Received-SPF: pass (google.com: domain of linux-kernel+bounces-10900-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Message-ID: <52ee95c2-1118-4f44-85e0-862ac5f83257@quicinc.com>
Date: Mon, 25 Dec 2023 14:13:14 +0800
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH v2] pinctrl: Add lock to ensure the state atomization
Content-Language: en-US
To: Linus Walleij <linus.walleij@linaro.org>
CC: <andersson@kernel.org>, <kernel@quicinc.com>, <linux-gpio@vger.kernel.org>,
        <linux-kernel@vger.kernel.org>, <linux-arm-msm@vger.kernel.org>
References: <20231212090611.950-1-quic_aiquny@quicinc.com>
 <CACRpkdb6dkw58GwkqYXTDAQtdLazOLyp1CEjnkxDX2v=TDvvMw@mail.gmail.com>
From: "Aiqun Yu (Maria)" <quic_aiquny@quicinc.com>
In-Reply-To: <CACRpkdb6dkw58GwkqYXTDAQtdLazOLyp1CEjnkxDX2v=TDvvMw@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit


On 12/20/2023 7:02 PM, Linus Walleij wrote:
> Hi Maria,
> 
> On Tue, Dec 12, 2023 at 10:06 AM Maria Yu <quic_aiquny@quicinc.com> wrote:
> 
>> Currently pinctrl_select_state is an export symbol and don't have
>> effective re-entrance protect design. During async probing of devices
>> it's possible to end up in pinctrl_select_state() from multiple
>> contexts simultaneously, so make it thread safe.
>> More over, when the real racy happened, the system frequently have
>> printk message like:
>>    "not freeing pin xx (xxx) as part of deactivating group xxx - it is
>> already used for some other setting".
>> Finally the system crashed after the flood log.
>> Add per pinctrl lock to ensure the old state and new state transition
>> atomization.
>> Also move dev error print message outside the region with interrupts
>> disabled.
>>
>> Fixes: 4198a9b57106 ("pinctrl: avoid reload of p state in list iteration")
>> Signed-off-by: Maria Yu <quic_aiquny@quicinc.com>
> 
> Overall this looks good!
> 
>> @@ -1262,9 +1263,12 @@ static void pinctrl_link_add(struct pinctrl_dev *pctldev,
>>   static int pinctrl_commit_state(struct pinctrl *p, struct pinctrl_state *state)
>>   {
>>          struct pinctrl_setting *setting, *setting2;
>> -       struct pinctrl_state *old_state = READ_ONCE(p->state);
>> +       struct pinctrl_state *old_state;
>>          int ret;
>> +       unsigned long flags;
>>
>> +       spin_lock_irqsave(&p->lock, flags);
> (...)
>> +       spin_unlock_irqrestore(&p->lock, flags);
> (...)
>> +       spin_unlock_irqrestore(&p->lock, flags);
> 
> Is it possible to use a scoped guard for pinctrl_commit_state()?
Good idea.
I will address this in next patchset.
> 
> #include <linux/cleanup.h>
> guard(spinlock_irqsave)(&p->lock);
> 
> It saves some code (and no need for flags) and avoid possible
> bugs when people add new errorpaths to the code.
> 
> Yours,
> Linus Walleij

-- 
Thx and BRs,
Aiqun(Maria) Yu