Re: [PATCH 2/2] uio: do not expose inode to uio open/release hooks

["Hans J. Koch" <hjk@hansjkoch.de>]

On Tue, Dec 11, 2012 at 08:46:48PM -0800, Greg KH wrote:
> On Wed, Dec 12, 2012 at 02:42:22AM +0100, Hans J. Koch wrote:
> > On Tue, Dec 11, 2012 at 03:20:32PM -0800, Greg KH wrote:
> > > On Wed, Dec 12, 2012 at 12:12:02AM +0100, Benedikt Spranger wrote:
> > > > The inode parameter is unused by in kernel users of UIO.
> > > 
> > > Ok.
> > > 
> > > > Also the inode parameter makes it hard to resolve the existing open(),
> > > > mmap() and close() difficulty.
> > > 
> > > I don't understand, what do you mean by this?  What is this parameter
> > > causing problems with?
> > 
> > The problem is that according to POSIX, it is guaranteed that in userspace
> > you can do
> > 
> > fd = open("/dev/uio0", ...)
> > ptr = mmap(...fd...)
> > close(fd)
> > 
> > with ptr still being valid and useable after that.
> 
> Yes, but what does that have to do with this in-kernel, internal api?

Ah, OK. You're right, the commit message is confusing.

Bene, it's enough to say we drop the inode parameter because nobody
ever needed it. I cannot see why this also helps with the other problem.

Thanks,
Hans