[SM-Grimoire-Bugs] [Bug 13114] simpleinit doesn't stop services during shutdown on recent kernels

bugzilla-daemon at bugs.sourcemage.org bugzilla-daemon at bugs.sourcemage.org
Tue Feb 27 10:57:56 EST 2007


http://bugs.sourcemage.org/show_bug.cgi?id=13114





------- Additional Comments From seth at swoolley.homeip.net  2007-02-27 09:57 -------
That there's a 12 second delay in the non-working example and a 3 second delay
in the working example seemed to invalidate the "it's simply not waiting"
theory.  Jeremy, can you confirm that the pause was here?

[pid 11220] rt_sigsuspend([])           = ? ERESTARTNOHAND (To be restarted)
                           ^
     (pause should be here ^ to help invalidate the simply not waiting theory)

Now, it's still possible the suspend didn't actually do any suspending and
didn't allow init to get any cycles during the pause, although highly unlikely
in my mind (SIGUSR1 was received), so a sleep would still be needed to
absolutely confirm this.

kill -1 doesn't send to the process itself or init, so init should have still
been able to spawn shutdown later, even after the kills, unless it were timed
perfectly to kill rc before rc got any messages out at all.

And my comments weren't attempting to indicate why there was no possibility that
Andrew's theory was the correct one.  I was merely analyzing independently so as
to give another possible theory, and I suggested that the reason init didn't do
its work wasn't clear enough yet and I requested more data.  Note that SIGUSR1
is received by initctl from somewhere in both cases, so I suspect there's
activity going on outside of initctl while it's trying to shutdown.  It's quite
possible that there was another independent change in the kernel that caused the
error, but we're not looking at it right now because this one change is
different, so we're focusing on it.  I just want to see what's actually
happening inside init.  BTW, Andrew, you're free to look at whatever you want,
as I don't presume to say you or anybody else can't look at something that I
won't look down yet without more complete analysis.  I explicitly said "perhaps"
because anything at this point is conjecture without more complete information.

strace takes a -p pid command to attach to a pid.  It can attach to up to 32
processes according to the manual.  I'm not going to try attaching to the box
I'm typing on, but that should work to attach to init.

-- 
Configure bugmail: http://bugs.sourcemage.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.



More information about the SM-Grimoire-Bugs mailing list