Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Thomas Gleixner <tglx@linutronix.de>
To:     Waiman Long <llong@redhat.com>,
        Xiaoming Ni <nixiaoming@huawei.com>,
        linux-kernel@vger.kernel.org, peterz@infradead.org,
        mingo@redhat.com, will@kernel.org, boqun.feng@gmail.com
Cc:     wangle6@huawei.com, xiaoqian9@huawei.com, shaolexi@huawei.com
Subject: Re: [PATCH] semaphore: Add might_sleep() to down_*() family
In-Reply-To: <48cddad4-0388-ae8b-f98b-1629b9ae590a@redhat.com>
References: <20210809021215.19991-1-nixiaoming@huawei.com>
 <48cddad4-0388-ae8b-f98b-1629b9ae590a@redhat.com>
Date:   Fri, 13 Aug 2021 19:27:37 +0200
Message-ID: <877dgpp852.ffs@tglx>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On Sun, Aug 08 2021 at 23:01, Waiman Long wrote:
> On 8/8/21 10:12 PM, Xiaoming Ni wrote:
>> Semaphore is sleeping lock. Add might_sleep() to down*() family
>> (with exception of down_trylock()) to detect atomic context sleep.
>> @@ -157,6 +160,7 @@ int down_timeout(struct semaphore *sem, long timeout)
>>   	unsigned long flags;
>>   	int result = 0;
>>   
>> +	might_sleep();
>>   	raw_spin_lock_irqsave(&sem->lock, flags);
>>   	if (likely(sem->count > 0))
>>   		sem->count--;
>
> I think it is simpler to just put a "might_sleep()" in __down_common() 
> which is the function where sleep can actually happen.

No. Putting it in __down_common() is wrong, because that covers only the
contended case.

But we want to cover the potential sleep, i.e. checking even in the
non-contended case, which is what might_sleep() is about.

Thanks,

        tglx