Multithreading Segfaults :(

[Steven D. Morrey]

So I've been studying the problem at great length now, experimenting with different methods of locking this up,  and have come to the conclusion that the only maintainable solution is to do pretty much what Andreas says.
I didn't want to do it this way right off because it requires a substantial refactor of key areas of code, but in the long run it will be worth it, especially if later profiling reveals that other threads could be useful.
Unfortunately Nagios uses lots of globals, however the macro stuff is all I'm concerned with for now.
A simple grep of the sources (Nagios 2.7) shows 617 references to the word macro.
Many of these are close together and are good candidates for extraction into their own functions anyways.
For those that aren't, most of the stuff is direct access to the object, which could be fixed by disallowing access to the underlaying object directly and instead create a series of get/set wrapper functions with built in mutexes, effectively an API (although wrapper is what I'd be tempted to call it).

Anyways I'll get to work on this and as soon as I can, I'll try to do the same in the 3.x branch so the majority of folks can have a go at it seeing how it works for them.

Sincerely,
Steve

________________________________________
From: Steven D. Morrey [smorrey at ldschurch.org]
Sent: Thursday, August 20, 2009 9:15 AM
To: Nagios Developers List
Subject: Re: [Nagios-devel] Multithreading Segfaults :(

Right but that moves out of the area of a simple patch and into something significantly more complex.
I'm trying to avoid doing that, because we're still dealing with 2.7 here, but  I agree with you, it's the proper way of doing it.

If this works, when my employer moves to the 3x branch, I'll look at reworking it that way.

Sincerely,
Steve
________________________________________
From: Andreas Ericsson [ae at op5.se]
Sent: Thursday, August 20, 2009 3:00 AM
To: Nagios Developers List
Subject: Re: [Nagios-devel] Multithreading Segfaults :(

Steven D. Morrey wrote:
> I think I've got it mostly worked out now. Turns out lots of
> functions call eachother, thereby producing a race.  The trick seems
> to be to ONLY lock the sections of code dealing directly with macro_x
> and not just the entire function dealing with macro_x itself. I'm
> going to give it  12hrs of run time and then I'll get a patch up for
> testing.
>

The "right" way of doing that is to lock down the global arrays in a
single file and then publish a thread-safe API to manipulate it. The
thread-safeness can be taken care of by the simple expedient of adding
locks to start with, and later one can modify the functions so that
hostnames etc are added in a function-local stack-allocated array
instead, which could potentially increase performance quite a lot
(and also reduce memory fragmentation quite a bit).

--
Andreas Ericsson                   andreas.ericsson at op5.se
OP5 AB                             www.op5.se
Tel: +46 8-230225                  Fax: +46 8-230231

Considering the successes of the wars on alcohol, poverty, drugs and
terror, I think we should give some serious thought to declaring war
on peace.

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Nagios-devel mailing list
Nagios-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nagios-devel


 NOTICE: This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message.



------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Nagios-devel mailing list
Nagios-devel at lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nagios-devel

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with 
Crystal Reports now.  http://p.sf.net/sfu/bobj-july