This patch-set fixes bug in states pruning logic hit in mailing list discussion [0]. The details of the fix are in patch #1. The main idea for the fix belongs to Yonghong Song, mine contribution is merely in review and test cases. There are some changes in verification performance: File Program Insns (DIFF) States (DIFF) ------------------------- ------------- --------------- -------------- pyperf600_bpf_loop.bpf.o on_event +15 (+0.42%) +0 (+0.00%) strobemeta_bpf_loop.bpf.o on_event +857 (+37.95%) +60 (+38.96%) xdp_synproxy_kern.bpf.o syncookie_tc +2892 (+30.39%) +109 (+36.33%) xdp_synproxy_kern.bpf.o syncookie_xdp +2892 (+30.01%) +109 (+36.09%) (when tested on a subset of selftests identified by selftests/bpf/veristat.cfg and Cilium bpf object files from [4]) Changelog: v2 [2] -> v3: - fixes for verifier.c commit message as suggested by Yonghong; - patch-set re-rerouted to 'bpf' tree as suggested in [2]; - patch for test_tcp_custom_syncookie is sent separately to 'bpf-next' [3]. - veristat results updated using 'bpf' tree as baseline and clang 16. v1 [1] -> v2: - patch #2 commit message updated to better reflect verifier behavior with regards to checkpoints tree (suggested by Yonghong); - veristat results added (suggested by Andrii). [0] https://lore.kernel.org/bpf/9b251840-7cb8-4d17-bd23-1fc8071d8eef@xxxxxxxxx/ [1] https://lore.kernel.org/bpf/20240212143832.28838-1-eddyz87@xxxxxxxxx/ [2] https://lore.kernel.org/bpf/20240216150334.31937-1-eddyz87@xxxxxxxxx/ [3] https://lore.kernel.org/bpf/20240222150300.14909-1-eddyz87@xxxxxxxxx/ [4] https://github.com/anakryiko/cilium Eduard Zingerman (2): bpf: check bpf_func_state->callback_depth when pruning states selftests/bpf: test case for callback_depth states pruning logic kernel/bpf/verifier.c | 3 + .../bpf/progs/verifier_iterating_callbacks.c | 70 +++++++++++++++++++ 2 files changed, 73 insertions(+) -- 2.43.0
