Subject: kern/1460: swap leakage
To: None <gnats-bugs@gnats.netbsd.org>
From: Simon J. Gerraty <sjg@zen.void.oz.au>
List: netbsd-bugs
Date: 09/14/1995 23:53:16
>Number: 1460
>Category: kern
>Synopsis: swap space is not recovered.
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: kern-bug-people (Kernel Bug People)
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Sep 14 10:05:02 1995
>Last-Modified:
>Originator: Simon J. Gerraty
>Organization:
Zen Programming...
>Release: 95-08-18
>Environment:
System: NetBSD zen.void.oz.au 1.0A NetBSD 1.0A (ZEN) #0: Fri Aug 18 18:59:14 EST 1995 root@zen.void.oz.au:/f1/usr.src/sys/arch/i386/compile/ZEN i386
>Description:
zen is an i486DX33 with 16M ram and just under 60M swap.
It has been up 27 days (since installing 95-08-18 kernel).
While investigating why zen was running like a dog... I noticed that I had
no swap space left:
sjg:9104$ pstat -s
Device 1K-blocks Used Avail Capacity Type
/dev/wd0b 32640 32628 12 100% Interleaved
/dev/sd0b 24488 24468 20 100% Interleaved
/dev/sd1b 0 *** not available for swapping ***
/dev/sd2b 32668 *** not available for swapping ***
Total 57128 57096 32 100%
A quick look through the process table showed netscape at 16M and
emacs at 6M as the only serious memory hogs. I terminated netscape
and recovered a small amount of swap.
sjg:9106$ pstat -s
Device 1K-blocks Used Avail Capacity Type
/dev/wd0b 32640 32036 604 98% Interleaved
/dev/sd0b 24488 22516 1972 92% Interleaved
/dev/sd1b 0 *** not available for swapping ***
/dev/sd2b 32668 *** not available for swapping ***
Total 57128 54552 2576 95%
No where near 16M.
I terminated emacs and got back some more...
$ pstat -s
Device 1K-blocks Used Avail Capacity Type
/dev/wd0b 32640 29268 3372 90% Interleaved
/dev/sd0b 24488 17556 6932 72% Interleaved
/dev/sd1b 0 *** not available for swapping ***
/dev/sd2b 32668 *** not available for swapping ***
Total 57128 46824 10304 82%
Much better. I stoped/started xconsole which at 3M was next.
And so on.
At this point, top and ps suggest that 28M worth of swap should be in
use. Yet pstat -s says 40M.
I've been running this sort of load for ages, and never noticed total
swap usage exceed 50% before.
It would appear there is a leak somewhere.
>How-To-Repeat:
>Fix:
>Audit-Trail:
>Unformatted: