UNIX for Dummies Questions & Answers

what happens when user does kill 1
and kill 0

does the system shutdown

This depend on the specific OS and version. These days, almost all unix versions protect a few processes from signals. But roughly in 1987, I did a kill 1 to a very early version of HP-UX. Note that I did not use -9! init exited and the OS sent a message to the console suggesting a reboot. There was no way to restart init. I was going to reboot anyway. And on BSD 4.2 you would edit a file called /etc/ttys which was sort of a predecessor to today's inittab. To tell init to reread the file, you would do "kill -1 1". My boss got that backwards and did "kill 1 -1" and she said "the system crashed". I did not get to view the exact sequence. Back then there was no process 0. I assume that stuff like this is now safe. A few years ago I verified a claim by HP that on HP-UX, init is immune to signals... even kill -9. But I haven't gone around and tested this extensively.

You first..

I have tried it on solaris 10

kill -9 1
It says "Not Owner" even when i tried to execute it from root user.

I see the ppid of init as 0
bash-3.00# ps -ef|grep init
root 1 0 0 Jul 28 ? 7:10 /sbin/init

and when i did
kill -9 1
it shows a message "Killed"
and exits from the bash prompt

You seem to be claiming two very different behaviors for "kill -9 1" on Solaris 10. Which is it? Or what was the difference between your two tests? I'm confused.

sorry for the confusion

Below is the actual test and result

kill -9 1
It says "Not Owner" even when i tried to execute it from root user.

kill -9 0
it shows a message "Killed"
and exits from the bash prompt

I gained some info on this the hard way back in 1999. I was working on the helpdesk and we had scripts to run as root to create new users. I was on the main DNS and NIS master for a 5000 person division of the company and the user creating script hung up. I did a "jobs" and found my script was job 1. So then I did "kill 1" forgetting that to specify a job number instead of PID you need to make it %1.

My session closed immediately and I couldn't log back in. I quickly headed toward the Unix admin's area and found one of the senior admins walking at a near run toward the server room. I hollered "Hey Dan", and without breaking stride he yelled back "WHAT DID YOU DO???"

He got it fixed fairly quickly. In the elevator on the way back down to our offices I said "Consider it a test of your slave server setup." He didn't think that was too funny.

Thus I can tell you for sure you can kill process 1 as root on Solaris 2.6 and 7 and it will crash your box. It will trigger a reboot if your auto-boot? in the OBP is set to true I believe. I'm pretty sure the protection to prevent that was included in Solaris 8, but it may have been 9.

By the way, with modern versions of unix you have no way to even try to signal process 0. The kill system call uses a process of -1 and 0 to mean two collections of processes. This is documented of the kill(2) man page and it's required by posix. This change occurred roughly about the same time as versions of unix had a process 0. I'm not sure if any version of unix ever had both a process 0 and a way to signal it.