Re: [PATCH 1/1] sched: fix cpu_down deadlock

[Lai Jiangshan <laijs@cn.fujitsu.com>]

Jiri Slaby wrote:
> On 09/11/2009 08:09 AM, Lai Jiangshan wrote:
>> Does this bug occur when a cpu is being offlined or
>> when the system is being suspended?
>> Or Both?
> 
> Hi, I tried echo 0/1 > /sys/devices/system/cpu/cpu1/online in a loop,
> but it didn't trigger the bug. It happened only on suspend/resume cycle
> (in the end I found even swsusp in qemu suffers from this).
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 
> 

OK, I knew where this bug is.

I thought the corresponding bit in cpu_active_mask is cleared before
_cpu_down(), but I missed the system-suspend path:disable_nonboot_cpus().

There is a bug in disable_nonboot_cpus() even if my patch is removed.

cpu_active_map is wrong during suspending.(scheduler system who uses
cpu_active_map are still working while suspending)

You need:

int disable_nonboot_cpus(void)
{
	....
	....

	/*
	 * You need adding 'set_cpu_active(cpu, false);' here
	 * to fix this bug and make my patch works well.
	 */
	error = _cpu_down(cpu, 1);

	....
	....
}

Lai.