perf trace: Print content of bpf-output event
authorWang Nan <wangnan0@huawei.com>
Fri, 26 Feb 2016 09:31:55 +0000 (09:31 +0000)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Fri, 26 Feb 2016 22:57:07 +0000 (19:57 -0300)
With this patch the contend of BPF output event is printed by
'perf trace'. For example:

 # ./perf trace -a --ev bpf-output/no-inherit,name=evt/ \
                   --ev ./test_bpf_trace.c/map:channel.event=evt/ \
                   usleep 100000
  ...
    1.787 ( 0.004 ms): usleep/3832 nanosleep(rqtp: 0x7ffc78b18980                                        ) ...
    1.787 (         ): evt:Raise a BPF event!..)
    1.788 (         ): perf_bpf_probe:func_begin:(ffffffff810e97d0))
  ...
  101.866 (87.038 ms): gmain/1654 poll(ufds: 0x7f57a80008c0, nfds: 2, timeout_msecs: 1000               ) ...
  101.866 (         ): evt:Raise a BPF event!..)
  101.867 (         ): perf_bpf_probe:func_end:(ffffffff810e97d0 <- ffffffff81796173))
  101.869 (100.087 ms): usleep/3832  ... [continued]: nanosleep()) = 0
  ...

 (There is an extra ')' at the end of several lines. However, it is
  another problem, unrelated to this commit.)

Where test_bpf_trace.c is:

  /************************ BEGIN **************************/
  #include <uapi/linux/bpf.h>
  struct bpf_map_def {
        unsigned int type;
        unsigned int key_size;
        unsigned int value_size;
        unsigned int max_entries;
  };
  #define SEC(NAME) __attribute__((section(NAME), used))
  static u64 (*ktime_get_ns)(void) =
        (void *)BPF_FUNC_ktime_get_ns;
  static int (*trace_printk)(const char *fmt, int fmt_size, ...) =
        (void *)BPF_FUNC_trace_printk;
  static int (*get_smp_processor_id)(void) =
        (void *)BPF_FUNC_get_smp_processor_id;
  static int (*perf_event_output)(void *, struct bpf_map_def *, int, void *, unsigned long) =
        (void *)BPF_FUNC_perf_event_output;

  struct bpf_map_def SEC("maps") channel = {
        .type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
        .key_size = sizeof(int),
        .value_size = sizeof(u32),
        .max_entries = __NR_CPUS__,
  };

  static inline int __attribute__((always_inline))
  func(void *ctx, int type)
  {
char output_str[] = "Raise a BPF event!";
char err_str[] = "BAD %d\n";
int err;

        err = perf_event_output(ctx, &channel, get_smp_processor_id(),
        &output_str, sizeof(output_str));
if (err)
trace_printk(err_str, sizeof(err_str), err);
        return 1;
  }
  SEC("func_begin=sys_nanosleep")
  int func_begin(void *ctx) {return func(ctx, 1);}
  SEC("func_end=sys_nanosleep%return")
  int func_end(void *ctx) { return func(ctx, 2);}
  char _license[] SEC("license") = "GPL";
  int _version SEC("version") = LINUX_VERSION_CODE;
  /************************* END ***************************/

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: pi3orama@163.com
Link: http://lkml.kernel.org/r/1456479154-136027-8-git-send-email-wangnan0@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/builtin-trace.c

index 254149ca5e1bcec8b0638a1d65b9ef786131aeb8..26a337f939d80e9e1d6de7bd3fdb2592ad4587ac 100644 (file)
@@ -2178,6 +2178,37 @@ out_dump:
        return 0;
 }
 
+static void bpf_output__printer(enum binary_printer_ops op,
+                               unsigned int val, void *extra)
+{
+       FILE *output = extra;
+       unsigned char ch = (unsigned char)val;
+
+       switch (op) {
+       case BINARY_PRINT_CHAR_DATA:
+               fprintf(output, "%c", isprint(ch) ? ch : '.');
+               break;
+       case BINARY_PRINT_DATA_BEGIN:
+       case BINARY_PRINT_LINE_BEGIN:
+       case BINARY_PRINT_ADDR:
+       case BINARY_PRINT_NUM_DATA:
+       case BINARY_PRINT_NUM_PAD:
+       case BINARY_PRINT_SEP:
+       case BINARY_PRINT_CHAR_PAD:
+       case BINARY_PRINT_LINE_END:
+       case BINARY_PRINT_DATA_END:
+       default:
+               break;
+       }
+}
+
+static void bpf_output__fprintf(struct trace *trace,
+                               struct perf_sample *sample)
+{
+       print_binary(sample->raw_data, sample->raw_size, 8,
+                    bpf_output__printer, trace->output);
+}
+
 static int trace__event_handler(struct trace *trace, struct perf_evsel *evsel,
                                union perf_event *event __maybe_unused,
                                struct perf_sample *sample)
@@ -2190,7 +2221,9 @@ static int trace__event_handler(struct trace *trace, struct perf_evsel *evsel,
 
        fprintf(trace->output, "%s:", evsel->name);
 
-       if (evsel->tp_format) {
+       if (perf_evsel__is_bpf_output(evsel)) {
+               bpf_output__fprintf(trace, sample);
+       } else if (evsel->tp_format) {
                event_format__fprintf(evsel->tp_format, sample->cpu,
                                      sample->raw_data, sample->raw_size,
                                      trace->output);