--- 2.4.4pre2aa/include/asm-alpha/atomic.h.~1~	Tue Nov 28 18:40:01 2000
+++ 2.4.4pre2aa/include/asm-alpha/atomic.h	Fri Apr 13 04:17:16 2001
@@ -106,4 +106,9 @@
 #define atomic_inc(v) atomic_add(1,(v))
 #define atomic_dec(v) atomic_sub(1,(v))
 
+#define smp_mb__before_atomic_inc()	smp_mb()
+#define smp_mb__after_atomic_inc()	smp_mb()
+#define smp_mb__before_atomic_dec()	smp_mb()
+#define smp_mb__after_atomic_dec()	smp_mb()
+
 #endif /* _ALPHA_ATOMIC_H */
--- 2.4.4pre2aa/include/asm-i386/atomic.h.~1~	Thu Apr 12 21:42:17 2001
+++ 2.4.4pre2aa/include/asm-i386/atomic.h	Fri Apr 13 05:09:17 2001
@@ -42,6 +42,15 @@
  */ 
 #define atomic_set(v,i)		(((v)->counter) = (i))
 
+/*
+ * It maybe we need an explicit serialization across inc/dec on
+ * recent chips, if not then just replace this with a barrier().
+ */
+#define smp_mb__before_atomic_inc()	smp_mb()
+#define smp_mb__after_atomic_inc()	smp_mb()
+#define smp_mb__before_atomic_dec()	smp_mb()
+#define smp_mb__after_atomic_dec()	smp_mb()
+
 /**
  * atomic_add - add integer to atomic variable
  * @i: integer value to add
--- 2.4.4pre2aa/include/linux/interrupt.h.~1~	Thu Apr 12 21:42:17 2001
+++ 2.4.4pre2aa/include/linux/interrupt.h	Fri Apr 13 04:41:19 2001
@@ -186,16 +186,20 @@
 static inline void tasklet_disable_nosync(struct tasklet_struct *t)
 {
 	atomic_inc(&t->count);
+	smp_mb__after_atomic_inc();
 }
 
 static inline void tasklet_disable(struct tasklet_struct *t)
 {
 	tasklet_disable_nosync(t);
 	tasklet_unlock_wait(t);
+	/* Make sure not to speculate reads until the tasklet resigns. */
+	smp_rmb();
 }
 
 static inline void tasklet_enable(struct tasklet_struct *t)
 {
+	smp_mb__before_atomic_dec();
 	atomic_dec(&t->count);
 }
 
--- 2.4.4pre2aa/kernel/softirq.c.~1~	Thu Apr 12 17:05:40 2001
+++ 2.4.4pre2aa/kernel/softirq.c	Fri Apr 13 04:35:54 2001
@@ -153,9 +153,15 @@
 
 				t->func(t->data);
 				/*
-				 * talklet_trylock() uses test_and_set_bit that imply
+				 * talklet_trylock() uses test_and_set_bit that _just_ imply
 				 * an mb when it returns zero, thus we need the explicit
 				 * mb only here: while closing the critical section.
+				 *
+				 * On UP the barrier() in not necessary because f->func
+				 * cannot be reordered by the compiler anyways because
+				 * of the C language. Unless the compiler is doing something
+				 * stupid dropping the #ifdef wouldn't make any difference in the
+				 * generatd asm though (the #ifdef is mostly a documenetation issue).
 				 */
 #ifdef CONFIG_SMP
 				smp_mb__before_clear_bit();
@@ -163,6 +169,13 @@
 				tasklet_unlock(t);
 				continue;
 			}
+			/*
+			 * NOTE: here we don't need the smb_mb() because the tasklet_unlock()
+			 * write depends on the read of the &t->count that is in turn _the_
+			 * variable that the lock is serializing. The CPU isn't allowed to
+			 * write to the memory before we checked t->count is not zero and
+			 * by that time we don't need the lock any longer.
+			 */
 			tasklet_unlock(t);
 		}
 		local_irq_disable();
@@ -197,6 +210,9 @@
 				clear_bit(TASKLET_STATE_SCHED, &t->state);
 
 				t->func(t->data);
+#ifdef CONFIG_SMP
+				smp_mb__before_clear_bit();
+#endif
 				tasklet_unlock(t);
 				continue;
 			}