From 2f1e85b1aee459b7d0fd981839042c6a38ffaf0c Mon Sep 17 00:00:00 2001 From: Tonghao Zhang Date: Sat, 16 Apr 2022 00:40:45 +0800 Subject: [PATCH] net: sched: use queue_mapping to pick tx queue This patch fixes issue: * If we install tc filters with act_skbedit in clsact hook. It doesn't work, because netdev_core_pick_tx() overwrites queue_mapping. $ tc filter ... action skbedit queue_mapping 1 And this patch is useful: * We can use FQ + EDT to implement efficient policies. Tx queues are picked by xps, ndo_select_queue of netdev driver, or skb hash in netdev_core_pick_tx(). In fact, the netdev driver, and skb hash are _not_ under control. xps uses the CPUs map to select Tx queues, but we can't figure out which task_struct of pod/containter running on this cpu in most case. We can use clsact filters to classify one pod/container traffic to one Tx queue. Why ? In containter networking environment, there are two kinds of pod/ containter/net-namespace. One kind (e.g. P1, P2), the high throughput is key in these applications. But avoid running out of network resource, the outbound traffic of these pods is limited, using or sharing one dedicated Tx queues assigned HTB/TBF/FQ Qdisc. Other kind of pods (e.g. Pn), the low latency of data access is key. And the traffic is not limited. Pods use or share other dedicated Tx queues assigned FIFO Qdisc. This choice provides two benefits. First, contention on the HTB/FQ Qdisc lock is significantly reduced since fewer CPUs contend for the same queue. More importantly, Qdisc contention can be eliminated completely if each CPU has its own FIFO Qdisc for the second kind of pods. There must be a mechanism in place to support classifying traffic based on pods/container to different Tx queues. Note that clsact is outside of Qdisc while Qdisc can run a classifier to select a sub-queue under the lock. In general recording the decision in the skb seems a little heavy handed. This patch introduces a per-CPU variable, suggested by Eric. The xmit.skip_txqueue flag is firstly cleared in __dev_queue_xmit(). - Tx Qdisc may install that skbedit actions, then xmit.skip_txqueue flag is set in qdisc->enqueue() though tx queue has been selected in netdev_tx_queue_mapping() or netdev_core_pick_tx(). That flag is cleared firstly in __dev_queue_xmit(), is useful: - Avoid picking Tx queue with netdev_tx_queue_mapping() in next netdev in such case: eth0 macvlan - eth0.3 vlan - eth0 ixgbe-phy: For example, eth0, macvlan in pod, which root Qdisc install skbedit queue_mapping, send packets to eth0.3, vlan in host. In __dev_queue_xmit() of eth0.3, clear the flag, does not select tx queue according to skb->queue_mapping because there is no filters in clsact or tx Qdisc of this netdev. Same action taked in eth0, ixgbe in Host. - Avoid picking Tx queue for next packet. If we set xmit.skip_txqueue in tx Qdisc (qdisc->enqueue()), the proper way to clear it is clearing it in __dev_queue_xmit when processing next packets. For performance reasons, use the static key. If user does not config the NET_EGRESS, the patch will not be compiled. +----+ +----+ +----+ | P1 | | P2 | | Pn | +----+ +----+ +----+ | | | +-----------+-----------+ | | clsact/skbedit | MQ v +-----------+-----------+ | q0 | q1 | qn v v v HTB/FQ HTB/FQ ... FIFO Cc: Jamal Hadi Salim Cc: Cong Wang Cc: Jiri Pirko Cc: "David S. Miller" Cc: Jakub Kicinski Cc: Jonathan Lemon Cc: Eric Dumazet Cc: Alexander Lobakin Cc: Paolo Abeni Cc: Talal Ahmad Cc: Kevin Hao Cc: Ilias Apalodimas Cc: Kees Cook Cc: Kumar Kartikeya Dwivedi Cc: Antoine Tenart Cc: Wei Wang Cc: Arnd Bergmann Suggested-by: Eric Dumazet Signed-off-by: Tonghao Zhang Acked-by: Jamal Hadi Salim Signed-off-by: Paolo Abeni --- include/linux/netdevice.h | 3 +++ include/linux/rtnetlink.h | 1 + net/core/dev.c | 31 +++++++++++++++++++++++++++++-- net/sched/act_skbedit.c | 6 +++++- 4 files changed, 38 insertions(+), 3 deletions(-) diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index a602f29365b0..7dccbfd1bf56 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -3061,6 +3061,9 @@ struct softnet_data { struct { u16 recursion; u8 more; +#ifdef CONFIG_NET_EGRESS + u8 skip_txqueue; +#endif } xmit; #ifdef CONFIG_RPS /* input_queue_head should be written by cpu owning this struct, diff --git a/include/linux/rtnetlink.h b/include/linux/rtnetlink.h index 7f970b16da3a..ae2c6a3cec5d 100644 --- a/include/linux/rtnetlink.h +++ b/include/linux/rtnetlink.h @@ -100,6 +100,7 @@ void net_dec_ingress_queue(void); #ifdef CONFIG_NET_EGRESS void net_inc_egress_queue(void); void net_dec_egress_queue(void); +void netdev_xmit_skip_txqueue(bool skip); #endif void rtnetlink_init(void); diff --git a/net/core/dev.c b/net/core/dev.c index ba853e878007..4a77ebda4fb1 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -3925,6 +3925,25 @@ sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev) return skb; } + +static struct netdev_queue * +netdev_tx_queue_mapping(struct net_device *dev, struct sk_buff *skb) +{ + int qm = skb_get_queue_mapping(skb); + + return netdev_get_tx_queue(dev, netdev_cap_txqueue(dev, qm)); +} + +static bool netdev_xmit_txqueue_skipped(void) +{ + return __this_cpu_read(softnet_data.xmit.skip_txqueue); +} + +void netdev_xmit_skip_txqueue(bool skip) +{ + __this_cpu_write(softnet_data.xmit.skip_txqueue, skip); +} +EXPORT_SYMBOL_GPL(netdev_xmit_skip_txqueue); #endif /* CONFIG_NET_EGRESS */ #ifdef CONFIG_XPS @@ -4095,7 +4114,7 @@ struct netdev_queue *netdev_core_pick_tx(struct net_device *dev, static int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev) { struct net_device *dev = skb->dev; - struct netdev_queue *txq; + struct netdev_queue *txq = NULL; struct Qdisc *q; int rc = -ENOMEM; bool again = false; @@ -4123,11 +4142,17 @@ static int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev) if (!skb) goto out; } + + netdev_xmit_skip_txqueue(false); + nf_skip_egress(skb, true); skb = sch_handle_egress(skb, &rc, dev); if (!skb) goto out; nf_skip_egress(skb, false); + + if (netdev_xmit_txqueue_skipped()) + txq = netdev_tx_queue_mapping(dev, skb); } #endif /* If device/qdisc don't need skb->dst, release it right now while @@ -4138,7 +4163,9 @@ static int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev) else skb_dst_force(skb); - txq = netdev_core_pick_tx(dev, skb, sb_dev); + if (!txq) + txq = netdev_core_pick_tx(dev, skb, sb_dev); + q = rcu_dereference_bh(txq->qdisc); trace_net_dev_queue(skb); diff --git a/net/sched/act_skbedit.c b/net/sched/act_skbedit.c index 92d0dc754207..1c5fdb6e7c2f 100644 --- a/net/sched/act_skbedit.c +++ b/net/sched/act_skbedit.c @@ -58,8 +58,12 @@ static int tcf_skbedit_act(struct sk_buff *skb, const struct tc_action *a, } } if (params->flags & SKBEDIT_F_QUEUE_MAPPING && - skb->dev->real_num_tx_queues > params->queue_mapping) + skb->dev->real_num_tx_queues > params->queue_mapping) { +#ifdef CONFIG_NET_EGRESS + netdev_xmit_skip_txqueue(true); +#endif skb_set_queue_mapping(skb, params->queue_mapping); + } if (params->flags & SKBEDIT_F_MARK) { skb->mark &= ~params->mask; skb->mark |= params->mark & params->mask;