Re: nmdm(4) does not call .l_close

On 7/26/07, Julian Elischer <julian_at_elischer.org> wrote:
> Maksim Yevmenkin wrote:
> > Dear All,
> >
> > it seems to me that nmdm(4) is not calling .l_close (i.e. does not
> > close whatever line discipline might be installed onto /dev/nmdmXX).
> > the problem is easy to reproduce - just open /dev/nmdm0A and install,
> > say, ng_tty(4) line discipline onto it. after that, simply close the
> > /dev/nmdm0A. in theory, the ng_tty(4) node should disappear when
> > device is closed, but it does not.
>
> when I originally wrote this I copied what the pty code was doing.
> This has all been redone since then however.
> I'm guessing it should do whatever the pty code is now doing :-) .

it looks like pty close calls ttyld_close() before calling
tty_close(). so at the very least, imo, nmdm close should call
ttyld_close() before tty_close().

ttyclose(), infact, calls ttyld_close() and tt_close(), so it seemed
like the right thing to do :)

> > the following patch fixes things for me
> >
> > ===
> >
> > beetle% diff -u nmdm.c.orig nmdm.c
> > --- nmdm.c.orig 2006-11-21 16:59:40.000000000 -0800
> > +++ nmdm.c      2007-07-26 17:01:47.000000000 -0700
> > _at__at_ -402,7 +402,7 _at__at_
> >   nmdmclose(struct cdev *dev, int flag, int mode, struct thread *td)
> >  {
> >
> > -       return (tty_close(dev->si_tty));
> > +       return (ttyclose(dev, flag, mode, td));
>
> sounds believable..
> but I don't really know offhand.

nor do i :) hence my question. can someone, with more tty fu, please look at it?

> does it work with other disciplines?

it does with ng_h4(4). since i cannot make my bluetooth xircom pccard
work, i decided to use my old trick - emulate bluetooth device. so i'm
testing my patches to make ng_h4(4) mp-safe by installing h4 line
discipline onto /dev/nmdm and feeding h4 packets to it.

i have not tried other disciplines yet, i.e. ppp and slip,

thanks,
max