Re: [PATCH] 2.5.44: lkcd (9/9): dump driver and build files

[Keith Owens]

On 22 Oct 2002 15:21:02 -0700, 
Piet Delaney <[e-mail]> wrote:
>Looks like your right, for parisc even these functions are using a
>spinlock. Is it really necessary for parisc to use spinlocks? Even
>Solaris doesn't use spinlocks for atomic set and reads of atomic
>variables.

Hardware restriction.  Somebody read that "all locks can be implemented
in terms of a single hardware primitive" and believed it :(.  The only
atomic operation on parisc is load and clear word.  According to the
comments in asm-parisc/atomic.h, atomic_set should not require a lock,
but it is coded to use one.  Atomic add/sub obviously needs a lock.