[Top] [All Lists]

ADD 804570 - The elevator bug

To: nb@xxxxxxx
Subject: ADD 804570 - The elevator bug
From: pv@xxxxxxxxxxxxx (chait@xxxxxxxxxxxx)
Date: Mon, 22 Jan 2001 18:37:07 -0800 (PST)
Cc: raa@xxxxxxx, huovinen@xxxxxxx, cattelan@xxxxxxx, mann@xxxxxxx, tbd@xxxxxxx, rrl@xxxxxxx, alaffin@xxxxxxx, cbrady@xxxxxxx, eac@xxxxxxx, hasib@xxxxxxx, coreym@xxxxxxx, steiner@xxxxxxx, linux-xfs@xxxxxxxxxxx
Reply-to: sgi.bugs.xfs@xxxxxxxxxxxxxxxxx
Sender: owner-linux-xfs@xxxxxxxxxxx
Webexec: webpvupdate,pvincident
Webpv: getafix.engr.sgi.com
View Incident: 

 Status : open                         Priority : 3                         
 Assigned Engineer : nb                Submitter : coreym                   
*Modified User : chait                *Modified User Domain : engr          
*Description :
I ran rwtest on a xfs-linux filesystem on the machine permit:

rwtest 100000000:/mnt1/file_1 >/tmp/file_1.out 2>&1 &

This caused df, ls -l, fdisk, and top all to hang.

Permit is running Redhat 6.2 with a 2.4.0-test5 kernel



From: chait@engr (BugWorks)
Date: Jan 22 2001 06:37:06PM

I just helped Ananth resolve some merging issues wrt. applying
the 2.4.1pre9 patch to the XFS tree. Most of the conflicts were
in the elevator.c file and ll_rw_blk.c file. So, I ended up doing
a code review of Jen's fix for the elevator problem. 

His patch provides the following additional functionality:
        1. consistent aging of requests,
        2. request aging based on incoming buffer sizes,
        3. eliminate redundant scans of the request queue
           for request insertion info.

1. and 2. primarily provide the fix we're looking for wrt. the
elevator problem. Change number 3. is performance icing.

Should be good to go for our SN/IA64 kernel. I'd wait for
Ananth to post an update on his tests with the new kernel too.


<Prev in Thread] Current Thread [Next in Thread>