kern/118626: Bad interaction between SIGPIPE and threads.

[Mark Andrews]

> David Schultz wrote:
> > On Thu, Dec 13, 2007, Mark Andrews wrote:
> >> The following reply was made to PR kern/118626; it has been noted by GNATS
> .
> >>
> >> From: Mark Andrews <Mark_Andrews at isc.org>
> >> To: bug-followup at FreeBSD.org, marka at isc.org
> >> Cc:  sobomax at freebsd.org
> >> Subject: Re: kern/118626: Bad interaction between SIGPIPE and threads.
> >> Date: Thu, 13 Dec 2007 11:41:16 +1100
> >>
> >>  I'll definitely try it.
> >>  
> >>  The man page for socket(2) should be updated to reference SO_NOSIGPIPE
> >>  where it talks about SIGPIPE being generated.
> > 
> > The commit log says SO_NOSIGPIPE exists for compatibility, but it
> > looks fully implemented and supported. I've cc'd the original
> > author to find out if there's any good reason not to document it.
> 
> It definitely should be documented fully.  Converting threaded 
> applications to use this can significantly reduce the lock contention 
> associated with signal delivery and associated processing, and may 
> improve performance.
> 
> Kris

FYI: libisc and libbind will both being using SO_NOSIGPIPE when it
is available as it is the only way to make the libraries safe in a
threaded envirionment without impacting on the applications use of
SIGPIPE.

Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: Mark_Andrews at isc.org