| Message ID | 20200214234146.2910011-1-songliubraving@fb.com |
|---|---|
| State | Accepted |
| Delegated to: | BPF Maintainers |
| Headers | show |
| Series | [bpf-next] bpf: allow bpf_perf_event_read_value in all BPF programs | expand |
On 2/15/20 12:41 AM, Song Liu wrote: > bpf_perf_event_read_value() is NMI safe. Enable it for all BPF programs. > This can be used in fentry/fexit to profile BPF program and individual > kernel function with hardware counters. > > Cc: Yonghong Song <yhs@fb.com> > Signed-off-by: Song Liu <songliubraving@fb.com> And ironically, we had that already for perf_event_read. Applied, thanks!
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c index 19e793aa441a..4ddd5ac46094 100644 --- a/kernel/trace/bpf_trace.c +++ b/kernel/trace/bpf_trace.c @@ -843,6 +843,8 @@ tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) return &bpf_send_signal_proto; case BPF_FUNC_send_signal_thread: return &bpf_send_signal_thread_proto; + case BPF_FUNC_perf_event_read_value: + return &bpf_perf_event_read_value_proto; default: return NULL; } @@ -858,8 +860,6 @@ kprobe_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) return &bpf_get_stackid_proto; case BPF_FUNC_get_stack: return &bpf_get_stack_proto; - case BPF_FUNC_perf_event_read_value: - return &bpf_perf_event_read_value_proto; #ifdef CONFIG_BPF_KPROBE_OVERRIDE case BPF_FUNC_override_return: return &bpf_override_return_proto;
bpf_perf_event_read_value() is NMI safe. Enable it for all BPF programs. This can be used in fentry/fexit to profile BPF program and individual kernel function with hardware counters. Cc: Yonghong Song <yhs@fb.com> Signed-off-by: Song Liu <songliubraving@fb.com> --- kernel/trace/bpf_trace.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)