Re: [PATCH] vfs: new O_NODE open flag

From: Miklos Szeredi
Date: Fri Sep 25 2009 - 08:18:32 EST


On Fri, 25 Sep 2009, Dr. David Alan Gilbert wrote:
> * Miklos Szeredi (miklos@xxxxxxxxxx) wrote:
> > On Fri, 25 Sep 2009, Andreas Gruenbacher wrote:
> > > On Thursday, 24 September 2009 16:51:58 Miklos Szeredi wrote:
> > > > This patch adds a new open flag, O_NODE. This flag means: open just
> > > > the filesystem node instead of the object referenced by the node.
> > >
> > > What is the intended use for O_NODE?
> >
> > It lets userspace file descriptors reference a inode without actually
> > "dereferencing" it to get the underlying object. This allows for a
> > couple of new things:
> >
> > - opening a special file (device/socket/fifo) without side effects
> >
> > - opening a symlink
> >
> > - opening any type of file without any permission is also possible
> > (of course the resuling file descriptor may not be read or written)
> >
> > The above allows fstat(), fchmod(), ioctl(), etc to be used for files
> > previously not possible.
>
> Given an fd opened in this way is it possible to reopen it normally and
> be guarenteed to get the same object?

For directories it should be possible with:

fd = open(path, O_NOACCESS | O_NODE);
fd2 = openat(fd, ".", O_RDWR);

We could implement something similar for non-directories as well, by
treating a NULL path parameter to openat() to mean - reopen:

fd2 = openat(fd, NULL, O_RDWR);

Comments?

Thanks,
Miklos
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/