HERBER@FNALA.FNAL.GOV ("Randolph J. Herber, CD/DCD/SP, x2966") (03/06/91)
>Date: Tue, 5 Mar 91 18:26:50 PST >From: Stefan Farestam <farestam@orion.cerfacs.fr> >Message-Id: <9103060226.AA15459@ultima.cerfacs.fr> >To: info-iris@BRL.MIL >Subject: processes with eternal life ? > > > I have a process which refuses to be killed. It is a piece of > graphics code running on a 4D25TG under IRIX 3.3.1. The window > comes up and then dies. Afterwards it is impossible to kill > the process (both as a normal user and the super-user). If I > reboot the machine, then these processes disappears, but that > is a somewhat awkward manoeuvre just to get rid of a process ! > > Any suggestions or insights ? > > Thanks, > Stefan Farestam All that kill() or the kernel equivalent does is set a bit in p_sig bit map in the process' process block. And, if the process is running on a another processor, a hardware inter-CPU interrupt is sent to that processor. The actual signal processing does not occur until that process is being dispatched to user state again. At that time the signal (the bit in p_sig) may be ignored, default actioned or control given to a signal handler in the user code (with changes being made to the stack to call the signal handler and if it returns, re-enter the kernel to support a return to the original point of interrupt in the user code.) If the process never comes to point of dispatch to user state, the process does not terminate. If a process is waiting at priority above priority PZERO, it will not be woke up when a signal is presented, not even for uid 0, root. Usually this condition occurs when a process is waiting for the completion of an I/O operation. The reason for this behavior is that the process has system critical resources locked that as the kernel is currently structured will only be released when the current operation completes. Most kernels do not have the necessary coding that would release the locked resources if a process was just interrupted and terminated anywhere in the kernel. If a process is waiting below priority PZERO in the kernel, the system call is terminated with errno 4, EINTR, and the process is marked runnable. In this case, the process will in time be dispatched and signal processing done. This is another Frequently Asked Question. Randolph J. Herber, herber@fnalf.fnal.gov
farestam@orion.cerfacs.fr (Stefan Farestam) (03/06/91)
I have a process which refuses to be killed. It is a piece of
graphics code running on a 4D25TG under IRIX 3.3.1. The window
comes up and then dies. Afterwards it is impossible to kill
the process (both as a normal user and the super-user). If I
reboot the machine, then these processes disappears, but that
is a somewhat awkward manoeuvre just to get rid of a process !
Any suggestions or insights ?
Thanks,
Stefan Farestam
.................................................................
. Stefan Farestam <farestam@cerfacs.fr> .
. __ __ __ _ _ _ .
. / |_ )|_ /_\/ ( European Centre for Research and .
. \_ |__\| / \__) Advanced Training in Scientific Computation .
................................................................. buchanan@cs.ubc.ca (John (juancho) buchanan) (03/07/91)
Does this means some one is working on evangelist processes? ========================================================================= | |===============================| | John Buchanan (juancho) | buchanan@cs.ubc.ca | | Imager Manager |===============================| | Imager | (604) 228-2218 | | Department of Computer Science |===============================| | University of British Columbia | Standard disclaimer | | Vancouver, BC, Canada | included in this | | | box, right here. | ========================================================================= | This from ;login: (The USENIX Association Newsletter), Sep/Oct 1990, | | in a report on a Usenix Standards BOF session: | | | | An overheard conversation: | | | | -"Mach was the greatest intellectual fraud in the last ten years." | | | | -"What about X?" | | | | -"I said `intellectual' " | | | =========================================================================
buchanan@cs.ubc.ca (John (juancho) buchanan) (03/07/91)
Does this mean some one is working on evangelist processes? ========================================================================= | |===============================| | John Buchanan (juancho) | buchanan@cs.ubc.ca | | Imager Manager |===============================| | Imager | (604) 228-2218 | | Department of Computer Science |===============================| | University of British Columbia | Standard disclaimer | | Vancouver, BC, Canada | included in this | | | box, right here. | ========================================================================= | This from ;login: (The USENIX Association Newsletter), Sep/Oct 1990, | | in a report on a Usenix Standards BOF session: | | | | An overheard conversation: | | | | -"Mach was the greatest intellectual fraud in the last ten years." | | | | -"What about X?" | | | | -"I said `intellectual' " | | | =========================================================================