← Back to team overview

registry team mailing list archive

[Bug 581291] [NEW] upstart (/sbin/init) fails to start many /etc/init and /etc/rc2.d files

 

*** This bug is a security vulnerability ***

You have been subscribed to a public security bug:

A virgin install of 10.4 was failing to start many of the items in /etc/init and /etc/rc2.d, or least that was the symptom.
Example:  at the end of boot the  getty  processes on /dev/tty[1-6] where  _often_  not present.
Example:  my home-made entries in /etc/rc2.d apparently did not run (and they have worked for years).

What's different?  This is may first AMD 64-bit quad-processor and it
runs very much faster than anything I've ever had before.

(Many days of debugging.)

I think the problem is that Upstart's tendency to run many things in
parallel means that sometimes subtle dependencies can get caught by race
conditions.  E.g  if event A completes before event B starts, all is
well, but the other order causes problems.

My get-around has been to add the  init='/sbin/int  --verbose'   to the  linux /boot/vmlinuz/...  line in  /boot/grub/grub.cfg  by
making my own  menuentry  item in  /etc/grub.d/08_custom .
See  man 8 init .
This seems to have the effect of slowing down Upstart, and possibly single-threading some of the startups (because they are held back by the queue of messages to the console?).
((Once again, adding the debugging printout _prevents_ the problem you are trying to debug.))

At this point I'm happy to just leave the  --verbose  option in, since
now my machine seems to boot reliably.


Suggestion:  Upstart needs a formal  --single-thread  option to get around just this sort of problem.  Give me a slow boot that always works any day!

I've marked the  security vulnerability  box because any security boot
script that does not run properly could be a problem.

** Affects: null
     Importance: Undecided
         Status: Invalid

-- 
upstart (/sbin/init) fails to start many /etc/init and /etc/rc2.d files
https://bugs.edge.launchpad.net/bugs/581291
You received this bug notification because you are a member of Registry Administrators, which is the registrant for NULL Project.