Received-SPF: pass (google.com: domain of linux-kernel+bounces-81212-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Message-ID: <ff4de2ac-ab2f-4c34-a49c-dd78ceb8add5@quicinc.com>
Date: Mon, 26 Feb 2024 16:19:16 +0530
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH] soc: qcom: llcc: Add llcc device availability check
Content-Language: en-US
To: Mukesh Ojha <quic_mojha@quicinc.com>, <andersson@kernel.org>,
        <konrad.dybcio@linaro.org>
CC: <linux-arm-msm@vger.kernel.org>, <linux-kernel@vger.kernel.org>
References: <20240220122805.9084-1-quic_mojha@quicinc.com>
 <c60638be-e12e-4148-8c20-1dbba9c817da@quicinc.com>
 <d617c991-b7a7-eb8f-0747-303c33cbf9a7@quicinc.com>
From: Sahil Chandna <quic_chandna@quicinc.com>
In-Reply-To: <d617c991-b7a7-eb8f-0747-303c33cbf9a7@quicinc.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit

On 2/26/2024 4:02 PM, Mukesh Ojha wrote:
> 
> 
> On 2/22/2024 11:37 PM, Sahil Chandna wrote:
>> On 2/20/2024 5:58 PM, Mukesh Ojha wrote:
>>> When llcc driver is enabled  and llcc device is not
>>> physically there on the SoC, client can get
>>> -EPROBE_DEFER on calling llcc_slice_getd() and it
>>> is possible they defer forever.
>>>
>>> Let's add a check device availabilty and set the
>>> appropriate applicable error in drv_data.
>>>
>>> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
>>> ---
>>>   drivers/soc/qcom/llcc-qcom.c | 23 ++++++++++++++++++++++-
>>>   1 file changed, 22 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/soc/qcom/llcc-qcom.c b/drivers/soc/qcom/llcc-qcom.c
>>> index 4ca88eaebf06..cb336b183bba 100644
>>> --- a/drivers/soc/qcom/llcc-qcom.c
>>> +++ b/drivers/soc/qcom/llcc-qcom.c
>>> @@ -769,6 +769,27 @@ static const struct qcom_sct_config 
>>> x1e80100_cfgs = {
>>>   };
>>>   static struct llcc_drv_data *drv_data = (void *) -EPROBE_DEFER;
>>> +static DEFINE_MUTEX(dev_avail);
>> what is the requirement for mutex lock here? Since we are only trying 
>> to find if node present or not
> 
> I was trying to avoid two parallel call from llcc_slice_getd() calling
> parallel call to of_find_node_by_name() as it should be one time search 
> for device presence to find a node and check if device is present or
> not.
> 
> -Mukesh
> 
Got it, but of_find_node_by_name () is holding a raw_spin_lock_irqsave 
() for concurrency, right ? please correct me if understanding is wrong.
>>> +
>>> +static bool is_llcc_device_available(void)
>>> +{
>>> +    static struct llcc_drv_data *ptr;
>>> +
>>> +    mutex_lock(&dev_avail);
>>> +    if (!ptr) {
>>> +        struct device_node *node;
>>> +
>>> +        node = of_find_node_by_name(NULL, "system-cache-controller");
>>> +        if (!of_device_is_available(node)) {
>>> +            pr_warn("llcc-qcom: system-cache-controller node not 
>>> found\n");
>>> +            drv_data = ERR_PTR(-ENODEV);
>>> +        }
>>> +        of_node_put(node);
>>> +        ptr = drv_data;
>>> +    }
>>> +    mutex_unlock(&dev_avail);
>>> +    return (PTR_ERR(ptr) != -ENODEV) ? true : false;
>>> +}
>>>   /**
>>>    * llcc_slice_getd - get llcc slice descriptor
>>> @@ -783,7 +804,7 @@ struct llcc_slice_desc *llcc_slice_getd(u32 uid)
>>>       struct llcc_slice_desc *desc;
>>>       u32 sz, count;
>>> -    if (IS_ERR(drv_data))
>>> +    if (!is_llcc_device_available() || IS_ERR(drv_data))
Also, thinking about this, should the status of device present or not be 
saved in static variable instead of function call for each client ?
>>>           return ERR_CAST(drv_data);
>>>       cfg = drv_data->cfg;
>>

Regards,
Sahil