balance_irq()'s move() while in machine_restart() hangs system

balance_irq()'s move() while in machine_restart() hangs system

[Matt Domsch]

Ingo, if one processor is in move() when another processor calls
machine_restart(), the system will hang.  All otherCPUs will move into
the hlt state, but one in move(), by virtue of smp_num_cpus getting set
to 1 in smp_send_stop() before the 4th CPU has completed its work and
gets out of move().  It will loop in move() in this chunk:

		if (direction == 1) {
			cpu++;
			if (cpu >= smp_num_cpus)
				cpu = 0;
		} else {
			cpu--;
			if (cpu == -1)
				cpu = smp_num_cpus-1;
		}
	} while (!IRQ_ALLOWED(cpu,allowed_mask) ||
			(search_idle && !IDLE_ENOUGH(cpu,now)));

with smp_num_cpus=1 and never manage to exit.

Is it fair to change smp_send_stop() to call
smp_call_function(stop_this_cpu, NULL, 1, 1) instead (make it wait for
the other CPUs to complete before continuing)?

Thanks,
Matt

-- 
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com

Re: balance_irq()'s move() while in machine_restart() hangs system

[Ingo Molnar]

On 9 Apr 2003, Matt Domsch wrote:

> Is it fair to change smp_send_stop() to call
> smp_call_function(stop_this_cpu, NULL, 1, 1) instead (make it wait for
> the other CPUs to complete before continuing)?

sure, it's fair - at that point all the notification has been done.

	Ingo