Big kernel lock in release functions

Big kernel lock in release functions

[David C. Hansen]

  I've been looking at use of the big kernel lock in a lot of different
places in the kernel.  I've noticed that there are quite a few uses in
devices' release functions, with no obvious purpose.  Take a look at
arch/m68k/atari/joystick.c:

    lock_kernel();
    joystick[minor].active = 0;
    joystick[minor].ready = 0;

    if ((joystick[0].active == 0) && (joystick[1].active == 0))
        ikbd_joystick_disable();
    unlock_kernel();

  But, there are other places in the same file where the same operations
are performed with no locking at all (except for
ikbd_joystick_disable()). Is there a reason to have locking in the
release function but not in the read or open functions?  
  This is a quite common practice in release functions throughout the
kernel.  In 2.4.10, I counted 108 different places where the BKL is used
in a release function.  I'm not claiming that all of these are
unnecessary, but I believe a big chunk of them are.  

--
David C. Hansen
haveblue@us.ibm.com
IBM LTC Base/OS Group
(503)578-4080