3 Commits

Author	SHA1	Message	Date
Andrii Nakryiko	4eaf0b5c5e	selftest/bpf: Fmod_ret prog and implement test_overhead as part of bench ... Add fmod_ret BPF program to existing test_overhead selftest. Also re-implement user-space benchmarking part into benchmark runner to compare results. Results with ./bench are consistently somewhat lower than test_overhead's, but relative performance of various types of BPF programs stay consisten (e.g., kretprobe is noticeably slower). This slowdown seems to be coming from the fact that test_overhead is single-threaded, while benchmark always spins off at least one thread for producer. This has been confirmed by hacking multi-threaded test_overhead variant and also single-threaded bench variant. Resutls are below. run_bench_rename.sh script from benchs/ subdirectory was used to produce results for ./bench. Single-threaded implementations =============================== /* bench: single-threaded, atomics */ base : 4.622 ± 0.049M/s kprobe : 3.673 ± 0.052M/s kretprobe : 2.625 ± 0.052M/s rawtp : 4.369 ± 0.089M/s fentry : 4.201 ± 0.558M/s fexit : 4.309 ± 0.148M/s fmodret : 4.314 ± 0.203M/s /* selftest: single-threaded, no atomics */ task_rename base 4555K events per sec task_rename kprobe 3643K events per sec task_rename kretprobe 2506K events per sec task_rename raw_tp 4303K events per sec task_rename fentry 4307K events per sec task_rename fexit 4010K events per sec task_rename fmod_ret 3984K events per sec Multi-threaded implementations ============================== /* bench: multi-threaded w/ atomics */ base : 3.910 ± 0.023M/s kprobe : 3.048 ± 0.037M/s kretprobe : 2.300 ± 0.015M/s rawtp : 3.687 ± 0.034M/s fentry : 3.740 ± 0.087M/s fexit : 3.510 ± 0.009M/s fmodret : 3.485 ± 0.050M/s /* selftest: multi-threaded w/ atomics */ task_rename base 3872K events per sec task_rename kprobe 3068K events per sec task_rename kretprobe 2350K events per sec task_rename raw_tp 3731K events per sec task_rename fentry 3639K events per sec task_rename fexit 3558K events per sec task_rename fmod_ret 3511K events per sec /* selftest: multi-threaded, no atomics */ task_rename base 3945K events per sec task_rename kprobe 3298K events per sec task_rename kretprobe 2451K events per sec task_rename raw_tp 3718K events per sec task_rename fentry 3782K events per sec task_rename fexit 3543K events per sec task_rename fmod_ret 3526K events per sec Note that the fact that ./bench benchmark always uses atomic increments for counting, while test_overhead doesn't, doesn't influence test results all that much. Signed-off-by: Andrii Nakryiko <andriin@fb.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: John Fastabend <john.fastabend@gmail.com> Acked-by: Yonghong Song <yhs@fb.com> Link: https://lore.kernel.org/bpf/20200512192445.2351848-4-andriin@fb.com	2020-05-13 12:19:38 -07:00
Stanislav Fomichev	e43002242a	selftests/bpf: Restore original comm in test_overhead ... test_overhead changes task comm in order to estimate BPF trampoline overhead but never sets the comm back to the original one. We have the tests (like core_reloc.c) that have 'test_progs' as hard-coded expected comm, so let's try to preserve the original comm. Currently, everything works because the order of execution is: first core_recloc, then test_overhead; but let's make it a bit future-proof. Other related changes: use 'test_overhead' as new comm instead of 'test' to make it easy to debug and drop '\n' at the end. Signed-off-by: Stanislav Fomichev <sdf@google.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Petar Penkov <ppenkov@google.com> Acked-by: Song Liu <songliubraving@fb.com> Link: https://lore.kernel.org/bpf/20200108192132.189221-1-sdf@google.com	2020-01-09 08:42:07 -08:00
Alexei Starovoitov	c4781e37c6	selftests/bpf: Add BPF trampoline performance test ... Add a test that benchmarks different ways of attaching BPF program to a kernel function. Here are the results for 2.4Ghz x86 cpu on a kernel without mitigations: $ ./test_progs -n 49 -v|grep events task_rename base 2743K events per sec task_rename kprobe 2419K events per sec task_rename kretprobe 1876K events per sec task_rename raw_tp 2578K events per sec task_rename fentry 2710K events per sec task_rename fexit 2685K events per sec On a kernel with retpoline: $ ./test_progs -n 49 -v|grep events task_rename base 2401K events per sec task_rename kprobe 1930K events per sec task_rename kretprobe 1485K events per sec task_rename raw_tp 2053K events per sec task_rename fentry 2351K events per sec task_rename fexit 2185K events per sec All 5 approaches: - kprobe/kretprobe in __set_task_comm() - raw tracepoint in trace_task_rename() - fentry/fexit in __set_task_comm() are roughly equivalent. __set_task_comm() by itself is quite fast, so any extra instructions add up. Until BPF trampoline was introduced the fastest mechanism was raw tracepoint. kprobe via ftrace was second best. kretprobe is slow due to trap. New fentry/fexit methods via BPF trampoline are clearly the fastest and the difference is more pronounced with retpoline on, since BPF trampoline doesn't use indirect jumps. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Acked-by: John Fastabend <john.fastabend@gmail.com> Link: https://lore.kernel.org/bpf/20191122011515.255371-1-ast@kernel.org	2019-11-24 16:58:46 -08:00