I have installed sys-fs/zfs-0.6.0_rc11 with USE flags rootfs and test-suite. Emerge message contains following lines: * You should add zfs to the boot runlevel. * You should add zfs-shutdown to the shutdown runlevel. so I added both init-scripts to the appropriate runlevel. My root fs is ext4. Every time I try to shutdown the system hangs with error messages like "Cannot remount / read-only because / is not ext4" (I'm too lazy to type the message letter by letter as this happens after syslog shutdown. If needed I will provide the error.) I think that either a) the ebuild should mention that adding zfs-shutdown to runlevel shutdown is only useful and necessary if your root partition is zfs, or b) the zfs-shutdown init script should check, if root partition is zfs, else only umount every other mounted zfs. Reproducible: Always Steps to Reproduce: 1. install sys-fs/zfs with USE rootfs test-suite 2. add init scripts to runlevels as mentioned by emerge ( and start /etc/init.d/zfs to load kernel module) 3. shutdown Actual Results: System hangs while shutting down, does not power off. Expected Results: Clean shutdown.
Confirming this. Situation: root on ext4 LVM volume, initramfs generated by sys-kernel/genkernel-3.4.24_p2, adding zfs-shutdown to shutdown runlevel - and shutdown/reboot stucks with: /linuxrc: line 10: expr: command not found
My apologies for not getting to this sooner. I was preoccupied with Hurricane Sandy when this was filed and I am still catching up on the backlog that caused. I believe that I fixed this issue a few weeks ago in =sys-fs/zfs-0.6.0_rc12-r1. If that is not the case, please let me know.
Hi Richard, thanks for working on this bug. I hope you and your family are all well. I'm sorry to report that this issue isn't resolved yet. But before presenting the error message, here's a short overview to my drives: device fs mountpoint options comment /dev/sda1 ntfs-3g /mnt/windows noauto my dual-boot windows /dev/sda2 ext4 / noatime Gentoo root /dev/sda3 swap - sw swap /dev/sda4 ntfs-3g /mnt/daten noatime data space, access with win and lin /dev/sdb1 ntfs-3g /mnt/sdb1 noatime data space, access with win and lin and sometimes /dev/sdc auto /mnt/usb external drive with changing FS including ZFS So this is the message coming when trying to reboot/shutdown: ######## [...] Stopping acpid ... [ok] Stopping syslog-ng ... [ok] Unmounting ZFS filesystems ... [ok] Unmounting loop devices ... Unmounting filesystems ... Unmounting /mnt/sdb1 ... [ok] Unmounting /mnt/daten ... [ok] Deactivating swap devices ... [ok] Stopping udev ... [ok] Setting hardware clock using system clock [Local Time] ... [ok] Terminating remaining processes ... [ok] Killing remaining processes ... [ok] Saving dependency cache ... [ok] Remounting remaining filesystems read-only ... Remounting / read only ... [ok] mount: Too many levels of symbolic links /linuxrc: line 10: expr: command not found mkdir: cannot create directory '/': File exists mount: special device /run does not exist /linuxrc: line 10: expr: command not found mkdir: cannot create directory '/': File exists mount: wrong fs type, bad option, bad superblock on /dev, missing codepage or helper program, or other error In some cases useful info is found in syslog - try dmesg | tail or so /linuxrc: line 10: expr: command not found mkdir: cannot create directory '/': File exists mount: special device /sys/fs/cgroup does not exist /linuxrc: line 10: expr: command not found Failed to load ZFS module stack. Load the module manually by running 'insmod <location>/zfs.ko' as root. internal error: unable to open /etc/mtab Failed to load ZFS module stack. Load the module manually by running 'insmod <location>/zfs.ko' as root. Failed to load ZFS module stack. Load the module manually by running 'insmod <location>/zfs.ko' as root. ######## Then the system hangs and only a long press on the power button powers off the system. Oh, one thing: I'm using =sys-kernel/gentoo-sources-3.5.7 but no initrd.
What does `df -TP / | awk 'NR>1{print $2}'` output on your system?
As I wrote in comment #3 my root partition (sda2) is ext4: > df -TP / | awk 'NR>1{print $2}' ext4
i have a problem on shutdown with a root ZFS system and /usr/portage etc as seperate datasets: when trying to shutdown i get: can't unmount /usr, still busy added zfs-shutdown to the shutdown runlevel
Hi Simon, please provide the output of "zfs list" and "rc-update show". Did your problem occur with or without zfs-shutdown being in the shutdown runlevel? Did adding it to the shutdown runlevel solve your problem?
Originally, I wrote the zfs-shutdown script because I felt that it would be a good idea to cleanly export the root pool, but that is unnecessary. No other ZFS implementation does this and I am now fairly certain that it was a mistake. Situations where the pool is cleanly exported will result in a label update that invalidates zpool.cache, requiring a force import on the next reboot. As a consequence, the zfs-shutdown script will never work as I had originally intended and I have removed it from the tree. If you update sys-fs/zfs and have zfs-shutdown in the boot runlevel, it will be automatically removed.
As an additional note, you will see some messages about not being able to unmount datasets at shutdown when using ZFS as a root filesystem. This is harmless. What happens is that during shutdown, OpenRC will invoke RC/init scripts' shutdown() functions in reverse order of dependencies. This means that zfs is called before localmount. When the rootfs is ZFS, zfs will naturally try to unmount it, but localmount has other stuff on top, which prevents that, causing information about the failure to unmount ZFS to be printed. That is a comsetic issue. I plan to address it in the future, but I consider it to be a low priority. I have opened a new bug to track it. I have also CCed the people in this bug.
(In reply to comment #9) > As an additional note, you will see some messages about not being able to > unmount datasets at shutdown when using ZFS as a root filesystem. This is > harmless. > > What happens is that during shutdown, OpenRC will invoke RC/init scripts' > shutdown() functions in reverse order of dependencies. This means that zfs > is called before localmount. When the rootfs is ZFS, zfs will naturally try > to unmount it, but localmount has other stuff on top, which prevents that, > causing information about the failure to unmount ZFS to be printed. > > That is a comsetic issue. I plan to address it in the future, but I consider > it to be a low priority. I have opened a new bug to track it. I have also > CCed the people in this bug.
Ignore that last comment. I was setting bug #450758 under See Also and accidentally hit enter without realizing that I had started a reply to myself.
