[PATCH] fix dynticks for voyager

[PATCH] fix dynticks for voyager

[James Bottomley]

As usual voyager tripped over an explicit boot CPU is zero assumption in
the dynticks code.  This is the fix I have queued in the voyager tree.

James

---

>From 5d3e489cf8fcf12eff92476098b9bfa0120ee3c7 Mon Sep 17 00:00:00 2001
From: James Bottomley <James.Bottomley@HansenPartnership.com>
Date: Mon, 30 Apr 2007 11:27:25 -0500
Subject: [VOYAGER] clockevents: correct boot cpu is zero assumption

This isn't true for voyager, so alter setup_pit_timer() to initialise
the cpumask from the current processor id (which should be the boot
processor) rather than defaulting to zero.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
---
 arch/i386/kernel/i8253.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/i386/kernel/i8253.c b/arch/i386/kernel/i8253.c
index 10cef5c..f8a3c40 100644
--- a/arch/i386/kernel/i8253.c
+++ b/arch/i386/kernel/i8253.c
@@ -110,7 +110,7 @@ void __init setup_pit_timer(void)
 	 * Start pit with the boot cpu mask and make it global after the
 	 * IO_APIC has been initialized.
 	 */
-	pit_clockevent.cpumask = cpumask_of_cpu(0);
+	pit_clockevent.cpumask = cpumask_of_cpu(smp_processor_id());
 	pit_clockevent.mult = div_sc(CLOCK_TICK_RATE, NSEC_PER_SEC, 32);
 	pit_clockevent.max_delta_ns =
 		clockevent_delta2ns(0x7FFF, &pit_clockevent);
-- 
1.4.4.4

Re: [PATCH] fix dynticks for voyager

[Thomas Gleixner]

On Mon, 2007-04-30 at 11:38 -0500, James Bottomley wrote:
> As usual voyager tripped over an explicit boot CPU is zero assumption in
> the dynticks code.  This is the fix I have queued in the voyager tree.

I expected some kind of voyager breakage. I desperately tried to find
one for testing in my museum of bizarre hardware to avoid this,
but ... :)

	tglx

> ---
> 
> >From 5d3e489cf8fcf12eff92476098b9bfa0120ee3c7 Mon Sep 17 00:00:00 2001
> From: James Bottomley <James.Bottomley@HansenPartnership.com>
> Date: Mon, 30 Apr 2007 11:27:25 -0500
> Subject: [VOYAGER] clockevents: correct boot cpu is zero assumption
> 
> This isn't true for voyager, so alter setup_pit_timer() to initialise
> the cpumask from the current processor id (which should be the boot
> processor) rather than defaulting to zero.
> 
> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

Acked-by: Thomas Gleixner <tglx@linutronix.de>

> ---
>  arch/i386/kernel/i8253.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/i386/kernel/i8253.c b/arch/i386/kernel/i8253.c
> index 10cef5c..f8a3c40 100644
> --- a/arch/i386/kernel/i8253.c
> +++ b/arch/i386/kernel/i8253.c
> @@ -110,7 +110,7 @@ void __init setup_pit_timer(void)
>  	 * Start pit with the boot cpu mask and make it global after the
>  	 * IO_APIC has been initialized.
>  	 */
> -	pit_clockevent.cpumask = cpumask_of_cpu(0);
> +	pit_clockevent.cpumask = cpumask_of_cpu(smp_processor_id());
>  	pit_clockevent.mult = div_sc(CLOCK_TICK_RATE, NSEC_PER_SEC, 32);
>  	pit_clockevent.max_delta_ns =
>  		clockevent_delta2ns(0x7FFF, &pit_clockevent);

Re: [PATCH] fix dynticks for voyager

[Matt Mackall]

On Mon, Apr 30, 2007 at 11:38:51AM -0500, James Bottomley wrote:
> As usual voyager tripped over an explicit boot CPU is zero assumption in
> the dynticks code.  This is the fix I have queued in the voyager tree.

Can we flush out all these assumptions by adding a constant offset
somewhere? Or reversing the ordering?

-- 
Mathematics is the supreme nostalgia of our time.

Re: [PATCH] fix dynticks for voyager

[Jeremy Fitzhardinge]

Matt Mackall wrote:
> On Mon, Apr 30, 2007 at 11:38:51AM -0500, James Bottomley wrote:
>   
>> As usual voyager tripped over an explicit boot CPU is zero assumption in
>> the dynticks code.  This is the fix I have queued in the voyager tree.
>>     
>
> Can we flush out all these assumptions by adding a constant offset
> somewhere? Or reversing the ordering?
>   

My understanding is that the boot CPU is random depending on who wins
the "I get to boot" bunfight.

    J

Re: [PATCH] fix dynticks for voyager

[James Bottomley]

On Tue, 2007-05-01 at 17:26 -0700, Jeremy Fitzhardinge wrote:
> Matt Mackall wrote:
> > On Mon, Apr 30, 2007 at 11:38:51AM -0500, James Bottomley wrote:
> >   
> >> As usual voyager tripped over an explicit boot CPU is zero assumption in
> >> the dynticks code.  This is the fix I have queued in the voyager tree.
> >>     
> >
> > Can we flush out all these assumptions by adding a constant offset
> > somewhere? Or reversing the ordering?

It might be useful to voyager, but unless the offset were random, people
would simply hardwire it.

> My understanding is that the boot CPU is random depending on who wins
> the "I get to boot" bunfight.

Not quite.  On apic systems, the boot CPU can be an arbitrary physical
id but is conventionally logical id 0.  On voyager, the CPU id actually
maps to a fixed location, so there's no sense in having a logical id.
The booting CPU on voyager is simply the lowest numbered CPU that comes
up through the SUS boot.

James