Re: [RFC] virtio_scsi: to poll and kick the virtqueue in timeout handler

[Hannes Reinecke <hare@suse.de>]

On 5/25/21 6:47 PM, Stefan Hajnoczi wrote:
> On Mon, May 24, 2021 at 11:33:33PM -0700, Dongli Zhang wrote:
>> On 5/24/21 6:24 AM, Stefan Hajnoczi wrote:
>>> On Sun, May 23, 2021 at 09:39:51AM +0200, Hannes Reinecke wrote:
>>>> On 5/23/21 8:38 AM, Dongli Zhang wrote:
>>>>> This RFC is to trigger the discussion about to poll and kick the
>>>>> virtqueue on purpose in virtio-scsi timeout handler.
>>>>>
>>>>> The virtio-scsi relies on the virtio vring shared between VM and host.
>>>>> The VM side produces requests to vring and kicks the virtqueue, while the
>>>>> host side produces responses to vring and interrupts the VM side.
>>>>>
>>>>> By default the virtio-scsi handler depends on the host timeout handler
>>>>> by BLK_EH_RESET_TIMER to give host a chance to perform EH.
>>>>>
>>>>> However, this is not helpful for the case that the responses are available
>>>>> on vring but the notification from host to VM is lost.
>>>>>
>>>> How can this happen?
>>>> If responses are lost the communication between VM and host is broken, and
>>>> we should rather reset the virtio rings themselves.
>>>
>>> I agree. In principle it's fine to poll the virtqueue at any time, but I
>>> don't understand the failure scenario here. It's not clear to me why the
>>> device-to-driver vq notification could be lost.
>>>
>>
>> One example is the CPU hotplug issue before the commit bf0beec0607d ("blk-mq:
>> drain I/O when all CPUs in a hctx are offline") was available. The issue is
>> equivalent to loss of interrupt. Without the CPU hotplug fix, while NVMe driver
>> relies on the timeout handler to complete inflight IO requests, the PV
>> virtio-scsi may hang permanently.
>>
>> In addition, as the virtio/vhost/QEMU are complex software, we are not able to
>> guarantee there is no further lost of interrupt/kick issue in the future. It is
>> really painful if we encounter such issue in production environment.
> 
> Any number of hardware or software bugs might exist that we don't know
> about, yet we don't pre-emptively add workarounds for them because where
> do you draw the line?
> 
> I checked other SCSI/block drivers and found it's rare to poll in the
> timeout function so there does not seem to be a consensus that it's
> useful to do this.
> 
Not only this; it's downright dangerous attempting to do that in SCSI.
In SCSI we don't have fixed lifetime guarantees that NVMe has, so there 
will be a race condition between timeout and command completion.
Plus there is no interface in SCSI allowing to 'poll' for completions in 
a meaningful manner.

> That said, it's technically fine to do it, the virtqueue APIs are there
> and can be used like this. So if you and others think this is necessary,
> then it's a pretty small change and I'm not against merging a patch like
> this.
> 
I would rather _not_ put more functionality into the virtio_scsi timeout 
handler; this only serves to assume that the timeout handler has some 
functionality in virtio.
Which it patently hasn't, as the prime reason for a timeout handler is 
to _abort_ a command, which we can't on virtio.
Well, we can on virtio, but qemu as the main user will re-route the I/O 
from virtio into doing async-I/O, and there is no way how we can abort 
outstanding asynchronous I/O.
Or any other ioctl, for that matter.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer