NetBSD Problem Report #9792

Received: (qmail 5581 invoked from network); 5 Apr 2000 20:18:14 -0000
Message-Id: <200004052016.VAA01807@shark1.cambridge.arm.com>
Date: Wed, 5 Apr 2000 21:16:00 +0100 (BST)
From: Richard Earnshaw <rearnsha@cambridge.arm.com>
Reply-To: rearnsha@cambridge.arm.com
To: gnats-bugs@gnats.netbsd.org
Subject: Kernel leaking buffers?
X-Send-Pr-Version: 3.95

>Number:         9792
>Category:       kern
>Synopsis:       Kernel leaking buffers?
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    chs
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Apr 05 20:19:00 +0000 2000
>Closed-Date:    Sun Apr 16 21:45:20 +0000 2000
>Last-Modified:  Sun Apr 16 21:45:20 +0000 2000
>Originator:     Richard Earnshaw
>Release:        NetBSD-current 2000/03/27<NetBSD-current source date>
>Organization:
ARM
-- 
>Environment:

System: NetBSD shark1 1.4X NetBSD 1.4X (SHARK) #52: Mon Mar 27 17:55:21 BST 2000 rearnsha@shark1:/usr/src/sys/arch/arm32/compile/SHARK arm32

	shark1:testsuite [537] $ df -k
	Filesystem  1K-blocks     Used    Avail Capacity  Mounted on
	/dev/wd0a               194983    37479   147754    20%    /
	/dev/wd0e              2028219   460698  1466110    23%    /usr
	mfs:49                  198183       22   188251     0%    /tmp
	/dev/wd0f              2028219   573062  1353746    29%    /usr/build
	/dev/wd0g              2028219  1583868   342940    82%    /work
	/dev/wd0h              2925218   834382  1944575    30%    /xx

	(/usr/build and /xx are mounted with soft-update enabled; all others
	are non-soft-update)

>Description:

	Sorry this report is about a kernel that is a week old, but it takes
	that long before the problem starts to manifest itself :-(

	Since I started using a kernel with soft-updates and trickle sync all
	my kernels have started to thrash the disk horribly after about a week
	of uptime.  I can't be entirely sure where the failure is (hints of
	what information you require would be appreciated), but "systat vmstat"
	is showing that the sys-cache hit ratio has dropped off alarmingly 
	(when the system first comes up I'm typically seeing 98->99% hits on
	approx 5k->10k calls; after a while this drops dramatically to about 
	50% hits on 500->2k calls).  In addition, the disk seems to start
	making a very large number of head movements (I can hear it).

	I'm not sure if this is directly related to the soft-updates code
	or trickle syncing (or either unrealated).  But it still occurs
	even if accessing only discs that aren't mounted with soft-updates.

	I haven't tried a kernel built without soft-updates.

	Final note: the machine only has 32M of memory, so I guess that if it
	is a buffer problem, I'm likely to start seeing problems sooner than
	folks with massive amounts of RAM (and hence system buffers of various
	types).


>How-To-Repeat:

	Leave a machine running for a week or so.

>Fix:
	No idea.

>Release-Note:
>Audit-Trail:

From: Richard Earnshaw <rearnsha@arm.com>
To: gnats-bugs@gnats.netbsd.org
Cc: rearnsha@arm.com
Subject: Re: kern/9792: Kernel leaking buffers? 
Date: Thu, 06 Apr 2000 11:45:40 +0100

 Some additional probing leads me to suspect that this is related to the
 vnode pool becoming significantly larger than the namei pool.  Running
 vmstat -m shows that the machine has more than twice as many vnodepl
 buffers allocated than ncachepl entries; which means (if I understand the
 interworking of these two correctly) that even if an entry is already in a
 vnode buffer, there is only a 50% chance of the name cache being able to
 hold a reference to it.

 sysctl shows that maxvnodes is 778 and that the ncachepl has maxed out at 
 this, but that vnodepl has continued to grow.  Using sysctl to forcibly 
 increase the maximum number of buffers solves the thrashing problem, but I 
 guess that if things really are leaking, it will only be for a period of 
 time.

 Memory resource pool statistics
 Name        Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg 
 Maxpg Idle

 vnodepl      208     1566    0        0    83     0    83    83     0   
 inf    0
 ncachepl      72      778    0        0    14     0    14    14     0   
 inf    0





State-Changed-From-To: open->closed 
State-Changed-By: chs 
State-Changed-When: Sun Apr 16 14:44:22 PDT 2000 
State-Changed-Why:  
fixed in rev 1.23 of vfs_cache.c 


Responsible-Changed-From-To: kern-bug-people->chs 
Responsible-Changed-By: chs 
Responsible-Changed-When: Sun Apr 16 14:44:22 PDT 2000 
Responsible-Changed-Why:  
I looked at this. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.