[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: networking in rcS



On Mon, 11 Jun 2001, Miquel van Smoorenburg wrote:

> > How's that?
> > 
> > Does init kill all running processes when going to single user? (Guess you
> > should know, since you wrote it ;-)
> 
> Yes. /etc/rc1.d/S20single

Aha.

Well, guess I learned something new here ;-)

> > > It sure is. No daemons run in single user mode, so remote access
> > > is impossible.
> > 
> > I'm not sure, but since most daemons spawn another instance of themselves
> > to actually handle an incoming connection, wouldn't you say that an
> > existing connection could stay alive when the "main" daemon gets killed?
> 
> *all* daemons are killed. There's just init running and some kernel
> threads. All other processes have been sent to the eternal bitfields.

Since there is something as killall5, they are. I just assumed 
start-stop-daemon was used to kill all running daemons, which seems to be 
wrong.

> > It shouldn't (and should go into zombie state), but I've seen situations
> > in which "kill -9" to a hung process simply doesn't work
> 
> I know of only 2 situations in which that can happen:
> 
> 1) kernel bug
> 2) process hung on an NFS mount

I've seen this happen to something else too -- not NFS, and too many
kernel versions to be a bug as well -- but I can't recall what it was.

Anyway, it doesn't matter. You're right, and I'm wrong ;-)

> > or where a
> > zombie process stayed alive for several minutes after its parent got
> > killed.
> 
> Zombies aren't alive. Zombies are dead. A zombie is a child process
> that *still has a parent* but which is dead - however the parent
> hasn't wait()ed for it to collect its exit status
> 
> A zombie without a parent doesn't exist. Unix 101

I won't bet on this one.

> > > Say you fscked up the box and bring it to single user mode to fix
> > > things, it is very handy (usually essential) to be able to use
> > > ftp to access debian packages, to save logfiles, what have you.
> > 
> > If it's really necessary (which, in this case it is), it's still possible
> > to bring up the network manually. I see no need to leave it up.
> 
> Yes - what if your box is diskless. Or, /usr is NFS mounted. I don't
> see a reason to explicitly bring the network down, it simply makes
> no sense.

After your arguments, I agree.

-- 
wouter dot verhelst at advalvas in belgium

Try does not exist. Believe that you will do it, else you will fail.

       -- Luke Skywalker,
       in the trilogy "The Jedi Academy", Kevin J. Anderson



Reply to: