Re: [PATCH v3] tuntap: add flow control to support back pressure

[Michael S. Tsirkin]

On Mon, Apr 14, 2014 at 09:21:47AM -0400, Steven Galgano wrote:
> Added optional per queue flow control support using IFF_FLOW_CONTROL. When the 
> IFF_FLOW_CONTROL TUNSETIFF flag is specified it will set a per queue flag to 
> indicate that the queue should be stopped using netif_tx_stop_queue(), rather 
> than discarding frames once full. After reading a frame from the respective 
> stopped queue, a netif_tx_wake_queue() is issued to signal resource 
> availability.
> 
> The per queue TUN_FLOW_CONTROL flag is stored in struct tun_file. This provides 
> the flexibility to enable flow control on all, none or some queues when using 
> IFF_MULTI_QUEUE. When not using IFF_MULTI_QUEUE, IFF_FLOW_CONTROL will apply to 
> the single queue. No changes were made to the default drop frame policy.
> 
> This change adds support for back pressure use cases.
> 
> Reported-by: Brian Adamson <brian.adamson@xxxxxxxxxxxx>
> Tested-by: Joseph Giovatto <jgiovatto@xxxxxxxxxxxxxxxx>
> Signed-off-by: Steven Galgano <sgalgano@xxxxxxxxxxxxxxxx>
> ---
> Version 3 patch reformatted commit message to be 80 columns max width. 
>  Corrected Tested-By email address.
> Version 2 patch added support for individual queues when applying flow
> control instead of using netif_tx_stop_all_queues()/netif_tx_wake_all_queues().

Any plans to try and address the more material comments on v2?

>  drivers/net/tun.c           | 32 ++++++++++++++++++++++++++++----
>  include/uapi/linux/if_tun.h |  2 ++
>  2 files changed, 30 insertions(+), 4 deletions(-)
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index ee328ba..3d09f5a 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -137,7 +137,7 @@ struct tun_file {
>  	struct tun_struct __rcu *tun;
>  	struct net *net;
>  	struct fasync_struct *fasync;
> -	/* only used for fasnyc */
> +	/* used for fasnyc and flow control */
>  	unsigned int flags;
>  	union {
>  		u16 queue_index;
> @@ -783,8 +783,19 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev)
>  	 * number of queues.
>  	 */
>  	if (skb_queue_len(&tfile->socket.sk->sk_receive_queue) * numqueues
> -			  >= dev->tx_queue_len)
> -		goto drop;
> +			>= dev->tx_queue_len) {
> +		if (tfile->flags & TUN_FLOW_CONTROL) {
> +			/* Resources unavailable stop queue */
> +			netif_tx_stop_queue(netdev_get_tx_queue(dev, txq));
> +
> +			/* We won't see all dropped packets individually, so
> +			 * over run error is more appropriate.
> +			 */
> +			dev->stats.tx_fifo_errors++;
> +		} else {
> +			goto drop;
> +		}
> +	}
>  
>  	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
>  		goto drop;
> @@ -1333,6 +1344,7 @@ static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
>  	DECLARE_WAITQUEUE(wait, current);
>  	struct sk_buff *skb;
>  	ssize_t ret = 0;
> +	struct netdev_queue *ntxq;
>  
>  	tun_debug(KERN_INFO, tun, "tun_do_read\n");
>  
> @@ -1362,6 +1374,12 @@ static ssize_t tun_do_read(struct tun_struct *tun, struct tun_file *tfile,
>  			continue;
>  		}
>  
> +		ntxq = netdev_get_tx_queue(tun->dev, tfile->queue_index);
> +
> +		if (tfile->flags & TUN_FLOW_CONTROL &&
> +		    netif_tx_queue_stopped(ntxq))
> +			netif_tx_wake_queue(ntxq);
> +
>  		ret = tun_put_user(tun, tfile, skb, iv, len);
>  		kfree_skb(skb);
>  		break;
> @@ -1732,6 +1750,11 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr)
>  	else
>  		tun->flags &= ~TUN_TAP_MQ;
>  
> +	if (ifr->ifr_flags & IFF_FLOW_CONTROL)
> +		tfile->flags |= TUN_FLOW_CONTROL;
> +	else
> +		tfile->flags &= ~TUN_FLOW_CONTROL;
> +
>  	/* Make sure persistent devices do not get stuck in
>  	 * xoff state.
>  	 */
> @@ -1900,7 +1923,8 @@ static long __tun_chr_ioctl(struct file *file, unsigned int cmd,
>  		 * This is needed because we never checked for invalid flags on
>  		 * TUNSETIFF. */
>  		return put_user(IFF_TUN | IFF_TAP | IFF_NO_PI | IFF_ONE_QUEUE |
> -				IFF_VNET_HDR | IFF_MULTI_QUEUE,
> +				IFF_VNET_HDR | IFF_MULTI_QUEUE |
> +				IFF_FLOW_CONTROL,
>  				(unsigned int __user*)argp);
>  	} else if (cmd == TUNSETQUEUE)
>  		return tun_set_queue(file, &ifr);
> diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> index e9502dd..bcf2790 100644
> --- a/include/uapi/linux/if_tun.h
> +++ b/include/uapi/linux/if_tun.h
> @@ -36,6 +36,7 @@
>  #define TUN_PERSIST 	0x0100	
>  #define TUN_VNET_HDR 	0x0200
>  #define TUN_TAP_MQ      0x0400
> +#define TUN_FLOW_CONTROL 0x0800
>  
>  /* Ioctl defines */
>  #define TUNSETNOCSUM  _IOW('T', 200, int) 
> @@ -70,6 +71,7 @@
>  #define IFF_MULTI_QUEUE 0x0100
>  #define IFF_ATTACH_QUEUE 0x0200
>  #define IFF_DETACH_QUEUE 0x0400
> +#define IFF_FLOW_CONTROL 0x0010
>  /* read-only flag */
>  #define IFF_PERSIST	0x0800
>  #define IFF_NOFILTER	0x1000
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/