Please read this before reporting a bug:
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
https://wiki.archlinux.org/title/Bug_reporting_guidelines
Do NOT report bugs when a package is just outdated, or it is in the AUR. Use the 'flag out of date' link on the package page, or the Mailing List.
REPEAT: Do NOT report bugs for outdated packages!
FS#22860 - [mkinitcpio] 0.6.8-1 broke autodetect fs detection
Attached to Project:
Arch Linux
Opened by Maximilian Kuhnke (mk) - Friday, 11 February 2011, 22:45 GMT
Last edited by Thomas Bächler (brain0) - Sunday, 27 March 2011, 11:06 GMT
Opened by Maximilian Kuhnke (mk) - Friday, 11 February 2011, 22:45 GMT
Last edited by Thomas Bächler (brain0) - Sunday, 27 March 2011, 11:06 GMT
|
DetailsDescription:
Consider the following partition layout: [sda1] (ext2) /boot [sda2] (luks -> ext2) /keys [sda3] (luks -> xfs) / On boot, I enter my passphrase to let the 'encrypt' hook decrypt and mount '/keys'. After that, a selfmade hook uses key files stored in '/keys' to decrypt and mount '/' and several other devices in the future. So when I generate the initramfs image using mkinitcpio, the 'install/autodetect' script is run. It will only add the filesystem kernel module required to mount '/', but not the one of '/keys' or any other device which will be mounted by my hook in the future. This obviously results in not being able to boot (without a fallback image). Note that this selfmade hook is used by several people. Additional info: since mkinitcpio 0.6.8-1 Commit which introduced the bug: http://projects.archlinux.org/mkinitcpio.git/patch/?id=05e0bc5987166573412f196ff932d76787d31957 Steps to reproduce: See description. Possible fix: Apply the following patch or invent some configuration procedure to let the user choose the devices which should be mountable on boot. But all in all this patch should be perfectly fine as it would basically be a revert of the changes done in the commit named under additional info. This patch will also correctly handle/ignore the bug described in https://bugs.archlinux.org/task/18585 It may add more filesystem modules to the images than necessary to boot but not as many as without 'autodetect', similar to what mkinitcpio did before 0.6.8-1. greetz :) |
This task depends upon
Closed by Thomas Bächler (brain0)
Sunday, 27 March 2011, 11:06 GMT
Reason for closing: Won't fix
Additional comments about closing: If another hook wants to support more file systems, the hook must take care of adding them.
Sunday, 27 March 2011, 11:06 GMT
Reason for closing: Won't fix
Additional comments about closing: If another hook wants to support more file systems, the hook must take care of adding them.
autodetect.patch
I personally think forcing me and others to put additional filesystem modules in the custom hooks would make the whole concept of hooks useless since the filesystem hook is supposed to add the modules. Don't get me wrong, I could accept this solution if the other one is not suitable for whatever reason.
Either way, it'll result in a hook which is no longer fully autodetecting.
If another hook wants to support more file systems, the hook must take care of adding them.