Re: why does rwlock prefer readers by default?

[Torvald Riegel <triegel at redhat dot com>]

On Wed, 2014-05-07 at 14:51 -0400, Rich Felker wrote:
> On Wed, May 07, 2014 at 12:33:51PM +0200, Torvald Riegel wrote:
> > POSIX makes it an implementation-defined choice whether readers or
> > writers are preferred.  Our current implementation's default is that
> > readers are to be preferred.  I couldn't find the rationale for this;
> > does anybody know what it was?
> > 
> > Otherwise, if this was an arbitrary choice, what do you all think the
> > default should be?  Can we change it?  Should we change it to preferring
> > writers?
> 
> As far as I know, there is no way to prefer writers but allow
> recursive locking by readers (which the standard requires be allowed)
> without unbounded memory usage (to track which threads already own a
> read lock). The problem is that you can't distinguish a new reader
> from an existing reader performing a recurive lock and thus you have
> to allow both, even if a writer is waiting. Please correct me if I'm
> wrong on this.

Yes you need space to track which readers have acquired which rwlock.
At some number of recursive/rwlocks, there will be a runtime overhead.

Irrespective of that, there's another case where we can prefer readers
or writers, as I've described elsewhere in the thread.