Re: [PATCH] kexec jump: fix compiling warning on xchg(&kexec_lock, 0) in kernel_kexec()

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 13 Aug 2008 10:01:13 -0700 (PDT) Linus Torvalds <torvalds@linux-foundation.org> wrote:

> 
> 
> On Wed, 13 Aug 2008, Andrew Morton wrote:
> >
> > We don't need to create that local.  I queued this:
> 
> No, please don't. 
> 
> Just don't take this whole patch-series until it's cleaned up.

We already took it - in 2.6.13!

> There is 
> absolutely no excuse for using xchg as a locking primitive. Nothing like 
> this should be queued anywhere, it should be burned and the ashes should 
> be scattered over the atlantic so that nobody will ever see them again.
> 
> F*ck me with a spoon, if you have to use xchg() to do a trylock, why the 
> hell isn't the unlock sequence then
> 
> 	smp_mb();
> 	var = 0;
> 
> instead? Not that that's really right either, but at least it avoids the 
> _ridiculous_ crap. The real solution is probably to use a spinlock and 
> trylock/unlock.
> 

Or test_and_set_bit().  That's what I've been saying too, only
differently ;)

But cleaning up the long-standing silly usage of xchg() is a different
activity from suppressing this recently-added compile warning.