Subject: kern/16323: null layer keeping vnodes active?
To: None <>
From: Frank Kardel <>
List: netbsd-bugs
Date: 04/12/2002 11:16:24
>Number:         16323
>Category:       kern
>Synopsis:       fs mounted via null seem to take a VERY long time until space is freed
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Apr 12 02:17:00 PDT 2002
>Originator:     Frank Kardel
>Release:        NetBSD 1.5ZC
System: NetBSD pip 1.5ZC NetBSD 1.5ZC (PIP) #0: Sun Apr 7 16:19:43 MEST 2002 kardel@pip:/fs/IC35L060-0-a/src/NetBSD/netbsd/sys/arch/i386/compile/PIP i386
Architecture: i386
Machine: i386
	When using file systems that are mount via the null layer it take a LONG time
	until space used by files that are deleted via the null layer is reclaimed
	by the underlying fs. This leads to fs full although there should be space
	the allocate. Note that this is not related to the async free meachinism.
	Space is not reclaimed for periods over serveral (>20) minutes. We run
	amanda backup across a null mount point - this is where we saw that
	Another observation is that quota entries are not updated for some
	time similar to the space reclamation problem above.
	run amanda with holding disk via null mount point.
	umounting and mount the null layer brings things in sync again.
	quotaoff/on will restore correct quota counts.