Bug 1551 - Deadlock in the ORB
Summary: Deadlock in the ORB
Status: RESOLVED FIXED
Alias: None
Product: TAO
Classification: Unclassified
Component: ORB (show other bugs)
Version: 1.3.3
Hardware: All All
: P3 normal
Assignee: DOC Center Support List (internal)
URL:
Depends on:
Blocks:
 
Reported: 2003-07-11 14:55 CDT by Nanbor Wang
Modified: 2003-07-21 15:08 CDT (History)
2 users (show)

See Also:

Attachments
Test case.. (9.41 KB, application/octet-stream)
2003-07-11 14:58 CDT, Nanbor Wang 		Details
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Nanbor Wang 2003-07-11 14:55:18 CDT 
This is an extension to bug 1482. The ORB can deadlock if one tries send
messages while handling exceptions in AMI. A test will be  attached shortly.
Thanks to Carlo for reporting the problem. 

The problem shows up when the following sequence of actions occur

- client after sending an AMI call tries to collect the reply
- peer dies. An event is waiting on the client's reactor indicating the death of
the peer.
- the client starts cleaning up and reaches
TAO_Transport::send_close_connection_notification ().
- the above call holds a lock and tries to clean up the  muxed strategy which in
turn delivers exception notification to the application.
- If the application tries sending out a message during exception notification
the application would deadlock trying to grab the lock which it already holds. 

There is a patch that Gan will integrate next week when we are at OMG's RTWS.
Comment 1 Nanbor Wang 2003-07-11 14:58:09 CDT 
Created attachment 220 [details]
Test case..
Comment 2 gan.deng 2003-07-21 15:08:44 CDT 
A work around has been added and here is the relevant ChangeLog entry:

TAO
-----
Tue Jul 15 10:18:07 2003  Gan Deng  <gan.deng@vanderbilt.edu>

Thanks again to Bala and Carlos O'Ryan and all other folks who helped with 
this.