Re: [PATCH] fix i386 interrupt re-enabling in die() (attempt 2)

["Jan Beulich" <JBeulich@novell.com>]

>>> Zwane Mwaikambo <zwane@arm.linux.org.uk> 08.09.05 19:37:20 >>>
>On Thu, 8 Sep 2005, Jan Beulich wrote:
>
>> diff -Npru 2.6.13/arch/i386/kernel/traps.c
>> 2.6.13-i386-die-irq/arch/i386/kernel/traps.c
>> --- 2.6.13/arch/i386/kernel/traps.c	2005-08-29 01:41:01.000000000
>> +0200
>> +++ 2.6.13-i386-die-irq/arch/i386/kernel/traps.c	2005-09-07
>> 11:39:40.000000000 +0200
>> @@ -304,6 +304,7 @@ void die(const char * str, struct pt_reg
>>  		spinlock_t lock;
>>  		u32 lock_owner;
>>  		int lock_owner_depth;
>> +		unsigned long flags;
>>  	} die = {
>>  		.lock =			SPIN_LOCK_UNLOCKED,
>>  		.lock_owner =		-1,
>> @@ -313,7 +314,7 @@ void die(const char * str, struct pt_reg
>>  
>>  	if (die.lock_owner != raw_smp_processor_id()) {
>>  		console_verbose();
>> -		spin_lock_irq(&die.lock);
>> +		spin_lock_irqsave(&die.lock, die.flags);
>
>This corrupts flags on contention, use a stack variable.

Indeed. Corrected below:

(Note: Patch also attached because the inline version is certain to
get
line wrapped.)

Rather than blindly re-enabling interrupts in die(), save their state
upon
entry and then restore that state.

Signed-off-by: Jan Beulich <jbeulich@novell.com>

diff -Npru 2.6.13/arch/i386/kernel/traps.c
2.6.13-i386-die-irq/arch/i386/kernel/traps.c
--- 2.6.13/arch/i386/kernel/traps.c	2005-08-29 01:41:01.000000000
+0200
+++ 2.6.13-i386-die-irq/arch/i386/kernel/traps.c	2005-09-09
11:33:30.202343904 +0200
@@ -310,14 +310,17 @@ void die(const char * str, struct pt_reg
 		.lock_owner_depth =	0
 	};
 	static int die_counter;
+	unsigned long flags;
 
 	if (die.lock_owner != raw_smp_processor_id()) {
 		console_verbose();
-		spin_lock_irq(&die.lock);
+		spin_lock_irqsave(&die.lock, flags);
 		die.lock_owner = smp_processor_id();
 		die.lock_owner_depth = 0;
 		bust_spinlocks(1);
 	}
+	else
+		local_save_flags(flags);
 
 	if (++die.lock_owner_depth < 3) {
 		int nl = 0;
@@ -344,7 +347,7 @@ void die(const char * str, struct pt_reg
 
 	bust_spinlocks(0);
 	die.lock_owner = -1;
-	spin_unlock_irq(&die.lock);
+	spin_unlock_irqrestore(&die.lock, flags);
 
 	if (kexec_should_crash(current))
 		crash_kexec(regs);

Attachment: linux-2.6.13-i386-die-irq.patch
Description: Binary data