From: Jussi Maki joamaki@gmail.com
mainline inclusion from mainline-v5.15-rc1 commit 9e2ee5c7e7c35d195e2aa0692a7241d47a433d1e category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I7NDRB CVE: NA
Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...
----------------------------------------
XDP is implemented in the bonding driver by transparently delegating the XDP program loading, removal and xmit operations to the bonding slave devices. The overall goal of this work is that XDP programs can be attached to a bond device *without* any further changes (or awareness) necessary to the program itself, meaning the same XDP program can be attached to a native device but also a bonding device.
Semantics of XDP_TX when attached to a bond are equivalent in such setting to the case when a tc/BPF program would be attached to the bond, meaning transmitting the packet out of the bond itself using one of the bond's configured xmit methods to select a slave device (rather than XDP_TX on the slave itself). Handling of XDP_TX to transmit using the configured bonding mechanism is therefore implemented by rewriting the BPF program return value in bpf_prog_run_xdp. To avoid performance impact this check is guarded by a static key, which is incremented when a XDP program is loaded onto a bond device. This approach was chosen to avoid changes to drivers implementing XDP. If the slave device does not match the receive device, then XDP_REDIRECT is transparently used to perform the redirection in order to have the network driver release the packet from its RX ring. The bonding driver hashing functions have been refactored to allow reuse with xdp_buff's to avoid code duplication.
The motivation for this change is to enable use of bonding (and 802.3ad) in hairpinning L4 load-balancers such as [1] implemented with XDP and also to transparently support bond devices for projects that use XDP given most modern NICs have dual port adapters. An alternative to this approach would be to implement 802.3ad in user-space and implement the bonding load-balancing in the XDP program itself, but is rather a cumbersome endeavor in terms of slave device management (e.g. by watching netlink) and requires separate programs for native vs bond cases for the orchestrator. A native in-kernel implementation overcomes these issues and provides more flexibility.
Below are benchmark results done on two machines with 100Gbit Intel E810 (ice) NIC and with 32-core 3970X on sending machine, and 16-core 3950X on receiving machine. 64 byte packets were sent with pktgen-dpdk at full rate. Two issues [2, 3] were identified with the ice driver, so the tests were performed with iommu=off and patch [2] applied. Additionally the bonding round robin algorithm was modified to use per-cpu tx counters as high CPU load (50% vs 10%) and high rate of cache misses were caused by the shared rr_tx_counter (see patch 2/3). The statistics were collected using "sar -n dev -u 1 10". On top of that, for ice, further work is in progress on improving the XDP_TX numbers [4].
-----------------------| CPU |--| rxpck/s |--| txpck/s |---- without patch (1 dev): XDP_DROP: 3.15% 48.6Mpps XDP_TX: 3.12% 18.3Mpps 18.3Mpps XDP_DROP (RSS): 9.47% 116.5Mpps XDP_TX (RSS): 9.67% 25.3Mpps 24.2Mpps ----------------------- with patch, bond (1 dev): XDP_DROP: 3.14% 46.7Mpps XDP_TX: 3.15% 13.9Mpps 13.9Mpps XDP_DROP (RSS): 10.33% 117.2Mpps XDP_TX (RSS): 10.64% 25.1Mpps 24.0Mpps ----------------------- with patch, bond (2 devs): XDP_DROP: 6.27% 92.7Mpps XDP_TX: 6.26% 17.6Mpps 17.5Mpps XDP_DROP (RSS): 11.38% 117.2Mpps XDP_TX (RSS): 14.30% 28.7Mpps 27.4Mpps --------------------------------------------------------------
RSS: Receive Side Scaling, e.g. the packets were sent to a range of destination IPs.
[1]: https://cilium.io/blog/2021/05/20/cilium-110#standalonelb [2]: https://lore.kernel.org/bpf/20210601113236.42651-1-maciej.fijalkowski@intel.... [3]: https://lore.kernel.org/bpf/CAHn8xckNXci+X_Eb2WMv4uVYjO2331UWB2JLtXr_58z0Av8... [4]: https://lore.kernel.org/bpf/20210805230046.28715-1-maciej.fijalkowski@intel....
Signed-off-by: Jussi Maki joamaki@gmail.com Signed-off-by: Daniel Borkmann daniel@iogearbox.net Cc: Jay Vosburgh j.vosburgh@gmail.com Cc: Veaceslav Falico vfalico@gmail.com Cc: Andy Gospodarek andy@greyhouse.net Cc: Maciej Fijalkowski maciej.fijalkowski@intel.com Cc: Magnus Karlsson magnus.karlsson@intel.com Link: https://lore.kernel.org/bpf/20210731055738.16820-4-joamaki@gmail.com
Conflicts: drivers/net/bonding/bond_main.c include/net/bonding.h
Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- drivers/net/bonding/bond_main.c | 309 +++++++++++++++++++++++++++++++- include/net/bonding.h | 1 + 2 files changed, 309 insertions(+), 1 deletion(-)
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c index ec5e55c53dde..38c33d2a0a0a 100644 --- a/drivers/net/bonding/bond_main.c +++ b/drivers/net/bonding/bond_main.c @@ -302,6 +302,19 @@ netdev_tx_t bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, return dev_queue_xmit(skb); }
+static bool bond_xdp_check(struct bonding *bond) +{ + switch (BOND_MODE(bond)) { + case BOND_MODE_ROUNDROBIN: + case BOND_MODE_ACTIVEBACKUP: + case BOND_MODE_8023AD: + case BOND_MODE_XOR: + return true; + default: + return false; + } +} + /*---------------------------------- VLAN -----------------------------------*/
/* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, @@ -2118,6 +2131,41 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev, bond_update_slave_arr(bond, NULL);
+ if (!slave_dev->netdev_ops->ndo_bpf || + !slave_dev->netdev_ops->ndo_xdp_xmit) { + if (bond->xdp_prog) { + NL_SET_ERR_MSG(extack, "Slave does not support XDP"); + slave_err(bond_dev, slave_dev, "Slave does not support XDP\n"); + res = -EOPNOTSUPP; + goto err_sysfs_del; + } + } else { + struct netdev_bpf xdp = { + .command = XDP_SETUP_PROG, + .flags = 0, + .prog = bond->xdp_prog, + .extack = extack, + }; + + if (dev_xdp_prog_count(slave_dev) > 0) { + NL_SET_ERR_MSG(extack, + "Slave has XDP program loaded, please unload before enslaving"); + slave_err(bond_dev, slave_dev, + "Slave has XDP program loaded, please unload before enslaving\n"); + res = -EOPNOTSUPP; + goto err_sysfs_del; + } + + res = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); + if (res < 0) { + /* ndo_bpf() sets extack error message */ + slave_dbg(bond_dev, slave_dev, "Error %d calling ndo_bpf\n", res); + goto err_sysfs_del; + } + if (bond->xdp_prog) + bpf_prog_inc(bond->xdp_prog); + } + slave_info(bond_dev, slave_dev, "Enslaving as %s interface with %s link\n", bond_is_active_slave(new_slave) ? "an active" : "a backup", new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); @@ -2235,6 +2283,17 @@ static int __bond_release_one(struct net_device *bond_dev, /* recompute stats just before removing the slave */ bond_get_stats(bond->dev, &bond->bond_stats);
+ if (bond->xdp_prog) { + struct netdev_bpf xdp = { + .command = XDP_SETUP_PROG, + .flags = 0, + .prog = NULL, + .extack = NULL, + }; + if (slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp)) + slave_warn(bond_dev, slave_dev, "failed to unload XDP program\n"); + } + /* unregister rx_handler early so bond_handle_frame wouldn't be called * for this slave anymore. */ @@ -3625,7 +3684,7 @@ static inline u32 bond_eth_hash(struct sk_buff *skb, void *data, int mhoff, int return 0;
ep = (struct ethhdr *)(data + mhoff); - return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; + return ep->h_dest[5] ^ ep->h_source[5] ^ be16_to_cpu(ep->h_proto); }
static bool bond_flow_ip(struct sk_buff *skb, struct flow_keys *fk, void *data, @@ -3754,6 +3813,26 @@ u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) skb_headlen(skb)); }
+/** + * bond_xmit_hash_xdp - generate a hash value based on the xmit policy + * @bond: bonding device + * @xdp: buffer to use for headers + * + * The XDP variant of bond_xmit_hash. + */ +static u32 bond_xmit_hash_xdp(struct bonding *bond, struct xdp_buff *xdp) +{ + struct ethhdr *eth; + + if (xdp->data + sizeof(struct ethhdr) > xdp->data_end) + return 0; + + eth = (struct ethhdr *)xdp->data; + + return __bond_xmit_hash(bond, NULL, xdp->data, eth->h_proto, 0, + sizeof(struct ethhdr), xdp->data_end - xdp->data); +} + /*-------------------------- Device entry points ----------------------------*/
void bond_work_init_all(struct bonding *bond) @@ -4387,6 +4466,47 @@ static struct slave *bond_xmit_roundrobin_slave_get(struct bonding *bond, return NULL; }
+static struct slave *bond_xdp_xmit_roundrobin_slave_get(struct bonding *bond, + struct xdp_buff *xdp) +{ + struct slave *slave; + int slave_cnt; + u32 slave_id; + const struct ethhdr *eth; + void *data = xdp->data; + + if (data + sizeof(struct ethhdr) > xdp->data_end) + goto non_igmp; + + eth = (struct ethhdr *)data; + data += sizeof(struct ethhdr); + + /* See comment on IGMP in bond_xmit_roundrobin_slave_get() */ + if (eth->h_proto == htons(ETH_P_IP)) { + const struct iphdr *iph; + + if (data + sizeof(struct iphdr) > xdp->data_end) + goto non_igmp; + + iph = (struct iphdr *)data; + + if (iph->protocol == IPPROTO_IGMP) { + slave = rcu_dereference(bond->curr_active_slave); + if (slave) + return slave; + return bond_get_slave_by_id(bond, 0); + } + } + +non_igmp: + slave_cnt = READ_ONCE(bond->slave_cnt); + if (likely(slave_cnt)) { + slave_id = bond_rr_gen_slave_id(bond) % slave_cnt; + return bond_get_slave_by_id(bond, slave_id); + } + return NULL; +} + static netdev_tx_t bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) { @@ -4635,6 +4755,22 @@ static struct slave *bond_xmit_3ad_xor_slave_get(struct bonding *bond, return slave; }
+static struct slave *bond_xdp_xmit_3ad_xor_slave_get(struct bonding *bond, + struct xdp_buff *xdp) +{ + struct bond_up_slave *slaves; + unsigned int count; + u32 hash; + + hash = bond_xmit_hash_xdp(bond, xdp); + slaves = rcu_dereference(bond->usable_slaves); + count = slaves ? READ_ONCE(slaves->count) : 0; + if (unlikely(!count)) + return NULL; + + return slaves->arr[hash % count]; +} + /* Use this Xmit function for 3AD as well as XOR modes. The current * usable slave array is formed in the control path. The xmit function * just calculates hash and sends the packet out. @@ -4843,6 +4979,174 @@ static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) return ret; }
+static struct net_device * +bond_xdp_get_xmit_slave(struct net_device *bond_dev, struct xdp_buff *xdp) +{ + struct bonding *bond = netdev_priv(bond_dev); + struct slave *slave; + + /* Caller needs to hold rcu_read_lock() */ + + switch (BOND_MODE(bond)) { + case BOND_MODE_ROUNDROBIN: + slave = bond_xdp_xmit_roundrobin_slave_get(bond, xdp); + break; + + case BOND_MODE_ACTIVEBACKUP: + slave = bond_xmit_activebackup_slave_get(bond); + break; + + case BOND_MODE_8023AD: + case BOND_MODE_XOR: + slave = bond_xdp_xmit_3ad_xor_slave_get(bond, xdp); + break; + + default: + /* Should never happen. Mode guarded by bond_xdp_check() */ + netdev_err(bond_dev, "Unknown bonding mode %d for xdp xmit\n", BOND_MODE(bond)); + WARN_ON_ONCE(1); + return NULL; + } + + if (slave) + return slave->dev; + + return NULL; +} + +static int bond_xdp_xmit(struct net_device *bond_dev, + int n, struct xdp_frame **frames, u32 flags) +{ + int nxmit, err = -ENXIO; + + rcu_read_lock(); + + for (nxmit = 0; nxmit < n; nxmit++) { + struct xdp_frame *frame = frames[nxmit]; + struct xdp_frame *frames1[] = {frame}; + struct net_device *slave_dev; + struct xdp_buff xdp; + + xdp_convert_frame_to_buff(frame, &xdp); + + slave_dev = bond_xdp_get_xmit_slave(bond_dev, &xdp); + if (!slave_dev) { + err = -ENXIO; + break; + } + + err = slave_dev->netdev_ops->ndo_xdp_xmit(slave_dev, 1, frames1, flags); + if (err < 1) + break; + } + + rcu_read_unlock(); + + /* If error happened on the first frame then we can pass the error up, otherwise + * report the number of frames that were xmitted. + */ + if (err < 0) + return (nxmit == 0 ? err : nxmit); + + return nxmit; +} + +static int bond_xdp_set(struct net_device *dev, struct bpf_prog *prog, + struct netlink_ext_ack *extack) +{ + struct bonding *bond = netdev_priv(dev); + struct list_head *iter; + struct slave *slave, *rollback_slave; + struct bpf_prog *old_prog; + struct netdev_bpf xdp = { + .command = XDP_SETUP_PROG, + .flags = 0, + .prog = prog, + .extack = extack, + }; + int err; + + ASSERT_RTNL(); + + if (!bond_xdp_check(bond)) + return -EOPNOTSUPP; + + old_prog = bond->xdp_prog; + bond->xdp_prog = prog; + + bond_for_each_slave(bond, slave, iter) { + struct net_device *slave_dev = slave->dev; + + if (!slave_dev->netdev_ops->ndo_bpf || + !slave_dev->netdev_ops->ndo_xdp_xmit) { + NL_SET_ERR_MSG(extack, "Slave device does not support XDP"); + slave_err(dev, slave_dev, "Slave does not support XDP\n"); + err = -EOPNOTSUPP; + goto err; + } + + if (dev_xdp_prog_count(slave_dev) > 0) { + NL_SET_ERR_MSG(extack, + "Slave has XDP program loaded, please unload before enslaving"); + slave_err(dev, slave_dev, + "Slave has XDP program loaded, please unload before enslaving\n"); + err = -EOPNOTSUPP; + goto err; + } + + err = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); + if (err < 0) { + /* ndo_bpf() sets extack error message */ + slave_err(dev, slave_dev, "Error %d calling ndo_bpf\n", err); + goto err; + } + if (prog) + bpf_prog_inc(prog); + } + + if (old_prog) + bpf_prog_put(old_prog); + + if (prog) + static_branch_inc(&bpf_master_redirect_enabled_key); + else + static_branch_dec(&bpf_master_redirect_enabled_key); + + return 0; + +err: + /* unwind the program changes */ + bond->xdp_prog = old_prog; + xdp.prog = old_prog; + xdp.extack = NULL; /* do not overwrite original error */ + + bond_for_each_slave(bond, rollback_slave, iter) { + struct net_device *slave_dev = rollback_slave->dev; + int err_unwind; + + if (slave == rollback_slave) + break; + + err_unwind = slave_dev->netdev_ops->ndo_bpf(slave_dev, &xdp); + if (err_unwind < 0) + slave_err(dev, slave_dev, + "Error %d when unwinding XDP program change\n", err_unwind); + else if (xdp.prog) + bpf_prog_inc(xdp.prog); + } + return err; +} + +static int bond_xdp(struct net_device *dev, struct netdev_bpf *xdp) +{ + switch (xdp->command) { + case XDP_SETUP_PROG: + return bond_xdp_set(dev, xdp->prog, xdp->extack); + default: + return -EINVAL; + } +} + static u32 bond_mode_bcast_speed(struct slave *slave, u32 speed) { if (speed == 0 || speed == SPEED_UNKNOWN) @@ -4928,6 +5232,9 @@ static const struct net_device_ops bond_netdev_ops = { .ndo_fix_features = bond_fix_features, .ndo_features_check = passthru_features_check, .ndo_get_xmit_slave = bond_xmit_get_slave, + .ndo_bpf = bond_xdp, + .ndo_xdp_xmit = bond_xdp_xmit, + .ndo_xdp_get_xmit_slave = bond_xdp_get_xmit_slave, };
static const struct device_type bond_type = { diff --git a/include/net/bonding.h b/include/net/bonding.h index 467c9dac43ca..bb238e322e75 100644 --- a/include/net/bonding.h +++ b/include/net/bonding.h @@ -258,6 +258,7 @@ struct bonding { /* protecting ipsec_list */ spinlock_t ipsec_lock; #endif /* CONFIG_XFRM_OFFLOAD */ + struct bpf_prog *xdp_prog; };
#define bond_slave_get_rcu(dev) \