I've tested LSUpdater.exe installations of (my own) armel rootfs, (created using
Andreas Bach Aaen's method) on a clean Buffalo LS Live restored to using 2.10 firmware,
the original partition scheme , and the 2.6.16.16-arm1 kernel, using lb_worms initrd and micro_evtd, and mkimage.
I have some suggestions for a public new release of Freelink.
(1) use the stock Buffalo uImage.buffalo, 2.6.16.16-arm1, as in 1.10 or later
LSPro firwmare, and 2.10 LsLive firmware. Make the upgrade to this latest
Buffalo fw obligatory for LSUpdater.exe installs. I've only tested on v2 hardware,
but I think that this update on v1 hardware perhaps also eliminates the problem that
that the initial "initrd-only" step of FreelinkRev2 solves . (The "initrd-only" step
is NOT needed on a v2 LSLive running "2.10" fw (=1.10 u-boot)
The user can optionally install a rebuilt 2.6.16 Buffalo GPL kernel with lots of extra modules
afterwards.
ftp://ftp.dmik.org/pub/common/buffalo/lspro/duncan_h/kernel-2.6.16.57-lsp_eabi-dh_v4.tar.gzExisting Freelink users should upgrade by copying hddrootfs,buffalo.updated to
/dev/sda6, backup their existing rootfs (or at least, /etc) (after mounting it a second time),
rebooting into EM mode, make a new fs (ext3?) on /dev/sda2,
and untarring the rootfs into it. Then optionally copy their old /etc/passwd, hosts, hostname, shadow,
group gshadow, ssh/* , fstab etc. to the new rootfilesystem.
(2) The XFS file system already present as shipped by Buffalo needs to be supported.
This means using the same xfsprogs that ship with buffalo, NOT the debian sid armel
ones. The programs are in the initrd filesystem. They (and the libs libxfs, libxlog, libdisk)
can be installed to /usr/local (with their man pages). I just added a tarball that the user
can untar as part of the post-installation steps, if they wants to use xfs.
The user can be advised to convert the rootfs to ext3 using EM mode, obtained using
a reboot after setting TIMEOUT=EM in boot_options. (mount /dev/sda2 a second time on
e.g. /tmp/mnt, make a tarball of its contents (without other filesystems mounted on it)
to /dev/sda6, boot to EM mode, mkfs.ext3 -f /dev/sda2, mount -t ext3 /dev/sda2, mount
/dev/sad6 and restore root filesystem, change boot_options back, reboot). This should be
simple enough for a reasonably linux-aware "newbie" user. (The WIki recipe for this is an
unnecessarily-complicated procedure) But in practice, (buffalo) xfs works.
I just tested again with the latest armel xfsprogs. debian armel mkfs.xfs still produces a filesystem that fails
the xfs_check test, while the buffalo one produces a "good" filesystem that passes both buffalo
and debian armel xfs_check. I uninstalled debian armel xfsprogs.
Perhaps they are incompatible with the Buffalo GPL 2.6.16 kernels.
Are 2.6.2x non-Buffalo kernels suitable for general use? Probably not yet.
@lb_worm:
(minor detail for pre-release cleaning: your recent initrd/root has a junk hidden .bash_history)