Re: [PATHC -tip 1/2] x86, nmi: Define a name for NMI control port

[Cyrill Gorcunov <gorcunov@openvz.org>]

On 03/01/2011 10:57 AM, Ingo Molnar wrote:
...
>
> FYI, the patch is whitespace damaged.
>
> 	Ingo

Ingo, this one should be fine

---
From: Cyrill Gorcunov <gorcunov@openvz.org>
Subject: [PATCH -tip 1/2] x86, nmi: Define a name for NMI control port

Also add a comment anout why we need in-out operations.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
---
  arch/x86/include/asm/mach_traps.h |   18 +++++++++++++-----
  1 file changed, 13 insertions(+), 5 deletions(-)

Index: linux-2.6.git/arch/x86/include/asm/mach_traps.h
=====================================================================
--- linux-2.6.git.orig/arch/x86/include/asm/mach_traps.h
+++ linux-2.6.git/arch/x86/include/asm/mach_traps.h
@@ -8,6 +8,7 @@
  #include <asm/mc146818rtc.h>

  #define NMI_REASON_PORT		0x61
+#define NMI_ENABLE_PORT		0x70 /* Real-Time Clock Address Register as well */

  #define NMI_REASON_SERR		0x80
  #define NMI_REASON_IOCHK	0x40
@@ -30,12 +31,19 @@ static inline void reassert_nmi(void)
  		old_reg = current_lock_cmos_reg();
  	else
  		lock_cmos(0); /* register doesn't matter here */
-	outb(0x8f, 0x70);
-	inb(0x71);		/* dummy */
-	outb(0x0f, 0x70);
-	inb(0x71);		/* dummy */
+
+	/*
+	 * This will cause the NMI output to transition low
+	 * then high if there are any pending NMI sources. The
+	 * CPU's NMI input logic will then register a new NMI.
+	 */
+	outb(0x8f, NMI_ENABLE_PORT);
+	inb(0x71);	/* dummy */
+	outb(0x0f, NMI_ENABLE_PORT);
+	inb(0x71);	/* dummy */
+
  	if (old_reg >= 0)
-		outb(old_reg, 0x70);
+		outb(old_reg, NMI_ENABLE_PORT);
  	else
  		unlock_cmos();
  }