Re: [BUG 2.6.25-rc3] scheduler/hotplug: some processes aredealocked when cpu is set to offline

[Suresh Siddha <suresh.b.siddha@intel.com>]

On Fri, Mar 07, 2008 at 07:50:02AM -0800, Gautham R Shenoy wrote:
> On Fri, Mar 07, 2008 at 07:25:37PM +0530, Gautham R Shenoy wrote:
> > On Fri, Mar 07, 2008 at 02:02:20PM +0100, Dmitry Adamushko wrote:
> > > Hi,
> > >
> > > 'watchdog' is of SCHED_FIFO class. The standard load-balancer doesn't
> > > move RT tasks between cpus anymore and there is a special mechanism in
> > > scher_rt.c instead (I think, it's .25 material).
> > >
> > > So I wonder, whether __migrate_task() is still capable of properly
> > > moving a RT task to another CPU (e.g. for the case when it's in
> > > TASK_RUNNING state) without breaking something in the rt migration
> > > mechanism (or whatever else) that would leave us with a runqueue in
> > > the 'inconsistent' state...
> > > (I've taken a quick look at the relevant code so can't confirm it yet)
> > >
> > > maybe it'd be faster if somebody could do a quick test now with the
> > > following line commented out in kernel/softlockup.c :: watchdog()
> > >
> > > -         sched_setscheduler(current, SCHED_FIFO, &param);
> >
> > Commenting out that like seems to work. Passed 500 iterations of
> > cpu-hotplug without any problems.
> 
> This seems to unearth another problem. After some 850 successful
> cpu-hotplug iterations I got this message.
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> ------------[ cut here ]------------
> BUG: spinlock recursion on CPU#2, kstopmachine/32521
> BUG: spinlock lockup on CPU#1, kstopmachine/32517, cc43db80
> Pid: 32517, comm: kstopmachine Not tainted 2.6.25-rc3 #44
>  [<c0284ebb>] <0>BUG: spinlock lockup on CPU#0, kstopmachine/32520,
> cc43db80
> _raw_spin_lock+0xd5/0xf9
>  [<c04e1a33>] <0>BUG: spinlock lockup on CPU#3, kstopmachine/32522,
> cc43db80
> _spin_lock+0x20/0x28
> Pid: 32522, comm: kstopmachine Not tainted 2.6.25-rc3 #44
>  [<c04dfa41>] ?  [<c0284ebb>] schedule+0xb0/0x5ab
>  [<c04dfa41>] schedule+0xb0/0x5ab
> _raw_spin_lock+0xd5/0xf9
>  [<c04e1d58>] ?  [<c04e1a33>] _spin_unlock_irqrestore+0x36/0x3c
>  [<c0119fdc>] ? _spin_lock+0x20/0x28
>  [<c0119806>] ? complete+0x34/0x3e
>  [<c0143646>] double_lock_balance+0x3a/0x57
>  [<c0119806>] do_stop+0xd4/0xfe

Well, there is another

	sched_setscheduler(p, SCHED_FIFO, &param);

in kernel/stop_machine.c

Perhaps we need to remove this aswell and try?

thanks,
suresh