Re: [PATCH v2 06/17] ibmvfc: add handlers to drain and complete Sub-CRQ responses

[Tyrel Datwyler <tyreld@linux.ibm.com>]

On 12/2/20 7:46 AM, Brian King wrote:
> On 12/1/20 6:53 PM, Tyrel Datwyler wrote:
>> The logic for iterating over the Sub-CRQ responses is similiar to that
>> of the primary CRQ. Add the necessary handlers for processing those
>> responses.
>>
>> Signed-off-by: Tyrel Datwyler <tyreld@linux.ibm.com>
>> ---
>>  drivers/scsi/ibmvscsi/ibmvfc.c | 77 ++++++++++++++++++++++++++++++++++
>>  1 file changed, 77 insertions(+)
>>
>> diff --git a/drivers/scsi/ibmvscsi/ibmvfc.c b/drivers/scsi/ibmvscsi/ibmvfc.c
>> index 97f00fefa809..e9da3f60c793 100644
>> --- a/drivers/scsi/ibmvscsi/ibmvfc.c
>> +++ b/drivers/scsi/ibmvscsi/ibmvfc.c
>> @@ -3381,6 +3381,83 @@ static int ibmvfc_toggle_scrq_irq(struct ibmvfc_sub_queue *scrq, int enable)
>>  	return rc;
>>  }
>>  
>> +static void ibmvfc_handle_scrq(struct ibmvfc_crq *crq, struct ibmvfc_host *vhost)
>> +{
>> +	struct ibmvfc_event *evt = (struct ibmvfc_event *)be64_to_cpu(crq->ioba);
>> +	unsigned long flags;
>> +
>> +	switch (crq->valid) {
>> +	case IBMVFC_CRQ_CMD_RSP:
>> +		break;
>> +	case IBMVFC_CRQ_XPORT_EVENT:
>> +		return;
>> +	default:
>> +		dev_err(vhost->dev, "Got and invalid message type 0x%02x\n", crq->valid);
>> +		return;
>> +	}
>> +
>> +	/* The only kind of payload CRQs we should get are responses to
>> +	 * things we send. Make sure this response is to something we
>> +	 * actually sent
>> +	 */
>> +	if (unlikely(!ibmvfc_valid_event(&vhost->pool, evt))) {
>> +		dev_err(vhost->dev, "Returned correlation_token 0x%08llx is invalid!\n",
>> +			crq->ioba);
>> +		return;
>> +	}
>> +
>> +	if (unlikely(atomic_read(&evt->free))) {
>> +		dev_err(vhost->dev, "Received duplicate correlation_token 0x%08llx!\n",
>> +			crq->ioba);
>> +		return;
>> +	}
>> +
>> +	spin_lock_irqsave(vhost->host->host_lock, flags);
>> +	del_timer(&evt->timer);
>> +	list_del(&evt->queue);
>> +	ibmvfc_trc_end(evt);
>> +	spin_unlock_irqrestore(vhost->host->host_lock, flags);
>> +	evt->done(evt);
>> +}
>> +
>> +static struct ibmvfc_crq *ibmvfc_next_scrq(struct ibmvfc_sub_queue *scrq)
>> +{
>> +	struct ibmvfc_crq *crq;
>> +
>> +	crq = &scrq->msgs[scrq->cur].crq;
>> +	if (crq->valid & 0x80) {
>> +		if (++scrq->cur == scrq->size)
> 
> You are incrementing the cur pointer without any locks held. Although
> unlikely, could you also be in ibmvfc_reset_crq in another thread?
> If so, you'd have a subtle race condition here where the cur pointer could
> be read, then ibmvfc_reset_crq writes it to zero, then this thread
> writes it to a non zero value, which would then cause you to be out of
> sync with the VIOS as to where the cur pointer is.

Oof, yeah I was previously holding the lock the whole time, but switched it up
once I realized I can't complete a scsi command with the lock held, and got a
little too loose with it.

-Tyrel
> 
>> +			scrq->cur = 0;
>> +		rmb();
>> +	} else
>> +		crq = NULL;
>> +
>> +	return crq;
>> +}
>> +
> 
> 
>