selftests/bpf: convert test_get_stack_raw_tp to perf_buffer API
authorAndrii Nakryiko <andriin@fb.com>
Tue, 23 Jul 2019 21:34:41 +0000 (14:34 -0700)
committerAlexei Starovoitov <ast@kernel.org>
Tue, 23 Jul 2019 23:05:42 +0000 (16:05 -0700)
Convert test_get_stack_raw_tp test to new perf_buffer API.

Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
tools/testing/selftests/bpf/prog_tests/get_stack_raw_tp.c
tools/testing/selftests/bpf/progs/test_get_stack_rawtp.c

index c2a0a9d5591b4b5ea3436bfe490689c7756e50db..9d73a8f932aca00b24353ed624d7b37ca6799489 100644 (file)
@@ -1,8 +1,15 @@
 // SPDX-License-Identifier: GPL-2.0
+#define _GNU_SOURCE
+#include <pthread.h>
+#include <sched.h>
+#include <sys/socket.h>
 #include <test_progs.h>
 
 #define MAX_CNT_RAWTP  10ull
 #define MAX_STACK_RAWTP        100
+
+static int duration = 0;
+
 struct get_stack_trace_t {
        int pid;
        int kern_stack_size;
@@ -13,7 +20,7 @@ struct get_stack_trace_t {
        struct bpf_stack_build_id user_stack_buildid[MAX_STACK_RAWTP];
 };
 
-static int get_stack_print_output(void *data, int size)
+static void get_stack_print_output(void *ctx, int cpu, void *data, __u32 size)
 {
        bool good_kern_stack = false, good_user_stack = false;
        const char *nonjit_func = "___bpf_prog_run";
@@ -65,75 +72,76 @@ static int get_stack_print_output(void *data, int size)
                if (e->user_stack_size > 0 && e->user_stack_buildid_size > 0)
                        good_user_stack = true;
        }
-       if (!good_kern_stack || !good_user_stack)
-               return LIBBPF_PERF_EVENT_ERROR;
 
-       if (cnt == MAX_CNT_RAWTP)
-               return LIBBPF_PERF_EVENT_DONE;
-
-       return LIBBPF_PERF_EVENT_CONT;
+       if (!good_kern_stack)
+           CHECK(!good_kern_stack, "kern_stack", "corrupted kernel stack\n");
+       if (!good_user_stack)
+           CHECK(!good_user_stack, "user_stack", "corrupted user stack\n");
 }
 
 void test_get_stack_raw_tp(void)
 {
        const char *file = "./test_get_stack_rawtp.o";
-       int i, efd, err, prog_fd, pmu_fd, perfmap_fd;
-       struct perf_event_attr attr = {};
+       const char *prog_name = "raw_tracepoint/sys_enter";
+       int i, err, prog_fd, exp_cnt = MAX_CNT_RAWTP;
+       struct perf_buffer_opts pb_opts = {};
+       struct perf_buffer *pb = NULL;
+       struct bpf_link *link = NULL;
        struct timespec tv = {0, 10};
-       __u32 key = 0, duration = 0;
+       struct bpf_program *prog;
        struct bpf_object *obj;
+       struct bpf_map *map;
+       cpu_set_t cpu_set;
 
        err = bpf_prog_load(file, BPF_PROG_TYPE_RAW_TRACEPOINT, &obj, &prog_fd);
        if (CHECK(err, "prog_load raw tp", "err %d errno %d\n", err, errno))
                return;
 
-       efd = bpf_raw_tracepoint_open("sys_enter", prog_fd);
-       if (CHECK(efd < 0, "raw_tp_open", "err %d errno %d\n", efd, errno))
+       prog = bpf_object__find_program_by_title(obj, prog_name);
+       if (CHECK(!prog, "find_probe", "prog '%s' not found\n", prog_name))
                goto close_prog;
 
-       perfmap_fd = bpf_find_map(__func__, obj, "perfmap");
-       if (CHECK(perfmap_fd < 0, "bpf_find_map", "err %d errno %d\n",
-                 perfmap_fd, errno))
+       map = bpf_object__find_map_by_name(obj, "perfmap");
+       if (CHECK(!map, "bpf_find_map", "not found\n"))
                goto close_prog;
 
        err = load_kallsyms();
        if (CHECK(err < 0, "load_kallsyms", "err %d errno %d\n", err, errno))
                goto close_prog;
 
-       attr.sample_type = PERF_SAMPLE_RAW;
-       attr.type = PERF_TYPE_SOFTWARE;
-       attr.config = PERF_COUNT_SW_BPF_OUTPUT;
-       pmu_fd = syscall(__NR_perf_event_open, &attr, getpid()/*pid*/, -1/*cpu*/,
-                        -1/*group_fd*/, 0);
-       if (CHECK(pmu_fd < 0, "perf_event_open", "err %d errno %d\n", pmu_fd,
-                 errno))
+       CPU_ZERO(&cpu_set);
+       CPU_SET(0, &cpu_set);
+       err = pthread_setaffinity_np(pthread_self(), sizeof(cpu_set), &cpu_set);
+       if (CHECK(err, "set_affinity", "err %d, errno %d\n", err, errno))
                goto close_prog;
 
-       err = bpf_map_update_elem(perfmap_fd, &key, &pmu_fd, BPF_ANY);
-       if (CHECK(err < 0, "bpf_map_update_elem", "err %d errno %d\n", err,
-                 errno))
+       link = bpf_program__attach_raw_tracepoint(prog, "sys_enter");
+       if (CHECK(IS_ERR(link), "attach_raw_tp", "err %ld\n", PTR_ERR(link)))
                goto close_prog;
 
-       err = ioctl(pmu_fd, PERF_EVENT_IOC_ENABLE, 0);
-       if (CHECK(err < 0, "ioctl PERF_EVENT_IOC_ENABLE", "err %d errno %d\n",
-                 err, errno))
-               goto close_prog;
-
-       err = perf_event_mmap(pmu_fd);
-       if (CHECK(err < 0, "perf_event_mmap", "err %d errno %d\n", err, errno))
+       pb_opts.sample_cb = get_stack_print_output;
+       pb = perf_buffer__new(bpf_map__fd(map), 8, &pb_opts);
+       if (CHECK(IS_ERR(pb), "perf_buf__new", "err %ld\n", PTR_ERR(pb)))
                goto close_prog;
 
        /* trigger some syscall action */
        for (i = 0; i < MAX_CNT_RAWTP; i++)
                nanosleep(&tv, NULL);
 
-       err = perf_event_poller(pmu_fd, get_stack_print_output);
-       if (CHECK(err < 0, "perf_event_poller", "err %d errno %d\n", err, errno))
-               goto close_prog;
+       while (exp_cnt > 0) {
+               err = perf_buffer__poll(pb, 100);
+               if (err < 0 && CHECK(err < 0, "pb__poll", "err %d\n", err))
+                       goto close_prog;
+               exp_cnt -= err;
+       }
 
        goto close_prog_noerr;
 close_prog:
        error_cnt++;
 close_prog_noerr:
+       if (!IS_ERR_OR_NULL(link))
+               bpf_link__destroy(link);
+       if (!IS_ERR_OR_NULL(pb))
+               perf_buffer__free(pb);
        bpf_object__close(obj);
 }
index 33254b771384e4d2c20a180c1c55b167ec37ca91..f8ffa3f3d44bb184c9e8eb2f9c694b3b71b85e19 100644 (file)
@@ -55,7 +55,7 @@ struct {
        __type(value, raw_stack_trace_t);
 } rawdata_map SEC(".maps");
 
-SEC("tracepoint/raw_syscalls/sys_enter")
+SEC("raw_tracepoint/sys_enter")
 int bpf_prog1(void *ctx)
 {
        int max_len, max_buildid_len, usize, ksize, total_size;