[RFC][PATCH] Inotify kernel API

[John McCutchan]

On Fri, 2005-08-26 at 10:37 -0500, Timothy R. Chavez wrote:
> On Wednesday 24 August 2005 15:24, Amy Griffis wrote:
> > On Wed, Aug 24, 2005 at 10:21:46AM -0700, Chris Wright wrote:
> > > * Timothy R. Chavez (tinytim at us.ibm.com) wrote:
> > > > The alternative approach is to embed Inotify watches in a per-client
> > > > specific watch, ie:
> > > 
> > > Exactly, this can eliminate typeless interface, and is how other
> > > subsystems do things.
> > 
> > It's my impression that John does not want to expose the inotify_watch
> > struct to kernel consumers.  But maybe there's a way to handle it
> > that's better than what I proposed.
> 
> Is there any particular reason why, if there were a kernel API for Inotify,
> that clients of that API should not be exposed to relevant Inotify structs
> like inotify_watch?

There is nothing in inotify_watch that is useful to other parts of the
kernel other than the inode, which will be passed to the callback.

-- 
John McCutchan <ttb at tentacle.dhs.org>