Message ID | 524f69650901090818w7f367832vb9d0979714c3c609@mail.gmail.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Fri, 9 Jan 2009 10:18:28 -0600 "Steve French" <smfrench@gmail.com> wrote: > On Fri, Jan 9, 2009 at 5:33 AM, Jeff Layton <jlayton@redhat.com> wrote: > > On Thu, 8 Jan 2009 21:32:09 -0600 > > "Steve French" <smfrench@gmail.com> wrote: > > This patch doesn't remove any capability of the current code. It just > > gets rid of this kthread that doesn't do anything useful. There's > > nothing stopping us from putting it back later once we have working dir > > notification, but until then it's just doing unnecessary wakeups. > Al's patch removes the ability for a network or cluster file system to > handle directory change notification in the future ... (AFAIK KDE, > GNOME still use this fcntl) so we won't be able to fix the fcntl to > work where it matters most (network or cluster environments) in the > future without reverting the patch. > Right. By "this patch", I meant the one that I sent. Al's concerns seem valid though. You may be able to convince him to add back in a dir_notify type interface to the VFS later but I imagine it will need some redesign to address the problems he noted.
--- a/fs/dnotify.c +++ b/fs/dnotify.c @@ -115,9 +115,6 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg) dn->dn_next = inode->i_dnotify; inode->i_dnotify = dn; spin_unlock(&inode->i_lock); - - if (filp->f_op && filp->f_op->dir_notify) - return filp->f_op->dir_notify(filp, arg); return 0;