Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 5 Mar 2018 21:48:57 +0100
From:      Bernd Walter <ticso@cicely7.cicely.de>
To:        tech-lists <tech-lists@zyxst.net>
Cc:        freebsd-arm@freebsd.org, ticso@cicely.de
Subject:   Re: growfs problem on current
Message-ID:  <20180305204857.GK31939@cicely7.cicely.de>
In-Reply-To: <d78feadd-2f5a-811d-7d13-cf9a9b0be585@zyxst.net>
References:  <20180305171128.GI31939@cicely7.cicely.de> <d78feadd-2f5a-811d-7d13-cf9a9b0be585@zyxst.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, Mar 05, 2018 at 07:53:12PM +0000, tech-lists wrote:
> On 05/03/2018 17:11, Bernd Walter wrote:
> > Not sure if this is stricly current, but so far I could see this
> > problem with current only.
> > The filesystem expands and everything seems to be fine.
> > But in fact the filesystem is broken and fails in a way, which can't
> > be solved by fsck anymore.
> > Somehow it tries to access data not reachable.
> > I havn't had the time to investigate it further.
> > Don't know if something in the partition tables or the filesystem is
> > wrong.
> > As a workaround I always expand the FS on my stable running desktop
> > before booting on the ARM.
> 
> Not noticed any issue. I built arm6 for rpi2 just yesterday. From these
> sources:
> 
> $ svnlite info fbsd-12-src/
> Path: fbsd-12-src
> Working Copy Root Path: /root/crochet/fbsd-12-src
> URL: https://svn.freebsd.org/base/head
> Relative URL: ^/head
> Repository Root: https://svn.freebsd.org/base
> Repository UUID: ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f
> Revision: 330287
> Node Kind: directory
> Schedule: normal
> Last Changed Author: kevans
> Last Changed Rev: 330287
> Last Changed Date: 2018-03-02 17:07:08 +0000 (Fri, 02 Mar 2018)
> 
> The built image:
> 
> -rw-r--r--  1 root  wheel   7.3G  4 Mar 23:31
> FreeBSD-armv6-12-GENERIC-NODEBUG-RaspberryPi2.img
> 
> written to a 32GB card:
> 
> Filesystem        Size    Used   Avail Capacity  Mounted on
> /dev/mmcsd0s2a     28G    3.0G     23G    11%    /
> devfs             1.0K    1.0K      0B   100%    /dev
> /dev/mmcsd0s1      50M    8.7M     41M    17%    /boot/msdos
> /dev/md1           14M     88K     13M     1%    /var/log
> /dev/md2           11M     12K     10M     0%    /var/tmp
> /dev/md3          242M    8.0K    222M     0%    /tmp
> 
> uname from installed system:
> 
> FreeBSD 12.0-CURRENT (GENERIC-NODEBUG) #0 f2f8a36(master): Sun Mar  4
> 23:30:06 GMT 2018
> 
> this is in /etc/rc.conf:
> 
> growfs_enable="YES"
> 
> I'm not seeing any errors at all.
> 
> What errors are you getting that make you think it's the filesystem, and
> what sources (revision) did you build from, and what method did you use
> to build? I used freebsd/crochet.

I've used countless prebuild current images.
The last one was this:
FreeBSD-12.0-CURRENT-arm64-aarch64-RPI3-20180226-r330034.img.xz

In that case everything seemed to be ok.
I've installed countless packages, including X.
Then I needed to preload a specific kernel module, so I've created a
/boot/loader.conf
The loader complained that the file is unreadable.
Otherwise the system booted ok and I could access the file without
problems.
After removing the file everything was fine again.
The I've added my lines to /boot/defaults/loader.conf and the loader
complained about this file.
After booting I could access the file without problems.
Removing those lines didn't avoid the problem with loader.
Then I've removed the card from the machine and did an fsck from desktop,
which spit some read errors on reading blocks.
I wasn't physical read error from a failing card, just an out of scope
block number.
I don't have the exact output anymore.

In a previous case I've had a filesystem panic and fsck failed.
Maybe it is something specific on the way the official images are build
to trigger this problem.
I've seen this problem with different cards, different boards, over
several weeks of current images and with armv7 as well as arm64.
Not a single system survived when the system increased the FS itself.

I know I should have more data about this and should have done more
low level tests.
I didn't expect that I'm the only one having problems with this.
Will do some tests with a self grown FS directly after it has been
increased.
So far I'd only seen the corruption after some heavy writes.

-- 
B.Walter <bernd@bwct.de> http://www.bwct.de
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180305204857.GK31939>