allow recursive die() on i386

allow recursive die() on i386

[Jeff Moyer]

Hello,

This patch allows for a recursive die() on i386.  This closely resembles
what is done on x86_64, fwiw.

-Jeff

--- linux-2.6.7/arch/i386/kernel/traps.c.orig	2004-07-28 08:08:21.000000000 -0400
+++ linux-2.6.7/arch/i386/kernel/traps.c	2004-07-28 08:10:54.000000000 -0400
@@ -294,6 +294,7 @@ bug:
 }
 
 spinlock_t die_lock = SPIN_LOCK_UNLOCKED;
+static int die_owner = -1;
 
 void die(const char * str, struct pt_regs * regs, long err)
 {
@@ -301,7 +302,13 @@ void die(const char * str, struct pt_reg
 	int nl = 0;
 
 	console_verbose();
-	spin_lock_irq(&die_lock);
+	local_irq_disable();
+	if (!spin_trylock(&die_lock)) {
+		if (smp_processor_id() != die_owner)
+			spin_lock(&die_lock);
+		/* allow recursive die to fall through */
+	}
+	die_owner = smp_processor_id();
 	bust_spinlocks(1);
 	handle_BUG(regs);
 	printk(KERN_ALERT "%s: %04lx [#%d]\n", str, err & 0xffff, ++die_counter);
@@ -321,6 +328,7 @@ void die(const char * str, struct pt_reg
 		printk("\n");
 	show_registers(regs);
 	bust_spinlocks(0);
+	die_owner = -1;
 	spin_unlock_irq(&die_lock);
 	if (in_interrupt())
 		panic("Fatal exception in interrupt");

Re: allow recursive die() on i386

[Keith Owens]

On Wed, 28 Jul 2004 08:35:39 -0400, 
Jeff Moyer <jmoyer@redhat.com> wrote:
>This patch allows for a recursive die() on i386.  This closely resembles
>what is done on x86_64, fwiw.

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.8-rc2/2.6.8-rc2-mm1/broken-out/make-i386-die-more-resilient-against-recursive-errors.patch

does this and more, it also guards against too many recursive calls to
die.  The patch is already in Andrew's tree.

Re: allow recursive die() on i386

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 791 bytes --]

On Wed, 2004-07-28 at 15:11, Keith Owens wrote:
> On Wed, 28 Jul 2004 08:35:39 -0400, 
> Jeff Moyer <jmoyer@redhat.com> wrote:
> >This patch allows for a recursive die() on i386.  This closely resembles
> >what is done on x86_64, fwiw.
> 
> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.8-rc2/2.6.8-rc2-mm1/broken-out/make-i386-die-more-resilient-against-recursive-errors.patch
> 
> does this and more, it also guards against too many recursive calls to
> die.  The patch is already in Andrew's tree.
+	static struct {
+		spinlock_t lock;
+		u32 lock_owner;
+		int lock_owner_depth;
+	} die = {
+		.lock =			SPIN_LOCK_UNLOCKED,
+		.lock_owner =		-1,
+		.lock_owner_depth =	0
+	};


humm... am I the only one who considers this a little bit ugly ???

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]