[PATCH bpf-next v1 0/7] selftests/bpf: cleanup bpf-gcc runs

[Vineet Gupta <vineet.gupta@linux.dev>]

This series came out of testing a still in-works __arena support for
BPF_GCC and anomalies seen in the A/B reports.

BPF_GCC selftest builds reports a number of things that are not true.
 - Tests the compiler cannot build do not appear at all, so the
   report is silent about them
 - Others are compiled down to an empty body and report OK despite
   not even been executed
 - Two programs that fail to verify take four unrelated tests down with
   them, because everything in that object is loaded together.
 - Unconditional gating on __clang__ when newer GCC support for those
   is coming online.

Only 7/7 needs the in-works arena support to be exercised at all; the
rest are measured below with a stock bpf-gcc.  Nothing in the series
changes what clang emits.

What each patch does, and what it measurably changes. Numbers are
test_progs on bpf-next with a stock bpf-gcc, before and after:

 1/7: Gate the callx tests on a CAN_USE_CALLX macro that accepts a GCC
      defining __BPF_FEATURE_CALLX, in addition to clang.
   -> 76 subtests run that were absent from the report entirely,
      75 OK and one @unpriv SKIP, no failures. The two dummy
      placeholders retire; callx_rodata_lskel SKIP -> OK.

 2/7: Move clang-only warning flags out of GCC builds.

 3/7: Add a __skip_without_arena tag, and give each
      #if __BPF_FEATURE_ADDR_SPACE_CAST guard an #else with a dummy
      carrying it, so a compiler that cannot build those tests says
      so instead of staying silent.
   -> 6 subtests newly visible, all SKIP; arena_kfunc_jit OK -> SKIP,
      every subtest it has being one of them.

 4/7: Make arena_base() opaque to GCC's -Warray-bounds, which fires
      because the macro reads a member of struct bpf_arena through the
      much smaller map definition.  A build failure today for any GCC
      with arena support, and for everyone once 5/7 drops the guards
      that were hiding those calls.  Behind #ifdef __clang__, since
      clang does not warn and the barrier is not free.
   -> no verdict change.  +35 instructions over 20 call sites in five
      progs under GCC; clang output unchanged, bit for bit.

 5/7: Put __skip_without_arena on the tests whose body dereferences an
      arena pointer -- and only those; drop the in-body guards.
   -> 6 OK -> SKIP, check_arena_arg_ret FAIL -> SKIP, and
      verifier_arena itself FAIL -> OK.  The 11 that never dereference
      stay OK, having gone from an empty body to actually running.

 6/7: Cast explicitly with bpf_addr_space_cast() in the arena subprog
      and callback faults, instead of leaving it to the compiler, as
      the sibling tests in the same file already do.
   -> 15 FAIL -> OK, every test that loads that object rather than
      just the arena ones, and 5 subtests newly visible.

 7/7: Replace a volatile register variable GCC is entitled to optimise
      away with the asm the mixed arena/scalar ALU tests actually need.
   -> not exercised by a stock bpf-gcc, which compiles both out with
      the rest of the arena block.

Totals for the series.  Nothing regresses:

  FAIL  473 -> 457        OK -> FAIL regressions:  0
  OK   7027 -> 7115
  SKIP  761 ->  775

Those numbers were taken on a revision that suppressed the warning
per-file instead of carrying 4/7; the series as posted was re-run and
gave the same result, which is what 4/7 changing no verdict predicts.

Two things worth re-iterating explicitly:

 - 1/7 needs a GCC defining __BPF_FEATURE_CALLX, in bleeding-edge
   but not released yet.

 - 7/7, need a GCC with arena support; that work is coming along
   but not yet ready for public review.

Vineet Gupta (7):
  selftests/bpf: Build callx tests with GCC that defines
    __BPF_FEATURE_CALLX
  selftests/bpf: don't pass clang-only warning flags to GCC
  selftests/bpf: report SKIP for arena tests a compiler cannot build
  selftests/bpf: make arena_base() opaque to GCC's -Warray-bounds
  selftests/bpf: don't report arena tests as passing with an empty body
  selftests/bpf: cast explicitly in the arena subprog and callback
    faults
  selftests/bpf: write the mixed arena/scalar ALU tests in asm

 tools/testing/selftests/bpf/Makefile          |   8 +-
 .../testing/selftests/bpf/Makefile.buildvars  |   6 +-
 .../bpf/libarena/include/bpf_arena_common.h   |  16 +++
 .../bpf/prog_tests/callx_rodata_lskel.c       |   2 +-
 .../selftests/bpf/progs/arena_kfunc_jit.c     |   9 ++
 tools/testing/selftests/bpf/progs/bpf_misc.h  |  21 +++
 .../selftests/bpf/progs/callx_rodata.c        |   4 +-
 .../bpf/progs/compute_live_registers.c        |  10 ++
 tools/testing/selftests/bpf/progs/stream.c    |   7 +-
 .../selftests/bpf/progs/verifier_arena.c      | 125 +++++++++---------
 .../bpf/progs/verifier_arena_large.c          |  24 ++--
 .../selftests/bpf/progs/verifier_callx.c      |   5 +-
 .../bpf/progs/verifier_callx_rodata.c         |   5 +-
 .../selftests/bpf/progs/verifier_ldsx.c       |   9 ++
 .../selftests/bpf/progs/verifier_zext.c       |   9 ++
 15 files changed, 163 insertions(+), 97 deletions(-)

-- 
2.53.0-Meta