[OpenAFS-devel] Was: [OpenAFS] set client to refuse byte-range locks?

[Jeffrey Hutzelman]

On Sunday, January 28, 2007 04:32:02 PM -0500 Matt Benjamin 
<matt@linuxbox.com> wrote:

> Somewhat tested, only on i386 Linux 2.6.  Not reviewed.  Uses C ioctl
> space (seems like should be O, but that's not really in 1.4.2, at least).
> Not what we'll be configuring in future, really, because Unix client
> locking options are increased with new features.  Still, people should
> have the freedom to disable "no-op" byte range locks.

The way to get pioctl numbers in the 'C' space is to send mail to 
registrar@grand.central.org (and CC me, so your request doesn't get lost in 
all the spam that address gets).  If you make up your own numbers, you risk 
colliding with values that have been assigned for other uses in code you 
don't see.  The same goes for Rx service ID's and procedure numbers.

FWIW, I'm inclined to agree that this should come from the 'O' space, since 
it's rather OpenAFS-specific.  On the other hand, there's also something to 
be said for adding a more general interface for toggling optional features, 
instead of inventing a new pioctl every time.

-- Jeff