Re: [PATCH v2 4/4] block/mq-deadline: Prioritize high-priority requests

[Bart Van Assche <bvanassche@acm.org>]

On 3/26/22 01:57, Oleksij Rempel wrote:
> On Fri, Mar 25, 2022 at 10:05:43AM -0700, Bart Van Assche wrote:
>> On 3/25/22 05:33, Oleksij Rempel wrote:
>>> I have regression on iMX6 + USB storage device with this patch.
>>> After plugging USB Flash driver (in my case USB3 Intenso 32GB) and
>>> running mount /dev/sda1 /mnt, the command will never exit.
>>>
>>> Reverting this patchs (322cff70d46) on top of v5.17 solves it for me.
>>>
>>> How can I help you to debug this issue?
>>
>> That is unexpected. Is there perhaps something special about the USB
>> stick for which the hang occurs, e.g. the queue depth it supports is
>> low? Can you please share the output of the following commands after
>> having inserted the USB stick that triggers the hang?
>>
>> (cd /sys/class && grep -aH . scsi_host/*/can_queue scsi_device/*/device/{blacklist,inquiry,model,queue*,vendor}) | sort
> 
> Here it is:
> root@test:/sys/class cat scsi_host/*/can_queue
> 1
> root@test:/sys/class cat scsi_device/*/device/blacklist
> root@test:/sys/class cat scsi_device/*/device/inquiry
> Intenso Speed Line      1.00
> root@test:/sys/class cat scsi_device/*/device/model
> Speed Line
> root@test:/sys/class cat scsi_device/*/device/queue*
> 1
> none
> root@test:/sys/class cat scsi_device/*/device/vendor
> Intenso
> 
> I do not know, if there is something special about it. If needed, i can
> take it apart to see what controller is used.

Thanks, this helps a lot. can_queue = 1 as I was suspecting. In a quick
test I noticed that I/O is about 10x slower for queue depth 1 and the
v5.17 mq-deadline scheduler. I will take a closer look at this tomorrow.

Bart.