linux-kernel - Re: [PATCH v1 1/2] scsi: ufs: Fix unbalanced scsi_block_reqs_cnt caused by ufshcd

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <09c5d4d31a0bd9bed99815cfbf51aaad@codeaurora.org>
Date:   Tue, 03 Nov 2020 18:01:01 +0800
From:   Can Guo <cang@...eaurora.org>
To:     Stanley Chu <stanley.chu@...iatek.com>
Cc:     asutoshd@...eaurora.org, nguyenb@...eaurora.org,
        hongwus@...eaurora.org, rnayak@...eaurora.org,
        linux-scsi@...r.kernel.org, kernel-team@...roid.com,
        saravanak@...gle.com, salyzyn@...gle.com,
        Alim Akhtar <alim.akhtar@...sung.com>,
        Avri Altman <avri.altman@....com>,
        "James E.J. Bottomley" <jejb@...ux.ibm.com>,
        "Martin K. Petersen" <martin.petersen@...cle.com>,
        Bean Huo <beanhuo@...ron.com>,
        Bart Van Assche <bvanassche@....org>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v1 1/2] scsi: ufs: Fix unbalanced scsi_block_reqs_cnt
 caused by ufshcd_hold()

On 2020-11-03 15:07, Stanley Chu wrote:
> Hi Can,
> 
> On Mon, 2020-11-02 at 22:24 -0800, Can Guo wrote:
>> The scsi_block_reqs_cnt increased in ufshcd_hold() is supposed to be
>> decreased back in ufshcd_ungate_work() in a paired way. However, if
>> specific ufshcd_hold/release sequences are met, it is possible that
>> scsi_block_reqs_cnt is increased twice but only one ungate work is
>> queued. To make sure scsi_block_reqs_cnt is handled by ufshcd_hold() 
>> and
> 
> Just curious that how could this be possible? Would you have some 
> failed
> examples?
> 

[1] One gate_work() is in the workqueue, not yet executed, now clk state 
== REQ_CLKS_OFF.
[2] ufshcd_queuecommand() calls ufshcd_hold(async == ture) -> 
active_req++ -> scsi_block_reqs_cnt++ -> REQ_CLKS_ON -> queue ungate 
work -> active_req-- -> return -EAGAIN.
[3] Now gate_work() starts to run, but since the clk state is 
REQ_CLKS_ON, gate_work() just sets clk state to CLKS_ON and bail.
[3] Someone calls ufshcd_hold(async == false) -> do something -> 
ufshcd_release() -> clk state is changed to REQ_CLKS_OFF. Note that, 
till now, ungate_work() is still in the work queue, not executed yet.
[4] Now, if someone calls ufshcd_hold(), we will hit the issue.

Above sequence is a very common clk gate/ungate sequence. The issue
is because ungate_work is queued but cannot be executed in time. In my
case, I see the ungate_work is somehow delayed for about 150ms. This
change has been tested by customers on multiple platforms. And you
can tell from the code that it won't break anything. :)

Thanks,

Can Guo.

>> ufshcd_ungate_work() in a paired way, increase it only if queue_work()
>> returns true.
>> 
>> Signed-off-by: Can Guo <cang@...eaurora.org>
>> Reviewed-by: Hongwu Su <hongwus@...eaurora.org>
>> ---
>>  drivers/scsi/ufs/ufshcd.c | 6 +++---
>>  1 file changed, 3 insertions(+), 3 deletions(-)
>> 
>> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
>> index 847f355..efa7d86 100644
>> --- a/drivers/scsi/ufs/ufshcd.c
>> +++ b/drivers/scsi/ufs/ufshcd.c
>> @@ -1634,12 +1634,12 @@ int ufshcd_hold(struct ufs_hba *hba, bool 
>> async)
>>  		 */
>>  		/* fallthrough */
>>  	case CLKS_OFF:
>> -		ufshcd_scsi_block_requests(hba);
>>  		hba->clk_gating.state = REQ_CLKS_ON;
>>  		trace_ufshcd_clk_gating(dev_name(hba->dev),
>>  					hba->clk_gating.state);
>> -		queue_work(hba->clk_gating.clk_gating_workq,
>> -			   &hba->clk_gating.ungate_work);
>> +		if (queue_work(hba->clk_gating.clk_gating_workq,
>> +			       &hba->clk_gating.ungate_work))
>> +			ufshcd_scsi_block_requests(hba);
>>  		/*
>>  		 * fall through to check if we should wait for this
>>  		 * work to be done or not.
> 
> Thanks,
> Stanley Chu