Re: filesystem change monitoring

[Masao Uebayashi <uebayasi%gmail.com@localhost>]

If I were to do this, I'd add kevent handling to struct mount, and
notify when page caches are committed to be written out to a disk
(genfs_putpages()).  Per-directory notification, being aware of
subdirectories, sounds a bit too complex to me.  I personally hope
that such a complex feature will not be added to VFS. :p

On Tue, Aug 11, 2015 at 11:44 PM, Emmanuel Dreyfus <manu%netbsd.org@localhost> wrote:
> Hello
>
> Do we have a smart way to monitor a whole filesystem hierarchy for
> changes? kqueue seems a good answer at a glance, but as far as I
> understand, I need to setup a filter for every single file and
> directory in the hierachy, and this can quickly hit kern.maxfiles.
> Of course kern.maxfiles can be increased, but that hints this solution
> may not scale very well.
>
> I also though about crawling on the hierarchy using FTS functions and
> look for ctime/mtime newer than previous crawl, but that can be CPU
> consuming.
>
> Any idea how this can be done efficiently?
>
> --
> Emmanuel Dreyfus
> manu%netbsd.org@localhost