Red Hat Bugzilla – Bug 1020398
cannot find all encrypted disks needed to boot
Last modified: 2013-10-24 12:23:52 EDT
I'm running an updated F20 x86_64 install, booting with UEFI (no secure boot) and the following disk layout:
/dev/sda - contains vg with encrypted lv for swap
/dev/sdb - contains nothing used for fedora install
/dev/sdc1 - ESP
/dev/sdc2 - /boot as unencrypted ext4
/dev/sdc3 - / as encrypted ext4
/dev/sdc4 - /home as encrypted ext4
When I boot with normal boot args, I end up in the dracut emergency shell with errors about not being able to find disks required for boot.
[ 195.269110] river.tirfa.net dracut-initqueue: Warning: /dev/disk/by-uuid/98b5df21-c4e6-44db-bec6-93e2b53dc751 does not exist
[ 195.269548] river.tirfa.net dracut-initqueue: Warning: /dev/mapper/luks-c626b926-7b66-41fd-9f1d-31ee2df530e1 does not exist
[ 195.269931] river.tirfa.net dracut-initqueue: Warning: /dev/mapper/luks-c626b926-7b66-41fd-9f1d-31ee2df530e1 does not exist
When I do some digging around, it turns out that the "non-existant" disks are the unlocked sdc3 and sdc4 partitions.
To make things even more interesting, if I add rd.debug to the boot args, my system boots to gdm without issue (reproduced twice).
Version-Release number of selected component (if applicable):
Command line: BOOT_IMAGE=/vmlinuz-3.11.5-300.fc20.x86_64 root=UUID=98b5df21-c4e6-44db-bec6-93e2b53dc751 ro vconsole.font=latarcyrheb-sun16 rd.lvm.lv=vg_storage/lv_swap vconsole.keymap=us rd.luks.uuid=luks-b0755693-19e0-4bbd-9091-5f7fa3e36d4f rd.luks.uuid=luks-c626b926-7b66-41fd-9f1d-31ee2df530e1 rhgb quiet LANG=en_US.UTF-8
Created attachment 813389 [details]
Created attachment 813390 [details]
Created attachment 813391 [details]
output of blkid -o udev
Created attachment 813421 [details]
boot.log from successful boot with rd.debug
[ 1.220850] river.tirfa.net systemd: Expecting device dev-disk-by\x2duuid-c626b926\x2d7b66\x2d41fd\x2d9f1d\x2d31ee2df530e1.device...
but it is never found although:
lrwxrwxrwx 1 root 0 10 Oct 17 08:07 c626b926-7b66-41fd-9f1d-31ee2df530e1 -> ../../sdc3
Can you run without rd.debug, but with "systemd.log_level=debug" and attach the rdsosreport ?
(In reply to Harald Hoyer from comment #5)
> Can you run without rd.debug, but with "systemd.log_level=debug" and attach
> the rdsosreport ?
Bah, I completely missed your comment until today. Apologies on the delay.
Of course, now that I'm trying to debug again, I can't reproduce it anymore. I'm closing this bug for now but will re-open if I start hitting the issue again.