bpf: Use bpf_mem_free_rcu when bpf_obj_dropping non-refcounted nodes
authorDave Marchevsky <davemarchevsky@fb.com>
Tue, 7 Nov 2023 08:56:36 +0000 (00:56 -0800)
committerAlexei Starovoitov <ast@kernel.org>
Fri, 10 Nov 2023 03:07:51 +0000 (19:07 -0800)
commit649924b76ab151a96bdd22a97a993fb0421f134c
treefbe3b9142af1969a7597d03cb57261d23534f6ef
parentf460e7bdb027d1da93f0c5090b239889cd46a33d
bpf: Use bpf_mem_free_rcu when bpf_obj_dropping non-refcounted nodes

The use of bpf_mem_free_rcu to free refcounted local kptrs was added
in commit 7e26cd12ad1c ("bpf: Use bpf_mem_free_rcu when
bpf_obj_dropping refcounted nodes"). In the cover letter for the
series containing that patch [0] I commented:

    Perhaps it makes sense to move to mem_free_rcu for _all_
    non-owning refs in the future, not just refcounted. This might
    allow custom non-owning ref lifetime + invalidation logic to be
    entirely subsumed by MEM_RCU handling. IMO this needs a bit more
    thought and should be tackled outside of a fix series, so it's not
    attempted here.

It's time to start moving in the "non-owning refs have MEM_RCU
lifetime" direction. As mentioned in that comment, using
bpf_mem_free_rcu for all local kptrs - not just refcounted - is
necessarily the first step towards that goal. This patch does so.

After this patch the memory pointed to by all local kptrs will not be
reused until RCU grace period elapses. The verifier's understanding of
non-owning ref validity and the clobbering logic it uses to enforce
that understanding are not changed here, that'll happen gradually in
future work, including further patches in the series.

  [0]: https://lore.kernel.org/all/20230821193311.3290257-1-davemarchevsky@fb.com/

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Link: https://lore.kernel.org/r/20231107085639.3016113-4-davemarchevsky@fb.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
kernel/bpf/helpers.c