Bug 560066 – >=sys-fs/udev-init-scripts-28 breaks the boot process

• Product: Gentoo Linux
• Status: RESOLVED OBSOLETE

[Description Richard Yao (RETIRED) 2015-09-09 18:56:58 UTC]

The following commit changed things to stop calling udevadm trigger and udevadm settle by default:

https://gitweb.gentoo.org/proj/udev-gentoo-scripts.git/commit/?id=ebb3f35bd06591bf7092cfc61e5cd51857ed888f

This breaks anything that needs those to work, such as localmount when symlinks in /dev/disk are specified. It also will break this week's ZoL 0.6.5 release which will stop calling udevadm settle itself because the udev script was originally supposed to do it. There is a report in IRC from Patrik Lauer that alsasound initialization is also broken.

I think we should revert that commit until the ideas can be implemented in a way that does not break the boot process.

[Comment 1 Mark Wagner 2015-10-20 00:08:39 UTC]

"It also will break this week's ZoL 0.6.5 release which will stop calling udevadm settle itself because the udev script was originally supposed to do it."

Are you sure about this? According to https://github.com/zfsonlinux/zfs/releases/tag/zfs-0.6.5 the commit is https://github.com/zfsonlinux/zfs/commit/2a34db1

The function udev_trigger() in etc/init.d/zfs-functions calls udevsettle.

The full call graph is etc/init.d/zfs-import:do_start() -> etc/init.d/zfs-functions:checksystem() -> etc/init.d/zfs-functions:udev_trigger()

[Comment 2 William Hubbs 2015-12-02 17:32:44 UTC]

I am looking for updates on this bug as follows:

(In reply to Richard Yao from comment #0)
> The following commit changed things to stop calling udevadm trigger and
> udevadm settle by default:
> 
> https://gitweb.gentoo.org/proj/udev-gentoo-scripts.git/commit/
> ?id=ebb3f35bd06591bf7092cfc61e5cd51857ed888f

Udevadm trigger is still called, but you are correct that udevadm settle is not.

> This breaks anything that needs those to work, such as localmount when
> symlinks in /dev/disk are specified. It also will break this week's ZoL

The preferred way to write fstab now is to use UID= instead of specifying anything in /dev. Is that broken with these scripts?

> 0.6.5 release which will stop calling udevadm settle itself because the udev
> script was originally supposed to do it. There is a report in IRC from
> Patrik Lauer that alsasound initialization is also broken.

Patrick has yet to file a bug about this.

(In reply to Mark Wagner from comment #1)
> "It also will break this week's ZoL 0.6.5 release which will stop calling
> udevadm settle itself because the udev script was originally supposed to do
> it."
> 
> Are you sure about this? According to
> https://github.com/zfsonlinux/zfs/releases/tag/zfs-0.6.5 the commit is
> https://github.com/zfsonlinux/zfs/commit/2a34db1
> 
> The function udev_trigger() in etc/init.d/zfs-functions calls udevsettle.
> 
> The full call graph is etc/init.d/zfs-import:do_start() ->
> etc/init.d/zfs-functions:checksystem() ->
> etc/init.d/zfs-functions:udev_trigger()

There has been no response to this question yet.

[Comment 3 William Hubbs 2016-11-10 18:02:09 UTC]

In the last couple of years, I have seen no comments on this bug.

We now officially direct people to use the tag syntax in fstab instead
of /dev/disk-by/*.

I never saw a report regarding the alsasound issue.

Also, wrt zol, I never saw an answer to the question I posed about it,
and there is a related bug, bug #565920.

If there are still issues, please feel free to let me know.