[PATCH] x86_64: fix die_lock nesting

[Corey Minyard]

I noticed this when poking around in this area.

BTW, the comments in oops_begin say the operation is racy, and the only
way I can think of that is races is if you get a non-NMI oops then
get an NMI oops within the oops_begin or oops_end functions.
That can actually be fixed using compare-and-swap, but, to tell you
the truth, it just doesn't seem worth it to me.  If you like, though,
I can attempt a fix at that, too.  Anyway, the patch...


The oops_begin() function in x86_64 would only conditionally claim
the die_lock if the call is nested, but oops_end() would always
release the spinlock. This patch adds a nest count for the die lock
so that the release of the lock is only done on the final oops_end().

Signed-off-by: Corey Minyard <minyard@xxxxxxx>

diff --git a/arch/x86_64/kernel/traps.c b/arch/x86_64/kernel/traps.c
index 6bda322..debd834 100644
--- a/arch/x86_64/kernel/traps.c
+++ b/arch/x86_64/kernel/traps.c
@@ -384,6 +384,7 @@ void out_of_line_bug(void)
 
 static DEFINE_SPINLOCK(die_lock);
 static int die_owner = -1;
+static unsigned int die_nest_count;
 
 unsigned __kprobes long oops_begin(void)
 {
@@ -398,6 +399,7 @@ unsigned __kprobes long oops_begin(void)
 		else
 			spin_lock(&die_lock);
 	}
+	die_nest_count++;
 	die_owner = cpu;
 	console_verbose();
 	bust_spinlocks(1);
@@ -408,7 +410,13 @@ void __kprobes oops_end(unsigned long fl
 { 
 	die_owner = -1;
 	bust_spinlocks(0);
-	spin_unlock_irqrestore(&die_lock, flags);
+	die_nest_count--;
+	if (die_nest_count)
+		/* We still own the lock */
+		local_irq_restore(flags);
+	else
+		/* Nest count reaches zero, release the lock. */
+		spin_unlock_irqrestore(&die_lock, flags);
 	if (panic_on_oops)
 		panic("Oops");
 }
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/