RE: [PATCH v2] locking/rtmutex: Fix incorrect spinning condition

From: "Zhang, Qiang1" <qiang1.zhang@intel.com>
To: Thomas Gleixner <tglx@linutronix.de>,
	"peterz@infradead.org" <peterz@infradead.org>,
	"mingo@redhat.com" <mingo@redhat.com>,
	"will@kernel.org" <will@kernel.org>,
	"longman@redhat.com" <longman@redhat.com>
Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"Zhang, Qiang1" <qiang1.zhang@intel.com>
Subject: RE: [PATCH v2] locking/rtmutex: Fix incorrect spinning condition
Date: Sat, 18 Dec 2021 07:24:07 +0000	[thread overview]
Message-ID: <PH0PR11MB5880A552E7A3856D8F9165FBDA799@PH0PR11MB5880.namprd11.prod.outlook.com> (raw)
In-Reply-To: <87k0g3ar7z.ffs@tglx>

-----Original Message-----
From: Thomas Gleixner <tglx@linutronix.de> 
Sent: 2021年12月18日 4:53
To: Zhang, Qiang1 <qiang1.zhang@intel.com>; peterz@infradead.org; mingo@redhat.com; will@kernel.org; longman@redhat.com
Cc: linux-kernel@vger.kernel.org; Zhang, Qiang1 <qiang1.zhang@intel.com>
Subject: Re: [PATCH v2] locking/rtmutex: Fix incorrect spinning condition

Zqiang,

On Fri, Dec 17 2021 at 15:42, Zqiang wrote:
> When the lock owner is on CPU and not need resched, the current waiter
> need to be checked, if it not longer top the waiter, stop spinning.
>
> Fixes: c3123c431447 ("locking/rtmutex: Dont dereference waiter lockless")
> Signed-off-by: Zqiang <qiang1.zhang@intel.com>
> ---
>  v1->v2:
>  Modify description information.
>
>  kernel/locking/rtmutex.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
> index 0c1f2e3f019a..8555c4efe97c 100644
> --- a/kernel/locking/rtmutex.c
> +++ b/kernel/locking/rtmutex.c
> @@ -1383,7 +1383,7 @@ static bool rtmutex_spin_on_owner(struct rt_mutex_base *lock,
>  		 *  - the VCPU on which owner runs is preempted
>  		 */
>  		if (!owner_on_cpu(owner) || need_resched() ||
> -		    rt_mutex_waiter_is_top_waiter(lock, waiter)) {
> +		    !rt_mutex_waiter_is_top_waiter(lock, waiter)) {
>  			res = false;
>  			break;
>
>good catch!
>
>Though this does not apply because the condition is incomplete. You
>somehow dropped this from the condition:
>
>                   vcpu_is_preempted(task_cpu(owner))) 
>
>Please make always sure that your patches apply against Linus tree
>before sending them out.

This commit c0bed69daf4b ("locking: Make owner_on_cpu() into <linux/sched.h>")
make the following modifications in latest linux-next.

+static inline bool owner_on_cpu(struct task_struct *owner)
+{
+       /*
+        * As lock holder preemption issue, we both skip spinning if
+        * task is not on cpu or its cpu is preempted
+        */
+       return owner->on_cpu && !vcpu_is_preempted(task_cpu(owner));
+}
+

Thanks

Zqiang

>
>Thanks,
>
>        tglx