diff -urN linux-2.6.0-test11.lockmeter.orig/include/asm-ia64/spinlock.h linux-2.6.0-test11.lockmeter/include/asm-ia64/spinlock.h
--- linux-2.6.0-test11.lockmeter.orig/include/asm-ia64/spinlock.h	2003-12-09 13:03:36.000000000 +0100
+++ linux-2.6.0-test11.lockmeter/include/asm-ia64/spinlock.h	2003-12-09 13:08:24.000000000 +0100
@@ -247,13 +247,33 @@
 extern void _metered_spin_unlock(spinlock_t *lock);
 
 /*
- *  Use a less efficient, and inline, atomic_dec_and_lock() if lockmetering
- *  so we can see the callerPC of who is actually doing the spin_lock().
- *  Otherwise, all we see is the generic rollup of all locks done by
- *  atomic_dec_and_lock().
+ *  Matches what is in arch/ia64/lib/dec_and_lock.c, except this one is
+ *  "static inline" so that the spin_lock(), if actually invoked, is charged
+ *  against the real caller, not against the catch-all atomic_dec_and_lock
  */
 static inline int atomic_dec_and_lock(atomic_t *atomic, spinlock_t *lock)
 {
+	int counter;
+	int newcount;
+
+repeat:
+	counter = atomic_read(atomic);
+	newcount = counter-1;
+
+	if (!newcount)
+		goto slow_path;
+
+	asm volatile("mov ar.ccv=%1;;\n\t"
+		     "cmpxchg4.acq %0=%2,%3,ar.ccv;;"
+		     :"=r" (newcount)
+		     :"r" (counter), "m" (atomic->counter), "r" (newcount)
+		     :"ar.ccv");
+
+	if (newcount != counter)
+		goto repeat;
+	return 0;
+
+slow_path:
 	_metered_spin_lock(lock);
 	if (atomic_dec_and_test(atomic))
 		return 1;