[comp.unix.questions] Kill -9

frech@mwraaa.army.mil (Norman R. Frech CPLS) (10/17/90)

Is there a stronger kill than kill -9?  I had a process which attached
itself to my tape drive and then locked up.  I couldn't kill it no
matter what I tried.  This caused me great concern, backups are very
important to me and I couldn't access the tape drive with this process
on it.  I shut down the system (two of my file systems didn't unmount
because of the process and shutting down didn't kill the process either) 
and rebooted.

Any suggestions.

Norm Frech < frech@mwraaa.army.mil >

gwyn@smoke.BRL.MIL (Doug Gwyn) (10/17/90)

In article <24788@adm.BRL.MIL> frech@mwraaa.army.mil (Norman R. Frech CPLS) writes:
>Is there a stronger kill than kill -9?

No.

>I had a process which attached itself to my tape drive and then locked up.
>I couldn't kill it no matter what I tried.

Actually, the signal was probably posted to the process, but as the process
was sleeping in the kernel at a "negative" priority, which is a common
problem with some device drivers in some UNIX kernels, it never reached a
state in which it could be terminated.

Sometimes, playing with the "hung" device (e.g., rewinding the tape, or
turning off power to the controller) will cause an interrupt that may
suffice to get the device driver unwedged.

avg@hq.demos.su (Vadim G. Antonov) (10/17/90)

In article <14166@smoke.BRL.MIL> gwyn@smoke.BRL.MIL (Doug Gwyn) writes:
>In article <24788@adm.BRL.MIL> frech@mwraaa.army.mil (Norman R. Frech CPLS) writes:
>>I had a process which attached itself to my tape drive and then locked up.
>>I couldn't kill it no matter what I tried.
>Sometimes, playing with the "hung" device (e.g., rewinding the tape, or
>turning off power to the controller) will cause an interrupt that may
>suffice to get the device driver unwedged.

	The simplier method to get an interrupt from the tape controller
	is to load tape, move in forward for some seconds, press REWIND
	and quickly switch drive ON-LINE. When it'd reach the physical
	marker it'll generate interrupt. This trick works as a rule.
	(Anyway the better solution is to hack up driver to make it
	to do timeout-s on operations ;-).

	Vadim Antonov
	DEMOS, Moscow, USSR

cpcahil@virtech.uucp (Conor P. Cahill) (10/17/90)

In article <24788@adm.BRL.MIL> frech@mwraaa.army.mil (Norman R. Frech CPLS) writes:
>Is there a stronger kill than kill -9?  I had a process which attached
>itself to my tape drive and then locked up.  I couldn't kill it no
>matter what I tried.  This caused me great concern, backups are very

There is no signal "stronger" than a 9.  The problem that caused the process
to be un-killable is due to a bug in the device driver for the tape drive.
Complain to the manufacturer of the driver.

-- 
Conor P. Cahill            (703)430-9247        Virtual Technologies, Inc.,
uunet!virtech!cpcahil                           46030 Manekin Plaza, Suite 160
                                                Sterling, VA 22170 

esa@tglobe2.tollpost-globe (Esa K Viitala) (10/18/90)

In article <14166@smoke.BRL.MIL> gwyn@smoke.BRL.MIL (Doug Gwyn) writes:
  >In article <24788@adm.BRL.MIL> frech@mwraaa.army.mil (Norman R. Frech CPLS) writes:
  >>Is there a stronger kill than kill -9?
  >
  >No.
  >

Wrong. Reboot.
-- 
Esa Viitala
TOLLPOST-GLOBE A/S, Systems Department, PO Box 100, N6301 ]ndalsnes, Norway
Tel: (+47 72) 21211 / 264, Fax: (+47 72) 22161 ...!mcsun!nuug!tglobe2!esa

rbottin@atl.calstate.edu (10/27/90)

This message is a reply from rbottin

===============================================

No there is no more powerful kill than -9. There is no real
hierarchy implied by the numbering - just different reasons
for sending a message to a process. Its just that it is
either impossible or very bad practice for a program to catch the -9
signal and ignore it....

On some UNIXes any hardware interupt hard a higher priority than the
signals that the software could send, and so once a process hangs up waiting
for a tape drive then ONLY an interupt from the TAPE wakes it up.
(your milage may vary)
In one case the processor in the tape drive was waiting for a signal
from the CPU and ignoring most of its buttons, while the
UNIX was wouldn't talk to the tape untilt the hung process was
not using the tape drive - deadly embrace.....
The only thing to do was to shutdown, ignore the complaints about
hung processes (or don't panic about them)....and reboot with tape
turned on...
dick botting (rbotting@atl.calstate.edu, dick@realsoonnow.csubnet.edu,...)
 


This is the original messsage
===============================================

Subject:  kill -9

Is there a stronger kill than kill -9?  I had a process which attached
itself to my tape drive and then locked up.  I couldn't kill it no
matter what I tried.  This caused me great concern, backups are very
important to me and I couldn't access the tape drive with this process
on it.  I shut down the system (two of my file systems didn't unmount
because of the process and shutting down didn't kill the process either) 
and rebooted.

Any suggestions.

Norm Frech < frech@mwraaa.army.mil >

 
This is a carbon copy of a message
sent by rbottin.