netdev - Re: [PATCH v4 bpf-next 7/9] selftests/bpf: switch test to new attach_perf

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: <6ecaf9a0-6af1-80e9-0ba5-e1faf37e5228@fb.com>
Date:   Mon, 1 Jul 2019 23:05:17 +0000
From:   Yonghong Song <yhs@...com>
To:     Andrii Nakryiko <andrii.nakryiko@...il.com>
CC:     Andrii Nakryiko <andriin@...com>,
        "bpf@...r.kernel.org" <bpf@...r.kernel.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Alexei Starovoitov <ast@...com>,
        "daniel@...earbox.net" <daniel@...earbox.net>,
        Kernel Team <Kernel-team@...com>,
        "sdf@...ichev.me" <sdf@...ichev.me>,
        Song Liu <songliubraving@...com>
Subject: Re: [PATCH v4 bpf-next 7/9] selftests/bpf: switch test to new
 attach_perf_event API



On 7/1/19 3:32 PM, Andrii Nakryiko wrote:
> On Mon, Jul 1, 2019 at 10:16 AM Yonghong Song <yhs@...com> wrote:
>>
>>
>>
>> On 6/28/19 8:49 PM, Andrii Nakryiko wrote:
>>> Use new bpf_program__attach_perf_event() in test previously relying on
>>> direct ioctl manipulations.
>>>
>>> Signed-off-by: Andrii Nakryiko <andriin@...com>
>>> Reviewed-by: Stanislav Fomichev <sdf@...gle.com>
>>> Acked-by: Song Liu <songliubraving@...com>
>>> ---
>>>    .../bpf/prog_tests/stacktrace_build_id_nmi.c  | 31 +++++++++----------
>>>    1 file changed, 15 insertions(+), 16 deletions(-)
>>>
>>> diff --git a/tools/testing/selftests/bpf/prog_tests/stacktrace_build_id_nmi.c b/tools/testing/selftests/bpf/prog_tests/stacktrace_build_id_nmi.c
>>> index 1c1a2f75f3d8..9557b7dfb782 100644
>>> --- a/tools/testing/selftests/bpf/prog_tests/stacktrace_build_id_nmi.c
>>> +++ b/tools/testing/selftests/bpf/prog_tests/stacktrace_build_id_nmi.c
>>> @@ -17,6 +17,7 @@ static __u64 read_perf_max_sample_freq(void)
>>>    void test_stacktrace_build_id_nmi(void)
>>>    {
>>>        int control_map_fd, stackid_hmap_fd, stackmap_fd, stack_amap_fd;
>>> +     const char *prog_name = "tracepoint/random/urandom_read";
>>>        const char *file = "./test_stacktrace_build_id.o";
>>>        int err, pmu_fd, prog_fd;
>>>        struct perf_event_attr attr = {
>>> @@ -25,7 +26,9 @@ void test_stacktrace_build_id_nmi(void)
>>>                .config = PERF_COUNT_HW_CPU_CYCLES,
>>>        };
>>>        __u32 key, previous_key, val, duration = 0;
>>> +     struct bpf_program *prog;
>>>        struct bpf_object *obj;
>>> +     struct bpf_link *link;
>>>        char buf[256];
>>>        int i, j;
>>>        struct bpf_stack_build_id id_offs[PERF_MAX_STACK_DEPTH];
>>> @@ -39,6 +42,10 @@ void test_stacktrace_build_id_nmi(void)
>>>        if (CHECK(err, "prog_load", "err %d errno %d\n", err, errno))
>>>                return;
>>>
>>> +     prog = bpf_object__find_program_by_title(obj, prog_name);
>>> +     if (CHECK(!prog, "find_prog", "prog '%s' not found\n", prog_name))
>>> +             goto close_prog;
>>> +
>>>        pmu_fd = syscall(__NR_perf_event_open, &attr, -1 /* pid */,
>>>                         0 /* cpu 0 */, -1 /* group id */,
>>>                         0 /* flags */);
>>> @@ -47,15 +54,12 @@ void test_stacktrace_build_id_nmi(void)
>>>                  pmu_fd, errno))
>>>                goto close_prog;
>>>
>>> -     err = ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE, 0);
>>> -     if (CHECK(err, "perf_event_ioc_enable", "err %d errno %d\n",
>>> -               err, errno))
>>> -             goto close_pmu;
>>> -
>>> -     err = ioctl(pmu_fd, PERF_EVENT_IOC_SET_BPF, prog_fd);
>>> -     if (CHECK(err, "perf_event_ioc_set_bpf", "err %d errno %d\n",
>>> -               err, errno))
>>> -             goto disable_pmu;
>>> +     link = bpf_program__attach_perf_event(prog, pmu_fd);
>>> +     if (CHECK(IS_ERR(link), "attach_perf_event",
>>> +               "err %ld\n", PTR_ERR(link))) {
>>> +             close(pmu_fd);
>>> +             goto close_prog;
>>> +     }
>>>
>>>        /* find map fds */
>>>        control_map_fd = bpf_find_map(__func__, obj, "control_map");
>>> @@ -134,8 +138,7 @@ void test_stacktrace_build_id_nmi(void)
>>>         * try it one more time.
>>>         */
>>>        if (build_id_matches < 1 && retry--) {
>>> -             ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
>>> -             close(pmu_fd);
>>> +             bpf_link__destroy(link);
>>>                bpf_object__close(obj);
>>>                printf("%s:WARN:Didn't find expected build ID from the map, retrying\n",
>>>                       __func__);
>>> @@ -154,11 +157,7 @@ void test_stacktrace_build_id_nmi(void)
>>>         */
>>>
>>>    disable_pmu:
>>> -     ioctl(pmu_fd, PERF_EVENT_IOC_DISABLE);
>>> -
>>> -close_pmu:
>>> -     close(pmu_fd);
>>> -
>>> +     bpf_link__destroy(link);
>>
>> There is a problem in bpf_link__destroy(link).
>> The "link = bpf_program__attach_perf_event(prog, pmu_fd)"
>> may be an error pointer (IS_ERR(link) is true), in which
>> case, link should be reset to NULL and then call
>> bpf_link__destroy(link). Otherwise, the program may
>> segfault or function incorrectly.
> 
> Not really, if bpf_program__attach_perf_event fails and IS_ERR(link)
> is true, we'll close pmu_fd explicitly and `goto close_prog` bypassing
> bpf_link__destroy. `goto disable_pmu` is done only after we
> successfully established attached link.
> 
> So unless I still miss something, I think this will work reliably.

Double checked again. You are correct. We do not have issues here.

> 
>>
>>>    close_prog:
>>>        bpf_object__close(obj);
>>>    }
>>>