Re: [External] Re: [PATCH bpf-next v2 2/2] selftests/bpf: add test case for bpf_map_lookup_percpu_elem

[Andrii Nakryiko]

On Wed, May 11, 2022 at 8:58 PM Feng Zhou <zhoufeng.zf@xxxxxxxxxxxxx> wrote:
>
> 在 2022/5/12 上午11:34, Andrii Nakryiko 写道:
> > On Wed, May 11, 2022 at 2:39 AM Feng zhou <zhoufeng.zf@xxxxxxxxxxxxx> wrote:
> >> From: Feng Zhou <zhoufeng.zf@xxxxxxxxxxxxx>
> >>
> >> test_progs:
> >> Tests new ebpf helpers bpf_map_lookup_percpu_elem.
> >>
> >> Signed-off-by: Feng Zhou <zhoufeng.zf@xxxxxxxxxxxxx>
> >> ---
> >>   .../bpf/prog_tests/map_lookup_percpu_elem.c   | 46 ++++++++++++++++
> >>   .../bpf/progs/test_map_lookup_percpu_elem.c   | 54 +++++++++++++++++++
> >>   2 files changed, 100 insertions(+)
> >>   create mode 100644 tools/testing/selftests/bpf/prog_tests/map_lookup_percpu_elem.c
> >>   create mode 100644 tools/testing/selftests/bpf/progs/test_map_lookup_percpu_elem.c
> >>
> >> diff --git a/tools/testing/selftests/bpf/prog_tests/map_lookup_percpu_elem.c b/tools/testing/selftests/bpf/prog_tests/map_lookup_percpu_elem.c
> >> new file mode 100644
> >> index 000000000000..58b24c2112b0
> >> --- /dev/null
> >> +++ b/tools/testing/selftests/bpf/prog_tests/map_lookup_percpu_elem.c
> >> @@ -0,0 +1,46 @@
> >> +// SPDX-License-Identifier: GPL-2.0
> >> +// Copyright (c) 2022 Bytedance
> > /* */ instead of //
>
> Ok, I will do. Thanks.
>
>
> >
> >> +
> >> +#include <test_progs.h>
> >> +
> >> +#include "test_map_lookup_percpu_elem.skel.h"
> >> +
> >> +#define TEST_VALUE  1
> >> +
> >> +void test_map_lookup_percpu_elem(void)
> >> +{
> >> +       struct test_map_lookup_percpu_elem *skel;
> >> +       int key = 0, ret;
> >> +       int nr_cpus = sysconf(_SC_NPROCESSORS_ONLN);
> > I think this is actually wrong and will break selftests on systems
> > with offline CPUs. Please use libbpf_num_possible_cpus() instead.
>
>
> Ok, I will do. Thanks.
>
>
> >
> >> +       int *buf;
> >> +
> >> +       buf = (int *)malloc(nr_cpus*sizeof(int));
> >> +       if (!ASSERT_OK_PTR(buf, "malloc"))
> >> +               return;
> >> +       memset(buf, 0, nr_cpus*sizeof(int));
> > this is wrong, kernel expects to have roundup(sz, 8) per each CPU,
> > while you have just 4 bytes per each element
> >
> > please also have spaces around multiplication operator here and above
>
>
> Ok, I will use 8 bytes for key and val. Thanks.
>
>
> >> +       buf[0] = TEST_VALUE;
> >> +
> >> +       skel = test_map_lookup_percpu_elem__open_and_load();
> >> +       if (!ASSERT_OK_PTR(skel, "test_map_lookup_percpu_elem__open_and_load"))
> >> +               return;
> > buf leaking here
>
>
> Yes, sorry for my negligence.
>
>
> >
> >> +       ret = test_map_lookup_percpu_elem__attach(skel);
> >> +       ASSERT_OK(ret, "test_map_lookup_percpu_elem__attach");
> >> +
> >> +       ret = bpf_map_update_elem(bpf_map__fd(skel->maps.percpu_array_map), &key, buf, 0);
> >> +       ASSERT_OK(ret, "percpu_array_map update");
> >> +
> >> +       ret = bpf_map_update_elem(bpf_map__fd(skel->maps.percpu_hash_map), &key, buf, 0);
> >> +       ASSERT_OK(ret, "percpu_hash_map update");
> >> +
> >> +       ret = bpf_map_update_elem(bpf_map__fd(skel->maps.percpu_lru_hash_map), &key, buf, 0);
> >> +       ASSERT_OK(ret, "percpu_lru_hash_map update");
> >> +
> >> +       syscall(__NR_getuid);
> >> +
> >> +       ret = skel->bss->percpu_array_elem_val == TEST_VALUE &&
> >> +             skel->bss->percpu_hash_elem_val == TEST_VALUE &&
> >> +             skel->bss->percpu_lru_hash_elem_val == TEST_VALUE;
> >> +       ASSERT_OK(!ret, "bpf_map_lookup_percpu_elem success");
> > this would be better done as three separate ASSERT_EQ(), combining
> > into opaque true/false isn't helpful if something breaks
>
>
> Good suggestion.
>
>
> >
> >> +
> >> +       test_map_lookup_percpu_elem__destroy(skel);
> >> +}
> >> diff --git a/tools/testing/selftests/bpf/progs/test_map_lookup_percpu_elem.c b/tools/testing/selftests/bpf/progs/test_map_lookup_percpu_elem.c
> >> new file mode 100644
> >> index 000000000000..5d4ef86cbf48
> >> --- /dev/null
> >> +++ b/tools/testing/selftests/bpf/progs/test_map_lookup_percpu_elem.c
> >> @@ -0,0 +1,54 @@
> >> +// SPDX-License-Identifier: GPL-2.0
> >> +// Copyright (c) 2022 Bytedance
> > /* */ instead of //
>
>
> Ok, I will do. Thanks.
>
>
> >
> >> +
> >> +#include "vmlinux.h"
> >> +#include <bpf/bpf_helpers.h>
> >> +
> >> +int percpu_array_elem_val = 0;
> >> +int percpu_hash_elem_val = 0;
> >> +int percpu_lru_hash_elem_val = 0;
> >> +
> >> +struct {
> >> +       __uint(type, BPF_MAP_TYPE_PERCPU_ARRAY);
> >> +       __uint(max_entries, 1);
> >> +       __type(key, __u32);
> >> +       __type(value, __u32);
> >> +} percpu_array_map SEC(".maps");
> >> +
> >> +struct {
> >> +       __uint(type, BPF_MAP_TYPE_PERCPU_HASH);
> >> +       __uint(max_entries, 1);
> >> +       __type(key, __u32);
> >> +       __type(value, __u32);
> >> +} percpu_hash_map SEC(".maps");
> >> +
> >> +struct {
> >> +       __uint(type, BPF_MAP_TYPE_LRU_PERCPU_HASH);
> >> +       __uint(max_entries, 1);
> >> +       __type(key, __u32);
> >> +       __type(value, __u32);
> >> +} percpu_lru_hash_map SEC(".maps");
> >> +
> >> +SEC("tp/syscalls/sys_enter_getuid")
> >> +int sysenter_getuid(const void *ctx)
> >> +{
> >> +       __u32 key = 0;
> >> +       __u32 cpu = 0;
> >> +       __u32 *value;
> >> +
> >> +       value = bpf_map_lookup_percpu_elem(&percpu_array_map, &key, cpu);
> >> +       if (value)
> >> +               percpu_array_elem_val = *value;
> >> +
> >> +       value = bpf_map_lookup_percpu_elem(&percpu_hash_map, &key, cpu);
> >> +       if (value)
> >> +               percpu_hash_elem_val = *value;
> >> +
> >> +       value = bpf_map_lookup_percpu_elem(&percpu_lru_hash_map, &key, cpu);
> >> +       if (value)
> >> +               percpu_lru_hash_elem_val = *value;
> >> +
> > if the test happens to run on CPU 0 then the test doesn't really test
> > much. It would be more interesting to have a bpf_loop() iteration that
> > would fetch values on each possible CPU instead and do something with
> > it.
>
>
> Good suggestion. I check the code and find no bpf helper function to get
> possible CPU nums.
>
> I think for the test function, read cpu0 elem value correctly should be
> considered to be no problem.
>
> Or is it necessary to add a new helper function to get num_possible_cpus ?
>
>

You can pass number of CPUs from user-space to BPF program through
read-only variable (search for `const volatile` under progs/ for
examples)

> >
> >> +       return 0;
> >> +}
> >> +
> >> +char _license[] SEC("license") = "GPL";
> >> --
> >> 2.20.1
> >>
>