[PATCH net 0/2] add checking sq is full inside xdp xmit

[PATCH net 0/2] add checking sq is full inside xdp xmit

[Xuan Zhuo]

If the queue of xdp xmit is not an independent queue, then when the xdp
xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
the following error.

net ens4: Unexpected TXQ (0) queue failure: -28

This patch adds a check whether sq is full in XDP Xmit.

Thanks.

Xuan Zhuo (2):
  virtio_net: separate the logic of checking whether sq is full
  virtio_net: add checking sq is full inside xdp xmit

 drivers/net/virtio_net.c | 78 ++++++++++++++++++++++++----------------
 1 file changed, 47 insertions(+), 31 deletions(-)

--
2.32.0.3.g01195cf9f

[PATCH net 1/2] virtio_net: separate the logic of checking whether sq is full

[Xuan Zhuo]

Separate the logic of checking whether sq is full. The subsequent patch
will reuse this func.

Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
---
 drivers/net/virtio_net.c | 59 ++++++++++++++++++++++++----------------
 1 file changed, 35 insertions(+), 24 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index fb5e68ed3ec2..777de0ec0b1b 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -1750,6 +1750,40 @@ static void free_old_xmit_skbs(struct send_queue *sq, bool in_napi)
 	u64_stats_update_end(&sq->stats.syncp);
 }
 
+static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
+			  struct send_queue *sq)
+{
+	bool use_napi = sq->napi.weight;
+	int qnum;
+
+	qnum = sq - vi->sq;
+
+	/* If running out of space, stop queue to avoid getting packets that we
+	 * are then unable to transmit.
+	 * An alternative would be to force queuing layer to requeue the skb by
+	 * returning NETDEV_TX_BUSY. However, NETDEV_TX_BUSY should not be
+	 * returned in a normal path of operation: it means that driver is not
+	 * maintaining the TX queue stop/start state properly, and causes
+	 * the stack to do a non-trivial amount of useless work.
+	 * Since most packets only take 1 or 2 ring slots, stopping the queue
+	 * early means 16 slots are typically wasted.
+	 */
+	if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
+		netif_stop_subqueue(dev, qnum);
+		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) {
+				netif_start_subqueue(dev, qnum);
+				virtqueue_disable_cb(sq->vq);
+			}
+		}
+	}
+}
+
 static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
 {
 	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
@@ -1989,30 +2023,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct net_device *dev)
 		nf_reset_ct(skb);
 	}
 
-	/* If running out of space, stop queue to avoid getting packets that we
-	 * are then unable to transmit.
-	 * An alternative would be to force queuing layer to requeue the skb by
-	 * returning NETDEV_TX_BUSY. However, NETDEV_TX_BUSY should not be
-	 * returned in a normal path of operation: it means that driver is not
-	 * maintaining the TX queue stop/start state properly, and causes
-	 * the stack to do a non-trivial amount of useless work.
-	 * Since most packets only take 1 or 2 ring slots, stopping the queue
-	 * early means 16 slots are typically wasted.
-	 */
-	if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
-		netif_stop_subqueue(dev, qnum);
-		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) {
-				netif_start_subqueue(dev, qnum);
-				virtqueue_disable_cb(sq->vq);
-			}
-		}
-	}
+	check_sq_full(vi, dev, sq);
 
 	if (kick || netif_xmit_stopped(txq)) {
 		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq)) {
-- 
2.32.0.3.g01195cf9f

Re: [PATCH net 1/2] virtio_net: separate the logic of checking whether sq is full

[Jason Wang]

On Mon, Mar 6, 2023 at 12:15 PM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
>
> Separate the logic of checking whether sq is full. The subsequent patch
> will reuse this func.
>
> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> ---
>  drivers/net/virtio_net.c | 59 ++++++++++++++++++++++++----------------
>  1 file changed, 35 insertions(+), 24 deletions(-)
>
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index fb5e68ed3ec2..777de0ec0b1b 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -1750,6 +1750,40 @@ static void free_old_xmit_skbs(struct send_queue *sq, bool in_napi)
>         u64_stats_update_end(&sq->stats.syncp);
>  }
>
> +static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
> +                         struct send_queue *sq)
> +{

Nit: we need a better name, it's not only the check but also can disable TX.

Thanks

> +       bool use_napi = sq->napi.weight;
> +       int qnum;
> +
> +       qnum = sq - vi->sq;
> +
> +       /* If running out of space, stop queue to avoid getting packets that we
> +        * are then unable to transmit.
> +        * An alternative would be to force queuing layer to requeue the skb by
> +        * returning NETDEV_TX_BUSY. However, NETDEV_TX_BUSY should not be
> +        * returned in a normal path of operation: it means that driver is not
> +        * maintaining the TX queue stop/start state properly, and causes
> +        * the stack to do a non-trivial amount of useless work.
> +        * Since most packets only take 1 or 2 ring slots, stopping the queue
> +        * early means 16 slots are typically wasted.
> +        */
> +       if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
> +               netif_stop_subqueue(dev, qnum);
> +               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) {
> +                               netif_start_subqueue(dev, qnum);
> +                               virtqueue_disable_cb(sq->vq);
> +                       }
> +               }
> +       }
> +}
> +
>  static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
>  {
>         if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
> @@ -1989,30 +2023,7 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct net_device *dev)
>                 nf_reset_ct(skb);
>         }
>
> -       /* If running out of space, stop queue to avoid getting packets that we
> -        * are then unable to transmit.
> -        * An alternative would be to force queuing layer to requeue the skb by
> -        * returning NETDEV_TX_BUSY. However, NETDEV_TX_BUSY should not be
> -        * returned in a normal path of operation: it means that driver is not
> -        * maintaining the TX queue stop/start state properly, and causes
> -        * the stack to do a non-trivial amount of useless work.
> -        * Since most packets only take 1 or 2 ring slots, stopping the queue
> -        * early means 16 slots are typically wasted.
> -        */
> -       if (sq->vq->num_free < 2+MAX_SKB_FRAGS) {
> -               netif_stop_subqueue(dev, qnum);
> -               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) {
> -                               netif_start_subqueue(dev, qnum);
> -                               virtqueue_disable_cb(sq->vq);
> -                       }
> -               }
> -       }
> +       check_sq_full(vi, dev, sq);
>
>         if (kick || netif_xmit_stopped(txq)) {
>                 if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq)) {
> --
> 2.32.0.3.g01195cf9f
>

[PATCH net 2/2] virtio_net: add checking sq is full inside xdp xmit

[Xuan Zhuo]

If the queue of xdp xmit is not an independent queue, then when the xdp
xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
the following error.

net ens4: Unexpected TXQ (0) queue failure: -28

This patch adds a check whether sq is full in XDP Xmit.

Reported-by: Yichun Zhang <yichun@openresty.com>
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
---
 drivers/net/virtio_net.c | 25 +++++++++++++++----------
 1 file changed, 15 insertions(+), 10 deletions(-)

diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index 777de0ec0b1b..3001b9a548e5 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -302,6 +302,8 @@ struct padded_vnet_hdr {
 
 static void virtnet_rq_free_unused_buf(struct virtqueue *vq, void *buf);
 static void virtnet_sq_free_unused_buf(struct virtqueue *vq, void *buf);
+static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
+			  struct send_queue *sq);
 
 static bool is_xdp_frame(void *ptr)
 {
@@ -341,6 +343,16 @@ static int rxq2vq(int rxq)
 	return rxq * 2;
 }
 
+static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
+{
+	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
+		return false;
+	else if (q < vi->curr_queue_pairs)
+		return true;
+	else
+		return false;
+}
+
 static inline struct virtio_net_hdr_mrg_rxbuf *skb_vnet_hdr(struct sk_buff *skb)
 {
 	return (struct virtio_net_hdr_mrg_rxbuf *)skb->cb;
@@ -686,6 +698,9 @@ static int virtnet_xdp_xmit(struct net_device *dev,
 	}
 	ret = nxmit;
 
+	if (!is_xdp_raw_buffer_queue(vi, sq - vi->sq))
+		check_sq_full(vi, dev, sq);
+
 	if (flags & XDP_XMIT_FLUSH) {
 		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq))
 			kicks = 1;
@@ -1784,16 +1799,6 @@ static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
 	}
 }
 
-static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
-{
-	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
-		return false;
-	else if (q < vi->curr_queue_pairs)
-		return true;
-	else
-		return false;
-}
-
 static void virtnet_poll_cleantx(struct receive_queue *rq)
 {
 	struct virtnet_info *vi = rq->vq->vdev->priv;
-- 
2.32.0.3.g01195cf9f

Re: [PATCH net 2/2] virtio_net: add checking sq is full inside xdp xmit

[Michael S. Tsirkin]

On Mon, Mar 06, 2023 at 12:15:35PM +0800, Xuan Zhuo wrote:
> If the queue of xdp xmit is not an independent queue, then when the xdp
> xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> the following error.
> 
> net ens4: Unexpected TXQ (0) queue failure: -28
> 
> This patch adds a check whether sq is full in XDP Xmit.
> 
> Reported-by: Yichun Zhang <yichun@openresty.com>
> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> ---
>  drivers/net/virtio_net.c | 25 +++++++++++++++----------
>  1 file changed, 15 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index 777de0ec0b1b..3001b9a548e5 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -302,6 +302,8 @@ struct padded_vnet_hdr {
>  
>  static void virtnet_rq_free_unused_buf(struct virtqueue *vq, void *buf);
>  static void virtnet_sq_free_unused_buf(struct virtqueue *vq, void *buf);
> +static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
> +			  struct send_queue *sq);
>  
>  static bool is_xdp_frame(void *ptr)
>  {
> @@ -341,6 +343,16 @@ static int rxq2vq(int rxq)
>  	return rxq * 2;
>  }
>  

I'd really rather we ordered functions reasonably so declarations
are not needed.

> +static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
> +{
> +	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
> +		return false;
> +	else if (q < vi->curr_queue_pairs)
> +		return true;
> +	else
> +		return false;
> +}
> +
>  static inline struct virtio_net_hdr_mrg_rxbuf *skb_vnet_hdr(struct sk_buff *skb)
>  {
>  	return (struct virtio_net_hdr_mrg_rxbuf *)skb->cb;
> @@ -686,6 +698,9 @@ static int virtnet_xdp_xmit(struct net_device *dev,
>  	}
>  	ret = nxmit;
>  
> +	if (!is_xdp_raw_buffer_queue(vi, sq - vi->sq))
> +		check_sq_full(vi, dev, sq);
> +
>  	if (flags & XDP_XMIT_FLUSH) {
>  		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq))
>  			kicks = 1;
> @@ -1784,16 +1799,6 @@ static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
>  	}
>  }
>  
> -static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
> -{
> -	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
> -		return false;
> -	else if (q < vi->curr_queue_pairs)
> -		return true;
> -	else
> -		return false;
> -}
> -
>  static void virtnet_poll_cleantx(struct receive_queue *rq)
>  {
>  	struct virtnet_info *vi = rq->vq->vdev->priv;
> -- 
> 2.32.0.3.g01195cf9f

Re: [PATCH net 2/2] virtio_net: add checking sq is full inside xdp xmit

[Xuan Zhuo]

On Mon, 6 Mar 2023 12:57:34 -0500, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> On Mon, Mar 06, 2023 at 12:15:35PM +0800, Xuan Zhuo wrote:
> > If the queue of xdp xmit is not an independent queue, then when the xdp
> > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > the following error.
> >
> > net ens4: Unexpected TXQ (0) queue failure: -28
> >
> > This patch adds a check whether sq is full in XDP Xmit.
> >
> > Reported-by: Yichun Zhang <yichun@openresty.com>
> > Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> > ---
> >  drivers/net/virtio_net.c | 25 +++++++++++++++----------
> >  1 file changed, 15 insertions(+), 10 deletions(-)
> >
> > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > index 777de0ec0b1b..3001b9a548e5 100644
> > --- a/drivers/net/virtio_net.c
> > +++ b/drivers/net/virtio_net.c
> > @@ -302,6 +302,8 @@ struct padded_vnet_hdr {
> >
> >  static void virtnet_rq_free_unused_buf(struct virtqueue *vq, void *buf);
> >  static void virtnet_sq_free_unused_buf(struct virtqueue *vq, void *buf);
> > +static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
> > +			  struct send_queue *sq);
> >
> >  static bool is_xdp_frame(void *ptr)
> >  {
> > @@ -341,6 +343,16 @@ static int rxq2vq(int rxq)
> >  	return rxq * 2;
> >  }
> >
>
> I'd really rather we ordered functions reasonably so declarations
> are not needed.

Yes, but this function depends on some other functions. If we put this function
above, then these dependent functions must be processed. The current method
should be a relatively simple way.

Thanks.


>
> > +static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
> > +{
> > +	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
> > +		return false;
> > +	else if (q < vi->curr_queue_pairs)
> > +		return true;
> > +	else
> > +		return false;
> > +}
> > +
> >  static inline struct virtio_net_hdr_mrg_rxbuf *skb_vnet_hdr(struct sk_buff *skb)
> >  {
> >  	return (struct virtio_net_hdr_mrg_rxbuf *)skb->cb;
> > @@ -686,6 +698,9 @@ static int virtnet_xdp_xmit(struct net_device *dev,
> >  	}
> >  	ret = nxmit;
> >
> > +	if (!is_xdp_raw_buffer_queue(vi, sq - vi->sq))
> > +		check_sq_full(vi, dev, sq);
> > +
> >  	if (flags & XDP_XMIT_FLUSH) {
> >  		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq))
> >  			kicks = 1;
> > @@ -1784,16 +1799,6 @@ static void check_sq_full(struct virtnet_info *vi, struct net_device *dev,
> >  	}
> >  }
> >
> > -static bool is_xdp_raw_buffer_queue(struct virtnet_info *vi, int q)
> > -{
> > -	if (q < (vi->curr_queue_pairs - vi->xdp_queue_pairs))
> > -		return false;
> > -	else if (q < vi->curr_queue_pairs)
> > -		return true;
> > -	else
> > -		return false;
> > -}
> > -
> >  static void virtnet_poll_cleantx(struct receive_queue *rq)
> >  {
> >  	struct virtnet_info *vi = rq->vq->vdev->priv;
> > --
> > 2.32.0.3.g01195cf9f
>

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Alexander H Duyck]

On Mon, 2023-03-06 at 12:15 +0800, Xuan Zhuo wrote:
> If the queue of xdp xmit is not an independent queue, then when the xdp
> xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> the following error.
> 
> net ens4: Unexpected TXQ (0) queue failure: -28
> 
> This patch adds a check whether sq is full in XDP Xmit.
> 
> Thanks.
> 
> Xuan Zhuo (2):
>   virtio_net: separate the logic of checking whether sq is full
>   virtio_net: add checking sq is full inside xdp xmit
> 
>  drivers/net/virtio_net.c | 78 ++++++++++++++++++++++++----------------
>  1 file changed, 47 insertions(+), 31 deletions(-)
> 
> --
> 2.32.0.3.g01195cf9f
> 

Series looks good to me.

Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Michael S. Tsirkin]

On Mon, Mar 06, 2023 at 12:15:33PM +0800, Xuan Zhuo wrote:
> If the queue of xdp xmit is not an independent queue, then when the xdp
> xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> the following error.
> 
> net ens4: Unexpected TXQ (0) queue failure: -28
> 
> This patch adds a check whether sq is full in XDP Xmit.
> 
> Thanks.

Acked-by: Michael S. Tsirkin <mst@redhat.com>

needed for stable?

> Xuan Zhuo (2):
>   virtio_net: separate the logic of checking whether sq is full
>   virtio_net: add checking sq is full inside xdp xmit
> 
>  drivers/net/virtio_net.c | 78 ++++++++++++++++++++++++----------------
>  1 file changed, 47 insertions(+), 31 deletions(-)
> 
> --
> 2.32.0.3.g01195cf9f

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Xuan Zhuo]

On Mon, 6 Mar 2023 12:58:22 -0500, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> On Mon, Mar 06, 2023 at 12:15:33PM +0800, Xuan Zhuo wrote:
> > If the queue of xdp xmit is not an independent queue, then when the xdp
> > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > the following error.
> >
> > net ens4: Unexpected TXQ (0) queue failure: -28
> >
> > This patch adds a check whether sq is full in XDP Xmit.
> >
> > Thanks.
>
> Acked-by: Michael S. Tsirkin <mst@redhat.com>
>
> needed for stable?


Yes i think.

Thanks.


>
> > Xuan Zhuo (2):
> >   virtio_net: separate the logic of checking whether sq is full
> >   virtio_net: add checking sq is full inside xdp xmit
> >
> >  drivers/net/virtio_net.c | 78 ++++++++++++++++++++++++----------------
> >  1 file changed, 47 insertions(+), 31 deletions(-)
> >
> > --
> > 2.32.0.3.g01195cf9f
>

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Paolo Abeni]

Hi,
On Tue, 2023-03-07 at 09:49 +0800, Xuan Zhuo wrote:
> On Mon, 6 Mar 2023 12:58:22 -0500, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > On Mon, Mar 06, 2023 at 12:15:33PM +0800, Xuan Zhuo wrote:
> > > If the queue of xdp xmit is not an independent queue, then when the xdp
> > > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > > the following error.
> > > 
> > > net ens4: Unexpected TXQ (0) queue failure: -28
> > > 
> > > This patch adds a check whether sq is full in XDP Xmit.
> > > 
> > > Thanks.
> > 
> > Acked-by: Michael S. Tsirkin <mst@redhat.com>
> > 
> > needed for stable?
> 
> Yes i think.

Could you please re-post including a suitable 'Fixes' tag? That would
address stable, too. Additionally you could rename check_sq_full() in
patch 1, perhaps 'check_disable_sq_full()' would do.

You can retain the already collected tags.

Thanks!

Paolo

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Xuan Zhuo]

On Tue, 07 Mar 2023 10:53:41 +0100, Paolo Abeni <pabeni@redhat.com> wrote:
> Hi,
> On Tue, 2023-03-07 at 09:49 +0800, Xuan Zhuo wrote:
> > On Mon, 6 Mar 2023 12:58:22 -0500, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > > On Mon, Mar 06, 2023 at 12:15:33PM +0800, Xuan Zhuo wrote:
> > > > If the queue of xdp xmit is not an independent queue, then when the xdp
> > > > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > > > the following error.
> > > >
> > > > net ens4: Unexpected TXQ (0) queue failure: -28
> > > >
> > > > This patch adds a check whether sq is full in XDP Xmit.
> > > >
> > > > Thanks.
> > >
> > > Acked-by: Michael S. Tsirkin <mst@redhat.com>
> > >
> > > needed for stable?
> >
> > Yes i think.
>
> Could you please re-post including a suitable 'Fixes' tag? That would
> address stable, too. Additionally you could rename check_sq_full() in
> patch 1, perhaps 'check_disable_sq_full()' would do.
>
> You can retain the already collected tags.

OK

Thanks.


>
> Thanks!
>
> Paolo
>

Re: [PATCH net 0/2] add checking sq is full inside xdp xmit

[Michael S. Tsirkin]

On Tue, Mar 07, 2023 at 10:53:41AM +0100, Paolo Abeni wrote:
> Hi,
> On Tue, 2023-03-07 at 09:49 +0800, Xuan Zhuo wrote:
> > On Mon, 6 Mar 2023 12:58:22 -0500, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > > On Mon, Mar 06, 2023 at 12:15:33PM +0800, Xuan Zhuo wrote:
> > > > If the queue of xdp xmit is not an independent queue, then when the xdp
> > > > xmit used all the desc, the xmit from the __dev_queue_xmit() may encounter
> > > > the following error.
> > > > 
> > > > net ens4: Unexpected TXQ (0) queue failure: -28
> > > > 
> > > > This patch adds a check whether sq is full in XDP Xmit.
> > > > 
> > > > Thanks.
> > > 
> > > Acked-by: Michael S. Tsirkin <mst@redhat.com>
> > > 
> > > needed for stable?
> > 
> > Yes i think.
> 
> Could you please re-post including a suitable 'Fixes' tag? That would
> address stable, too. Additionally you could rename check_sq_full() in
> patch 1, perhaps 'check_disable_sq_full()' would do.

Not that's even more confusing it sounds like it's checking that
it's checking that sq is disabled unless one looks closely.

I think check_sq_full_and_disable() is better.

> You can retain the already collected tags.
> 
> Thanks!
> 
> Paolo