Re: [patch 0/3] Re: tty contention resulting from tty_open_by_device export

From: Okash Khawaja
Date: Mon Jul 10 2017 - 08:33:16 EST

On Mon, Jul 10, 2017 at 12:52:33PM +0100, Alan Cox wrote:
> On Sun, 09 Jul 2017 12:41:53 +0100
> Okash Khawaja <okash.khawaja@xxxxxxxxx> wrote:
> > On Sat, Jul 08, 2017 at 10:38:03AM +0200, Greg Kroah-Hartman wrote:
> > > Overall, the idea looks sane to me. Keeping userspace from opening a
> > > tty that the kernel has opened internally makes sense, hopefully
> > > userspace doesn't get too confused when that happens. I don't think we
> > > normally return -EBUSY from an open call, have you seen what happens
> > > with apps when you do this (like minicom?)
> > >
> > I tested this wil minincom, picocom and commands like "echo foo >
> > /dev/ttyS0". They all correctly report "Device or resource busy".
> >
> > I have addressed all the comments you made. I have also split the patch
> > into three. Following is summary of each.
> If the tty counts are being misreported then it would be better to fix
> the code to actually manage the counts properly. The core tty code is
> telling you that the tty is not in a valid state. While this is of
> itself a good API to have, the underlying reference miscounting ought
> IMHO to be fixed as well.
When opening from kernel, we don't use file pointer. The count mismatch
is between tty->count and #fd's. So opening from kernel leads to #fd's
being less than tty->count. I thought this difference is relevant to
user-space opening of tty, and not to kernel opening of tty. Can you
suggest how to address this mismatch?

> Also you don't need a new TTY_KOPENED flag as far as I can see. All tty's
> have a usage count and active bit flags already. Use those.
Ah may be I didn't notice the active bit. Is it one of the #defines in
tty.h? Can usage count and active bit be used to differentiate between
whether the tty was opened by kernel or user?