NetBSD Problem Report #45709

From campbell@mumble.net  Tue Dec 13 17:32:04 2011
Return-Path: <campbell@mumble.net>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
	by www.NetBSD.org (Postfix) with ESMTP id DC80863BC61
	for <gnats-bugs@gnats.NetBSD.org>; Tue, 13 Dec 2011 17:32:03 +0000 (UTC)
Message-Id: <20111213173204.247A5982AB@pluto.mumble.net>
Date: Tue, 13 Dec 2011 17:32:04 +0000 (UTC)
From: Taylor R Campbell <campbell+netbsd@mumble.net>
Reply-To: Taylor R Campbell <campbell+netbsd@mumble.net>
To: gnats-bugs@gnats.NetBSD.org
Subject: tmpfs size went nuts
X-Send-Pr-Version: 3.95

>Number:         45709
>Category:       kern
>Synopsis:       tmpfs size went nuts
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          analyzed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Dec 13 17:35:00 +0000 2011
>Closed-Date:    
>Last-Modified:  Thu Dec 05 17:20:03 +0000 2013
>Originator:     Taylor R Campbell <campbell+netbsd@mumble.net>
>Release:        NetBSD 5.99.58
>Organization:
>Environment:
System: NetBSD oberon.local 5.99.58 NetBSD 5.99.58 (RIAMONOHACK) #4: Sat Dec 10 23:18:29 UTC 2011 root@oberon.local:/home/riastradh/netbsd/current/obj/sys/arch/i386/compile/RIAMONOHACK i386
Architecture: i386
Machine: i386
>Description:

	I have three tmpfses mounted, at

		/tmp (1 GB)
		/var/chroot/pbulk/2011Q3/tmp (2 GB)
		/var/chroot/pbulk/current/tmp (1 GB)

	Last night I started a bulk build chrooted in
	/var/chroot/pbulk/2011Q3.  A lot of packages started failing,
	and I noticed that `df -h' reported the tmpfses to have a size
	(not available space, but size) of 10--20 MB!  The exact size
	reported by df for each of the tmpfses varied over time between
	around ten and twenty megabytes, sometimes ten, sometimes
	fifteen, sometimes sixteen, &c.  After an hour or so, the
	tmpfses returned to their correct sizes of 1 GB or 2 GB.

>How-To-Repeat:

	Do a bulk build with wrkobjdir set to a tmpfs in a chroot,
	maybe?

>Fix:

	Yes, please!

>Release-Note:

>Audit-Trail:
From: dieter roelants <dieter.NetBSD@pandora.be>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: kern/45709: tmpfs size went nuts
Date: Thu, 15 Dec 2011 10:49:28 +0100

 On Tue, 13 Dec 2011 17:35:00 +0000 (UTC)
 Taylor R Campbell <campbell+netbsd@mumble.net> wrote:

 > 	Last night I started a bulk build chrooted in
 > 	/var/chroot/pbulk/2011Q3.  A lot of packages started failing,
 > 	and I noticed that `df -h' reported the tmpfses to have a size
 > 	(not available space, but size) of 10--20 MB!  The exact size
 > 	reported by df for each of the tmpfses varied over time between
 > 	around ten and twenty megabytes, sometimes ten, sometimes
 > 	fifteen, sometimes sixteen, &c.  After an hour or so, the
 > 	tmpfses returned to their correct sizes of 1 GB or 2 GB.

 Could it be you didn't have enough free swap? I.e. the tmpfs file
 system size seems to be capped by the amount of free memory.

 dieter

From: Taylor R Campbell <campbell+netbsd@mumble.net>
To: dieter roelants <dieter.NetBSD@pandora.be>
Cc: kern-bug-people@netbsd.org, gnats-admin@netbsd.org,
	netbsd-bugs@netbsd.org, gnats-bugs@NetBSD.org
Subject: Re: kern/45709: tmpfs size went nuts
Date: Fri, 16 Dec 2011 22:59:31 +0000

    Date: Thu, 15 Dec 2011 10:05:03 +0000 (UTC)
    From: dieter roelants <dieter.NetBSD@pandora.be>

    Could it be you didn't have enough free swap? I.e. the tmpfs file
    system size seems to be capped by the amount of free memory.

 Unlikely.  I had two or three gigabytes of swap handy, and there was
 only about a gigabyte of RAM in use at the time, plus a few dozens of
 megabytes stored in the tmpfses.  So there should have been four or
 five gigabytes of memory free, not ten to twenty megabytes.

From: David Holland <dholland-bugs@netbsd.org>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: kern/45709: tmpfs size went nuts
Date: Sat, 17 Dec 2011 02:59:10 +0000

 On Fri, Dec 16, 2011 at 11:00:10PM +0000, Taylor R Campbell wrote:
  > From: Taylor R Campbell <campbell+netbsd@mumble.net>
  > To: dieter roelants <dieter.NetBSD@pandora.be>
  > Cc: kern-bug-people@netbsd.org, gnats-admin@netbsd.org,
  > 	netbsd-bugs@netbsd.org, gnats-bugs@NetBSD.org
  > Subject: Re: kern/45709: tmpfs size went nuts
  > Date: Fri, 16 Dec 2011 22:59:31 +0000
  > 
  >     Date: Thu, 15 Dec 2011 10:05:03 +0000 (UTC)
  >     From: dieter roelants <dieter.NetBSD@pandora.be>
  >  
  >     Could it be you didn't have enough free swap? I.e. the tmpfs file
  >     system size seems to be capped by the amount of free memory.
  >  
  >  Unlikely.  I had two or three gigabytes of swap handy, and there was
  >  only about a gigabyte of RAM in use at the time, plus a few dozens of
  >  megabytes stored in the tmpfses.  So there should have been four or
  >  five gigabytes of memory free, not ten to twenty megabytes.

 So, something was truncating to 32 bits? :-/

 -- 
 David A. Holland
 dholland@netbsd.org

From: Taylor R Campbell <campbell+netbsd@mumble.net>
To: David Holland <dholland-bugs@netbsd.org>
Cc: kern-bug-people@netbsd.org, gnats-admin@netbsd.org,
	netbsd-bugs@netbsd.org, gnats-bugs@NetBSD.org
Subject: Re: kern/45709: tmpfs size went nuts
Date: Sat, 17 Dec 2011 03:09:48 +0000

    Date: Sat, 17 Dec 2011 03:00:06 +0000 (UTC)
    From: David Holland <dholland-bugs@netbsd.org>

    On Fri, Dec 16, 2011 at 11:00:10PM +0000, Taylor R Campbell wrote:
     >  Unlikely.  I had two or three gigabytes of swap handy, and there was
     >  only about a gigabyte of RAM in use at the time, plus a few dozens of
     >  megabytes stored in the tmpfses.  So there should have been four or
     >  five gigabytes of memory free, not ten to twenty megabytes.

    So, something was truncating to 32 bits? :-/

 Hmm, sounds plausible, particularly since I'm on i386.  The
 combination of size_t and uint64_t in tmpfs_bytes_max in tmpfs_subr.c
 looks suspect, but it wouldn't surprise me if there were other
 problems with this elsewhere in the system.

State-Changed-From-To: open->analyzed
State-Changed-By: rmind@NetBSD.org
State-Changed-When: Thu, 05 Dec 2013 17:20:03 +0000
State-Changed-Why:
Did your tmpfs mounts have any size (memory limit) specified?  If there 
is no size/limit specified, it currently uses the amount of free memory,
which is, obviously, fluctuating.  

How do we want to fix this?


>Unformatted:

Home	PR Database Search