From 792caccc4526bb489e054f9ab61d7c024b15dea2 Mon Sep 17 00:00:00 2001 From: Song Liu Date: Wed, 30 Sep 2020 15:49:26 -0700 Subject: [PATCH 1/2] bpf: Introduce BPF_F_PRESERVE_ELEMS for perf event array Currently, perf event in perf event array is removed from the array when the map fd used to add the event is closed. This behavior makes it difficult to the share perf events with perf event array. Introduce perf event map that keeps the perf event open with a new flag BPF_F_PRESERVE_ELEMS. With this flag set, perf events in the array are not removed when the original map fd is closed. Instead, the perf event will stay in the map until 1) it is explicitly removed from the array; or 2) the array is freed. Signed-off-by: Song Liu Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20200930224927.1936644-2-songliubraving@fb.com --- include/uapi/linux/bpf.h | 3 +++ kernel/bpf/arraymap.c | 19 +++++++++++++++++-- tools/include/uapi/linux/bpf.h | 3 +++ 3 files changed, 23 insertions(+), 2 deletions(-) diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 1f17c6752deb..4f556cfcbfbe 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -414,6 +414,9 @@ enum { /* Enable memory-mapping BPF map */ BPF_F_MMAPABLE = (1U << 10), + +/* Share perf_event among processes */ + BPF_F_PRESERVE_ELEMS = (1U << 11), }; /* Flags for BPF_PROG_QUERY. */ diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c index e5fd31268ae0..bd777dd6f967 100644 --- a/kernel/bpf/arraymap.c +++ b/kernel/bpf/arraymap.c @@ -15,7 +15,8 @@ #include "map_in_map.h" #define ARRAY_CREATE_FLAG_MASK \ - (BPF_F_NUMA_NODE | BPF_F_MMAPABLE | BPF_F_ACCESS_MASK) + (BPF_F_NUMA_NODE | BPF_F_MMAPABLE | BPF_F_ACCESS_MASK | \ + BPF_F_PRESERVE_ELEMS) static void bpf_array_free_percpu(struct bpf_array *array) { @@ -64,6 +65,10 @@ int array_map_alloc_check(union bpf_attr *attr) attr->map_flags & BPF_F_MMAPABLE) return -EINVAL; + if (attr->map_type != BPF_MAP_TYPE_PERF_EVENT_ARRAY && + attr->map_flags & BPF_F_PRESERVE_ELEMS) + return -EINVAL; + if (attr->value_size > KMALLOC_MAX_SIZE) /* if value_size is bigger, the user space won't be able to * access the elements. @@ -1134,6 +1139,9 @@ static void perf_event_fd_array_release(struct bpf_map *map, struct bpf_event_entry *ee; int i; + if (map->map_flags & BPF_F_PRESERVE_ELEMS) + return; + rcu_read_lock(); for (i = 0; i < array->map.max_entries; i++) { ee = READ_ONCE(array->ptrs[i]); @@ -1143,12 +1151,19 @@ static void perf_event_fd_array_release(struct bpf_map *map, rcu_read_unlock(); } +static void perf_event_fd_array_map_free(struct bpf_map *map) +{ + if (map->map_flags & BPF_F_PRESERVE_ELEMS) + bpf_fd_array_map_clear(map); + fd_array_map_free(map); +} + static int perf_event_array_map_btf_id; const struct bpf_map_ops perf_event_array_map_ops = { .map_meta_equal = bpf_map_meta_equal, .map_alloc_check = fd_array_map_alloc_check, .map_alloc = array_map_alloc, - .map_free = fd_array_map_free, + .map_free = perf_event_fd_array_map_free, .map_get_next_key = array_map_get_next_key, .map_lookup_elem = fd_array_map_lookup_elem, .map_delete_elem = fd_array_map_delete_elem, diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 1f17c6752deb..4f556cfcbfbe 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -414,6 +414,9 @@ enum { /* Enable memory-mapping BPF map */ BPF_F_MMAPABLE = (1U << 10), + +/* Share perf_event among processes */ + BPF_F_PRESERVE_ELEMS = (1U << 11), }; /* Flags for BPF_PROG_QUERY. */ From d6b42068412bd17d94b3b7b51d3787120ba8c738 Mon Sep 17 00:00:00 2001 From: Song Liu Date: Wed, 30 Sep 2020 15:49:27 -0700 Subject: [PATCH 2/2] selftests/bpf: Add tests for BPF_F_PRESERVE_ELEMS Add tests for perf event array with and without BPF_F_PRESERVE_ELEMS. Add a perf event to array via fd mfd. Without BPF_F_PRESERVE_ELEMS, the perf event is removed when mfd is closed. With BPF_F_PRESERVE_ELEMS, the perf event is removed when the map is freed. Signed-off-by: Song Liu Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20200930224927.1936644-3-songliubraving@fb.com --- .../bpf/prog_tests/pe_preserve_elems.c | 66 +++++++++++++++++++ .../bpf/progs/test_pe_preserve_elems.c | 38 +++++++++++ 2 files changed, 104 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/pe_preserve_elems.c create mode 100644 tools/testing/selftests/bpf/progs/test_pe_preserve_elems.c diff --git a/tools/testing/selftests/bpf/prog_tests/pe_preserve_elems.c b/tools/testing/selftests/bpf/prog_tests/pe_preserve_elems.c new file mode 100644 index 000000000000..673d38395253 --- /dev/null +++ b/tools/testing/selftests/bpf/prog_tests/pe_preserve_elems.c @@ -0,0 +1,66 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright (c) 2019 Facebook */ +#include +#include +#include "test_pe_preserve_elems.skel.h" + +static int duration; + +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); + close(pfd); + if (CHECK(err < 0, "bpf_map_update_elem", "failed\n")) + return; + + err = bpf_prog_test_run_opts(bpf_program__fd(prog), &opts); + if (CHECK(err < 0, "bpf_prog_test_run_opts", "failed\n")) + return; + if (CHECK(opts.retval != 0, "bpf_perf_event_read_value", + "failed with %d\n", opts.retval)) + return; + + /* 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")) + return; + + 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); + } +} + +void test_pe_preserve_elems(void) +{ + struct test_pe_preserve_elems *skel; + + skel = test_pe_preserve_elems__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_pe_preserve_elems__destroy(skel); +} diff --git a/tools/testing/selftests/bpf/progs/test_pe_preserve_elems.c b/tools/testing/selftests/bpf/progs/test_pe_preserve_elems.c new file mode 100644 index 000000000000..fb22de7c365d --- /dev/null +++ b/tools/testing/selftests/bpf/progs/test_pe_preserve_elems.c @@ -0,0 +1,38 @@ +// SPDX-License-Identifier: GPL-2.0 +// Copyright (c) 2020 Facebook +#include +#include +#include + +struct { + __uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY); + __uint(max_entries, 1); + __uint(key_size, sizeof(int)); + __uint(value_size, sizeof(int)); +} array_1 SEC(".maps"); + +struct { + __uint(type, BPF_MAP_TYPE_PERF_EVENT_ARRAY); + __uint(max_entries, 1); + __uint(key_size, sizeof(int)); + __uint(value_size, sizeof(int)); + __uint(map_flags, BPF_F_PRESERVE_ELEMS); +} array_2 SEC(".maps"); + +SEC("raw_tp/sched_switch") +int BPF_PROG(read_array_1) +{ + struct bpf_perf_event_value val; + + return bpf_perf_event_read_value(&array_1, 0, &val, sizeof(val)); +} + +SEC("raw_tp/task_rename") +int BPF_PROG(read_array_2) +{ + struct bpf_perf_event_value val; + + return bpf_perf_event_read_value(&array_2, 0, &val, sizeof(val)); +} + +char LICENSE[] SEC("license") = "GPL";