[patch 7/8] lock validator: s390 use raw_spinlock in mcck handler

[patch 7/8] lock validator: s390 use raw_spinlock in mcck handler

[Heiko Carstens]

From: Heiko Carstens <heiko.carstens@de.ibm.com>

Machine checks on s390 are always enabled (except in the machine check handler
itself). Therefore use a raw_spinlock in the machine check handler to avoid
deadlocks in the lock validator.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
---

 drivers/s390/s390mach.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff -purN a/drivers/s390/s390mach.c b/drivers/s390/s390mach.c
--- a/drivers/s390/s390mach.c	2006-06-14 13:44:52.000000000 +0200
+++ b/drivers/s390/s390mach.c	2006-06-14 13:45:01.000000000 +0200
@@ -372,7 +372,7 @@ s390_revalidate_registers(struct mci *mc
 void
 s390_do_machine_check(struct pt_regs *regs)
 {
-	static DEFINE_SPINLOCK(ipd_lock);
+	static raw_spinlock_t ipd_lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
 	static unsigned long long last_ipd;
 	static int ipd_count;
 	unsigned long long tmp;
@@ -418,7 +418,7 @@ s390_do_machine_check(struct pt_regs *re
 			 * retry this instruction.
 			 */
 
-			spin_lock(&ipd_lock);
+			__raw_spin_lock(&ipd_lock);
 
 			tmp = get_clock();
 
@@ -432,7 +432,7 @@ s390_do_machine_check(struct pt_regs *re
 			if (ipd_count == MAX_IPD_COUNT)
 				s390_handle_damage("too many ipd retries.");
 
-			spin_unlock(&ipd_lock);
+			__raw_spin_unlock(&ipd_lock);
 		}
 		else {
 			/* Processing damage -> stopping machine */

Re: [patch 7/8] lock validator: s390 use raw_spinlock in mcck handler

[Ingo Molnar]

* Heiko Carstens <heiko.carstens@de.ibm.com> wrote:

> From: Heiko Carstens <heiko.carstens@de.ibm.com>
> 
> Machine checks on s390 are always enabled (except in the machine check 
> handler itself). Therefore use a raw_spinlock in the machine check 
> handler to avoid deadlocks in the lock validator.

hm, couldnt you use the ->lockdep_recursion mechanism to take care of 
such cases? Just call lockdep_off() when entering a machine exception 
handler, and lockdep_on() when exiting it.

	Ingo

Re: [patch 7/8] lock validator: s390 use raw_spinlock in mcck handler

[Heiko Carstens]

On Wed, Jun 14, 2006 at 09:25:16PM +0200, Ingo Molnar wrote:
> 
> * Heiko Carstens <heiko.carstens@de.ibm.com> wrote:
> 
> > From: Heiko Carstens <heiko.carstens@de.ibm.com>
> > 
> > Machine checks on s390 are always enabled (except in the machine check 
> > handler itself). Therefore use a raw_spinlock in the machine check 
> > handler to avoid deadlocks in the lock validator.
> 
> hm, couldnt you use the ->lockdep_recursion mechanism to take care of 
> such cases? Just call lockdep_off() when entering a machine exception 
> handler, and lockdep_on() when exiting it.

Ah, sure. Wasn't aware that there is such an interface. Will do.