From 307f660d056b5eb8f5bb2328fac3915ab75b5007 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 7 May 2020 09:32:18 -0700 Subject: [PATCH 1/5] netpoll: remove dev argument from netpoll_send_skb_on_dev() netpoll_send_skb_on_dev() can get the device pointer directly from np->dev Rename it to __netpoll_send_skb() Following patch will move netpoll_send_skb() out-of-line. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/netpoll.h | 5 ++--- net/core/netpoll.c | 10 ++++++---- 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h index 676f1ff161a9..00e0bae3d402 100644 --- a/include/linux/netpoll.h +++ b/include/linux/netpoll.h @@ -63,13 +63,12 @@ int netpoll_setup(struct netpoll *np); void __netpoll_cleanup(struct netpoll *np); void __netpoll_free(struct netpoll *np); void netpoll_cleanup(struct netpoll *np); -void netpoll_send_skb_on_dev(struct netpoll *np, struct sk_buff *skb, - struct net_device *dev); +void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb); static inline void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) { unsigned long flags; local_irq_save(flags); - netpoll_send_skb_on_dev(np, skb, np->dev); + __netpoll_send_skb(np, skb); local_irq_restore(flags); } diff --git a/net/core/netpoll.c b/net/core/netpoll.c index 15b366a1a958..c5059b7ffc94 100644 --- a/net/core/netpoll.c +++ b/net/core/netpoll.c @@ -305,17 +305,19 @@ static int netpoll_owner_active(struct net_device *dev) } /* call with IRQ disabled */ -void netpoll_send_skb_on_dev(struct netpoll *np, struct sk_buff *skb, - struct net_device *dev) +void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) { netdev_tx_t status = NETDEV_TX_BUSY; + struct net_device *dev; unsigned long tries; /* It is up to the caller to keep npinfo alive. */ struct netpoll_info *npinfo; lockdep_assert_irqs_disabled(); - npinfo = rcu_dereference_bh(np->dev->npinfo); + dev = np->dev; + npinfo = rcu_dereference_bh(dev->npinfo); + if (!npinfo || !netif_running(dev) || !netif_device_present(dev)) { dev_kfree_skb_irq(skb); return; @@ -358,7 +360,7 @@ void netpoll_send_skb_on_dev(struct netpoll *np, struct sk_buff *skb, schedule_delayed_work(&npinfo->tx_work,0); } } -EXPORT_SYMBOL(netpoll_send_skb_on_dev); +EXPORT_SYMBOL(__netpoll_send_skb); void netpoll_send_udp(struct netpoll *np, const char *msg, int len) { From fb1eee476b0d3be3e58dac1a3a96f726c6278bed Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 7 May 2020 09:32:19 -0700 Subject: [PATCH 2/5] netpoll: move netpoll_send_skb() out of line There is no need to inline this helper, as we intend to add more code in this function. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/netpoll.h | 9 +-------- net/core/netpoll.c | 13 +++++++++++-- 2 files changed, 12 insertions(+), 10 deletions(-) diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h index 00e0bae3d402..e466ddffef61 100644 --- a/include/linux/netpoll.h +++ b/include/linux/netpoll.h @@ -63,14 +63,7 @@ int netpoll_setup(struct netpoll *np); void __netpoll_cleanup(struct netpoll *np); void __netpoll_free(struct netpoll *np); void netpoll_cleanup(struct netpoll *np); -void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb); -static inline void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) -{ - unsigned long flags; - local_irq_save(flags); - __netpoll_send_skb(np, skb); - local_irq_restore(flags); -} +void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb); #ifdef CONFIG_NETPOLL static inline void *netpoll_poll_lock(struct napi_struct *napi) diff --git a/net/core/netpoll.c b/net/core/netpoll.c index c5059b7ffc94..34cd34f24423 100644 --- a/net/core/netpoll.c +++ b/net/core/netpoll.c @@ -305,7 +305,7 @@ static int netpoll_owner_active(struct net_device *dev) } /* call with IRQ disabled */ -void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) +static void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) { netdev_tx_t status = NETDEV_TX_BUSY; struct net_device *dev; @@ -360,7 +360,16 @@ void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) schedule_delayed_work(&npinfo->tx_work,0); } } -EXPORT_SYMBOL(__netpoll_send_skb); + +void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) +{ + unsigned long flags; + + local_irq_save(flags); + __netpoll_send_skb(np, skb); + local_irq_restore(flags); +} +EXPORT_SYMBOL(netpoll_send_skb); void netpoll_send_udp(struct netpoll *np, const char *msg, int len) { From 1ddabdfaf70c202b88925edd74c66f4707dbd92e Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 7 May 2020 09:32:20 -0700 Subject: [PATCH 3/5] netpoll: netpoll_send_skb() returns transmit status Some callers want to know if the packet has been sent or dropped, to inform upper stacks. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/linux/netpoll.h | 2 +- net/core/netpoll.c | 11 +++++++---- 2 files changed, 8 insertions(+), 5 deletions(-) diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h index e466ddffef61..f47af135bd56 100644 --- a/include/linux/netpoll.h +++ b/include/linux/netpoll.h @@ -63,7 +63,7 @@ int netpoll_setup(struct netpoll *np); void __netpoll_cleanup(struct netpoll *np); void __netpoll_free(struct netpoll *np); void netpoll_cleanup(struct netpoll *np); -void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb); +netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb); #ifdef CONFIG_NETPOLL static inline void *netpoll_poll_lock(struct napi_struct *napi) diff --git a/net/core/netpoll.c b/net/core/netpoll.c index 34cd34f24423..40d2753aa47d 100644 --- a/net/core/netpoll.c +++ b/net/core/netpoll.c @@ -305,7 +305,7 @@ static int netpoll_owner_active(struct net_device *dev) } /* call with IRQ disabled */ -static void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) +static netdev_tx_t __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) { netdev_tx_t status = NETDEV_TX_BUSY; struct net_device *dev; @@ -320,7 +320,7 @@ static void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) if (!npinfo || !netif_running(dev) || !netif_device_present(dev)) { dev_kfree_skb_irq(skb); - return; + return NET_XMIT_DROP; } /* don't get messages out of order, and no recursion */ @@ -359,15 +359,18 @@ static void __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) skb_queue_tail(&npinfo->txq, skb); schedule_delayed_work(&npinfo->tx_work,0); } + return NETDEV_TX_OK; } -void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) +netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) { unsigned long flags; + netdev_tx_t ret; local_irq_save(flags); - __netpoll_send_skb(np, skb); + ret = __netpoll_send_skb(np, skb); local_irq_restore(flags); + return ret; } EXPORT_SYMBOL(netpoll_send_skb); From f78ed2204db9fc35b545d693865bddbe0149aa1f Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 7 May 2020 09:32:21 -0700 Subject: [PATCH 4/5] netpoll: accept NULL np argument in netpoll_send_skb() netpoll_send_skb() callers seem to leak skb if the np pointer is NULL. While this should not happen, we can make the code more robust. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- drivers/net/macvlan.c | 5 ++--- include/linux/if_team.h | 5 +---- include/net/bonding.h | 5 +---- net/8021q/vlan_dev.c | 5 ++--- net/bridge/br_private.h | 5 +---- net/core/netpoll.c | 11 ++++++++--- net/dsa/slave.c | 5 ++--- 7 files changed, 17 insertions(+), 24 deletions(-) diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c index 34eb073cdd74..9a419d5102ce 100644 --- a/drivers/net/macvlan.c +++ b/drivers/net/macvlan.c @@ -542,12 +542,11 @@ xmit_world: static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb) { #ifdef CONFIG_NET_POLL_CONTROLLER - if (vlan->netpoll) - netpoll_send_skb(vlan->netpoll, skb); + return netpoll_send_skb(vlan->netpoll, skb); #else BUG(); -#endif return NETDEV_TX_OK; +#endif } static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb, diff --git a/include/linux/if_team.h b/include/linux/if_team.h index ec7e4bd07f82..537dc2b8c879 100644 --- a/include/linux/if_team.h +++ b/include/linux/if_team.h @@ -102,10 +102,7 @@ static inline bool team_port_dev_txable(const struct net_device *port_dev) static inline void team_netpoll_send_skb(struct team_port *port, struct sk_buff *skb) { - struct netpoll *np = port->np; - - if (np) - netpoll_send_skb(np, skb); + netpoll_send_skb(port->np, skb); } #else static inline void team_netpoll_send_skb(struct team_port *port, diff --git a/include/net/bonding.h b/include/net/bonding.h index 0b696da5c115..f211983cd52a 100644 --- a/include/net/bonding.h +++ b/include/net/bonding.h @@ -507,10 +507,7 @@ static inline unsigned long slave_last_rx(struct bonding *bond, static inline void bond_netpoll_send_skb(const struct slave *slave, struct sk_buff *skb) { - struct netpoll *np = slave->np; - - if (np) - netpoll_send_skb(np, skb); + netpoll_send_skb(slave->np, skb); } #else static inline void bond_netpoll_send_skb(const struct slave *slave, diff --git a/net/8021q/vlan_dev.c b/net/8021q/vlan_dev.c index 319220b2341d..f00bb57f0f60 100644 --- a/net/8021q/vlan_dev.c +++ b/net/8021q/vlan_dev.c @@ -88,12 +88,11 @@ static int vlan_dev_hard_header(struct sk_buff *skb, struct net_device *dev, static inline netdev_tx_t vlan_netpoll_send_skb(struct vlan_dev_priv *vlan, struct sk_buff *skb) { #ifdef CONFIG_NET_POLL_CONTROLLER - if (vlan->netpoll) - netpoll_send_skb(vlan->netpoll, skb); + return netpoll_send_skb(vlan->netpoll, skb); #else BUG(); -#endif return NETDEV_TX_OK; +#endif } static netdev_tx_t vlan_dev_hard_start_xmit(struct sk_buff *skb, diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h index 78d3a951180d..4dc21e8f7e33 100644 --- a/net/bridge/br_private.h +++ b/net/bridge/br_private.h @@ -598,10 +598,7 @@ netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev); static inline void br_netpoll_send_skb(const struct net_bridge_port *p, struct sk_buff *skb) { - struct netpoll *np = p->np; - - if (np) - netpoll_send_skb(np, skb); + netpoll_send_skb(p->np, skb); } int br_netpoll_enable(struct net_bridge_port *p); diff --git a/net/core/netpoll.c b/net/core/netpoll.c index 40d2753aa47d..093e90e52bc2 100644 --- a/net/core/netpoll.c +++ b/net/core/netpoll.c @@ -367,9 +367,14 @@ netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) unsigned long flags; netdev_tx_t ret; - local_irq_save(flags); - ret = __netpoll_send_skb(np, skb); - local_irq_restore(flags); + if (unlikely(!np)) { + dev_kfree_skb_irq(skb); + ret = NET_XMIT_DROP; + } else { + local_irq_save(flags); + ret = __netpoll_send_skb(np, skb); + local_irq_restore(flags); + } return ret; } EXPORT_SYMBOL(netpoll_send_skb); diff --git a/net/dsa/slave.c b/net/dsa/slave.c index dfb4282fc339..61b0de52040a 100644 --- a/net/dsa/slave.c +++ b/net/dsa/slave.c @@ -445,12 +445,11 @@ static inline netdev_tx_t dsa_slave_netpoll_send_skb(struct net_device *dev, #ifdef CONFIG_NET_POLL_CONTROLLER struct dsa_slave_priv *p = netdev_priv(dev); - if (p->netpoll) - netpoll_send_skb(p->netpoll, skb); + return netpoll_send_skb(p->netpoll, skb); #else BUG(); -#endif return NETDEV_TX_OK; +#endif } static void dsa_skb_tx_timestamp(struct dsa_slave_priv *p, From ae46f184bc1fb15bf2de47114c29236e61ca4bbc Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Thu, 7 May 2020 09:32:22 -0700 Subject: [PATCH 5/5] bonding: propagate transmit status Currently, bonding always returns NETDEV_TX_OK to its caller. It is worth trying to be more accurate : TCP for instance can have different recovery strategies if it can have more precise status, if packet was dropped by slave qdisc. This is especially important when host is under stress. Signed-off-by: Eric Dumazet Cc: Jay Vosburgh Cc: Veaceslav Falico Cc: Andy Gospodarek Signed-off-by: David S. Miller --- drivers/net/bonding/bond_alb.c | 7 ++-- drivers/net/bonding/bond_main.c | 60 ++++++++++++--------------------- include/net/bonding.h | 13 ++++--- 3 files changed, 32 insertions(+), 48 deletions(-) diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c index c81698550e5a..3a598d04b156 100644 --- a/drivers/net/bonding/bond_alb.c +++ b/drivers/net/bonding/bond_alb.c @@ -1318,8 +1318,7 @@ static netdev_tx_t bond_do_alb_xmit(struct sk_buff *skb, struct bonding *bond, tx_slave->dev->dev_addr); } - bond_dev_queue_xmit(bond, skb, tx_slave->dev); - goto out; + return bond_dev_queue_xmit(bond, skb, tx_slave->dev); } if (tx_slave && bond->params.tlb_dynamic_lb) { @@ -1329,9 +1328,7 @@ static netdev_tx_t bond_do_alb_xmit(struct sk_buff *skb, struct bonding *bond, } /* no suitable interface, frame not sent */ - bond_tx_drop(bond->dev, skb); -out: - return NETDEV_TX_OK; + return bond_tx_drop(bond->dev, skb); } netdev_tx_t bond_tlb_xmit(struct sk_buff *skb, struct net_device *bond_dev) diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index baa93191dfdd..4f9e7c421f57 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -287,7 +287,7 @@ const char *bond_mode_name(int mode) * @skb: hw accel VLAN tagged skb to transmit * @slave_dev: slave that is supposed to xmit this skbuff */ -void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, +netdev_tx_t bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev) { skb->dev = slave_dev; @@ -297,9 +297,9 @@ void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, skb_set_queue_mapping(skb, qdisc_skb_cb(skb)->slave_dev_queue_mapping); if (unlikely(netpoll_tx_running(bond->dev))) - bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); - else - dev_queue_xmit(skb); + return bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); + + return dev_queue_xmit(skb); } /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, @@ -3932,7 +3932,7 @@ unwind: * it fails, it tries to find the first available slave for transmission. * The skb is consumed in all cases, thus the function is void. */ -static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) +static netdev_tx_t bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) { struct list_head *iter; struct slave *slave; @@ -3941,10 +3941,8 @@ static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int sl /* Here we start from the slave with slave_id */ bond_for_each_slave_rcu(bond, slave, iter) { if (--i < 0) { - if (bond_slave_can_tx(slave)) { - bond_dev_queue_xmit(bond, skb, slave->dev); - return; - } + if (bond_slave_can_tx(slave)) + return bond_dev_queue_xmit(bond, skb, slave->dev); } } @@ -3953,13 +3951,11 @@ static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int sl bond_for_each_slave_rcu(bond, slave, iter) { if (--i < 0) break; - if (bond_slave_can_tx(slave)) { - bond_dev_queue_xmit(bond, skb, slave->dev); - return; - } + if (bond_slave_can_tx(slave)) + return bond_dev_queue_xmit(bond, skb, slave->dev); } /* no slave that can tx has been found */ - bond_tx_drop(bond->dev, skb); + return bond_tx_drop(bond->dev, skb); } /** @@ -4020,10 +4016,8 @@ static netdev_tx_t bond_xmit_roundrobin(struct sk_buff *skb, if (iph->protocol == IPPROTO_IGMP) { slave = rcu_dereference(bond->curr_active_slave); if (slave) - bond_dev_queue_xmit(bond, skb, slave->dev); - else - bond_xmit_slave_id(bond, skb, 0); - return NETDEV_TX_OK; + return bond_dev_queue_xmit(bond, skb, slave->dev); + return bond_xmit_slave_id(bond, skb, 0); } } @@ -4031,11 +4025,9 @@ non_igmp: slave_cnt = READ_ONCE(bond->slave_cnt); if (likely(slave_cnt)) { slave_id = bond_rr_gen_slave_id(bond); - bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); - } else { - bond_tx_drop(bond_dev, skb); + return bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); } - return NETDEV_TX_OK; + return bond_tx_drop(bond_dev, skb); } /* In active-backup mode, we know that bond->curr_active_slave is always valid if @@ -4049,11 +4041,9 @@ static netdev_tx_t bond_xmit_activebackup(struct sk_buff *skb, slave = rcu_dereference(bond->curr_active_slave); if (slave) - bond_dev_queue_xmit(bond, skb, slave->dev); - else - bond_tx_drop(bond_dev, skb); + return bond_dev_queue_xmit(bond, skb, slave->dev); - return NETDEV_TX_OK; + return bond_tx_drop(bond_dev, skb); } /* Use this to update slave_array when (a) it's not appropriate to update @@ -4196,12 +4186,9 @@ static netdev_tx_t bond_3ad_xor_xmit(struct sk_buff *skb, count = slaves ? READ_ONCE(slaves->count) : 0; if (likely(count)) { slave = slaves->arr[bond_xmit_hash(bond, skb) % count]; - bond_dev_queue_xmit(bond, skb, slave->dev); - } else { - bond_tx_drop(dev, skb); + return bond_dev_queue_xmit(bond, skb, slave->dev); } - - return NETDEV_TX_OK; + return bond_tx_drop(dev, skb); } /* in broadcast mode, we send everything to all usable interfaces. */ @@ -4227,11 +4214,9 @@ static netdev_tx_t bond_xmit_broadcast(struct sk_buff *skb, } } if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) - bond_dev_queue_xmit(bond, skb, slave->dev); - else - bond_tx_drop(bond_dev, skb); + return bond_dev_queue_xmit(bond, skb, slave->dev); - return NETDEV_TX_OK; + return bond_tx_drop(bond_dev, skb); } /*------------------------- Device initialization ---------------------------*/ @@ -4310,8 +4295,7 @@ static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev /* Should never happen, mode already checked */ netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); WARN_ON_ONCE(1); - bond_tx_drop(dev, skb); - return NETDEV_TX_OK; + return bond_tx_drop(dev, skb); } } @@ -4330,7 +4314,7 @@ static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) if (bond_has_slaves(bond)) ret = __bond_start_xmit(skb, dev); else - bond_tx_drop(dev, skb); + ret = bond_tx_drop(dev, skb); rcu_read_unlock(); return ret; diff --git a/include/net/bonding.h b/include/net/bonding.h index f211983cd52a..9b1e76515a9c 100644 --- a/include/net/bonding.h +++ b/include/net/bonding.h @@ -504,15 +504,17 @@ static inline unsigned long slave_last_rx(struct bonding *bond, } #ifdef CONFIG_NET_POLL_CONTROLLER -static inline void bond_netpoll_send_skb(const struct slave *slave, +static inline netdev_tx_t bond_netpoll_send_skb(const struct slave *slave, struct sk_buff *skb) { - netpoll_send_skb(slave->np, skb); + return netpoll_send_skb(slave->np, skb); } #else -static inline void bond_netpoll_send_skb(const struct slave *slave, +static inline netdev_tx_t bond_netpoll_send_skb(const struct slave *slave, struct sk_buff *skb) { + BUG(); + return NETDEV_TX_OK; } #endif @@ -606,7 +608,7 @@ struct bond_net { }; int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave); -void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev); +netdev_tx_t bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev); int bond_create(struct net *net, const char *name); int bond_create_sysfs(struct bond_net *net); void bond_destroy_sysfs(struct bond_net *net); @@ -739,10 +741,11 @@ extern struct bond_parm_tbl ad_select_tbl[]; /* exported from bond_netlink.c */ extern struct rtnl_link_ops bond_link_ops; -static inline void bond_tx_drop(struct net_device *dev, struct sk_buff *skb) +static inline netdev_tx_t bond_tx_drop(struct net_device *dev, struct sk_buff *skb) { atomic_long_inc(&dev->tx_dropped); dev_kfree_skb_any(skb); + return NET_XMIT_DROP; } #endif /* _NET_BONDING_H */