From: Kei Tokunaga <ktokunag@redhat.com>
Date: Thu, 20 Nov 2008 09:44:21 -0500
Subject: [ia64] replace printk with mprintk in MCA/INIT context
Message-id: 492577C5.7080305@redhat.com
O-Subject: [RHEL5.3][PATCH] replace printk with mprintk in ia64_mca_modify_original_stack
Bugzilla: 471970
RH-Acked-by: Prarit Bhargava <prarit@redhat.com>
RH-Acked-by: Dave Anderson <anderson@redhat.com>

bz471970
https://bugzilla.redhat.com/show_bug.cgi?id=471970

Description
===========
 Using printk from MCA/INIT context is unsafe since it can cause
 deadlock.  The ia64_mca_modify_original_stack is called from both
 a MCA handler and an INIT handler, so it should use mprintk instead
 of printk.  This problem would block execution of kdump from INIT
 button.  If INIT is asserted while one of cpus is processing printk,
 and if a printk is called from the INIT handler, then system will
 hang up and 2nd kernel will never start up.

kABI status
===========
 No kABI breakage found.

Brew status
===========
 Built on all platforms.
  https://brewweb.devel.redhat.com/taskinfo?taskID=1580701

Upstream status
===============
 commit 150e323dcb4f1d8906c480162ae4b24a83453d44
 Author: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
 Date:   Mon, 17 Nov 2008 09:46:08 +0900

Test status
===========
 It's really hard to reproduce the bug as we've never experienced it
 in real life.  Fujitsu found the bug from the source code review
 and it's obvious bug and the fix is a one-liner.  For this posting,
 I confirmed it built, booted OK, and kdump worked on PRIMEQUEST
 located in Westford.

Additional notes
================
 - The patch applies to 2.6.18-124.el5.
 - This only effects ia64.

Please review and ACK,
Kei
--
Kei Tokunaga
Fujitsu on-site partner

diff --git a/arch/ia64/kernel/mca.c b/arch/ia64/kernel/mca.c
index 39ed501..272f3f3 100644
--- a/arch/ia64/kernel/mca.c
+++ b/arch/ia64/kernel/mca.c
@@ -1115,7 +1115,7 @@ ia64_mca_modify_original_stack(struct pt_regs *regs,
 	return previous_current;
 
 no_mod:
-	printk(KERN_INFO "cpu %d, %s %s, original stack not modified\n",
+	mprintk(KERN_INFO "cpu %d, %s %s, original stack not modified\n",
 			smp_processor_id(), type, msg);
 	return previous_current;
 }