zfs/contrib/initramfs
Garrett Fields e8b199faec Force systems with kernel option "quiet" to display prompt for password
On systems that utilize TTY for password entry, if the kernel 
option "quiet" is set, the system would appear to freeze on a 
blank screen, when in fact it is waiting for password entry 
from the user.

Since TTY is the fallback method, this has no effect on systemd 
or plymouth password prompting.

By temporarily setting "printk" to "7", running the command, 
then resuming with the original "printk" state, the user can 
see the password prompt.

Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Garrett Fields <ghfields@gmail.com>
Closes #9731
2019-12-17 17:45:06 -08:00
..
conf-hooks.d initramfs: setup keymapping and video for prompts 2019-12-16 16:54:51 -08:00
conf.d contrib/initramfs: add missing conf.d/zfs 2018-02-12 11:40:00 -08:00
hooks Restore :: in Makefile.am 2019-08-26 11:48:31 -07:00
scripts Force systems with kernel option "quiet" to display prompt for password 2019-12-17 17:45:06 -08:00
Makefile.am initramfs: fixes for (debian) initramfs 2019-08-16 09:02:32 -06:00
README.initramfs.markdown Update README.initramfs.markdown 2018-01-26 09:55:16 -08:00

README.initramfs.markdown

DESCRIPTION These scripts are intended to be used with initramfs-tools, which is a similar software product to "dracut" (which is used in RedHat based distributions), and is mainly used by Debian GNU/Linux and derivatives to create an initramfs so that the system can be booted off a ZFS filesystem. If you have no need or interest in this, then it can safely be ignored.

These script were written with the primary intention of being portable and usable on as many systems as possible.

This is, in practice, usually not possible. But the intention is there. And it is a good one.

They have been tested successfully on:

* Debian GNU/Linux Wheezy
* Debian GNU/Linux Jessie

It uses some functionality common with the SYSV init scripts, primarily the "/etc/zfs/zfs-functions" script.

FUNCTIONALITY

  • Supports booting of a ZFS snapshot. Do this by cloning the snapshot into a dataset. If this, the resulting dataset, already exists, destroy it. Then mount it as the root filesystem.

    • If snapshot does not exist, use base dataset (the part before '@') as boot filesystem instead.
    • Clone with 'mountpoint=none' and 'canmount=noauto' - we mount manually and explicitly.
    • Allow rollback of snapshots instead of clone it and boot from the clone.
    • If no snapshot is specified on the 'root=' kernel command line, but there is an '@', then get a list of snapshots below that filesystem and ask the user which to use.
  • Support all currently used kernel command line arguments

    • Core options: All the different distributions have their own standard on what to specify on the kernel command line to boot of a ZFS filesystem.

      Supports the following kernel command line argument combinations (in this order - first match win):

      • rpool= (tries to finds bootfs automatically)
      • bootfs=/ (uses this for rpool - first part)
      • rpool= bootfs=/
      • -B zfs-bootfs=/ (uses this for rpool - first part)
      • rpool=rpool (default if none of the above is used)
      • root=/ (uses this for rpool - first part)
      • root=ZFS=/ (uses this for rpool - first part, without 'ZFS=')
      • root=zfs:AUTO (tries to detect both pool and rootfs
      • root=zfs:/ (uses this for rpool - first part, without 'zfs:')

      Option could also be

    • Extra (control) options:

      • zfsdebug=(on,yes,1) Show extra debugging information
      • zfsforce=(on,yes,1) Force import the pool
      • rollback=(on,yes,1) Rollback (instead of clone) the snapshot
  • 'Smarter' way to import pools. Don't just try cache file or /dev.

    • Try to use /dev/disk/by-vdev (if /etc/zfs/vdev_id.conf exists),
    • Try /dev/mapper (to be able to use LUKS backed pools as well as multi-path devices).
    • /dev/disk/by-id and any other /dev/disk/by-* directory that may exist.
    • Use /dev as a last ditch attempt.
    • Fallback to using the cache file if that exist if nothing else worked.
    • Only try to import pool if it haven't already been imported
      • This will negate the need to force import a pool that have not been exported cleanly.
      • Support exclusion of pools to import by setting ZFS_POOL_EXCEPTIONS in /etc/default/zfs.

    Controlling in which order devices is searched for is controlled by ZPOOL_IMPORT_PATH variable set in /etc/defaults/zfs.

  • Support additional configuration variable ZFS_INITRD_ADDITIONAL_DATASETS to mount additional filesystems not located under your root dataset.

    For example, if the root fs is specified as 'rpool/ROOT/rootfs', it will automatically and without specific configuration mount any filesystems below this on the mount point specified in the 'mountpoint' property. Such as 'rpool/root/rootfs/var', 'rpool/root/rootfs/usr' etc)

    However, if one prefer to have separate filesystems, not located below the root fs (such as 'rpool/var', 'rpool/ROOT/opt' etc), special configuration needs to be done. This is what the variable, set in /etc/defaults/zfs file, needs to be configured. The 'mountpoint' property needs to be correct for this to work though.

  • Allows mounting a rootfs with mountpoint=legacy set.

  • Include /etc/modprobe.d/{zfs,spl}.conf in the initrd if it/they exist.

  • Include the udev rule to use by-vdev for pool imports.

  • Include the /etc/default/zfs file to the initrd.