On Mon, Oct 13, 2014 at 11:23 AM, John Florian <john.florian(a)dart.biz> wrote:
I have a script that causes a new initrd to be built for my live images (that have been
written to flash drives). This script works great on Fedora 18 but causes the error (see
subject) when run a Fedora 20 system – both having been created by livecd-tools. My
script calls _LiveImageCreatorBase__write_dracut_conf() and I can see that dmsquash-live
# cat /etc/dracut.conf
filesystems+="vfat msdos isofs ext4 xfs btrfs "
drivers+="sr_mod sd_mod ide-cd cdrom =ata sym53c8xx aic7xxx ehci_hcd uhci_hcd
ohci_hcd usb_storage usbhid firewire-sbp2 firewire-ohci sbp2 ohci1394 ieee1394 mmc_block
sdhci sdhci-pci pata_pcmcia mptsas udf virtio_blk virtio_pci virtio_scsi virtio_net
virtio_mmio virtio_balloon virtio-rng ums_realtek "
add_dracutmodules+=" dmsquash-live pollcdrom "
Interestingly, the pollcdrom module on that same line is included in the initrd without
any problem. Both modules appear to be available for inclusion:
# dracut --list-modules | sort
Does anyone have an easy answer to this one? I’ll keep digging deeper but thought someone
here might already know what’s going on.
Sorry for the late reply.
Does this link help?
Thanks for the reply, although I've already resolved the issue.
I had to force the HostOnly mode to be disabled. python-imgcreate, which generates my
/etc/dracut.conf file already has an entry for "hostonly=no". However, a
related file named /usr/lib/dracut/dracut.conf.d/01-dist.conf has "hostonly=yes"
and according to DRACUT.CONF(5), "it will overwrite parameters set in
/etc/dracut.conf." The same man page states that, "Command line parameters will
overwrite any values set in those config files." Apparently a command-line option
specifying an alternate config file will not be treated as if the contents of that config
file were stated directly on the command-line themselves. Thus it's necessary to
explicitly disable the hostonly mode on the dracut command line, otherwise the critical
dmsquash-live module will be excluded and the resultant initramfs will be unable to make
the squashed root file system available.
Reading the link you sent seems to confirm what I found. Thanks anyway Fred!