Virtualization - Jan 2023 - [PATCH net V3] virtio-net: correctly enable callback during start_xmit

Commit a7766ef18b33("virtio_net: disable cb aggressively") enables
virtqueue callback via the following statement:

        do {
		if (use_napi)
			virtqueue_disable_cb(sq->vq);

		free_old_xmit_skbs(sq, false);

	} while (use_napi && kick &&
               unlikely(!virtqueue_enable_cb_delayed(sq->vq)));

When NAPI is used and kick is false, the callback won't be enabled
here. And when the virtqueue is about to be full, the tx will be
disabled, but we still don't enable tx interrupt which will cause a TX
hang. This could be observed when using pktgen with burst enabled.

TO be consistent with the logic that tries to disable cb only for
NAPI, fixing this by trying to enable delayed callback only when NAPI
is enabled when the queue is about to be full.

Fixes: a7766ef18b33 ("virtio_net: disable cb aggressively")
Signed-off-by: Jason Wang <jasowang at redhat.com>
---
The patch is needed for -stable.
Changes since V2:
- try to enabled delayed callback and schedule NAPI instead of try to
  poll as when TX NAPI is disabled
Changes since V1:
- enable tx interrupt after we disable TX
---
 drivers/net/virtio_net.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 7723b2a49d8e..18b3de854aeb 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -1877,8 +1877,10 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct
net_device *dev)
 	 */
 	if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
 		netif_stop_subqueue(dev, qnum);
-		if (!use_napi &&
-		    unlikely(!virtqueue_enable_cb_delayed(sq->vq))) {
+		if (use_napi) {
+			if (unlikely(!virtqueue_enable_cb_delayed(sq->vq)))
+				virtqueue_napi_schedule(&sq->napi, sq->vq);
+		} else if (unlikely(!virtqueue_enable_cb_delayed(sq->vq))) {
 			/* More just got used, free them then recheck. */
 			free_old_xmit_skbs(sq, false);
 			if (sq->vq->num_free >= 2+MAX_SKB_FRAGS) {
-- 
2.25.1

On 1/17/23 04:47, Jason Wang wrote:
> Commit a7766ef18b33("virtio_net: disable cb aggressively")
enables
> virtqueue callback via the following statement:
> 
>          do {
> 		if (use_napi)
> 			virtqueue_disable_cb(sq->vq);
> 
> 		free_old_xmit_skbs(sq, false);
> 
> 	} while (use_napi && kick &&
>                 unlikely(!virtqueue_enable_cb_delayed(sq->vq)));
> 
> When NAPI is used and kick is false, the callback won't be enabled
> here. And when the virtqueue is about to be full, the tx will be
> disabled, but we still don't enable tx interrupt which will cause a TX
> hang. This could be observed when using pktgen with burst enabled.
> 
> TO be consistent with the logic that tries to disable cb only for
> NAPI, fixing this by trying to enable delayed callback only when NAPI
> is enabled when the queue is about to be full.
> 
> Fixes: a7766ef18b33 ("virtio_net: disable cb aggressively")
> Signed-off-by: Jason Wang <jasowang at redhat.com>
> ---
> The patch is needed for -stable.
> Changes since V2:
> - try to enabled delayed callback and schedule NAPI instead of try to
>    poll as when TX NAPI is disabled
> Changes since V1:
> - enable tx interrupt after we disable TX
> ---
>   drivers/net/virtio_net.c | 6 ++++--
>   1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 7723b2a49d8e..18b3de854aeb 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -1877,8 +1877,10 @@ static netdev_tx_t start_xmit(struct sk_buff *skb,
struct net_device *dev)
>   	 */
>   	if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
>   		netif_stop_subqueue(dev, qnum);
> -		if (!use_napi &&
> -		    unlikely(!virtqueue_enable_cb_delayed(sq->vq))) {
> +		if (use_napi) {
> +			if (unlikely(!virtqueue_enable_cb_delayed(sq->vq)))
> +				virtqueue_napi_schedule(&sq->napi, sq->vq);
> +		} else if (unlikely(!virtqueue_enable_cb_delayed(sq->vq))) {
>   			/* More just got used, free them then recheck. */
>   			free_old_xmit_skbs(sq, false);
>   			if (sq->vq->num_free >= 2+MAX_SKB_FRAGS) {
This fix works fine with my test case (netdev stream + passt)

Tested-by: Laurent Vivier <lvivier at redhat.com>

Thanks,
Laurent