View Incident:
http://co-op.engr.sgi.com/BugWorks/code/bwxquery.cgi?search=Search&wlong=1&view_type=Bug&wi=813113
*Assigned Engineer : raa *Assigned Domain : sgi.com
*Assigned Group : xfs-linux *Modified Date : 03/14/01
*Modified User : nb *Modified User Domain : sgi.com
*Description :
Below is the state of a system that has kswapd (pid = 3)
and a process performing I/O (pid = 1120) that are deadlocked
waiting for each other. I'm sure that the wakeup_kswapd()
changes in 2.4.1preX change this situation somewhat. However,
in the long run, allocating memory under IOLOCK could
still leave avenues for similar problems.
Steve, we'll use this bug to record problems in this area.
This instance is the fdatasync case, while running a 4-threaded
doio on a 64MB machine (takes about an hour to hit it).
.....
==========================
ADDITIONAL INFORMATION (REASSIGN)
From: nb@xxxxxxx (BWX)
Date: Mar 14 2001 01:40:45PM
==========================
RE-assigning to Bob
|