Date: Fri, 28 Mar 2014 16:39:49 -0600 From: John Nielsen <lists@jnielsen.net> To: Karl Pielorz <kpielorz_lst@tdx.co.uk> Cc: freebsd-fs@freebsd.org Subject: Re: File system corruption with 9.2-R on PC Engines Alix boards Message-ID: <0D92C2CC-3640-42AA-B589-33929641E984@jnielsen.net> In-Reply-To: <C7112211D77A121BD752C9B7@Mail-PC.tdx.co.uk> References: <C7112211D77A121BD752C9B7@Mail-PC.tdx.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mar 28, 2014, at 11:02 AM, Karl Pielorz <kpielorz_lst@tdx.co.uk> wrote: > We have a number of PC Engines Alix boards, running FreeBSD 8.2. They boot off of an onboard CF cards. > > I recently installed a new one of these with 9.2-Release (i386) - only to discover that it silently (i.e. with no errors) destroys the file system when in use. > > Typically we install these systems then flip the file system over to 'read-only' when sending them out. The corruption happens while we're installing various packages etc. > > We don't run journal soft-updates on these boxes - just regular soft-updates . > > No console errors are logged, no syslog messages are logged. Just after a while you might go to edit '/etc/rc.conf' - to find when you vi it - it's now become a copy of '/etc/ntp.conf' - or other oddities. > > A reboot runs fsck - which will usually fail then. Running a foreground check reals off thousands of duplicate errors. If you foreground check the file system, you're usually left with "not a lot" when it's finished (i.e. if you run 'fsck -y /'). > > 8.2 runs fine (we have systems that have been running embedded for years) - 9.2 doesn't. > > I found a similar thread: > > <http://lists.freebsd.org/pipermail/freebsd-stable/2012-March/066697.html> > > This eludes to CF card quality etc. - the cards we've been using have worked fine for years - and a 9.2 'flaky' system reformatted to 8.2 then runs fine. > > Anyone else running later than 8.2 on PC Engine Alix kit? I have a pair of alix3dw systems that I use as wireless access points. Both are running 10-STABLE on internal CF without problems. (I do world and package builds on a faster host.) One has SU+J, the other has just SU. I know I ran some version of 9-STABLE on the same hardware as well (probably 9.1-ish) and don't remember any issues like those you are describing. FWIW, both use this CF card: http://www.newegg.com/Product/Product.aspx?Item=N82E16820134575 You might try to see when the badness appeared in the FreeBSD 9 branch by building kernels from various points and running them. If you use a different build host and just build a custom kernel without (many) modules or world it shouldn't be too onerous. I'd be curious to see if a different CF card has the same problems too. JN
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?0D92C2CC-3640-42AA-B589-33929641E984>
