr_value_type.c
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_int_kind.c
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_int_sz.c
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_int_type.c
 delete mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___err_wrong_struct_type.c
 create mode 100644 tools/testing/selftests/bpf/progs/btf__core_reloc_existence___wrong_field_defs.c

On Mon, Apr 26, 2021 at 12:30 PM Andrii Nakryiko <andrii@kernel.org> wrote:
>

> Lorenz Bauer noticed that core_reloc selftest has two inverted CHECK()

> conditions, allowing failing tests to pass unnoticed. Fixing that opened up

> few long-standing (field existence and direct memory bitfields) and one recent

> failures (BTF_KIND_FLOAT relos).

>

> This patch set fixes core_reloc selftest to capture such failures reliably in

> the future. It also fixes all the newly failing tests. See individual patches

> for details.

>

> This patch set also completes a set of ASSERT_xxx() macros, so now there

> should be a very little reason to use verbose and error-prone generic CHECK()

> macro.

>

> v1->v2:

>   - updated bpf_core_fields_are_compat() comment to mention FLOAT (Lorenz).


Applied.