Re: [PATCH v5 03/15] scsi: ufs: implement scsi host timeout handler

[Hannes Reinecke]

On 03/03/2016 05:10 PM, ygardi@xxxxxxxxxxxxxx wrote:
>> On 03/01/2016 09:25 PM, ygardi@xxxxxxxxxxxxxx wrote:
>>>> On 02/28/2016 09:32 PM, Yaniv Gardi wrote:
>>>>> A race condition exists between request requeueing and scsi layer
>>>>> error handling:
>>>>> When UFS driver queuecommand returns a busy status for a request,
>>>>> it will be requeued and its tag will be freed and set to -1.
>>>>> At the same time it is possible that the request will timeout and
>>>>> scsi layer will start error handling for it. The scsi layer reuses
>>>>> the request and its tag to send error related commands to the device,
>>>>> however its tag is no longer valid.
>>>> Hmm. How can the host return a 'busy' status for a request?
>>>> From my understanding we have three possibilities:
>>>>
>>>> 1) queuecommand returns busy; however, that means that the command has
>>>> never been send and this issue shouldn't occur
>>>> 2) The command returns with BUSY status. But in this case it has
>>>> already
>>>> been returned, so there cannot be any timeout coming in.
>>>> 3) The host receives a command with a tag which is already in-use.
>>>> However, that should have been prevented by the block-layer, which
>>>> really should ensure that this situation never happens.
>>>>
>>>> So either way I look at it, it really looks like a bug and adding a
>>>> timeout handler will just paper over it.
>>>> (Not that a timeout handler is a bad idea, in fact I'm convinced that
>>>> you need one. Just not for this purpose.)
>>>>
>>>> So can you elaborate how this 'busy' status comes about?
>>>> Is the command sent to the device?
>>>>
>>>> Cheers,
>>>>
>>>> Hannes
>>>
>>>
>>> Hi Hannes,
>>>
>>> it's going to be a bit long :)
>>> I think you are missing the point.
>>> I will describe a race condition happened to us a while ago, that was
>>> quite difficult to understand and fix.
>>> So, this patch is not about the "busy" returning to the scsi dispatch
>>> routine. it's about the abort triggered after 30 seconds.
>>>
>>> imagine a request being queued and sent to the scsi, and then to the
>>> ufs.
>>> a timer, initialized to 30 seconds start ticking.
>>> but the request is never sent to the ufs device, as queuecommand()
>>> returns
>>> with "SCSI_MLQUEUE_HOST_BUSY"
>>> by looking at the code, this could happen, for example:
>>> 	err = ufshcd_hold(hba, true);
>>> 	if (err) {
>>> 		err = SCSI_MLQUEUE_HOST_BUSY;
>>> 		goto out;
>>> 	}
>>>
>> Uuhhh.
>> You probably should not have pointed me to that piece of code ...
>> open-coding loops in ufshcd_hold() ... shudder.
>> (Did I ever review that one? Must've ...)
>> _Anyway_: sleeping in queuecommand is always a bad idea, as then
>> precisely those issues you've just described will happen.
>>
>> Couldn't you just call
>> ufshcd_hold(hba, false)
>> instead of
>> ufshcd_hold(hba, true)
>> ?
>> The request will be requeued more-or-less immediately, avoiding the
>> issue with timeout handler kicking in.
>> And the queue will remain blocked until the ungate work item returns, at
>> which point I/O submission will continue.
>> As the request will be requeued to the head of the queue there won't be
>> other I/O competing with tags, so it shouldn't have any adverse effects.
>>
>> Wouldn't that work?
>>
>> Cheers,
>>
>> Hannes
> 
> Hi Hannes
> 
> This is a bug, and it should be fixed.
Oh, definitely agreed. The question is _where_.


> if you choose to bypass it, by calling ufshcd_hold(hba, false), not only
> the race condition is still there, and can pop-out at any other point in
> the future, but also, not sure what are the consequences of
> ufshcd_hold(hba, false) unstead of "true".
Well ... seeing it's your driver, I would've thought _you_ should know ...

> so, changing the already tested and working code, (not to return BUSY from
> queuecommand) is not a fix.
Hey, I did _not_ suggest not to retury BUSY from queuecommand.

I was suggesting this patch:

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 9c1b94b..b9295ad 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -1388,7 +1388,7 @@ static int ufshcd_queuecommand(struct Scsi_Host
*host, struct scsi_cmnd *cmd)
                goto out;
        }

-       err = ufshcd_hold(hba, true);
+       err = ufshcd_hold(hba, false);
        if (err) {
                err = SCSI_MLQUEUE_HOST_BUSY;
                clear_bit_unlock(tag, &hba->lrb_in_use);

which, by reading the code, should be avoiding this issue.
I was just asking you if you could give this patch a spin and see if it
works. If not (for whatever reason) I'm happy to accept your patch.
But first I would like to have an explanation why the above would _not_
work.

Unfortunately I don't have the hardware otherwise I'd be running the
tests myself.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      zSeries & Storage
hare@xxxxxxx			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)