perf: Make perf_event_output() propagate the output() return

For the original mode of operation it isn't needed, since we report back
errors via PERF_RECORD_LOST records in the ring buffer, but for use in
bpf_perf_event_output() it is convenient to return the errors, basically
-ENOSPC.

Currently bpf_perf_event_output() returns an error indication, the last
thing it does, which is to push it to the ring buffer is that can fail
and if so, this failure won't be reported back to its users, fix it.

Reported-by: Jamal Hadi Salim <jhs@mojatatu.com>
Tested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/r/20190118150938.GN5823@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/examples/bpf/etcsnoop.c

@@ -49,11 +49,11 @@ int syscall_enter(syscall)(struct syscall_enter_##syscall##_args *args)				\
 						      args->filename_ptr); 			\
 	if (__builtin_memcmp(augmented_args.filename.value, etc, 4) != 0)			\
 		return 0;									\
-	perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, 			\
-			  &augmented_args, 							\
-			  (sizeof(augmented_args) - sizeof(augmented_args.filename.value) +	\
-			   augmented_args.filename.size));					\
-	return 0;										\
+	/* If perf_event_output fails, return non-zero so that it gets recorded unaugmented */	\
+	return perf_event_output(args, &__augmented_syscalls__, BPF_F_CURRENT_CPU, 		\
+				 &augmented_args,						\
+				 (sizeof(augmented_args) - sizeof(augmented_args.filename.value) + \
+				 augmented_args.filename.size));				\
 }
 
 struct syscall_enter_openat_args {