Re: [PATCH bpf-next v10 4/4] selftests: bpf: crypto: add benchmark for crypto functions

[Martin KaFai Lau <martin.lau@linux.dev>]

On 4/22/24 3:50 PM, Vadim Fedorenko wrote:
> diff --git a/tools/testing/selftests/bpf/benchs/bench_bpf_crypto.c b/tools/testing/selftests/bpf/benchs/bench_bpf_crypto.c
> new file mode 100644
> index 000000000000..0b8c1f2fe7e6
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/benchs/bench_bpf_crypto.c
> @@ -0,0 +1,185 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Copyright (c) 2024 Meta Platforms, Inc. and affiliates. */
> +
> +#include <argp.h>
> +#include "bench.h"
> +#include "crypto_bench.skel.h"
> +
> +#define MAX_CIPHER_LEN 32
> +static char *input;

[ ... ]

> +static void *crypto_producer(void *input)

The bench result has all 0s in the output:

$> ./bench -p 4 crypto-decrypt
Setting up benchmark 'crypto-decrypt'...
Benchmark 'crypto-decrypt' started.
Iter   0 (209.082us): hits    0.000M/s (  0.000M/prod), drops    0.000M/s, total operations    0.000M/s
Iter   1 (154.618us): hits    0.000M/s (  0.000M/prod), drops    0.000M/s, total operations    0.000M/s
Iter   2 (-36.658us): hits    0.000M/s (  0.000M/prod), drops    0.000M/s, total operations    0.000M/s

This "void *input" arg shadowed the global variable.

> +{
> +	LIBBPF_OPTS(bpf_test_run_opts, opts,
> +		.repeat = 64,
> +		.data_in = input,
> +		.data_size_in = args.crypto_len,
> +	);
> +
> +	while (true)
> +		(void)bpf_prog_test_run_opts(ctx.pfd, &opts);
> +	return NULL;
> +}