Fw: Re: Implemented OPLOCK for FreeBsd

[Jeremy Allison]

On Fri, Sep 07, 2001 at 08:20:25PM -0500, Jonathan Lemon wrote:

> No.  Essentially what happens is when you register the event
> with the kernel, (the equivalent of fcntl(fd, F_GETLEASE,...)),
> resources are allocated at that time, and failures reported to
> the app.  From that point on, no failures will happen because 
> all required resources have been allocated.

Fair point. This could be done with a very smart POSIX.4
implementation that treated the 'request lease' fcntl in
a special way (ensured that there were enough signal queue
slots for the incoming signal) but that would be really
ugly inside a kernel :-(.

Jeremy.