From: Aristeu Rozanski <aris@redhat.com>
Date: Wed, 11 Mar 2009 15:18:32 -0400
Subject: [x86_64] mce: do not clear an unrecoverable error status
Message-id: 20090311191831.GK2706@redhat.com
O-Subject: [RHEL5.4 PATCH] mce: do not clear the status if the error is unrecoverable
Bugzilla: 489692
RH-Acked-by: Rik van Riel <riel@redhat.com>
RH-Acked-by: Pete Zaitcev <zaitcev@redhat.com>
RH-Acked-by: Prarit Bhargava <prarit@redhat.com>

https://bugzilla.redhat.com/show_bug.cgi?id=489692

Currently in all cases do_machine_check() clears all the MCE status registers.
This is a problem for Fujitsu because in case of unrecoverable error events,
the kernel will panic and the BIOS won't be able to use those registers to
determine the cause.

Upstream: part of bd78432c8f209a1028f4e5bada8b1da1d8e4da09
Tested by Fujitsu.
Patch by Martin Wilck.

diff --git a/arch/x86_64/kernel/mce.c b/arch/x86_64/kernel/mce.c
index 8c76505..7bc24a1 100644
--- a/arch/x86_64/kernel/mce.c
+++ b/arch/x86_64/kernel/mce.c
@@ -218,7 +218,6 @@ void do_machine_check(struct pt_regs * regs, long error_code)
 		mce_get_rip(&m, regs);
 		if (error_code >= 0)
 			rdtscll(m.tsc);
-		wrmsrl(MSR_IA32_MC0_STATUS + i*4, 0);
 		if (error_code != -2)
 			mce_log(&m);
 
@@ -270,6 +269,9 @@ void do_machine_check(struct pt_regs * regs, long error_code)
  out:
 	/* Last thing done in the machine check exception to clear state. */
 	wrmsrl(MSR_IA32_MCG_STATUS, 0);
+	for (i = 0; i < banks; i++)
+		wrmsrl(MSR_IA32_MC0_STATUS + i*4, 0);
+
  out2:
 	atomic_dec(&mce_entry);
 }