Re: [PATCH] generalized spin_lock_bit, take two

[Robert Love <rml@tech9.net>]

On Wed, 2002-07-24 at 19:25, Andrew Morton wrote:

> I did some testing yesterday with fork/exec/exit-intensive
> workloads and the contention rate on pte_chain_lock was 0.3%,
> so the efficiency problems which Linus described are unlikely
> to bite us in this particular application.  But if the usage
> of spin_lock_bit() were to widen, some platforms may be impacted.

At this point (with Linus asking me to keep the "pte_chain_lock()"
interface) I do not really care so long as we not compile the actual
locking on UP.

Will you take this patch, then?

	Robert Love

diff -urN linux-2.5.28/include/linux/page-flags.h linux/include/linux/page-flags.h
--- linux-2.5.28/include/linux/page-flags.h	Wed Jul 24 14:03:21 2002
+++ linux/include/linux/page-flags.h	Thu Jul 25 09:27:01 2002
@@ -228,6 +228,8 @@
 #define ClearPageDirect(page)		clear_bit(PG_direct, &(page)->flags)
 #define TestClearPageDirect(page)	test_and_clear_bit(PG_direct, &(page)->flags)
 
+#ifdef CONFIG_SMP
+
 /*
  * inlines for acquisition and release of PG_chainlock
  */
@@ -253,6 +255,20 @@
 	preempt_enable();
 }
 
+#else
+
+static inline void pte_chain_lock(struct page *page)
+{
+	preempt_disable();
+}
+
+static inline void pte_chain_unlock(struct page *page)
+{
+	preempt_enable();
+}
+
+#endif
+
 /*
  * The PageSwapCache predicate doesn't use a PG_flag at this time,
  * but it may again do so one day.