Re: [PATCH bpf-next v3 5/5] selftests/bpf: Add some tests with new bpf_program__attach_sockmap() APIs

[Yonghong Song <yonghong.song@linux.dev>]

On 4/2/24 6:17 AM, Eduard Zingerman wrote:
> On Mon, 2024-03-25 at 19:22 -0700, Yonghong Song wrote:
>> Add a few more tests in sockmap_basic.c and sockmap_listen.c to
>> test bpf_link based APIs for SK_MSG and SK_SKB programs.
>> Link attach/detach/update are all tested.
>>
>> All tests are passed.
>>
>> Signed-off-by: Yonghong Song <yonghong.song@linux.dev>
>> ---
> Acked-by: Eduard Zingerman <eddyz87@gmail.com>
>
> As far as I understand patch #1 there are several error conditions
> that are not covered by these tests:
> - update link using program with wrong expected attach type;
> - bpf_program__attach_sockmap() _after_ bpf_prog_attach();
> - creation of a second link between same prog and map
>    (if I understand 'link && link != progs->msg_parser_link' checks in
>     sock_map_link_lookup() correctly).
>
> Are these important enough to have a dedicated test?

All these are missed. I can add them in the next revision.

>
> [...]
>
>> diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c b/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c
>> index 63fb2da7930a..b4b38f92b864 100644
>> --- a/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c
>> +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_basic.c
>> @@ -131,6 +131,40 @@ static void test_skmsg_helpers(enum bpf_map_type map_type)
>>   	test_skmsg_load_helpers__destroy(skel);
>>   }
>>   
>> +static void test_skmsg_helpers_with_link(enum bpf_map_type map_type)
>> +{
>> +	struct test_skmsg_load_helpers *skel;
>> +	struct bpf_link *link, *link2;
>> +	struct bpf_program *prog;
>> +	int err, map;
>> +
>> +	skel = test_skmsg_load_helpers__open_and_load();
>> +	if (!ASSERT_OK_PTR(skel, "test_skmsg_load_helpers__open_and_load"))
>> +		return;
>> +
>> +	prog = skel->progs.prog_msg_verdict;
>> +	map = bpf_map__fd(skel->maps.sock_map);
>> +
>> +	link = bpf_program__attach_sockmap(prog, map);
>> +	if (!ASSERT_OK_PTR(link, "bpf_program__attach_sockmap"))
>> +		goto out;
>> +
>> +	err = bpf_prog_attach(bpf_program__fd(prog), map, BPF_SK_MSG_VERDICT, 0);
>> +	if (!ASSERT_ERR(err, "bpf_prog_attach"))
>> +		goto out;
>> +
>> +	link2 = bpf_program__attach_sockmap(skel->progs.prog_msg_verdict_clone, map);
>> +	if (!ASSERT_ERR_PTR(link2, "bpf_program__attach_sockmap"))
>> +		goto out;
> Nit: should bpf_link__detach(link2) be called before 'goto out' ?

Yes, we should do this although it will auto detach when test_progs exists.

>
>> +
>> +	err = bpf_link__update_program(link, skel->progs.prog_msg_verdict_clone);
>> +	ASSERT_OK(err, "bpf_link__update_program");
>> +
>> +out:
>> +	bpf_link__detach(link);
>> +	test_skmsg_load_helpers__destroy(skel);
>> +}
>> +
>>   static void test_sockmap_update(enum bpf_map_type map_type)
>>   {
>>   	int err, prog, src;