[Btrfs-devel] btrfs on /

[Chris Mason]

On Thursday 07 February 2008, Chris Mason wrote:
> On Wed, Feb 06, 2008 at 05:49:35PM +0100, Pär Andersson wrote:
> > On Tuesday 05 February 2008 21:36:27 Christian Hesse wrote:
> > > now that I have a live CD with btrfs support I decided to give btrfs on
> > > root partition a try. I have good backups. ;) Here is what I have to
> > > tell so far:
> >
> > Here is one more data point. As an experiment I have used btrfs
> > on / since 0.11 was released. Initially I believed that it
> > wouldn't survive more than a few days, but I was wrong and now I
> > must agree that it is quite stable. Great work!
> >
> > I have btrfs / on a usb stick, so the old filesystems are still
> > safe on the hard drive if btrfs should eat my data.
> >
> > I have only found two issues:
> >
> > * No ENOSPC handling really sucks when your root file system is
> >   on a small usb stick. :-)
> >
> > * APT (Debian/Ubuntu package management) crashes when it tries to
> >   read its package lists. I have not had time to debug this yet,
> >   but a simple "aptitude update" is enough to reproduce the
> >   problem:
> >
> > root at garbage:~# aptitude update
> > -- snip --
> > Fetched 3B in 0s (11B/s)
> > Reading package lists... Error!
> > E: Unable to determine the file size - fstat (75 Value too large for
> > defined data type)
> > E: Can't mmap an empty file
> > W: You may want to run apt-get update to correct these problems
> > E: Empty package cache
> > E: Couldn't rebuild package cache
> > root at garbage:~#
>
> Ah ok, this is probably an underflow of i_blocks.  If you strace it
> you'll probably find the offending file.

You can also try the code I just pushed to btrfs-unstable, it seems to fix 
i_blocks accounting problems so far.

http://oss.oracle.com/mercurial/mason/btrfs-unstable/rev/a503a3f82684

-chris