bpf: Cleanup check_refcount_ok
authorDave Marchevsky <davemarchevsky@fb.com>
Mon, 8 Aug 2022 17:15:59 +0000 (10:15 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Tue, 9 Aug 2022 16:43:31 +0000 (09:43 -0700)
commitb2d8ef19c6e7ed71ba5092feb0710063a751834f
tree486ef2b27fe42453fbd101bd3e38c0f4384310ec
parent6e116280b41b0cbfd90dfe9fa66e07ff348d50d5
bpf: Cleanup check_refcount_ok

Discussion around a recently-submitted patch provided historical
context for check_refcount_ok [0]. Specifically, the function and its
helpers - may_be_acquire_function and arg_type_may_be_refcounted -
predate the OBJ_RELEASE type flag and the addition of many more helpers
with acquire/release semantics.

The purpose of check_refcount_ok is to ensure:
  1) Helper doesn't have multiple uses of return reg's ref_obj_id
  2) Helper with release semantics only has one arg needing to be
  released, since that's tracked using meta->ref_obj_id

With current verifier, it's safe to remove check_refcount_ok and its
helpers. Since addition of OBJ_RELEASE type flag, case 2) has been
handled by the arg_type_is_release check in check_func_arg. To ensure
case 1) won't result in verifier silently prioritizing one use of
ref_obj_id, this patch adds a helper_multiple_ref_obj_use check which
fails loudly if a helper passes > 1 test for use of ref_obj_id.

  [0]: lore.kernel.org/bpf/20220713234529.4154673-1-davemarchevsky@fb.com

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Joanne Koong <joannelkoong@gmail.com>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Link: https://lore.kernel.org/r/20220808171559.3251090-1-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
kernel/bpf/verifier.c