Number: 1113

Date: 18-May-84 11':39':58

Submitter: LE.PASA

Source:  Eric Schoen -- Schlumberger (Dering.pasa)

Subject: Aborted Process does not release Monitor Lock?

Assigned To: 


Status: Incomplete


Problem Type: Bug

Impact: Serious

Difficulty: Hard

Frequency: Everytime

Priority: Hopefully

System: Operating System

Subsystem: Processes



Lisp Version:  Fugue4

Source Files: 

Microcode Version: 

Memory Size: 

File Server: 

Server Software Version: 


Description: '
Eric Schoen reports the following ':'
One of our users got himself in a situation where he could not open a Leaf connection to a particular file server host.  His process would hang on OBTAIN.MONITORLOCK forever.  I broke the process and backtraced to find the lock he was waiting on (associated with the SEQUIN connection to this'
server) was owned by a processes which had STACK OVERFLOWed and was aborted via ↑ from the break window.  Once I manually freed the lock, he was able to connect to this host.'
It seems there''s a case where a process can die and not release the locks it owns.  Very dangerous.'
Date': 22 May 84 16':02 PDT'
Subject': Re': AR#1113': Aborted Process does not release Monitor Lock'
To': 1100Support.pasa'
When a process is killed (or aborts itself), any monitorlocks it still owns DO get released, so that is not the problem.'
The main way I know of a monitor lock not getting released is if you do a RETFROM around the code that obtained the lock (well-known bug concerning RESETLST).  But ↑ doesn''t do that, so this does not explain the AR.'
Anyway, if customer has a case where he can demonstrate a monitorlock not getting released without resorting to the previous paragraph, I would like to see it.'
There is a slight possibility that the statements above are not true for Fugue4, so I would want to see an example in a newer release if possible.'


Test Case: 

Edit-By: vanMelle

Edit-Date: 12-Jul-84 15':09':09