bonding: add IPv6 NS/NA monitor support
Subsystems (5)
cover bonding: add IPv6 NS/NA monitor support
This is an RFC of adding IPv6 NS/NA monitor support for bonding. I posted a draft patch before[1]. But that patch is too big and David Ahern suggested to split it smaller. So I split the previous patch to 5 small ones, maybe not very good :)
The iproute2 patch is here [2].
This patch add bond IPv6 NS/NA monitor support. A new option ns_ip6_target is added, which is similar with arp_ip_target. The IPv6 NS/NA monitor will take effect when there is a valid IPv6 address. And ARP monitor will stop working.
A new field struct in6_addr ip6_addr is added to struct bond_opt_value for IPv6 support. Thus __bond_opt_init() is also updated to check string, addr first.
Function bond_handle_vlan() is split from bond_arp_send() for both IPv4/IPv6 usage.
To alloc NS message and send out. ndisc_ns_create() and ndisc_send_skb() are exported.
[1] https://lore.kernel.org/netdev/[email protected]
[2] https://lore.kernel.org/netdev/[email protected]
Hangbin Liu (5):
ipv6: separate ndisc_ns_create() from ndisc_send_ns()
Bonding: split bond_handle_vlan from bond_arp_send
bonding: add ip6_addr for bond_opt_value
bonding: add new parameter ns_targets
bonding: add new option ns_ip6_target
Documentation/networking/bonding.rst | 11 ++ drivers/net/bonding/bond_main.c | 266 ++++++++++++++++++++++++--- drivers/net/bonding/bond_netlink.c | 55 ++++++ drivers/net/bonding/bond_options.c | 142 +++++++++++++- drivers/net/bonding/bond_sysfs.c | 22 +++ include/net/bond_options.h | 14 +- include/net/bonding.h | 36 ++++ include/net/ndisc.h | 5 + include/uapi/linux/if_link.h | 1 + net/ipv6/ndisc.c | 45 +++-- tools/include/uapi/linux/if_link.h | 1 + 11 files changed, 549 insertions(+), 49 deletions(-)
Signature · 1 line
2.31.1
RFC PATCH 1/5 ipv6: separate ndisc_ns_create() from ndisc_send_ns()
This patch separate NS message allocation steps from ndisc_send_ns(), so it could be used in other places, like bonding, to allocate and send IPv6 NS message.
Also export ndisc_send_skb() and ndisc_ns_create() for later bonding usage.
--- include/net/ndisc.h | 5 +++++ net/ipv6/ndisc.c | 45 ++++++++++++++++++++++++++++++--------------- 2 files changed, 35 insertions(+), 15 deletions(-)
include/net/ndisc.h +5 -0
@@ -447,10 +447,15 @@ void ndisc_cleanup(void);int ndisc_rcv (struct sk_buff *skb); +struct sk_buff *ndisc_ns_create (struct net_device *dev,const struct in6_addr *solicit, +const struct in6_addr *saddr,u64 nonce);void ndisc_send_ns (struct net_device *dev,const struct in6_addr *solicit,const struct in6_addr *daddr,const struct in6_addr *saddr,u64 nonce); +void ndisc_send_skb (struct sk_buff *skb,const struct in6_addr *daddr, +const struct in6_addr *saddr); +void ndisc_send_rs (struct net_device *dev,const struct in6_addr *saddr,const struct in6_addr *daddr); void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr,
net/ipv6/ndisc.c +30 -15
81 lines not on this page · raw message
Signature · 1 line
2.31.1
RFC PATCH 2/5 Bonding: split bond_handle_vlan from bond_arp_send
In reply to Hangbin LiuFunction bond_handle_vlan() is split from bond_arp_send() for later IPv6 usage.
--- drivers/net/bonding/bond_main.c | 57 +++++++++++++++++++-------------- 1 file changed, 33 insertions(+), 24 deletions(-)
drivers/net/bonding/bond_main.c +33 -24
@@ -2794,31 +2794,15 @@ static bool bond_has_this_ip(struct bonding *bond, __be32 ip)return ret; } -/* We go to the (large) trouble of VLAN tagging ARP frames because -* switches in VLAN mode (especially if ports are configured as -* "native" to a VLAN) might not pass non-tagged frames. -*/ -static void bond_arp_send (struct slave *slave,int arp_op,__be32 dest_ip, -__be32 src_ip,struct bond_vlan_tag *tags) +static bool bond_handle_vlan (struct slave *slave,struct bond_vlan_tag *tags, +struct sk_buff *skb) { -struct sk_buff *skb; -struct bond_vlan_tag *outer_tag = tags; -struct net_device *slave_dev = slave->dev;struct net_device *bond_dev = slave->bond->dev; - -slave_dbg (bond_dev, slave_dev,"arp %d on slave: dst %pI4 src %pI4\n" , - arp_op, &dest_ip, &src_ip); - - skb =arp_create (arp_op,ETH_P_ARP , dest_ip, slave_dev, src_ip, -NULL , slave_dev->dev_addr,NULL ); - -if (!skb) { -net_err_ratelimited ("ARP packet allocation failed\n" ); -return ; - } +struct net_device *slave_dev = slave->dev; +struct bond_vlan_tag *outer_tag = tags;if (!tags || tags->vlan_proto ==VLAN_N_VID ) - goto xmit; +return true; tags++;
@@ -2835,7 +2819,7 @@ static void bond_arp_send(struct slave *slave, int arp_op, __be32 dest_ip, tags->vlan_id);if (!skb) {net_err_ratelimited ("failed to insert inner VLAN tag\n" ); -return ; +return false; } tags++;
@@ -2848,8 +2832,33 @@ static void bond_arp_send(struct slave *slave, int arp_op, __be32 dest_ip, outer_tag->vlan_id); } -xmit: -arp_xmit (skb); +return true; +} +/* We go to the (large) trouble of VLAN tagging ARP frames because +* switches in VLAN mode (especially if ports are configured as +* "native" to a VLAN) might not pass non-tagged frames. +*/ +static void bond_arp_send (struct slave *slave,int arp_op,__be32 dest_ip, +__be32 src_ip,struct bond_vlan_tag *tags) +{ +struct sk_buff *skb; +struct net_device *slave_dev = slave->dev; +struct net_device *bond_dev = slave->bond->dev; + +slave_dbg (bond_dev, slave_dev,"arp %d on slave: dst %pI4 src %pI4\n" , + arp_op, &dest_ip, &src_ip); + + skb =arp_create (arp_op,ETH_P_ARP , dest_ip, slave_dev, src_ip, +NULL , slave_dev->dev_addr,NULL ); + +if (!skb) { +net_err_ratelimited ("ARP packet allocation failed\n" ); +return ; + } + +if (bond_handle_vlan (slave, tags, skb)) +arp_xmit (skb); +return ; }/* Validate the device path between the @start_dev and the @end_dev.
Signature · 1 line
2.31.1
RFC PATCH 3/5 bonding: add ip6_addr for bond_opt_value
In reply to Hangbin LiuAdding a new field ip6_addr for bond_opt_value so we can get IPv6 address in future.
Also change the checking logic of __bond_opt_init(). Set string or addr when there is, otherwise set the value.
Is there a need to update bond_opt_parse() for IPv6 address? I think the checking in patch 05 should be enough.
--- include/net/bond_options.h | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)
include/net/bond_options.h +8 -4
@@ -79,6 +79,7 @@ struct bond_opt_value {char *string;u64 value;u32 flags; +struct in6_addr ip6_addr; };struct bonding ;
@@ -118,17 +119,20 @@ const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val);* When value is ULLONG_MAX then string will be used. */ static inline void __bond_opt_init (struct bond_opt_value *optval, -char *string,u64 value) +char *string,u64 value,struct in6_addr *addr) {memset (optval,0 ,sizeof (*optval)); optval->value =ULLONG_MAX ; -if (value ==ULLONG_MAX ) +if (string) optval->string = string; +else if (addr) + optval->ip6_addr = *addr;else optval->value = value; } -#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value) -#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX) +#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value, NULL) +#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX, NULL) +#define bond_opt_initaddr (optval, addr) __bond_opt_init(optval, NULL, ULLONG_MAX, addr)void bond_option_arp_ip_targets_clear (struct bonding *bond);
Signature · 1 line
2.31.1
RFC PATCH 4/5 bonding: add new parameter ns_targets
In reply to Hangbin LiuAdd a new bonding parameter ns_targets to store IPv6 address. Add required bond_ns_send/rcv functions first before adding IPv6 address option setting.
The bond_do_ns_validate() will check if there is any IPv6 address in bond parameter ns_targets. If yes, we will use IPv6 NS/NA probe instead of ARP probe.
--- drivers/net/bonding/bond_main.c | 209 ++++++++++++++++++++++++++++- drivers/net/bonding/bond_options.c | 5 +- include/net/bonding.h | 29 ++++ 3 files changed, 237 insertions(+), 6 deletions(-)
drivers/net/bonding/bond_main.c +204 -5
@@ -88,6 +88,7 @@#if IS_ENABLED (CONFIG_TLS_DEVICE )#include <net/tls.h> #endif +#include <net/ip6_route.h> #include "bonding_priv.h"
@@ -3069,6 +3070,190 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,return RX_HANDLER_ANOTHER ; } +static void bond_ns_send (struct slave *slave,const struct in6_addr *daddr, +const struct in6_addr *saddr,struct bond_vlan_tag *tags) +{ +struct sk_buff *skb; +struct net_device *slave_dev = slave->dev; +struct net_device *bond_dev = slave->bond->dev; +struct in6_addr mcaddr; + +slave_dbg (bond_dev, slave_dev,"NS on slave: dst %pI6 src %pI6\n" , + &daddr, &saddr); + + skb =ndisc_ns_create (slave_dev, daddr, saddr,0 ); +if (!skb) { +net_err_ratelimited ("NS packet allocation failed\n" ); +return ; + } + +addrconf_addr_solict_mult (daddr, &mcaddr); +if (bond_handle_vlan (slave, tags, skb)) +ndisc_send_skb (skb, &mcaddr, saddr); +return ; +} + +static void bond_ns_send_all (struct bonding *bond,struct slave *slave) +{ +struct in6_addr *targets = bond->params.ns_targets; +struct bond_vlan_tag *tags; +struct in6_addr saddr; +struct dst_entry *dst; +struct flowi6 fl6; +int i; + +for (i =0 ; i <BOND_MAX_ARP_TARGETS && !ipv6_addr_any (&targets[i]); i++) { +slave_dbg (bond->dev, slave->dev,"%s: target %pI6\n" , + __func__, &targets[i]); + tags =NULL ; + +/* Find out through which dev should the packet go */ +memset (&fl6,0 ,sizeof (struct flowi6 )); + fl6.daddr = targets[i]; + fl6.flowi6_oif = bond->dev->ifindex; + + dst =ip6_route_output (dev_net (bond->dev),NULL , &fl6); +if (dst->error) { +dst_release (dst); +/* there's no route to target - try to send arp +* probe to generate any traffic (arp_validate=0) +*/ +if (bond->params.arp_validate) +pr_warn_once ("%s: no route to ns_ip6_target %pI6 and arp_validate is set\n" , + bond->dev->name, + &targets[i]); +bond_ns_send (slave, &targets[i], &in6addr_any, tags); +continue ; + } + +/* bond device itself */ +if (dst->dev == bond->dev) + goto found; + +rcu_read_lock (); + tags =bond_verify_device_path (bond->dev, dst->dev,0 ); +rcu_read_unlock (); + +if (!IS_ERR_OR_NULL (tags)) + goto found; + +/* Not our device - skip */ +slave_dbg (bond->dev, slave->dev,"no path to ns_ip6_target %pI6 via dst->dev %s\n" , + &targets[i], dst->dev ? dst->dev->name :"NULL" ); + +dst_release (dst); +continue ; + +found: +if (!ipv6_dev_get_saddr (dev_net (dst->dev), dst->dev, &targets[i],0 , &saddr)) +bond_ns_send (slave, &targets[i], &saddr, tags); +dst_release (dst); +kfree (tags); + } +} + +static int bond_confirm_addr6 (struct net_device *dev, +struct netdev_nested_priv *priv) +{ +struct in6_addr *addr = (struct in6_addr *)priv->data; + +return ipv6_chk_addr (dev_net (dev), addr, dev,0 ); +} + +static bool bond_has_this_ip6 (struct bonding *bond,struct in6_addr *addr) +{ +struct netdev_nested_priv priv = { + .data = addr, + }; +int ret = false; + +if (bond_confirm_addr6 (bond->dev, &priv)) +return true; + +rcu_read_lock (); +if (netdev_walk_all_upper_dev_rcu (bond->dev, bond_confirm_addr6, &priv)) + ret = true; +rcu_read_unlock (); + +return ret; +} + +static void bond_validate_ns (struct bonding *bond,struct slave *slave, +struct in6_addr *saddr,struct in6_addr *daddr) +{ +int i; + +if (ipv6_addr_any (saddr) || !bond_has_this_ip6 (bond, daddr)) { +slave_dbg (bond->dev, slave->dev,"%s: sip %pI6 tip %pI6 not found\n" , + __func__, saddr, daddr); +return ; + } + + i =bond_get_targets_ip6 (bond->params.ns_targets, saddr); +if (i ==-1 ) { +slave_dbg (bond->dev, slave->dev,"%s: sip %pI6 not found in targets\n" , + __func__, saddr); +return ; + } + slave->last_rx = jiffies; + slave->target_last_arp_rx[i] = jiffies; +} + +int bond_na_rcv (const struct sk_buff *skb,struct bonding *bond, +struct slave *slave) +{ +struct icmp6hdr *hdr =icmp6_hdr (skb); +struct slave *curr_active_slave, *curr_arp_slave; +struct in6_addr *saddr, *daddr; +bool is_ipv6 = skb->protocol ==__cpu_to_be16 (ETH_P_IPV6 ); + +/* Use arp validate logic, should we add a NS one? or an alise? */ +if (!slave_do_arp_validate (bond, slave)) { +if ((slave_do_arp_validate_only (bond) && is_ipv6) || + !slave_do_arp_validate_only (bond)) + slave->last_rx = jiffies; +return RX_HANDLER_ANOTHER ; + }else if (!is_ipv6) { +return RX_HANDLER_ANOTHER ; + } + +slave_dbg (bond->dev, slave->dev,"%s: skb->dev %s\n" , + __func__, skb->dev->name); + +if (skb->pkt_type ==PACKET_OTHERHOST || + skb->pkt_type ==PACKET_LOOPBACK || + hdr->icmp6_type !=NDISC_NEIGHBOUR_ADVERTISEMENT ) + goto out; + + saddr = &ipv6_hdr (skb)->saddr; + daddr = &ipv6_hdr (skb)->daddr; + +slave_dbg (bond->dev, slave->dev,"%s: %s/%d av %d sv %d sip %pI6 tip %pI6\n" , + __func__, slave->dev->name,bond_slave_state (slave), + bond->params.arp_validate,slave_do_arp_validate (bond, slave), + saddr, daddr); + + curr_active_slave =rcu_dereference (bond->curr_active_slave); + curr_arp_slave =rcu_dereference (bond->current_arp_slave); + +/* We 'trust' the received ARP enough to validate it if: +* see bond_arp_rcv(). +*/ +if (bond_is_active_slave (slave)) +bond_validate_ns (bond, slave, saddr, daddr); +else if (curr_active_slave && +time_after (slave_last_rx (bond, curr_active_slave), + curr_active_slave->last_link_up)) +bond_validate_ns (bond, slave, saddr, daddr); +else if (curr_arp_slave && +bond_time_in_interval (bond, +dev_trans_start (curr_arp_slave->dev),1 )) +bond_validate_ns (bond, slave, saddr, daddr); + +out: +return RX_HANDLER_ANOTHER ; +} +/* function to verify if we're in the arp_interval timeslice, returns true if * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
@@ -3163,8 +3348,12 @@ static void bond_loadbalance_arp_mon(struct bonding *bond)* do - all replies will be rx'ed on same link causing slaves * to be unstable during low/no traffic periods */ -if (bond_slave_is_up (slave)) -bond_arp_send_all (bond, slave); +if (bond_slave_is_up (slave)) { +if (bond_do_ns_validate (bond)) +bond_ns_send_all (bond, slave); +else +bond_arp_send_all (bond, slave); + } }rcu_read_unlock ();
@@ -3378,7 +3567,10 @@ static bool bond_ab_arp_probe(struct bonding *bond) curr_active_slave->dev->name);if (curr_active_slave) { -bond_arp_send_all (bond, curr_active_slave); +if (bond_do_ns_validate (bond)) +bond_ns_send_all (bond, curr_active_slave); +else +bond_arp_send_all (bond, curr_active_slave);return should_notify_rtnl; }
@@ -3430,7 +3622,10 @@ static bool bond_ab_arp_probe(struct bonding *bond)bond_set_slave_link_state (new_slave,BOND_LINK_BACK ,BOND_SLAVE_NOTIFY_LATER );bond_set_slave_active_flags (new_slave,BOND_SLAVE_NOTIFY_LATER ); -bond_arp_send_all (bond, new_slave); +if (bond_do_ns_validate (bond)) +bond_ns_send_all (bond, new_slave); +else +bond_arp_send_all (bond, new_slave); new_slave->last_link_up = jiffies;rcu_assign_pointer (bond->current_arp_slave, new_slave);
@@ -3966,7 +4161,10 @@ static int bond_open(struct net_device *bond_dev)if (bond->params.arp_interval) {/* arp interval, in milliseconds. */ queue_delayed_work (bond->wq, &bond->arp_work,0 ); - bond->recv_probe = bond_arp_rcv; +if (bond_do_ns_validate (bond)) + bond->recv_probe = bond_na_rcv; +else + bond->recv_probe = bond_arp_rcv; }if (BOND_MODE (bond) ==BOND_MODE_8023AD ) {
@@ -5937,6 +6135,7 @@ static int bond_check_params(struct bond_params *params)strscpy_pad (params->primary, primary,sizeof (params->primary));memcpy (params->arp_targets, arp_target,sizeof (arp_target)); +memset (params->ns_targets,0 ,sizeof (struct in6_addr ) *BOND_MAX_ARP_TARGETS );return 0 ; }
drivers/net/bonding/bond_options.c +4 -1
@@ -1052,7 +1052,10 @@ static int bond_option_arp_interval_set(struct bonding *bond,cancel_delayed_work_sync (&bond->arp_work); }else {/* arp_validate can be set only in active-backup mode */ - bond->recv_probe = bond_arp_rcv; +if (bond_do_ns_validate (bond)) + bond->recv_probe = bond_na_rcv; +else + bond->recv_probe = bond_arp_rcv;cancel_delayed_work_sync (&bond->mii_work);queue_delayed_work (bond->wq, &bond->arp_work,0 ); }
include/net/bonding.h +29 -0
@@ -29,6 +29,8 @@#include <net/bond_3ad.h> #include <net/bond_alb.h> #include <net/bond_options.h> +#include <net/ipv6.h> +#include <net/addrconf.h> #define BOND_MAX_ARP_TARGETS 16
@@ -146,6 +148,7 @@ struct bond_params {struct reciprocal_value reciprocal_packets_per_slave;u16 ad_actor_sys_prio;u16 ad_user_port_key; +struct in6_addr ns_targets[BOND_MAX_ARP_TARGETS ];/* 2 bytes of padding : see ether_addr_equal_64bits() */ u8 ad_actor_system[ETH_ALEN +2 ];
@@ -619,6 +622,18 @@ static inline __be32 bond_confirm_addr(struct net_device *dev, __be32 dst, __be3return addr; } +static inline bool bond_do_ns_validate (struct bonding *bond) +{ +int i; + +for (i =0 ; i <BOND_MAX_ARP_TARGETS ; i++) { +if (!ipv6_addr_any (&bond->params.ns_targets[i])) +return true; + } + +return false; +} +struct bond_net {struct net *net;/* Associated network namespace */ struct list_head dev_list;
@@ -629,6 +644,7 @@ struct bond_net { };int bond_arp_rcv (const struct sk_buff *skb,struct bonding *bond,struct slave *slave); +int bond_na_rcv (const struct sk_buff *skb,struct bonding *bond,struct slave *slave);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);
@@ -749,6 +765,19 @@ static inline int bond_get_targets_ip(__be32 *targets, __be32 ip)return -1 ; } +static inline int bond_get_targets_ip6 (struct in6_addr *targets,struct in6_addr *ip) +{ +int i; + +for (i =0 ; i <BOND_MAX_ARP_TARGETS ; i++) +if (ipv6_addr_equal (&targets[i], ip)) +return i; +else if (ipv6_addr_any (&targets[i])) +break ; + +return -1 ; +} +/* exported from bond_main.c */ extern unsigned int bond_net_id;
Signature · 1 line
2.31.1
RFC PATCH 5/5 bonding: add new option ns_ip6_target
In reply to Hangbin LiuThis patch add a new bonding option ns_ip6_target, which correspond to the arp_ip_target. With this we set IPv6 targets and send IPv6 NS request to determine the health of the link.
For other related options like the validation, we still use arp_validate, and will change to ns_validate later.
--- Documentation/networking/bonding.rst | 11 +++ drivers/net/bonding/bond_netlink.c | 55 +++++++++++ drivers/net/bonding/bond_options.c | 137 +++++++++++++++++++++++++++ drivers/net/bonding/bond_sysfs.c | 22 +++++ include/net/bond_options.h | 2 + include/net/bonding.h | 7 ++ include/uapi/linux/if_link.h | 1 + tools/include/uapi/linux/if_link.h | 1 + 8 files changed, 236 insertions(+)
Documentation/networking/bonding.rst +11 -0
@@ -313,6 +313,17 @@ arp_ip_target maximum number of targets that can be specified is 16. The default value is no IP addresses. +ns_ip6_target + + Specifies the IPv6 addresses to use as IPv6 monitoring peers when + arp_interval is > 0. These are the targets of the NS request + sent to determine the health of the link to the targets. + Specify these values in ffff:ffff::ffff:ffff format. Multiple IPv6 + addresses must be separated by a comma. At least one IPv6 + address must be given for NS/NA monitoring to function. The + maximum number of targets that can be specified is 16. The + default value is no IPv6 addresses. + arp_validate Specifies whether or not ARP probes and replies should be
drivers/net/bonding/bond_netlink.c +55 -0
85 lines not on this page · raw message
drivers/net/bonding/bond_options.c +137 -0
155 lines not on this page · raw message
drivers/net/bonding/bond_sysfs.c +22 -0
@@ -25,6 +25,7 @@#include <linux/nsproxy.h> #include <net/bonding.h> +#include <net/ipv6.h> #define to_bond (cd) ((struct bonding *)(netdev_priv(to_net_dev(cd))))
@@ -315,6 +316,26 @@ static ssize_t bonding_show_missed_max(struct device *d, static DEVICE_ATTR(arp_missed_max, 0644, bonding_show_missed_max, bonding_sysfs_store_option); +static ssize_t bonding_show_ns_targets (struct device *d, +struct device_attribute *attr, +char *buf) +{ +struct bonding *bond =to_bond (d); +int i, res =0 ; + +for (i =0 ; i <BOND_MAX_ARP_TARGETS ; i++) { +if (!ipv6_addr_any (&bond->params.ns_targets[i])) + res +=sprintf (buf + res,"%pI6 " , + &bond->params.ns_targets[i]); + } +if (res) + buf[res-1 ] ='\n' ;/* eat the leftover space */ + +return res; +} +static DEVICE_ATTR(ns_ip6_target, 0644, + bonding_show_ns_targets, bonding_sysfs_store_option); +/* Show the up and down delays. */ static ssize_t bonding_show_downdelay(struct device *d, struct device_attribute *attr,
@@ -761,6 +782,7 @@ static struct attribute *per_bond_attrs[] = { &dev_attr_arp_all_targets.attr, &dev_attr_arp_interval.attr, &dev_attr_arp_ip_target.attr, + &dev_attr_ns_ip6_target.attr, &dev_attr_downdelay.attr, &dev_attr_updelay.attr, &dev_attr_peer_notif_delay.attr,
include/net/bond_options.h +2 -0
@@ -66,6 +66,7 @@ enum {BOND_OPT_PEER_NOTIF_DELAY ,BOND_OPT_LACP_ACTIVE ,BOND_OPT_MISSED_MAX , +BOND_OPT_NS_TARGETS , BOND_OPT_LAST };
@@ -135,5 +136,6 @@ static inline void __bond_opt_init(struct bond_opt_value *optval,#define bond_opt_initaddr (optval, addr) __bond_opt_init(optval, NULL, ULLONG_MAX, addr)void bond_option_arp_ip_targets_clear (struct bonding *bond); +void bond_option_ns_ip6_targets_clear (struct bonding *bond);#endif /* _NET_BOND_OPTIONS_H */
include/net/bonding.h +7 -0
@@ -502,6 +502,13 @@ static inline int bond_is_ip_target_ok(__be32 addr)return !ipv4_is_lbcast (addr) && !ipv4_is_zeronet (addr); } +static inline int bond_is_ip6_target_ok (struct in6_addr *addr) +{ +return !ipv6_addr_any (addr) && + !ipv6_addr_loopback (addr) && + !ipv6_addr_is_multicast (addr); +} +/* Get the oldest arp which we've received on this slave for bond's * arp_targets. */
include/uapi/linux/if_link.h +1 -0
@@ -860,6 +860,7 @@ enum {IFLA_BOND_PEER_NOTIF_DELAY ,IFLA_BOND_AD_LACP_ACTIVE ,IFLA_BOND_MISSED_MAX , +IFLA_BOND_NS_IP6_TARGET , __IFLA_BOND_MAX, };
tools/include/uapi/linux/if_link.h +1 -0
@@ -860,6 +860,7 @@ enum {IFLA_BOND_PEER_NOTIF_DELAY ,IFLA_BOND_AD_LACP_ACTIVE ,IFLA_BOND_MISSED_MAX , +IFLA_BOND_NS_IP6_TARGET , __IFLA_BOND_MAX, };
Signature · 1 line
2.31.1
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
In reply to Hangbin LiuQuoted 41 lines · Hangbin Liu
Adding a new field
ip6_addrforbond_opt_valueso we can get IPv6 address in future.Also change the checking logic of
__bond_opt_init(). Set string or addr when there is, otherwise set the value.Is there a need to update
bond_opt_parse()for IPv6 address? I think the checking in patch 05 should be enough.Signed-off-by Hangbin Liu <[email protected]>--- include/net/bond_options.h | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)include/net/bond_options.h +5
-2@@ -79,6 +79,7 @@ struct bond_opt_value {char *string;u64 value;u32 flags; +struct in6_addr ip6_addr; };struct bonding ;@@ -118,17 +119,20 @@ const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val);* When value is ULLONG_MAX then string will be used. */ static inline void __bond_opt_init (struct bond_opt_value *optval, -char *string,u64 value) +char *string,u64 value,struct in6_addr *addr) {memset (optval,0 ,sizeof (*optval)); optval->value =ULLONG_MAX ; -if (value ==ULLONG_MAX ) +if (string) optval->string = string; +else if (addr) + optval->ip6_addr = *addr;else
include/net/bond_options.h const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val); · line 138
Quoted 10 lines
include/net/bond_options.h +3
-2optval->value = value; } -#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value) -#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX) +#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value, NULL) +#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX, NULL) +#define bond_opt_initaddr (optval, addr) __bond_opt_init(optval, NULL, ULLONG_MAX, addr)void bond_option_arp_ip_targets_clear (struct bonding *bond);
Please don't add arbitrary fields to struct bond_opt_value. As the comment above it states:
/* This structure is used for storing option values and for passing option
* values when changing an option. The logic when used as an arg is as follows:
* - if string != NULL -> parse it, if the opt is RAW type then return it, else
* return the parse result
* - if string == NULL -> parse value
*/
You can use an anonymous union to extend value's size and use the extra room for storage only, that should keep most of the current logic intact.
Thanks,
Nik
Quoted 32 lines · Hangbin Liu
This is an RFC of adding IPv6 NS/NA monitor support for bonding. I posted a draft patch before[1]. But that patch is too big and David Ahern suggested to split it smaller. So I split the previous patch to 5 small ones, maybe not very good :)
The iproute2 patch is here [2].
This patch add bond IPv6 NS/NA monitor support. A new option
ns_ip6_targetis added, which is similar witharp_ip_target. The IPv6 NS/NA monitor will take effect when there is a valid IPv6 address. And ARP monitor will stop working.A new field struct
in6_addrip6_addris added to structbond_opt_valuefor IPv6 support. Thus__bond_opt_init()is also updated to check string, addr first.Function
bond_handle_vlan()is split frombond_arp_send()for both IPv4/IPv6 usage.To alloc NS message and send out.
ndisc_ns_create()andndisc_send_skb()are exported.[1] https://lore.kernel.org/netdev/[email protected]
[2] https://lore.kernel.org/netdev/[email protected]Hangbin Liu (5):
ipv6: separatendisc_ns_create()fromndisc_send_ns()
Bonding: splitbond_handle_vlanfrombond_arp_send
bonding: addip6_addrforbond_opt_value
bonding: add new parameterns_targets
bonding: add new optionns_ip6_target
Quoted 13 lines
Documentation/networking/bonding.rst | 11 ++ drivers/net/bonding/bond_main.c | 266 ++++++++++++++++++++++++--- drivers/net/bonding/bond_netlink.c | 55 ++++++ drivers/net/bonding/bond_options.c | 142 +++++++++++++- drivers/net/bonding/bond_sysfs.c | 22 +++ include/net/bond_options.h | 14 +- include/net/bonding.h | 36 ++++ include/net/ndisc.h | 5 + include/uapi/linux/if_link.h | 1 + net/ipv6/ndisc.c | 45 +++-- tools/include/uapi/linux/if_link.h | 1 + 11 files changed, 549 insertions(+), 49 deletions(-)
Hi,
I'd imagine such option to work alongside ARP, i.e. to be able to have both ARP and ND targets at the same time. On Rx you can choose which one to check based on the protocol, at Tx the same. Then you can reuse and extend most of the current arp procedures to handle IPv6 as well. And most of all remove these ifs all around the code:
+ if (bond_slave_is_up(slave)) {
+ if (bond_do_ns_validate(bond))
+ bond_ns_send_all(bond, slave);
+ else
+ bond_arp_send_all(bond, slave);
+ }
and just have one procedure that handles both if there are any targets for that protocol. That will completely remove the need for bond_do_ns_validate() helper.
Also define BOND_MAX_ND_TARGETS as BOND_MAX_ARP_TARGETS just for the namesake.
Another cosmetic nit: adjust for reverse xmas tree ordering of local variables all over.
Cheers,
Nik
kernel test robot
build botRe: [PATCH RFC net-next 4/5] bonding: add new parameter ns_targets
In reply to Hangbin Liureproduce (this is a W=1 build): # https://github.com/0day-ci/linux/commit/dfbbac3b2cfd5e862f9bb2b79173ba57af5df2fe git remote add linux-review https://github.com/0day-ci/linux git fetch --no-tags…
Hi Hangbin,
[FYI, it's a private test report for your RFC patch.] [auto build test ERROR on net-next/master]
url: https://github.com/0day-ci/linux/commits/Hangbin-Liu/bonding-add-IPv6-NS-NA-monitor-support/20220126-153745 base: https://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git ab14f1802cfb2d7ca120bbf48e3ba6712314ffc3 config: i386-randconfig-a002-20220124 (https://download.01.org/0day-ci/archive/20220127/202201270856.rccZqMC7-lkp(a)intel.com/config) compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce (this is a W=1 build):
# https://github.com/0day-ci/linux/commit/dfbbac3b2cfd5e862f9bb2b79173ba57af5df2fe
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Hangbin-Liu/bonding-add-IPv6-NS-NA-monitor-support/20220126-153745 git checkout dfbbac3b2cfd5e862f9bb2b79173ba57af5df2fe
# save the config file to linux build tree
mkdir build_dir
make W=1 O=build_dir ARCH=i386 SHELL=/bin/bash
If you fix the issue, kindly add following tag as appropriate
All errors (new ones prefixed by >>, old ones prefixed by <<):
>> ERROR: modpost: "ndisc_ns_create" [drivers/net/bonding/bonding.ko] undefined! >> ERROR: modpost: "ipv6_chk_addr" [drivers/net/bonding/bonding.ko] undefined! >> ERROR: modpost: "ndisc_send_skb" [drivers/net/bonding/bonding.ko] undefined! >> ERROR: modpost: "ipv6_dev_get_saddr" [drivers/net/bonding/bonding.ko] undefined! >> ERROR: modpost: "ip6_route_output_flags" [drivers/net/bonding/bonding.ko] undefined!
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
In reply to Nikolay AleksandrovQuoted 46 lines · Nikolay Aleksandrov
Quoted 27 lines
include/net/bond_options.h +5
-2@@ -79,6 +79,7 @@ struct bond_opt_value {char *string;u64 value;u32 flags; +struct in6_addr ip6_addr; };struct bonding ;@@ -118,17 +119,20 @@ const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val);* When value is ULLONG_MAX then string will be used. */ static inline void __bond_opt_init (struct bond_opt_value *optval, -char *string,u64 value) +char *string,u64 value,struct in6_addr *addr) {memset (optval,0 ,sizeof (*optval)); optval->value =ULLONG_MAX ; -if (value ==ULLONG_MAX ) +if (string) optval->string = string; +else if (addr) + optval->ip6_addr = *addr;else include/net/bond_options.h const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val); · line 138
Quoted 10 lines
include/net/bond_options.h +3
-2optval->value = value; } -#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value) -#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX) +#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value, NULL) +#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX, NULL) +#define bond_opt_initaddr (optval, addr) __bond_opt_init(optval, NULL, ULLONG_MAX, addr)void bond_option_arp_ip_targets_clear (struct bonding *bond);Please don't add arbitrary fields to struct
bond_opt_value. As the comment above it states:
/* This structure is used for storing option values and for passing option
* values when changing an option. The logic when used as an arg is as follows:
* - if string != NULL -> parse it, if the opt is RAW type then return it, else
* return the parse result
* - if string == NULL -> parse value
*/
Quoted 2 lines
You can use an anonymous union to extend value's size and use the extra room for storage only, that should keep most of the current logic intact.
Hi Nikolay,
The current checking for string is (value == ULLONG_MAX). If we use a union for IPv6 address and value, what about the address like
ffff:ffff:ffff:ffff::/64?
Thanks
Hangbin
Quoted 13 lines · Nikolay Aleksandrov
Hi,
I'd imagine such option to work alongside ARP, i.e. to be able to have both ARP and ND targets at the same time. On Rx you can choose which one to check based on the protocol, at Tx the same. Then you can reuse and extend most of the current arp procedures to handle IPv6 as well. And most of all remove these ifs all around the code:(quoted diff) +6
-0+ if (bond_slave_is_up(slave)) { + if (bond_do_ns_validate(bond)) + bond_ns_send_all(bond, slave); + else + bond_arp_send_all(bond, slave); + }
Quoted 2 lines
and just have one procedure that handles both if there are any targets for that protocol. That will completely remove the need for
bond_do_ns_validate()helper.
Thanks, I will have a try.
Quoted 2 lines
Also define
BOND_MAX_ND_TARGETSasBOND_MAX_ARP_TARGETSjust for the namesake.
Ah, yes.
Quoted 2 lines
Another cosmetic nit: adjust for reverse xmas tree ordering of local variables all over.
OK, I will.
Thanks
Hangbin
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
In reply to Hangbin LiuQuoted 52 lines · Hangbin Liu
Quoted 46 lines · Nikolay Aleksandrov
Quoted 27 lines
include/net/bond_options.h +5
-2@@ -79,6 +79,7 @@ struct bond_opt_value {char *string;u64 value;u32 flags; +struct in6_addr ip6_addr; };struct bonding ;@@ -118,17 +119,20 @@ const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val);* When value is ULLONG_MAX then string will be used. */ static inline void __bond_opt_init (struct bond_opt_value *optval, -char *string,u64 value) +char *string,u64 value,struct in6_addr *addr) {memset (optval,0 ,sizeof (*optval)); optval->value =ULLONG_MAX ; -if (value ==ULLONG_MAX ) +if (string) optval->string = string; +else if (addr) + optval->ip6_addr = *addr;else include/net/bond_options.h const struct bond_opt_value *bond_opt_get_val(unsigned int option, u64 val); · line 138
Quoted 10 lines
include/net/bond_options.h +3
-2optval->value = value; } -#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value) -#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX) +#define bond_opt_initval (optval, value) __bond_opt_init(optval, NULL, value, NULL) +#define bond_opt_initstr (optval, str) __bond_opt_init(optval, str, ULLONG_MAX, NULL) +#define bond_opt_initaddr (optval, addr) __bond_opt_init(optval, NULL, ULLONG_MAX, addr)void bond_option_arp_ip_targets_clear (struct bonding *bond);Please don't add arbitrary fields to struct
bond_opt_value. As the comment above it states:
/* This structure is used for storing option values and for passing option
* values when changing an option. The logic when used as an arg is as follows:
* - if string != NULL -> parse it, if the opt is RAW type then return it, else
* return the parse result
* - if string == NULL -> parse value
*/Quoted 2 lines
You can use an anonymous union to extend value's size and use the extra room for storage only, that should keep most of the current logic intact.
Hi Nikolay,
Quoted 7 lines
The current checking for string is (value ==
ULLONG_MAX). If we use a union for IPv6 address and value, what about the address likeffff:ffff:ffff:ffff::/64?
Thanks
Hangbin
You're right in that we shouldn't overload value. My point was that bond_opt_val is supposed to be generic, also it wouldn't work as expected for bond_opt_parse(). Perhaps a better solution would be to add a generic extra storage field and length and initialize them with a helper that copies the needed bytes there. As for value in that case you can just set it to 0, since all of this would be used internally the options which support this new extra storage would expect it and should error out if it's missing (wrong/zero length). Maybe something like:
static inline void __bond_opt_init(struct bond_opt_value *optval,
- char *string, u64 value) + char *string, u64 value, + void *extra, size_t extra_len)
with sanity and length checks of course, and:
+#define bond_opt_initextra(optval, extra, len) __bond_opt_init(optval, NULL, 0, extra, len)
It is similar to your solution, but it can be used by other options to store larger values and it uses the value field as indicator that string shouldn't be parsed.
There are other alternatives like using the bond_opt_val flags to denote what has been set instead of using the current struct field checks, but they would cause much more changes that seems unnecessary just for this case.
Cheers,
Nik
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
Quoted 3 lines · Nikolay Aleksandrov
You're right in that we shouldn't overload value. My point was that
bond_opt_valis supposed to be generic, also it wouldn't work as expected forbond_opt_parse(). Perhaps a better solution would be to add a generic extra storage field and length and initialize them with a helper that copies the needed bytes there. As for
Not sure if I understand your suggestion correctly. Do you mean add a field in bond_opt_value like:
#define MAX_LEN 128
struct bond_opt_value {
char *string;
u64 value;
u32 flags;
char extra[MAX_LEN];
};And init it before using?
or define a char *extra and alloc/init the memory when using it?
Thanks
Hangbin
Quoted 24 lines
value in that case you can just set it to 0, since all of this would be used internally the options which support this new extra storage would expect it and should error out if it's missing (wrong/zero length). Maybe something like:
static inline void __bond_opt_init(structbond_opt_value*optval,(quoted diff) +2
-1- char *string, u64 value) + char *string, u64 value, + void *extra, size_t extra_len)with sanity and length checks of course, and:
+#definebond_opt_initextra(optval, extra, len)__bond_opt_init(optval, NULL, 0, extra, len)It is similar to your solution, but it can be used by other options to store larger values and it uses the value field as indicator that string shouldn't be parsed.
There are other alternatives like using the
bond_opt_valflags to denote what has been set instead of using the current struct field checks, but they would cause much more changes that seems unnecessary just for this case.Cheers,
Nik
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
Quoted 17 lines · Hangbin Liu
Quoted 3 lines · Nikolay Aleksandrov
You're right in that we shouldn't overload value. My point was that
bond_opt_valis supposed to be generic, also it wouldn't work as expected forbond_opt_parse(). Perhaps a better solution would be to add a generic extra storage field and length and initialize them with a helper that copies the needed bytes there. As forNot sure if I understand your suggestion correctly. Do you mean add a field in
bond_opt_valuelike:#define
MAX_LEN128struct bond_opt_value { char *string; u64 value; u32 flags; char extra[MAX_LEN]; };
Quoted 7 lines
And init it before using?
or define a char *extra and alloc/init the memory when using it?
Thanks
Hangbin
Yeah, something like that so you can pass around larger items in the extra storage, but please keep it to the minimum e.g. 16 bytes for this case as we have many static bond_opt_values and pass it around a lot.
Quoted 24 lines
Quoted 24 lines
value in that case you can just set it to 0, since all of this would be used internally the options which support this new extra storage would expect it and should error out if it's missing (wrong/zero length). Maybe something like:
static inline void __bond_opt_init(structbond_opt_value*optval,(quoted diff) +2
-1- char *string, u64 value) + char *string, u64 value, + void *extra, size_t extra_len)with sanity and length checks of course, and:
+#definebond_opt_initextra(optval, extra, len)__bond_opt_init(optval, NULL, 0, extra, len)It is similar to your solution, but it can be used by other options to store larger values and it uses the value field as indicator that string shouldn't be parsed.
There are other alternatives like using the
bond_opt_valflags to denote what has been set instead of using the current struct field checks, but they would cause much more changes that seems unnecessary just for this case.Cheers,
Nik
Re: [PATCH RFC net-next 3/5] bonding: add ip6_addr for bond_opt_value
Quoted 3 lines · Nikolay Aleksandrov
Yeah, something like that so you can pass around larger items in the extra storage, but please keep it to the minimum e.g. 16 bytes for this case as we have many static
bond_opt_valuesand pass it around a lot.
OK, I will update the patch after Chinese new year holiday.
Thanks
Hangbin