Re: SIGSTOP/CONT behaviour

[Alan Cox]

> If I send a SIGSTOP to another process, I can NOT rely on him not
> running anymore after the kill syscall returns. Is that right? I think

Correct in specific narrow windows.  A signal is an asynchronous event

> Additionally I could imagine there isn't even a guarantee that the
> process won't execute userland code anymore in the case of SMP?

Yep.

> (What I wanted to do is stop several file serving daemons
> (ftp,samba,netatalk) from fiddling with the filesystem while I traverse
> through a file tree. I need to make sure I don't get files twice because
> they have been moved while traversing and so on. I would not mind that
> much about file *contents* changing, though. I'd welcome other
> suggestions that don't imply changing the source code of these servers.)

I guess if you arent doing it often you could kill them then check they
are in stopped state but thats icky. 

Alan