Bug 681586

Summary:	=sys-fs/udev-init-scripts-33 with sys-fs/udev - system in unstable state after udev restart
Product:	Gentoo Linux	Reporter:	Tomáš Mózes <hydrapolic>
Component:	Current packages	Assignee:	udev maintainers <udev-bugs>
Status:	RESOLVED FIXED
Severity:	normal	CC:	dan, gentoo_eshoes
Priority:	Normal
Version:	unspecified
Hardware:	All
OS:	Linux
See Also:	https://bugs.gentoo.org/show_bug.cgi?id=678638
Whiteboard:
Package list:		Runtime testing required:	---

Description Tomáš Mózes 2019-03-23 23:27:59 UTC

I've distfiles on a glusterfs server and is mounted to save bandwidth during regular updates. As part of regular maintenance after a @world update I restart the services (via app-admin/restart-services). After upgrading to =sys-fs/udev-init-scripts-33 this causes random process killing (most notably syslog-ng and sshd).

How to reproduce:
mount /usr/portage/distfiles
emerge util-linux
restart-services (restarts udev)
umount /usr/portage/distfiles
Connection to ssh.example.com closed by remote host.
Connection to ssh.example.com closed.

The killing combination is =sys-fs/udev-init-scripts-33 with sys-fs/udev
=sys-fs/udev-init-scripts-32 with sys-fs/udev = OK
=sys-fs/udev-init-scripts-3[23] with sys-fs/eudev = OK

Comment 1 Daniel M. Weeks 2019-04-03 21:19:06 UTC

I saw the same behavior where sshd died when unmounting an NFS filesystem. It is no longer an issue after downgrading to sys-fs/udev-init-scripts-32.

Comment 2 Tomáš Mózes 2019-04-03 21:37:24 UTC

I "resolved" it by migrating to eudev. Now even udev-init-scripts-33 works fine.

Comment 3 David Flogeras 2019-05-06 19:16:08 UTC

This has been plaguing me since Feb as well and I just found this bug.  I can't give proof positive it is fixed, but I can say that it only seemed to happen on machines that were old enough the udev was still the default when they were installed.  My newer hardware runs eudev and none of them had issues.

Most of mine manifested in ways like
 - random lvmetad crashes
 - sshd would crash when (among other things) mounting filesystems.  This was particularly painful on headless machines.

Thanks Tomáš for the fix/workaround!

Comment 4 Mike Gilbert gentoo-dev

2020-08-12 14:45:57 UTC

In udev-init-scripts-33, the --daemon option was removed from the udevd command line. Perhaps this triggers some change in behavior in systemd-udevd.

https://gitweb.gentoo.org/proj/udev-gentoo-scripts.git/commit/?id=6f98cf89f54a9ad1c9625577d100f6ca8b8a2b0b

Comment 5 Mike Gilbert gentoo-dev

2020-08-12 14:51:24 UTC

Are you able to reproduce this with >=sys-fs/udev-241?

https://github.com/systemd/systemd/commit/31cbd2025359e1e8435a2dc371e439591846d8c4

Comment 6 Tomáš Mózes 2020-08-14 05:31:26 UTC

I'm unable to reproduce on a current (~)amd64 system.

sys-fs/udev-245.5-r1 + sys-fs/udev-init-scripts-33 OK
sys-fs/udev-246-r1 + sys-fs/udev-init-scripts-34 OK