Re: [RFC] O(1) proc_pid_readdir

[William Lee Irwin III <wli@holomorphy.com>]

On Mon, Mar 17, 2003 at 07:17:08PM +0100, Manfred Spraul wrote:
>>> Could you check if the attached test app triggers the NMI oopser?

On Mon, Mar 17, 2003 at 04:14:05PM -0800, William Lee Irwin III wrote:
>> Sure, no problem.

On Mon, Mar 17, 2003 at 04:48:50PM -0800, William Lee Irwin III wrote:
> Gee, this is bright. I think I remember why I haven't done testing of
> tasklist_lock NMI oopses for several releases now.
[...]
> timer doesn't work through the IO-APIC - disabling NMI Watchdog!
> ...trying to set up timer as Virtual Wire IRQ...Uhhuh. NMI received for unknown reason 35 on CPU 0.

NMI_LOCAL_APIC should not care whether the timer works through IO-APIC's.

--- linux-2.5.64/arch/i386/kernel/io_apic.c.orig	Mon Mar 17 17:19:02 2003
+++ linux-2.5.64/arch/i386/kernel/io_apic.c	Mon Mar 17 17:19:08 2003
@@ -1990,7 +1990,7 @@ static inline void check_timer(void)
 	}
 	printk(" failed.\n");
 
-	if (nmi_watchdog) {
+	if (nmi_watchdog == NMI_IO_APIC) {
 		printk(KERN_WARNING "timer doesn't work through the IO-APIC - disabling NMI Watchdog!\n");
 		nmi_watchdog = 0;
 	}