On Tue, Feb 26, 2008 at 08:57:07PM +0100, José L. Redrejo Rodríguez wrote:
> El mar, 26-02-2008 a las 10:43 -0800, [EMAIL PROTECTED] escribió:
> > On Tue, Feb 19, 2008 at 12:03:09AM +0100, José L. Redrejo Rodríguez wrote:
> > > I can make more tests if some ideas, options or workarounds are
> > > possible.
> > you probably have much more experience with this mode of operation than
> > i do...
> Maybe with the operation I have more experience but I do not have a good
> knowledge of how the image is uncompressed and mounted from the initrd
> when the ltsp terminal starts. The fix could be something as stupid as
> setting some parameter for squashfs.ko.
ok, here's the code run from the initramfs:
# mount the readonly root
nbd-client ${NBD_ROOT_SERVER} ${NBD_ROOT_PORT} /dev/nbd0 && mount -o ro -t
squashfs /dev/nbd0 /rofs
# merge ro and rw filesystems
mount -t unionfs -o dirs=/cow=rw:/rofs=ro unionfs ${rootmnt}
this is in the client's /usr/share/initramfs-tools/scripts/ltsp_nbd
script...
i'm wondering if running it through NBD somehow is breaking it? can you
try on the server itself to mount the squashfs via NBD?
nbd-client 127.0.0.1 2000 /dev/nbd0
mount -o ro -t squashfs /dev/nbd0 /mnt
you're saying the mount works fine mounting the image directly with
something like:
mount -o ro,loop -t squashfs /opt/ltsp/images/i386.img /mnt
the mount being used in the initramfs may be from busybox or
klibc-utils, i'm not sure which. i wonder if that might behave
differently, too.
what kernel versions have you tried with?
live well,
vagrant
