Re: [PATCH 0/2] SIGWINCH problem with terminal apps still alive

["Adam Tlałka" <atlka@pg.gda.pl>]

Sun, 12 Oct 2008 19:03:12 +0100 - Alan Cox <alan@lxorguk.ukuu.org.uk>:
> Then we end up with both using real_tty.
> 
> > Anyway I think that you miss the point. Why using
> > real_tty->termios_mutex instead of tty->termios_mutex in
> > tty_do_resize
> 
> To avoid deadlocks if you took both as you updated both structures.

Actually to avoid race and not valid winsize reading because
ioctl(TIOCGWINSZ) is done with tty == real_tty so we have to use this
mutex and not the other one.
What is more because we are using real_tty we could safely use only
real_tty->winsize and forget about master tty->winsize.
 
> > So it seems that tty->termios_mutex could point to different
> > location in different calls but real_tty->termios_mutex always
> > points to the same location.
> 
> You've finally got there - we always work off real_tty. That is why we
> can safely use the mutex on the real_tty side.

typically ioctl(,TIOCSWINSZ,) is called on master side but
ioctl(,TIOCGWINSZ,) is called on slave side so tty is not the same in
both calls inside ioctl() in tty_io.c. Only real_tty variable has the
same value for the same master/slave pair. So we must use real_tty
mutex all the time because we not always work from the same side IMHO.

That is why I propose usage of real_tty in ioctl() handling:

case TIOCGWINSZ:
                return tiocgwinsz(real_tty, p);


Regards

-- 
Adam Tlałka       mailto:atlka@pg.gda.pl    ^v^ ^v^ ^v^
System  & Network Administration Group       - - - ~~~~~~
Computer Center, Gdańsk University of Technology, Poland