Fw: Re: Implemented OPLOCK for FreeBsd

Garrett Wollman wollman at khavrinen.lcs.mit.edu
Fri Sep 7 20:03:52 GMT 2001

<<On Fri, 7 Sep 2001 18:14:19 -0700, jra at samba.org (Jeremy Allison) said:

> Yep - see my other message. So this really boils down to
> POSIX.4 signals have a resource limit tunable, kqueues
> don't.

Two corrections:

1) There is no such thing as POSIX.4.  It's one of the lettered 1003.1
amendments, IIRC .1d.  (All of the real-time work was moved to one or
more .1 amendements.)  This functionality becomes the Realtime Signals
option of 1003.1-200x.

2) There is no requirement anywhere that {SIGQUEUE_MAX} be tunable.
The only requirement is that {SIGQUEUE_MAX} be no less than
{_POSIX_SIGQUEUE_MAX}, which is 32.  (The minimum number of signals
available to applications using this feature is {_POSIX_RTSIG_MAX} or
8.)  I can easily imagine an implementor wishing to constrain memory
allocation for realtime signals by restricting {SIGQUEUE_MAX}.

It's been suggested in other messages (but I have not verified) that
SIGINFO, the signal, is being used; this is of course not portable, so
I hope those other people are wrong.


More information about the samba-technical mailing list