[lldb] [Process/FreeBSD] Do not send SIGSTOP to stopped process

[mgorny]

Summary

Do not send SIGSTOP when requested to halt a process that's already
stopped. This results in the signal being queued for delivery once
the process is resumed, and unexpectedly stopping it again.

This is necessary for non-stop protocol patches to land.

Sponsored by: The FreeBSD Foundation

[labath]

This still leaves the possibility of a race where the process stops on its own while you're in the process of sending the signal, right?

I believe that on linux we handling this by checking the stopped events: If the application stopped due to a signal from us (siginfo->si_pid == getpid()) *and* we're not currently trying to stop a process, then we transparently resume it. Maybe you need to do the same here?

[labath]

(Actually, I don't think we do it for the main Halt operation (though maybe we should) -- only for the SIGSTOPs send to individual threads when we're performing an all-stop.)

lldb/source/Plugins/Process/FreeBSD/NativeProcessFreeBSD.cpp

Status NativeProcessFreeBSD::Resume(const ResumeActionList &resume_actions) {
   if (ret.Success())
     SetState(eStateRunning, true);
   return ret;
 }
 
 Status NativeProcessFreeBSD::Halt() {
   Status error;
 
+  // Do not try to stop a process that's already stopped, this may cause
+  // the SIGSTOP to get queued and stop the process again once resumed.
+  if (StateIsStoppedState(m_state, false))
+    return error;
   if (kill(GetID(), SIGSTOP) != 0)
     error.SetErrorToErrno();
   return error;
 }
 
 Status NativeProcessFreeBSD::Detach() {
   Status error;