Re: [PATCH v4 2/7] s390: vfio-ap: wait for queue empty on queue reset

[Harald Freudenberger <freude@linux.ibm.com>]

On 13.06.19 21:39, Tony Krowiak wrote:
> Refactors the AP queue reset function to wait until the queue is empty
> after the PQAP(ZAPQ) instruction is executed to zero out the queue as
> required by the AP architecture.
>
> Signed-off-by: Tony Krowiak <akrowiak@linux.ibm.com>
> ---
>  drivers/s390/crypto/vfio_ap_ops.c | 49 +++++++++++++++++++++++++++------------
>  1 file changed, 34 insertions(+), 15 deletions(-)
>
> diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
> index bf2ab02b9a0b..60efd3d7896d 100644
> --- a/drivers/s390/crypto/vfio_ap_ops.c
> +++ b/drivers/s390/crypto/vfio_ap_ops.c
> @@ -1128,23 +1128,46 @@ static void vfio_ap_irq_disable_apqn(int apqn)
>  	}
>  }
>  
> -int vfio_ap_mdev_reset_queue(unsigned int apid, unsigned int apqi,
> -			     unsigned int retry)
> +static void vfio_ap_mdev_wait_for_qempty(ap_qid_t qid)
> +{
> +	struct ap_queue_status status;
> +	int retry = 5;
> +
> +	do {
> +		status = ap_tapq(qid, NULL);
> +		switch (status.response_code) {
> +		case AP_RESPONSE_NORMAL:
> +			if (status.queue_empty)
> +				return;
> +		case AP_RESPONSE_RESET_IN_PROGRESS:
> +		case AP_RESPONSE_BUSY:
> +			msleep(20);
> +			break;
> +		default:
> +			pr_warn("%s: tapq response %02x waiting for queue %04x.%02x empty\n",
> +				__func__, status.response_code,
> +				AP_QID_CARD(qid), AP_QID_QUEUE(qid));
The ap and zcrypt code uses %02x.%04x for displaying an APQN.
I would also recommend to handle 0x01 (AP_RESPONSE_Q_NOT_AVAIL) and
0x03 (AP_RESPONSE_DECONFIGURED) as this code may run when a APQN
is removed from the configuration of the LPAR. However, it's up to you if you
want to handle these with the default statement and issue an sysfs warning.
> +			return;
> +		}
> +	} while (--retry);
> +
> +	WARN_ON_ONCE(retry <= 0);
> +}
> +
> +int vfio_ap_mdev_reset_queue(unsigned int apid, unsigned int apqi)
>  {
>  	struct ap_queue_status status;
> -	int retry2 = 2;
>  	int apqn = AP_MKQID(apid, apqi);
> +	int retry = 5;
>  
>  	do {
>  		status = ap_zapq(apqn);
>  		switch (status.response_code) {
>  		case AP_RESPONSE_NORMAL:
> -			while (!status.queue_empty && retry2--) {
> -				msleep(20);
> -				status = ap_tapq(apqn, NULL);
> -			}
> -			WARN_ON_ONCE(retry <= 0);
> +			vfio_ap_mdev_wait_for_qempty(AP_MKQID(apid, apqi));
>  			return 0;
> +		case AP_RESPONSE_DECONFIGURED:
> +			return -ENODEV;
>  		case AP_RESPONSE_RESET_IN_PROGRESS:
>  		case AP_RESPONSE_BUSY:
>  			msleep(20);
> @@ -1169,14 +1192,10 @@ static int vfio_ap_mdev_reset_queues(struct mdev_device *mdev)
>  			     matrix_mdev->matrix.apm_max + 1) {
>  		for_each_set_bit_inv(apqi, matrix_mdev->matrix.aqm,
>  				     matrix_mdev->matrix.aqm_max + 1) {
> -			ret = vfio_ap_mdev_reset_queue(apid, apqi, 1);
> -			/*
> -			 * Regardless whether a queue turns out to be busy, or
> -			 * is not operational, we need to continue resetting
> -			 * the remaining queues.
> -			 */
> +			ret = vfio_ap_mdev_reset_queue(apid, apqi);
>  			if (ret)
>  				rc = ret;
> +
>  			vfio_ap_irq_disable_apqn(AP_MKQID(apid, apqi));
>  		}
>  	}
> @@ -1326,7 +1345,7 @@ void vfio_ap_mdev_remove_queue(struct ap_queue *queue)
>  	dev_set_drvdata(&queue->ap_dev.device, NULL);
>  	apid = AP_QID_CARD(q->apqn);
>  	apqi = AP_QID_QUEUE(q->apqn);
> -	vfio_ap_mdev_reset_queue(apid, apqi, 1);
> +	vfio_ap_mdev_reset_queue(apid, apqi);
>  	vfio_ap_irq_disable(q);
>  	kfree(q);
>  }