[systemd-bugs] [Bug 90051] udevd have no measure to avoid exiting before all jobs are done

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=90051

--- Comment #3 from Tom Yan <tom.ty89 at gmail.com> ---
(In reply to Tom Gundersen from comment #2)
> What you should see is that once you call exit, udev exits "almost
> immediately".
This is true for `udevadm control --exit`, but if udevd is started and stopped
as a service, this is generally false. Here is an example, which you can see
that it waits for exactly 30s:

[tom at localhost ~]$ systemctl status -l systemd-udevd
● systemd-udevd.service - udev Kernel Device Manager
   Loaded: loaded (/usr/lib/systemd/system/systemd-udevd.service; static;
vendor preset: disabled)
   Active: active (running) since Fri 2015-04-24 19:36:02 HKT; 46s ago
     Docs: man:systemd-udevd.service(8)
           man:udev(7)
 Main PID: 194 (systemd-udevd)
   CGroup: /system.slice/systemd-udevd.service
           └─194 /usr/lib/systemd/systemd-udevd

Apr 24 19:36:02 localhost systemd-udevd[194]: starting version 219
[tom at localhost ~]$ systemctl status -l systemd-udevd
● systemd-udevd.service - udev Kernel Device Manager
   Loaded: loaded (/usr/lib/systemd/system/systemd-udevd.service; static;
vendor preset: disabled)
   Active: inactive (dead) since Fri 2015-04-24 19:37:26 HKT; 2s ago
     Docs: man:systemd-udevd.service(8)
           man:udev(7)
  Process: 659 ExecStart=/usr/lib/systemd/systemd-udevd (code=exited,
status=0/SUCCESS)
 Main PID: 659 (code=exited, status=0/SUCCESS)

Apr 24 19:36:55 localhost systemd-udevd[659]: starting version 219
Apr 24 19:36:55 localhost systemd[1]: Started udev Kernel Device Manager.
Apr 24 19:36:55 localhost systemd[1]: Stopping udev Kernel Device Manager...
Apr 24 19:36:55 localhost systemd-udevd[659]: worker [668] did not accept
message -1 (Connection refused), kill it
Apr 24 19:36:55 localhost systemd-udevd[659]: worker [672] did not accept
message -1 (Connection refused), kill it
Apr 24 19:36:55 localhost systemd-udevd[659]: worker [677] did not accept
message -1 (Connection refused), kill it
Apr 24 19:36:55 localhost systemd-udevd[659]: worker [682] did not accept
message -1 (Connection refused), kill it
Apr 24 19:36:55 localhost systemd-udevd[659]: worker [682] terminated by signal
9 (Killed)
Apr 24 19:37:25 localhost systemd-udevd[659]: timeout, giving up waiting for
workers to finish
Apr 24 19:37:26 localhost systemd[1]: Stopped udev Kernel Device Manager.

But during the wait, I only see the main pid with `watch -n 0.1 pgrep udev`.
(Well most of the wait, like 29.x seconds out of 30)

> Lastly, if if you want to make sure that udevd only exits after all the
> events caused by trigger have been processed, that is precisely what settle
> is for. It waits for udevd to receive the triggered events and then it waits
> for udevd to finish procssing them all (new events may appear at any time,
> but we make no guarantees about those).

It doesn't seem to work in the way you said. For example in this script:
https://projects.archlinux.org/svntogit/packages.git/tree/trunk/initcpio-hook-udev?h=packages/systemd
I assume that `udevadm settle` should make sure that events triggered by the
two commands before it should be finished? But from what I am experiencing it
does not.

For example, a number of kernel modules is included in the initramfs. Suppose
given enough time (with `sleep` command), 26 of them would be loaded in early
user space, but if only `udevadm settle` is used, the number could be random.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/systemd-bugs/attachments/20150424/5ad8863c/attachment.html>