Bug 3967

Summary:	Kernels with dracut do not boot
Product:	Mageia	Reporter:	Olav Vitters <olav>
Component:	RPM Packages	Assignee:	Colin Guthrie <mageia>
Status:	RESOLVED FIXED	QA Contact:
Severity:	critical
Priority:	Normal
Version:	Cauldron
Target Milestone:	---
Hardware:	x86_64
OS:	Linux
Whiteboard:
Source RPM:	dracut-014-9.mga2.src.rpm	CVE:
Status comment:
Attachments:	Error message by dracut

Description Olav Vitters 2011-12-31 13:34:09 CET

Description of problem:
Any kernel with an initrd made with dracut doesn't boot. It says it cannot find the root device, while a ls -l actually shows the device in /dev. Only guess is perhaps it is due to the block: before the root device. Maybe it searches that path?


Version-Release number of selected component (if applicable):
014-9.mga2

How reproducible:
Try to boot latest kernel

Comment 1 Colin Guthrie 2011-12-31 13:36:20 CET

This is just kernel command line issues. Not tracked it down yet, but just remove the resume= bit from the command line and it should boot fine.

Comment 2 Olav Vitters 2011-12-31 13:38:51 CET

Created attachment 1313 [details]
Error message by dracut

If you want me to try some commands, tell me.

Comment 3 Olav Vitters 2011-12-31 13:49:46 CET

oh.. you already responded :) damn quick. will try

Comment 4 Colin Guthrie 2011-12-31 13:53:14 CET

Actually one thing you could try:

 a) Remove the whole resume= argument and check that it then boots.
 b) Double check what the UUID in the resume= argument says and compare it to the output from the blkid command. It should match your swap, but I suspect it does not.... changing it to match that of your swap partition should then make it boot.

The primary problem I noticed is that the kernel command line UUID simply didn't match my swap space UUID and thus it was always waiting for a device that just doesn't exist. I think this is valid, but it should just continue with regular boot if the swap partition can not be found rather than bail hard!

Comment 5 Olav Vitters 2011-12-31 13:59:26 CET

I removed the whole resume= stuff and it boots!

The UUID in the resume is 327c5e86-2993-4f0f-baba-58462e3afa21. I don't have such an UUID partition (nothing in /dev/disk/by-uuid, nor /etc/fstab). Pretty sure that is some old disk (have switched hard drives).

So somewhat of a self-created problem. But would be nice if something would catch this (either fix the resume=, or warn me).

Comment 6 Colin Guthrie 2011-12-31 14:18:48 CET

Yeah, others have reported this as well. I even ran into it myself while generally doing bad things on my VM... so it does need to be fixed. I know how to do it, so it should be easy enough :)

Comment 7 Colin Guthrie 2011-12-31 14:47:19 CET

OK, so this should be fixed in latest dracut (10.mga2). I also fixed the bogus error message about the root device not being found.

There are still a bunch of warning spat out due to udevadm command syntax that need to be fixed, but the general principle of waiting for the device, timing out, issuing a warning, removing the wait for device and then continuing with normal boot works fine. Will submit upstream.

Status: NEW => RESOLVED
Resolution: (none) => FIXED