netdev - Re: [PATCH bpf-next 2/2] selftests/bpf: add tests for BPF_F_SHARE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <462d5ed2-727b-d932-a5f3-333c4b2c4806@iogearbox.net>
Date:   Tue, 29 Sep 2020 16:08:05 +0200
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Song Liu <songliubraving@...com>, netdev@...r.kernel.org,
        bpf@...r.kernel.org
Cc:     kernel-team@...com, ast@...nel.org, john.fastabend@...il.com,
        kpsingh@...omium.org
Subject: Re: [PATCH bpf-next 2/2] selftests/bpf: add tests for BPF_F_SHARE_PE

On 9/29/20 10:47 AM, Song Liu wrote:
> Add tests for perf event array with and without BPF_F_SHARE_PE.
> 
> Add a perf event to array via fd mfd. Without BPF_F_SHARE_PE, the perf
> event is removed when mfd is closed. With BPF_F_SHARE_PE, the perf event
> is removed when the map is freed.
> 
> Signed-off-by: Song Liu <songliubraving@...com>
[...]
> +static void test_one_map(struct bpf_map *map, struct bpf_program *prog,
> +			 bool has_share_pe)
> +{
> +	int err, key = 0, pfd = -1, mfd = bpf_map__fd(map);
> +	DECLARE_LIBBPF_OPTS(bpf_test_run_opts, opts);
> +	struct perf_event_attr attr = {
> +		.size = sizeof(struct perf_event_attr),
> +		.type = PERF_TYPE_SOFTWARE,
> +		.config = PERF_COUNT_SW_CPU_CLOCK,
> +	};
> +
> +	pfd = syscall(__NR_perf_event_open, &attr, 0 /* pid */,
> +		      -1 /* cpu 0 */, -1 /* group id */, 0 /* flags */);
> +	if (CHECK(pfd < 0, "perf_event_open", "failed\n"))
> +		return;
> +
> +	err = bpf_map_update_elem(mfd, &key, &pfd, BPF_ANY);
> +	if (CHECK(err < 0, "bpf_map_update_elem", "failed\n"))
> +		goto cleanup;
> +
> +	err = bpf_prog_test_run_opts(bpf_program__fd(prog), &opts);
> +	if (CHECK(err < 0, "bpf_prog_test_run_opts", "failed\n"))
> +		goto cleanup;
> +	if (CHECK(opts.retval != 0, "bpf_perf_event_read_value",
> +		  "failed with %d\n", opts.retval))
> +		goto cleanup;
> +
> +	/* closing mfd, prog still holds a reference on map */
> +	close(mfd);
> +
> +	err = bpf_prog_test_run_opts(bpf_program__fd(prog), &opts);
> +	if (CHECK(err < 0, "bpf_prog_test_run_opts", "failed\n"))
> +		goto cleanup;
> +
> +	if (has_share_pe) {
> +		CHECK(opts.retval != 0, "bpf_perf_event_read_value",
> +		      "failed with %d\n", opts.retval);
> +	} else {
> +		CHECK(opts.retval != -ENOENT, "bpf_perf_event_read_value",
> +		      "should have failed with %d, but got %d\n", -ENOENT,
> +		      opts.retval);
> +	}
> +
> +cleanup:
> +	close(pfd);

Why holding pfd until the end? The map should already hold a ref after update.
So you should be able to just do ...

err = bpf_map_update_elem(mfd, &key, &pfd, BPF_ANY);
close(pfd);

... and simplify cleanup.

> +}
> +
> +void test_perf_event_share(void)
> +{
> +	struct test_perf_event_share *skel;
> +
> +	skel = test_perf_event_share__open_and_load();
> +	if (CHECK(!skel, "skel_open", "failed to open skeleton\n"))
> +		return;
> +
> +	test_one_map(skel->maps.array_1, skel->progs.read_array_1, false);
> +	test_one_map(skel->maps.array_2, skel->progs.read_array_2, true);
> +
> +	test_perf_event_share__destroy(skel);
> +}