Re: Killing clones
Richard Gooch (rgooch@atnf.CSIRO.AU)
Sun, 10 Aug 1997 23:08:38 +1000
Ingo Molnar writes:
>
> On Sun, 10 Aug 1997, Richard Gooch wrote:
>
> > Hi, all. I make heavy use of the clone syscall in a number of
> > applications, and everything works well except when the parent process
> > dies (say from a SEGV or a call to abort). The problem is the cloned
> > processes hang around until I manually kill them. I note that there
> > seems to be a mechanism to choose which signal is sent to the parent
> > if a child dies (CSIGNAL mask), but not the other way around.
> > Is there a solution to this problem? I tried cloning the PID but that
> > stuffs up sending signals to the clones.
>
> LinuxThreads uses a 'manager thread' (which is not the main thread) that
> can do exactly this kind of cleanup stuff.
But how does this manager know when the parent has died? Does it
periodically wake up and check if its ppid == 1? Yuk.
> the real solution will be Posix.1b RT signal handling ...
Yeah, well, that would be nice. But until then...
Regards,
Richard....