Re: [PATCH] scsi: core: Run queue first after running device.

[John Garry <john.garry@huawei.com>]

On 06/08/2021 09:58, John Garry wrote:

And the patch subject is ambiguous

> On 05/08/2021 15:32, lijinlin3@huawei.com wrote:
>> From: Li Jinlin<lijinlin3@huawei.com>
>>
>> We found a hang issue, the test steps are as follows:
>>    1. echo "blocked" >/sys/block/sda/device/state
>>    2. dd if=/dev/sda of=/mnt/t.log bs=1M count=10
>>    3. echo none > /sys/block/sda/queue/scheduler
>>    4. echo "running" >/sys/block/sda/device/state
>>
>> Step3 and Step4 should finish this work after Step4, but them hangs.
>>
>>    CPU#0               CPU#1                CPU#2
>>    ---------------     ----------------     ----------------
>>                                             Step1: blocking device
>>
>>                                             Step2: dd xxxx
>>                                                    ^^^^^^ get request
>>                                                           
>> q_usage_counter++
>>
>>                        Step3: switching scheculer
>>                        elv_iosched_store
>>                          elevator_switch
>>                            blk_mq_freeze_queue
>>                              blk_freeze_queue
>>                                > blk_freeze_queue_start
>>                                  ^^^^^^ mq_freeze_depth++
>>
>>                                > blk_mq_run_hw_queues
>>                                  ^^^^^^ can't run queue when dev blocked
>>
>>                                > blk_mq_freeze_queue_wait
>>                                  ^^^^^^ Hang here!!!
>>                                         wait q_usage_counter==0
>>
>>    Step4: running device
>>    store_state_field
>>      scsi_rescan_device
>>        scsi_attach_vpd
>>          scsi_vpd_inquiry
>>            __scsi_execute
>>              blk_get_request
>>                blk_mq_alloc_request
>>                  blk_queue_enter
>>                  ^^^^^^ Hang here!!!
>>                         wait mq_freeze_depth==0
>>
>>      blk_mq_run_hw_queues
>>      ^^^^^^ dispatch IO, q_usage_counter will reduce to zero
>>
>>                              blk_mq_unfreeze_queue
>>                              ^^^^^ mq_freeze_depth--
>>
>> Step3 and Step4 wait for each other, caused hangs.
>>
>> This requires run queue frist to fix this issue when the device state
> 
> frist ?
> 
>> changes to SDEV_RUNNING.
>>
>> Fixes: f0f82e2476f6 ("scsi: core: Fix capacity set to zero after 
>> offlinining device")
>> Signed-off-by: Li Jinlin<lijinlin3@huawei.com>
>> Signed-off-by: Qiu Laibin<qiulaibin@huawei.com>
>> Signed-off-by: Wu Bo<wubo40@huawei.com>
> 
> what kind of SoB is this?
> 
>> ---
>>   drivers/scsi/scsi_sysfs.c | 6 +++---
>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
>> index c3a710bceba0..aa701582c950 100644
>> --- a/drivers/scsi/scsi_sysfs.c
>> +++ b/drivers/scsi/scsi_sysfs.c
>> @@ -809,12 +809,12 @@ store_state_field(struct device *dev, struct 
>> device_attribute *attr,
>>       ret = scsi_device_set_state(sdev, state);
>>       /*
>>        * If the device state changes to SDEV_RUNNING, we need to
>> -     * rescan the device to revalidate it, and run the queue to
>> -     * avoid I/O hang.
>> +     * run the queue to avoid I/O hang, and rescan the device
>> +     * to revalidate it.
> 
> A bit more description of the IO hang would be useful
> 
>>        */
>>       if (ret == 0 && state == SDEV_RUNNING) {
>> -        scsi_rescan_device(dev);
>>           blk_mq_run_hw_queues(sdev->request_queue, true);
>> +        scsi_rescan_device(dev);
> 
> This would not have happened if scsi_rescan_device() was ran outside the 
> mutex lock region, like I suggested originally.
> 
> Indeed, I doubt blk_mq_run_hw_queues() needs to be run with the sdev 
> state_mutex held either.
> 
>>       }
>>       mutex_unlock(&sdev->state_mutex);
>> -- 
> 
> .