Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Tue, 18 Sep 2007 10:10:24 -0600

> Is this a reason to put a cookie in the epicsEventOSD structure, to
> check that it's a valid structure before using it? If the consensus is
> that it's worth it, I'll get hacking.

There are pluses and minuses with this sort of protection. 

Examining the minuses, even if there is a magic cookie in the data structure
it won't help us one bit if thread A clears the cookie before calling free
for the epicsEventOSD pointed to by pA, thread B calls allocate and receives
the same block initializing it with a epicsEventOSD cookie, and then thread
A flagrantly continues to use pA as an epicsEventId. With best fit
implementations of malloc and our heavy use of free list based memory
management reuse of the same block for its original purpose becomes
increasingly likely.

Another minus is of course increased data structure size and runtime
overhead. You could add checks like this to almost any data structure, and
at some point if everything is checked the overhead might actually become
significant.

Another contrary consideration is that programs like purify, bounds checker,
valgrind, and others are arguably superior at detecting use of memory after
it has been de-allocated. However, we must enable build options that
temporarily turn off use of free lists for these types of error checking
environments to be effective (they detect attempts to us memory after its
de-allocated by replacing the implementation of malloc, free, etc).

Looking at a plus consideration, perhaps, a good compromise might be to
place the sanity checking code inside of "assert(xxx)" expressions and
bracket any extra data structure members with #ifndef NDEBUG. We are
currently not turning off assert checks in optimized builds, but this
approach would allow for some level of control should the overhead become
oppressive ;-)

Jeff

> -----Original Message-----
> From: [email protected] [mailto:tech-talk-
> [email protected]] On Behalf Of Denison, PN (Peter)
> Sent: Tuesday, September 18, 2007 4:20 AM
> To: Denison, PN (Peter); TechTalk EPICS
> Subject: RE: Deadlock on epicsEvent
> 
> > I'm trying to debug a deadlock on Linux under R3.14.8.2, on a
> > cut down IOC with only asyn, motor and autosave installed.
> >
> > What appears to be happening is that one thread is waiting on
> > an epicsEvent, whilst another thread is trying to signal that
> > event, but getting stuck on the event's internal lock. The
> > pthread man pages imply that this shouldn't happen.
> 
> There's nothing quite like asking in public to find out you've been
> stupid, is there? Emerging from my brown-paper-bag, I've found the
> cause
> of the problem. However, I think there is still room for improvement,
> as
> the library code could have caught it, rather than deadlocking.
> 
> I was trying to signal an epicsEvent after having called
> epicsEventDestroy() on it. Therefore the lock it was trying to acquire
> was 0.
> 
> Is this a reason to put a cookie in the epicsEventOSD structure, to
> check that it's a valid structure before using it? If the consensus is
> that it's worth it, I'll get hacking.
> 
> --
> Peter Denison, Senior Software Engineer, Diamond Light Source Ltd.
> Tel: +44 1235 778511
> (apologies in advance for the lines below. Some bits are a legal
> requirement and I have no control over them)
> 
> <DIV><FONT size="1" color="gray">This e-mail and any attachments may
> contain confidential, copyright and or privileged material, and are for
> the use of the intended addressee only. If you are not the intended
> addressee or an authorised recipient of the addressee please notify us
> of receipt by returning the e-mail and do not use, copy, retain,
> distribute or disclose the information in or attached to the e-mail.
> Any opinions expressed within this e-mail are those of the individual
> and not necessarily of Diamond Light Source Ltd.
> Diamond Light Source Ltd. cannot guarantee that this e-mail or any
> attachments are free from viruses and we cannot accept liability for
> any damage which you may sustain as a result of software viruses which
> may be transmitted in or with the message.
> Diamond Light Source Limited (company no. 4375679). Registered in
> England and Wales with its registered office at Diamond House, Harwell
> Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United
> Kingdom
> </FONT></DIV>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System