Bug 290629

Summary:	sys-fs/cryptsetup: manually opened devices are not closed, causing sys-fs/lvm2 "read failed" on shutdown
Product:	Gentoo Linux	Reporter:	Toralf Förster <toralf>
Component:	Current packages	Assignee:	Gentoo's Team for Core System packages <base-system>
Status:	RESOLVED FIXED
Severity:	normal	CC:	proxy-maint
Priority:	High
Version:	unspecified
Hardware:	All
OS:	Linux
Whiteboard:
Package list:		Runtime testing required:	---
Attachments:	screen shot

Description Toralf Förster gentoo-dev

2009-10-26 20:37:13 UTC

While helping w/ http://www.faulhammer.org/archiv-mainmenu-31/35-gentoo/307-testers-wanted-for-x86 I observed an issue (see attached photo).


Reproducible: Always

Steps to Reproduce:
1. boot my ThinkPad T400 with "reboot=pci init 2"
2. modprobe dm-crypt
3. cryptsetup luksOpen /dev/sdb2 media
4. mount -o noatime /dev/mapper/media /mnt/media
5. umount /mnt/media
6. reboot

Comment 1 Toralf Förster gentoo-dev

2009-10-26 20:37:55 UTC

Created attachment 208351 [details]
screen shot

Comment 2 Toralf Förster gentoo-dev

2009-10-26 21:03:25 UTC

important to reproduce it : use an external USB drive, plug it in after login as root, plug out it before reboot

Comment 3 Robin Johnson archtester

2009-10-30 19:01:04 UTC

if, right before you reboot, you shutdown the luks, does the problem go away?
I'm thinking it's due to dm-crypt mappings not being removed, and then when device-mapper tries to finally stop, those mappings still exist.

Comment 4 Toralf Förster gentoo-dev

2009-10-31 10:45:00 UTC

(In reply to comment #3)
> if, right before you reboot, you shutdown the luks, does the problem go away?
yes, luksClose removes thuis issue

Comment 5 Robin Johnson archtester

2009-11-04 07:29:58 UTC

This is not an LVM problem then. The dm-crypt scripts should probably be closing ALL dm-crypt devices, not just those registered in /etc/conf.d/dm-crypt.

vapier: ACK on closing all dm-crypt devices?

Comment 6 Robin Johnson archtester

2010-08-03 18:17:18 UTC

vapier: reping

Comment 7 SpanKY gentoo-dev

2011-01-07 23:32:19 UTC

yes, that makes sense