[PATCH RT 04/10] x86/mm/cpa: avoid wbinvd() for PREEMPT

[Steven Rostedt]

4.4.50-rt63-rc1 stable review patch.
If anyone has any objections, please let me know.

------------------

From: John Ogness <john.ogness@xxxxxxxxxxxxx>

Although wbinvd() is faster than flushing many individual pages, it
blocks the memory bus for "long" periods of time (>100us), thus
directly causing unusually large latencies on all CPUs, regardless
of any CPU isolation features that may be active.

For 1024 pages, flushing those pages individually can take up to
2200us, but the task remains fully preemptible during that time.

Cc: stable-rt@xxxxxxxxxxxxxxx
Acked-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Signed-off-by: John Ogness <john.ogness@xxxxxxxxxxxxx>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
Signed-off-by: Steven Rostedt (VMware) <rostedt@xxxxxxxxxxx>
---
 arch/x86/mm/pageattr.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c
index b599a780a5a9..2e85c4117daf 100644
--- a/arch/x86/mm/pageattr.c
+++ b/arch/x86/mm/pageattr.c
@@ -208,7 +208,15 @@ static void cpa_flush_array(unsigned long *start, int numpages, int cache,
 			    int in_flags, struct page **pages)
 {
 	unsigned int i, level;
+#ifdef CONFIG_PREEMPT
+	/*
+	 * Avoid wbinvd() because it causes latencies on all CPUs,
+	 * regardless of any CPU isolation that may be in effect.
+	 */
+	unsigned long do_wbinvd = 0;
+#else
 	unsigned long do_wbinvd = cache && numpages >= 1024; /* 4M threshold */
+#endif
 
 	BUG_ON(irqs_disabled());
 
-- 
2.10.2