[Bug 1083723] Re: 'telinit u' has a cage fight with busybox init

[Steve Langasek]

Since we are hoping to obsolete the /var/run/init.upgraded interface now
that we have stateful re-exec capabilities, this should be fixed for
raring.

** Changed in: upstart
       Status: New => Triaged

** Changed in: upstart
   Importance: Undecided => Medium

** Also affects: upstart (Ubuntu)
   Importance: Undecided
       Status: New

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to upstart in Ubuntu.
https://bugs.launchpad.net/bugs/1083723

Title:
  'telinit u' has a cage fight with busybox init

Status in Upstart:
  Triaged
Status in “upstart” package in Ubuntu:
  New

Bug description:
  A number of packages (e.g. eglibc, libsepol) use 'telinit -u' in their
  postinst to request that the init daemon be restarted.  Now, in Ubuntu
  we've generally hacked these to use 'touch /var/run/init.upgraded'
  instead to avoid issues such as bug 672177, but with stateful-reexec
  hopefully we'll be able to get rid of these hacks once things like bug
  1079715 have been fixed.

  However, there's another problem.  'telinit -u' sends SIGTERM to pid
  1.  This amounts to a private IPC method, but it is not one agreed
  upon by all init daemons.  busybox init (not unreasonably, I think)
  defines SIGTERM to mean "reboot", and busybox init is used by debian-
  installer.  This means that if we were ever to produce a Debian
  installer image that installed Upstart by default - which is certainly
  something we want to attempt once it's worthwhile - it will reboot as
  soon as it tries to configure any of the affected packages.

  Perhaps telinit should check whether pid 1 is in fact Upstart before
  sending these signals?  Or perhaps it should use some other IPC
  mechanism.

To manage notifications about this bug go to:
https://bugs.launchpad.net/upstart/+bug/1083723/+subscriptions