Tests: Cleanup: notification: `assert()` that `app_pid` is set
authorFrancis Deslauriers <francis.deslauriers@efficios.com>
Wed, 12 Feb 2020 20:30:16 +0000 (15:30 -0500)
committerJérémie Galarneau <jeremie.galarneau@efficios.com>
Thu, 4 Mar 2021 16:59:19 +0000 (11:59 -0500)
According to the kill(2) manpage:
  If pid equals -1, then sig is sent to every process for which the
  calling  process  has  permission  to  send  signals,  except for
  process 1 (init),

Since our current default value is -1, if we call the
`{suspend,resume}_application()` functions before setting the `app_pid`
we end up sending a signal to all processes owned by the user which can
be pretty dramatic.

If we are about to call `kill()` with -1, it's a logic error so we can
assert().

Signed-off-by: Francis Deslauriers <francis.deslauriers@efficios.com>
Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Change-Id: I211f75d44e6cfdd2f2e35ef342b930eb683dbbe5
Depends-on: lttng-ust: I5a800fc92e588c2a6a0e26282b0ad5f31c044479

tests/regression/tools/notification/notification.c

index d3917b4784d0eccb421af7e38474bfd5db255639..631948cda93d4a440875a8146fb306d4147edf5a 100644 (file)
@@ -43,7 +43,7 @@
 
 int nb_args = 0;
 int named_pipe_args_start = 0;
-pid_t app_pid = -1;
+pid_t app_pid = 0;
 const char *app_state_file = NULL;
 
 static
@@ -156,6 +156,8 @@ int suspend_application(void)
        /*
         * Send SIGUSR1 to application instructing it to bypass tracepoint.
         */
+       assert(app_pid > 1);
+
        ret = kill(app_pid, SIGUSR1);
        if (ret) {
                fail("SIGUSR1 failed. errno %d", errno);
@@ -186,6 +188,8 @@ int resume_application(void)
                goto error;
        }
 
+       assert(app_pid > 1);
+
        ret = kill(app_pid, SIGUSR1);
        if (ret) {
                fail("SIGUSR1 failed. errno %d", errno);
This page took 0.025938 seconds and 4 git commands to generate.