Date: Wed, 16 Dec 2009 16:40:31 +0100 From: Grzegorz Bernacki <gjb@semihalf.com> To: RuiDC <ruidc@yahoo.com> Cc: freebsd-arm@freebsd.org Subject: Re: fetch data corruption on local fs Message-ID: <4B28FF6F.4090601@semihalf.com> In-Reply-To: <26811801.post@talk.nabble.com> References: <26803523.post@talk.nabble.com> <4B28C608.1070802@gmail.com> <4B28CFCD.3000401@semihalf.com> <26811801.post@talk.nabble.com>
next in thread | previous in thread | raw e-mail | index | archive | help
RuiDC wrote: > Thanks! That's definitely the problem, I used an md filesystem instead of nfs > as it was easier, and reproduced the problem, this time on a smaller .gz > file (11MB): ftp://ftp.gnu.org/gnu/gettext/gettext-0.17.tar.gz > which was also failing consistently. If I mount it without -o sync I > reproduce the problem, if I mount it with -o sync it works. > > so the only questions open are that: > 1. this happens for relatively small files as above rather than the 300MB > referred to in your quoted thread, It all depends on how memory is use and which cache lines are evicted. In my case it happened always for files bigger then 300MB and sometimes for smaller files. > 2. what is the safest to use: just -o sync or also -o noclusterr -o > noclusterw ? As I understand setting just "-o noclusterr -o noclusterw" didn't help you, right? If so use just "-o sync". > > 3. I then need to seek how to compile these mount options into the kernel, > as this is for my root filesystem compiled using: option > ROOTDEVNAME=\"ufs:/dev/da0s2\" I don't know how to compile it into kernel, but you can try to add line: vfs.root.mountfrom.options="sync" to your loader.conf Grzesiek
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4B28FF6F.4090601>