Re: [PATCH v6 net-next] virtio-net: support XDP when not more queues

[Jason Wang <jasowang@redhat.com>]

On 2021/3/8 1:43 下午, Xuan Zhuo wrote:
> On Mon, 8 Mar 2021 11:41:03 +0800, Jason Wang <jasowang@redhat.com> wrote:
>> On 2021/3/5 6:29 下午, Xuan Zhuo wrote:
>>> The number of queues implemented by many virtio backends is limited,
>>> especially some machines have a large number of CPUs. In this case, it
>>> is often impossible to allocate a separate queue for
>>> XDP_TX/XDP_REDIRECT, then xdp cannot be loaded to work, even xdp does
>>> not use the XDP_TX/XDP_REDIRECT.
>>>
>>> This patch allows XDP_TX/XDP_REDIRECT to run by reuse the existing SQ
>>> with __netif_tx_lock() hold when there are not enough queues.
>>>
>>> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
>>> Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
>>> ---
>>> v6: 1. use __netif_tx_acquire()/__netif_tx_release(). (suggested by Jason Wang)
>>>       2. add note for why not lock. (suggested by Jason Wang)
>>>       3. Use variable 'flag' to record with or without locked.  It is not safe to
>>>          use curr_queue_pairs in "virtnet_put_xdp_sq", because it may changed after
>>>          "virtnet_get_xdp_sq".
>>>
>>> v5: change subject from 'support XDP_TX when not more queues'
>>>
>>> v4: make sparse happy
>>>       suggested by Jakub Kicinski
>>>
>>> v3: add warning when no more queues
>>>       suggested by Jesper Dangaard Brouer
>>>
>>>    drivers/net/virtio_net.c | 63 ++++++++++++++++++++++++++++++++++++++++--------
>>>    1 file changed, 53 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
>>> index ba8e637..f9e024d 100644
>>> --- a/drivers/net/virtio_net.c
>>> +++ b/drivers/net/virtio_net.c
>>> @@ -195,6 +195,9 @@ struct virtnet_info {
>>>    	/* # of XDP queue pairs currently used by the driver */
>>>    	u16 xdp_queue_pairs;
>>>
>>> +	/* xdp_queue_pairs may be 0, when xdp is already loaded. So add this. */
>>> +	bool xdp_enabled;
>>> +
>>>    	/* I like... big packets and I cannot lie! */
>>>    	bool big_packets;
>>>
>>> @@ -481,14 +484,48 @@ static int __virtnet_xdp_xmit_one(struct virtnet_info *vi,
>>>    	return 0;
>>>    }
>>>
>>> -static struct send_queue *virtnet_xdp_sq(struct virtnet_info *vi)
>>> +static struct send_queue *virtnet_get_xdp_sq(struct virtnet_info *vi, int *flag)
>>> +	__acquires(txq->_xmit_lock)
>>>    {
>>> +	struct netdev_queue *txq;
>>>    	unsigned int qp;
>>>
>>> -	qp = vi->curr_queue_pairs - vi->xdp_queue_pairs + smp_processor_id();
>>> +	if (vi->curr_queue_pairs > nr_cpu_ids) {
>>> +		qp = vi->curr_queue_pairs - vi->xdp_queue_pairs + smp_processor_id();
>>> +		txq = netdev_get_tx_queue(vi->dev, qp);
>>> +
>>> +		/* In this case, this txq is only used for xdp tx on the current
>>> +		 * cpu, so it does not need to be locked.
>>> +		 * __netif_tx_acquire is for sparse.
>>> +		 */
>>> +		__netif_tx_acquire(txq);
>>> +		*flag = false;
>>> +	} else {
>>> +		qp = smp_processor_id() % vi->curr_queue_pairs;
>>> +		txq = netdev_get_tx_queue(vi->dev, qp);
>>> +		__netif_tx_lock(txq, raw_smp_ƒprocessor_id());
>>> +		*flag = true;
>>> +	}
>>> +
>>>    	return &vi->sq[qp];
>>
>> Two questions:
>>
>> 1) Can we simply check xdp_queue_paris against 0 then we don't need flag?
>> 2) Can we pass txq to virtnet_get_xdp_sq() then the annotation looks
>> even more better?
>>
>> Thanks
> In this patch, I added xdp_enabled to determine the status of xdp, because
> xdp_queue_pairs may be 0 when no more queues.
>
> But we can't use these to determine whether to unlock. Because after lock, this
> variable may change, so we may make an error when we use this variable to
> determine whether to unlock.
>
> When virtnet_get_xdp_sq is called by virtnet_xdp_xmit, it is actually safe to be
> protected by rcu, but it is still called by virtnet_poll. This time is not safe.
> It is possible that xdp_enabled will change after virtnet_get_xdp_sq.


I think NAPI is disabled when we want to set/unset XDP. So xdp_qp can't 
be changed in virtnet_poll()?

Thanks


>   So I think
> it is a better choice to add a local variable to record whether to unlock.
>
> Regarding the second question, I think it's okay. In this way, the code for
> get/put is very simple, I think it is better to use macros directly.
>
> ```
> static struct send_queue *virtnet_xdp_sq(struct virtnet_info *vi, int *flag)
> {
> 	unsigned int qp;
>
> 	if (vi->curr_queue_pairs > nr_cpu_ids) {
> 		qp = vi->curr_queue_pairs - vi->xdp_queue_pairs + smp_processor_id();
> 		*flag = false;
> 	} else {
> 		qp = smp_processor_id() % vi->curr_queue_pairs;
> 		*flag = true;
> 	}
>
> 	return &vi->sq[qp];
> }
>
> #define virtnet_xdp_get_txq(vi, sq, flag)  \
> 	if (flag) \
> 		__netif_tx_acquire(netdev_get_tx_queue(vi->dev, vq2txq(sq->vq))); \
> 	else \
> 		__netif_tx_lock(netdev_get_tx_queue(vi->dev, vq2txq(sq->vq)), \
> 				raw_smp_processor_id());
>
> #define virtnet_xdp_put_txq(vi, sq, flag)  \
> 	if (flag) \
> 		__netif_tx_unlock(netdev_get_tx_queue(vi->dev, vq2txq(sq->vq))); \
> 	else \
> 		__netif_tx_release(netdev_get_tx_queue(vi->dev, vq2txq(sq->vq)));
>
>
>
> 	sq = virtnet_xdp_sq(vi, &sq_flag);
> 	virtnet_xdp_get_txq(vi, sq, sq_flag);
> 	.......
> 	virtnet_xdp_put_txq(vi, sq, sq_flag);
> ```
>
> Do you think this is ok? @Jason
>
> thanks.
>
>>
>>>    }
>>>
>>> +static void virtnet_put_xdp_sq(struct virtnet_info *vi, struct send_queue *sq,
>>> +			       int flag)
>>> +	__releases(txq->_xmit_lock)
>>> +{
>>> +	struct netdev_queue *txq;
>>> +	unsigned int qp;
>>> +
>>> +	qp = sq - vi->sq;
>>> +	txq = netdev_get_tx_queue(vi->dev, qp);
>>> +
>>> +	if (flag)
>>> +		__netif_tx_unlock(txq);
>>> +	else
>>> +		__netif_tx_release(txq);
>>> +}
>>> +
>>>    static int virtnet_xdp_xmit(struct net_device *dev,
>>>    			    int n, struct xdp_frame **frames, u32 flags)
>>>    {
>>> @@ -496,12 +533,12 @@ static int virtnet_xdp_xmit(struct net_device *dev,
>>>    	struct receive_queue *rq = vi->rq;
>>>    	struct bpf_prog *xdp_prog;
>>>    	struct send_queue *sq;
>>> +	int ret, err, sq_flag;
>>>    	unsigned int len;
>>>    	int packets = 0;
>>>    	int bytes = 0;
>>>    	int drops = 0;
>>>    	int kicks = 0;
>>> -	int ret, err;
>>>    	void *ptr;
>>>    	int i;
>>>
>>> @@ -512,7 +549,7 @@ static int virtnet_xdp_xmit(struct net_device *dev,
>>>    	if (!xdp_prog)
>>>    		return -ENXIO;
>>>
>>> -	sq = virtnet_xdp_sq(vi);
>>> +	sq = virtnet_get_xdp_sq(vi, &sq_flag);
>>>
>>>    	if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK)) {
>>>    		ret = -EINVAL;
>>> @@ -560,12 +597,13 @@ static int virtnet_xdp_xmit(struct net_device *dev,
>>>    	sq->stats.kicks += kicks;
>>>    	u64_stats_update_end(&sq->stats.syncp);
>>>
>>> +	virtnet_put_xdp_sq(vi, sq, sq_flag);
>>>    	return ret;
>>>    }
>>>
>>>    static unsigned int virtnet_get_headroom(struct virtnet_info *vi)
>>>    {
>>> -	return vi->xdp_queue_pairs ? VIRTIO_XDP_HEADROOM : 0;
>>> +	return vi->xdp_enabled ? VIRTIO_XDP_HEADROOM : 0;
>>>    }
>>>
>>>    /* We copy the packet for XDP in the following cases:
>>> @@ -1457,12 +1495,15 @@ static int virtnet_poll(struct napi_struct *napi, int budget)
>>>    		xdp_do_flush();
>>>
>>>    	if (xdp_xmit & VIRTIO_XDP_TX) {
>>> -		sq = virtnet_xdp_sq(vi);
>>> +		int sq_flag;
>>> +
>>> +		sq = virtnet_get_xdp_sq(vi, &sq_flag);
>>>    		if (virtqueue_kick_prepare(sq->vq) && virtqueue_notify(sq->vq)) {
>>>    			u64_stats_update_begin(&sq->stats.syncp);
>>>    			sq->stats.kicks++;
>>>    			u64_stats_update_end(&sq->stats.syncp);
>>>    		}
>>> +		virtnet_put_xdp_sq(vi, sq, sq_flag);
>>>    	}
>>>
>>>    	return received;
>>> @@ -2417,10 +2458,9 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
>>>
>>>    	/* XDP requires extra queues for XDP_TX */
>>>    	if (curr_qp + xdp_qp > vi->max_queue_pairs) {
>>> -		NL_SET_ERR_MSG_MOD(extack, "Too few free TX rings available");
>>> -		netdev_warn(dev, "request %i queues but max is %i\n",
>>> +		netdev_warn(dev, "XDP request %i queues but max is %i. XDP_TX and XDP_REDIRECT will operate in a slower locked tx mode.\n",
>>>    			    curr_qp + xdp_qp, vi->max_queue_pairs);
>>> -		return -ENOMEM;
>>> +		xdp_qp = 0;
>>>    	}
>>>
>>>    	old_prog = rtnl_dereference(vi->rq[0].xdp_prog);
>>> @@ -2454,11 +2494,14 @@ static int virtnet_xdp_set(struct net_device *dev, struct bpf_prog *prog,
>>>    	vi->xdp_queue_pairs = xdp_qp;
>>>
>>>    	if (prog) {
>>> +		vi->xdp_enabled = true;
>>>    		for (i = 0; i < vi->max_queue_pairs; i++) {
>>>    			rcu_assign_pointer(vi->rq[i].xdp_prog, prog);
>>>    			if (i == 0 && !old_prog)
>>>    				virtnet_clear_guest_offloads(vi);
>>>    		}
>>> +	} else {
>>> +		vi->xdp_enabled = false;
>>>    	}
>>>
>>>    	for (i = 0; i < vi->max_queue_pairs; i++) {
>>> @@ -2526,7 +2569,7 @@ static int virtnet_set_features(struct net_device *dev,
>>>    	int err;
>>>
>>>    	if ((dev->features ^ features) & NETIF_F_LRO) {
>>> -		if (vi->xdp_queue_pairs)
>>> +		if (vi->xdp_enabled)
>>>    			return -EBUSY;
>>>
>>>    		if (features & NETIF_F_LRO)
>>> --
>>> 1.8.3.1
>>>