[bpf-next,0/5] introduce bpf_strncmp() helper

Message

Hou Tao Nov. 30, 2021, 2:22 p.m. UTC

Hi,

The motivation for introducing bpf_strncmp() helper comes from
two aspects:

(1) clang doesn't always replace strncmp() automatically
In tracing program, sometimes we need to using a home-made
strncmp() to check whether or not the file name is expected.

(2) the performance of home-made strncmp is not so good
As shown in the benchmark in patch #4, the performance of
bpf_strncmp() helper is 18% or 33% better than home-made strncmp()
under x86-64 or arm64 when the compared string length is 64. When
the string length grows to 4095, the performance win will be
179% or 600% under x86-64 or arm64.

The prototype of bpf_strncmp() has changed from

  bpf_strncmp(const char *s1, const char *s2, u32 s2_sz)

to

  bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)

The main reason is readability and there is nearly no performance
difference between these two APIs (refer to the data attached below
[1]).

Any comments are welcome.
Regards,
Tao

Change Log:
v1:
 * change API to bpf_strncmp(const char *s1, u32 s1_sz, const char *s2)
 * add benchmark refactor and benchmark between bpf_strncmp() and strncmp()

RFC: https://lore.kernel.org/bpf/[e-mail]/

[1] Performance difference between two APIs under x86-64:

helper_rfc-X: use bpf_strncmp in RFC to compare X-sized string
helper-Y: use bpf_strncmp in v1 to compare Y-sized string

helper_rfc-1         3.482 ± 0.002M/s (drops 0.000 ± 0.000M/s)
helper-1             3.485 ± 0.001M/s (drops 0.000 ± 0.000M/s)

helper_rfc-8         3.428 ± 0.001M/s (drops 0.000 ± 0.000M/s)
helper-8             3.434 ± 0.001M/s (drops 0.000 ± 0.000M/s)

helper_rfc-32        3.253 ± 0.002M/s (drops 0.000 ± 0.000M/s)
helper-32            3.234 ± 0.001M/s (drops 0.000 ± 0.000M/s)

helper_rfc-64        3.039 ± 0.000M/s (drops 0.000 ± 0.000M/s)
helper-64            3.042 ± 0.001M/s (drops 0.000 ± 0.000M/s)

helper_rfc-128       2.640 ± 0.000M/s (drops 0.000 ± 0.000M/s)
helper-128           2.633 ± 0.000M/s (drops 0.000 ± 0.000M/s)

helper_rfc-512       1.576 ± 0.000M/s (drops 0.000 ± 0.000M/s)
helper-512           1.574 ± 0.000M/s (drops 0.000 ± 0.000M/s)

helper_rfc-2048      0.602 ± 0.000M/s (drops 0.000 ± 0.000M/s)
helper-2048          0.602 ± 0.000M/s (drops 0.000 ± 0.000M/s)

helper_rfc-4095      0.328 ± 0.000M/s (drops 0.000 ± 0.000M/s)
helper-4095          0.328 ± 0.000M/s (drops 0.000 ± 0.000M/s)

Hou Tao (5):
  bpf: add bpf_strncmp helper
  selftests/bpf: fix checkpatch error on empty function parameter
  selftests/bpf: factor out common helpers for benchmarks
  selftests/bpf: add benchmark for bpf_strncmp() helper
  selftests/bpf: add test cases for bpf_strncmp()

 include/linux/bpf.h                           |   1 +
 include/uapi/linux/bpf.h                      |  11 ++
 kernel/bpf/helpers.c                          |  16 ++
 tools/include/uapi/linux/bpf.h                |  11 ++
 tools/testing/selftests/bpf/Makefile          |   4 +-
 tools/testing/selftests/bpf/bench.c           |  21 ++-
 tools/testing/selftests/bpf/bench.h           |  34 +++-
 .../bpf/benchs/bench_bloom_filter_map.c       |  44 ++---
 .../selftests/bpf/benchs/bench_count.c        |  16 +-
 .../selftests/bpf/benchs/bench_rename.c       |  43 ++---
 .../selftests/bpf/benchs/bench_ringbufs.c     |  21 +--
 .../selftests/bpf/benchs/bench_strncmp.c      | 150 ++++++++++++++++
 .../selftests/bpf/benchs/bench_trigger.c      |  79 ++++----
 .../selftests/bpf/benchs/run_bench_strncmp.sh |  12 ++
 .../selftests/bpf/prog_tests/test_strncmp.c   | 170 ++++++++++++++++++
 .../selftests/bpf/progs/strncmp_bench.c       |  50 ++++++
 .../selftests/bpf/progs/strncmp_test.c        |  59 ++++++
 17 files changed, 604 insertions(+), 138 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/benchs/bench_strncmp.c
 create mode 100755 tools/testing/selftests/bpf/benchs/run_bench_strncmp.sh
 create mode 100644 tools/testing/selftests/bpf/prog_tests/test_strncmp.c
 create mode 100644 tools/testing/selftests/bpf/progs/strncmp_bench.c
 create mode 100644 tools/testing/selftests/bpf/progs/strncmp_test.c

Comments

Alexei Starovoitov Dec. 3, 2021, 2:09 a.m. UTC | #1

On Tue, Nov 30, 2021 at 6:07 AM Hou Tao <[e-mail]> wrote:
>
> Hi,
>
> The motivation for introducing bpf_strncmp() helper comes from
> two aspects:
>
> (1) clang doesn't always replace strncmp() automatically
> In tracing program, sometimes we need to using a home-made
> strncmp() to check whether or not the file name is expected.
>
> (2) the performance of home-made strncmp is not so good
> As shown in the benchmark in patch #4, the performance of
> bpf_strncmp() helper is 18% or 33% better than home-made strncmp()
> under x86-64 or arm64 when the compared string length is 64. When
> the string length grows to 4095, the performance win will be
> 179% or 600% under x86-64 or arm64.

I think 'home made' strncmp could have been written
differently. I bet in bpf assembly it would be much closer
in performance if not the same,
but the helper is useful.
The patch set doesn't apply cleanly.
Pls respin.