Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
Message-ID: <c12b0ed6-4189-4625-84bf-abd835b9b364@quicinc.com>
Date:   Thu, 7 Dec 2023 17:20:51 +0800
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH v4 4/4] bus: mhi: host: Take irqsave lock after TRE is
 generated
Content-Language: en-US
To:     Manivannan Sadhasivam <mani@kernel.org>
CC:     <quic_jhugo@quicinc.com>, <mhi@lists.linux.dev>,
        <linux-arm-msm@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <quic_cang@quicinc.com>, <quic_mrana@quicinc.com>,
        Hemant Kumar <quic_hemantk@quicinc.com>,
        Lazarus Motha <quic_lmotha@quicinc.com>
References: <1699939661-7385-1-git-send-email-quic_qianyu@quicinc.com>
 <1699939661-7385-5-git-send-email-quic_qianyu@quicinc.com>
 <20231124100916.GB4536@thinkpad>
 <ae1514d1-1cdf-42f8-ac8c-6e6c2005922f@quicinc.com>
 <20231207063851.GD2932@thinkpad>
From:   Qiang Yu <quic_qianyu@quicinc.com>
In-Reply-To: <20231207063851.GD2932@thinkpad>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 12/7/2023 2:38 PM, Manivannan Sadhasivam wrote:
> On Mon, Nov 27, 2023 at 03:19:49PM +0800, Qiang Yu wrote:
>> On 11/24/2023 6:09 PM, Manivannan Sadhasivam wrote:
>>> On Tue, Nov 14, 2023 at 01:27:41PM +0800, Qiang Yu wrote:
>>>> From: Hemant Kumar <quic_hemantk@quicinc.com>
>>>>
>>>> If CONFIG_TRACE_IRQFLAGS is enabled, irq will be enabled once __local_bh_
>>>> enable_ip is called as part of write_unlock_bh. Hence, let's take irqsave
>>> "__local_bh_enable_ip" is a function name, so you should not break it.
>> Thanks for let me know, will note this in following patch.
>>>> lock after TRE is generated to avoid running write_unlock_bh when irqsave
>>>> lock is held.
>>>>
>>> I still don't understand this commit message. Where is the write_unlock_bh()
>>> being called?
>>>
>>> - Mani
>> Write_unlock_bh() is invoked in mhi_gen_te()
>> The calling flow is like
>> mhi_queue
>>      read_lock_irqsave(&mhi_cntrl->pm_lock, flags)
>>      mhi_gen_tre
>>          write_lock_bh(&mhi_chan->lock)
>>          write_unlock_bh(&mhi_chan->lock)   // Will enable irq if
>> CONFIG_TRACE_IRQFLAGS is enabled
>>      read_lock_irqsave(&mhi_cntrl->pm_lock, flags)
>>
>> after adding this patch, the calling flow becomes
>>
>> mhi_queue
>>      mhi_gen_tre
>>          write_lock_bh(&mhi_chan->lock)
>>          write_unlock_bh(&mhi_chan->lock)
>>      read_lock_irqsave(&mhi_cntrl->pm_lock, flags)
>>      read_lock_irqsave(&mhi_cntrl->pm_lock, flags)
> So this patch essentially fixes the issue caused by patch 1? If so, this should
> be squashed into patch 1.
>
> - Mani

Yes, this patch is to fix the issue caused by patch 1. Will squash patch 
1 and this patch into one patch

in next version.

>
>>>> Signed-off-by: Hemant Kumar <quic_hemantk@quicinc.com>
>>>> Signed-off-by: Lazarus Motha <quic_lmotha@quicinc.com>
>>>> Signed-off-by: Qiang Yu <quic_qianyu@quicinc.com>
>>>> ---
>>>>    drivers/bus/mhi/host/main.c | 13 +++++--------
>>>>    1 file changed, 5 insertions(+), 8 deletions(-)
>>>>
>>>> diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c
>>>> index 33f27e2..d7abd0b 100644
>>>> --- a/drivers/bus/mhi/host/main.c
>>>> +++ b/drivers/bus/mhi/host/main.c
>>>> @@ -1128,17 +1128,15 @@ static int mhi_queue(struct mhi_device *mhi_dev, struct mhi_buf_info *buf_info,
>>>>    	if (unlikely(MHI_PM_IN_ERROR_STATE(mhi_cntrl->pm_state)))
>>>>    		return -EIO;
>>>> -	read_lock_irqsave(&mhi_cntrl->pm_lock, flags);
>>>> -
>>>>    	ret = mhi_is_ring_full(mhi_cntrl, tre_ring);
>>>> -	if (unlikely(ret)) {
>>>> -		ret = -EAGAIN;
>>>> -		goto exit_unlock;
>>>> -	}
>>>> +	if (unlikely(ret))
>>>> +		return -EAGAIN;
>>>>    	ret = mhi_gen_tre(mhi_cntrl, mhi_chan, buf_info, mflags);
>>>>    	if (unlikely(ret))
>>>> -		goto exit_unlock;
>>>> +		return ret;
>>>> +
>>>> +	read_lock_irqsave(&mhi_cntrl->pm_lock, flags);
>>>>    	/* Packet is queued, take a usage ref to exit M3 if necessary
>>>>    	 * for host->device buffer, balanced put is done on buffer completion
>>>> @@ -1158,7 +1156,6 @@ static int mhi_queue(struct mhi_device *mhi_dev, struct mhi_buf_info *buf_info,
>>>>    	if (dir == DMA_FROM_DEVICE)
>>>>    		mhi_cntrl->runtime_put(mhi_cntrl);
>>>> -exit_unlock:
>>>>    	read_unlock_irqrestore(&mhi_cntrl->pm_lock, flags);
>>>>    	return ret;
>>>> -- 
>>>> 2.7.4
>>>>
>>>>