Subject: kern/35577: unionfs and nullfs spending too much I/O time
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: None <juan@xtrarom.org>
List: netbsd-bugs
Date: 02/08/2007 17:35:00
>Number: 35577
>Category: kern
>Synopsis: unionfs and nullfs spending too much I/O time
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: kern-bug-people
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Feb 08 17:35:00 +0000 2007
>Originator: Juan RP
>Release: NetBSD 4.99.9
>Organization:
The NetBSD Project
>Environment:
System: NetBSD nocturno 4.99.9 NetBSD 4.99.9 (Nocturno) #17: Wed Feb 7 17:59:28 CET 2007 juan@nocturno:/home/juan/build/obj/sys/arch/i386/compile/Nocturno i386
Architecture: i386
Machine: i386
>Description:
With last vnode locking changes made by chs@, working
with a stackable filesystem (I tried nullfs and unionfs)
spends a reasonable amount of time in I/O
(hard disk is continuosly reading something).
I was trying to test the newlock2 branch, so I downloaded
it into ~/src-newlock2, I mounted it with nullfs like:
$ mount_null ~/src-newlock2/sys /usr/src/sys
and ran build.sh to build a new kernel, the depend process
took >5 minutes! when 2 weeks ago it took <1m, same
for unionfs.
>How-To-Repeat:
Just try to use a null or union mount and see how slow
is.
>Fix:
Yes, please.