perf trace: Avoid early exit due to running SIGCHLD handler before it makes sense to

When running 'perf trace' with an BPF object like:

  # perf trace -e openat,tools/perf/examples/bpf/hello.c

the event parsing eventually calls llvm__get_kbuild_opts() that runs a
script and that ends up with SIGCHLD delivered to the 'perf trace'
handler, which assumes the workload process is done and quits 'perf
trace'.

Move the SIGCHLD handler setup directly to trace__run(), where the event
is parsed and the object is already compiled.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Christy Lee <christyc.y.lee@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Michael Petlan <mpetlan@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lore.kernel.org/lkml/20220106222030.227499-1-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
index 0b52e08e558e3c971c48a1bd56ab0ae8b2e9a2ab..ef94388e8323436a87f8be7f10443c005b4a233d 100644 (file)
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -3925,6 +3925,7 @@ static int trace__run(struct trace *trace, int argc, const char **argv)
        bool draining = false;
 
        trace->live = true;
+       signal(SIGCHLD, sig_handler);
 
        if (!trace->raw_augmented_syscalls) {
                if (trace->trace_syscalls && trace__add_syscall_newtp(trace))
@@ -4873,7 +4874,6 @@ int cmd_trace(int argc, const char **argv)
 
        signal(SIGSEGV, sighandler_dump_stack);
        signal(SIGFPE, sighandler_dump_stack);
-       signal(SIGCHLD, sig_handler);
        signal(SIGINT, sig_handler);
 
        trace.evlist = evlist__new();