Re: tail + inotify over nfs

[Pádraig Brady]

On 16/12/09 12:36, Jim Meyering wrote:
> Pádraig Brady wrote:
> > I got a few minutes to look at this today,
> > and the attached patch seems to work with a very quick test.
> >
> > It doesn't handle the above remount case though
> > as if I mount the parent dir of a file or bind mount the file itself
> > then there are no inotify notifications. This remounting issue is
> > independent of nfs anyway. So can inotify handle this or will we
> > have to periodically check with a select rather than a blocking read?
>
> Thanks for starting on this.
> Please filter through cppi to indent the new cpp directives.
> I should make syntax-check automate that check.  There used to be
> a cvs commit hook to enforce it.
>
> Also, please spell "file system" with two words, not one,
> to get by the syntax-check for that.
>
> Not to look the gift horse in the mouth, but what do you
> think about adding a test for this?  Maybe we can exercise
> it via a FUSE-based file system.  That should be easier than
> setting up NFS.

I had a quick look at this and it seemed messy as
I couldn't see any builtin fuse filesystems I could use.

Attached is the latest version with the unlikely edge cases that
aren't handled mentioned in this comment:

  FIXME: inotify doesn't give any notification when a new
  (remote) file or directory is mounted on top a watched file.
  When follow_mode == Follow_name we would ideally like to detect that.
  Note if there is a change to the original file then we'll
  recheck it and follow the new file, or ignore it if the
  file has changed to being remote.

cheers,
Pádraig.

tail-nfs.diff
Description: Text Data