[svlug] (forw) [DNG] Linux system can be brought down by sending SIGILL to Systemd

Rick Moen rick at linuxmafia.com
Sat May 25 10:53:03 PDT 2019


Quoting Michael Eager (eager at eagerm.com):

> Kill(2) sends an unaccepted signal to proc 1, in apparent contradiction 
> to its man page, and some versions of init die and some don't.  That 
> doesn't sound like a problem with the init process.

Phrased that way, my off-the-cuff reaction is that I'd be a bit worried
about both of them -- but I'd be particularly glad to have an init that 
doesn't react that way, when sent an unaccepted signal.  

Among leading init systems, I happen to be fond of OpenRC, and am
warming to s6.  It's a fair point that SysVInit is crufty and ripe for
being given the heave-ho (as is BSD init, albeit IMO less so).  As an
aside, that isn't actually an argument for systemd specifically, any
more than not liking pastrami is an argument specifically for pepperoni.
(Just sayin'.  Not a comment inspired by anything _you_ said, BTW.)  

We used to get that rhetorical dodge from DJB acolytes:  'BIND sucks,
therefore you should run djbdns.  Sendmail sucks, so you should run
qmail.'  It was always a bit insulting to infer the speaker though we
were just too dim to be aware of Postfix, Exim4, Courier-MTA, NSD,
Unbound, PowerDNS, MaraDNS, Deadwood, etc., etc.   I believe the
technical term is 'Fallacy of the excluded middle', or, less formally,
'false dilemma'.

Eh, now I really _do need to get AFK and start cooking for company.



More information about the svlug mailing list