RE: [PATCH] sched/rt: don't enable runtime if already enabled

[Mike Galbraith <bitbucket@online.de>]

On Tue, 2012-12-11 at 21:38 -0800, Neil Zhang wrote: 
> Ingo and All,

Hm, /me wonders why we would even need update_runtime() handler.  We
disable at CPU_DOWN_PREPARE in update_runtime(), again at CPU_DYING in
migration_call(), and enable at CPU_ONLINE in both.

> -----Original Message-----
> From: Neil Zhang 
> Sent: 2012年12月4日 17:31
> To: Neil Zhang; peterz@infradead.org; mingo@kernel.org
> Cc: Chao Xie; linux-kernel@vger.kernel.org
> Subject: RE: [PATCH] sched/rt: don't enable runtime if already enabled
> 
> Hi All,
> 
> -----Original Message-----
> From: Neil Zhang [mailto:zhangwm@marvell.com] 
> Sent: 2012年12月3日 16:01
> To: mingo@redhat.com; peterz@infradead.org
> Cc: Chao Xie; linux-kernel@vger.kernel.org; Neil Zhang
> Subject: [PATCH] sched/rt: don't enable runtime if already enabled
> 
> There are two paths will call __enable_runtime, they are enable_runtime and rq_online. Don't enable runtime again if it is already enabled.
> It can fix the following kernel BUG when do cpu hotplug while there are realtime threads running.
> 
> [ 3347.091644] kernel BUG at kernel/sched/rt.c:687!
> [ 3347.091674] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM [ 3347.091674] Modules linked in: cidatattydev(O) gs_diag(O) diag(O) gs_modem(O) ccinetdev(O) cci_datastub(O) citty(O) msocke
> [ 3347.091705] CPU: 1    Tainted: G           O  (3.4.5+ #439)
> [ 3347.091735] PC is at __disable_runtime+0x1e4/0x240 [ 3347.091735] LR is at __disable_runtime+0x1dc/0x240
> [ 3347.091735] pc : [<c0162124>]    lr : [<c016211c>]    psr:
> 	a0000093
> [ 3347.091735] sp : e0061dc8  ip : c0807861  fp : c07a9740 [ 3347.091766] r10: c1a0b740  r9 : ffffffff  r8 : fd050f80 [ 3347.091766] r7 : 00000002  r6 : 000003f8  r5 : c1a13740  r4 : c1a137d8 [ 3347.091766] r3 : ffffffff  r2 : e0061dbc  r1 : c06b11b6  r0 : 00000037 [ 3347.091796] Flags: NzCv  IRQs off  FIQs on  Mode SVC_32  ISA ARM  Segment kernel [ 3347.091796] Control: 10c53c7d  Table: 21e5004a  DAC: 00000015
> 
> Signed-off-by: Neil Zhang <zhangwm@marvell.com>
> ---
>  kernel/sched/rt.c |    8 +++++---
>  1 files changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index 418feb0..6525e5b 100644
> --- a/kernel/sched/rt.c
> +++ b/kernel/sched/rt.c
> @@ -722,9 +722,11 @@ static void __enable_runtime(struct rq *rq)
>  
>  		raw_spin_lock(&rt_b->rt_runtime_lock);
>  		raw_spin_lock(&rt_rq->rt_runtime_lock);
> -		rt_rq->rt_runtime = rt_b->rt_runtime;
> -		rt_rq->rt_time = 0;
> -		rt_rq->rt_throttled = 0;
> +		if (rt_rq->rt_runtime == RUNTIME_INF) {
> +			rt_rq->rt_runtime = rt_b->rt_runtime;
> +			rt_rq->rt_time = 0;
> +			rt_rq->rt_throttled = 0;
> +		}
>  		raw_spin_unlock(&rt_rq->rt_runtime_lock);
>  		raw_spin_unlock(&rt_b->rt_runtime_lock);
>  	}
> --
> 1.7.4.1
> 
> Ingo seems use another mail now, update the To list.
> 
> Best Regards,
> Neil Zhang
> 
> Please help review it.
> 
> Best Regards,
> Neil Zhang
> NР骒rybX肚v^)藓{.n+伐{赙zXФ≤}财z&j:+v赙zZ++zf＂h~izwア?ㄨ&)撷f^j谦ym@Aa囤\x7f0鹅hi\x7f