Re: any chance of 2.6.0-test*?

[Andi Kleen]

On Sat, Jan 11, 2003 at 02:01:51PM +0100, Russell King wrote:
> > --- linux-2.5.56-work/drivers/char/tty_io.c-o	2003-01-02 05:13:12.000000000 +0100
> > +++ linux-2.5.56-work/drivers/char/tty_io.c	2003-01-11 13:23:15.000000000 +0100
> > @@ -1329,6 +1329,8 @@
> >  		int major, minor;
> >  		struct tty_driver *driver;
> >  
> > +		lock_kernel(); 
> > +
> 
> Deadlock.  chrdev_open() calls lock_kernel() and then the fops->open
> method, which is tty_open().

No problem, lock_kernel is recursive and dropped on schedule.

It is very very hard to get a BKL deadlock.

> This one needs deeper review.

I agree, but one has to start somewhere. Please submit any fixes,
perhaps we can take then close these issues for good.

Was looking at n_tty.c now, looks like it has some more race 
problems.

-Andi