Re: [PATCH] net: qrtr: mhi: synchronize qrtr and mhi preparation

[Bhaumik Bhatt <bbhatt@codeaurora.org>]

On 2021-07-21 03:27 PM, hemantk@codeaurora.org wrote:
> On 2021-07-21 11:07, Bhaumik Bhatt wrote:
>> On 2021-07-21 10:52 AM, hemantk@codeaurora.org wrote:
>>> On 2021-07-20 18:42, Bhaumik Bhatt wrote:
>>>> A dl callback can be received anytime after mhi_prepare_for_transfer
>>>> has been called. There is a window where the callback may happen
>>>> before the probe initializes the qrtr_mhi_dev state. Move the
>>>> mhi_prepare_for_transfer call after the registering the endpoint.
>>>> 
>>>> Once moved, the reverse can happen where qrtr will try to send a 
>>>> packet
>>>> before the channels are prepared. Add a wait in the sending path to
>>>> ensure the channels are prepared before trying to do a ul transfer.
>>>> 
>>>> Fixes: a2e2cc0dbb11 ("net: qrtr: Start MHI channels during init")
>>>> Signed-off-by: Bhaumik Bhatt <bbhatt@codeaurora.org>
>>>> ---
>>>>  net/qrtr/mhi.c | 20 +++++++++++++++-----
>>>>  1 file changed, 15 insertions(+), 5 deletions(-)
>>>> 
>>>> diff --git a/net/qrtr/mhi.c b/net/qrtr/mhi.c
>>>> index 29b4fa3..22b0395 100644
>>>> --- a/net/qrtr/mhi.c
>>>> +++ b/net/qrtr/mhi.c
>>>> @@ -15,6 +15,7 @@ struct qrtr_mhi_dev {
>>>>  	struct qrtr_endpoint ep;
>>>>  	struct mhi_device *mhi_dev;
>>>>  	struct device *dev;
>>>> +	struct completion ready;
>>>>  };
>>>> 
>>>>  /* From MHI to QRTR */
>>>> @@ -50,6 +51,10 @@ static int qcom_mhi_qrtr_send(struct 
>>>> qrtr_endpoint
>>>> *ep, struct sk_buff *skb)
>>>>  	struct qrtr_mhi_dev *qdev = container_of(ep, struct qrtr_mhi_dev, 
>>>> ep);
>>>>  	int rc;
>>>> 
>>>> +	rc = wait_for_completion_interruptible(&qdev->ready);
>>>> +	if (rc)
>>>> +		goto free_skb;
>>>> +
>>>>  	if (skb->sk)
>>>>  		sock_hold(skb->sk);
>>>> 
>>>> @@ -78,11 +83,6 @@ static int qcom_mhi_qrtr_probe(struct mhi_device 
>>>> *mhi_dev,
>>>>  	struct qrtr_mhi_dev *qdev;
>>>>  	int rc;
>>>> 
>>>> -	/* start channels */
>>>> -	rc = mhi_prepare_for_transfer(mhi_dev, MHI_CH_INBOUND_ALLOC_BUFS);
>>>> -	if (rc)
>>>> -		return rc;
>>>> -
>>>>  	qdev = devm_kzalloc(&mhi_dev->dev, sizeof(*qdev), GFP_KERNEL);
>>>>  	if (!qdev)
>>>>  		return -ENOMEM;
>>> would it be good to init completion variable here (call 
>>> init_completion) ?
>> You mean just before setting qdev->mhi_dev? I don't see why that would
>> make a difference
>> mainly because the qcom_mhi_qrtr_send() will only happen after 
>> endpoint is
>> registered and DL xfer cb will also only come in after we have 
>> prepared the
>> channels and completed ready with dev_data already set.
> looks like qcom_mhi_qrtr_send is not going to get called directly. i
> was thinking
> what if this api is called before init_completion() returns. if it is
> only possible
> through ep.xmit call back only, can you move it right above
> qdev->ep.xmit = qcom_mhi_qrtr_send; ?
>> 
Ah. OK. I see your point. I will do that and upload a v2.

>>>> @@ -90,12 +90,22 @@ static int qcom_mhi_qrtr_probe(struct mhi_device 
>>>> *mhi_dev,
>>>>  	qdev->mhi_dev = mhi_dev;
>>>>  	qdev->dev = &mhi_dev->dev;
>>>>  	qdev->ep.xmit = qcom_mhi_qrtr_send;
>>>> +	init_completion(&qdev->ready);
>>>> 
>>> 
>>>> 
>>>>  	return 0;
>> 
>> Thanks,
>> Bhaumik
>> ---
>> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora 
>> Forum,
>> a Linux Foundation Collaborative Project
> 
> Thanks,
> Hemant
> ---
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora 
> Forum,
> a Linux Foundation Collaborative Project

Thanks,
Bhaumik
---
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora 
Forum,
a Linux Foundation Collaborative Project