cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

addrconf.c (183212B)


      1// SPDX-License-Identifier: GPL-2.0-or-later
      2/*
      3 *	IPv6 Address [auto]configuration
      4 *	Linux INET6 implementation
      5 *
      6 *	Authors:
      7 *	Pedro Roque		<roque@di.fc.ul.pt>
      8 *	Alexey Kuznetsov	<kuznet@ms2.inr.ac.ru>
      9 */
     10
     11/*
     12 *	Changes:
     13 *
     14 *	Janos Farkas			:	delete timer on ifdown
     15 *	<chexum@bankinf.banki.hu>
     16 *	Andi Kleen			:	kill double kfree on module
     17 *						unload.
     18 *	Maciej W. Rozycki		:	FDDI support
     19 *	sekiya@USAGI			:	Don't send too many RS
     20 *						packets.
     21 *	yoshfuji@USAGI			:       Fixed interval between DAD
     22 *						packets.
     23 *	YOSHIFUJI Hideaki @USAGI	:	improved accuracy of
     24 *						address validation timer.
     25 *	YOSHIFUJI Hideaki @USAGI	:	Privacy Extensions (RFC3041)
     26 *						support.
     27 *	Yuji SEKIYA @USAGI		:	Don't assign a same IPv6
     28 *						address on a same interface.
     29 *	YOSHIFUJI Hideaki @USAGI	:	ARCnet support
     30 *	YOSHIFUJI Hideaki @USAGI	:	convert /proc/net/if_inet6 to
     31 *						seq_file.
     32 *	YOSHIFUJI Hideaki @USAGI	:	improved source address
     33 *						selection; consider scope,
     34 *						status etc.
     35 */
     36
     37#define pr_fmt(fmt) "IPv6: " fmt
     38
     39#include <linux/errno.h>
     40#include <linux/types.h>
     41#include <linux/kernel.h>
     42#include <linux/sched/signal.h>
     43#include <linux/socket.h>
     44#include <linux/sockios.h>
     45#include <linux/net.h>
     46#include <linux/inet.h>
     47#include <linux/in6.h>
     48#include <linux/netdevice.h>
     49#include <linux/if_addr.h>
     50#include <linux/if_arp.h>
     51#include <linux/if_arcnet.h>
     52#include <linux/if_infiniband.h>
     53#include <linux/route.h>
     54#include <linux/inetdevice.h>
     55#include <linux/init.h>
     56#include <linux/slab.h>
     57#ifdef CONFIG_SYSCTL
     58#include <linux/sysctl.h>
     59#endif
     60#include <linux/capability.h>
     61#include <linux/delay.h>
     62#include <linux/notifier.h>
     63#include <linux/string.h>
     64#include <linux/hash.h>
     65
     66#include <net/net_namespace.h>
     67#include <net/sock.h>
     68#include <net/snmp.h>
     69
     70#include <net/6lowpan.h>
     71#include <net/firewire.h>
     72#include <net/ipv6.h>
     73#include <net/protocol.h>
     74#include <net/ndisc.h>
     75#include <net/ip6_route.h>
     76#include <net/addrconf.h>
     77#include <net/tcp.h>
     78#include <net/ip.h>
     79#include <net/netlink.h>
     80#include <net/pkt_sched.h>
     81#include <net/l3mdev.h>
     82#include <linux/if_tunnel.h>
     83#include <linux/rtnetlink.h>
     84#include <linux/netconf.h>
     85#include <linux/random.h>
     86#include <linux/uaccess.h>
     87#include <asm/unaligned.h>
     88
     89#include <linux/proc_fs.h>
     90#include <linux/seq_file.h>
     91#include <linux/export.h>
     92#include <linux/ioam6.h>
     93
     94#define	INFINITY_LIFE_TIME	0xFFFFFFFF
     95
     96#define IPV6_MAX_STRLEN \
     97	sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255")
     98
     99static inline u32 cstamp_delta(unsigned long cstamp)
    100{
    101	return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
    102}
    103
    104static inline s32 rfc3315_s14_backoff_init(s32 irt)
    105{
    106	/* multiply 'initial retransmission time' by 0.9 .. 1.1 */
    107	u64 tmp = (900000 + prandom_u32() % 200001) * (u64)irt;
    108	do_div(tmp, 1000000);
    109	return (s32)tmp;
    110}
    111
    112static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt)
    113{
    114	/* multiply 'retransmission timeout' by 1.9 .. 2.1 */
    115	u64 tmp = (1900000 + prandom_u32() % 200001) * (u64)rt;
    116	do_div(tmp, 1000000);
    117	if ((s32)tmp > mrt) {
    118		/* multiply 'maximum retransmission time' by 0.9 .. 1.1 */
    119		tmp = (900000 + prandom_u32() % 200001) * (u64)mrt;
    120		do_div(tmp, 1000000);
    121	}
    122	return (s32)tmp;
    123}
    124
    125#ifdef CONFIG_SYSCTL
    126static int addrconf_sysctl_register(struct inet6_dev *idev);
    127static void addrconf_sysctl_unregister(struct inet6_dev *idev);
    128#else
    129static inline int addrconf_sysctl_register(struct inet6_dev *idev)
    130{
    131	return 0;
    132}
    133
    134static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
    135{
    136}
    137#endif
    138
    139static void ipv6_gen_rnd_iid(struct in6_addr *addr);
    140
    141static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
    142static int ipv6_count_addresses(const struct inet6_dev *idev);
    143static int ipv6_generate_stable_address(struct in6_addr *addr,
    144					u8 dad_count,
    145					const struct inet6_dev *idev);
    146
    147#define IN6_ADDR_HSIZE_SHIFT	8
    148#define IN6_ADDR_HSIZE		(1 << IN6_ADDR_HSIZE_SHIFT)
    149
    150static void addrconf_verify(struct net *net);
    151static void addrconf_verify_rtnl(struct net *net);
    152
    153static struct workqueue_struct *addrconf_wq;
    154
    155static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
    156static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
    157
    158static void addrconf_type_change(struct net_device *dev,
    159				 unsigned long event);
    160static int addrconf_ifdown(struct net_device *dev, bool unregister);
    161
    162static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
    163						  int plen,
    164						  const struct net_device *dev,
    165						  u32 flags, u32 noflags,
    166						  bool no_gw);
    167
    168static void addrconf_dad_start(struct inet6_ifaddr *ifp);
    169static void addrconf_dad_work(struct work_struct *w);
    170static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
    171				   bool send_na);
    172static void addrconf_dad_run(struct inet6_dev *idev, bool restart);
    173static void addrconf_rs_timer(struct timer_list *t);
    174static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
    175static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
    176
    177static void inet6_prefix_notify(int event, struct inet6_dev *idev,
    178				struct prefix_info *pinfo);
    179
    180static struct ipv6_devconf ipv6_devconf __read_mostly = {
    181	.forwarding		= 0,
    182	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
    183	.mtu6			= IPV6_MIN_MTU,
    184	.accept_ra		= 1,
    185	.accept_redirects	= 1,
    186	.autoconf		= 1,
    187	.force_mld_version	= 0,
    188	.mldv1_unsolicited_report_interval = 10 * HZ,
    189	.mldv2_unsolicited_report_interval = HZ,
    190	.dad_transmits		= 1,
    191	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
    192	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
    193	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
    194	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
    195	.use_tempaddr		= 0,
    196	.temp_valid_lft		= TEMP_VALID_LIFETIME,
    197	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
    198	.regen_max_retry	= REGEN_MAX_RETRY,
    199	.max_desync_factor	= MAX_DESYNC_FACTOR,
    200	.max_addresses		= IPV6_MAX_ADDRESSES,
    201	.accept_ra_defrtr	= 1,
    202	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
    203	.accept_ra_from_local	= 0,
    204	.accept_ra_min_hop_limit= 1,
    205	.accept_ra_pinfo	= 1,
    206#ifdef CONFIG_IPV6_ROUTER_PREF
    207	.accept_ra_rtr_pref	= 1,
    208	.rtr_probe_interval	= 60 * HZ,
    209#ifdef CONFIG_IPV6_ROUTE_INFO
    210	.accept_ra_rt_info_min_plen = 0,
    211	.accept_ra_rt_info_max_plen = 0,
    212#endif
    213#endif
    214	.proxy_ndp		= 0,
    215	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
    216	.disable_ipv6		= 0,
    217	.accept_dad		= 0,
    218	.suppress_frag_ndisc	= 1,
    219	.accept_ra_mtu		= 1,
    220	.stable_secret		= {
    221		.initialized = false,
    222	},
    223	.use_oif_addrs_only	= 0,
    224	.ignore_routes_with_linkdown = 0,
    225	.keep_addr_on_down	= 0,
    226	.seg6_enabled		= 0,
    227#ifdef CONFIG_IPV6_SEG6_HMAC
    228	.seg6_require_hmac	= 0,
    229#endif
    230	.enhanced_dad           = 1,
    231	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
    232	.disable_policy		= 0,
    233	.rpl_seg_enabled	= 0,
    234	.ioam6_enabled		= 0,
    235	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
    236	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
    237	.ndisc_evict_nocarrier	= 1,
    238};
    239
    240static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
    241	.forwarding		= 0,
    242	.hop_limit		= IPV6_DEFAULT_HOPLIMIT,
    243	.mtu6			= IPV6_MIN_MTU,
    244	.accept_ra		= 1,
    245	.accept_redirects	= 1,
    246	.autoconf		= 1,
    247	.force_mld_version	= 0,
    248	.mldv1_unsolicited_report_interval = 10 * HZ,
    249	.mldv2_unsolicited_report_interval = HZ,
    250	.dad_transmits		= 1,
    251	.rtr_solicits		= MAX_RTR_SOLICITATIONS,
    252	.rtr_solicit_interval	= RTR_SOLICITATION_INTERVAL,
    253	.rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL,
    254	.rtr_solicit_delay	= MAX_RTR_SOLICITATION_DELAY,
    255	.use_tempaddr		= 0,
    256	.temp_valid_lft		= TEMP_VALID_LIFETIME,
    257	.temp_prefered_lft	= TEMP_PREFERRED_LIFETIME,
    258	.regen_max_retry	= REGEN_MAX_RETRY,
    259	.max_desync_factor	= MAX_DESYNC_FACTOR,
    260	.max_addresses		= IPV6_MAX_ADDRESSES,
    261	.accept_ra_defrtr	= 1,
    262	.ra_defrtr_metric	= IP6_RT_PRIO_USER,
    263	.accept_ra_from_local	= 0,
    264	.accept_ra_min_hop_limit= 1,
    265	.accept_ra_pinfo	= 1,
    266#ifdef CONFIG_IPV6_ROUTER_PREF
    267	.accept_ra_rtr_pref	= 1,
    268	.rtr_probe_interval	= 60 * HZ,
    269#ifdef CONFIG_IPV6_ROUTE_INFO
    270	.accept_ra_rt_info_min_plen = 0,
    271	.accept_ra_rt_info_max_plen = 0,
    272#endif
    273#endif
    274	.proxy_ndp		= 0,
    275	.accept_source_route	= 0,	/* we do not accept RH0 by default. */
    276	.disable_ipv6		= 0,
    277	.accept_dad		= 1,
    278	.suppress_frag_ndisc	= 1,
    279	.accept_ra_mtu		= 1,
    280	.stable_secret		= {
    281		.initialized = false,
    282	},
    283	.use_oif_addrs_only	= 0,
    284	.ignore_routes_with_linkdown = 0,
    285	.keep_addr_on_down	= 0,
    286	.seg6_enabled		= 0,
    287#ifdef CONFIG_IPV6_SEG6_HMAC
    288	.seg6_require_hmac	= 0,
    289#endif
    290	.enhanced_dad           = 1,
    291	.addr_gen_mode		= IN6_ADDR_GEN_MODE_EUI64,
    292	.disable_policy		= 0,
    293	.rpl_seg_enabled	= 0,
    294	.ioam6_enabled		= 0,
    295	.ioam6_id               = IOAM6_DEFAULT_IF_ID,
    296	.ioam6_id_wide		= IOAM6_DEFAULT_IF_ID_WIDE,
    297	.ndisc_evict_nocarrier	= 1,
    298};
    299
    300/* Check if link is ready: is it up and is a valid qdisc available */
    301static inline bool addrconf_link_ready(const struct net_device *dev)
    302{
    303	return netif_oper_up(dev) && !qdisc_tx_is_noop(dev);
    304}
    305
    306static void addrconf_del_rs_timer(struct inet6_dev *idev)
    307{
    308	if (del_timer(&idev->rs_timer))
    309		__in6_dev_put(idev);
    310}
    311
    312static void addrconf_del_dad_work(struct inet6_ifaddr *ifp)
    313{
    314	if (cancel_delayed_work(&ifp->dad_work))
    315		__in6_ifa_put(ifp);
    316}
    317
    318static void addrconf_mod_rs_timer(struct inet6_dev *idev,
    319				  unsigned long when)
    320{
    321	if (!timer_pending(&idev->rs_timer))
    322		in6_dev_hold(idev);
    323	mod_timer(&idev->rs_timer, jiffies + when);
    324}
    325
    326static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp,
    327				   unsigned long delay)
    328{
    329	in6_ifa_hold(ifp);
    330	if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay))
    331		in6_ifa_put(ifp);
    332}
    333
    334static int snmp6_alloc_dev(struct inet6_dev *idev)
    335{
    336	int i;
    337
    338	idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT);
    339	if (!idev->stats.ipv6)
    340		goto err_ip;
    341
    342	for_each_possible_cpu(i) {
    343		struct ipstats_mib *addrconf_stats;
    344		addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i);
    345		u64_stats_init(&addrconf_stats->syncp);
    346	}
    347
    348
    349	idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
    350					GFP_KERNEL);
    351	if (!idev->stats.icmpv6dev)
    352		goto err_icmp;
    353	idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
    354					   GFP_KERNEL_ACCOUNT);
    355	if (!idev->stats.icmpv6msgdev)
    356		goto err_icmpmsg;
    357
    358	return 0;
    359
    360err_icmpmsg:
    361	kfree(idev->stats.icmpv6dev);
    362err_icmp:
    363	free_percpu(idev->stats.ipv6);
    364err_ip:
    365	return -ENOMEM;
    366}
    367
    368static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
    369{
    370	struct inet6_dev *ndev;
    371	int err = -ENOMEM;
    372
    373	ASSERT_RTNL();
    374
    375	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)
    376		return ERR_PTR(-EINVAL);
    377
    378	ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT);
    379	if (!ndev)
    380		return ERR_PTR(err);
    381
    382	rwlock_init(&ndev->lock);
    383	ndev->dev = dev;
    384	INIT_LIST_HEAD(&ndev->addr_list);
    385	timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0);
    386	memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
    387
    388	if (ndev->cnf.stable_secret.initialized)
    389		ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
    390
    391	ndev->cnf.mtu6 = dev->mtu;
    392	ndev->ra_mtu = 0;
    393	ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
    394	if (!ndev->nd_parms) {
    395		kfree(ndev);
    396		return ERR_PTR(err);
    397	}
    398	if (ndev->cnf.forwarding)
    399		dev_disable_lro(dev);
    400	/* We refer to the device */
    401	dev_hold_track(dev, &ndev->dev_tracker, GFP_KERNEL);
    402
    403	if (snmp6_alloc_dev(ndev) < 0) {
    404		netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",
    405			   __func__);
    406		neigh_parms_release(&nd_tbl, ndev->nd_parms);
    407		dev_put_track(dev, &ndev->dev_tracker);
    408		kfree(ndev);
    409		return ERR_PTR(err);
    410	}
    411
    412	if (dev != blackhole_netdev) {
    413		if (snmp6_register_dev(ndev) < 0) {
    414			netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n",
    415				   __func__, dev->name);
    416			goto err_release;
    417		}
    418	}
    419	/* One reference from device. */
    420	refcount_set(&ndev->refcnt, 1);
    421
    422	if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
    423		ndev->cnf.accept_dad = -1;
    424
    425#if IS_ENABLED(CONFIG_IPV6_SIT)
    426	if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
    427		pr_info("%s: Disabled Multicast RS\n", dev->name);
    428		ndev->cnf.rtr_solicits = 0;
    429	}
    430#endif
    431
    432	INIT_LIST_HEAD(&ndev->tempaddr_list);
    433	ndev->desync_factor = U32_MAX;
    434	if ((dev->flags&IFF_LOOPBACK) ||
    435	    dev->type == ARPHRD_TUNNEL ||
    436	    dev->type == ARPHRD_TUNNEL6 ||
    437	    dev->type == ARPHRD_SIT ||
    438	    dev->type == ARPHRD_NONE) {
    439		ndev->cnf.use_tempaddr = -1;
    440	}
    441
    442	ndev->token = in6addr_any;
    443
    444	if (netif_running(dev) && addrconf_link_ready(dev))
    445		ndev->if_flags |= IF_READY;
    446
    447	ipv6_mc_init_dev(ndev);
    448	ndev->tstamp = jiffies;
    449	if (dev != blackhole_netdev) {
    450		err = addrconf_sysctl_register(ndev);
    451		if (err) {
    452			ipv6_mc_destroy_dev(ndev);
    453			snmp6_unregister_dev(ndev);
    454			goto err_release;
    455		}
    456	}
    457	/* protected by rtnl_lock */
    458	rcu_assign_pointer(dev->ip6_ptr, ndev);
    459
    460	if (dev != blackhole_netdev) {
    461		/* Join interface-local all-node multicast group */
    462		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
    463
    464		/* Join all-node multicast group */
    465		ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
    466
    467		/* Join all-router multicast group if forwarding is set */
    468		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
    469			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
    470	}
    471	return ndev;
    472
    473err_release:
    474	neigh_parms_release(&nd_tbl, ndev->nd_parms);
    475	ndev->dead = 1;
    476	in6_dev_finish_destroy(ndev);
    477	return ERR_PTR(err);
    478}
    479
    480static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
    481{
    482	struct inet6_dev *idev;
    483
    484	ASSERT_RTNL();
    485
    486	idev = __in6_dev_get(dev);
    487	if (!idev) {
    488		idev = ipv6_add_dev(dev);
    489		if (IS_ERR(idev))
    490			return idev;
    491	}
    492
    493	if (dev->flags&IFF_UP)
    494		ipv6_mc_up(idev);
    495	return idev;
    496}
    497
    498static int inet6_netconf_msgsize_devconf(int type)
    499{
    500	int size =  NLMSG_ALIGN(sizeof(struct netconfmsg))
    501		    + nla_total_size(4);	/* NETCONFA_IFINDEX */
    502	bool all = false;
    503
    504	if (type == NETCONFA_ALL)
    505		all = true;
    506
    507	if (all || type == NETCONFA_FORWARDING)
    508		size += nla_total_size(4);
    509#ifdef CONFIG_IPV6_MROUTE
    510	if (all || type == NETCONFA_MC_FORWARDING)
    511		size += nla_total_size(4);
    512#endif
    513	if (all || type == NETCONFA_PROXY_NEIGH)
    514		size += nla_total_size(4);
    515
    516	if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN)
    517		size += nla_total_size(4);
    518
    519	return size;
    520}
    521
    522static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex,
    523				      struct ipv6_devconf *devconf, u32 portid,
    524				      u32 seq, int event, unsigned int flags,
    525				      int type)
    526{
    527	struct nlmsghdr  *nlh;
    528	struct netconfmsg *ncm;
    529	bool all = false;
    530
    531	nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg),
    532			flags);
    533	if (!nlh)
    534		return -EMSGSIZE;
    535
    536	if (type == NETCONFA_ALL)
    537		all = true;
    538
    539	ncm = nlmsg_data(nlh);
    540	ncm->ncm_family = AF_INET6;
    541
    542	if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0)
    543		goto nla_put_failure;
    544
    545	if (!devconf)
    546		goto out;
    547
    548	if ((all || type == NETCONFA_FORWARDING) &&
    549	    nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0)
    550		goto nla_put_failure;
    551#ifdef CONFIG_IPV6_MROUTE
    552	if ((all || type == NETCONFA_MC_FORWARDING) &&
    553	    nla_put_s32(skb, NETCONFA_MC_FORWARDING,
    554			atomic_read(&devconf->mc_forwarding)) < 0)
    555		goto nla_put_failure;
    556#endif
    557	if ((all || type == NETCONFA_PROXY_NEIGH) &&
    558	    nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0)
    559		goto nla_put_failure;
    560
    561	if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) &&
    562	    nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
    563			devconf->ignore_routes_with_linkdown) < 0)
    564		goto nla_put_failure;
    565
    566out:
    567	nlmsg_end(skb, nlh);
    568	return 0;
    569
    570nla_put_failure:
    571	nlmsg_cancel(skb, nlh);
    572	return -EMSGSIZE;
    573}
    574
    575void inet6_netconf_notify_devconf(struct net *net, int event, int type,
    576				  int ifindex, struct ipv6_devconf *devconf)
    577{
    578	struct sk_buff *skb;
    579	int err = -ENOBUFS;
    580
    581	skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL);
    582	if (!skb)
    583		goto errout;
    584
    585	err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0,
    586					 event, 0, type);
    587	if (err < 0) {
    588		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
    589		WARN_ON(err == -EMSGSIZE);
    590		kfree_skb(skb);
    591		goto errout;
    592	}
    593	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL);
    594	return;
    595errout:
    596	rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err);
    597}
    598
    599static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = {
    600	[NETCONFA_IFINDEX]	= { .len = sizeof(int) },
    601	[NETCONFA_FORWARDING]	= { .len = sizeof(int) },
    602	[NETCONFA_PROXY_NEIGH]	= { .len = sizeof(int) },
    603	[NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN]	= { .len = sizeof(int) },
    604};
    605
    606static int inet6_netconf_valid_get_req(struct sk_buff *skb,
    607				       const struct nlmsghdr *nlh,
    608				       struct nlattr **tb,
    609				       struct netlink_ext_ack *extack)
    610{
    611	int i, err;
    612
    613	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) {
    614		NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request");
    615		return -EINVAL;
    616	}
    617
    618	if (!netlink_strict_get_check(skb))
    619		return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg),
    620					      tb, NETCONFA_MAX,
    621					      devconf_ipv6_policy, extack);
    622
    623	err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg),
    624					    tb, NETCONFA_MAX,
    625					    devconf_ipv6_policy, extack);
    626	if (err)
    627		return err;
    628
    629	for (i = 0; i <= NETCONFA_MAX; i++) {
    630		if (!tb[i])
    631			continue;
    632
    633		switch (i) {
    634		case NETCONFA_IFINDEX:
    635			break;
    636		default:
    637			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request");
    638			return -EINVAL;
    639		}
    640	}
    641
    642	return 0;
    643}
    644
    645static int inet6_netconf_get_devconf(struct sk_buff *in_skb,
    646				     struct nlmsghdr *nlh,
    647				     struct netlink_ext_ack *extack)
    648{
    649	struct net *net = sock_net(in_skb->sk);
    650	struct nlattr *tb[NETCONFA_MAX+1];
    651	struct inet6_dev *in6_dev = NULL;
    652	struct net_device *dev = NULL;
    653	struct sk_buff *skb;
    654	struct ipv6_devconf *devconf;
    655	int ifindex;
    656	int err;
    657
    658	err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack);
    659	if (err < 0)
    660		return err;
    661
    662	if (!tb[NETCONFA_IFINDEX])
    663		return -EINVAL;
    664
    665	err = -EINVAL;
    666	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
    667	switch (ifindex) {
    668	case NETCONFA_IFINDEX_ALL:
    669		devconf = net->ipv6.devconf_all;
    670		break;
    671	case NETCONFA_IFINDEX_DEFAULT:
    672		devconf = net->ipv6.devconf_dflt;
    673		break;
    674	default:
    675		dev = dev_get_by_index(net, ifindex);
    676		if (!dev)
    677			return -EINVAL;
    678		in6_dev = in6_dev_get(dev);
    679		if (!in6_dev)
    680			goto errout;
    681		devconf = &in6_dev->cnf;
    682		break;
    683	}
    684
    685	err = -ENOBUFS;
    686	skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
    687	if (!skb)
    688		goto errout;
    689
    690	err = inet6_netconf_fill_devconf(skb, ifindex, devconf,
    691					 NETLINK_CB(in_skb).portid,
    692					 nlh->nlmsg_seq, RTM_NEWNETCONF, 0,
    693					 NETCONFA_ALL);
    694	if (err < 0) {
    695		/* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */
    696		WARN_ON(err == -EMSGSIZE);
    697		kfree_skb(skb);
    698		goto errout;
    699	}
    700	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
    701errout:
    702	if (in6_dev)
    703		in6_dev_put(in6_dev);
    704	dev_put(dev);
    705	return err;
    706}
    707
    708static int inet6_netconf_dump_devconf(struct sk_buff *skb,
    709				      struct netlink_callback *cb)
    710{
    711	const struct nlmsghdr *nlh = cb->nlh;
    712	struct net *net = sock_net(skb->sk);
    713	int h, s_h;
    714	int idx, s_idx;
    715	struct net_device *dev;
    716	struct inet6_dev *idev;
    717	struct hlist_head *head;
    718
    719	if (cb->strict_check) {
    720		struct netlink_ext_ack *extack = cb->extack;
    721		struct netconfmsg *ncm;
    722
    723		if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) {
    724			NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request");
    725			return -EINVAL;
    726		}
    727
    728		if (nlmsg_attrlen(nlh, sizeof(*ncm))) {
    729			NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request");
    730			return -EINVAL;
    731		}
    732	}
    733
    734	s_h = cb->args[0];
    735	s_idx = idx = cb->args[1];
    736
    737	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
    738		idx = 0;
    739		head = &net->dev_index_head[h];
    740		rcu_read_lock();
    741		cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^
    742			  net->dev_base_seq;
    743		hlist_for_each_entry_rcu(dev, head, index_hlist) {
    744			if (idx < s_idx)
    745				goto cont;
    746			idev = __in6_dev_get(dev);
    747			if (!idev)
    748				goto cont;
    749
    750			if (inet6_netconf_fill_devconf(skb, dev->ifindex,
    751						       &idev->cnf,
    752						       NETLINK_CB(cb->skb).portid,
    753						       nlh->nlmsg_seq,
    754						       RTM_NEWNETCONF,
    755						       NLM_F_MULTI,
    756						       NETCONFA_ALL) < 0) {
    757				rcu_read_unlock();
    758				goto done;
    759			}
    760			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
    761cont:
    762			idx++;
    763		}
    764		rcu_read_unlock();
    765	}
    766	if (h == NETDEV_HASHENTRIES) {
    767		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL,
    768					       net->ipv6.devconf_all,
    769					       NETLINK_CB(cb->skb).portid,
    770					       nlh->nlmsg_seq,
    771					       RTM_NEWNETCONF, NLM_F_MULTI,
    772					       NETCONFA_ALL) < 0)
    773			goto done;
    774		else
    775			h++;
    776	}
    777	if (h == NETDEV_HASHENTRIES + 1) {
    778		if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT,
    779					       net->ipv6.devconf_dflt,
    780					       NETLINK_CB(cb->skb).portid,
    781					       nlh->nlmsg_seq,
    782					       RTM_NEWNETCONF, NLM_F_MULTI,
    783					       NETCONFA_ALL) < 0)
    784			goto done;
    785		else
    786			h++;
    787	}
    788done:
    789	cb->args[0] = h;
    790	cb->args[1] = idx;
    791
    792	return skb->len;
    793}
    794
    795#ifdef CONFIG_SYSCTL
    796static void dev_forward_change(struct inet6_dev *idev)
    797{
    798	struct net_device *dev;
    799	struct inet6_ifaddr *ifa;
    800	LIST_HEAD(tmp_addr_list);
    801
    802	if (!idev)
    803		return;
    804	dev = idev->dev;
    805	if (idev->cnf.forwarding)
    806		dev_disable_lro(dev);
    807	if (dev->flags & IFF_MULTICAST) {
    808		if (idev->cnf.forwarding) {
    809			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
    810			ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters);
    811			ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters);
    812		} else {
    813			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
    814			ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters);
    815			ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters);
    816		}
    817	}
    818
    819	read_lock_bh(&idev->lock);
    820	list_for_each_entry(ifa, &idev->addr_list, if_list) {
    821		if (ifa->flags&IFA_F_TENTATIVE)
    822			continue;
    823		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
    824	}
    825	read_unlock_bh(&idev->lock);
    826
    827	while (!list_empty(&tmp_addr_list)) {
    828		ifa = list_first_entry(&tmp_addr_list,
    829				       struct inet6_ifaddr, if_list_aux);
    830		list_del(&ifa->if_list_aux);
    831		if (idev->cnf.forwarding)
    832			addrconf_join_anycast(ifa);
    833		else
    834			addrconf_leave_anycast(ifa);
    835	}
    836
    837	inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
    838				     NETCONFA_FORWARDING,
    839				     dev->ifindex, &idev->cnf);
    840}
    841
    842
    843static void addrconf_forward_change(struct net *net, __s32 newf)
    844{
    845	struct net_device *dev;
    846	struct inet6_dev *idev;
    847
    848	for_each_netdev(net, dev) {
    849		idev = __in6_dev_get(dev);
    850		if (idev) {
    851			int changed = (!idev->cnf.forwarding) ^ (!newf);
    852			idev->cnf.forwarding = newf;
    853			if (changed)
    854				dev_forward_change(idev);
    855		}
    856	}
    857}
    858
    859static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
    860{
    861	struct net *net;
    862	int old;
    863
    864	if (!rtnl_trylock())
    865		return restart_syscall();
    866
    867	net = (struct net *)table->extra2;
    868	old = *p;
    869	*p = newf;
    870
    871	if (p == &net->ipv6.devconf_dflt->forwarding) {
    872		if ((!newf) ^ (!old))
    873			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
    874						     NETCONFA_FORWARDING,
    875						     NETCONFA_IFINDEX_DEFAULT,
    876						     net->ipv6.devconf_dflt);
    877		rtnl_unlock();
    878		return 0;
    879	}
    880
    881	if (p == &net->ipv6.devconf_all->forwarding) {
    882		int old_dflt = net->ipv6.devconf_dflt->forwarding;
    883
    884		net->ipv6.devconf_dflt->forwarding = newf;
    885		if ((!newf) ^ (!old_dflt))
    886			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
    887						     NETCONFA_FORWARDING,
    888						     NETCONFA_IFINDEX_DEFAULT,
    889						     net->ipv6.devconf_dflt);
    890
    891		addrconf_forward_change(net, newf);
    892		if ((!newf) ^ (!old))
    893			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
    894						     NETCONFA_FORWARDING,
    895						     NETCONFA_IFINDEX_ALL,
    896						     net->ipv6.devconf_all);
    897	} else if ((!newf) ^ (!old))
    898		dev_forward_change((struct inet6_dev *)table->extra1);
    899	rtnl_unlock();
    900
    901	if (newf)
    902		rt6_purge_dflt_routers(net);
    903	return 1;
    904}
    905
    906static void addrconf_linkdown_change(struct net *net, __s32 newf)
    907{
    908	struct net_device *dev;
    909	struct inet6_dev *idev;
    910
    911	for_each_netdev(net, dev) {
    912		idev = __in6_dev_get(dev);
    913		if (idev) {
    914			int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf);
    915
    916			idev->cnf.ignore_routes_with_linkdown = newf;
    917			if (changed)
    918				inet6_netconf_notify_devconf(dev_net(dev),
    919							     RTM_NEWNETCONF,
    920							     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
    921							     dev->ifindex,
    922							     &idev->cnf);
    923		}
    924	}
    925}
    926
    927static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf)
    928{
    929	struct net *net;
    930	int old;
    931
    932	if (!rtnl_trylock())
    933		return restart_syscall();
    934
    935	net = (struct net *)table->extra2;
    936	old = *p;
    937	*p = newf;
    938
    939	if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) {
    940		if ((!newf) ^ (!old))
    941			inet6_netconf_notify_devconf(net,
    942						     RTM_NEWNETCONF,
    943						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
    944						     NETCONFA_IFINDEX_DEFAULT,
    945						     net->ipv6.devconf_dflt);
    946		rtnl_unlock();
    947		return 0;
    948	}
    949
    950	if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) {
    951		net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf;
    952		addrconf_linkdown_change(net, newf);
    953		if ((!newf) ^ (!old))
    954			inet6_netconf_notify_devconf(net,
    955						     RTM_NEWNETCONF,
    956						     NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN,
    957						     NETCONFA_IFINDEX_ALL,
    958						     net->ipv6.devconf_all);
    959	}
    960	rtnl_unlock();
    961
    962	return 1;
    963}
    964
    965#endif
    966
    967/* Nobody refers to this ifaddr, destroy it */
    968void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
    969{
    970	WARN_ON(!hlist_unhashed(&ifp->addr_lst));
    971
    972#ifdef NET_REFCNT_DEBUG
    973	pr_debug("%s\n", __func__);
    974#endif
    975
    976	in6_dev_put(ifp->idev);
    977
    978	if (cancel_delayed_work(&ifp->dad_work))
    979		pr_notice("delayed DAD work was pending while freeing ifa=%p\n",
    980			  ifp);
    981
    982	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
    983		pr_warn("Freeing alive inet6 address %p\n", ifp);
    984		return;
    985	}
    986
    987	kfree_rcu(ifp, rcu);
    988}
    989
    990static void
    991ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
    992{
    993	struct list_head *p;
    994	int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
    995
    996	/*
    997	 * Each device address list is sorted in order of scope -
    998	 * global before linklocal.
    999	 */
   1000	list_for_each(p, &idev->addr_list) {
   1001		struct inet6_ifaddr *ifa
   1002			= list_entry(p, struct inet6_ifaddr, if_list);
   1003		if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
   1004			break;
   1005	}
   1006
   1007	list_add_tail_rcu(&ifp->if_list, p);
   1008}
   1009
   1010static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr)
   1011{
   1012	u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net);
   1013
   1014	return hash_32(val, IN6_ADDR_HSIZE_SHIFT);
   1015}
   1016
   1017static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
   1018			       struct net_device *dev, unsigned int hash)
   1019{
   1020	struct inet6_ifaddr *ifp;
   1021
   1022	hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
   1023		if (ipv6_addr_equal(&ifp->addr, addr)) {
   1024			if (!dev || ifp->idev->dev == dev)
   1025				return true;
   1026		}
   1027	}
   1028	return false;
   1029}
   1030
   1031static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa)
   1032{
   1033	struct net *net = dev_net(dev);
   1034	unsigned int hash = inet6_addr_hash(net, &ifa->addr);
   1035	int err = 0;
   1036
   1037	spin_lock(&net->ipv6.addrconf_hash_lock);
   1038
   1039	/* Ignore adding duplicate addresses on an interface */
   1040	if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) {
   1041		netdev_dbg(dev, "ipv6_add_addr: already assigned\n");
   1042		err = -EEXIST;
   1043	} else {
   1044		hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]);
   1045	}
   1046
   1047	spin_unlock(&net->ipv6.addrconf_hash_lock);
   1048
   1049	return err;
   1050}
   1051
   1052/* On success it returns ifp with increased reference count */
   1053
   1054static struct inet6_ifaddr *
   1055ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg,
   1056	      bool can_block, struct netlink_ext_ack *extack)
   1057{
   1058	gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC;
   1059	int addr_type = ipv6_addr_type(cfg->pfx);
   1060	struct net *net = dev_net(idev->dev);
   1061	struct inet6_ifaddr *ifa = NULL;
   1062	struct fib6_info *f6i = NULL;
   1063	int err = 0;
   1064
   1065	if (addr_type == IPV6_ADDR_ANY ||
   1066	    (addr_type & IPV6_ADDR_MULTICAST &&
   1067	     !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) ||
   1068	    (!(idev->dev->flags & IFF_LOOPBACK) &&
   1069	     !netif_is_l3_master(idev->dev) &&
   1070	     addr_type & IPV6_ADDR_LOOPBACK))
   1071		return ERR_PTR(-EADDRNOTAVAIL);
   1072
   1073	if (idev->dead) {
   1074		err = -ENODEV;			/*XXX*/
   1075		goto out;
   1076	}
   1077
   1078	if (idev->cnf.disable_ipv6) {
   1079		err = -EACCES;
   1080		goto out;
   1081	}
   1082
   1083	/* validator notifier needs to be blocking;
   1084	 * do not call in atomic context
   1085	 */
   1086	if (can_block) {
   1087		struct in6_validator_info i6vi = {
   1088			.i6vi_addr = *cfg->pfx,
   1089			.i6vi_dev = idev,
   1090			.extack = extack,
   1091		};
   1092
   1093		err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi);
   1094		err = notifier_to_errno(err);
   1095		if (err < 0)
   1096			goto out;
   1097	}
   1098
   1099	ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT);
   1100	if (!ifa) {
   1101		err = -ENOBUFS;
   1102		goto out;
   1103	}
   1104
   1105	f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags);
   1106	if (IS_ERR(f6i)) {
   1107		err = PTR_ERR(f6i);
   1108		f6i = NULL;
   1109		goto out;
   1110	}
   1111
   1112	neigh_parms_data_state_setall(idev->nd_parms);
   1113
   1114	ifa->addr = *cfg->pfx;
   1115	if (cfg->peer_pfx)
   1116		ifa->peer_addr = *cfg->peer_pfx;
   1117
   1118	spin_lock_init(&ifa->lock);
   1119	INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work);
   1120	INIT_HLIST_NODE(&ifa->addr_lst);
   1121	ifa->scope = cfg->scope;
   1122	ifa->prefix_len = cfg->plen;
   1123	ifa->rt_priority = cfg->rt_priority;
   1124	ifa->flags = cfg->ifa_flags;
   1125	ifa->ifa_proto = cfg->ifa_proto;
   1126	/* No need to add the TENTATIVE flag for addresses with NODAD */
   1127	if (!(cfg->ifa_flags & IFA_F_NODAD))
   1128		ifa->flags |= IFA_F_TENTATIVE;
   1129	ifa->valid_lft = cfg->valid_lft;
   1130	ifa->prefered_lft = cfg->preferred_lft;
   1131	ifa->cstamp = ifa->tstamp = jiffies;
   1132	ifa->tokenized = false;
   1133
   1134	ifa->rt = f6i;
   1135
   1136	ifa->idev = idev;
   1137	in6_dev_hold(idev);
   1138
   1139	/* For caller */
   1140	refcount_set(&ifa->refcnt, 1);
   1141
   1142	rcu_read_lock_bh();
   1143
   1144	err = ipv6_add_addr_hash(idev->dev, ifa);
   1145	if (err < 0) {
   1146		rcu_read_unlock_bh();
   1147		goto out;
   1148	}
   1149
   1150	write_lock(&idev->lock);
   1151
   1152	/* Add to inet6_dev unicast addr list. */
   1153	ipv6_link_dev_addr(idev, ifa);
   1154
   1155	if (ifa->flags&IFA_F_TEMPORARY) {
   1156		list_add(&ifa->tmp_list, &idev->tempaddr_list);
   1157		in6_ifa_hold(ifa);
   1158	}
   1159
   1160	in6_ifa_hold(ifa);
   1161	write_unlock(&idev->lock);
   1162
   1163	rcu_read_unlock_bh();
   1164
   1165	inet6addr_notifier_call_chain(NETDEV_UP, ifa);
   1166out:
   1167	if (unlikely(err < 0)) {
   1168		fib6_info_release(f6i);
   1169
   1170		if (ifa) {
   1171			if (ifa->idev)
   1172				in6_dev_put(ifa->idev);
   1173			kfree(ifa);
   1174		}
   1175		ifa = ERR_PTR(err);
   1176	}
   1177
   1178	return ifa;
   1179}
   1180
   1181enum cleanup_prefix_rt_t {
   1182	CLEANUP_PREFIX_RT_NOP,    /* no cleanup action for prefix route */
   1183	CLEANUP_PREFIX_RT_DEL,    /* delete the prefix route */
   1184	CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */
   1185};
   1186
   1187/*
   1188 * Check, whether the prefix for ifp would still need a prefix route
   1189 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_*
   1190 * constants.
   1191 *
   1192 * 1) we don't purge prefix if address was not permanent.
   1193 *    prefix is managed by its own lifetime.
   1194 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE.
   1195 * 3) if there are no addresses, delete prefix.
   1196 * 4) if there are still other permanent address(es),
   1197 *    corresponding prefix is still permanent.
   1198 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE,
   1199 *    don't purge the prefix, assume user space is managing it.
   1200 * 6) otherwise, update prefix lifetime to the
   1201 *    longest valid lifetime among the corresponding
   1202 *    addresses on the device.
   1203 *    Note: subsequent RA will update lifetime.
   1204 **/
   1205static enum cleanup_prefix_rt_t
   1206check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires)
   1207{
   1208	struct inet6_ifaddr *ifa;
   1209	struct inet6_dev *idev = ifp->idev;
   1210	unsigned long lifetime;
   1211	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL;
   1212
   1213	*expires = jiffies;
   1214
   1215	list_for_each_entry(ifa, &idev->addr_list, if_list) {
   1216		if (ifa == ifp)
   1217			continue;
   1218		if (ifa->prefix_len != ifp->prefix_len ||
   1219		    !ipv6_prefix_equal(&ifa->addr, &ifp->addr,
   1220				       ifp->prefix_len))
   1221			continue;
   1222		if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE))
   1223			return CLEANUP_PREFIX_RT_NOP;
   1224
   1225		action = CLEANUP_PREFIX_RT_EXPIRE;
   1226
   1227		spin_lock(&ifa->lock);
   1228
   1229		lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
   1230		/*
   1231		 * Note: Because this address is
   1232		 * not permanent, lifetime <
   1233		 * LONG_MAX / HZ here.
   1234		 */
   1235		if (time_before(*expires, ifa->tstamp + lifetime * HZ))
   1236			*expires = ifa->tstamp + lifetime * HZ;
   1237		spin_unlock(&ifa->lock);
   1238	}
   1239
   1240	return action;
   1241}
   1242
   1243static void
   1244cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires,
   1245		     bool del_rt, bool del_peer)
   1246{
   1247	struct fib6_info *f6i;
   1248
   1249	f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr,
   1250					ifp->prefix_len,
   1251					ifp->idev->dev, 0, RTF_DEFAULT, true);
   1252	if (f6i) {
   1253		if (del_rt)
   1254			ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
   1255		else {
   1256			if (!(f6i->fib6_flags & RTF_EXPIRES))
   1257				fib6_set_expires(f6i, expires);
   1258			fib6_info_release(f6i);
   1259		}
   1260	}
   1261}
   1262
   1263
   1264/* This function wants to get referenced ifp and releases it before return */
   1265
   1266static void ipv6_del_addr(struct inet6_ifaddr *ifp)
   1267{
   1268	enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP;
   1269	struct net *net = dev_net(ifp->idev->dev);
   1270	unsigned long expires;
   1271	int state;
   1272
   1273	ASSERT_RTNL();
   1274
   1275	spin_lock_bh(&ifp->lock);
   1276	state = ifp->state;
   1277	ifp->state = INET6_IFADDR_STATE_DEAD;
   1278	spin_unlock_bh(&ifp->lock);
   1279
   1280	if (state == INET6_IFADDR_STATE_DEAD)
   1281		goto out;
   1282
   1283	spin_lock_bh(&net->ipv6.addrconf_hash_lock);
   1284	hlist_del_init_rcu(&ifp->addr_lst);
   1285	spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
   1286
   1287	write_lock_bh(&ifp->idev->lock);
   1288
   1289	if (ifp->flags&IFA_F_TEMPORARY) {
   1290		list_del(&ifp->tmp_list);
   1291		if (ifp->ifpub) {
   1292			in6_ifa_put(ifp->ifpub);
   1293			ifp->ifpub = NULL;
   1294		}
   1295		__in6_ifa_put(ifp);
   1296	}
   1297
   1298	if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE))
   1299		action = check_cleanup_prefix_route(ifp, &expires);
   1300
   1301	list_del_rcu(&ifp->if_list);
   1302	__in6_ifa_put(ifp);
   1303
   1304	write_unlock_bh(&ifp->idev->lock);
   1305
   1306	addrconf_del_dad_work(ifp);
   1307
   1308	ipv6_ifa_notify(RTM_DELADDR, ifp);
   1309
   1310	inet6addr_notifier_call_chain(NETDEV_DOWN, ifp);
   1311
   1312	if (action != CLEANUP_PREFIX_RT_NOP) {
   1313		cleanup_prefix_route(ifp, expires,
   1314			action == CLEANUP_PREFIX_RT_DEL, false);
   1315	}
   1316
   1317	/* clean up prefsrc entries */
   1318	rt6_remove_prefsrc(ifp);
   1319out:
   1320	in6_ifa_put(ifp);
   1321}
   1322
   1323static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block)
   1324{
   1325	struct inet6_dev *idev = ifp->idev;
   1326	unsigned long tmp_tstamp, age;
   1327	unsigned long regen_advance;
   1328	unsigned long now = jiffies;
   1329	s32 cnf_temp_preferred_lft;
   1330	struct inet6_ifaddr *ift;
   1331	struct ifa6_config cfg;
   1332	long max_desync_factor;
   1333	struct in6_addr addr;
   1334	int ret = 0;
   1335
   1336	write_lock_bh(&idev->lock);
   1337
   1338retry:
   1339	in6_dev_hold(idev);
   1340	if (idev->cnf.use_tempaddr <= 0) {
   1341		write_unlock_bh(&idev->lock);
   1342		pr_info("%s: use_tempaddr is disabled\n", __func__);
   1343		in6_dev_put(idev);
   1344		ret = -1;
   1345		goto out;
   1346	}
   1347	spin_lock_bh(&ifp->lock);
   1348	if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
   1349		idev->cnf.use_tempaddr = -1;	/*XXX*/
   1350		spin_unlock_bh(&ifp->lock);
   1351		write_unlock_bh(&idev->lock);
   1352		pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
   1353			__func__);
   1354		in6_dev_put(idev);
   1355		ret = -1;
   1356		goto out;
   1357	}
   1358	in6_ifa_hold(ifp);
   1359	memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
   1360	ipv6_gen_rnd_iid(&addr);
   1361
   1362	age = (now - ifp->tstamp) / HZ;
   1363
   1364	regen_advance = idev->cnf.regen_max_retry *
   1365			idev->cnf.dad_transmits *
   1366			max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
   1367
   1368	/* recalculate max_desync_factor each time and update
   1369	 * idev->desync_factor if it's larger
   1370	 */
   1371	cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft);
   1372	max_desync_factor = min_t(__u32,
   1373				  idev->cnf.max_desync_factor,
   1374				  cnf_temp_preferred_lft - regen_advance);
   1375
   1376	if (unlikely(idev->desync_factor > max_desync_factor)) {
   1377		if (max_desync_factor > 0) {
   1378			get_random_bytes(&idev->desync_factor,
   1379					 sizeof(idev->desync_factor));
   1380			idev->desync_factor %= max_desync_factor;
   1381		} else {
   1382			idev->desync_factor = 0;
   1383		}
   1384	}
   1385
   1386	memset(&cfg, 0, sizeof(cfg));
   1387	cfg.valid_lft = min_t(__u32, ifp->valid_lft,
   1388			      idev->cnf.temp_valid_lft + age);
   1389	cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor;
   1390	cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft);
   1391
   1392	cfg.plen = ifp->prefix_len;
   1393	tmp_tstamp = ifp->tstamp;
   1394	spin_unlock_bh(&ifp->lock);
   1395
   1396	write_unlock_bh(&idev->lock);
   1397
   1398	/* A temporary address is created only if this calculated Preferred
   1399	 * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
   1400	 * an implementation must not create a temporary address with a zero
   1401	 * Preferred Lifetime.
   1402	 * Use age calculation as in addrconf_verify to avoid unnecessary
   1403	 * temporary addresses being generated.
   1404	 */
   1405	age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
   1406	if (cfg.preferred_lft <= regen_advance + age) {
   1407		in6_ifa_put(ifp);
   1408		in6_dev_put(idev);
   1409		ret = -1;
   1410		goto out;
   1411	}
   1412
   1413	cfg.ifa_flags = IFA_F_TEMPORARY;
   1414	/* set in addrconf_prefix_rcv() */
   1415	if (ifp->flags & IFA_F_OPTIMISTIC)
   1416		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
   1417
   1418	cfg.pfx = &addr;
   1419	cfg.scope = ipv6_addr_scope(cfg.pfx);
   1420
   1421	ift = ipv6_add_addr(idev, &cfg, block, NULL);
   1422	if (IS_ERR(ift)) {
   1423		in6_ifa_put(ifp);
   1424		in6_dev_put(idev);
   1425		pr_info("%s: retry temporary address regeneration\n", __func__);
   1426		write_lock_bh(&idev->lock);
   1427		goto retry;
   1428	}
   1429
   1430	spin_lock_bh(&ift->lock);
   1431	ift->ifpub = ifp;
   1432	ift->cstamp = now;
   1433	ift->tstamp = tmp_tstamp;
   1434	spin_unlock_bh(&ift->lock);
   1435
   1436	addrconf_dad_start(ift);
   1437	in6_ifa_put(ift);
   1438	in6_dev_put(idev);
   1439out:
   1440	return ret;
   1441}
   1442
   1443/*
   1444 *	Choose an appropriate source address (RFC3484)
   1445 */
   1446enum {
   1447	IPV6_SADDR_RULE_INIT = 0,
   1448	IPV6_SADDR_RULE_LOCAL,
   1449	IPV6_SADDR_RULE_SCOPE,
   1450	IPV6_SADDR_RULE_PREFERRED,
   1451#ifdef CONFIG_IPV6_MIP6
   1452	IPV6_SADDR_RULE_HOA,
   1453#endif
   1454	IPV6_SADDR_RULE_OIF,
   1455	IPV6_SADDR_RULE_LABEL,
   1456	IPV6_SADDR_RULE_PRIVACY,
   1457	IPV6_SADDR_RULE_ORCHID,
   1458	IPV6_SADDR_RULE_PREFIX,
   1459#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   1460	IPV6_SADDR_RULE_NOT_OPTIMISTIC,
   1461#endif
   1462	IPV6_SADDR_RULE_MAX
   1463};
   1464
   1465struct ipv6_saddr_score {
   1466	int			rule;
   1467	int			addr_type;
   1468	struct inet6_ifaddr	*ifa;
   1469	DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
   1470	int			scopedist;
   1471	int			matchlen;
   1472};
   1473
   1474struct ipv6_saddr_dst {
   1475	const struct in6_addr *addr;
   1476	int ifindex;
   1477	int scope;
   1478	int label;
   1479	unsigned int prefs;
   1480};
   1481
   1482static inline int ipv6_saddr_preferred(int type)
   1483{
   1484	if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
   1485		return 1;
   1486	return 0;
   1487}
   1488
   1489static bool ipv6_use_optimistic_addr(struct net *net,
   1490				     struct inet6_dev *idev)
   1491{
   1492#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   1493	if (!idev)
   1494		return false;
   1495	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
   1496		return false;
   1497	if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic)
   1498		return false;
   1499
   1500	return true;
   1501#else
   1502	return false;
   1503#endif
   1504}
   1505
   1506static bool ipv6_allow_optimistic_dad(struct net *net,
   1507				      struct inet6_dev *idev)
   1508{
   1509#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   1510	if (!idev)
   1511		return false;
   1512	if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad)
   1513		return false;
   1514
   1515	return true;
   1516#else
   1517	return false;
   1518#endif
   1519}
   1520
   1521static int ipv6_get_saddr_eval(struct net *net,
   1522			       struct ipv6_saddr_score *score,
   1523			       struct ipv6_saddr_dst *dst,
   1524			       int i)
   1525{
   1526	int ret;
   1527
   1528	if (i <= score->rule) {
   1529		switch (i) {
   1530		case IPV6_SADDR_RULE_SCOPE:
   1531			ret = score->scopedist;
   1532			break;
   1533		case IPV6_SADDR_RULE_PREFIX:
   1534			ret = score->matchlen;
   1535			break;
   1536		default:
   1537			ret = !!test_bit(i, score->scorebits);
   1538		}
   1539		goto out;
   1540	}
   1541
   1542	switch (i) {
   1543	case IPV6_SADDR_RULE_INIT:
   1544		/* Rule 0: remember if hiscore is not ready yet */
   1545		ret = !!score->ifa;
   1546		break;
   1547	case IPV6_SADDR_RULE_LOCAL:
   1548		/* Rule 1: Prefer same address */
   1549		ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
   1550		break;
   1551	case IPV6_SADDR_RULE_SCOPE:
   1552		/* Rule 2: Prefer appropriate scope
   1553		 *
   1554		 *      ret
   1555		 *       ^
   1556		 *    -1 |  d 15
   1557		 *    ---+--+-+---> scope
   1558		 *       |
   1559		 *       |             d is scope of the destination.
   1560		 *  B-d  |  \
   1561		 *       |   \      <- smaller scope is better if
   1562		 *  B-15 |    \        if scope is enough for destination.
   1563		 *       |             ret = B - scope (-1 <= scope >= d <= 15).
   1564		 * d-C-1 | /
   1565		 *       |/         <- greater is better
   1566		 *   -C  /             if scope is not enough for destination.
   1567		 *      /|             ret = scope - C (-1 <= d < scope <= 15).
   1568		 *
   1569		 * d - C - 1 < B -15 (for all -1 <= d <= 15).
   1570		 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
   1571		 * Assume B = 0 and we get C > 29.
   1572		 */
   1573		ret = __ipv6_addr_src_scope(score->addr_type);
   1574		if (ret >= dst->scope)
   1575			ret = -ret;
   1576		else
   1577			ret -= 128;	/* 30 is enough */
   1578		score->scopedist = ret;
   1579		break;
   1580	case IPV6_SADDR_RULE_PREFERRED:
   1581	    {
   1582		/* Rule 3: Avoid deprecated and optimistic addresses */
   1583		u8 avoid = IFA_F_DEPRECATED;
   1584
   1585		if (!ipv6_use_optimistic_addr(net, score->ifa->idev))
   1586			avoid |= IFA_F_OPTIMISTIC;
   1587		ret = ipv6_saddr_preferred(score->addr_type) ||
   1588		      !(score->ifa->flags & avoid);
   1589		break;
   1590	    }
   1591#ifdef CONFIG_IPV6_MIP6
   1592	case IPV6_SADDR_RULE_HOA:
   1593	    {
   1594		/* Rule 4: Prefer home address */
   1595		int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
   1596		ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
   1597		break;
   1598	    }
   1599#endif
   1600	case IPV6_SADDR_RULE_OIF:
   1601		/* Rule 5: Prefer outgoing interface */
   1602		ret = (!dst->ifindex ||
   1603		       dst->ifindex == score->ifa->idev->dev->ifindex);
   1604		break;
   1605	case IPV6_SADDR_RULE_LABEL:
   1606		/* Rule 6: Prefer matching label */
   1607		ret = ipv6_addr_label(net,
   1608				      &score->ifa->addr, score->addr_type,
   1609				      score->ifa->idev->dev->ifindex) == dst->label;
   1610		break;
   1611	case IPV6_SADDR_RULE_PRIVACY:
   1612	    {
   1613		/* Rule 7: Prefer public address
   1614		 * Note: prefer temporary address if use_tempaddr >= 2
   1615		 */
   1616		int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
   1617				!!(dst->prefs & IPV6_PREFER_SRC_TMP) :
   1618				score->ifa->idev->cnf.use_tempaddr >= 2;
   1619		ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
   1620		break;
   1621	    }
   1622	case IPV6_SADDR_RULE_ORCHID:
   1623		/* Rule 8-: Prefer ORCHID vs ORCHID or
   1624		 *	    non-ORCHID vs non-ORCHID
   1625		 */
   1626		ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
   1627			ipv6_addr_orchid(dst->addr));
   1628		break;
   1629	case IPV6_SADDR_RULE_PREFIX:
   1630		/* Rule 8: Use longest matching prefix */
   1631		ret = ipv6_addr_diff(&score->ifa->addr, dst->addr);
   1632		if (ret > score->ifa->prefix_len)
   1633			ret = score->ifa->prefix_len;
   1634		score->matchlen = ret;
   1635		break;
   1636#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   1637	case IPV6_SADDR_RULE_NOT_OPTIMISTIC:
   1638		/* Optimistic addresses still have lower precedence than other
   1639		 * preferred addresses.
   1640		 */
   1641		ret = !(score->ifa->flags & IFA_F_OPTIMISTIC);
   1642		break;
   1643#endif
   1644	default:
   1645		ret = 0;
   1646	}
   1647
   1648	if (ret)
   1649		__set_bit(i, score->scorebits);
   1650	score->rule = i;
   1651out:
   1652	return ret;
   1653}
   1654
   1655static int __ipv6_dev_get_saddr(struct net *net,
   1656				struct ipv6_saddr_dst *dst,
   1657				struct inet6_dev *idev,
   1658				struct ipv6_saddr_score *scores,
   1659				int hiscore_idx)
   1660{
   1661	struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx];
   1662
   1663	list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) {
   1664		int i;
   1665
   1666		/*
   1667		 * - Tentative Address (RFC2462 section 5.4)
   1668		 *  - A tentative address is not considered
   1669		 *    "assigned to an interface" in the traditional
   1670		 *    sense, unless it is also flagged as optimistic.
   1671		 * - Candidate Source Address (section 4)
   1672		 *  - In any case, anycast addresses, multicast
   1673		 *    addresses, and the unspecified address MUST
   1674		 *    NOT be included in a candidate set.
   1675		 */
   1676		if ((score->ifa->flags & IFA_F_TENTATIVE) &&
   1677		    (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
   1678			continue;
   1679
   1680		score->addr_type = __ipv6_addr_type(&score->ifa->addr);
   1681
   1682		if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
   1683			     score->addr_type & IPV6_ADDR_MULTICAST)) {
   1684			net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s",
   1685					    idev->dev->name);
   1686			continue;
   1687		}
   1688
   1689		score->rule = -1;
   1690		bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
   1691
   1692		for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
   1693			int minihiscore, miniscore;
   1694
   1695			minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i);
   1696			miniscore = ipv6_get_saddr_eval(net, score, dst, i);
   1697
   1698			if (minihiscore > miniscore) {
   1699				if (i == IPV6_SADDR_RULE_SCOPE &&
   1700				    score->scopedist > 0) {
   1701					/*
   1702					 * special case:
   1703					 * each remaining entry
   1704					 * has too small (not enough)
   1705					 * scope, because ifa entries
   1706					 * are sorted by their scope
   1707					 * values.
   1708					 */
   1709					goto out;
   1710				}
   1711				break;
   1712			} else if (minihiscore < miniscore) {
   1713				swap(hiscore, score);
   1714				hiscore_idx = 1 - hiscore_idx;
   1715
   1716				/* restore our iterator */
   1717				score->ifa = hiscore->ifa;
   1718
   1719				break;
   1720			}
   1721		}
   1722	}
   1723out:
   1724	return hiscore_idx;
   1725}
   1726
   1727static int ipv6_get_saddr_master(struct net *net,
   1728				 const struct net_device *dst_dev,
   1729				 const struct net_device *master,
   1730				 struct ipv6_saddr_dst *dst,
   1731				 struct ipv6_saddr_score *scores,
   1732				 int hiscore_idx)
   1733{
   1734	struct inet6_dev *idev;
   1735
   1736	idev = __in6_dev_get(dst_dev);
   1737	if (idev)
   1738		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
   1739						   scores, hiscore_idx);
   1740
   1741	idev = __in6_dev_get(master);
   1742	if (idev)
   1743		hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev,
   1744						   scores, hiscore_idx);
   1745
   1746	return hiscore_idx;
   1747}
   1748
   1749int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev,
   1750		       const struct in6_addr *daddr, unsigned int prefs,
   1751		       struct in6_addr *saddr)
   1752{
   1753	struct ipv6_saddr_score scores[2], *hiscore;
   1754	struct ipv6_saddr_dst dst;
   1755	struct inet6_dev *idev;
   1756	struct net_device *dev;
   1757	int dst_type;
   1758	bool use_oif_addr = false;
   1759	int hiscore_idx = 0;
   1760	int ret = 0;
   1761
   1762	dst_type = __ipv6_addr_type(daddr);
   1763	dst.addr = daddr;
   1764	dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
   1765	dst.scope = __ipv6_addr_src_scope(dst_type);
   1766	dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
   1767	dst.prefs = prefs;
   1768
   1769	scores[hiscore_idx].rule = -1;
   1770	scores[hiscore_idx].ifa = NULL;
   1771
   1772	rcu_read_lock();
   1773
   1774	/* Candidate Source Address (section 4)
   1775	 *  - multicast and link-local destination address,
   1776	 *    the set of candidate source address MUST only
   1777	 *    include addresses assigned to interfaces
   1778	 *    belonging to the same link as the outgoing
   1779	 *    interface.
   1780	 * (- For site-local destination addresses, the
   1781	 *    set of candidate source addresses MUST only
   1782	 *    include addresses assigned to interfaces
   1783	 *    belonging to the same site as the outgoing
   1784	 *    interface.)
   1785	 *  - "It is RECOMMENDED that the candidate source addresses
   1786	 *    be the set of unicast addresses assigned to the
   1787	 *    interface that will be used to send to the destination
   1788	 *    (the 'outgoing' interface)." (RFC 6724)
   1789	 */
   1790	if (dst_dev) {
   1791		idev = __in6_dev_get(dst_dev);
   1792		if ((dst_type & IPV6_ADDR_MULTICAST) ||
   1793		    dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL ||
   1794		    (idev && idev->cnf.use_oif_addrs_only)) {
   1795			use_oif_addr = true;
   1796		}
   1797	}
   1798
   1799	if (use_oif_addr) {
   1800		if (idev)
   1801			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
   1802	} else {
   1803		const struct net_device *master;
   1804		int master_idx = 0;
   1805
   1806		/* if dst_dev exists and is enslaved to an L3 device, then
   1807		 * prefer addresses from dst_dev and then the master over
   1808		 * any other enslaved devices in the L3 domain.
   1809		 */
   1810		master = l3mdev_master_dev_rcu(dst_dev);
   1811		if (master) {
   1812			master_idx = master->ifindex;
   1813
   1814			hiscore_idx = ipv6_get_saddr_master(net, dst_dev,
   1815							    master, &dst,
   1816							    scores, hiscore_idx);
   1817
   1818			if (scores[hiscore_idx].ifa)
   1819				goto out;
   1820		}
   1821
   1822		for_each_netdev_rcu(net, dev) {
   1823			/* only consider addresses on devices in the
   1824			 * same L3 domain
   1825			 */
   1826			if (l3mdev_master_ifindex_rcu(dev) != master_idx)
   1827				continue;
   1828			idev = __in6_dev_get(dev);
   1829			if (!idev)
   1830				continue;
   1831			hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx);
   1832		}
   1833	}
   1834
   1835out:
   1836	hiscore = &scores[hiscore_idx];
   1837	if (!hiscore->ifa)
   1838		ret = -EADDRNOTAVAIL;
   1839	else
   1840		*saddr = hiscore->ifa->addr;
   1841
   1842	rcu_read_unlock();
   1843	return ret;
   1844}
   1845EXPORT_SYMBOL(ipv6_dev_get_saddr);
   1846
   1847static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
   1848			      u32 banned_flags)
   1849{
   1850	struct inet6_ifaddr *ifp;
   1851	int err = -EADDRNOTAVAIL;
   1852
   1853	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
   1854		if (ifp->scope > IFA_LINK)
   1855			break;
   1856		if (ifp->scope == IFA_LINK &&
   1857		    !(ifp->flags & banned_flags)) {
   1858			*addr = ifp->addr;
   1859			err = 0;
   1860			break;
   1861		}
   1862	}
   1863	return err;
   1864}
   1865
   1866int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
   1867		    u32 banned_flags)
   1868{
   1869	struct inet6_dev *idev;
   1870	int err = -EADDRNOTAVAIL;
   1871
   1872	rcu_read_lock();
   1873	idev = __in6_dev_get(dev);
   1874	if (idev) {
   1875		read_lock_bh(&idev->lock);
   1876		err = __ipv6_get_lladdr(idev, addr, banned_flags);
   1877		read_unlock_bh(&idev->lock);
   1878	}
   1879	rcu_read_unlock();
   1880	return err;
   1881}
   1882
   1883static int ipv6_count_addresses(const struct inet6_dev *idev)
   1884{
   1885	const struct inet6_ifaddr *ifp;
   1886	int cnt = 0;
   1887
   1888	rcu_read_lock();
   1889	list_for_each_entry_rcu(ifp, &idev->addr_list, if_list)
   1890		cnt++;
   1891	rcu_read_unlock();
   1892	return cnt;
   1893}
   1894
   1895int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
   1896		  const struct net_device *dev, int strict)
   1897{
   1898	return ipv6_chk_addr_and_flags(net, addr, dev, !dev,
   1899				       strict, IFA_F_TENTATIVE);
   1900}
   1901EXPORT_SYMBOL(ipv6_chk_addr);
   1902
   1903/* device argument is used to find the L3 domain of interest. If
   1904 * skip_dev_check is set, then the ifp device is not checked against
   1905 * the passed in dev argument. So the 2 cases for addresses checks are:
   1906 *   1. does the address exist in the L3 domain that dev is part of
   1907 *      (skip_dev_check = true), or
   1908 *
   1909 *   2. does the address exist on the specific device
   1910 *      (skip_dev_check = false)
   1911 */
   1912static struct net_device *
   1913__ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
   1914			  const struct net_device *dev, bool skip_dev_check,
   1915			  int strict, u32 banned_flags)
   1916{
   1917	unsigned int hash = inet6_addr_hash(net, addr);
   1918	struct net_device *l3mdev, *ndev;
   1919	struct inet6_ifaddr *ifp;
   1920	u32 ifp_flags;
   1921
   1922	rcu_read_lock();
   1923
   1924	l3mdev = l3mdev_master_dev_rcu(dev);
   1925	if (skip_dev_check)
   1926		dev = NULL;
   1927
   1928	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
   1929		ndev = ifp->idev->dev;
   1930
   1931		if (l3mdev_master_dev_rcu(ndev) != l3mdev)
   1932			continue;
   1933
   1934		/* Decouple optimistic from tentative for evaluation here.
   1935		 * Ban optimistic addresses explicitly, when required.
   1936		 */
   1937		ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC)
   1938			    ? (ifp->flags&~IFA_F_TENTATIVE)
   1939			    : ifp->flags;
   1940		if (ipv6_addr_equal(&ifp->addr, addr) &&
   1941		    !(ifp_flags&banned_flags) &&
   1942		    (!dev || ndev == dev ||
   1943		     !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
   1944			rcu_read_unlock();
   1945			return ndev;
   1946		}
   1947	}
   1948
   1949	rcu_read_unlock();
   1950	return NULL;
   1951}
   1952
   1953int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr,
   1954			    const struct net_device *dev, bool skip_dev_check,
   1955			    int strict, u32 banned_flags)
   1956{
   1957	return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check,
   1958					 strict, banned_flags) ? 1 : 0;
   1959}
   1960EXPORT_SYMBOL(ipv6_chk_addr_and_flags);
   1961
   1962
   1963/* Compares an address/prefix_len with addresses on device @dev.
   1964 * If one is found it returns true.
   1965 */
   1966bool ipv6_chk_custom_prefix(const struct in6_addr *addr,
   1967	const unsigned int prefix_len, struct net_device *dev)
   1968{
   1969	const struct inet6_ifaddr *ifa;
   1970	const struct inet6_dev *idev;
   1971	bool ret = false;
   1972
   1973	rcu_read_lock();
   1974	idev = __in6_dev_get(dev);
   1975	if (idev) {
   1976		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
   1977			ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len);
   1978			if (ret)
   1979				break;
   1980		}
   1981	}
   1982	rcu_read_unlock();
   1983
   1984	return ret;
   1985}
   1986EXPORT_SYMBOL(ipv6_chk_custom_prefix);
   1987
   1988int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
   1989{
   1990	const struct inet6_ifaddr *ifa;
   1991	const struct inet6_dev *idev;
   1992	int	onlink;
   1993
   1994	onlink = 0;
   1995	rcu_read_lock();
   1996	idev = __in6_dev_get(dev);
   1997	if (idev) {
   1998		list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) {
   1999			onlink = ipv6_prefix_equal(addr, &ifa->addr,
   2000						   ifa->prefix_len);
   2001			if (onlink)
   2002				break;
   2003		}
   2004	}
   2005	rcu_read_unlock();
   2006	return onlink;
   2007}
   2008EXPORT_SYMBOL(ipv6_chk_prefix);
   2009
   2010/**
   2011 * ipv6_dev_find - find the first device with a given source address.
   2012 * @net: the net namespace
   2013 * @addr: the source address
   2014 * @dev: used to find the L3 domain of interest
   2015 *
   2016 * The caller should be protected by RCU, or RTNL.
   2017 */
   2018struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr,
   2019				 struct net_device *dev)
   2020{
   2021	return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1,
   2022					 IFA_F_TENTATIVE);
   2023}
   2024EXPORT_SYMBOL(ipv6_dev_find);
   2025
   2026struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
   2027				     struct net_device *dev, int strict)
   2028{
   2029	unsigned int hash = inet6_addr_hash(net, addr);
   2030	struct inet6_ifaddr *ifp, *result = NULL;
   2031
   2032	rcu_read_lock();
   2033	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
   2034		if (ipv6_addr_equal(&ifp->addr, addr)) {
   2035			if (!dev || ifp->idev->dev == dev ||
   2036			    !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
   2037				result = ifp;
   2038				in6_ifa_hold(ifp);
   2039				break;
   2040			}
   2041		}
   2042	}
   2043	rcu_read_unlock();
   2044
   2045	return result;
   2046}
   2047
   2048/* Gets referenced address, destroys ifaddr */
   2049
   2050static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
   2051{
   2052	if (dad_failed)
   2053		ifp->flags |= IFA_F_DADFAILED;
   2054
   2055	if (ifp->flags&IFA_F_TEMPORARY) {
   2056		struct inet6_ifaddr *ifpub;
   2057		spin_lock_bh(&ifp->lock);
   2058		ifpub = ifp->ifpub;
   2059		if (ifpub) {
   2060			in6_ifa_hold(ifpub);
   2061			spin_unlock_bh(&ifp->lock);
   2062			ipv6_create_tempaddr(ifpub, true);
   2063			in6_ifa_put(ifpub);
   2064		} else {
   2065			spin_unlock_bh(&ifp->lock);
   2066		}
   2067		ipv6_del_addr(ifp);
   2068	} else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) {
   2069		spin_lock_bh(&ifp->lock);
   2070		addrconf_del_dad_work(ifp);
   2071		ifp->flags |= IFA_F_TENTATIVE;
   2072		if (dad_failed)
   2073			ifp->flags &= ~IFA_F_OPTIMISTIC;
   2074		spin_unlock_bh(&ifp->lock);
   2075		if (dad_failed)
   2076			ipv6_ifa_notify(0, ifp);
   2077		in6_ifa_put(ifp);
   2078	} else {
   2079		ipv6_del_addr(ifp);
   2080	}
   2081}
   2082
   2083static int addrconf_dad_end(struct inet6_ifaddr *ifp)
   2084{
   2085	int err = -ENOENT;
   2086
   2087	spin_lock_bh(&ifp->lock);
   2088	if (ifp->state == INET6_IFADDR_STATE_DAD) {
   2089		ifp->state = INET6_IFADDR_STATE_POSTDAD;
   2090		err = 0;
   2091	}
   2092	spin_unlock_bh(&ifp->lock);
   2093
   2094	return err;
   2095}
   2096
   2097void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp)
   2098{
   2099	struct inet6_dev *idev = ifp->idev;
   2100	struct net *net = dev_net(idev->dev);
   2101
   2102	if (addrconf_dad_end(ifp)) {
   2103		in6_ifa_put(ifp);
   2104		return;
   2105	}
   2106
   2107	net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n",
   2108			     ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source);
   2109
   2110	spin_lock_bh(&ifp->lock);
   2111
   2112	if (ifp->flags & IFA_F_STABLE_PRIVACY) {
   2113		struct in6_addr new_addr;
   2114		struct inet6_ifaddr *ifp2;
   2115		int retries = ifp->stable_privacy_retry + 1;
   2116		struct ifa6_config cfg = {
   2117			.pfx = &new_addr,
   2118			.plen = ifp->prefix_len,
   2119			.ifa_flags = ifp->flags,
   2120			.valid_lft = ifp->valid_lft,
   2121			.preferred_lft = ifp->prefered_lft,
   2122			.scope = ifp->scope,
   2123		};
   2124
   2125		if (retries > net->ipv6.sysctl.idgen_retries) {
   2126			net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n",
   2127					     ifp->idev->dev->name);
   2128			goto errdad;
   2129		}
   2130
   2131		new_addr = ifp->addr;
   2132		if (ipv6_generate_stable_address(&new_addr, retries,
   2133						 idev))
   2134			goto errdad;
   2135
   2136		spin_unlock_bh(&ifp->lock);
   2137
   2138		if (idev->cnf.max_addresses &&
   2139		    ipv6_count_addresses(idev) >=
   2140		    idev->cnf.max_addresses)
   2141			goto lock_errdad;
   2142
   2143		net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n",
   2144				     ifp->idev->dev->name);
   2145
   2146		ifp2 = ipv6_add_addr(idev, &cfg, false, NULL);
   2147		if (IS_ERR(ifp2))
   2148			goto lock_errdad;
   2149
   2150		spin_lock_bh(&ifp2->lock);
   2151		ifp2->stable_privacy_retry = retries;
   2152		ifp2->state = INET6_IFADDR_STATE_PREDAD;
   2153		spin_unlock_bh(&ifp2->lock);
   2154
   2155		addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay);
   2156		in6_ifa_put(ifp2);
   2157lock_errdad:
   2158		spin_lock_bh(&ifp->lock);
   2159	}
   2160
   2161errdad:
   2162	/* transition from _POSTDAD to _ERRDAD */
   2163	ifp->state = INET6_IFADDR_STATE_ERRDAD;
   2164	spin_unlock_bh(&ifp->lock);
   2165
   2166	addrconf_mod_dad_work(ifp, 0);
   2167	in6_ifa_put(ifp);
   2168}
   2169
   2170/* Join to solicited addr multicast group.
   2171 * caller must hold RTNL */
   2172void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
   2173{
   2174	struct in6_addr maddr;
   2175
   2176	if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
   2177		return;
   2178
   2179	addrconf_addr_solict_mult(addr, &maddr);
   2180	ipv6_dev_mc_inc(dev, &maddr);
   2181}
   2182
   2183/* caller must hold RTNL */
   2184void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
   2185{
   2186	struct in6_addr maddr;
   2187
   2188	if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
   2189		return;
   2190
   2191	addrconf_addr_solict_mult(addr, &maddr);
   2192	__ipv6_dev_mc_dec(idev, &maddr);
   2193}
   2194
   2195/* caller must hold RTNL */
   2196static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
   2197{
   2198	struct in6_addr addr;
   2199
   2200	if (ifp->prefix_len >= 127) /* RFC 6164 */
   2201		return;
   2202	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
   2203	if (ipv6_addr_any(&addr))
   2204		return;
   2205	__ipv6_dev_ac_inc(ifp->idev, &addr);
   2206}
   2207
   2208/* caller must hold RTNL */
   2209static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
   2210{
   2211	struct in6_addr addr;
   2212
   2213	if (ifp->prefix_len >= 127) /* RFC 6164 */
   2214		return;
   2215	ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
   2216	if (ipv6_addr_any(&addr))
   2217		return;
   2218	__ipv6_dev_ac_dec(ifp->idev, &addr);
   2219}
   2220
   2221static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev)
   2222{
   2223	switch (dev->addr_len) {
   2224	case ETH_ALEN:
   2225		memcpy(eui, dev->dev_addr, 3);
   2226		eui[3] = 0xFF;
   2227		eui[4] = 0xFE;
   2228		memcpy(eui + 5, dev->dev_addr + 3, 3);
   2229		break;
   2230	case EUI64_ADDR_LEN:
   2231		memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN);
   2232		eui[0] ^= 2;
   2233		break;
   2234	default:
   2235		return -1;
   2236	}
   2237
   2238	return 0;
   2239}
   2240
   2241static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev)
   2242{
   2243	const union fwnet_hwaddr *ha;
   2244
   2245	if (dev->addr_len != FWNET_ALEN)
   2246		return -1;
   2247
   2248	ha = (const union fwnet_hwaddr *)dev->dev_addr;
   2249
   2250	memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id));
   2251	eui[0] ^= 2;
   2252	return 0;
   2253}
   2254
   2255static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
   2256{
   2257	/* XXX: inherit EUI-64 from other interface -- yoshfuji */
   2258	if (dev->addr_len != ARCNET_ALEN)
   2259		return -1;
   2260	memset(eui, 0, 7);
   2261	eui[7] = *(u8 *)dev->dev_addr;
   2262	return 0;
   2263}
   2264
   2265static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
   2266{
   2267	if (dev->addr_len != INFINIBAND_ALEN)
   2268		return -1;
   2269	memcpy(eui, dev->dev_addr + 12, 8);
   2270	eui[0] |= 2;
   2271	return 0;
   2272}
   2273
   2274static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
   2275{
   2276	if (addr == 0)
   2277		return -1;
   2278	eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
   2279		  ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
   2280		  ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
   2281		  ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
   2282		  ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
   2283		  ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
   2284	eui[1] = 0;
   2285	eui[2] = 0x5E;
   2286	eui[3] = 0xFE;
   2287	memcpy(eui + 4, &addr, 4);
   2288	return 0;
   2289}
   2290
   2291static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
   2292{
   2293	if (dev->priv_flags & IFF_ISATAP)
   2294		return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
   2295	return -1;
   2296}
   2297
   2298static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
   2299{
   2300	return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
   2301}
   2302
   2303static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev)
   2304{
   2305	memcpy(eui, dev->perm_addr, 3);
   2306	memcpy(eui + 5, dev->perm_addr + 3, 3);
   2307	eui[3] = 0xFF;
   2308	eui[4] = 0xFE;
   2309	eui[0] ^= 2;
   2310	return 0;
   2311}
   2312
   2313static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
   2314{
   2315	switch (dev->type) {
   2316	case ARPHRD_ETHER:
   2317	case ARPHRD_FDDI:
   2318		return addrconf_ifid_eui48(eui, dev);
   2319	case ARPHRD_ARCNET:
   2320		return addrconf_ifid_arcnet(eui, dev);
   2321	case ARPHRD_INFINIBAND:
   2322		return addrconf_ifid_infiniband(eui, dev);
   2323	case ARPHRD_SIT:
   2324		return addrconf_ifid_sit(eui, dev);
   2325	case ARPHRD_IPGRE:
   2326	case ARPHRD_TUNNEL:
   2327		return addrconf_ifid_gre(eui, dev);
   2328	case ARPHRD_6LOWPAN:
   2329		return addrconf_ifid_6lowpan(eui, dev);
   2330	case ARPHRD_IEEE1394:
   2331		return addrconf_ifid_ieee1394(eui, dev);
   2332	case ARPHRD_TUNNEL6:
   2333	case ARPHRD_IP6GRE:
   2334	case ARPHRD_RAWIP:
   2335		return addrconf_ifid_ip6tnl(eui, dev);
   2336	}
   2337	return -1;
   2338}
   2339
   2340static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
   2341{
   2342	int err = -1;
   2343	struct inet6_ifaddr *ifp;
   2344
   2345	read_lock_bh(&idev->lock);
   2346	list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) {
   2347		if (ifp->scope > IFA_LINK)
   2348			break;
   2349		if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
   2350			memcpy(eui, ifp->addr.s6_addr+8, 8);
   2351			err = 0;
   2352			break;
   2353		}
   2354	}
   2355	read_unlock_bh(&idev->lock);
   2356	return err;
   2357}
   2358
   2359/* Generation of a randomized Interface Identifier
   2360 * draft-ietf-6man-rfc4941bis, Section 3.3.1
   2361 */
   2362
   2363static void ipv6_gen_rnd_iid(struct in6_addr *addr)
   2364{
   2365regen:
   2366	get_random_bytes(&addr->s6_addr[8], 8);
   2367
   2368	/* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1:
   2369	 * check if generated address is not inappropriate:
   2370	 *
   2371	 * - Reserved IPv6 Interface Identifiers
   2372	 * - XXX: already assigned to an address on the device
   2373	 */
   2374
   2375	/* Subnet-router anycast: 0000:0000:0000:0000 */
   2376	if (!(addr->s6_addr32[2] | addr->s6_addr32[3]))
   2377		goto regen;
   2378
   2379	/* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212
   2380	 * Proxy Mobile IPv6:   0200:5EFF:FE00:5213
   2381	 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF
   2382	 */
   2383	if (ntohl(addr->s6_addr32[2]) == 0x02005eff &&
   2384	    (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000)
   2385		goto regen;
   2386
   2387	/* Reserved subnet anycast addresses */
   2388	if (ntohl(addr->s6_addr32[2]) == 0xfdffffff &&
   2389	    ntohl(addr->s6_addr32[3]) >= 0Xffffff80)
   2390		goto regen;
   2391}
   2392
   2393/*
   2394 *	Add prefix route.
   2395 */
   2396
   2397static void
   2398addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric,
   2399		      struct net_device *dev, unsigned long expires,
   2400		      u32 flags, gfp_t gfp_flags)
   2401{
   2402	struct fib6_config cfg = {
   2403		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX,
   2404		.fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF,
   2405		.fc_ifindex = dev->ifindex,
   2406		.fc_expires = expires,
   2407		.fc_dst_len = plen,
   2408		.fc_flags = RTF_UP | flags,
   2409		.fc_nlinfo.nl_net = dev_net(dev),
   2410		.fc_protocol = RTPROT_KERNEL,
   2411		.fc_type = RTN_UNICAST,
   2412	};
   2413
   2414	cfg.fc_dst = *pfx;
   2415
   2416	/* Prevent useless cloning on PtP SIT.
   2417	   This thing is done here expecting that the whole
   2418	   class of non-broadcast devices need not cloning.
   2419	 */
   2420#if IS_ENABLED(CONFIG_IPV6_SIT)
   2421	if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
   2422		cfg.fc_flags |= RTF_NONEXTHOP;
   2423#endif
   2424
   2425	ip6_route_add(&cfg, gfp_flags, NULL);
   2426}
   2427
   2428
   2429static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
   2430						  int plen,
   2431						  const struct net_device *dev,
   2432						  u32 flags, u32 noflags,
   2433						  bool no_gw)
   2434{
   2435	struct fib6_node *fn;
   2436	struct fib6_info *rt = NULL;
   2437	struct fib6_table *table;
   2438	u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX;
   2439
   2440	table = fib6_get_table(dev_net(dev), tb_id);
   2441	if (!table)
   2442		return NULL;
   2443
   2444	rcu_read_lock();
   2445	fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true);
   2446	if (!fn)
   2447		goto out;
   2448
   2449	for_each_fib6_node_rt_rcu(fn) {
   2450		/* prefix routes only use builtin fib6_nh */
   2451		if (rt->nh)
   2452			continue;
   2453
   2454		if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex)
   2455			continue;
   2456		if (no_gw && rt->fib6_nh->fib_nh_gw_family)
   2457			continue;
   2458		if ((rt->fib6_flags & flags) != flags)
   2459			continue;
   2460		if ((rt->fib6_flags & noflags) != 0)
   2461			continue;
   2462		if (!fib6_info_hold_safe(rt))
   2463			continue;
   2464		break;
   2465	}
   2466out:
   2467	rcu_read_unlock();
   2468	return rt;
   2469}
   2470
   2471
   2472/* Create "default" multicast route to the interface */
   2473
   2474static void addrconf_add_mroute(struct net_device *dev)
   2475{
   2476	struct fib6_config cfg = {
   2477		.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL,
   2478		.fc_metric = IP6_RT_PRIO_ADDRCONF,
   2479		.fc_ifindex = dev->ifindex,
   2480		.fc_dst_len = 8,
   2481		.fc_flags = RTF_UP,
   2482		.fc_type = RTN_MULTICAST,
   2483		.fc_nlinfo.nl_net = dev_net(dev),
   2484		.fc_protocol = RTPROT_KERNEL,
   2485	};
   2486
   2487	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
   2488
   2489	ip6_route_add(&cfg, GFP_KERNEL, NULL);
   2490}
   2491
   2492static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
   2493{
   2494	struct inet6_dev *idev;
   2495
   2496	ASSERT_RTNL();
   2497
   2498	idev = ipv6_find_idev(dev);
   2499	if (IS_ERR(idev))
   2500		return idev;
   2501
   2502	if (idev->cnf.disable_ipv6)
   2503		return ERR_PTR(-EACCES);
   2504
   2505	/* Add default multicast route */
   2506	if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev))
   2507		addrconf_add_mroute(dev);
   2508
   2509	return idev;
   2510}
   2511
   2512static void manage_tempaddrs(struct inet6_dev *idev,
   2513			     struct inet6_ifaddr *ifp,
   2514			     __u32 valid_lft, __u32 prefered_lft,
   2515			     bool create, unsigned long now)
   2516{
   2517	u32 flags;
   2518	struct inet6_ifaddr *ift;
   2519
   2520	read_lock_bh(&idev->lock);
   2521	/* update all temporary addresses in the list */
   2522	list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) {
   2523		int age, max_valid, max_prefered;
   2524
   2525		if (ifp != ift->ifpub)
   2526			continue;
   2527
   2528		/* RFC 4941 section 3.3:
   2529		 * If a received option will extend the lifetime of a public
   2530		 * address, the lifetimes of temporary addresses should
   2531		 * be extended, subject to the overall constraint that no
   2532		 * temporary addresses should ever remain "valid" or "preferred"
   2533		 * for a time longer than (TEMP_VALID_LIFETIME) or
   2534		 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively.
   2535		 */
   2536		age = (now - ift->cstamp) / HZ;
   2537		max_valid = idev->cnf.temp_valid_lft - age;
   2538		if (max_valid < 0)
   2539			max_valid = 0;
   2540
   2541		max_prefered = idev->cnf.temp_prefered_lft -
   2542			       idev->desync_factor - age;
   2543		if (max_prefered < 0)
   2544			max_prefered = 0;
   2545
   2546		if (valid_lft > max_valid)
   2547			valid_lft = max_valid;
   2548
   2549		if (prefered_lft > max_prefered)
   2550			prefered_lft = max_prefered;
   2551
   2552		spin_lock(&ift->lock);
   2553		flags = ift->flags;
   2554		ift->valid_lft = valid_lft;
   2555		ift->prefered_lft = prefered_lft;
   2556		ift->tstamp = now;
   2557		if (prefered_lft > 0)
   2558			ift->flags &= ~IFA_F_DEPRECATED;
   2559
   2560		spin_unlock(&ift->lock);
   2561		if (!(flags&IFA_F_TENTATIVE))
   2562			ipv6_ifa_notify(0, ift);
   2563	}
   2564
   2565	if ((create || list_empty(&idev->tempaddr_list)) &&
   2566	    idev->cnf.use_tempaddr > 0) {
   2567		/* When a new public address is created as described
   2568		 * in [ADDRCONF], also create a new temporary address.
   2569		 * Also create a temporary address if it's enabled but
   2570		 * no temporary address currently exists.
   2571		 */
   2572		read_unlock_bh(&idev->lock);
   2573		ipv6_create_tempaddr(ifp, false);
   2574	} else {
   2575		read_unlock_bh(&idev->lock);
   2576	}
   2577}
   2578
   2579static bool is_addr_mode_generate_stable(struct inet6_dev *idev)
   2580{
   2581	return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY ||
   2582	       idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM;
   2583}
   2584
   2585int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev,
   2586				 const struct prefix_info *pinfo,
   2587				 struct inet6_dev *in6_dev,
   2588				 const struct in6_addr *addr, int addr_type,
   2589				 u32 addr_flags, bool sllao, bool tokenized,
   2590				 __u32 valid_lft, u32 prefered_lft)
   2591{
   2592	struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1);
   2593	int create = 0, update_lft = 0;
   2594
   2595	if (!ifp && valid_lft) {
   2596		int max_addresses = in6_dev->cnf.max_addresses;
   2597		struct ifa6_config cfg = {
   2598			.pfx = addr,
   2599			.plen = pinfo->prefix_len,
   2600			.ifa_flags = addr_flags,
   2601			.valid_lft = valid_lft,
   2602			.preferred_lft = prefered_lft,
   2603			.scope = addr_type & IPV6_ADDR_SCOPE_MASK,
   2604			.ifa_proto = IFAPROT_KERNEL_RA
   2605		};
   2606
   2607#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   2608		if ((net->ipv6.devconf_all->optimistic_dad ||
   2609		     in6_dev->cnf.optimistic_dad) &&
   2610		    !net->ipv6.devconf_all->forwarding && sllao)
   2611			cfg.ifa_flags |= IFA_F_OPTIMISTIC;
   2612#endif
   2613
   2614		/* Do not allow to create too much of autoconfigured
   2615		 * addresses; this would be too easy way to crash kernel.
   2616		 */
   2617		if (!max_addresses ||
   2618		    ipv6_count_addresses(in6_dev) < max_addresses)
   2619			ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL);
   2620
   2621		if (IS_ERR_OR_NULL(ifp))
   2622			return -1;
   2623
   2624		create = 1;
   2625		spin_lock_bh(&ifp->lock);
   2626		ifp->flags |= IFA_F_MANAGETEMPADDR;
   2627		ifp->cstamp = jiffies;
   2628		ifp->tokenized = tokenized;
   2629		spin_unlock_bh(&ifp->lock);
   2630		addrconf_dad_start(ifp);
   2631	}
   2632
   2633	if (ifp) {
   2634		u32 flags;
   2635		unsigned long now;
   2636		u32 stored_lft;
   2637
   2638		/* update lifetime (RFC2462 5.5.3 e) */
   2639		spin_lock_bh(&ifp->lock);
   2640		now = jiffies;
   2641		if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
   2642			stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
   2643		else
   2644			stored_lft = 0;
   2645		if (!create && stored_lft) {
   2646			const u32 minimum_lft = min_t(u32,
   2647				stored_lft, MIN_VALID_LIFETIME);
   2648			valid_lft = max(valid_lft, minimum_lft);
   2649
   2650			/* RFC4862 Section 5.5.3e:
   2651			 * "Note that the preferred lifetime of the
   2652			 *  corresponding address is always reset to
   2653			 *  the Preferred Lifetime in the received
   2654			 *  Prefix Information option, regardless of
   2655			 *  whether the valid lifetime is also reset or
   2656			 *  ignored."
   2657			 *
   2658			 * So we should always update prefered_lft here.
   2659			 */
   2660			update_lft = 1;
   2661		}
   2662
   2663		if (update_lft) {
   2664			ifp->valid_lft = valid_lft;
   2665			ifp->prefered_lft = prefered_lft;
   2666			ifp->tstamp = now;
   2667			flags = ifp->flags;
   2668			ifp->flags &= ~IFA_F_DEPRECATED;
   2669			spin_unlock_bh(&ifp->lock);
   2670
   2671			if (!(flags&IFA_F_TENTATIVE))
   2672				ipv6_ifa_notify(0, ifp);
   2673		} else
   2674			spin_unlock_bh(&ifp->lock);
   2675
   2676		manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft,
   2677				 create, now);
   2678
   2679		in6_ifa_put(ifp);
   2680		addrconf_verify(net);
   2681	}
   2682
   2683	return 0;
   2684}
   2685EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr);
   2686
   2687void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
   2688{
   2689	struct prefix_info *pinfo;
   2690	__u32 valid_lft;
   2691	__u32 prefered_lft;
   2692	int addr_type, err;
   2693	u32 addr_flags = 0;
   2694	struct inet6_dev *in6_dev;
   2695	struct net *net = dev_net(dev);
   2696
   2697	pinfo = (struct prefix_info *) opt;
   2698
   2699	if (len < sizeof(struct prefix_info)) {
   2700		netdev_dbg(dev, "addrconf: prefix option too short\n");
   2701		return;
   2702	}
   2703
   2704	/*
   2705	 *	Validation checks ([ADDRCONF], page 19)
   2706	 */
   2707
   2708	addr_type = ipv6_addr_type(&pinfo->prefix);
   2709
   2710	if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
   2711		return;
   2712
   2713	valid_lft = ntohl(pinfo->valid);
   2714	prefered_lft = ntohl(pinfo->prefered);
   2715
   2716	if (prefered_lft > valid_lft) {
   2717		net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
   2718		return;
   2719	}
   2720
   2721	in6_dev = in6_dev_get(dev);
   2722
   2723	if (!in6_dev) {
   2724		net_dbg_ratelimited("addrconf: device %s not configured\n",
   2725				    dev->name);
   2726		return;
   2727	}
   2728
   2729	/*
   2730	 *	Two things going on here:
   2731	 *	1) Add routes for on-link prefixes
   2732	 *	2) Configure prefixes with the auto flag set
   2733	 */
   2734
   2735	if (pinfo->onlink) {
   2736		struct fib6_info *rt;
   2737		unsigned long rt_expires;
   2738
   2739		/* Avoid arithmetic overflow. Really, we could
   2740		 * save rt_expires in seconds, likely valid_lft,
   2741		 * but it would require division in fib gc, that it
   2742		 * not good.
   2743		 */
   2744		if (HZ > USER_HZ)
   2745			rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
   2746		else
   2747			rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
   2748
   2749		if (addrconf_finite_timeout(rt_expires))
   2750			rt_expires *= HZ;
   2751
   2752		rt = addrconf_get_prefix_route(&pinfo->prefix,
   2753					       pinfo->prefix_len,
   2754					       dev,
   2755					       RTF_ADDRCONF | RTF_PREFIX_RT,
   2756					       RTF_DEFAULT, true);
   2757
   2758		if (rt) {
   2759			/* Autoconf prefix route */
   2760			if (valid_lft == 0) {
   2761				ip6_del_rt(net, rt, false);
   2762				rt = NULL;
   2763			} else if (addrconf_finite_timeout(rt_expires)) {
   2764				/* not infinity */
   2765				fib6_set_expires(rt, jiffies + rt_expires);
   2766			} else {
   2767				fib6_clean_expires(rt);
   2768			}
   2769		} else if (valid_lft) {
   2770			clock_t expires = 0;
   2771			int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
   2772			if (addrconf_finite_timeout(rt_expires)) {
   2773				/* not infinity */
   2774				flags |= RTF_EXPIRES;
   2775				expires = jiffies_to_clock_t(rt_expires);
   2776			}
   2777			addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
   2778					      0, dev, expires, flags,
   2779					      GFP_ATOMIC);
   2780		}
   2781		fib6_info_release(rt);
   2782	}
   2783
   2784	/* Try to figure out our local address for this prefix */
   2785
   2786	if (pinfo->autoconf && in6_dev->cnf.autoconf) {
   2787		struct in6_addr addr;
   2788		bool tokenized = false, dev_addr_generated = false;
   2789
   2790		if (pinfo->prefix_len == 64) {
   2791			memcpy(&addr, &pinfo->prefix, 8);
   2792
   2793			if (!ipv6_addr_any(&in6_dev->token)) {
   2794				read_lock_bh(&in6_dev->lock);
   2795				memcpy(addr.s6_addr + 8,
   2796				       in6_dev->token.s6_addr + 8, 8);
   2797				read_unlock_bh(&in6_dev->lock);
   2798				tokenized = true;
   2799			} else if (is_addr_mode_generate_stable(in6_dev) &&
   2800				   !ipv6_generate_stable_address(&addr, 0,
   2801								 in6_dev)) {
   2802				addr_flags |= IFA_F_STABLE_PRIVACY;
   2803				goto ok;
   2804			} else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
   2805				   ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
   2806				goto put;
   2807			} else {
   2808				dev_addr_generated = true;
   2809			}
   2810			goto ok;
   2811		}
   2812		net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
   2813				    pinfo->prefix_len);
   2814		goto put;
   2815
   2816ok:
   2817		err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev,
   2818						   &addr, addr_type,
   2819						   addr_flags, sllao,
   2820						   tokenized, valid_lft,
   2821						   prefered_lft);
   2822		if (err)
   2823			goto put;
   2824
   2825		/* Ignore error case here because previous prefix add addr was
   2826		 * successful which will be notified.
   2827		 */
   2828		ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr,
   2829					      addr_type, addr_flags, sllao,
   2830					      tokenized, valid_lft,
   2831					      prefered_lft,
   2832					      dev_addr_generated);
   2833	}
   2834	inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
   2835put:
   2836	in6_dev_put(in6_dev);
   2837}
   2838
   2839static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev,
   2840		struct in6_ifreq *ireq)
   2841{
   2842	struct ip_tunnel_parm p = { };
   2843	int err;
   2844
   2845	if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4))
   2846		return -EADDRNOTAVAIL;
   2847
   2848	p.iph.daddr = ireq->ifr6_addr.s6_addr32[3];
   2849	p.iph.version = 4;
   2850	p.iph.ihl = 5;
   2851	p.iph.protocol = IPPROTO_IPV6;
   2852	p.iph.ttl = 64;
   2853
   2854	if (!dev->netdev_ops->ndo_tunnel_ctl)
   2855		return -EOPNOTSUPP;
   2856	err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL);
   2857	if (err)
   2858		return err;
   2859
   2860	dev = __dev_get_by_name(net, p.name);
   2861	if (!dev)
   2862		return -ENOBUFS;
   2863	return dev_open(dev, NULL);
   2864}
   2865
   2866/*
   2867 *	Set destination address.
   2868 *	Special case for SIT interfaces where we create a new "virtual"
   2869 *	device.
   2870 */
   2871int addrconf_set_dstaddr(struct net *net, void __user *arg)
   2872{
   2873	struct net_device *dev;
   2874	struct in6_ifreq ireq;
   2875	int err = -ENODEV;
   2876
   2877	if (!IS_ENABLED(CONFIG_IPV6_SIT))
   2878		return -ENODEV;
   2879	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
   2880		return -EFAULT;
   2881
   2882	rtnl_lock();
   2883	dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
   2884	if (dev && dev->type == ARPHRD_SIT)
   2885		err = addrconf_set_sit_dstaddr(net, dev, &ireq);
   2886	rtnl_unlock();
   2887	return err;
   2888}
   2889
   2890static int ipv6_mc_config(struct sock *sk, bool join,
   2891			  const struct in6_addr *addr, int ifindex)
   2892{
   2893	int ret;
   2894
   2895	ASSERT_RTNL();
   2896
   2897	lock_sock(sk);
   2898	if (join)
   2899		ret = ipv6_sock_mc_join(sk, ifindex, addr);
   2900	else
   2901		ret = ipv6_sock_mc_drop(sk, ifindex, addr);
   2902	release_sock(sk);
   2903
   2904	return ret;
   2905}
   2906
   2907/*
   2908 *	Manual configuration of address on an interface
   2909 */
   2910static int inet6_addr_add(struct net *net, int ifindex,
   2911			  struct ifa6_config *cfg,
   2912			  struct netlink_ext_ack *extack)
   2913{
   2914	struct inet6_ifaddr *ifp;
   2915	struct inet6_dev *idev;
   2916	struct net_device *dev;
   2917	unsigned long timeout;
   2918	clock_t expires;
   2919	u32 flags;
   2920
   2921	ASSERT_RTNL();
   2922
   2923	if (cfg->plen > 128)
   2924		return -EINVAL;
   2925
   2926	/* check the lifetime */
   2927	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
   2928		return -EINVAL;
   2929
   2930	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64)
   2931		return -EINVAL;
   2932
   2933	dev = __dev_get_by_index(net, ifindex);
   2934	if (!dev)
   2935		return -ENODEV;
   2936
   2937	idev = addrconf_add_dev(dev);
   2938	if (IS_ERR(idev))
   2939		return PTR_ERR(idev);
   2940
   2941	if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
   2942		int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk,
   2943					 true, cfg->pfx, ifindex);
   2944
   2945		if (ret < 0)
   2946			return ret;
   2947	}
   2948
   2949	cfg->scope = ipv6_addr_scope(cfg->pfx);
   2950
   2951	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
   2952	if (addrconf_finite_timeout(timeout)) {
   2953		expires = jiffies_to_clock_t(timeout * HZ);
   2954		cfg->valid_lft = timeout;
   2955		flags = RTF_EXPIRES;
   2956	} else {
   2957		expires = 0;
   2958		flags = 0;
   2959		cfg->ifa_flags |= IFA_F_PERMANENT;
   2960	}
   2961
   2962	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
   2963	if (addrconf_finite_timeout(timeout)) {
   2964		if (timeout == 0)
   2965			cfg->ifa_flags |= IFA_F_DEPRECATED;
   2966		cfg->preferred_lft = timeout;
   2967	}
   2968
   2969	ifp = ipv6_add_addr(idev, cfg, true, extack);
   2970	if (!IS_ERR(ifp)) {
   2971		if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
   2972			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
   2973					      ifp->rt_priority, dev, expires,
   2974					      flags, GFP_KERNEL);
   2975		}
   2976
   2977		/* Send a netlink notification if DAD is enabled and
   2978		 * optimistic flag is not set
   2979		 */
   2980		if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD)))
   2981			ipv6_ifa_notify(0, ifp);
   2982		/*
   2983		 * Note that section 3.1 of RFC 4429 indicates
   2984		 * that the Optimistic flag should not be set for
   2985		 * manually configured addresses
   2986		 */
   2987		addrconf_dad_start(ifp);
   2988		if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR)
   2989			manage_tempaddrs(idev, ifp, cfg->valid_lft,
   2990					 cfg->preferred_lft, true, jiffies);
   2991		in6_ifa_put(ifp);
   2992		addrconf_verify_rtnl(net);
   2993		return 0;
   2994	} else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) {
   2995		ipv6_mc_config(net->ipv6.mc_autojoin_sk, false,
   2996			       cfg->pfx, ifindex);
   2997	}
   2998
   2999	return PTR_ERR(ifp);
   3000}
   3001
   3002static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags,
   3003			  const struct in6_addr *pfx, unsigned int plen)
   3004{
   3005	struct inet6_ifaddr *ifp;
   3006	struct inet6_dev *idev;
   3007	struct net_device *dev;
   3008
   3009	if (plen > 128)
   3010		return -EINVAL;
   3011
   3012	dev = __dev_get_by_index(net, ifindex);
   3013	if (!dev)
   3014		return -ENODEV;
   3015
   3016	idev = __in6_dev_get(dev);
   3017	if (!idev)
   3018		return -ENXIO;
   3019
   3020	read_lock_bh(&idev->lock);
   3021	list_for_each_entry(ifp, &idev->addr_list, if_list) {
   3022		if (ifp->prefix_len == plen &&
   3023		    ipv6_addr_equal(pfx, &ifp->addr)) {
   3024			in6_ifa_hold(ifp);
   3025			read_unlock_bh(&idev->lock);
   3026
   3027			if (!(ifp->flags & IFA_F_TEMPORARY) &&
   3028			    (ifa_flags & IFA_F_MANAGETEMPADDR))
   3029				manage_tempaddrs(idev, ifp, 0, 0, false,
   3030						 jiffies);
   3031			ipv6_del_addr(ifp);
   3032			addrconf_verify_rtnl(net);
   3033			if (ipv6_addr_is_multicast(pfx)) {
   3034				ipv6_mc_config(net->ipv6.mc_autojoin_sk,
   3035					       false, pfx, dev->ifindex);
   3036			}
   3037			return 0;
   3038		}
   3039	}
   3040	read_unlock_bh(&idev->lock);
   3041	return -EADDRNOTAVAIL;
   3042}
   3043
   3044
   3045int addrconf_add_ifaddr(struct net *net, void __user *arg)
   3046{
   3047	struct ifa6_config cfg = {
   3048		.ifa_flags = IFA_F_PERMANENT,
   3049		.preferred_lft = INFINITY_LIFE_TIME,
   3050		.valid_lft = INFINITY_LIFE_TIME,
   3051	};
   3052	struct in6_ifreq ireq;
   3053	int err;
   3054
   3055	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
   3056		return -EPERM;
   3057
   3058	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
   3059		return -EFAULT;
   3060
   3061	cfg.pfx = &ireq.ifr6_addr;
   3062	cfg.plen = ireq.ifr6_prefixlen;
   3063
   3064	rtnl_lock();
   3065	err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL);
   3066	rtnl_unlock();
   3067	return err;
   3068}
   3069
   3070int addrconf_del_ifaddr(struct net *net, void __user *arg)
   3071{
   3072	struct in6_ifreq ireq;
   3073	int err;
   3074
   3075	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
   3076		return -EPERM;
   3077
   3078	if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
   3079		return -EFAULT;
   3080
   3081	rtnl_lock();
   3082	err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr,
   3083			     ireq.ifr6_prefixlen);
   3084	rtnl_unlock();
   3085	return err;
   3086}
   3087
   3088static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
   3089		     int plen, int scope, u8 proto)
   3090{
   3091	struct inet6_ifaddr *ifp;
   3092	struct ifa6_config cfg = {
   3093		.pfx = addr,
   3094		.plen = plen,
   3095		.ifa_flags = IFA_F_PERMANENT,
   3096		.valid_lft = INFINITY_LIFE_TIME,
   3097		.preferred_lft = INFINITY_LIFE_TIME,
   3098		.scope = scope,
   3099		.ifa_proto = proto
   3100	};
   3101
   3102	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
   3103	if (!IS_ERR(ifp)) {
   3104		spin_lock_bh(&ifp->lock);
   3105		ifp->flags &= ~IFA_F_TENTATIVE;
   3106		spin_unlock_bh(&ifp->lock);
   3107		rt_genid_bump_ipv6(dev_net(idev->dev));
   3108		ipv6_ifa_notify(RTM_NEWADDR, ifp);
   3109		in6_ifa_put(ifp);
   3110	}
   3111}
   3112
   3113#if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
   3114static void add_v4_addrs(struct inet6_dev *idev)
   3115{
   3116	struct in6_addr addr;
   3117	struct net_device *dev;
   3118	struct net *net = dev_net(idev->dev);
   3119	int scope, plen, offset = 0;
   3120	u32 pflags = 0;
   3121
   3122	ASSERT_RTNL();
   3123
   3124	memset(&addr, 0, sizeof(struct in6_addr));
   3125	/* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */
   3126	if (idev->dev->addr_len == sizeof(struct in6_addr))
   3127		offset = sizeof(struct in6_addr) - 4;
   3128	memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4);
   3129
   3130	if (idev->dev->flags&IFF_POINTOPOINT) {
   3131		if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE)
   3132			return;
   3133
   3134		addr.s6_addr32[0] = htonl(0xfe800000);
   3135		scope = IFA_LINK;
   3136		plen = 64;
   3137	} else {
   3138		scope = IPV6_ADDR_COMPATv4;
   3139		plen = 96;
   3140		pflags |= RTF_NONEXTHOP;
   3141	}
   3142
   3143	if (addr.s6_addr32[3]) {
   3144		add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC);
   3145		addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags,
   3146				      GFP_KERNEL);
   3147		return;
   3148	}
   3149
   3150	for_each_netdev(net, dev) {
   3151		struct in_device *in_dev = __in_dev_get_rtnl(dev);
   3152		if (in_dev && (dev->flags & IFF_UP)) {
   3153			struct in_ifaddr *ifa;
   3154			int flag = scope;
   3155
   3156			in_dev_for_each_ifa_rtnl(ifa, in_dev) {
   3157				addr.s6_addr32[3] = ifa->ifa_local;
   3158
   3159				if (ifa->ifa_scope == RT_SCOPE_LINK)
   3160					continue;
   3161				if (ifa->ifa_scope >= RT_SCOPE_HOST) {
   3162					if (idev->dev->flags&IFF_POINTOPOINT)
   3163						continue;
   3164					flag |= IFA_HOST;
   3165				}
   3166
   3167				add_addr(idev, &addr, plen, flag,
   3168					 IFAPROT_UNSPEC);
   3169				addrconf_prefix_route(&addr, plen, 0, idev->dev,
   3170						      0, pflags, GFP_KERNEL);
   3171			}
   3172		}
   3173	}
   3174}
   3175#endif
   3176
   3177static void init_loopback(struct net_device *dev)
   3178{
   3179	struct inet6_dev  *idev;
   3180
   3181	/* ::1 */
   3182
   3183	ASSERT_RTNL();
   3184
   3185	idev = ipv6_find_idev(dev);
   3186	if (IS_ERR(idev)) {
   3187		pr_debug("%s: add_dev failed\n", __func__);
   3188		return;
   3189	}
   3190
   3191	add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO);
   3192}
   3193
   3194void addrconf_add_linklocal(struct inet6_dev *idev,
   3195			    const struct in6_addr *addr, u32 flags)
   3196{
   3197	struct ifa6_config cfg = {
   3198		.pfx = addr,
   3199		.plen = 64,
   3200		.ifa_flags = flags | IFA_F_PERMANENT,
   3201		.valid_lft = INFINITY_LIFE_TIME,
   3202		.preferred_lft = INFINITY_LIFE_TIME,
   3203		.scope = IFA_LINK,
   3204		.ifa_proto = IFAPROT_KERNEL_LL
   3205	};
   3206	struct inet6_ifaddr *ifp;
   3207
   3208#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   3209	if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad ||
   3210	     idev->cnf.optimistic_dad) &&
   3211	    !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
   3212		cfg.ifa_flags |= IFA_F_OPTIMISTIC;
   3213#endif
   3214
   3215	ifp = ipv6_add_addr(idev, &cfg, true, NULL);
   3216	if (!IS_ERR(ifp)) {
   3217		addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev,
   3218				      0, 0, GFP_ATOMIC);
   3219		addrconf_dad_start(ifp);
   3220		in6_ifa_put(ifp);
   3221	}
   3222}
   3223EXPORT_SYMBOL_GPL(addrconf_add_linklocal);
   3224
   3225static bool ipv6_reserved_interfaceid(struct in6_addr address)
   3226{
   3227	if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0)
   3228		return true;
   3229
   3230	if (address.s6_addr32[2] == htonl(0x02005eff) &&
   3231	    ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000)))
   3232		return true;
   3233
   3234	if (address.s6_addr32[2] == htonl(0xfdffffff) &&
   3235	    ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80)))
   3236		return true;
   3237
   3238	return false;
   3239}
   3240
   3241static int ipv6_generate_stable_address(struct in6_addr *address,
   3242					u8 dad_count,
   3243					const struct inet6_dev *idev)
   3244{
   3245	static DEFINE_SPINLOCK(lock);
   3246	static __u32 digest[SHA1_DIGEST_WORDS];
   3247	static __u32 workspace[SHA1_WORKSPACE_WORDS];
   3248
   3249	static union {
   3250		char __data[SHA1_BLOCK_SIZE];
   3251		struct {
   3252			struct in6_addr secret;
   3253			__be32 prefix[2];
   3254			unsigned char hwaddr[MAX_ADDR_LEN];
   3255			u8 dad_count;
   3256		} __packed;
   3257	} data;
   3258
   3259	struct in6_addr secret;
   3260	struct in6_addr temp;
   3261	struct net *net = dev_net(idev->dev);
   3262
   3263	BUILD_BUG_ON(sizeof(data.__data) != sizeof(data));
   3264
   3265	if (idev->cnf.stable_secret.initialized)
   3266		secret = idev->cnf.stable_secret.secret;
   3267	else if (net->ipv6.devconf_dflt->stable_secret.initialized)
   3268		secret = net->ipv6.devconf_dflt->stable_secret.secret;
   3269	else
   3270		return -1;
   3271
   3272retry:
   3273	spin_lock_bh(&lock);
   3274
   3275	sha1_init(digest);
   3276	memset(&data, 0, sizeof(data));
   3277	memset(workspace, 0, sizeof(workspace));
   3278	memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len);
   3279	data.prefix[0] = address->s6_addr32[0];
   3280	data.prefix[1] = address->s6_addr32[1];
   3281	data.secret = secret;
   3282	data.dad_count = dad_count;
   3283
   3284	sha1_transform(digest, data.__data, workspace);
   3285
   3286	temp = *address;
   3287	temp.s6_addr32[2] = (__force __be32)digest[0];
   3288	temp.s6_addr32[3] = (__force __be32)digest[1];
   3289
   3290	spin_unlock_bh(&lock);
   3291
   3292	if (ipv6_reserved_interfaceid(temp)) {
   3293		dad_count++;
   3294		if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries)
   3295			return -1;
   3296		goto retry;
   3297	}
   3298
   3299	*address = temp;
   3300	return 0;
   3301}
   3302
   3303static void ipv6_gen_mode_random_init(struct inet6_dev *idev)
   3304{
   3305	struct ipv6_stable_secret *s = &idev->cnf.stable_secret;
   3306
   3307	if (s->initialized)
   3308		return;
   3309	s = &idev->cnf.stable_secret;
   3310	get_random_bytes(&s->secret, sizeof(s->secret));
   3311	s->initialized = true;
   3312}
   3313
   3314static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route)
   3315{
   3316	struct in6_addr addr;
   3317
   3318	/* no link local addresses on L3 master devices */
   3319	if (netif_is_l3_master(idev->dev))
   3320		return;
   3321
   3322	/* no link local addresses on devices flagged as slaves */
   3323	if (idev->dev->flags & IFF_SLAVE)
   3324		return;
   3325
   3326	ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
   3327
   3328	switch (idev->cnf.addr_gen_mode) {
   3329	case IN6_ADDR_GEN_MODE_RANDOM:
   3330		ipv6_gen_mode_random_init(idev);
   3331		fallthrough;
   3332	case IN6_ADDR_GEN_MODE_STABLE_PRIVACY:
   3333		if (!ipv6_generate_stable_address(&addr, 0, idev))
   3334			addrconf_add_linklocal(idev, &addr,
   3335					       IFA_F_STABLE_PRIVACY);
   3336		else if (prefix_route)
   3337			addrconf_prefix_route(&addr, 64, 0, idev->dev,
   3338					      0, 0, GFP_KERNEL);
   3339		break;
   3340	case IN6_ADDR_GEN_MODE_EUI64:
   3341		/* addrconf_add_linklocal also adds a prefix_route and we
   3342		 * only need to care about prefix routes if ipv6_generate_eui64
   3343		 * couldn't generate one.
   3344		 */
   3345		if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0)
   3346			addrconf_add_linklocal(idev, &addr, 0);
   3347		else if (prefix_route)
   3348			addrconf_prefix_route(&addr, 64, 0, idev->dev,
   3349					      0, 0, GFP_KERNEL);
   3350		break;
   3351	case IN6_ADDR_GEN_MODE_NONE:
   3352	default:
   3353		/* will not add any link local address */
   3354		break;
   3355	}
   3356}
   3357
   3358static void addrconf_dev_config(struct net_device *dev)
   3359{
   3360	struct inet6_dev *idev;
   3361
   3362	ASSERT_RTNL();
   3363
   3364	if ((dev->type != ARPHRD_ETHER) &&
   3365	    (dev->type != ARPHRD_FDDI) &&
   3366	    (dev->type != ARPHRD_ARCNET) &&
   3367	    (dev->type != ARPHRD_INFINIBAND) &&
   3368	    (dev->type != ARPHRD_IEEE1394) &&
   3369	    (dev->type != ARPHRD_TUNNEL6) &&
   3370	    (dev->type != ARPHRD_6LOWPAN) &&
   3371	    (dev->type != ARPHRD_TUNNEL) &&
   3372	    (dev->type != ARPHRD_NONE) &&
   3373	    (dev->type != ARPHRD_RAWIP)) {
   3374		/* Alas, we support only Ethernet autoconfiguration. */
   3375		idev = __in6_dev_get(dev);
   3376		if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP &&
   3377		    dev->flags & IFF_MULTICAST)
   3378			ipv6_mc_up(idev);
   3379		return;
   3380	}
   3381
   3382	idev = addrconf_add_dev(dev);
   3383	if (IS_ERR(idev))
   3384		return;
   3385
   3386	/* this device type has no EUI support */
   3387	if (dev->type == ARPHRD_NONE &&
   3388	    idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)
   3389		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM;
   3390
   3391	addrconf_addr_gen(idev, false);
   3392}
   3393
   3394#if IS_ENABLED(CONFIG_IPV6_SIT)
   3395static void addrconf_sit_config(struct net_device *dev)
   3396{
   3397	struct inet6_dev *idev;
   3398
   3399	ASSERT_RTNL();
   3400
   3401	/*
   3402	 * Configure the tunnel with one of our IPv4
   3403	 * addresses... we should configure all of
   3404	 * our v4 addrs in the tunnel
   3405	 */
   3406
   3407	idev = ipv6_find_idev(dev);
   3408	if (IS_ERR(idev)) {
   3409		pr_debug("%s: add_dev failed\n", __func__);
   3410		return;
   3411	}
   3412
   3413	if (dev->priv_flags & IFF_ISATAP) {
   3414		addrconf_addr_gen(idev, false);
   3415		return;
   3416	}
   3417
   3418	add_v4_addrs(idev);
   3419
   3420	if (dev->flags&IFF_POINTOPOINT)
   3421		addrconf_add_mroute(dev);
   3422}
   3423#endif
   3424
   3425#if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
   3426static void addrconf_gre_config(struct net_device *dev)
   3427{
   3428	struct inet6_dev *idev;
   3429
   3430	ASSERT_RTNL();
   3431
   3432	idev = ipv6_find_idev(dev);
   3433	if (IS_ERR(idev)) {
   3434		pr_debug("%s: add_dev failed\n", __func__);
   3435		return;
   3436	}
   3437
   3438	if (dev->type == ARPHRD_ETHER) {
   3439		addrconf_addr_gen(idev, true);
   3440		return;
   3441	}
   3442
   3443	add_v4_addrs(idev);
   3444
   3445	if (dev->flags & IFF_POINTOPOINT)
   3446		addrconf_add_mroute(dev);
   3447}
   3448#endif
   3449
   3450static int fixup_permanent_addr(struct net *net,
   3451				struct inet6_dev *idev,
   3452				struct inet6_ifaddr *ifp)
   3453{
   3454	/* !fib6_node means the host route was removed from the
   3455	 * FIB, for example, if 'lo' device is taken down. In that
   3456	 * case regenerate the host route.
   3457	 */
   3458	if (!ifp->rt || !ifp->rt->fib6_node) {
   3459		struct fib6_info *f6i, *prev;
   3460
   3461		f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false,
   3462					 GFP_ATOMIC);
   3463		if (IS_ERR(f6i))
   3464			return PTR_ERR(f6i);
   3465
   3466		/* ifp->rt can be accessed outside of rtnl */
   3467		spin_lock(&ifp->lock);
   3468		prev = ifp->rt;
   3469		ifp->rt = f6i;
   3470		spin_unlock(&ifp->lock);
   3471
   3472		fib6_info_release(prev);
   3473	}
   3474
   3475	if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) {
   3476		addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
   3477				      ifp->rt_priority, idev->dev, 0, 0,
   3478				      GFP_ATOMIC);
   3479	}
   3480
   3481	if (ifp->state == INET6_IFADDR_STATE_PREDAD)
   3482		addrconf_dad_start(ifp);
   3483
   3484	return 0;
   3485}
   3486
   3487static void addrconf_permanent_addr(struct net *net, struct net_device *dev)
   3488{
   3489	struct inet6_ifaddr *ifp, *tmp;
   3490	struct inet6_dev *idev;
   3491
   3492	idev = __in6_dev_get(dev);
   3493	if (!idev)
   3494		return;
   3495
   3496	write_lock_bh(&idev->lock);
   3497
   3498	list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) {
   3499		if ((ifp->flags & IFA_F_PERMANENT) &&
   3500		    fixup_permanent_addr(net, idev, ifp) < 0) {
   3501			write_unlock_bh(&idev->lock);
   3502			in6_ifa_hold(ifp);
   3503			ipv6_del_addr(ifp);
   3504			write_lock_bh(&idev->lock);
   3505
   3506			net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n",
   3507					     idev->dev->name, &ifp->addr);
   3508		}
   3509	}
   3510
   3511	write_unlock_bh(&idev->lock);
   3512}
   3513
   3514static int addrconf_notify(struct notifier_block *this, unsigned long event,
   3515			   void *ptr)
   3516{
   3517	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
   3518	struct netdev_notifier_change_info *change_info;
   3519	struct netdev_notifier_changeupper_info *info;
   3520	struct inet6_dev *idev = __in6_dev_get(dev);
   3521	struct net *net = dev_net(dev);
   3522	int run_pending = 0;
   3523	int err;
   3524
   3525	switch (event) {
   3526	case NETDEV_REGISTER:
   3527		if (!idev && dev->mtu >= IPV6_MIN_MTU) {
   3528			idev = ipv6_add_dev(dev);
   3529			if (IS_ERR(idev))
   3530				return notifier_from_errno(PTR_ERR(idev));
   3531		}
   3532		break;
   3533
   3534	case NETDEV_CHANGEMTU:
   3535		/* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */
   3536		if (dev->mtu < IPV6_MIN_MTU) {
   3537			addrconf_ifdown(dev, dev != net->loopback_dev);
   3538			break;
   3539		}
   3540
   3541		if (idev) {
   3542			rt6_mtu_change(dev, dev->mtu);
   3543			idev->cnf.mtu6 = dev->mtu;
   3544			break;
   3545		}
   3546
   3547		/* allocate new idev */
   3548		idev = ipv6_add_dev(dev);
   3549		if (IS_ERR(idev))
   3550			break;
   3551
   3552		/* device is still not ready */
   3553		if (!(idev->if_flags & IF_READY))
   3554			break;
   3555
   3556		run_pending = 1;
   3557		fallthrough;
   3558	case NETDEV_UP:
   3559	case NETDEV_CHANGE:
   3560		if (dev->flags & IFF_SLAVE)
   3561			break;
   3562
   3563		if (idev && idev->cnf.disable_ipv6)
   3564			break;
   3565
   3566		if (event == NETDEV_UP) {
   3567			/* restore routes for permanent addresses */
   3568			addrconf_permanent_addr(net, dev);
   3569
   3570			if (!addrconf_link_ready(dev)) {
   3571				/* device is not ready yet. */
   3572				pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
   3573					 dev->name);
   3574				break;
   3575			}
   3576
   3577			if (!idev && dev->mtu >= IPV6_MIN_MTU)
   3578				idev = ipv6_add_dev(dev);
   3579
   3580			if (!IS_ERR_OR_NULL(idev)) {
   3581				idev->if_flags |= IF_READY;
   3582				run_pending = 1;
   3583			}
   3584		} else if (event == NETDEV_CHANGE) {
   3585			if (!addrconf_link_ready(dev)) {
   3586				/* device is still not ready. */
   3587				rt6_sync_down_dev(dev, event);
   3588				break;
   3589			}
   3590
   3591			if (!IS_ERR_OR_NULL(idev)) {
   3592				if (idev->if_flags & IF_READY) {
   3593					/* device is already configured -
   3594					 * but resend MLD reports, we might
   3595					 * have roamed and need to update
   3596					 * multicast snooping switches
   3597					 */
   3598					ipv6_mc_up(idev);
   3599					change_info = ptr;
   3600					if (change_info->flags_changed & IFF_NOARP)
   3601						addrconf_dad_run(idev, true);
   3602					rt6_sync_up(dev, RTNH_F_LINKDOWN);
   3603					break;
   3604				}
   3605				idev->if_flags |= IF_READY;
   3606			}
   3607
   3608			pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
   3609				dev->name);
   3610
   3611			run_pending = 1;
   3612		}
   3613
   3614		switch (dev->type) {
   3615#if IS_ENABLED(CONFIG_IPV6_SIT)
   3616		case ARPHRD_SIT:
   3617			addrconf_sit_config(dev);
   3618			break;
   3619#endif
   3620#if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE)
   3621		case ARPHRD_IP6GRE:
   3622		case ARPHRD_IPGRE:
   3623			addrconf_gre_config(dev);
   3624			break;
   3625#endif
   3626		case ARPHRD_LOOPBACK:
   3627			init_loopback(dev);
   3628			break;
   3629
   3630		default:
   3631			addrconf_dev_config(dev);
   3632			break;
   3633		}
   3634
   3635		if (!IS_ERR_OR_NULL(idev)) {
   3636			if (run_pending)
   3637				addrconf_dad_run(idev, false);
   3638
   3639			/* Device has an address by now */
   3640			rt6_sync_up(dev, RTNH_F_DEAD);
   3641
   3642			/*
   3643			 * If the MTU changed during the interface down,
   3644			 * when the interface up, the changed MTU must be
   3645			 * reflected in the idev as well as routers.
   3646			 */
   3647			if (idev->cnf.mtu6 != dev->mtu &&
   3648			    dev->mtu >= IPV6_MIN_MTU) {
   3649				rt6_mtu_change(dev, dev->mtu);
   3650				idev->cnf.mtu6 = dev->mtu;
   3651			}
   3652			idev->tstamp = jiffies;
   3653			inet6_ifinfo_notify(RTM_NEWLINK, idev);
   3654
   3655			/*
   3656			 * If the changed mtu during down is lower than
   3657			 * IPV6_MIN_MTU stop IPv6 on this interface.
   3658			 */
   3659			if (dev->mtu < IPV6_MIN_MTU)
   3660				addrconf_ifdown(dev, dev != net->loopback_dev);
   3661		}
   3662		break;
   3663
   3664	case NETDEV_DOWN:
   3665	case NETDEV_UNREGISTER:
   3666		/*
   3667		 *	Remove all addresses from this interface.
   3668		 */
   3669		addrconf_ifdown(dev, event != NETDEV_DOWN);
   3670		break;
   3671
   3672	case NETDEV_CHANGENAME:
   3673		if (idev) {
   3674			snmp6_unregister_dev(idev);
   3675			addrconf_sysctl_unregister(idev);
   3676			err = addrconf_sysctl_register(idev);
   3677			if (err)
   3678				return notifier_from_errno(err);
   3679			err = snmp6_register_dev(idev);
   3680			if (err) {
   3681				addrconf_sysctl_unregister(idev);
   3682				return notifier_from_errno(err);
   3683			}
   3684		}
   3685		break;
   3686
   3687	case NETDEV_PRE_TYPE_CHANGE:
   3688	case NETDEV_POST_TYPE_CHANGE:
   3689		if (idev)
   3690			addrconf_type_change(dev, event);
   3691		break;
   3692
   3693	case NETDEV_CHANGEUPPER:
   3694		info = ptr;
   3695
   3696		/* flush all routes if dev is linked to or unlinked from
   3697		 * an L3 master device (e.g., VRF)
   3698		 */
   3699		if (info->upper_dev && netif_is_l3_master(info->upper_dev))
   3700			addrconf_ifdown(dev, false);
   3701	}
   3702
   3703	return NOTIFY_OK;
   3704}
   3705
   3706/*
   3707 *	addrconf module should be notified of a device going up
   3708 */
   3709static struct notifier_block ipv6_dev_notf = {
   3710	.notifier_call = addrconf_notify,
   3711	.priority = ADDRCONF_NOTIFY_PRIORITY,
   3712};
   3713
   3714static void addrconf_type_change(struct net_device *dev, unsigned long event)
   3715{
   3716	struct inet6_dev *idev;
   3717	ASSERT_RTNL();
   3718
   3719	idev = __in6_dev_get(dev);
   3720
   3721	if (event == NETDEV_POST_TYPE_CHANGE)
   3722		ipv6_mc_remap(idev);
   3723	else if (event == NETDEV_PRE_TYPE_CHANGE)
   3724		ipv6_mc_unmap(idev);
   3725}
   3726
   3727static bool addr_is_local(const struct in6_addr *addr)
   3728{
   3729	return ipv6_addr_type(addr) &
   3730		(IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
   3731}
   3732
   3733static int addrconf_ifdown(struct net_device *dev, bool unregister)
   3734{
   3735	unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN;
   3736	struct net *net = dev_net(dev);
   3737	struct inet6_dev *idev;
   3738	struct inet6_ifaddr *ifa;
   3739	LIST_HEAD(tmp_addr_list);
   3740	bool keep_addr = false;
   3741	bool was_ready;
   3742	int state, i;
   3743
   3744	ASSERT_RTNL();
   3745
   3746	rt6_disable_ip(dev, event);
   3747
   3748	idev = __in6_dev_get(dev);
   3749	if (!idev)
   3750		return -ENODEV;
   3751
   3752	/*
   3753	 * Step 1: remove reference to ipv6 device from parent device.
   3754	 *	   Do not dev_put!
   3755	 */
   3756	if (unregister) {
   3757		idev->dead = 1;
   3758
   3759		/* protected by rtnl_lock */
   3760		RCU_INIT_POINTER(dev->ip6_ptr, NULL);
   3761
   3762		/* Step 1.5: remove snmp6 entry */
   3763		snmp6_unregister_dev(idev);
   3764
   3765	}
   3766
   3767	/* combine the user config with event to determine if permanent
   3768	 * addresses are to be removed from address hash table
   3769	 */
   3770	if (!unregister && !idev->cnf.disable_ipv6) {
   3771		/* aggregate the system setting and interface setting */
   3772		int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down;
   3773
   3774		if (!_keep_addr)
   3775			_keep_addr = idev->cnf.keep_addr_on_down;
   3776
   3777		keep_addr = (_keep_addr > 0);
   3778	}
   3779
   3780	/* Step 2: clear hash table */
   3781	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
   3782		struct hlist_head *h = &net->ipv6.inet6_addr_lst[i];
   3783
   3784		spin_lock_bh(&net->ipv6.addrconf_hash_lock);
   3785restart:
   3786		hlist_for_each_entry_rcu(ifa, h, addr_lst) {
   3787			if (ifa->idev == idev) {
   3788				addrconf_del_dad_work(ifa);
   3789				/* combined flag + permanent flag decide if
   3790				 * address is retained on a down event
   3791				 */
   3792				if (!keep_addr ||
   3793				    !(ifa->flags & IFA_F_PERMANENT) ||
   3794				    addr_is_local(&ifa->addr)) {
   3795					hlist_del_init_rcu(&ifa->addr_lst);
   3796					goto restart;
   3797				}
   3798			}
   3799		}
   3800		spin_unlock_bh(&net->ipv6.addrconf_hash_lock);
   3801	}
   3802
   3803	write_lock_bh(&idev->lock);
   3804
   3805	addrconf_del_rs_timer(idev);
   3806
   3807	/* Step 2: clear flags for stateless addrconf, repeated down
   3808	 *         detection
   3809	 */
   3810	was_ready = idev->if_flags & IF_READY;
   3811	if (!unregister)
   3812		idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
   3813
   3814	/* Step 3: clear tempaddr list */
   3815	while (!list_empty(&idev->tempaddr_list)) {
   3816		ifa = list_first_entry(&idev->tempaddr_list,
   3817				       struct inet6_ifaddr, tmp_list);
   3818		list_del(&ifa->tmp_list);
   3819		write_unlock_bh(&idev->lock);
   3820		spin_lock_bh(&ifa->lock);
   3821
   3822		if (ifa->ifpub) {
   3823			in6_ifa_put(ifa->ifpub);
   3824			ifa->ifpub = NULL;
   3825		}
   3826		spin_unlock_bh(&ifa->lock);
   3827		in6_ifa_put(ifa);
   3828		write_lock_bh(&idev->lock);
   3829	}
   3830
   3831	list_for_each_entry(ifa, &idev->addr_list, if_list)
   3832		list_add_tail(&ifa->if_list_aux, &tmp_addr_list);
   3833	write_unlock_bh(&idev->lock);
   3834
   3835	while (!list_empty(&tmp_addr_list)) {
   3836		struct fib6_info *rt = NULL;
   3837		bool keep;
   3838
   3839		ifa = list_first_entry(&tmp_addr_list,
   3840				       struct inet6_ifaddr, if_list_aux);
   3841		list_del(&ifa->if_list_aux);
   3842
   3843		addrconf_del_dad_work(ifa);
   3844
   3845		keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) &&
   3846			!addr_is_local(&ifa->addr);
   3847
   3848		spin_lock_bh(&ifa->lock);
   3849
   3850		if (keep) {
   3851			/* set state to skip the notifier below */
   3852			state = INET6_IFADDR_STATE_DEAD;
   3853			ifa->state = INET6_IFADDR_STATE_PREDAD;
   3854			if (!(ifa->flags & IFA_F_NODAD))
   3855				ifa->flags |= IFA_F_TENTATIVE;
   3856
   3857			rt = ifa->rt;
   3858			ifa->rt = NULL;
   3859		} else {
   3860			state = ifa->state;
   3861			ifa->state = INET6_IFADDR_STATE_DEAD;
   3862		}
   3863
   3864		spin_unlock_bh(&ifa->lock);
   3865
   3866		if (rt)
   3867			ip6_del_rt(net, rt, false);
   3868
   3869		if (state != INET6_IFADDR_STATE_DEAD) {
   3870			__ipv6_ifa_notify(RTM_DELADDR, ifa);
   3871			inet6addr_notifier_call_chain(NETDEV_DOWN, ifa);
   3872		} else {
   3873			if (idev->cnf.forwarding)
   3874				addrconf_leave_anycast(ifa);
   3875			addrconf_leave_solict(ifa->idev, &ifa->addr);
   3876		}
   3877
   3878		if (!keep) {
   3879			write_lock_bh(&idev->lock);
   3880			list_del_rcu(&ifa->if_list);
   3881			write_unlock_bh(&idev->lock);
   3882			in6_ifa_put(ifa);
   3883		}
   3884	}
   3885
   3886	/* Step 5: Discard anycast and multicast list */
   3887	if (unregister) {
   3888		ipv6_ac_destroy_dev(idev);
   3889		ipv6_mc_destroy_dev(idev);
   3890	} else if (was_ready) {
   3891		ipv6_mc_down(idev);
   3892	}
   3893
   3894	idev->tstamp = jiffies;
   3895	idev->ra_mtu = 0;
   3896
   3897	/* Last: Shot the device (if unregistered) */
   3898	if (unregister) {
   3899		addrconf_sysctl_unregister(idev);
   3900		neigh_parms_release(&nd_tbl, idev->nd_parms);
   3901		neigh_ifdown(&nd_tbl, dev);
   3902		in6_dev_put(idev);
   3903	}
   3904	return 0;
   3905}
   3906
   3907static void addrconf_rs_timer(struct timer_list *t)
   3908{
   3909	struct inet6_dev *idev = from_timer(idev, t, rs_timer);
   3910	struct net_device *dev = idev->dev;
   3911	struct in6_addr lladdr;
   3912
   3913	write_lock(&idev->lock);
   3914	if (idev->dead || !(idev->if_flags & IF_READY))
   3915		goto out;
   3916
   3917	if (!ipv6_accept_ra(idev))
   3918		goto out;
   3919
   3920	/* Announcement received after solicitation was sent */
   3921	if (idev->if_flags & IF_RA_RCVD)
   3922		goto out;
   3923
   3924	if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) {
   3925		write_unlock(&idev->lock);
   3926		if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
   3927			ndisc_send_rs(dev, &lladdr,
   3928				      &in6addr_linklocal_allrouters);
   3929		else
   3930			goto put;
   3931
   3932		write_lock(&idev->lock);
   3933		idev->rs_interval = rfc3315_s14_backoff_update(
   3934			idev->rs_interval, idev->cnf.rtr_solicit_max_interval);
   3935		/* The wait after the last probe can be shorter */
   3936		addrconf_mod_rs_timer(idev, (idev->rs_probes ==
   3937					     idev->cnf.rtr_solicits) ?
   3938				      idev->cnf.rtr_solicit_delay :
   3939				      idev->rs_interval);
   3940	} else {
   3941		/*
   3942		 * Note: we do not support deprecated "all on-link"
   3943		 * assumption any longer.
   3944		 */
   3945		pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
   3946	}
   3947
   3948out:
   3949	write_unlock(&idev->lock);
   3950put:
   3951	in6_dev_put(idev);
   3952}
   3953
   3954/*
   3955 *	Duplicate Address Detection
   3956 */
   3957static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
   3958{
   3959	unsigned long rand_num;
   3960	struct inet6_dev *idev = ifp->idev;
   3961	u64 nonce;
   3962
   3963	if (ifp->flags & IFA_F_OPTIMISTIC)
   3964		rand_num = 0;
   3965	else
   3966		rand_num = prandom_u32() % (idev->cnf.rtr_solicit_delay ? : 1);
   3967
   3968	nonce = 0;
   3969	if (idev->cnf.enhanced_dad ||
   3970	    dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) {
   3971		do
   3972			get_random_bytes(&nonce, 6);
   3973		while (nonce == 0);
   3974	}
   3975	ifp->dad_nonce = nonce;
   3976	ifp->dad_probes = idev->cnf.dad_transmits;
   3977	addrconf_mod_dad_work(ifp, rand_num);
   3978}
   3979
   3980static void addrconf_dad_begin(struct inet6_ifaddr *ifp)
   3981{
   3982	struct inet6_dev *idev = ifp->idev;
   3983	struct net_device *dev = idev->dev;
   3984	bool bump_id, notify = false;
   3985	struct net *net;
   3986
   3987	addrconf_join_solict(dev, &ifp->addr);
   3988
   3989	read_lock_bh(&idev->lock);
   3990	spin_lock(&ifp->lock);
   3991	if (ifp->state == INET6_IFADDR_STATE_DEAD)
   3992		goto out;
   3993
   3994	net = dev_net(dev);
   3995	if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
   3996	    (net->ipv6.devconf_all->accept_dad < 1 &&
   3997	     idev->cnf.accept_dad < 1) ||
   3998	    !(ifp->flags&IFA_F_TENTATIVE) ||
   3999	    ifp->flags & IFA_F_NODAD) {
   4000		bool send_na = false;
   4001
   4002		if (ifp->flags & IFA_F_TENTATIVE &&
   4003		    !(ifp->flags & IFA_F_OPTIMISTIC))
   4004			send_na = true;
   4005		bump_id = ifp->flags & IFA_F_TENTATIVE;
   4006		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
   4007		spin_unlock(&ifp->lock);
   4008		read_unlock_bh(&idev->lock);
   4009
   4010		addrconf_dad_completed(ifp, bump_id, send_na);
   4011		return;
   4012	}
   4013
   4014	if (!(idev->if_flags & IF_READY)) {
   4015		spin_unlock(&ifp->lock);
   4016		read_unlock_bh(&idev->lock);
   4017		/*
   4018		 * If the device is not ready:
   4019		 * - keep it tentative if it is a permanent address.
   4020		 * - otherwise, kill it.
   4021		 */
   4022		in6_ifa_hold(ifp);
   4023		addrconf_dad_stop(ifp, 0);
   4024		return;
   4025	}
   4026
   4027	/*
   4028	 * Optimistic nodes can start receiving
   4029	 * Frames right away
   4030	 */
   4031	if (ifp->flags & IFA_F_OPTIMISTIC) {
   4032		ip6_ins_rt(net, ifp->rt);
   4033		if (ipv6_use_optimistic_addr(net, idev)) {
   4034			/* Because optimistic nodes can use this address,
   4035			 * notify listeners. If DAD fails, RTM_DELADDR is sent.
   4036			 */
   4037			notify = true;
   4038		}
   4039	}
   4040
   4041	addrconf_dad_kick(ifp);
   4042out:
   4043	spin_unlock(&ifp->lock);
   4044	read_unlock_bh(&idev->lock);
   4045	if (notify)
   4046		ipv6_ifa_notify(RTM_NEWADDR, ifp);
   4047}
   4048
   4049static void addrconf_dad_start(struct inet6_ifaddr *ifp)
   4050{
   4051	bool begin_dad = false;
   4052
   4053	spin_lock_bh(&ifp->lock);
   4054	if (ifp->state != INET6_IFADDR_STATE_DEAD) {
   4055		ifp->state = INET6_IFADDR_STATE_PREDAD;
   4056		begin_dad = true;
   4057	}
   4058	spin_unlock_bh(&ifp->lock);
   4059
   4060	if (begin_dad)
   4061		addrconf_mod_dad_work(ifp, 0);
   4062}
   4063
   4064static void addrconf_dad_work(struct work_struct *w)
   4065{
   4066	struct inet6_ifaddr *ifp = container_of(to_delayed_work(w),
   4067						struct inet6_ifaddr,
   4068						dad_work);
   4069	struct inet6_dev *idev = ifp->idev;
   4070	bool bump_id, disable_ipv6 = false;
   4071	struct in6_addr mcaddr;
   4072
   4073	enum {
   4074		DAD_PROCESS,
   4075		DAD_BEGIN,
   4076		DAD_ABORT,
   4077	} action = DAD_PROCESS;
   4078
   4079	rtnl_lock();
   4080
   4081	spin_lock_bh(&ifp->lock);
   4082	if (ifp->state == INET6_IFADDR_STATE_PREDAD) {
   4083		action = DAD_BEGIN;
   4084		ifp->state = INET6_IFADDR_STATE_DAD;
   4085	} else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) {
   4086		action = DAD_ABORT;
   4087		ifp->state = INET6_IFADDR_STATE_POSTDAD;
   4088
   4089		if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 ||
   4090		     idev->cnf.accept_dad > 1) &&
   4091		    !idev->cnf.disable_ipv6 &&
   4092		    !(ifp->flags & IFA_F_STABLE_PRIVACY)) {
   4093			struct in6_addr addr;
   4094
   4095			addr.s6_addr32[0] = htonl(0xfe800000);
   4096			addr.s6_addr32[1] = 0;
   4097
   4098			if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
   4099			    ipv6_addr_equal(&ifp->addr, &addr)) {
   4100				/* DAD failed for link-local based on MAC */
   4101				idev->cnf.disable_ipv6 = 1;
   4102
   4103				pr_info("%s: IPv6 being disabled!\n",
   4104					ifp->idev->dev->name);
   4105				disable_ipv6 = true;
   4106			}
   4107		}
   4108	}
   4109	spin_unlock_bh(&ifp->lock);
   4110
   4111	if (action == DAD_BEGIN) {
   4112		addrconf_dad_begin(ifp);
   4113		goto out;
   4114	} else if (action == DAD_ABORT) {
   4115		in6_ifa_hold(ifp);
   4116		addrconf_dad_stop(ifp, 1);
   4117		if (disable_ipv6)
   4118			addrconf_ifdown(idev->dev, false);
   4119		goto out;
   4120	}
   4121
   4122	if (!ifp->dad_probes && addrconf_dad_end(ifp))
   4123		goto out;
   4124
   4125	write_lock_bh(&idev->lock);
   4126	if (idev->dead || !(idev->if_flags & IF_READY)) {
   4127		write_unlock_bh(&idev->lock);
   4128		goto out;
   4129	}
   4130
   4131	spin_lock(&ifp->lock);
   4132	if (ifp->state == INET6_IFADDR_STATE_DEAD) {
   4133		spin_unlock(&ifp->lock);
   4134		write_unlock_bh(&idev->lock);
   4135		goto out;
   4136	}
   4137
   4138	if (ifp->dad_probes == 0) {
   4139		bool send_na = false;
   4140
   4141		/*
   4142		 * DAD was successful
   4143		 */
   4144
   4145		if (ifp->flags & IFA_F_TENTATIVE &&
   4146		    !(ifp->flags & IFA_F_OPTIMISTIC))
   4147			send_na = true;
   4148		bump_id = ifp->flags & IFA_F_TENTATIVE;
   4149		ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
   4150		spin_unlock(&ifp->lock);
   4151		write_unlock_bh(&idev->lock);
   4152
   4153		addrconf_dad_completed(ifp, bump_id, send_na);
   4154
   4155		goto out;
   4156	}
   4157
   4158	ifp->dad_probes--;
   4159	addrconf_mod_dad_work(ifp,
   4160			      max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME),
   4161				  HZ/100));
   4162	spin_unlock(&ifp->lock);
   4163	write_unlock_bh(&idev->lock);
   4164
   4165	/* send a neighbour solicitation for our addr */
   4166	addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
   4167	ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any,
   4168		      ifp->dad_nonce);
   4169out:
   4170	in6_ifa_put(ifp);
   4171	rtnl_unlock();
   4172}
   4173
   4174/* ifp->idev must be at least read locked */
   4175static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp)
   4176{
   4177	struct inet6_ifaddr *ifpiter;
   4178	struct inet6_dev *idev = ifp->idev;
   4179
   4180	list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) {
   4181		if (ifpiter->scope > IFA_LINK)
   4182			break;
   4183		if (ifp != ifpiter && ifpiter->scope == IFA_LINK &&
   4184		    (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|
   4185				       IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) ==
   4186		    IFA_F_PERMANENT)
   4187			return false;
   4188	}
   4189	return true;
   4190}
   4191
   4192static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id,
   4193				   bool send_na)
   4194{
   4195	struct net_device *dev = ifp->idev->dev;
   4196	struct in6_addr lladdr;
   4197	bool send_rs, send_mld;
   4198
   4199	addrconf_del_dad_work(ifp);
   4200
   4201	/*
   4202	 *	Configure the address for reception. Now it is valid.
   4203	 */
   4204
   4205	ipv6_ifa_notify(RTM_NEWADDR, ifp);
   4206
   4207	/* If added prefix is link local and we are prepared to process
   4208	   router advertisements, start sending router solicitations.
   4209	 */
   4210
   4211	read_lock_bh(&ifp->idev->lock);
   4212	send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp);
   4213	send_rs = send_mld &&
   4214		  ipv6_accept_ra(ifp->idev) &&
   4215		  ifp->idev->cnf.rtr_solicits != 0 &&
   4216		  (dev->flags & IFF_LOOPBACK) == 0 &&
   4217		  (dev->type != ARPHRD_TUNNEL);
   4218	read_unlock_bh(&ifp->idev->lock);
   4219
   4220	/* While dad is in progress mld report's source address is in6_addrany.
   4221	 * Resend with proper ll now.
   4222	 */
   4223	if (send_mld)
   4224		ipv6_mc_dad_complete(ifp->idev);
   4225
   4226	/* send unsolicited NA if enabled */
   4227	if (send_na &&
   4228	    (ifp->idev->cnf.ndisc_notify ||
   4229	     dev_net(dev)->ipv6.devconf_all->ndisc_notify)) {
   4230		ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr,
   4231			      /*router=*/ !!ifp->idev->cnf.forwarding,
   4232			      /*solicited=*/ false, /*override=*/ true,
   4233			      /*inc_opt=*/ true);
   4234	}
   4235
   4236	if (send_rs) {
   4237		/*
   4238		 *	If a host as already performed a random delay
   4239		 *	[...] as part of DAD [...] there is no need
   4240		 *	to delay again before sending the first RS
   4241		 */
   4242		if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE))
   4243			return;
   4244		ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters);
   4245
   4246		write_lock_bh(&ifp->idev->lock);
   4247		spin_lock(&ifp->lock);
   4248		ifp->idev->rs_interval = rfc3315_s14_backoff_init(
   4249			ifp->idev->cnf.rtr_solicit_interval);
   4250		ifp->idev->rs_probes = 1;
   4251		ifp->idev->if_flags |= IF_RS_SENT;
   4252		addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval);
   4253		spin_unlock(&ifp->lock);
   4254		write_unlock_bh(&ifp->idev->lock);
   4255	}
   4256
   4257	if (bump_id)
   4258		rt_genid_bump_ipv6(dev_net(dev));
   4259
   4260	/* Make sure that a new temporary address will be created
   4261	 * before this temporary address becomes deprecated.
   4262	 */
   4263	if (ifp->flags & IFA_F_TEMPORARY)
   4264		addrconf_verify_rtnl(dev_net(dev));
   4265}
   4266
   4267static void addrconf_dad_run(struct inet6_dev *idev, bool restart)
   4268{
   4269	struct inet6_ifaddr *ifp;
   4270
   4271	read_lock_bh(&idev->lock);
   4272	list_for_each_entry(ifp, &idev->addr_list, if_list) {
   4273		spin_lock(&ifp->lock);
   4274		if ((ifp->flags & IFA_F_TENTATIVE &&
   4275		     ifp->state == INET6_IFADDR_STATE_DAD) || restart) {
   4276			if (restart)
   4277				ifp->state = INET6_IFADDR_STATE_PREDAD;
   4278			addrconf_dad_kick(ifp);
   4279		}
   4280		spin_unlock(&ifp->lock);
   4281	}
   4282	read_unlock_bh(&idev->lock);
   4283}
   4284
   4285#ifdef CONFIG_PROC_FS
   4286struct if6_iter_state {
   4287	struct seq_net_private p;
   4288	int bucket;
   4289	int offset;
   4290};
   4291
   4292static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
   4293{
   4294	struct if6_iter_state *state = seq->private;
   4295	struct net *net = seq_file_net(seq);
   4296	struct inet6_ifaddr *ifa = NULL;
   4297	int p = 0;
   4298
   4299	/* initial bucket if pos is 0 */
   4300	if (pos == 0) {
   4301		state->bucket = 0;
   4302		state->offset = 0;
   4303	}
   4304
   4305	for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
   4306		hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket],
   4307					 addr_lst) {
   4308			/* sync with offset */
   4309			if (p < state->offset) {
   4310				p++;
   4311				continue;
   4312			}
   4313			return ifa;
   4314		}
   4315
   4316		/* prepare for next bucket */
   4317		state->offset = 0;
   4318		p = 0;
   4319	}
   4320	return NULL;
   4321}
   4322
   4323static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
   4324					 struct inet6_ifaddr *ifa)
   4325{
   4326	struct if6_iter_state *state = seq->private;
   4327	struct net *net = seq_file_net(seq);
   4328
   4329	hlist_for_each_entry_continue_rcu(ifa, addr_lst) {
   4330		state->offset++;
   4331		return ifa;
   4332	}
   4333
   4334	state->offset = 0;
   4335	while (++state->bucket < IN6_ADDR_HSIZE) {
   4336		hlist_for_each_entry_rcu(ifa,
   4337				     &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) {
   4338			return ifa;
   4339		}
   4340	}
   4341
   4342	return NULL;
   4343}
   4344
   4345static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
   4346	__acquires(rcu)
   4347{
   4348	rcu_read_lock();
   4349	return if6_get_first(seq, *pos);
   4350}
   4351
   4352static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
   4353{
   4354	struct inet6_ifaddr *ifa;
   4355
   4356	ifa = if6_get_next(seq, v);
   4357	++*pos;
   4358	return ifa;
   4359}
   4360
   4361static void if6_seq_stop(struct seq_file *seq, void *v)
   4362	__releases(rcu)
   4363{
   4364	rcu_read_unlock();
   4365}
   4366
   4367static int if6_seq_show(struct seq_file *seq, void *v)
   4368{
   4369	struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
   4370	seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
   4371		   &ifp->addr,
   4372		   ifp->idev->dev->ifindex,
   4373		   ifp->prefix_len,
   4374		   ifp->scope,
   4375		   (u8) ifp->flags,
   4376		   ifp->idev->dev->name);
   4377	return 0;
   4378}
   4379
   4380static const struct seq_operations if6_seq_ops = {
   4381	.start	= if6_seq_start,
   4382	.next	= if6_seq_next,
   4383	.show	= if6_seq_show,
   4384	.stop	= if6_seq_stop,
   4385};
   4386
   4387static int __net_init if6_proc_net_init(struct net *net)
   4388{
   4389	if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops,
   4390			sizeof(struct if6_iter_state)))
   4391		return -ENOMEM;
   4392	return 0;
   4393}
   4394
   4395static void __net_exit if6_proc_net_exit(struct net *net)
   4396{
   4397	remove_proc_entry("if_inet6", net->proc_net);
   4398}
   4399
   4400static struct pernet_operations if6_proc_net_ops = {
   4401	.init = if6_proc_net_init,
   4402	.exit = if6_proc_net_exit,
   4403};
   4404
   4405int __init if6_proc_init(void)
   4406{
   4407	return register_pernet_subsys(&if6_proc_net_ops);
   4408}
   4409
   4410void if6_proc_exit(void)
   4411{
   4412	unregister_pernet_subsys(&if6_proc_net_ops);
   4413}
   4414#endif	/* CONFIG_PROC_FS */
   4415
   4416#if IS_ENABLED(CONFIG_IPV6_MIP6)
   4417/* Check if address is a home address configured on any interface. */
   4418int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
   4419{
   4420	unsigned int hash = inet6_addr_hash(net, addr);
   4421	struct inet6_ifaddr *ifp = NULL;
   4422	int ret = 0;
   4423
   4424	rcu_read_lock();
   4425	hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
   4426		if (ipv6_addr_equal(&ifp->addr, addr) &&
   4427		    (ifp->flags & IFA_F_HOMEADDRESS)) {
   4428			ret = 1;
   4429			break;
   4430		}
   4431	}
   4432	rcu_read_unlock();
   4433	return ret;
   4434}
   4435#endif
   4436
   4437/* RFC6554 has some algorithm to avoid loops in segment routing by
   4438 * checking if the segments contains any of a local interface address.
   4439 *
   4440 * Quote:
   4441 *
   4442 * To detect loops in the SRH, a router MUST determine if the SRH
   4443 * includes multiple addresses assigned to any interface on that router.
   4444 * If such addresses appear more than once and are separated by at least
   4445 * one address not assigned to that router.
   4446 */
   4447int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs,
   4448			  unsigned char nsegs)
   4449{
   4450	const struct in6_addr *addr;
   4451	int i, ret = 0, found = 0;
   4452	struct inet6_ifaddr *ifp;
   4453	bool separated = false;
   4454	unsigned int hash;
   4455	bool hash_found;
   4456
   4457	rcu_read_lock();
   4458	for (i = 0; i < nsegs; i++) {
   4459		addr = &segs[i];
   4460		hash = inet6_addr_hash(net, addr);
   4461
   4462		hash_found = false;
   4463		hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) {
   4464
   4465			if (ipv6_addr_equal(&ifp->addr, addr)) {
   4466				hash_found = true;
   4467				break;
   4468			}
   4469		}
   4470
   4471		if (hash_found) {
   4472			if (found > 1 && separated) {
   4473				ret = 1;
   4474				break;
   4475			}
   4476
   4477			separated = false;
   4478			found++;
   4479		} else {
   4480			separated = true;
   4481		}
   4482	}
   4483	rcu_read_unlock();
   4484
   4485	return ret;
   4486}
   4487
   4488/*
   4489 *	Periodic address status verification
   4490 */
   4491
   4492static void addrconf_verify_rtnl(struct net *net)
   4493{
   4494	unsigned long now, next, next_sec, next_sched;
   4495	struct inet6_ifaddr *ifp;
   4496	int i;
   4497
   4498	ASSERT_RTNL();
   4499
   4500	rcu_read_lock_bh();
   4501	now = jiffies;
   4502	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
   4503
   4504	cancel_delayed_work(&net->ipv6.addr_chk_work);
   4505
   4506	for (i = 0; i < IN6_ADDR_HSIZE; i++) {
   4507restart:
   4508		hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) {
   4509			unsigned long age;
   4510
   4511			/* When setting preferred_lft to a value not zero or
   4512			 * infinity, while valid_lft is infinity
   4513			 * IFA_F_PERMANENT has a non-infinity life time.
   4514			 */
   4515			if ((ifp->flags & IFA_F_PERMANENT) &&
   4516			    (ifp->prefered_lft == INFINITY_LIFE_TIME))
   4517				continue;
   4518
   4519			spin_lock(&ifp->lock);
   4520			/* We try to batch several events at once. */
   4521			age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
   4522
   4523			if (ifp->valid_lft != INFINITY_LIFE_TIME &&
   4524			    age >= ifp->valid_lft) {
   4525				spin_unlock(&ifp->lock);
   4526				in6_ifa_hold(ifp);
   4527				rcu_read_unlock_bh();
   4528				ipv6_del_addr(ifp);
   4529				rcu_read_lock_bh();
   4530				goto restart;
   4531			} else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
   4532				spin_unlock(&ifp->lock);
   4533				continue;
   4534			} else if (age >= ifp->prefered_lft) {
   4535				/* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
   4536				int deprecate = 0;
   4537
   4538				if (!(ifp->flags&IFA_F_DEPRECATED)) {
   4539					deprecate = 1;
   4540					ifp->flags |= IFA_F_DEPRECATED;
   4541				}
   4542
   4543				if ((ifp->valid_lft != INFINITY_LIFE_TIME) &&
   4544				    (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)))
   4545					next = ifp->tstamp + ifp->valid_lft * HZ;
   4546
   4547				spin_unlock(&ifp->lock);
   4548
   4549				if (deprecate) {
   4550					in6_ifa_hold(ifp);
   4551
   4552					ipv6_ifa_notify(0, ifp);
   4553					in6_ifa_put(ifp);
   4554					goto restart;
   4555				}
   4556			} else if ((ifp->flags&IFA_F_TEMPORARY) &&
   4557				   !(ifp->flags&IFA_F_TENTATIVE)) {
   4558				unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
   4559					ifp->idev->cnf.dad_transmits *
   4560					max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ;
   4561
   4562				if (age >= ifp->prefered_lft - regen_advance) {
   4563					struct inet6_ifaddr *ifpub = ifp->ifpub;
   4564					if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
   4565						next = ifp->tstamp + ifp->prefered_lft * HZ;
   4566					if (!ifp->regen_count && ifpub) {
   4567						ifp->regen_count++;
   4568						in6_ifa_hold(ifp);
   4569						in6_ifa_hold(ifpub);
   4570						spin_unlock(&ifp->lock);
   4571
   4572						spin_lock(&ifpub->lock);
   4573						ifpub->regen_count = 0;
   4574						spin_unlock(&ifpub->lock);
   4575						rcu_read_unlock_bh();
   4576						ipv6_create_tempaddr(ifpub, true);
   4577						in6_ifa_put(ifpub);
   4578						in6_ifa_put(ifp);
   4579						rcu_read_lock_bh();
   4580						goto restart;
   4581					}
   4582				} else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
   4583					next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
   4584				spin_unlock(&ifp->lock);
   4585			} else {
   4586				/* ifp->prefered_lft <= ifp->valid_lft */
   4587				if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
   4588					next = ifp->tstamp + ifp->prefered_lft * HZ;
   4589				spin_unlock(&ifp->lock);
   4590			}
   4591		}
   4592	}
   4593
   4594	next_sec = round_jiffies_up(next);
   4595	next_sched = next;
   4596
   4597	/* If rounded timeout is accurate enough, accept it. */
   4598	if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
   4599		next_sched = next_sec;
   4600
   4601	/* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
   4602	if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
   4603		next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
   4604
   4605	pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
   4606		 now, next, next_sec, next_sched);
   4607	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now);
   4608	rcu_read_unlock_bh();
   4609}
   4610
   4611static void addrconf_verify_work(struct work_struct *w)
   4612{
   4613	struct net *net = container_of(to_delayed_work(w), struct net,
   4614				       ipv6.addr_chk_work);
   4615
   4616	rtnl_lock();
   4617	addrconf_verify_rtnl(net);
   4618	rtnl_unlock();
   4619}
   4620
   4621static void addrconf_verify(struct net *net)
   4622{
   4623	mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0);
   4624}
   4625
   4626static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local,
   4627				     struct in6_addr **peer_pfx)
   4628{
   4629	struct in6_addr *pfx = NULL;
   4630
   4631	*peer_pfx = NULL;
   4632
   4633	if (addr)
   4634		pfx = nla_data(addr);
   4635
   4636	if (local) {
   4637		if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
   4638			*peer_pfx = pfx;
   4639		pfx = nla_data(local);
   4640	}
   4641
   4642	return pfx;
   4643}
   4644
   4645static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
   4646	[IFA_ADDRESS]		= { .len = sizeof(struct in6_addr) },
   4647	[IFA_LOCAL]		= { .len = sizeof(struct in6_addr) },
   4648	[IFA_CACHEINFO]		= { .len = sizeof(struct ifa_cacheinfo) },
   4649	[IFA_FLAGS]		= { .len = sizeof(u32) },
   4650	[IFA_RT_PRIORITY]	= { .len = sizeof(u32) },
   4651	[IFA_TARGET_NETNSID]	= { .type = NLA_S32 },
   4652	[IFA_PROTO]		= { .type = NLA_U8 },
   4653};
   4654
   4655static int
   4656inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh,
   4657		  struct netlink_ext_ack *extack)
   4658{
   4659	struct net *net = sock_net(skb->sk);
   4660	struct ifaddrmsg *ifm;
   4661	struct nlattr *tb[IFA_MAX+1];
   4662	struct in6_addr *pfx, *peer_pfx;
   4663	u32 ifa_flags;
   4664	int err;
   4665
   4666	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
   4667				     ifa_ipv6_policy, extack);
   4668	if (err < 0)
   4669		return err;
   4670
   4671	ifm = nlmsg_data(nlh);
   4672	pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
   4673	if (!pfx)
   4674		return -EINVAL;
   4675
   4676	ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags;
   4677
   4678	/* We ignore other flags so far. */
   4679	ifa_flags &= IFA_F_MANAGETEMPADDR;
   4680
   4681	return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx,
   4682			      ifm->ifa_prefixlen);
   4683}
   4684
   4685static int modify_prefix_route(struct inet6_ifaddr *ifp,
   4686			       unsigned long expires, u32 flags,
   4687			       bool modify_peer)
   4688{
   4689	struct fib6_info *f6i;
   4690	u32 prio;
   4691
   4692	f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
   4693					ifp->prefix_len,
   4694					ifp->idev->dev, 0, RTF_DEFAULT, true);
   4695	if (!f6i)
   4696		return -ENOENT;
   4697
   4698	prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF;
   4699	if (f6i->fib6_metric != prio) {
   4700		/* delete old one */
   4701		ip6_del_rt(dev_net(ifp->idev->dev), f6i, false);
   4702
   4703		/* add new one */
   4704		addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr,
   4705				      ifp->prefix_len,
   4706				      ifp->rt_priority, ifp->idev->dev,
   4707				      expires, flags, GFP_KERNEL);
   4708	} else {
   4709		if (!expires)
   4710			fib6_clean_expires(f6i);
   4711		else
   4712			fib6_set_expires(f6i, expires);
   4713
   4714		fib6_info_release(f6i);
   4715	}
   4716
   4717	return 0;
   4718}
   4719
   4720static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp,
   4721			     struct ifa6_config *cfg)
   4722{
   4723	u32 flags;
   4724	clock_t expires;
   4725	unsigned long timeout;
   4726	bool was_managetempaddr;
   4727	bool had_prefixroute;
   4728	bool new_peer = false;
   4729
   4730	ASSERT_RTNL();
   4731
   4732	if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft)
   4733		return -EINVAL;
   4734
   4735	if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR &&
   4736	    (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64))
   4737		return -EINVAL;
   4738
   4739	if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED)
   4740		cfg->ifa_flags &= ~IFA_F_OPTIMISTIC;
   4741
   4742	timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ);
   4743	if (addrconf_finite_timeout(timeout)) {
   4744		expires = jiffies_to_clock_t(timeout * HZ);
   4745		cfg->valid_lft = timeout;
   4746		flags = RTF_EXPIRES;
   4747	} else {
   4748		expires = 0;
   4749		flags = 0;
   4750		cfg->ifa_flags |= IFA_F_PERMANENT;
   4751	}
   4752
   4753	timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ);
   4754	if (addrconf_finite_timeout(timeout)) {
   4755		if (timeout == 0)
   4756			cfg->ifa_flags |= IFA_F_DEPRECATED;
   4757		cfg->preferred_lft = timeout;
   4758	}
   4759
   4760	if (cfg->peer_pfx &&
   4761	    memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) {
   4762		if (!ipv6_addr_any(&ifp->peer_addr))
   4763			cleanup_prefix_route(ifp, expires, true, true);
   4764		new_peer = true;
   4765	}
   4766
   4767	spin_lock_bh(&ifp->lock);
   4768	was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR;
   4769	had_prefixroute = ifp->flags & IFA_F_PERMANENT &&
   4770			  !(ifp->flags & IFA_F_NOPREFIXROUTE);
   4771	ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD |
   4772			IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR |
   4773			IFA_F_NOPREFIXROUTE);
   4774	ifp->flags |= cfg->ifa_flags;
   4775	ifp->tstamp = jiffies;
   4776	ifp->valid_lft = cfg->valid_lft;
   4777	ifp->prefered_lft = cfg->preferred_lft;
   4778	ifp->ifa_proto = cfg->ifa_proto;
   4779
   4780	if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority)
   4781		ifp->rt_priority = cfg->rt_priority;
   4782
   4783	if (new_peer)
   4784		ifp->peer_addr = *cfg->peer_pfx;
   4785
   4786	spin_unlock_bh(&ifp->lock);
   4787	if (!(ifp->flags&IFA_F_TENTATIVE))
   4788		ipv6_ifa_notify(0, ifp);
   4789
   4790	if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) {
   4791		int rc = -ENOENT;
   4792
   4793		if (had_prefixroute)
   4794			rc = modify_prefix_route(ifp, expires, flags, false);
   4795
   4796		/* prefix route could have been deleted; if so restore it */
   4797		if (rc == -ENOENT) {
   4798			addrconf_prefix_route(&ifp->addr, ifp->prefix_len,
   4799					      ifp->rt_priority, ifp->idev->dev,
   4800					      expires, flags, GFP_KERNEL);
   4801		}
   4802
   4803		if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr))
   4804			rc = modify_prefix_route(ifp, expires, flags, true);
   4805
   4806		if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) {
   4807			addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len,
   4808					      ifp->rt_priority, ifp->idev->dev,
   4809					      expires, flags, GFP_KERNEL);
   4810		}
   4811	} else if (had_prefixroute) {
   4812		enum cleanup_prefix_rt_t action;
   4813		unsigned long rt_expires;
   4814
   4815		write_lock_bh(&ifp->idev->lock);
   4816		action = check_cleanup_prefix_route(ifp, &rt_expires);
   4817		write_unlock_bh(&ifp->idev->lock);
   4818
   4819		if (action != CLEANUP_PREFIX_RT_NOP) {
   4820			cleanup_prefix_route(ifp, rt_expires,
   4821				action == CLEANUP_PREFIX_RT_DEL, false);
   4822		}
   4823	}
   4824
   4825	if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) {
   4826		if (was_managetempaddr &&
   4827		    !(ifp->flags & IFA_F_MANAGETEMPADDR)) {
   4828			cfg->valid_lft = 0;
   4829			cfg->preferred_lft = 0;
   4830		}
   4831		manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft,
   4832				 cfg->preferred_lft, !was_managetempaddr,
   4833				 jiffies);
   4834	}
   4835
   4836	addrconf_verify_rtnl(net);
   4837
   4838	return 0;
   4839}
   4840
   4841static int
   4842inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
   4843		  struct netlink_ext_ack *extack)
   4844{
   4845	struct net *net = sock_net(skb->sk);
   4846	struct ifaddrmsg *ifm;
   4847	struct nlattr *tb[IFA_MAX+1];
   4848	struct in6_addr *peer_pfx;
   4849	struct inet6_ifaddr *ifa;
   4850	struct net_device *dev;
   4851	struct inet6_dev *idev;
   4852	struct ifa6_config cfg;
   4853	int err;
   4854
   4855	err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
   4856				     ifa_ipv6_policy, extack);
   4857	if (err < 0)
   4858		return err;
   4859
   4860	memset(&cfg, 0, sizeof(cfg));
   4861
   4862	ifm = nlmsg_data(nlh);
   4863	cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx);
   4864	if (!cfg.pfx)
   4865		return -EINVAL;
   4866
   4867	cfg.peer_pfx = peer_pfx;
   4868	cfg.plen = ifm->ifa_prefixlen;
   4869	if (tb[IFA_RT_PRIORITY])
   4870		cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]);
   4871
   4872	if (tb[IFA_PROTO])
   4873		cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]);
   4874
   4875	cfg.valid_lft = INFINITY_LIFE_TIME;
   4876	cfg.preferred_lft = INFINITY_LIFE_TIME;
   4877
   4878	if (tb[IFA_CACHEINFO]) {
   4879		struct ifa_cacheinfo *ci;
   4880
   4881		ci = nla_data(tb[IFA_CACHEINFO]);
   4882		cfg.valid_lft = ci->ifa_valid;
   4883		cfg.preferred_lft = ci->ifa_prefered;
   4884	}
   4885
   4886	dev =  __dev_get_by_index(net, ifm->ifa_index);
   4887	if (!dev)
   4888		return -ENODEV;
   4889
   4890	if (tb[IFA_FLAGS])
   4891		cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]);
   4892	else
   4893		cfg.ifa_flags = ifm->ifa_flags;
   4894
   4895	/* We ignore other flags so far. */
   4896	cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS |
   4897			 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE |
   4898			 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC;
   4899
   4900	idev = ipv6_find_idev(dev);
   4901	if (IS_ERR(idev))
   4902		return PTR_ERR(idev);
   4903
   4904	if (!ipv6_allow_optimistic_dad(net, idev))
   4905		cfg.ifa_flags &= ~IFA_F_OPTIMISTIC;
   4906
   4907	if (cfg.ifa_flags & IFA_F_NODAD &&
   4908	    cfg.ifa_flags & IFA_F_OPTIMISTIC) {
   4909		NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive");
   4910		return -EINVAL;
   4911	}
   4912
   4913	ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1);
   4914	if (!ifa) {
   4915		/*
   4916		 * It would be best to check for !NLM_F_CREATE here but
   4917		 * userspace already relies on not having to provide this.
   4918		 */
   4919		return inet6_addr_add(net, ifm->ifa_index, &cfg, extack);
   4920	}
   4921
   4922	if (nlh->nlmsg_flags & NLM_F_EXCL ||
   4923	    !(nlh->nlmsg_flags & NLM_F_REPLACE))
   4924		err = -EEXIST;
   4925	else
   4926		err = inet6_addr_modify(net, ifa, &cfg);
   4927
   4928	in6_ifa_put(ifa);
   4929
   4930	return err;
   4931}
   4932
   4933static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags,
   4934			  u8 scope, int ifindex)
   4935{
   4936	struct ifaddrmsg *ifm;
   4937
   4938	ifm = nlmsg_data(nlh);
   4939	ifm->ifa_family = AF_INET6;
   4940	ifm->ifa_prefixlen = prefixlen;
   4941	ifm->ifa_flags = flags;
   4942	ifm->ifa_scope = scope;
   4943	ifm->ifa_index = ifindex;
   4944}
   4945
   4946static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
   4947			 unsigned long tstamp, u32 preferred, u32 valid)
   4948{
   4949	struct ifa_cacheinfo ci;
   4950
   4951	ci.cstamp = cstamp_delta(cstamp);
   4952	ci.tstamp = cstamp_delta(tstamp);
   4953	ci.ifa_prefered = preferred;
   4954	ci.ifa_valid = valid;
   4955
   4956	return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
   4957}
   4958
   4959static inline int rt_scope(int ifa_scope)
   4960{
   4961	if (ifa_scope & IFA_HOST)
   4962		return RT_SCOPE_HOST;
   4963	else if (ifa_scope & IFA_LINK)
   4964		return RT_SCOPE_LINK;
   4965	else if (ifa_scope & IFA_SITE)
   4966		return RT_SCOPE_SITE;
   4967	else
   4968		return RT_SCOPE_UNIVERSE;
   4969}
   4970
   4971static inline int inet6_ifaddr_msgsize(void)
   4972{
   4973	return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
   4974	       + nla_total_size(16) /* IFA_LOCAL */
   4975	       + nla_total_size(16) /* IFA_ADDRESS */
   4976	       + nla_total_size(sizeof(struct ifa_cacheinfo))
   4977	       + nla_total_size(4)  /* IFA_FLAGS */
   4978	       + nla_total_size(1)  /* IFA_PROTO */
   4979	       + nla_total_size(4)  /* IFA_RT_PRIORITY */;
   4980}
   4981
   4982enum addr_type_t {
   4983	UNICAST_ADDR,
   4984	MULTICAST_ADDR,
   4985	ANYCAST_ADDR,
   4986};
   4987
   4988struct inet6_fill_args {
   4989	u32 portid;
   4990	u32 seq;
   4991	int event;
   4992	unsigned int flags;
   4993	int netnsid;
   4994	int ifindex;
   4995	enum addr_type_t type;
   4996};
   4997
   4998static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
   4999			     struct inet6_fill_args *args)
   5000{
   5001	struct nlmsghdr  *nlh;
   5002	u32 preferred, valid;
   5003
   5004	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
   5005			sizeof(struct ifaddrmsg), args->flags);
   5006	if (!nlh)
   5007		return -EMSGSIZE;
   5008
   5009	put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
   5010		      ifa->idev->dev->ifindex);
   5011
   5012	if (args->netnsid >= 0 &&
   5013	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid))
   5014		goto error;
   5015
   5016	spin_lock_bh(&ifa->lock);
   5017	if (!((ifa->flags&IFA_F_PERMANENT) &&
   5018	      (ifa->prefered_lft == INFINITY_LIFE_TIME))) {
   5019		preferred = ifa->prefered_lft;
   5020		valid = ifa->valid_lft;
   5021		if (preferred != INFINITY_LIFE_TIME) {
   5022			long tval = (jiffies - ifa->tstamp)/HZ;
   5023			if (preferred > tval)
   5024				preferred -= tval;
   5025			else
   5026				preferred = 0;
   5027			if (valid != INFINITY_LIFE_TIME) {
   5028				if (valid > tval)
   5029					valid -= tval;
   5030				else
   5031					valid = 0;
   5032			}
   5033		}
   5034	} else {
   5035		preferred = INFINITY_LIFE_TIME;
   5036		valid = INFINITY_LIFE_TIME;
   5037	}
   5038	spin_unlock_bh(&ifa->lock);
   5039
   5040	if (!ipv6_addr_any(&ifa->peer_addr)) {
   5041		if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 ||
   5042		    nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0)
   5043			goto error;
   5044	} else
   5045		if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0)
   5046			goto error;
   5047
   5048	if (ifa->rt_priority &&
   5049	    nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority))
   5050		goto error;
   5051
   5052	if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
   5053		goto error;
   5054
   5055	if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0)
   5056		goto error;
   5057
   5058	if (ifa->ifa_proto &&
   5059	    nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto))
   5060		goto error;
   5061
   5062	nlmsg_end(skb, nlh);
   5063	return 0;
   5064
   5065error:
   5066	nlmsg_cancel(skb, nlh);
   5067	return -EMSGSIZE;
   5068}
   5069
   5070static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
   5071			       struct inet6_fill_args *args)
   5072{
   5073	struct nlmsghdr  *nlh;
   5074	u8 scope = RT_SCOPE_UNIVERSE;
   5075	int ifindex = ifmca->idev->dev->ifindex;
   5076
   5077	if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
   5078		scope = RT_SCOPE_SITE;
   5079
   5080	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
   5081			sizeof(struct ifaddrmsg), args->flags);
   5082	if (!nlh)
   5083		return -EMSGSIZE;
   5084
   5085	if (args->netnsid >= 0 &&
   5086	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
   5087		nlmsg_cancel(skb, nlh);
   5088		return -EMSGSIZE;
   5089	}
   5090
   5091	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
   5092	if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 ||
   5093	    put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
   5094			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
   5095		nlmsg_cancel(skb, nlh);
   5096		return -EMSGSIZE;
   5097	}
   5098
   5099	nlmsg_end(skb, nlh);
   5100	return 0;
   5101}
   5102
   5103static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
   5104			       struct inet6_fill_args *args)
   5105{
   5106	struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt);
   5107	int ifindex = dev ? dev->ifindex : 1;
   5108	struct nlmsghdr  *nlh;
   5109	u8 scope = RT_SCOPE_UNIVERSE;
   5110
   5111	if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
   5112		scope = RT_SCOPE_SITE;
   5113
   5114	nlh = nlmsg_put(skb, args->portid, args->seq, args->event,
   5115			sizeof(struct ifaddrmsg), args->flags);
   5116	if (!nlh)
   5117		return -EMSGSIZE;
   5118
   5119	if (args->netnsid >= 0 &&
   5120	    nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) {
   5121		nlmsg_cancel(skb, nlh);
   5122		return -EMSGSIZE;
   5123	}
   5124
   5125	put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
   5126	if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 ||
   5127	    put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
   5128			  INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
   5129		nlmsg_cancel(skb, nlh);
   5130		return -EMSGSIZE;
   5131	}
   5132
   5133	nlmsg_end(skb, nlh);
   5134	return 0;
   5135}
   5136
   5137/* called with rcu_read_lock() */
   5138static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
   5139			  struct netlink_callback *cb, int s_ip_idx,
   5140			  struct inet6_fill_args *fillargs)
   5141{
   5142	struct ifmcaddr6 *ifmca;
   5143	struct ifacaddr6 *ifaca;
   5144	int ip_idx = 0;
   5145	int err = 1;
   5146
   5147	read_lock_bh(&idev->lock);
   5148	switch (fillargs->type) {
   5149	case UNICAST_ADDR: {
   5150		struct inet6_ifaddr *ifa;
   5151		fillargs->event = RTM_NEWADDR;
   5152
   5153		/* unicast address incl. temp addr */
   5154		list_for_each_entry(ifa, &idev->addr_list, if_list) {
   5155			if (ip_idx < s_ip_idx)
   5156				goto next;
   5157			err = inet6_fill_ifaddr(skb, ifa, fillargs);
   5158			if (err < 0)
   5159				break;
   5160			nl_dump_check_consistent(cb, nlmsg_hdr(skb));
   5161next:
   5162			ip_idx++;
   5163		}
   5164		break;
   5165	}
   5166	case MULTICAST_ADDR:
   5167		read_unlock_bh(&idev->lock);
   5168		fillargs->event = RTM_GETMULTICAST;
   5169
   5170		/* multicast address */
   5171		for (ifmca = rtnl_dereference(idev->mc_list);
   5172		     ifmca;
   5173		     ifmca = rtnl_dereference(ifmca->next), ip_idx++) {
   5174			if (ip_idx < s_ip_idx)
   5175				continue;
   5176			err = inet6_fill_ifmcaddr(skb, ifmca, fillargs);
   5177			if (err < 0)
   5178				break;
   5179		}
   5180		read_lock_bh(&idev->lock);
   5181		break;
   5182	case ANYCAST_ADDR:
   5183		fillargs->event = RTM_GETANYCAST;
   5184		/* anycast address */
   5185		for (ifaca = idev->ac_list; ifaca;
   5186		     ifaca = ifaca->aca_next, ip_idx++) {
   5187			if (ip_idx < s_ip_idx)
   5188				continue;
   5189			err = inet6_fill_ifacaddr(skb, ifaca, fillargs);
   5190			if (err < 0)
   5191				break;
   5192		}
   5193		break;
   5194	default:
   5195		break;
   5196	}
   5197	read_unlock_bh(&idev->lock);
   5198	cb->args[2] = ip_idx;
   5199	return err;
   5200}
   5201
   5202static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh,
   5203				       struct inet6_fill_args *fillargs,
   5204				       struct net **tgt_net, struct sock *sk,
   5205				       struct netlink_callback *cb)
   5206{
   5207	struct netlink_ext_ack *extack = cb->extack;
   5208	struct nlattr *tb[IFA_MAX+1];
   5209	struct ifaddrmsg *ifm;
   5210	int err, i;
   5211
   5212	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
   5213		NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request");
   5214		return -EINVAL;
   5215	}
   5216
   5217	ifm = nlmsg_data(nlh);
   5218	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
   5219		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request");
   5220		return -EINVAL;
   5221	}
   5222
   5223	fillargs->ifindex = ifm->ifa_index;
   5224	if (fillargs->ifindex) {
   5225		cb->answer_flags |= NLM_F_DUMP_FILTERED;
   5226		fillargs->flags |= NLM_F_DUMP_FILTERED;
   5227	}
   5228
   5229	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
   5230					    ifa_ipv6_policy, extack);
   5231	if (err < 0)
   5232		return err;
   5233
   5234	for (i = 0; i <= IFA_MAX; ++i) {
   5235		if (!tb[i])
   5236			continue;
   5237
   5238		if (i == IFA_TARGET_NETNSID) {
   5239			struct net *net;
   5240
   5241			fillargs->netnsid = nla_get_s32(tb[i]);
   5242			net = rtnl_get_net_ns_capable(sk, fillargs->netnsid);
   5243			if (IS_ERR(net)) {
   5244				fillargs->netnsid = -1;
   5245				NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id");
   5246				return PTR_ERR(net);
   5247			}
   5248			*tgt_net = net;
   5249		} else {
   5250			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request");
   5251			return -EINVAL;
   5252		}
   5253	}
   5254
   5255	return 0;
   5256}
   5257
   5258static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
   5259			   enum addr_type_t type)
   5260{
   5261	const struct nlmsghdr *nlh = cb->nlh;
   5262	struct inet6_fill_args fillargs = {
   5263		.portid = NETLINK_CB(cb->skb).portid,
   5264		.seq = cb->nlh->nlmsg_seq,
   5265		.flags = NLM_F_MULTI,
   5266		.netnsid = -1,
   5267		.type = type,
   5268	};
   5269	struct net *tgt_net = sock_net(skb->sk);
   5270	int idx, s_idx, s_ip_idx;
   5271	int h, s_h;
   5272	struct net_device *dev;
   5273	struct inet6_dev *idev;
   5274	struct hlist_head *head;
   5275	int err = 0;
   5276
   5277	s_h = cb->args[0];
   5278	s_idx = idx = cb->args[1];
   5279	s_ip_idx = cb->args[2];
   5280
   5281	if (cb->strict_check) {
   5282		err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net,
   5283						  skb->sk, cb);
   5284		if (err < 0)
   5285			goto put_tgt_net;
   5286
   5287		err = 0;
   5288		if (fillargs.ifindex) {
   5289			dev = __dev_get_by_index(tgt_net, fillargs.ifindex);
   5290			if (!dev) {
   5291				err = -ENODEV;
   5292				goto put_tgt_net;
   5293			}
   5294			idev = __in6_dev_get(dev);
   5295			if (idev) {
   5296				err = in6_dump_addrs(idev, skb, cb, s_ip_idx,
   5297						     &fillargs);
   5298				if (err > 0)
   5299					err = 0;
   5300			}
   5301			goto put_tgt_net;
   5302		}
   5303	}
   5304
   5305	rcu_read_lock();
   5306	cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq;
   5307	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
   5308		idx = 0;
   5309		head = &tgt_net->dev_index_head[h];
   5310		hlist_for_each_entry_rcu(dev, head, index_hlist) {
   5311			if (idx < s_idx)
   5312				goto cont;
   5313			if (h > s_h || idx > s_idx)
   5314				s_ip_idx = 0;
   5315			idev = __in6_dev_get(dev);
   5316			if (!idev)
   5317				goto cont;
   5318
   5319			if (in6_dump_addrs(idev, skb, cb, s_ip_idx,
   5320					   &fillargs) < 0)
   5321				goto done;
   5322cont:
   5323			idx++;
   5324		}
   5325	}
   5326done:
   5327	rcu_read_unlock();
   5328	cb->args[0] = h;
   5329	cb->args[1] = idx;
   5330put_tgt_net:
   5331	if (fillargs.netnsid >= 0)
   5332		put_net(tgt_net);
   5333
   5334	return skb->len ? : err;
   5335}
   5336
   5337static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
   5338{
   5339	enum addr_type_t type = UNICAST_ADDR;
   5340
   5341	return inet6_dump_addr(skb, cb, type);
   5342}
   5343
   5344static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
   5345{
   5346	enum addr_type_t type = MULTICAST_ADDR;
   5347
   5348	return inet6_dump_addr(skb, cb, type);
   5349}
   5350
   5351
   5352static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
   5353{
   5354	enum addr_type_t type = ANYCAST_ADDR;
   5355
   5356	return inet6_dump_addr(skb, cb, type);
   5357}
   5358
   5359static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb,
   5360				       const struct nlmsghdr *nlh,
   5361				       struct nlattr **tb,
   5362				       struct netlink_ext_ack *extack)
   5363{
   5364	struct ifaddrmsg *ifm;
   5365	int i, err;
   5366
   5367	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
   5368		NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request");
   5369		return -EINVAL;
   5370	}
   5371
   5372	if (!netlink_strict_get_check(skb))
   5373		return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX,
   5374					      ifa_ipv6_policy, extack);
   5375
   5376	ifm = nlmsg_data(nlh);
   5377	if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) {
   5378		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request");
   5379		return -EINVAL;
   5380	}
   5381
   5382	err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX,
   5383					    ifa_ipv6_policy, extack);
   5384	if (err)
   5385		return err;
   5386
   5387	for (i = 0; i <= IFA_MAX; i++) {
   5388		if (!tb[i])
   5389			continue;
   5390
   5391		switch (i) {
   5392		case IFA_TARGET_NETNSID:
   5393		case IFA_ADDRESS:
   5394		case IFA_LOCAL:
   5395			break;
   5396		default:
   5397			NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request");
   5398			return -EINVAL;
   5399		}
   5400	}
   5401
   5402	return 0;
   5403}
   5404
   5405static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
   5406			     struct netlink_ext_ack *extack)
   5407{
   5408	struct net *tgt_net = sock_net(in_skb->sk);
   5409	struct inet6_fill_args fillargs = {
   5410		.portid = NETLINK_CB(in_skb).portid,
   5411		.seq = nlh->nlmsg_seq,
   5412		.event = RTM_NEWADDR,
   5413		.flags = 0,
   5414		.netnsid = -1,
   5415	};
   5416	struct ifaddrmsg *ifm;
   5417	struct nlattr *tb[IFA_MAX+1];
   5418	struct in6_addr *addr = NULL, *peer;
   5419	struct net_device *dev = NULL;
   5420	struct inet6_ifaddr *ifa;
   5421	struct sk_buff *skb;
   5422	int err;
   5423
   5424	err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack);
   5425	if (err < 0)
   5426		return err;
   5427
   5428	if (tb[IFA_TARGET_NETNSID]) {
   5429		fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]);
   5430
   5431		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk,
   5432						  fillargs.netnsid);
   5433		if (IS_ERR(tgt_net))
   5434			return PTR_ERR(tgt_net);
   5435	}
   5436
   5437	addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer);
   5438	if (!addr)
   5439		return -EINVAL;
   5440
   5441	ifm = nlmsg_data(nlh);
   5442	if (ifm->ifa_index)
   5443		dev = dev_get_by_index(tgt_net, ifm->ifa_index);
   5444
   5445	ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1);
   5446	if (!ifa) {
   5447		err = -EADDRNOTAVAIL;
   5448		goto errout;
   5449	}
   5450
   5451	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
   5452	if (!skb) {
   5453		err = -ENOBUFS;
   5454		goto errout_ifa;
   5455	}
   5456
   5457	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
   5458	if (err < 0) {
   5459		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
   5460		WARN_ON(err == -EMSGSIZE);
   5461		kfree_skb(skb);
   5462		goto errout_ifa;
   5463	}
   5464	err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid);
   5465errout_ifa:
   5466	in6_ifa_put(ifa);
   5467errout:
   5468	dev_put(dev);
   5469	if (fillargs.netnsid >= 0)
   5470		put_net(tgt_net);
   5471
   5472	return err;
   5473}
   5474
   5475static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
   5476{
   5477	struct sk_buff *skb;
   5478	struct net *net = dev_net(ifa->idev->dev);
   5479	struct inet6_fill_args fillargs = {
   5480		.portid = 0,
   5481		.seq = 0,
   5482		.event = event,
   5483		.flags = 0,
   5484		.netnsid = -1,
   5485	};
   5486	int err = -ENOBUFS;
   5487
   5488	skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
   5489	if (!skb)
   5490		goto errout;
   5491
   5492	err = inet6_fill_ifaddr(skb, ifa, &fillargs);
   5493	if (err < 0) {
   5494		/* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
   5495		WARN_ON(err == -EMSGSIZE);
   5496		kfree_skb(skb);
   5497		goto errout;
   5498	}
   5499	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
   5500	return;
   5501errout:
   5502	if (err < 0)
   5503		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
   5504}
   5505
   5506static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
   5507				__s32 *array, int bytes)
   5508{
   5509	BUG_ON(bytes < (DEVCONF_MAX * 4));
   5510
   5511	memset(array, 0, bytes);
   5512	array[DEVCONF_FORWARDING] = cnf->forwarding;
   5513	array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
   5514	array[DEVCONF_MTU6] = cnf->mtu6;
   5515	array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
   5516	array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
   5517	array[DEVCONF_AUTOCONF] = cnf->autoconf;
   5518	array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
   5519	array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
   5520	array[DEVCONF_RTR_SOLICIT_INTERVAL] =
   5521		jiffies_to_msecs(cnf->rtr_solicit_interval);
   5522	array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] =
   5523		jiffies_to_msecs(cnf->rtr_solicit_max_interval);
   5524	array[DEVCONF_RTR_SOLICIT_DELAY] =
   5525		jiffies_to_msecs(cnf->rtr_solicit_delay);
   5526	array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
   5527	array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] =
   5528		jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval);
   5529	array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] =
   5530		jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval);
   5531	array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
   5532	array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
   5533	array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
   5534	array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
   5535	array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
   5536	array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
   5537	array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
   5538	array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric;
   5539	array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
   5540	array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
   5541#ifdef CONFIG_IPV6_ROUTER_PREF
   5542	array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
   5543	array[DEVCONF_RTR_PROBE_INTERVAL] =
   5544		jiffies_to_msecs(cnf->rtr_probe_interval);
   5545#ifdef CONFIG_IPV6_ROUTE_INFO
   5546	array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen;
   5547	array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
   5548#endif
   5549#endif
   5550	array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
   5551	array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
   5552#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   5553	array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
   5554	array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic;
   5555#endif
   5556#ifdef CONFIG_IPV6_MROUTE
   5557	array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding);
   5558#endif
   5559	array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
   5560	array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
   5561	array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
   5562	array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify;
   5563	array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc;
   5564	array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local;
   5565	array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu;
   5566	array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown;
   5567	/* we omit DEVCONF_STABLE_SECRET for now */
   5568	array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only;
   5569	array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast;
   5570	array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na;
   5571	array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down;
   5572	array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled;
   5573#ifdef CONFIG_IPV6_SEG6_HMAC
   5574	array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac;
   5575#endif
   5576	array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad;
   5577	array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode;
   5578	array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy;
   5579	array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass;
   5580	array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled;
   5581	array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled;
   5582	array[DEVCONF_IOAM6_ID] = cnf->ioam6_id;
   5583	array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide;
   5584	array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier;
   5585	array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na;
   5586}
   5587
   5588static inline size_t inet6_ifla6_size(void)
   5589{
   5590	return nla_total_size(4) /* IFLA_INET6_FLAGS */
   5591	     + nla_total_size(sizeof(struct ifla_cacheinfo))
   5592	     + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
   5593	     + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
   5594	     + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
   5595	     + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */
   5596	     + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */
   5597	     + nla_total_size(4) /* IFLA_INET6_RA_MTU */
   5598	     + 0;
   5599}
   5600
   5601static inline size_t inet6_if_nlmsg_size(void)
   5602{
   5603	return NLMSG_ALIGN(sizeof(struct ifinfomsg))
   5604	       + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
   5605	       + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
   5606	       + nla_total_size(4) /* IFLA_MTU */
   5607	       + nla_total_size(4) /* IFLA_LINK */
   5608	       + nla_total_size(1) /* IFLA_OPERSTATE */
   5609	       + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
   5610}
   5611
   5612static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
   5613					int bytes)
   5614{
   5615	int i;
   5616	int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX;
   5617	BUG_ON(pad < 0);
   5618
   5619	/* Use put_unaligned() because stats may not be aligned for u64. */
   5620	put_unaligned(ICMP6_MIB_MAX, &stats[0]);
   5621	for (i = 1; i < ICMP6_MIB_MAX; i++)
   5622		put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
   5623
   5624	memset(&stats[ICMP6_MIB_MAX], 0, pad);
   5625}
   5626
   5627static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib,
   5628					int bytes, size_t syncpoff)
   5629{
   5630	int i, c;
   5631	u64 buff[IPSTATS_MIB_MAX];
   5632	int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX;
   5633
   5634	BUG_ON(pad < 0);
   5635
   5636	memset(buff, 0, sizeof(buff));
   5637	buff[0] = IPSTATS_MIB_MAX;
   5638
   5639	for_each_possible_cpu(c) {
   5640		for (i = 1; i < IPSTATS_MIB_MAX; i++)
   5641			buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff);
   5642	}
   5643
   5644	memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64));
   5645	memset(&stats[IPSTATS_MIB_MAX], 0, pad);
   5646}
   5647
   5648static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
   5649			     int bytes)
   5650{
   5651	switch (attrtype) {
   5652	case IFLA_INET6_STATS:
   5653		__snmp6_fill_stats64(stats, idev->stats.ipv6, bytes,
   5654				     offsetof(struct ipstats_mib, syncp));
   5655		break;
   5656	case IFLA_INET6_ICMP6STATS:
   5657		__snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes);
   5658		break;
   5659	}
   5660}
   5661
   5662static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev,
   5663				  u32 ext_filter_mask)
   5664{
   5665	struct nlattr *nla;
   5666	struct ifla_cacheinfo ci;
   5667
   5668	if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
   5669		goto nla_put_failure;
   5670	ci.max_reasm_len = IPV6_MAXPLEN;
   5671	ci.tstamp = cstamp_delta(idev->tstamp);
   5672	ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
   5673	ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME));
   5674	if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
   5675		goto nla_put_failure;
   5676	nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
   5677	if (!nla)
   5678		goto nla_put_failure;
   5679	ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
   5680
   5681	/* XXX - MC not implemented */
   5682
   5683	if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS)
   5684		return 0;
   5685
   5686	nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
   5687	if (!nla)
   5688		goto nla_put_failure;
   5689	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
   5690
   5691	nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
   5692	if (!nla)
   5693		goto nla_put_failure;
   5694	snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
   5695
   5696	nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr));
   5697	if (!nla)
   5698		goto nla_put_failure;
   5699	read_lock_bh(&idev->lock);
   5700	memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla));
   5701	read_unlock_bh(&idev->lock);
   5702
   5703	if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode))
   5704		goto nla_put_failure;
   5705
   5706	if (idev->ra_mtu &&
   5707	    nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu))
   5708		goto nla_put_failure;
   5709
   5710	return 0;
   5711
   5712nla_put_failure:
   5713	return -EMSGSIZE;
   5714}
   5715
   5716static size_t inet6_get_link_af_size(const struct net_device *dev,
   5717				     u32 ext_filter_mask)
   5718{
   5719	if (!__in6_dev_get(dev))
   5720		return 0;
   5721
   5722	return inet6_ifla6_size();
   5723}
   5724
   5725static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev,
   5726			      u32 ext_filter_mask)
   5727{
   5728	struct inet6_dev *idev = __in6_dev_get(dev);
   5729
   5730	if (!idev)
   5731		return -ENODATA;
   5732
   5733	if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0)
   5734		return -EMSGSIZE;
   5735
   5736	return 0;
   5737}
   5738
   5739static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token,
   5740			     struct netlink_ext_ack *extack)
   5741{
   5742	struct inet6_ifaddr *ifp;
   5743	struct net_device *dev = idev->dev;
   5744	bool clear_token, update_rs = false;
   5745	struct in6_addr ll_addr;
   5746
   5747	ASSERT_RTNL();
   5748
   5749	if (!token)
   5750		return -EINVAL;
   5751
   5752	if (dev->flags & IFF_LOOPBACK) {
   5753		NL_SET_ERR_MSG_MOD(extack, "Device is loopback");
   5754		return -EINVAL;
   5755	}
   5756
   5757	if (dev->flags & IFF_NOARP) {
   5758		NL_SET_ERR_MSG_MOD(extack,
   5759				   "Device does not do neighbour discovery");
   5760		return -EINVAL;
   5761	}
   5762
   5763	if (!ipv6_accept_ra(idev)) {
   5764		NL_SET_ERR_MSG_MOD(extack,
   5765				   "Router advertisement is disabled on device");
   5766		return -EINVAL;
   5767	}
   5768
   5769	if (idev->cnf.rtr_solicits == 0) {
   5770		NL_SET_ERR_MSG(extack,
   5771			       "Router solicitation is disabled on device");
   5772		return -EINVAL;
   5773	}
   5774
   5775	write_lock_bh(&idev->lock);
   5776
   5777	BUILD_BUG_ON(sizeof(token->s6_addr) != 16);
   5778	memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8);
   5779
   5780	write_unlock_bh(&idev->lock);
   5781
   5782	clear_token = ipv6_addr_any(token);
   5783	if (clear_token)
   5784		goto update_lft;
   5785
   5786	if (!idev->dead && (idev->if_flags & IF_READY) &&
   5787	    !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE |
   5788			     IFA_F_OPTIMISTIC)) {
   5789		/* If we're not ready, then normal ifup will take care
   5790		 * of this. Otherwise, we need to request our rs here.
   5791		 */
   5792		ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters);
   5793		update_rs = true;
   5794	}
   5795
   5796update_lft:
   5797	write_lock_bh(&idev->lock);
   5798
   5799	if (update_rs) {
   5800		idev->if_flags |= IF_RS_SENT;
   5801		idev->rs_interval = rfc3315_s14_backoff_init(
   5802			idev->cnf.rtr_solicit_interval);
   5803		idev->rs_probes = 1;
   5804		addrconf_mod_rs_timer(idev, idev->rs_interval);
   5805	}
   5806
   5807	/* Well, that's kinda nasty ... */
   5808	list_for_each_entry(ifp, &idev->addr_list, if_list) {
   5809		spin_lock(&ifp->lock);
   5810		if (ifp->tokenized) {
   5811			ifp->valid_lft = 0;
   5812			ifp->prefered_lft = 0;
   5813		}
   5814		spin_unlock(&ifp->lock);
   5815	}
   5816
   5817	write_unlock_bh(&idev->lock);
   5818	inet6_ifinfo_notify(RTM_NEWLINK, idev);
   5819	addrconf_verify_rtnl(dev_net(dev));
   5820	return 0;
   5821}
   5822
   5823static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = {
   5824	[IFLA_INET6_ADDR_GEN_MODE]	= { .type = NLA_U8 },
   5825	[IFLA_INET6_TOKEN]		= { .len = sizeof(struct in6_addr) },
   5826	[IFLA_INET6_RA_MTU]		= { .type = NLA_REJECT,
   5827					    .reject_message =
   5828						"IFLA_INET6_RA_MTU can not be set" },
   5829};
   5830
   5831static int check_addr_gen_mode(int mode)
   5832{
   5833	if (mode != IN6_ADDR_GEN_MODE_EUI64 &&
   5834	    mode != IN6_ADDR_GEN_MODE_NONE &&
   5835	    mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
   5836	    mode != IN6_ADDR_GEN_MODE_RANDOM)
   5837		return -EINVAL;
   5838	return 1;
   5839}
   5840
   5841static int check_stable_privacy(struct inet6_dev *idev, struct net *net,
   5842				int mode)
   5843{
   5844	if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY &&
   5845	    !idev->cnf.stable_secret.initialized &&
   5846	    !net->ipv6.devconf_dflt->stable_secret.initialized)
   5847		return -EINVAL;
   5848	return 1;
   5849}
   5850
   5851static int inet6_validate_link_af(const struct net_device *dev,
   5852				  const struct nlattr *nla,
   5853				  struct netlink_ext_ack *extack)
   5854{
   5855	struct nlattr *tb[IFLA_INET6_MAX + 1];
   5856	struct inet6_dev *idev = NULL;
   5857	int err;
   5858
   5859	if (dev) {
   5860		idev = __in6_dev_get(dev);
   5861		if (!idev)
   5862			return -EAFNOSUPPORT;
   5863	}
   5864
   5865	err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla,
   5866					  inet6_af_policy, extack);
   5867	if (err)
   5868		return err;
   5869
   5870	if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE])
   5871		return -EINVAL;
   5872
   5873	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
   5874		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
   5875
   5876		if (check_addr_gen_mode(mode) < 0)
   5877			return -EINVAL;
   5878		if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0)
   5879			return -EINVAL;
   5880	}
   5881
   5882	return 0;
   5883}
   5884
   5885static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla,
   5886			     struct netlink_ext_ack *extack)
   5887{
   5888	struct inet6_dev *idev = __in6_dev_get(dev);
   5889	struct nlattr *tb[IFLA_INET6_MAX + 1];
   5890	int err;
   5891
   5892	if (!idev)
   5893		return -EAFNOSUPPORT;
   5894
   5895	if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0)
   5896		return -EINVAL;
   5897
   5898	if (tb[IFLA_INET6_TOKEN]) {
   5899		err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]),
   5900					extack);
   5901		if (err)
   5902			return err;
   5903	}
   5904
   5905	if (tb[IFLA_INET6_ADDR_GEN_MODE]) {
   5906		u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]);
   5907
   5908		idev->cnf.addr_gen_mode = mode;
   5909	}
   5910
   5911	return 0;
   5912}
   5913
   5914static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
   5915			     u32 portid, u32 seq, int event, unsigned int flags)
   5916{
   5917	struct net_device *dev = idev->dev;
   5918	struct ifinfomsg *hdr;
   5919	struct nlmsghdr *nlh;
   5920	void *protoinfo;
   5921
   5922	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags);
   5923	if (!nlh)
   5924		return -EMSGSIZE;
   5925
   5926	hdr = nlmsg_data(nlh);
   5927	hdr->ifi_family = AF_INET6;
   5928	hdr->__ifi_pad = 0;
   5929	hdr->ifi_type = dev->type;
   5930	hdr->ifi_index = dev->ifindex;
   5931	hdr->ifi_flags = dev_get_flags(dev);
   5932	hdr->ifi_change = 0;
   5933
   5934	if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
   5935	    (dev->addr_len &&
   5936	     nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
   5937	    nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
   5938	    (dev->ifindex != dev_get_iflink(dev) &&
   5939	     nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) ||
   5940	    nla_put_u8(skb, IFLA_OPERSTATE,
   5941		       netif_running(dev) ? dev->operstate : IF_OPER_DOWN))
   5942		goto nla_put_failure;
   5943	protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO);
   5944	if (!protoinfo)
   5945		goto nla_put_failure;
   5946
   5947	if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0)
   5948		goto nla_put_failure;
   5949
   5950	nla_nest_end(skb, protoinfo);
   5951	nlmsg_end(skb, nlh);
   5952	return 0;
   5953
   5954nla_put_failure:
   5955	nlmsg_cancel(skb, nlh);
   5956	return -EMSGSIZE;
   5957}
   5958
   5959static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh,
   5960				   struct netlink_ext_ack *extack)
   5961{
   5962	struct ifinfomsg *ifm;
   5963
   5964	if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
   5965		NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request");
   5966		return -EINVAL;
   5967	}
   5968
   5969	if (nlmsg_attrlen(nlh, sizeof(*ifm))) {
   5970		NL_SET_ERR_MSG_MOD(extack, "Invalid data after header");
   5971		return -EINVAL;
   5972	}
   5973
   5974	ifm = nlmsg_data(nlh);
   5975	if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
   5976	    ifm->ifi_change || ifm->ifi_index) {
   5977		NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request");
   5978		return -EINVAL;
   5979	}
   5980
   5981	return 0;
   5982}
   5983
   5984static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
   5985{
   5986	struct net *net = sock_net(skb->sk);
   5987	int h, s_h;
   5988	int idx = 0, s_idx;
   5989	struct net_device *dev;
   5990	struct inet6_dev *idev;
   5991	struct hlist_head *head;
   5992
   5993	/* only requests using strict checking can pass data to
   5994	 * influence the dump
   5995	 */
   5996	if (cb->strict_check) {
   5997		int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack);
   5998
   5999		if (err < 0)
   6000			return err;
   6001	}
   6002
   6003	s_h = cb->args[0];
   6004	s_idx = cb->args[1];
   6005
   6006	rcu_read_lock();
   6007	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
   6008		idx = 0;
   6009		head = &net->dev_index_head[h];
   6010		hlist_for_each_entry_rcu(dev, head, index_hlist) {
   6011			if (idx < s_idx)
   6012				goto cont;
   6013			idev = __in6_dev_get(dev);
   6014			if (!idev)
   6015				goto cont;
   6016			if (inet6_fill_ifinfo(skb, idev,
   6017					      NETLINK_CB(cb->skb).portid,
   6018					      cb->nlh->nlmsg_seq,
   6019					      RTM_NEWLINK, NLM_F_MULTI) < 0)
   6020				goto out;
   6021cont:
   6022			idx++;
   6023		}
   6024	}
   6025out:
   6026	rcu_read_unlock();
   6027	cb->args[1] = idx;
   6028	cb->args[0] = h;
   6029
   6030	return skb->len;
   6031}
   6032
   6033void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
   6034{
   6035	struct sk_buff *skb;
   6036	struct net *net = dev_net(idev->dev);
   6037	int err = -ENOBUFS;
   6038
   6039	skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
   6040	if (!skb)
   6041		goto errout;
   6042
   6043	err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
   6044	if (err < 0) {
   6045		/* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
   6046		WARN_ON(err == -EMSGSIZE);
   6047		kfree_skb(skb);
   6048		goto errout;
   6049	}
   6050	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
   6051	return;
   6052errout:
   6053	if (err < 0)
   6054		rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
   6055}
   6056
   6057static inline size_t inet6_prefix_nlmsg_size(void)
   6058{
   6059	return NLMSG_ALIGN(sizeof(struct prefixmsg))
   6060	       + nla_total_size(sizeof(struct in6_addr))
   6061	       + nla_total_size(sizeof(struct prefix_cacheinfo));
   6062}
   6063
   6064static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
   6065			     struct prefix_info *pinfo, u32 portid, u32 seq,
   6066			     int event, unsigned int flags)
   6067{
   6068	struct prefixmsg *pmsg;
   6069	struct nlmsghdr *nlh;
   6070	struct prefix_cacheinfo	ci;
   6071
   6072	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags);
   6073	if (!nlh)
   6074		return -EMSGSIZE;
   6075
   6076	pmsg = nlmsg_data(nlh);
   6077	pmsg->prefix_family = AF_INET6;
   6078	pmsg->prefix_pad1 = 0;
   6079	pmsg->prefix_pad2 = 0;
   6080	pmsg->prefix_ifindex = idev->dev->ifindex;
   6081	pmsg->prefix_len = pinfo->prefix_len;
   6082	pmsg->prefix_type = pinfo->type;
   6083	pmsg->prefix_pad3 = 0;
   6084	pmsg->prefix_flags = 0;
   6085	if (pinfo->onlink)
   6086		pmsg->prefix_flags |= IF_PREFIX_ONLINK;
   6087	if (pinfo->autoconf)
   6088		pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
   6089
   6090	if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
   6091		goto nla_put_failure;
   6092	ci.preferred_time = ntohl(pinfo->prefered);
   6093	ci.valid_time = ntohl(pinfo->valid);
   6094	if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
   6095		goto nla_put_failure;
   6096	nlmsg_end(skb, nlh);
   6097	return 0;
   6098
   6099nla_put_failure:
   6100	nlmsg_cancel(skb, nlh);
   6101	return -EMSGSIZE;
   6102}
   6103
   6104static void inet6_prefix_notify(int event, struct inet6_dev *idev,
   6105			 struct prefix_info *pinfo)
   6106{
   6107	struct sk_buff *skb;
   6108	struct net *net = dev_net(idev->dev);
   6109	int err = -ENOBUFS;
   6110
   6111	skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
   6112	if (!skb)
   6113		goto errout;
   6114
   6115	err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
   6116	if (err < 0) {
   6117		/* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
   6118		WARN_ON(err == -EMSGSIZE);
   6119		kfree_skb(skb);
   6120		goto errout;
   6121	}
   6122	rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
   6123	return;
   6124errout:
   6125	if (err < 0)
   6126		rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
   6127}
   6128
   6129static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
   6130{
   6131	struct net *net = dev_net(ifp->idev->dev);
   6132
   6133	if (event)
   6134		ASSERT_RTNL();
   6135
   6136	inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
   6137
   6138	switch (event) {
   6139	case RTM_NEWADDR:
   6140		/*
   6141		 * If the address was optimistic we inserted the route at the
   6142		 * start of our DAD process, so we don't need to do it again.
   6143		 * If the device was taken down in the middle of the DAD
   6144		 * cycle there is a race where we could get here without a
   6145		 * host route, so nothing to insert. That will be fixed when
   6146		 * the device is brought up.
   6147		 */
   6148		if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) {
   6149			ip6_ins_rt(net, ifp->rt);
   6150		} else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) {
   6151			pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n",
   6152				&ifp->addr, ifp->idev->dev->name);
   6153		}
   6154
   6155		if (ifp->idev->cnf.forwarding)
   6156			addrconf_join_anycast(ifp);
   6157		if (!ipv6_addr_any(&ifp->peer_addr))
   6158			addrconf_prefix_route(&ifp->peer_addr, 128,
   6159					      ifp->rt_priority, ifp->idev->dev,
   6160					      0, 0, GFP_ATOMIC);
   6161		break;
   6162	case RTM_DELADDR:
   6163		if (ifp->idev->cnf.forwarding)
   6164			addrconf_leave_anycast(ifp);
   6165		addrconf_leave_solict(ifp->idev, &ifp->addr);
   6166		if (!ipv6_addr_any(&ifp->peer_addr)) {
   6167			struct fib6_info *rt;
   6168
   6169			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
   6170						       ifp->idev->dev, 0, 0,
   6171						       false);
   6172			if (rt)
   6173				ip6_del_rt(net, rt, false);
   6174		}
   6175		if (ifp->rt) {
   6176			ip6_del_rt(net, ifp->rt, false);
   6177			ifp->rt = NULL;
   6178		}
   6179		rt_genid_bump_ipv6(net);
   6180		break;
   6181	}
   6182	atomic_inc(&net->ipv6.dev_addr_genid);
   6183}
   6184
   6185static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
   6186{
   6187	if (likely(ifp->idev->dead == 0))
   6188		__ipv6_ifa_notify(event, ifp);
   6189}
   6190
   6191#ifdef CONFIG_SYSCTL
   6192
   6193static int addrconf_sysctl_forward(struct ctl_table *ctl, int write,
   6194		void *buffer, size_t *lenp, loff_t *ppos)
   6195{
   6196	int *valp = ctl->data;
   6197	int val = *valp;
   6198	loff_t pos = *ppos;
   6199	struct ctl_table lctl;
   6200	int ret;
   6201
   6202	/*
   6203	 * ctl->data points to idev->cnf.forwarding, we should
   6204	 * not modify it until we get the rtnl lock.
   6205	 */
   6206	lctl = *ctl;
   6207	lctl.data = &val;
   6208
   6209	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
   6210
   6211	if (write)
   6212		ret = addrconf_fixup_forwarding(ctl, valp, val);
   6213	if (ret)
   6214		*ppos = pos;
   6215	return ret;
   6216}
   6217
   6218static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
   6219		void *buffer, size_t *lenp, loff_t *ppos)
   6220{
   6221	struct inet6_dev *idev = ctl->extra1;
   6222	int min_mtu = IPV6_MIN_MTU;
   6223	struct ctl_table lctl;
   6224
   6225	lctl = *ctl;
   6226	lctl.extra1 = &min_mtu;
   6227	lctl.extra2 = idev ? &idev->dev->mtu : NULL;
   6228
   6229	return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
   6230}
   6231
   6232static void dev_disable_change(struct inet6_dev *idev)
   6233{
   6234	struct netdev_notifier_info info;
   6235
   6236	if (!idev || !idev->dev)
   6237		return;
   6238
   6239	netdev_notifier_info_init(&info, idev->dev);
   6240	if (idev->cnf.disable_ipv6)
   6241		addrconf_notify(NULL, NETDEV_DOWN, &info);
   6242	else
   6243		addrconf_notify(NULL, NETDEV_UP, &info);
   6244}
   6245
   6246static void addrconf_disable_change(struct net *net, __s32 newf)
   6247{
   6248	struct net_device *dev;
   6249	struct inet6_dev *idev;
   6250
   6251	for_each_netdev(net, dev) {
   6252		idev = __in6_dev_get(dev);
   6253		if (idev) {
   6254			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
   6255			idev->cnf.disable_ipv6 = newf;
   6256			if (changed)
   6257				dev_disable_change(idev);
   6258		}
   6259	}
   6260}
   6261
   6262static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
   6263{
   6264	struct net *net;
   6265	int old;
   6266
   6267	if (!rtnl_trylock())
   6268		return restart_syscall();
   6269
   6270	net = (struct net *)table->extra2;
   6271	old = *p;
   6272	*p = newf;
   6273
   6274	if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
   6275		rtnl_unlock();
   6276		return 0;
   6277	}
   6278
   6279	if (p == &net->ipv6.devconf_all->disable_ipv6) {
   6280		net->ipv6.devconf_dflt->disable_ipv6 = newf;
   6281		addrconf_disable_change(net, newf);
   6282	} else if ((!newf) ^ (!old))
   6283		dev_disable_change((struct inet6_dev *)table->extra1);
   6284
   6285	rtnl_unlock();
   6286	return 0;
   6287}
   6288
   6289static int addrconf_sysctl_disable(struct ctl_table *ctl, int write,
   6290		void *buffer, size_t *lenp, loff_t *ppos)
   6291{
   6292	int *valp = ctl->data;
   6293	int val = *valp;
   6294	loff_t pos = *ppos;
   6295	struct ctl_table lctl;
   6296	int ret;
   6297
   6298	/*
   6299	 * ctl->data points to idev->cnf.disable_ipv6, we should
   6300	 * not modify it until we get the rtnl lock.
   6301	 */
   6302	lctl = *ctl;
   6303	lctl.data = &val;
   6304
   6305	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
   6306
   6307	if (write)
   6308		ret = addrconf_disable_ipv6(ctl, valp, val);
   6309	if (ret)
   6310		*ppos = pos;
   6311	return ret;
   6312}
   6313
   6314static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write,
   6315		void *buffer, size_t *lenp, loff_t *ppos)
   6316{
   6317	int *valp = ctl->data;
   6318	int ret;
   6319	int old, new;
   6320
   6321	old = *valp;
   6322	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
   6323	new = *valp;
   6324
   6325	if (write && old != new) {
   6326		struct net *net = ctl->extra2;
   6327
   6328		if (!rtnl_trylock())
   6329			return restart_syscall();
   6330
   6331		if (valp == &net->ipv6.devconf_dflt->proxy_ndp)
   6332			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
   6333						     NETCONFA_PROXY_NEIGH,
   6334						     NETCONFA_IFINDEX_DEFAULT,
   6335						     net->ipv6.devconf_dflt);
   6336		else if (valp == &net->ipv6.devconf_all->proxy_ndp)
   6337			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
   6338						     NETCONFA_PROXY_NEIGH,
   6339						     NETCONFA_IFINDEX_ALL,
   6340						     net->ipv6.devconf_all);
   6341		else {
   6342			struct inet6_dev *idev = ctl->extra1;
   6343
   6344			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
   6345						     NETCONFA_PROXY_NEIGH,
   6346						     idev->dev->ifindex,
   6347						     &idev->cnf);
   6348		}
   6349		rtnl_unlock();
   6350	}
   6351
   6352	return ret;
   6353}
   6354
   6355static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write,
   6356					 void *buffer, size_t *lenp,
   6357					 loff_t *ppos)
   6358{
   6359	int ret = 0;
   6360	u32 new_val;
   6361	struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
   6362	struct net *net = (struct net *)ctl->extra2;
   6363	struct ctl_table tmp = {
   6364		.data = &new_val,
   6365		.maxlen = sizeof(new_val),
   6366		.mode = ctl->mode,
   6367	};
   6368
   6369	if (!rtnl_trylock())
   6370		return restart_syscall();
   6371
   6372	new_val = *((u32 *)ctl->data);
   6373
   6374	ret = proc_douintvec(&tmp, write, buffer, lenp, ppos);
   6375	if (ret != 0)
   6376		goto out;
   6377
   6378	if (write) {
   6379		if (check_addr_gen_mode(new_val) < 0) {
   6380			ret = -EINVAL;
   6381			goto out;
   6382		}
   6383
   6384		if (idev) {
   6385			if (check_stable_privacy(idev, net, new_val) < 0) {
   6386				ret = -EINVAL;
   6387				goto out;
   6388			}
   6389
   6390			if (idev->cnf.addr_gen_mode != new_val) {
   6391				idev->cnf.addr_gen_mode = new_val;
   6392				addrconf_dev_config(idev->dev);
   6393			}
   6394		} else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) {
   6395			struct net_device *dev;
   6396
   6397			net->ipv6.devconf_dflt->addr_gen_mode = new_val;
   6398			for_each_netdev(net, dev) {
   6399				idev = __in6_dev_get(dev);
   6400				if (idev &&
   6401				    idev->cnf.addr_gen_mode != new_val) {
   6402					idev->cnf.addr_gen_mode = new_val;
   6403					addrconf_dev_config(idev->dev);
   6404				}
   6405			}
   6406		}
   6407
   6408		*((u32 *)ctl->data) = new_val;
   6409	}
   6410
   6411out:
   6412	rtnl_unlock();
   6413
   6414	return ret;
   6415}
   6416
   6417static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write,
   6418					 void *buffer, size_t *lenp,
   6419					 loff_t *ppos)
   6420{
   6421	int err;
   6422	struct in6_addr addr;
   6423	char str[IPV6_MAX_STRLEN];
   6424	struct ctl_table lctl = *ctl;
   6425	struct net *net = ctl->extra2;
   6426	struct ipv6_stable_secret *secret = ctl->data;
   6427
   6428	if (&net->ipv6.devconf_all->stable_secret == ctl->data)
   6429		return -EIO;
   6430
   6431	lctl.maxlen = IPV6_MAX_STRLEN;
   6432	lctl.data = str;
   6433
   6434	if (!rtnl_trylock())
   6435		return restart_syscall();
   6436
   6437	if (!write && !secret->initialized) {
   6438		err = -EIO;
   6439		goto out;
   6440	}
   6441
   6442	err = snprintf(str, sizeof(str), "%pI6", &secret->secret);
   6443	if (err >= sizeof(str)) {
   6444		err = -EIO;
   6445		goto out;
   6446	}
   6447
   6448	err = proc_dostring(&lctl, write, buffer, lenp, ppos);
   6449	if (err || !write)
   6450		goto out;
   6451
   6452	if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) {
   6453		err = -EIO;
   6454		goto out;
   6455	}
   6456
   6457	secret->initialized = true;
   6458	secret->secret = addr;
   6459
   6460	if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) {
   6461		struct net_device *dev;
   6462
   6463		for_each_netdev(net, dev) {
   6464			struct inet6_dev *idev = __in6_dev_get(dev);
   6465
   6466			if (idev) {
   6467				idev->cnf.addr_gen_mode =
   6468					IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
   6469			}
   6470		}
   6471	} else {
   6472		struct inet6_dev *idev = ctl->extra1;
   6473
   6474		idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY;
   6475	}
   6476
   6477out:
   6478	rtnl_unlock();
   6479
   6480	return err;
   6481}
   6482
   6483static
   6484int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl,
   6485						int write, void *buffer,
   6486						size_t *lenp,
   6487						loff_t *ppos)
   6488{
   6489	int *valp = ctl->data;
   6490	int val = *valp;
   6491	loff_t pos = *ppos;
   6492	struct ctl_table lctl;
   6493	int ret;
   6494
   6495	/* ctl->data points to idev->cnf.ignore_routes_when_linkdown
   6496	 * we should not modify it until we get the rtnl lock.
   6497	 */
   6498	lctl = *ctl;
   6499	lctl.data = &val;
   6500
   6501	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
   6502
   6503	if (write)
   6504		ret = addrconf_fixup_linkdown(ctl, valp, val);
   6505	if (ret)
   6506		*ppos = pos;
   6507	return ret;
   6508}
   6509
   6510static
   6511void addrconf_set_nopolicy(struct rt6_info *rt, int action)
   6512{
   6513	if (rt) {
   6514		if (action)
   6515			rt->dst.flags |= DST_NOPOLICY;
   6516		else
   6517			rt->dst.flags &= ~DST_NOPOLICY;
   6518	}
   6519}
   6520
   6521static
   6522void addrconf_disable_policy_idev(struct inet6_dev *idev, int val)
   6523{
   6524	struct inet6_ifaddr *ifa;
   6525
   6526	read_lock_bh(&idev->lock);
   6527	list_for_each_entry(ifa, &idev->addr_list, if_list) {
   6528		spin_lock(&ifa->lock);
   6529		if (ifa->rt) {
   6530			/* host routes only use builtin fib6_nh */
   6531			struct fib6_nh *nh = ifa->rt->fib6_nh;
   6532			int cpu;
   6533
   6534			rcu_read_lock();
   6535			ifa->rt->dst_nopolicy = val ? true : false;
   6536			if (nh->rt6i_pcpu) {
   6537				for_each_possible_cpu(cpu) {
   6538					struct rt6_info **rtp;
   6539
   6540					rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu);
   6541					addrconf_set_nopolicy(*rtp, val);
   6542				}
   6543			}
   6544			rcu_read_unlock();
   6545		}
   6546		spin_unlock(&ifa->lock);
   6547	}
   6548	read_unlock_bh(&idev->lock);
   6549}
   6550
   6551static
   6552int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val)
   6553{
   6554	struct inet6_dev *idev;
   6555	struct net *net;
   6556
   6557	if (!rtnl_trylock())
   6558		return restart_syscall();
   6559
   6560	*valp = val;
   6561
   6562	net = (struct net *)ctl->extra2;
   6563	if (valp == &net->ipv6.devconf_dflt->disable_policy) {
   6564		rtnl_unlock();
   6565		return 0;
   6566	}
   6567
   6568	if (valp == &net->ipv6.devconf_all->disable_policy)  {
   6569		struct net_device *dev;
   6570
   6571		for_each_netdev(net, dev) {
   6572			idev = __in6_dev_get(dev);
   6573			if (idev)
   6574				addrconf_disable_policy_idev(idev, val);
   6575		}
   6576	} else {
   6577		idev = (struct inet6_dev *)ctl->extra1;
   6578		addrconf_disable_policy_idev(idev, val);
   6579	}
   6580
   6581	rtnl_unlock();
   6582	return 0;
   6583}
   6584
   6585static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write,
   6586				   void *buffer, size_t *lenp, loff_t *ppos)
   6587{
   6588	int *valp = ctl->data;
   6589	int val = *valp;
   6590	loff_t pos = *ppos;
   6591	struct ctl_table lctl;
   6592	int ret;
   6593
   6594	lctl = *ctl;
   6595	lctl.data = &val;
   6596	ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
   6597
   6598	if (write && (*valp != val))
   6599		ret = addrconf_disable_policy(ctl, valp, val);
   6600
   6601	if (ret)
   6602		*ppos = pos;
   6603
   6604	return ret;
   6605}
   6606
   6607static int minus_one = -1;
   6608static const int two_five_five = 255;
   6609static u32 ioam6_if_id_max = U16_MAX;
   6610
   6611static const struct ctl_table addrconf_sysctl[] = {
   6612	{
   6613		.procname	= "forwarding",
   6614		.data		= &ipv6_devconf.forwarding,
   6615		.maxlen		= sizeof(int),
   6616		.mode		= 0644,
   6617		.proc_handler	= addrconf_sysctl_forward,
   6618	},
   6619	{
   6620		.procname	= "hop_limit",
   6621		.data		= &ipv6_devconf.hop_limit,
   6622		.maxlen		= sizeof(int),
   6623		.mode		= 0644,
   6624		.proc_handler	= proc_dointvec_minmax,
   6625		.extra1		= (void *)SYSCTL_ONE,
   6626		.extra2		= (void *)&two_five_five,
   6627	},
   6628	{
   6629		.procname	= "mtu",
   6630		.data		= &ipv6_devconf.mtu6,
   6631		.maxlen		= sizeof(int),
   6632		.mode		= 0644,
   6633		.proc_handler	= addrconf_sysctl_mtu,
   6634	},
   6635	{
   6636		.procname	= "accept_ra",
   6637		.data		= &ipv6_devconf.accept_ra,
   6638		.maxlen		= sizeof(int),
   6639		.mode		= 0644,
   6640		.proc_handler	= proc_dointvec,
   6641	},
   6642	{
   6643		.procname	= "accept_redirects",
   6644		.data		= &ipv6_devconf.accept_redirects,
   6645		.maxlen		= sizeof(int),
   6646		.mode		= 0644,
   6647		.proc_handler	= proc_dointvec,
   6648	},
   6649	{
   6650		.procname	= "autoconf",
   6651		.data		= &ipv6_devconf.autoconf,
   6652		.maxlen		= sizeof(int),
   6653		.mode		= 0644,
   6654		.proc_handler	= proc_dointvec,
   6655	},
   6656	{
   6657		.procname	= "dad_transmits",
   6658		.data		= &ipv6_devconf.dad_transmits,
   6659		.maxlen		= sizeof(int),
   6660		.mode		= 0644,
   6661		.proc_handler	= proc_dointvec,
   6662	},
   6663	{
   6664		.procname	= "router_solicitations",
   6665		.data		= &ipv6_devconf.rtr_solicits,
   6666		.maxlen		= sizeof(int),
   6667		.mode		= 0644,
   6668		.proc_handler	= proc_dointvec_minmax,
   6669		.extra1		= &minus_one,
   6670	},
   6671	{
   6672		.procname	= "router_solicitation_interval",
   6673		.data		= &ipv6_devconf.rtr_solicit_interval,
   6674		.maxlen		= sizeof(int),
   6675		.mode		= 0644,
   6676		.proc_handler	= proc_dointvec_jiffies,
   6677	},
   6678	{
   6679		.procname	= "router_solicitation_max_interval",
   6680		.data		= &ipv6_devconf.rtr_solicit_max_interval,
   6681		.maxlen		= sizeof(int),
   6682		.mode		= 0644,
   6683		.proc_handler	= proc_dointvec_jiffies,
   6684	},
   6685	{
   6686		.procname	= "router_solicitation_delay",
   6687		.data		= &ipv6_devconf.rtr_solicit_delay,
   6688		.maxlen		= sizeof(int),
   6689		.mode		= 0644,
   6690		.proc_handler	= proc_dointvec_jiffies,
   6691	},
   6692	{
   6693		.procname	= "force_mld_version",
   6694		.data		= &ipv6_devconf.force_mld_version,
   6695		.maxlen		= sizeof(int),
   6696		.mode		= 0644,
   6697		.proc_handler	= proc_dointvec,
   6698	},
   6699	{
   6700		.procname	= "mldv1_unsolicited_report_interval",
   6701		.data		=
   6702			&ipv6_devconf.mldv1_unsolicited_report_interval,
   6703		.maxlen		= sizeof(int),
   6704		.mode		= 0644,
   6705		.proc_handler	= proc_dointvec_ms_jiffies,
   6706	},
   6707	{
   6708		.procname	= "mldv2_unsolicited_report_interval",
   6709		.data		=
   6710			&ipv6_devconf.mldv2_unsolicited_report_interval,
   6711		.maxlen		= sizeof(int),
   6712		.mode		= 0644,
   6713		.proc_handler	= proc_dointvec_ms_jiffies,
   6714	},
   6715	{
   6716		.procname	= "use_tempaddr",
   6717		.data		= &ipv6_devconf.use_tempaddr,
   6718		.maxlen		= sizeof(int),
   6719		.mode		= 0644,
   6720		.proc_handler	= proc_dointvec,
   6721	},
   6722	{
   6723		.procname	= "temp_valid_lft",
   6724		.data		= &ipv6_devconf.temp_valid_lft,
   6725		.maxlen		= sizeof(int),
   6726		.mode		= 0644,
   6727		.proc_handler	= proc_dointvec,
   6728	},
   6729	{
   6730		.procname	= "temp_prefered_lft",
   6731		.data		= &ipv6_devconf.temp_prefered_lft,
   6732		.maxlen		= sizeof(int),
   6733		.mode		= 0644,
   6734		.proc_handler	= proc_dointvec,
   6735	},
   6736	{
   6737		.procname	= "regen_max_retry",
   6738		.data		= &ipv6_devconf.regen_max_retry,
   6739		.maxlen		= sizeof(int),
   6740		.mode		= 0644,
   6741		.proc_handler	= proc_dointvec,
   6742	},
   6743	{
   6744		.procname	= "max_desync_factor",
   6745		.data		= &ipv6_devconf.max_desync_factor,
   6746		.maxlen		= sizeof(int),
   6747		.mode		= 0644,
   6748		.proc_handler	= proc_dointvec,
   6749	},
   6750	{
   6751		.procname	= "max_addresses",
   6752		.data		= &ipv6_devconf.max_addresses,
   6753		.maxlen		= sizeof(int),
   6754		.mode		= 0644,
   6755		.proc_handler	= proc_dointvec,
   6756	},
   6757	{
   6758		.procname	= "accept_ra_defrtr",
   6759		.data		= &ipv6_devconf.accept_ra_defrtr,
   6760		.maxlen		= sizeof(int),
   6761		.mode		= 0644,
   6762		.proc_handler	= proc_dointvec,
   6763	},
   6764	{
   6765		.procname	= "ra_defrtr_metric",
   6766		.data		= &ipv6_devconf.ra_defrtr_metric,
   6767		.maxlen		= sizeof(u32),
   6768		.mode		= 0644,
   6769		.proc_handler	= proc_douintvec_minmax,
   6770		.extra1		= (void *)SYSCTL_ONE,
   6771	},
   6772	{
   6773		.procname	= "accept_ra_min_hop_limit",
   6774		.data		= &ipv6_devconf.accept_ra_min_hop_limit,
   6775		.maxlen		= sizeof(int),
   6776		.mode		= 0644,
   6777		.proc_handler	= proc_dointvec,
   6778	},
   6779	{
   6780		.procname	= "accept_ra_pinfo",
   6781		.data		= &ipv6_devconf.accept_ra_pinfo,
   6782		.maxlen		= sizeof(int),
   6783		.mode		= 0644,
   6784		.proc_handler	= proc_dointvec,
   6785	},
   6786#ifdef CONFIG_IPV6_ROUTER_PREF
   6787	{
   6788		.procname	= "accept_ra_rtr_pref",
   6789		.data		= &ipv6_devconf.accept_ra_rtr_pref,
   6790		.maxlen		= sizeof(int),
   6791		.mode		= 0644,
   6792		.proc_handler	= proc_dointvec,
   6793	},
   6794	{
   6795		.procname	= "router_probe_interval",
   6796		.data		= &ipv6_devconf.rtr_probe_interval,
   6797		.maxlen		= sizeof(int),
   6798		.mode		= 0644,
   6799		.proc_handler	= proc_dointvec_jiffies,
   6800	},
   6801#ifdef CONFIG_IPV6_ROUTE_INFO
   6802	{
   6803		.procname	= "accept_ra_rt_info_min_plen",
   6804		.data		= &ipv6_devconf.accept_ra_rt_info_min_plen,
   6805		.maxlen		= sizeof(int),
   6806		.mode		= 0644,
   6807		.proc_handler	= proc_dointvec,
   6808	},
   6809	{
   6810		.procname	= "accept_ra_rt_info_max_plen",
   6811		.data		= &ipv6_devconf.accept_ra_rt_info_max_plen,
   6812		.maxlen		= sizeof(int),
   6813		.mode		= 0644,
   6814		.proc_handler	= proc_dointvec,
   6815	},
   6816#endif
   6817#endif
   6818	{
   6819		.procname	= "proxy_ndp",
   6820		.data		= &ipv6_devconf.proxy_ndp,
   6821		.maxlen		= sizeof(int),
   6822		.mode		= 0644,
   6823		.proc_handler	= addrconf_sysctl_proxy_ndp,
   6824	},
   6825	{
   6826		.procname	= "accept_source_route",
   6827		.data		= &ipv6_devconf.accept_source_route,
   6828		.maxlen		= sizeof(int),
   6829		.mode		= 0644,
   6830		.proc_handler	= proc_dointvec,
   6831	},
   6832#ifdef CONFIG_IPV6_OPTIMISTIC_DAD
   6833	{
   6834		.procname	= "optimistic_dad",
   6835		.data		= &ipv6_devconf.optimistic_dad,
   6836		.maxlen		= sizeof(int),
   6837		.mode		= 0644,
   6838		.proc_handler   = proc_dointvec,
   6839	},
   6840	{
   6841		.procname	= "use_optimistic",
   6842		.data		= &ipv6_devconf.use_optimistic,
   6843		.maxlen		= sizeof(int),
   6844		.mode		= 0644,
   6845		.proc_handler	= proc_dointvec,
   6846	},
   6847#endif
   6848#ifdef CONFIG_IPV6_MROUTE
   6849	{
   6850		.procname	= "mc_forwarding",
   6851		.data		= &ipv6_devconf.mc_forwarding,
   6852		.maxlen		= sizeof(int),
   6853		.mode		= 0444,
   6854		.proc_handler	= proc_dointvec,
   6855	},
   6856#endif
   6857	{
   6858		.procname	= "disable_ipv6",
   6859		.data		= &ipv6_devconf.disable_ipv6,
   6860		.maxlen		= sizeof(int),
   6861		.mode		= 0644,
   6862		.proc_handler	= addrconf_sysctl_disable,
   6863	},
   6864	{
   6865		.procname	= "accept_dad",
   6866		.data		= &ipv6_devconf.accept_dad,
   6867		.maxlen		= sizeof(int),
   6868		.mode		= 0644,
   6869		.proc_handler	= proc_dointvec,
   6870	},
   6871	{
   6872		.procname	= "force_tllao",
   6873		.data		= &ipv6_devconf.force_tllao,
   6874		.maxlen		= sizeof(int),
   6875		.mode		= 0644,
   6876		.proc_handler	= proc_dointvec
   6877	},
   6878	{
   6879		.procname	= "ndisc_notify",
   6880		.data		= &ipv6_devconf.ndisc_notify,
   6881		.maxlen		= sizeof(int),
   6882		.mode		= 0644,
   6883		.proc_handler	= proc_dointvec
   6884	},
   6885	{
   6886		.procname	= "suppress_frag_ndisc",
   6887		.data		= &ipv6_devconf.suppress_frag_ndisc,
   6888		.maxlen		= sizeof(int),
   6889		.mode		= 0644,
   6890		.proc_handler	= proc_dointvec
   6891	},
   6892	{
   6893		.procname	= "accept_ra_from_local",
   6894		.data		= &ipv6_devconf.accept_ra_from_local,
   6895		.maxlen		= sizeof(int),
   6896		.mode		= 0644,
   6897		.proc_handler	= proc_dointvec,
   6898	},
   6899	{
   6900		.procname	= "accept_ra_mtu",
   6901		.data		= &ipv6_devconf.accept_ra_mtu,
   6902		.maxlen		= sizeof(int),
   6903		.mode		= 0644,
   6904		.proc_handler	= proc_dointvec,
   6905	},
   6906	{
   6907		.procname	= "stable_secret",
   6908		.data		= &ipv6_devconf.stable_secret,
   6909		.maxlen		= IPV6_MAX_STRLEN,
   6910		.mode		= 0600,
   6911		.proc_handler	= addrconf_sysctl_stable_secret,
   6912	},
   6913	{
   6914		.procname	= "use_oif_addrs_only",
   6915		.data		= &ipv6_devconf.use_oif_addrs_only,
   6916		.maxlen		= sizeof(int),
   6917		.mode		= 0644,
   6918		.proc_handler	= proc_dointvec,
   6919	},
   6920	{
   6921		.procname	= "ignore_routes_with_linkdown",
   6922		.data		= &ipv6_devconf.ignore_routes_with_linkdown,
   6923		.maxlen		= sizeof(int),
   6924		.mode		= 0644,
   6925		.proc_handler	= addrconf_sysctl_ignore_routes_with_linkdown,
   6926	},
   6927	{
   6928		.procname	= "drop_unicast_in_l2_multicast",
   6929		.data		= &ipv6_devconf.drop_unicast_in_l2_multicast,
   6930		.maxlen		= sizeof(int),
   6931		.mode		= 0644,
   6932		.proc_handler	= proc_dointvec,
   6933	},
   6934	{
   6935		.procname	= "drop_unsolicited_na",
   6936		.data		= &ipv6_devconf.drop_unsolicited_na,
   6937		.maxlen		= sizeof(int),
   6938		.mode		= 0644,
   6939		.proc_handler	= proc_dointvec,
   6940	},
   6941	{
   6942		.procname	= "keep_addr_on_down",
   6943		.data		= &ipv6_devconf.keep_addr_on_down,
   6944		.maxlen		= sizeof(int),
   6945		.mode		= 0644,
   6946		.proc_handler	= proc_dointvec,
   6947
   6948	},
   6949	{
   6950		.procname	= "seg6_enabled",
   6951		.data		= &ipv6_devconf.seg6_enabled,
   6952		.maxlen		= sizeof(int),
   6953		.mode		= 0644,
   6954		.proc_handler	= proc_dointvec,
   6955	},
   6956#ifdef CONFIG_IPV6_SEG6_HMAC
   6957	{
   6958		.procname	= "seg6_require_hmac",
   6959		.data		= &ipv6_devconf.seg6_require_hmac,
   6960		.maxlen		= sizeof(int),
   6961		.mode		= 0644,
   6962		.proc_handler	= proc_dointvec,
   6963	},
   6964#endif
   6965	{
   6966		.procname       = "enhanced_dad",
   6967		.data           = &ipv6_devconf.enhanced_dad,
   6968		.maxlen         = sizeof(int),
   6969		.mode           = 0644,
   6970		.proc_handler   = proc_dointvec,
   6971	},
   6972	{
   6973		.procname	= "addr_gen_mode",
   6974		.data		= &ipv6_devconf.addr_gen_mode,
   6975		.maxlen		= sizeof(int),
   6976		.mode		= 0644,
   6977		.proc_handler	= addrconf_sysctl_addr_gen_mode,
   6978	},
   6979	{
   6980		.procname       = "disable_policy",
   6981		.data           = &ipv6_devconf.disable_policy,
   6982		.maxlen         = sizeof(int),
   6983		.mode           = 0644,
   6984		.proc_handler   = addrconf_sysctl_disable_policy,
   6985	},
   6986	{
   6987		.procname	= "ndisc_tclass",
   6988		.data		= &ipv6_devconf.ndisc_tclass,
   6989		.maxlen		= sizeof(int),
   6990		.mode		= 0644,
   6991		.proc_handler	= proc_dointvec_minmax,
   6992		.extra1		= (void *)SYSCTL_ZERO,
   6993		.extra2		= (void *)&two_five_five,
   6994	},
   6995	{
   6996		.procname	= "rpl_seg_enabled",
   6997		.data		= &ipv6_devconf.rpl_seg_enabled,
   6998		.maxlen		= sizeof(int),
   6999		.mode		= 0644,
   7000		.proc_handler	= proc_dointvec,
   7001	},
   7002	{
   7003		.procname	= "ioam6_enabled",
   7004		.data		= &ipv6_devconf.ioam6_enabled,
   7005		.maxlen		= sizeof(u8),
   7006		.mode		= 0644,
   7007		.proc_handler	= proc_dou8vec_minmax,
   7008		.extra1		= (void *)SYSCTL_ZERO,
   7009		.extra2		= (void *)SYSCTL_ONE,
   7010	},
   7011	{
   7012		.procname	= "ioam6_id",
   7013		.data		= &ipv6_devconf.ioam6_id,
   7014		.maxlen		= sizeof(u32),
   7015		.mode		= 0644,
   7016		.proc_handler	= proc_douintvec_minmax,
   7017		.extra1		= (void *)SYSCTL_ZERO,
   7018		.extra2		= (void *)&ioam6_if_id_max,
   7019	},
   7020	{
   7021		.procname	= "ioam6_id_wide",
   7022		.data		= &ipv6_devconf.ioam6_id_wide,
   7023		.maxlen		= sizeof(u32),
   7024		.mode		= 0644,
   7025		.proc_handler	= proc_douintvec,
   7026	},
   7027	{
   7028		.procname	= "ndisc_evict_nocarrier",
   7029		.data		= &ipv6_devconf.ndisc_evict_nocarrier,
   7030		.maxlen		= sizeof(u8),
   7031		.mode		= 0644,
   7032		.proc_handler	= proc_dou8vec_minmax,
   7033		.extra1		= (void *)SYSCTL_ZERO,
   7034		.extra2		= (void *)SYSCTL_ONE,
   7035	},
   7036	{
   7037		.procname	= "accept_untracked_na",
   7038		.data		= &ipv6_devconf.accept_untracked_na,
   7039		.maxlen		= sizeof(int),
   7040		.mode		= 0644,
   7041		.proc_handler	= proc_dointvec_minmax,
   7042		.extra1		= (void *)SYSCTL_ZERO,
   7043		.extra2		= (void *)SYSCTL_ONE,
   7044	},
   7045	{
   7046		/* sentinel */
   7047	}
   7048};
   7049
   7050static int __addrconf_sysctl_register(struct net *net, char *dev_name,
   7051		struct inet6_dev *idev, struct ipv6_devconf *p)
   7052{
   7053	int i, ifindex;
   7054	struct ctl_table *table;
   7055	char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
   7056
   7057	table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT);
   7058	if (!table)
   7059		goto out;
   7060
   7061	for (i = 0; table[i].data; i++) {
   7062		table[i].data += (char *)p - (char *)&ipv6_devconf;
   7063		/* If one of these is already set, then it is not safe to
   7064		 * overwrite either of them: this makes proc_dointvec_minmax
   7065		 * usable.
   7066		 */
   7067		if (!table[i].extra1 && !table[i].extra2) {
   7068			table[i].extra1 = idev; /* embedded; no ref */
   7069			table[i].extra2 = net;
   7070		}
   7071	}
   7072
   7073	snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
   7074
   7075	p->sysctl_header = register_net_sysctl(net, path, table);
   7076	if (!p->sysctl_header)
   7077		goto free;
   7078
   7079	if (!strcmp(dev_name, "all"))
   7080		ifindex = NETCONFA_IFINDEX_ALL;
   7081	else if (!strcmp(dev_name, "default"))
   7082		ifindex = NETCONFA_IFINDEX_DEFAULT;
   7083	else
   7084		ifindex = idev->dev->ifindex;
   7085	inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL,
   7086				     ifindex, p);
   7087	return 0;
   7088
   7089free:
   7090	kfree(table);
   7091out:
   7092	return -ENOBUFS;
   7093}
   7094
   7095static void __addrconf_sysctl_unregister(struct net *net,
   7096					 struct ipv6_devconf *p, int ifindex)
   7097{
   7098	struct ctl_table *table;
   7099
   7100	if (!p->sysctl_header)
   7101		return;
   7102
   7103	table = p->sysctl_header->ctl_table_arg;
   7104	unregister_net_sysctl_table(p->sysctl_header);
   7105	p->sysctl_header = NULL;
   7106	kfree(table);
   7107
   7108	inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL);
   7109}
   7110
   7111static int addrconf_sysctl_register(struct inet6_dev *idev)
   7112{
   7113	int err;
   7114
   7115	if (!sysctl_dev_name_is_allowed(idev->dev->name))
   7116		return -EINVAL;
   7117
   7118	err = neigh_sysctl_register(idev->dev, idev->nd_parms,
   7119				    &ndisc_ifinfo_sysctl_change);
   7120	if (err)
   7121		return err;
   7122	err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
   7123					 idev, &idev->cnf);
   7124	if (err)
   7125		neigh_sysctl_unregister(idev->nd_parms);
   7126
   7127	return err;
   7128}
   7129
   7130static void addrconf_sysctl_unregister(struct inet6_dev *idev)
   7131{
   7132	__addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf,
   7133				     idev->dev->ifindex);
   7134	neigh_sysctl_unregister(idev->nd_parms);
   7135}
   7136
   7137
   7138#endif
   7139
   7140static int __net_init addrconf_init_net(struct net *net)
   7141{
   7142	int err = -ENOMEM;
   7143	struct ipv6_devconf *all, *dflt;
   7144
   7145	spin_lock_init(&net->ipv6.addrconf_hash_lock);
   7146	INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work);
   7147	net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE,
   7148					   sizeof(struct hlist_head),
   7149					   GFP_KERNEL);
   7150	if (!net->ipv6.inet6_addr_lst)
   7151		goto err_alloc_addr;
   7152
   7153	all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
   7154	if (!all)
   7155		goto err_alloc_all;
   7156
   7157	dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
   7158	if (!dflt)
   7159		goto err_alloc_dflt;
   7160
   7161	if (IS_ENABLED(CONFIG_SYSCTL) &&
   7162	    !net_eq(net, &init_net)) {
   7163		switch (sysctl_devconf_inherit_init_net) {
   7164		case 1:  /* copy from init_net */
   7165			memcpy(all, init_net.ipv6.devconf_all,
   7166			       sizeof(ipv6_devconf));
   7167			memcpy(dflt, init_net.ipv6.devconf_dflt,
   7168			       sizeof(ipv6_devconf_dflt));
   7169			break;
   7170		case 3: /* copy from the current netns */
   7171			memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all,
   7172			       sizeof(ipv6_devconf));
   7173			memcpy(dflt,
   7174			       current->nsproxy->net_ns->ipv6.devconf_dflt,
   7175			       sizeof(ipv6_devconf_dflt));
   7176			break;
   7177		case 0:
   7178		case 2:
   7179			/* use compiled values */
   7180			break;
   7181		}
   7182	}
   7183
   7184	/* these will be inherited by all namespaces */
   7185	dflt->autoconf = ipv6_defaults.autoconf;
   7186	dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
   7187
   7188	dflt->stable_secret.initialized = false;
   7189	all->stable_secret.initialized = false;
   7190
   7191	net->ipv6.devconf_all = all;
   7192	net->ipv6.devconf_dflt = dflt;
   7193
   7194#ifdef CONFIG_SYSCTL
   7195	err = __addrconf_sysctl_register(net, "all", NULL, all);
   7196	if (err < 0)
   7197		goto err_reg_all;
   7198
   7199	err = __addrconf_sysctl_register(net, "default", NULL, dflt);
   7200	if (err < 0)
   7201		goto err_reg_dflt;
   7202#endif
   7203	return 0;
   7204
   7205#ifdef CONFIG_SYSCTL
   7206err_reg_dflt:
   7207	__addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL);
   7208err_reg_all:
   7209	kfree(dflt);
   7210#endif
   7211err_alloc_dflt:
   7212	kfree(all);
   7213err_alloc_all:
   7214	kfree(net->ipv6.inet6_addr_lst);
   7215err_alloc_addr:
   7216	return err;
   7217}
   7218
   7219static void __net_exit addrconf_exit_net(struct net *net)
   7220{
   7221	int i;
   7222
   7223#ifdef CONFIG_SYSCTL
   7224	__addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt,
   7225				     NETCONFA_IFINDEX_DEFAULT);
   7226	__addrconf_sysctl_unregister(net, net->ipv6.devconf_all,
   7227				     NETCONFA_IFINDEX_ALL);
   7228#endif
   7229	kfree(net->ipv6.devconf_dflt);
   7230	net->ipv6.devconf_dflt = NULL;
   7231	kfree(net->ipv6.devconf_all);
   7232	net->ipv6.devconf_all = NULL;
   7233
   7234	cancel_delayed_work_sync(&net->ipv6.addr_chk_work);
   7235	/*
   7236	 *	Check hash table, then free it.
   7237	 */
   7238	for (i = 0; i < IN6_ADDR_HSIZE; i++)
   7239		WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i]));
   7240
   7241	kfree(net->ipv6.inet6_addr_lst);
   7242	net->ipv6.inet6_addr_lst = NULL;
   7243}
   7244
   7245static struct pernet_operations addrconf_ops = {
   7246	.init = addrconf_init_net,
   7247	.exit = addrconf_exit_net,
   7248};
   7249
   7250static struct rtnl_af_ops inet6_ops __read_mostly = {
   7251	.family		  = AF_INET6,
   7252	.fill_link_af	  = inet6_fill_link_af,
   7253	.get_link_af_size = inet6_get_link_af_size,
   7254	.validate_link_af = inet6_validate_link_af,
   7255	.set_link_af	  = inet6_set_link_af,
   7256};
   7257
   7258/*
   7259 *	Init / cleanup code
   7260 */
   7261
   7262int __init addrconf_init(void)
   7263{
   7264	struct inet6_dev *idev;
   7265	int err;
   7266
   7267	err = ipv6_addr_label_init();
   7268	if (err < 0) {
   7269		pr_crit("%s: cannot initialize default policy table: %d\n",
   7270			__func__, err);
   7271		goto out;
   7272	}
   7273
   7274	err = register_pernet_subsys(&addrconf_ops);
   7275	if (err < 0)
   7276		goto out_addrlabel;
   7277
   7278	addrconf_wq = create_workqueue("ipv6_addrconf");
   7279	if (!addrconf_wq) {
   7280		err = -ENOMEM;
   7281		goto out_nowq;
   7282	}
   7283
   7284	rtnl_lock();
   7285	idev = ipv6_add_dev(blackhole_netdev);
   7286	rtnl_unlock();
   7287	if (IS_ERR(idev)) {
   7288		err = PTR_ERR(idev);
   7289		goto errlo;
   7290	}
   7291
   7292	ip6_route_init_special_entries();
   7293
   7294	register_netdevice_notifier(&ipv6_dev_notf);
   7295
   7296	addrconf_verify(&init_net);
   7297
   7298	rtnl_af_register(&inet6_ops);
   7299
   7300	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK,
   7301				   NULL, inet6_dump_ifinfo, 0);
   7302	if (err < 0)
   7303		goto errout;
   7304
   7305	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR,
   7306				   inet6_rtm_newaddr, NULL, 0);
   7307	if (err < 0)
   7308		goto errout;
   7309	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR,
   7310				   inet6_rtm_deladdr, NULL, 0);
   7311	if (err < 0)
   7312		goto errout;
   7313	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR,
   7314				   inet6_rtm_getaddr, inet6_dump_ifaddr,
   7315				   RTNL_FLAG_DOIT_UNLOCKED);
   7316	if (err < 0)
   7317		goto errout;
   7318	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST,
   7319				   NULL, inet6_dump_ifmcaddr, 0);
   7320	if (err < 0)
   7321		goto errout;
   7322	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST,
   7323				   NULL, inet6_dump_ifacaddr, 0);
   7324	if (err < 0)
   7325		goto errout;
   7326	err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF,
   7327				   inet6_netconf_get_devconf,
   7328				   inet6_netconf_dump_devconf,
   7329				   RTNL_FLAG_DOIT_UNLOCKED);
   7330	if (err < 0)
   7331		goto errout;
   7332	err = ipv6_addr_label_rtnl_register();
   7333	if (err < 0)
   7334		goto errout;
   7335
   7336	return 0;
   7337errout:
   7338	rtnl_unregister_all(PF_INET6);
   7339	rtnl_af_unregister(&inet6_ops);
   7340	unregister_netdevice_notifier(&ipv6_dev_notf);
   7341errlo:
   7342	destroy_workqueue(addrconf_wq);
   7343out_nowq:
   7344	unregister_pernet_subsys(&addrconf_ops);
   7345out_addrlabel:
   7346	ipv6_addr_label_cleanup();
   7347out:
   7348	return err;
   7349}
   7350
   7351void addrconf_cleanup(void)
   7352{
   7353	struct net_device *dev;
   7354
   7355	unregister_netdevice_notifier(&ipv6_dev_notf);
   7356	unregister_pernet_subsys(&addrconf_ops);
   7357	ipv6_addr_label_cleanup();
   7358
   7359	rtnl_af_unregister(&inet6_ops);
   7360
   7361	rtnl_lock();
   7362
   7363	/* clean dev list */
   7364	for_each_netdev(&init_net, dev) {
   7365		if (__in6_dev_get(dev) == NULL)
   7366			continue;
   7367		addrconf_ifdown(dev, true);
   7368	}
   7369	addrconf_ifdown(init_net.loopback_dev, true);
   7370
   7371	rtnl_unlock();
   7372
   7373	destroy_workqueue(addrconf_wq);
   7374}