Re: [PATCH] keyspan: init termios properly

[Borislav Petkov <bbpetkov@yahoo.de>]

On Sun, Dec 02, 2007 at 01:57:34PM +0000, Alan Cox wrote:
> On Sun, 2 Dec 2007 09:03:40 +0100
> Borislav Petkov <yme@gollum.tnic> wrote:
> 
> > On Fri, Nov 30, 2007 at 09:23:43AM -0800, Lucy McCoy wrote:
> > > Hi All,
> > >
> > > I've been too busy to get to this but i'd rather not use the code fragment 
> > > i sent Boris to try.  It would be better to go ahead with the tty setup if 
> > > the pointer is not NULL, otherwise use the defaults and not reference the 
> > > NULL pointer. I'll see if I can get to this today.
> > you might also want to take a look at usa90_indat_callback() - there is
> > port->tty accessed again and this time it kills the machine completely when
> > something is sent over the serial line from the other end (through minicom, for
> > example). Oops is at
> > 
> > http://www.screenshots.cc/view_image/ab9a1837/cimg00772.jpg
> 
> Change it to if (tty && urb->actual_length) in that function and it'll
> probably cure it
yep, this does the trick, however, this is just a brown paper bag over the issue, afaict.
Most of the usb serial drivers do something in the likes of

	struct usb_serial_port *port = (struct usb_serial_port *) urb->context;
	...
	tty = port->tty;
	...
	tty_insert_flip_{char,string,..}(tty,..)

in their _indat_callback()s.

In the keyspan case the port->tty is being accessed already in the ->open() function
from usb_console_setup(), before setting the termios settings and later again in the
_indat callback. I'm clearly missing the big picture here so what is the proper way
to initialize termios settings correctly in a driver?

-- 
Regards/Gruß,
    Boris.