Re: [PATCH v4 1/8] scsi: ufs: Flush exception event before suspend

From: Can Guo <cang@codeaurora.org>
To: Bart Van Assche <bvanassche@acm.org>
Cc: asutoshd@codeaurora.org, nguyenb@codeaurora.org,
	hongwus@codeaurora.org, rnayak@codeaurora.org,
	linux-scsi@vger.kernel.org, kernel-team@android.com,
	saravanak@google.com, salyzyn@google.com,
	Sayali Lokhande <sayalil@codeaurora.org>,
	Alim Akhtar <alim.akhtar@samsung.com>,
	Avri Altman <avri.altman@wdc.com>,
	Pedro Sousa <pedrom.sousa@synopsys.com>,
	"James E.J. Bottomley" <jejb@linux.ibm.com>,
	"Martin K. Petersen" <martin.petersen@oracle.com>,
	Stanley Chu <stanley.chu@mediatek.com>,
	Bean Huo <beanhuo@micron.com>,
	Venkat Gopalakrishnan <venkatg@codeaurora.org>,
	Tomas Winkler <tomas.winkler@intel.com>,
	open list <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v4 1/8] scsi: ufs: Flush exception event before suspend
Date: Mon, 03 Feb 2020 14:23:15 +0800	[thread overview]
Message-ID: <82723efc44714e8677505cb7999d3fd5@codeaurora.org> (raw)
In-Reply-To: <525e4f67-f471-54a6-aaea-b3772a550af1@acm.org>

On 2020-01-26 11:29, Bart Van Assche wrote:
> On 2020-01-22 23:25, Can Guo wrote:
>> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
>> index 1201578..c2de29f 100644
>> --- a/drivers/scsi/ufs/ufshcd.c
>> +++ b/drivers/scsi/ufs/ufshcd.c
>> @@ -4760,8 +4760,15 @@ static void ufshcd_slave_destroy(struct 
>> scsi_device *sdev)
>>  			 * UFS device needs urgent BKOPs.
>>  			 */
>>  			if (!hba->pm_op_in_progress &&
>> -			    ufshcd_is_exception_event(lrbp->ucd_rsp_ptr))
>> -				schedule_work(&hba->eeh_work);
>> +			    ufshcd_is_exception_event(lrbp->ucd_rsp_ptr)) {
>> +				/*
>> +				 * Prevent suspend once eeh_work is scheduled
>> +				 * to avoid deadlock between ufshcd_suspend
>> +				 * and exception event handler.
>> +				 */
>> +				if (schedule_work(&hba->eeh_work))
>> +					pm_runtime_get_noresume(hba->dev);
>> +			}
> 
> Please combine the two logical tests with "&&" instead of nesting two
> if-statements inside each other.
> 
>>  			break;
>>  		case UPIU_TRANSACTION_REJECT_UPIU:
>>  			/* TODO: handle Reject UPIU Response */
>> @@ -5215,7 +5222,14 @@ static void 
>> ufshcd_exception_event_handler(struct work_struct *work)
>> 
>>  out:
>>  	scsi_unblock_requests(hba->host);
>> -	pm_runtime_put_sync(hba->dev);
>> +	/*
>> +	 * pm_runtime_get_noresume is called while scheduling
>> +	 * eeh_work to avoid suspend racing with exception work.
>> +	 * Hence decrement usage counter using pm_runtime_put_noidle
>> +	 * to allow suspend on completion of exception event handler.
>> +	 */
>> +	pm_runtime_put_noidle(hba->dev);
>> +	pm_runtime_put(hba->dev);
>>  	return;
>>  }
>> 
>> @@ -7901,6 +7915,7 @@ static int ufshcd_suspend(struct ufs_hba *hba, 
>> enum ufs_pm_op pm_op)
>>  			goto enable_gating;
>>  	}
>> 
>> +	flush_work(&hba->eeh_work);
>>  	ret = ufshcd_link_state_transition(hba, req_link_state, 1);
>>  	if (ret)
>>  		goto set_dev_active;
> 
> I think this patch introduces a new race condition, namely the 
> following:
> - ufshcd_slave_destroy() tests pm_op_in_progress and reads the value
>   zero from that variable.
> - ufshcd_suspend() sets hba->pm_op_in_progress to one.
> - ufshcd_slave_destroy() calls schedule_work().
> 
> How about fixing this race condition by calling
> pm_runtime_get_noresume() before checking pm_op_in_progress and by
> reallowing resume if no work is scheduled?
> 
> Thanks,
> 
> Bart.

Hi Bart,

If you apply this patch, you will find the change is not in
ufshcd_slave_destroy(), but in ufshcd_transfer_rsp_status().
So the racing you mentioned above does not exist.

Thanks,

Can Guo.