| 1 |
// SPDX-License-Identifier: GPL-2.0-or-later |
| 2 |
/* |
| 3 |
* Linux INET6 implementation |
| 4 |
* FIB front-end. |
| 5 |
* |
| 6 |
* Authors: |
| 7 |
* Pedro Roque <[email protected]> |
| 8 |
*/ |
| 9 |
|
| 10 |
/* Changes: |
| 11 |
* |
| 12 |
* YOSHIFUJI Hideaki @USAGI |
| 13 |
* reworked default router selection. |
| 14 |
* - respect outgoing interface |
| 15 |
* - select from (probably) reachable routers (i.e. |
| 16 |
* routers in REACHABLE, STALE, DELAY or PROBE states). |
| 17 |
* - always select the same router if it is (probably) |
| 18 |
* reachable. otherwise, round-robin the list. |
| 19 |
* Ville Nuorvala |
| 20 |
* Fixed routing subtrees. |
| 21 |
*/ |
| 22 |
|
| 23 |
#define pr_fmt(fmt) "IPv6: " fmt |
| 24 |
|
| 25 |
#include <linux/capability.h> |
| 26 |
#include <linux/errno.h> |
| 27 |
#include <linux/export.h> |
| 28 |
#include <linux/types.h> |
| 29 |
#include <linux/times.h> |
| 30 |
#include <linux/socket.h> |
| 31 |
#include <linux/sockios.h> |
| 32 |
#include <linux/net.h> |
| 33 |
#include <linux/route.h> |
| 34 |
#include <linux/netdevice.h> |
| 35 |
#include <linux/in6.h> |
| 36 |
#include <linux/mroute6.h> |
| 37 |
#include <linux/init.h> |
| 38 |
#include <linux/if_arp.h> |
| 39 |
#include <linux/proc_fs.h> |
| 40 |
#include <linux/seq_file.h> |
| 41 |
#include <linux/nsproxy.h> |
| 42 |
#include <linux/slab.h> |
| 43 |
#include <linux/jhash.h> |
| 44 |
#include <linux/siphash.h> |
| 45 |
#include <net/net_namespace.h> |
| 46 |
#include <net/snmp.h> |
| 47 |
#include <net/ipv6.h> |
| 48 |
#include <net/ip6_fib.h> |
| 49 |
#include <net/ip6_route.h> |
| 50 |
#include <net/ndisc.h> |
| 51 |
#include <net/addrconf.h> |
| 52 |
#include <net/tcp.h> |
| 53 |
#include <linux/rtnetlink.h> |
| 54 |
#include <net/dst.h> |
| 55 |
#include <net/dst_metadata.h> |
| 56 |
#include <net/xfrm.h> |
| 57 |
#include <net/netevent.h> |
| 58 |
#include <net/netlink.h> |
| 59 |
#include <net/rtnh.h> |
| 60 |
#include <net/lwtunnel.h> |
| 61 |
#include <net/ip_tunnels.h> |
| 62 |
#include <net/l3mdev.h> |
| 63 |
#include <net/ip.h> |
| 64 |
#include <linux/uaccess.h> |
| 65 |
#include <linux/btf_ids.h> |
| 66 |
|
| 67 |
#ifdef CONFIG_SYSCTL |
| 68 |
#include <linux/sysctl.h> |
| 69 |
#endif |
| 70 |
|
| 71 |
static int ip6_rt_type_to_error(u8 fib6_type); |
| 72 |
|
| 73 |
#define CREATE_TRACE_POINTS |
| 74 |
#include <trace/events/fib6.h> |
| 75 |
EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup); |
| 76 |
#undef CREATE_TRACE_POINTS |
| 77 |
|
| 78 |
enum rt6_nud_state { |
| 79 |
RT6_NUD_FAIL_HARD = -3, |
| 80 |
RT6_NUD_FAIL_PROBE = -2, |
| 81 |
RT6_NUD_FAIL_DO_RR = -1, |
| 82 |
RT6_NUD_SUCCEED = 1 |
| 83 |
}; |
| 84 |
|
| 85 |
INDIRECT_CALLABLE_SCOPE |
| 86 |
struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie); |
| 87 |
static unsigned int ip6_default_advmss(const struct dst_entry *dst); |
| 88 |
INDIRECT_CALLABLE_SCOPE |
| 89 |
unsigned int ip6_mtu(const struct dst_entry *dst); |
| 90 |
static void ip6_negative_advice(struct sock *sk, |
| 91 |
struct dst_entry *dst); |
| 92 |
static void ip6_dst_destroy(struct dst_entry *); |
| 93 |
static void ip6_dst_ifdown(struct dst_entry *, |
| 94 |
struct net_device *dev); |
| 95 |
static void ip6_dst_gc(struct dst_ops *ops); |
| 96 |
|
| 97 |
static int ip6_pkt_discard(struct sk_buff *skb); |
| 98 |
static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb); |
| 99 |
static int ip6_pkt_prohibit(struct sk_buff *skb); |
| 100 |
static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb); |
| 101 |
static void ip6_link_failure(struct sk_buff *skb); |
| 102 |
static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
| 103 |
struct sk_buff *skb, u32 mtu, |
| 104 |
bool confirm_neigh); |
| 105 |
static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, |
| 106 |
struct sk_buff *skb); |
| 107 |
static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif, |
| 108 |
int strict); |
| 109 |
static size_t rt6_nlmsg_size(struct fib6_info *f6i); |
| 110 |
static int rt6_fill_node(struct net *net, struct sk_buff *skb, |
| 111 |
struct fib6_info *rt, struct dst_entry *dst, |
| 112 |
struct in6_addr *dest, struct in6_addr *src, |
| 113 |
int iif, int type, u32 portid, u32 seq, |
| 114 |
unsigned int flags); |
| 115 |
static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res, |
| 116 |
const struct in6_addr *daddr, |
| 117 |
const struct in6_addr *saddr); |
| 118 |
|
| 119 |
#ifdef CONFIG_IPV6_ROUTE_INFO |
| 120 |
static struct fib6_info *rt6_add_route_info(struct net *net, |
| 121 |
const struct in6_addr *prefix, int prefixlen, |
| 122 |
const struct in6_addr *gwaddr, |
| 123 |
struct net_device *dev, |
| 124 |
unsigned int pref); |
| 125 |
static struct fib6_info *rt6_get_route_info(struct net *net, |
| 126 |
const struct in6_addr *prefix, int prefixlen, |
| 127 |
const struct in6_addr *gwaddr, |
| 128 |
struct net_device *dev); |
| 129 |
#endif |
| 130 |
|
| 131 |
struct uncached_list { |
| 132 |
spinlock_t lock; |
| 133 |
struct list_head head; |
| 134 |
}; |
| 135 |
|
| 136 |
static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list); |
| 137 |
|
| 138 |
void rt6_uncached_list_add(struct rt6_info *rt) |
| 139 |
{ |
| 140 |
struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list); |
| 141 |
|
| 142 |
rt->dst.rt_uncached_list = ul; |
| 143 |
|
| 144 |
spin_lock_bh(&ul->lock); |
| 145 |
list_add_tail(&rt->dst.rt_uncached, &ul->head); |
| 146 |
spin_unlock_bh(&ul->lock); |
| 147 |
} |
| 148 |
|
| 149 |
void rt6_uncached_list_del(struct rt6_info *rt) |
| 150 |
{ |
| 151 |
struct uncached_list *ul = rt->dst.rt_uncached_list; |
| 152 |
|
| 153 |
if (ul) { |
| 154 |
spin_lock_bh(&ul->lock); |
| 155 |
list_del_init(&rt->dst.rt_uncached); |
| 156 |
spin_unlock_bh(&ul->lock); |
| 157 |
} |
| 158 |
} |
| 159 |
|
| 160 |
static void rt6_uncached_list_flush_dev(struct net_device *dev) |
| 161 |
{ |
| 162 |
int cpu; |
| 163 |
|
| 164 |
for_each_possible_cpu(cpu) { |
| 165 |
struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu); |
| 166 |
struct rt6_info *rt, *safe; |
| 167 |
|
| 168 |
if (list_empty(&ul->head)) |
| 169 |
continue; |
| 170 |
|
| 171 |
spin_lock_bh(&ul->lock); |
| 172 |
list_for_each_entry_safe(rt, safe, &ul->head, dst.rt_uncached) { |
| 173 |
struct inet6_dev *rt_idev = rt->rt6i_idev; |
| 174 |
struct net_device *rt_dev = rt->dst.dev; |
| 175 |
bool handled = false; |
| 176 |
|
| 177 |
if (rt_idev && rt_idev->dev == dev) { |
| 178 |
rt->rt6i_idev = in6_dev_get(blackhole_netdev); |
| 179 |
in6_dev_put(rt_idev); |
| 180 |
handled = true; |
| 181 |
} |
| 182 |
|
| 183 |
if (rt_dev == dev) { |
| 184 |
rt->dst.dev = blackhole_netdev; |
| 185 |
netdev_ref_replace(rt_dev, blackhole_netdev, |
| 186 |
&rt->dst.dev_tracker, |
| 187 |
GFP_ATOMIC); |
| 188 |
handled = true; |
| 189 |
} |
| 190 |
if (handled) |
| 191 |
list_del_init(&rt->dst.rt_uncached); |
| 192 |
} |
| 193 |
spin_unlock_bh(&ul->lock); |
| 194 |
} |
| 195 |
} |
| 196 |
|
| 197 |
static inline const void *choose_neigh_daddr(const struct in6_addr *p, |
| 198 |
struct sk_buff *skb, |
| 199 |
const void *daddr) |
| 200 |
{ |
| 201 |
if (!ipv6_addr_any(p)) |
| 202 |
return (const void *) p; |
| 203 |
else if (skb) |
| 204 |
return &ipv6_hdr(skb)->daddr; |
| 205 |
return daddr; |
| 206 |
} |
| 207 |
|
| 208 |
struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw, |
| 209 |
struct net_device *dev, |
| 210 |
struct sk_buff *skb, |
| 211 |
const void *daddr) |
| 212 |
{ |
| 213 |
struct neighbour *n; |
| 214 |
|
| 215 |
daddr = choose_neigh_daddr(gw, skb, daddr); |
| 216 |
n = __ipv6_neigh_lookup(dev, daddr); |
| 217 |
if (n) |
| 218 |
return n; |
| 219 |
|
| 220 |
n = neigh_create(&nd_tbl, daddr, dev); |
| 221 |
return IS_ERR(n) ? NULL : n; |
| 222 |
} |
| 223 |
|
| 224 |
static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst, |
| 225 |
struct sk_buff *skb, |
| 226 |
const void *daddr) |
| 227 |
{ |
| 228 |
const struct rt6_info *rt = dst_rt6_info(dst); |
| 229 |
|
| 230 |
return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any), |
| 231 |
dst_dev(dst), skb, daddr); |
| 232 |
} |
| 233 |
|
| 234 |
static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr) |
| 235 |
{ |
| 236 |
const struct rt6_info *rt = dst_rt6_info(dst); |
| 237 |
struct net_device *dev = dst_dev(dst); |
| 238 |
|
| 239 |
daddr = choose_neigh_daddr(rt6_nexthop(rt, &in6addr_any), NULL, daddr); |
| 240 |
if (!daddr) |
| 241 |
return; |
| 242 |
if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) |
| 243 |
return; |
| 244 |
if (ipv6_addr_is_multicast((const struct in6_addr *)daddr)) |
| 245 |
return; |
| 246 |
__ipv6_confirm_neigh(dev, daddr); |
| 247 |
} |
| 248 |
|
| 249 |
static struct dst_ops ip6_dst_ops_template = { |
| 250 |
.family = AF_INET6, |
| 251 |
.gc = ip6_dst_gc, |
| 252 |
.gc_thresh = 1024, |
| 253 |
.check = ip6_dst_check, |
| 254 |
.default_advmss = ip6_default_advmss, |
| 255 |
.mtu = ip6_mtu, |
| 256 |
.cow_metrics = dst_cow_metrics_generic, |
| 257 |
.destroy = ip6_dst_destroy, |
| 258 |
.ifdown = ip6_dst_ifdown, |
| 259 |
.negative_advice = ip6_negative_advice, |
| 260 |
.link_failure = ip6_link_failure, |
| 261 |
.update_pmtu = ip6_rt_update_pmtu, |
| 262 |
.redirect = rt6_do_redirect, |
| 263 |
.local_out = __ip6_local_out, |
| 264 |
.neigh_lookup = ip6_dst_neigh_lookup, |
| 265 |
.confirm_neigh = ip6_confirm_neigh, |
| 266 |
}; |
| 267 |
|
| 268 |
static struct dst_ops ip6_dst_blackhole_ops = { |
| 269 |
.family = AF_INET6, |
| 270 |
.default_advmss = ip6_default_advmss, |
| 271 |
.neigh_lookup = ip6_dst_neigh_lookup, |
| 272 |
.check = ip6_dst_check, |
| 273 |
.destroy = ip6_dst_destroy, |
| 274 |
.cow_metrics = dst_cow_metrics_generic, |
| 275 |
.update_pmtu = dst_blackhole_update_pmtu, |
| 276 |
.redirect = dst_blackhole_redirect, |
| 277 |
.mtu = dst_blackhole_mtu, |
| 278 |
}; |
| 279 |
|
| 280 |
static const u32 ip6_template_metrics[RTAX_MAX] = { |
| 281 |
[RTAX_HOPLIMIT - 1] = 0, |
| 282 |
}; |
| 283 |
|
| 284 |
static const struct fib6_info fib6_null_entry_template = { |
| 285 |
.fib6_flags = (RTF_REJECT | RTF_NONEXTHOP), |
| 286 |
.fib6_protocol = RTPROT_KERNEL, |
| 287 |
.fib6_metric = ~(u32)0, |
| 288 |
.fib6_ref = REFCOUNT_INIT(1), |
| 289 |
.fib6_type = RTN_UNREACHABLE, |
| 290 |
.fib6_metrics = (struct dst_metrics *)&dst_default_metrics, |
| 291 |
}; |
| 292 |
|
| 293 |
static const struct rt6_info ip6_null_entry_template = { |
| 294 |
.dst = { |
| 295 |
.__rcuref = RCUREF_INIT(1), |
| 296 |
.__use = 1, |
| 297 |
.obsolete = DST_OBSOLETE_FORCE_CHK, |
| 298 |
.error = -ENETUNREACH, |
| 299 |
.input = ip6_pkt_discard, |
| 300 |
.output = ip6_pkt_discard_out, |
| 301 |
}, |
| 302 |
.rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP), |
| 303 |
}; |
| 304 |
|
| 305 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 306 |
|
| 307 |
static const struct rt6_info ip6_prohibit_entry_template = { |
| 308 |
.dst = { |
| 309 |
.__rcuref = RCUREF_INIT(1), |
| 310 |
.__use = 1, |
| 311 |
.obsolete = DST_OBSOLETE_FORCE_CHK, |
| 312 |
.error = -EACCES, |
| 313 |
.input = ip6_pkt_prohibit, |
| 314 |
.output = ip6_pkt_prohibit_out, |
| 315 |
}, |
| 316 |
.rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP), |
| 317 |
}; |
| 318 |
|
| 319 |
static const struct rt6_info ip6_blk_hole_entry_template = { |
| 320 |
.dst = { |
| 321 |
.__rcuref = RCUREF_INIT(1), |
| 322 |
.__use = 1, |
| 323 |
.obsolete = DST_OBSOLETE_FORCE_CHK, |
| 324 |
.error = -EINVAL, |
| 325 |
.input = dst_discard, |
| 326 |
.output = dst_discard_out, |
| 327 |
}, |
| 328 |
.rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP), |
| 329 |
}; |
| 330 |
|
| 331 |
#endif |
| 332 |
|
| 333 |
static void rt6_info_init(struct rt6_info *rt) |
| 334 |
{ |
| 335 |
memset_after(rt, 0, dst); |
| 336 |
} |
| 337 |
|
| 338 |
/* allocate dst with ip6_dst_ops */ |
| 339 |
struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev, |
| 340 |
int flags) |
| 341 |
{ |
| 342 |
struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev, |
| 343 |
DST_OBSOLETE_FORCE_CHK, flags); |
| 344 |
|
| 345 |
if (rt) { |
| 346 |
rt6_info_init(rt); |
| 347 |
atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc); |
| 348 |
} |
| 349 |
|
| 350 |
return rt; |
| 351 |
} |
| 352 |
EXPORT_SYMBOL(ip6_dst_alloc); |
| 353 |
|
| 354 |
static void ip6_dst_destroy(struct dst_entry *dst) |
| 355 |
{ |
| 356 |
struct rt6_info *rt = dst_rt6_info(dst); |
| 357 |
struct fib6_info *from; |
| 358 |
struct inet6_dev *idev; |
| 359 |
|
| 360 |
ip_dst_metrics_put(dst); |
| 361 |
rt6_uncached_list_del(rt); |
| 362 |
|
| 363 |
idev = rt->rt6i_idev; |
| 364 |
if (idev) { |
| 365 |
rt->rt6i_idev = NULL; |
| 366 |
in6_dev_put(idev); |
| 367 |
} |
| 368 |
|
| 369 |
from = unrcu_pointer(xchg(&rt->from, NULL)); |
| 370 |
fib6_info_release(from); |
| 371 |
} |
| 372 |
|
| 373 |
static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev) |
| 374 |
{ |
| 375 |
struct rt6_info *rt = dst_rt6_info(dst); |
| 376 |
struct inet6_dev *idev = rt->rt6i_idev; |
| 377 |
struct fib6_info *from; |
| 378 |
|
| 379 |
if (idev && idev->dev != blackhole_netdev) { |
| 380 |
struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev); |
| 381 |
|
| 382 |
if (blackhole_idev) { |
| 383 |
rt->rt6i_idev = blackhole_idev; |
| 384 |
in6_dev_put(idev); |
| 385 |
} |
| 386 |
} |
| 387 |
from = unrcu_pointer(xchg(&rt->from, NULL)); |
| 388 |
fib6_info_release(from); |
| 389 |
} |
| 390 |
|
| 391 |
static bool __rt6_check_expired(const struct rt6_info *rt) |
| 392 |
{ |
| 393 |
if (rt->rt6i_flags & RTF_EXPIRES) |
| 394 |
return time_after(jiffies, READ_ONCE(rt->dst.expires)); |
| 395 |
return false; |
| 396 |
} |
| 397 |
|
| 398 |
static bool rt6_check_expired(const struct rt6_info *rt) |
| 399 |
{ |
| 400 |
struct fib6_info *from; |
| 401 |
|
| 402 |
from = rcu_dereference(rt->from); |
| 403 |
|
| 404 |
if (rt->rt6i_flags & RTF_EXPIRES) { |
| 405 |
if (time_after(jiffies, READ_ONCE(rt->dst.expires))) |
| 406 |
return true; |
| 407 |
} else if (from) { |
| 408 |
return READ_ONCE(rt->dst.obsolete) != DST_OBSOLETE_FORCE_CHK || |
| 409 |
fib6_check_expired(from); |
| 410 |
} |
| 411 |
return false; |
| 412 |
} |
| 413 |
|
| 414 |
static struct fib6_info * |
| 415 |
rt6_multipath_first_sibling_rcu(const struct fib6_info *rt) |
| 416 |
{ |
| 417 |
struct fib6_info *iter; |
| 418 |
struct fib6_node *fn; |
| 419 |
|
| 420 |
fn = rcu_dereference(rt->fib6_node); |
| 421 |
if (!fn) |
| 422 |
goto out; |
| 423 |
iter = rcu_dereference(fn->leaf); |
| 424 |
if (!iter) |
| 425 |
goto out; |
| 426 |
|
| 427 |
while (iter) { |
| 428 |
if (iter->fib6_metric == rt->fib6_metric && |
| 429 |
rt6_qualify_for_ecmp(iter)) |
| 430 |
return iter; |
| 431 |
iter = rcu_dereference(iter->fib6_next); |
| 432 |
} |
| 433 |
|
| 434 |
out: |
| 435 |
return NULL; |
| 436 |
} |
| 437 |
|
| 438 |
void fib6_select_path(const struct net *net, struct fib6_result *res, |
| 439 |
struct flowi6 *fl6, int oif, bool have_oif_match, |
| 440 |
const struct sk_buff *skb, int strict) |
| 441 |
{ |
| 442 |
struct fib6_info *first, *match = res->f6i; |
| 443 |
struct fib6_info *sibling; |
| 444 |
int hash; |
| 445 |
|
| 446 |
if (!match->nh && (!match->fib6_nsiblings || have_oif_match)) |
| 447 |
goto out; |
| 448 |
|
| 449 |
if (match->nh && have_oif_match && res->nh) |
| 450 |
return; |
| 451 |
|
| 452 |
if (skb) |
| 453 |
IP6CB(skb)->flags |= IP6SKB_MULTIPATH; |
| 454 |
|
| 455 |
/* We might have already computed the hash for ICMPv6 errors. In such |
| 456 |
* case it will always be non-zero. Otherwise now is the time to do it. |
| 457 |
*/ |
| 458 |
if (!fl6->mp_hash && |
| 459 |
(!match->nh || nexthop_is_multipath(match->nh))) |
| 460 |
fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL); |
| 461 |
|
| 462 |
if (unlikely(match->nh)) { |
| 463 |
nexthop_path_fib6_result(res, fl6->mp_hash); |
| 464 |
return; |
| 465 |
} |
| 466 |
|
| 467 |
first = rt6_multipath_first_sibling_rcu(match); |
| 468 |
if (!first) |
| 469 |
goto out; |
| 470 |
|
| 471 |
hash = fl6->mp_hash; |
| 472 |
if (hash <= atomic_read(&first->fib6_nh->fib_nh_upper_bound)) { |
| 473 |
if (rt6_score_route(first->fib6_nh, first->fib6_flags, oif, |
| 474 |
strict) >= 0) |
| 475 |
match = first; |
| 476 |
goto out; |
| 477 |
} |
| 478 |
|
| 479 |
list_for_each_entry_rcu(sibling, &first->fib6_siblings, |
| 480 |
fib6_siblings) { |
| 481 |
const struct fib6_nh *nh = sibling->fib6_nh; |
| 482 |
int nh_upper_bound; |
| 483 |
|
| 484 |
nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound); |
| 485 |
if (hash > nh_upper_bound) |
| 486 |
continue; |
| 487 |
if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0) |
| 488 |
break; |
| 489 |
match = sibling; |
| 490 |
break; |
| 491 |
} |
| 492 |
|
| 493 |
out: |
| 494 |
res->f6i = match; |
| 495 |
res->nh = match->fib6_nh; |
| 496 |
} |
| 497 |
|
| 498 |
/* |
| 499 |
* Route lookup. rcu_read_lock() should be held. |
| 500 |
*/ |
| 501 |
|
| 502 |
static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh, |
| 503 |
const struct in6_addr *saddr, int oif, int flags) |
| 504 |
{ |
| 505 |
const struct net_device *dev; |
| 506 |
|
| 507 |
if (nh->fib_nh_flags & RTNH_F_DEAD) |
| 508 |
return false; |
| 509 |
|
| 510 |
dev = nh->fib_nh_dev; |
| 511 |
if (oif) { |
| 512 |
if (dev->ifindex == oif) |
| 513 |
return true; |
| 514 |
} else { |
| 515 |
if (ipv6_chk_addr(net, saddr, dev, |
| 516 |
flags & RT6_LOOKUP_F_IFACE)) |
| 517 |
return true; |
| 518 |
} |
| 519 |
|
| 520 |
return false; |
| 521 |
} |
| 522 |
|
| 523 |
struct fib6_nh_dm_arg { |
| 524 |
struct net *net; |
| 525 |
const struct in6_addr *saddr; |
| 526 |
int oif; |
| 527 |
int flags; |
| 528 |
struct fib6_nh *nh; |
| 529 |
}; |
| 530 |
|
| 531 |
static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg) |
| 532 |
{ |
| 533 |
struct fib6_nh_dm_arg *arg = _arg; |
| 534 |
|
| 535 |
arg->nh = nh; |
| 536 |
return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif, |
| 537 |
arg->flags); |
| 538 |
} |
| 539 |
|
| 540 |
/* returns fib6_nh from nexthop or NULL */ |
| 541 |
static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh, |
| 542 |
struct fib6_result *res, |
| 543 |
const struct in6_addr *saddr, |
| 544 |
int oif, int flags) |
| 545 |
{ |
| 546 |
struct fib6_nh_dm_arg arg = { |
| 547 |
.net = net, |
| 548 |
.saddr = saddr, |
| 549 |
.oif = oif, |
| 550 |
.flags = flags, |
| 551 |
}; |
| 552 |
|
| 553 |
if (nexthop_is_blackhole(nh)) |
| 554 |
return NULL; |
| 555 |
|
| 556 |
if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg)) |
| 557 |
return arg.nh; |
| 558 |
|
| 559 |
return NULL; |
| 560 |
} |
| 561 |
|
| 562 |
static void rt6_device_match(struct net *net, struct fib6_result *res, |
| 563 |
const struct in6_addr *saddr, int oif, int flags) |
| 564 |
{ |
| 565 |
struct fib6_info *f6i = res->f6i; |
| 566 |
struct fib6_info *spf6i; |
| 567 |
struct fib6_nh *nh; |
| 568 |
|
| 569 |
if (!oif && ipv6_addr_any(saddr)) { |
| 570 |
if (unlikely(f6i->nh)) { |
| 571 |
nh = nexthop_fib6_nh(f6i->nh); |
| 572 |
if (nexthop_is_blackhole(f6i->nh)) |
| 573 |
goto out_blackhole; |
| 574 |
} else { |
| 575 |
nh = f6i->fib6_nh; |
| 576 |
} |
| 577 |
if (!(nh->fib_nh_flags & RTNH_F_DEAD)) |
| 578 |
goto out; |
| 579 |
} |
| 580 |
|
| 581 |
for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) { |
| 582 |
bool matched = false; |
| 583 |
|
| 584 |
if (unlikely(spf6i->nh)) { |
| 585 |
nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr, |
| 586 |
oif, flags); |
| 587 |
if (nh) |
| 588 |
matched = true; |
| 589 |
} else { |
| 590 |
nh = spf6i->fib6_nh; |
| 591 |
if (__rt6_device_match(net, nh, saddr, oif, flags)) |
| 592 |
matched = true; |
| 593 |
} |
| 594 |
if (matched) { |
| 595 |
res->f6i = spf6i; |
| 596 |
goto out; |
| 597 |
} |
| 598 |
} |
| 599 |
|
| 600 |
if (oif && flags & RT6_LOOKUP_F_IFACE) { |
| 601 |
res->f6i = net->ipv6.fib6_null_entry; |
| 602 |
nh = res->f6i->fib6_nh; |
| 603 |
goto out; |
| 604 |
} |
| 605 |
|
| 606 |
if (unlikely(f6i->nh)) { |
| 607 |
nh = nexthop_fib6_nh(f6i->nh); |
| 608 |
if (nexthop_is_blackhole(f6i->nh)) |
| 609 |
goto out_blackhole; |
| 610 |
} else { |
| 611 |
nh = f6i->fib6_nh; |
| 612 |
} |
| 613 |
|
| 614 |
if (nh->fib_nh_flags & RTNH_F_DEAD) { |
| 615 |
res->f6i = net->ipv6.fib6_null_entry; |
| 616 |
nh = res->f6i->fib6_nh; |
| 617 |
} |
| 618 |
out: |
| 619 |
res->nh = nh; |
| 620 |
res->fib6_type = res->f6i->fib6_type; |
| 621 |
res->fib6_flags = res->f6i->fib6_flags; |
| 622 |
return; |
| 623 |
|
| 624 |
out_blackhole: |
| 625 |
res->fib6_flags |= RTF_REJECT; |
| 626 |
res->fib6_type = RTN_BLACKHOLE; |
| 627 |
res->nh = nh; |
| 628 |
} |
| 629 |
|
| 630 |
#ifdef CONFIG_IPV6_ROUTER_PREF |
| 631 |
struct __rt6_probe_work { |
| 632 |
struct work_struct work; |
| 633 |
struct in6_addr target; |
| 634 |
struct net_device *dev; |
| 635 |
netdevice_tracker dev_tracker; |
| 636 |
}; |
| 637 |
|
| 638 |
static void rt6_probe_deferred(struct work_struct *w) |
| 639 |
{ |
| 640 |
struct in6_addr mcaddr; |
| 641 |
struct __rt6_probe_work *work = |
| 642 |
container_of(w, struct __rt6_probe_work, work); |
| 643 |
|
| 644 |
addrconf_addr_solict_mult(&work->target, &mcaddr); |
| 645 |
ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0); |
| 646 |
netdev_put(work->dev, &work->dev_tracker); |
| 647 |
kfree(work); |
| 648 |
} |
| 649 |
|
| 650 |
static void rt6_probe(struct fib6_nh *fib6_nh) |
| 651 |
{ |
| 652 |
struct __rt6_probe_work *work = NULL; |
| 653 |
const struct in6_addr *nh_gw; |
| 654 |
unsigned long last_probe; |
| 655 |
struct neighbour *neigh; |
| 656 |
struct net_device *dev; |
| 657 |
struct inet6_dev *idev; |
| 658 |
|
| 659 |
/* |
| 660 |
* Okay, this does not seem to be appropriate |
| 661 |
* for now, however, we need to check if it |
| 662 |
* is really so; aka Router Reachability Probing. |
| 663 |
* |
| 664 |
* Router Reachability Probe MUST be rate-limited |
| 665 |
* to no more than one per minute. |
| 666 |
*/ |
| 667 |
if (!fib6_nh->fib_nh_gw_family) |
| 668 |
return; |
| 669 |
|
| 670 |
nh_gw = &fib6_nh->fib_nh_gw6; |
| 671 |
dev = fib6_nh->fib_nh_dev; |
| 672 |
rcu_read_lock(); |
| 673 |
last_probe = READ_ONCE(fib6_nh->last_probe); |
| 674 |
idev = __in6_dev_get(dev); |
| 675 |
if (!idev) |
| 676 |
goto out; |
| 677 |
neigh = __ipv6_neigh_lookup_noref(dev, nh_gw); |
| 678 |
if (neigh) { |
| 679 |
if (READ_ONCE(neigh->nud_state) & NUD_VALID) |
| 680 |
goto out; |
| 681 |
|
| 682 |
write_lock_bh(&neigh->lock); |
| 683 |
if (!(neigh->nud_state & NUD_VALID) && |
| 684 |
time_after(jiffies, |
| 685 |
neigh->updated + |
| 686 |
READ_ONCE(idev->cnf.rtr_probe_interval))) { |
| 687 |
work = kmalloc_obj(*work, GFP_ATOMIC); |
| 688 |
if (work) |
| 689 |
__neigh_set_probe_once(neigh); |
| 690 |
} |
| 691 |
write_unlock_bh(&neigh->lock); |
| 692 |
} else if (time_after(jiffies, last_probe + |
| 693 |
READ_ONCE(idev->cnf.rtr_probe_interval))) { |
| 694 |
work = kmalloc_obj(*work, GFP_ATOMIC); |
| 695 |
} |
| 696 |
|
| 697 |
if (!work || cmpxchg(&fib6_nh->last_probe, |
| 698 |
last_probe, jiffies) != last_probe) { |
| 699 |
kfree(work); |
| 700 |
} else { |
| 701 |
INIT_WORK(&work->work, rt6_probe_deferred); |
| 702 |
work->target = *nh_gw; |
| 703 |
netdev_hold(dev, &work->dev_tracker, GFP_ATOMIC); |
| 704 |
work->dev = dev; |
| 705 |
schedule_work(&work->work); |
| 706 |
} |
| 707 |
|
| 708 |
out: |
| 709 |
rcu_read_unlock(); |
| 710 |
} |
| 711 |
#else |
| 712 |
static inline void rt6_probe(struct fib6_nh *fib6_nh) |
| 713 |
{ |
| 714 |
} |
| 715 |
#endif |
| 716 |
|
| 717 |
/* |
| 718 |
* Default Router Selection (RFC 2461 6.3.6) |
| 719 |
*/ |
| 720 |
static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh) |
| 721 |
{ |
| 722 |
enum rt6_nud_state ret = RT6_NUD_FAIL_HARD; |
| 723 |
struct neighbour *neigh; |
| 724 |
|
| 725 |
rcu_read_lock(); |
| 726 |
neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev, |
| 727 |
&fib6_nh->fib_nh_gw6); |
| 728 |
if (neigh) { |
| 729 |
u8 nud_state = READ_ONCE(neigh->nud_state); |
| 730 |
|
| 731 |
if (nud_state & NUD_VALID) |
| 732 |
ret = RT6_NUD_SUCCEED; |
| 733 |
#ifdef CONFIG_IPV6_ROUTER_PREF |
| 734 |
else if (!(nud_state & NUD_FAILED)) |
| 735 |
ret = RT6_NUD_SUCCEED; |
| 736 |
else |
| 737 |
ret = RT6_NUD_FAIL_PROBE; |
| 738 |
#endif |
| 739 |
} else { |
| 740 |
ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ? |
| 741 |
RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR; |
| 742 |
} |
| 743 |
rcu_read_unlock(); |
| 744 |
|
| 745 |
return ret; |
| 746 |
} |
| 747 |
|
| 748 |
static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif, |
| 749 |
int strict) |
| 750 |
{ |
| 751 |
int m = 0; |
| 752 |
|
| 753 |
if (!oif || nh->fib_nh_dev->ifindex == oif) |
| 754 |
m = 2; |
| 755 |
|
| 756 |
if (!m && (strict & RT6_LOOKUP_F_IFACE)) |
| 757 |
return RT6_NUD_FAIL_HARD; |
| 758 |
#ifdef CONFIG_IPV6_ROUTER_PREF |
| 759 |
m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2; |
| 760 |
#endif |
| 761 |
if ((strict & RT6_LOOKUP_F_REACHABLE) && |
| 762 |
!(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) { |
| 763 |
int n = rt6_check_neigh(nh); |
| 764 |
if (n < 0) |
| 765 |
return n; |
| 766 |
} |
| 767 |
return m; |
| 768 |
} |
| 769 |
|
| 770 |
static bool find_match(struct fib6_nh *nh, u32 fib6_flags, |
| 771 |
int oif, int strict, int *mpri, bool *do_rr) |
| 772 |
{ |
| 773 |
bool match_do_rr = false; |
| 774 |
bool rc = false; |
| 775 |
int m; |
| 776 |
|
| 777 |
if (nh->fib_nh_flags & RTNH_F_DEAD) |
| 778 |
goto out; |
| 779 |
|
| 780 |
if (ip6_ignore_linkdown(nh->fib_nh_dev) && |
| 781 |
nh->fib_nh_flags & RTNH_F_LINKDOWN && |
| 782 |
!(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE)) |
| 783 |
goto out; |
| 784 |
|
| 785 |
m = rt6_score_route(nh, fib6_flags, oif, strict); |
| 786 |
if (m == RT6_NUD_FAIL_DO_RR) { |
| 787 |
match_do_rr = true; |
| 788 |
m = 0; /* lowest valid score */ |
| 789 |
} else if (m == RT6_NUD_FAIL_HARD) { |
| 790 |
goto out; |
| 791 |
} |
| 792 |
|
| 793 |
if (strict & RT6_LOOKUP_F_REACHABLE) |
| 794 |
rt6_probe(nh); |
| 795 |
|
| 796 |
/* note that m can be RT6_NUD_FAIL_PROBE at this point */ |
| 797 |
if (m > *mpri) { |
| 798 |
*do_rr = match_do_rr; |
| 799 |
*mpri = m; |
| 800 |
rc = true; |
| 801 |
} |
| 802 |
out: |
| 803 |
return rc; |
| 804 |
} |
| 805 |
|
| 806 |
struct fib6_nh_frl_arg { |
| 807 |
u32 flags; |
| 808 |
int oif; |
| 809 |
int strict; |
| 810 |
int *mpri; |
| 811 |
bool *do_rr; |
| 812 |
struct fib6_nh *nh; |
| 813 |
}; |
| 814 |
|
| 815 |
static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg) |
| 816 |
{ |
| 817 |
struct fib6_nh_frl_arg *arg = _arg; |
| 818 |
|
| 819 |
arg->nh = nh; |
| 820 |
return find_match(nh, arg->flags, arg->oif, arg->strict, |
| 821 |
arg->mpri, arg->do_rr); |
| 822 |
} |
| 823 |
|
| 824 |
static void __find_rr_leaf(struct fib6_info *f6i_start, |
| 825 |
struct fib6_info *nomatch, u32 metric, |
| 826 |
struct fib6_result *res, struct fib6_info **cont, |
| 827 |
int oif, int strict, bool *do_rr, int *mpri) |
| 828 |
{ |
| 829 |
struct fib6_info *f6i; |
| 830 |
|
| 831 |
for (f6i = f6i_start; |
| 832 |
f6i && f6i != nomatch; |
| 833 |
f6i = rcu_dereference(f6i->fib6_next)) { |
| 834 |
bool matched = false; |
| 835 |
struct fib6_nh *nh; |
| 836 |
|
| 837 |
if (cont && f6i->fib6_metric != metric) { |
| 838 |
*cont = f6i; |
| 839 |
return; |
| 840 |
} |
| 841 |
|
| 842 |
if (fib6_check_expired(f6i)) |
| 843 |
continue; |
| 844 |
|
| 845 |
if (unlikely(f6i->nh)) { |
| 846 |
struct fib6_nh_frl_arg arg = { |
| 847 |
.flags = f6i->fib6_flags, |
| 848 |
.oif = oif, |
| 849 |
.strict = strict, |
| 850 |
.mpri = mpri, |
| 851 |
.do_rr = do_rr |
| 852 |
}; |
| 853 |
|
| 854 |
if (nexthop_is_blackhole(f6i->nh)) { |
| 855 |
res->fib6_flags = RTF_REJECT; |
| 856 |
res->fib6_type = RTN_BLACKHOLE; |
| 857 |
res->f6i = f6i; |
| 858 |
res->nh = nexthop_fib6_nh(f6i->nh); |
| 859 |
return; |
| 860 |
} |
| 861 |
if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match, |
| 862 |
&arg)) { |
| 863 |
matched = true; |
| 864 |
nh = arg.nh; |
| 865 |
} |
| 866 |
} else { |
| 867 |
nh = f6i->fib6_nh; |
| 868 |
if (find_match(nh, f6i->fib6_flags, oif, strict, |
| 869 |
mpri, do_rr)) |
| 870 |
matched = true; |
| 871 |
} |
| 872 |
if (matched) { |
| 873 |
res->f6i = f6i; |
| 874 |
res->nh = nh; |
| 875 |
res->fib6_flags = f6i->fib6_flags; |
| 876 |
res->fib6_type = f6i->fib6_type; |
| 877 |
} |
| 878 |
} |
| 879 |
} |
| 880 |
|
| 881 |
static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf, |
| 882 |
struct fib6_info *rr_head, int oif, int strict, |
| 883 |
bool *do_rr, struct fib6_result *res) |
| 884 |
{ |
| 885 |
u32 metric = rr_head->fib6_metric; |
| 886 |
struct fib6_info *cont = NULL; |
| 887 |
int mpri = -1; |
| 888 |
|
| 889 |
__find_rr_leaf(rr_head, NULL, metric, res, &cont, |
| 890 |
oif, strict, do_rr, &mpri); |
| 891 |
|
| 892 |
__find_rr_leaf(leaf, rr_head, metric, res, &cont, |
| 893 |
oif, strict, do_rr, &mpri); |
| 894 |
|
| 895 |
if (res->f6i || !cont) |
| 896 |
return; |
| 897 |
|
| 898 |
__find_rr_leaf(cont, NULL, metric, res, NULL, |
| 899 |
oif, strict, do_rr, &mpri); |
| 900 |
} |
| 901 |
|
| 902 |
static void rt6_select(struct net *net, struct fib6_node *fn, int oif, |
| 903 |
struct fib6_result *res, int strict) |
| 904 |
{ |
| 905 |
struct fib6_info *leaf = rcu_dereference(fn->leaf); |
| 906 |
struct fib6_info *rt0; |
| 907 |
bool do_rr = false; |
| 908 |
int key_plen; |
| 909 |
|
| 910 |
/* make sure this function or its helpers sets f6i */ |
| 911 |
res->f6i = NULL; |
| 912 |
|
| 913 |
if (!leaf || leaf == net->ipv6.fib6_null_entry) |
| 914 |
goto out; |
| 915 |
|
| 916 |
rt0 = rcu_dereference(fn->rr_ptr); |
| 917 |
if (!rt0) |
| 918 |
rt0 = leaf; |
| 919 |
|
| 920 |
/* Double check to make sure fn is not an intermediate node |
| 921 |
* and fn->leaf does not points to its child's leaf |
| 922 |
* (This might happen if all routes under fn are deleted from |
| 923 |
* the tree and fib6_repair_tree() is called on the node.) |
| 924 |
*/ |
| 925 |
key_plen = rt0->fib6_dst.plen; |
| 926 |
#ifdef CONFIG_IPV6_SUBTREES |
| 927 |
if (rt0->fib6_src.plen) |
| 928 |
key_plen = rt0->fib6_src.plen; |
| 929 |
#endif |
| 930 |
if (fn->fn_bit != key_plen) |
| 931 |
goto out; |
| 932 |
|
| 933 |
find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res); |
| 934 |
if (do_rr) { |
| 935 |
struct fib6_info *next = rcu_dereference(rt0->fib6_next); |
| 936 |
|
| 937 |
/* no entries matched; do round-robin */ |
| 938 |
if (!next || next->fib6_metric != rt0->fib6_metric) |
| 939 |
next = leaf; |
| 940 |
|
| 941 |
if (next != rt0) { |
| 942 |
spin_lock_bh(&leaf->fib6_table->tb6_lock); |
| 943 |
/* make sure next is not being deleted from the tree */ |
| 944 |
if (next->fib6_node) |
| 945 |
rcu_assign_pointer(fn->rr_ptr, next); |
| 946 |
spin_unlock_bh(&leaf->fib6_table->tb6_lock); |
| 947 |
} |
| 948 |
} |
| 949 |
|
| 950 |
out: |
| 951 |
if (!res->f6i) { |
| 952 |
res->f6i = net->ipv6.fib6_null_entry; |
| 953 |
res->nh = res->f6i->fib6_nh; |
| 954 |
res->fib6_flags = res->f6i->fib6_flags; |
| 955 |
res->fib6_type = res->f6i->fib6_type; |
| 956 |
} |
| 957 |
} |
| 958 |
|
| 959 |
static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res) |
| 960 |
{ |
| 961 |
return (res->f6i->fib6_flags & RTF_NONEXTHOP) || |
| 962 |
res->nh->fib_nh_gw_family; |
| 963 |
} |
| 964 |
|
| 965 |
#ifdef CONFIG_IPV6_ROUTE_INFO |
| 966 |
int rt6_route_rcv(struct net_device *dev, u8 *opt, int len, |
| 967 |
const struct in6_addr *gwaddr) |
| 968 |
{ |
| 969 |
struct net *net = dev_net(dev); |
| 970 |
struct route_info *rinfo = (struct route_info *) opt; |
| 971 |
struct in6_addr prefix_buf, *prefix; |
| 972 |
struct fib6_table *table; |
| 973 |
unsigned int pref; |
| 974 |
unsigned long lifetime; |
| 975 |
struct fib6_info *rt; |
| 976 |
|
| 977 |
if (len < sizeof(struct route_info)) { |
| 978 |
return -EINVAL; |
| 979 |
} |
| 980 |
|
| 981 |
/* Sanity check for prefix_len and length */ |
| 982 |
if (rinfo->length > 3) { |
| 983 |
return -EINVAL; |
| 984 |
} else if (rinfo->prefix_len > 128) { |
| 985 |
return -EINVAL; |
| 986 |
} else if (rinfo->prefix_len > 64) { |
| 987 |
if (rinfo->length < 2) { |
| 988 |
return -EINVAL; |
| 989 |
} |
| 990 |
} else if (rinfo->prefix_len > 0) { |
| 991 |
if (rinfo->length < 1) { |
| 992 |
return -EINVAL; |
| 993 |
} |
| 994 |
} |
| 995 |
|
| 996 |
pref = rinfo->route_pref; |
| 997 |
if (pref == ICMPV6_ROUTER_PREF_INVALID) |
| 998 |
return -EINVAL; |
| 999 |
|
| 1000 |
lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ); |
| 1001 |
|
| 1002 |
if (rinfo->length == 3) |
| 1003 |
prefix = (struct in6_addr *)rinfo->prefix; |
| 1004 |
else { |
| 1005 |
/* this function is safe */ |
| 1006 |
ipv6_addr_prefix(&prefix_buf, |
| 1007 |
(struct in6_addr *)rinfo->prefix, |
| 1008 |
rinfo->prefix_len); |
| 1009 |
prefix = &prefix_buf; |
| 1010 |
} |
| 1011 |
|
| 1012 |
if (rinfo->prefix_len == 0) |
| 1013 |
rt = rt6_get_dflt_router(net, gwaddr, dev); |
| 1014 |
else |
| 1015 |
rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, |
| 1016 |
gwaddr, dev); |
| 1017 |
|
| 1018 |
if (rt && !lifetime) { |
| 1019 |
ip6_del_rt(net, rt, false); |
| 1020 |
rt = NULL; |
| 1021 |
} |
| 1022 |
|
| 1023 |
if (!rt && lifetime) |
| 1024 |
rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, |
| 1025 |
dev, pref); |
| 1026 |
else if (rt) |
| 1027 |
rt->fib6_flags = RTF_ROUTEINFO | |
| 1028 |
(rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref); |
| 1029 |
|
| 1030 |
if (rt) { |
| 1031 |
table = rt->fib6_table; |
| 1032 |
spin_lock_bh(&table->tb6_lock); |
| 1033 |
|
| 1034 |
if (!addrconf_finite_timeout(lifetime)) { |
| 1035 |
fib6_clean_expires(rt); |
| 1036 |
fib6_may_remove_gc_list(net, rt); |
| 1037 |
} else { |
| 1038 |
fib6_set_expires(rt, jiffies + HZ * lifetime); |
| 1039 |
fib6_add_gc_list(rt); |
| 1040 |
} |
| 1041 |
|
| 1042 |
spin_unlock_bh(&table->tb6_lock); |
| 1043 |
|
| 1044 |
fib6_info_release(rt); |
| 1045 |
} |
| 1046 |
return 0; |
| 1047 |
} |
| 1048 |
#endif |
| 1049 |
|
| 1050 |
/* |
| 1051 |
* Misc support functions |
| 1052 |
*/ |
| 1053 |
|
| 1054 |
/* called with rcu_lock held */ |
| 1055 |
static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res) |
| 1056 |
{ |
| 1057 |
struct net_device *dev = res->nh->fib_nh_dev; |
| 1058 |
|
| 1059 |
if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) { |
| 1060 |
/* for copies of local routes, dst->dev needs to be the |
| 1061 |
* device if it is a master device, the master device if |
| 1062 |
* device is enslaved, and the loopback as the default |
| 1063 |
*/ |
| 1064 |
if (netif_is_l3_slave(dev) && |
| 1065 |
!rt6_need_strict(&res->f6i->fib6_dst.addr)) |
| 1066 |
dev = l3mdev_master_dev_rcu(dev) ? : |
| 1067 |
dev_net(dev)->loopback_dev; |
| 1068 |
else if (!netif_is_l3_master(dev)) |
| 1069 |
dev = dev_net(dev)->loopback_dev; |
| 1070 |
/* last case is netif_is_l3_master(dev) is true in which |
| 1071 |
* case we want dev returned to be dev |
| 1072 |
*/ |
| 1073 |
} |
| 1074 |
|
| 1075 |
return dev; |
| 1076 |
} |
| 1077 |
|
| 1078 |
static const int fib6_prop[RTN_MAX + 1] = { |
| 1079 |
[RTN_UNSPEC] = 0, |
| 1080 |
[RTN_UNICAST] = 0, |
| 1081 |
[RTN_LOCAL] = 0, |
| 1082 |
[RTN_BROADCAST] = 0, |
| 1083 |
[RTN_ANYCAST] = 0, |
| 1084 |
[RTN_MULTICAST] = 0, |
| 1085 |
[RTN_BLACKHOLE] = -EINVAL, |
| 1086 |
[RTN_UNREACHABLE] = -EHOSTUNREACH, |
| 1087 |
[RTN_PROHIBIT] = -EACCES, |
| 1088 |
[RTN_THROW] = -EAGAIN, |
| 1089 |
[RTN_NAT] = -EINVAL, |
| 1090 |
[RTN_XRESOLVE] = -EINVAL, |
| 1091 |
}; |
| 1092 |
|
| 1093 |
static int ip6_rt_type_to_error(u8 fib6_type) |
| 1094 |
{ |
| 1095 |
return fib6_prop[fib6_type]; |
| 1096 |
} |
| 1097 |
|
| 1098 |
static unsigned short fib6_info_dst_flags(struct fib6_info *rt) |
| 1099 |
{ |
| 1100 |
unsigned short flags = 0; |
| 1101 |
|
| 1102 |
if (rt->dst_nocount) |
| 1103 |
flags |= DST_NOCOUNT; |
| 1104 |
if (rt->dst_nopolicy) |
| 1105 |
flags |= DST_NOPOLICY; |
| 1106 |
|
| 1107 |
return flags; |
| 1108 |
} |
| 1109 |
|
| 1110 |
static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type) |
| 1111 |
{ |
| 1112 |
rt->dst.error = ip6_rt_type_to_error(fib6_type); |
| 1113 |
|
| 1114 |
switch (fib6_type) { |
| 1115 |
case RTN_BLACKHOLE: |
| 1116 |
rt->dst.output = dst_discard_out; |
| 1117 |
rt->dst.input = dst_discard; |
| 1118 |
break; |
| 1119 |
case RTN_PROHIBIT: |
| 1120 |
rt->dst.output = ip6_pkt_prohibit_out; |
| 1121 |
rt->dst.input = ip6_pkt_prohibit; |
| 1122 |
break; |
| 1123 |
case RTN_THROW: |
| 1124 |
case RTN_UNREACHABLE: |
| 1125 |
default: |
| 1126 |
rt->dst.output = ip6_pkt_discard_out; |
| 1127 |
rt->dst.input = ip6_pkt_discard; |
| 1128 |
break; |
| 1129 |
} |
| 1130 |
} |
| 1131 |
|
| 1132 |
static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res) |
| 1133 |
{ |
| 1134 |
struct fib6_info *f6i = res->f6i; |
| 1135 |
|
| 1136 |
if (res->fib6_flags & RTF_REJECT) { |
| 1137 |
ip6_rt_init_dst_reject(rt, res->fib6_type); |
| 1138 |
return; |
| 1139 |
} |
| 1140 |
|
| 1141 |
rt->dst.error = 0; |
| 1142 |
rt->dst.output = ip6_output; |
| 1143 |
|
| 1144 |
if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) { |
| 1145 |
rt->dst.input = ip6_input; |
| 1146 |
} else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) { |
| 1147 |
rt->dst.input = ip6_mc_input; |
| 1148 |
rt->dst.output = ip6_mr_output; |
| 1149 |
} else { |
| 1150 |
rt->dst.input = ip6_forward; |
| 1151 |
} |
| 1152 |
|
| 1153 |
if (res->nh->fib_nh_lws) { |
| 1154 |
rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws); |
| 1155 |
lwtunnel_set_redirect(&rt->dst); |
| 1156 |
} |
| 1157 |
|
| 1158 |
rt->dst.lastuse = jiffies; |
| 1159 |
} |
| 1160 |
|
| 1161 |
/* Caller must already hold reference to @from */ |
| 1162 |
static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from) |
| 1163 |
{ |
| 1164 |
rt->rt6i_flags &= ~RTF_EXPIRES; |
| 1165 |
rcu_assign_pointer(rt->from, from); |
| 1166 |
ip_dst_init_metrics(&rt->dst, from->fib6_metrics); |
| 1167 |
} |
| 1168 |
|
| 1169 |
/* Caller must already hold reference to f6i in result */ |
| 1170 |
static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res) |
| 1171 |
{ |
| 1172 |
const struct fib6_nh *nh = res->nh; |
| 1173 |
const struct net_device *dev = nh->fib_nh_dev; |
| 1174 |
struct fib6_info *f6i = res->f6i; |
| 1175 |
|
| 1176 |
ip6_rt_init_dst(rt, res); |
| 1177 |
|
| 1178 |
rt->rt6i_dst = f6i->fib6_dst; |
| 1179 |
rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL; |
| 1180 |
rt->rt6i_flags = res->fib6_flags; |
| 1181 |
if (nh->fib_nh_gw_family) { |
| 1182 |
rt->rt6i_gateway = nh->fib_nh_gw6; |
| 1183 |
rt->rt6i_flags |= RTF_GATEWAY; |
| 1184 |
} |
| 1185 |
rt6_set_from(rt, f6i); |
| 1186 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1187 |
rt->rt6i_src = f6i->fib6_src; |
| 1188 |
#endif |
| 1189 |
} |
| 1190 |
|
| 1191 |
static struct fib6_node* fib6_backtrack(struct fib6_node *fn, |
| 1192 |
struct in6_addr *saddr) |
| 1193 |
{ |
| 1194 |
struct fib6_node *pn, *sn; |
| 1195 |
while (1) { |
| 1196 |
if (fn->fn_flags & RTN_TL_ROOT) |
| 1197 |
return NULL; |
| 1198 |
pn = rcu_dereference(fn->parent); |
| 1199 |
sn = FIB6_SUBTREE(pn); |
| 1200 |
if (sn && sn != fn) |
| 1201 |
fn = fib6_node_lookup(sn, NULL, saddr); |
| 1202 |
else |
| 1203 |
fn = pn; |
| 1204 |
if (fn->fn_flags & RTN_RTINFO) |
| 1205 |
return fn; |
| 1206 |
} |
| 1207 |
} |
| 1208 |
|
| 1209 |
static bool ip6_hold_safe(struct net *net, struct rt6_info **prt) |
| 1210 |
{ |
| 1211 |
struct rt6_info *rt = *prt; |
| 1212 |
|
| 1213 |
if (dst_hold_safe(&rt->dst)) |
| 1214 |
return true; |
| 1215 |
if (net) { |
| 1216 |
rt = net->ipv6.ip6_null_entry; |
| 1217 |
dst_hold(&rt->dst); |
| 1218 |
} else { |
| 1219 |
rt = NULL; |
| 1220 |
} |
| 1221 |
*prt = rt; |
| 1222 |
return false; |
| 1223 |
} |
| 1224 |
|
| 1225 |
/* called with rcu_lock held */ |
| 1226 |
static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res) |
| 1227 |
{ |
| 1228 |
struct net_device *dev = res->nh->fib_nh_dev; |
| 1229 |
struct fib6_info *f6i = res->f6i; |
| 1230 |
unsigned short flags; |
| 1231 |
struct rt6_info *nrt; |
| 1232 |
|
| 1233 |
if (!fib6_info_hold_safe(f6i)) |
| 1234 |
goto fallback; |
| 1235 |
|
| 1236 |
flags = fib6_info_dst_flags(f6i); |
| 1237 |
nrt = ip6_dst_alloc(dev_net(dev), dev, flags); |
| 1238 |
if (!nrt) { |
| 1239 |
fib6_info_release(f6i); |
| 1240 |
goto fallback; |
| 1241 |
} |
| 1242 |
|
| 1243 |
ip6_rt_copy_init(nrt, res); |
| 1244 |
return nrt; |
| 1245 |
|
| 1246 |
fallback: |
| 1247 |
nrt = dev_net(dev)->ipv6.ip6_null_entry; |
| 1248 |
dst_hold(&nrt->dst); |
| 1249 |
return nrt; |
| 1250 |
} |
| 1251 |
|
| 1252 |
INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_lookup(struct net *net, |
| 1253 |
struct fib6_table *table, |
| 1254 |
struct flowi6 *fl6, |
| 1255 |
const struct sk_buff *skb, |
| 1256 |
int flags) |
| 1257 |
{ |
| 1258 |
struct fib6_result res = {}; |
| 1259 |
struct fib6_node *fn; |
| 1260 |
struct rt6_info *rt; |
| 1261 |
|
| 1262 |
rcu_read_lock(); |
| 1263 |
fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr); |
| 1264 |
restart: |
| 1265 |
res.f6i = rcu_dereference(fn->leaf); |
| 1266 |
if (!res.f6i) |
| 1267 |
res.f6i = net->ipv6.fib6_null_entry; |
| 1268 |
else |
| 1269 |
rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif, |
| 1270 |
flags); |
| 1271 |
|
| 1272 |
if (res.f6i == net->ipv6.fib6_null_entry) { |
| 1273 |
fn = fib6_backtrack(fn, &fl6->saddr); |
| 1274 |
if (fn) |
| 1275 |
goto restart; |
| 1276 |
|
| 1277 |
rt = net->ipv6.ip6_null_entry; |
| 1278 |
dst_hold(&rt->dst); |
| 1279 |
goto out; |
| 1280 |
} else if (res.fib6_flags & RTF_REJECT) { |
| 1281 |
goto do_create; |
| 1282 |
} |
| 1283 |
|
| 1284 |
fib6_select_path(net, &res, fl6, fl6->flowi6_oif, |
| 1285 |
fl6->flowi6_oif != 0, skb, flags); |
| 1286 |
|
| 1287 |
/* Search through exception table */ |
| 1288 |
rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr); |
| 1289 |
if (rt) { |
| 1290 |
if (ip6_hold_safe(net, &rt)) |
| 1291 |
dst_use_noref(&rt->dst, jiffies); |
| 1292 |
} else { |
| 1293 |
do_create: |
| 1294 |
rt = ip6_create_rt_rcu(&res); |
| 1295 |
} |
| 1296 |
|
| 1297 |
out: |
| 1298 |
trace_fib6_table_lookup(net, &res, table, fl6); |
| 1299 |
|
| 1300 |
rcu_read_unlock(); |
| 1301 |
|
| 1302 |
return rt; |
| 1303 |
} |
| 1304 |
|
| 1305 |
struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6, |
| 1306 |
const struct sk_buff *skb, int flags) |
| 1307 |
{ |
| 1308 |
return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup); |
| 1309 |
} |
| 1310 |
EXPORT_SYMBOL_GPL(ip6_route_lookup); |
| 1311 |
|
| 1312 |
struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr, |
| 1313 |
const struct in6_addr *saddr, int oif, |
| 1314 |
const struct sk_buff *skb, int strict) |
| 1315 |
{ |
| 1316 |
struct flowi6 fl6 = { |
| 1317 |
.flowi6_oif = oif, |
| 1318 |
.daddr = *daddr, |
| 1319 |
}; |
| 1320 |
struct dst_entry *dst; |
| 1321 |
int flags = strict ? RT6_LOOKUP_F_IFACE : 0; |
| 1322 |
|
| 1323 |
if (saddr) { |
| 1324 |
memcpy(&fl6.saddr, saddr, sizeof(*saddr)); |
| 1325 |
flags |= RT6_LOOKUP_F_HAS_SADDR; |
| 1326 |
} |
| 1327 |
|
| 1328 |
dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup); |
| 1329 |
if (dst->error == 0) |
| 1330 |
return dst_rt6_info(dst); |
| 1331 |
|
| 1332 |
dst_release(dst); |
| 1333 |
|
| 1334 |
return NULL; |
| 1335 |
} |
| 1336 |
EXPORT_SYMBOL(rt6_lookup); |
| 1337 |
|
| 1338 |
/* ip6_ins_rt is called with FREE table->tb6_lock. |
| 1339 |
* It takes new route entry, the addition fails by any reason the |
| 1340 |
* route is released. |
| 1341 |
* Caller must hold dst before calling it. |
| 1342 |
*/ |
| 1343 |
|
| 1344 |
static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info, |
| 1345 |
struct netlink_ext_ack *extack) |
| 1346 |
{ |
| 1347 |
int err; |
| 1348 |
struct fib6_table *table; |
| 1349 |
|
| 1350 |
table = rt->fib6_table; |
| 1351 |
spin_lock_bh(&table->tb6_lock); |
| 1352 |
err = fib6_add(&table->tb6_root, rt, info, extack); |
| 1353 |
spin_unlock_bh(&table->tb6_lock); |
| 1354 |
|
| 1355 |
return err; |
| 1356 |
} |
| 1357 |
|
| 1358 |
int ip6_ins_rt(struct net *net, struct fib6_info *rt) |
| 1359 |
{ |
| 1360 |
struct nl_info info = { .nl_net = net, }; |
| 1361 |
|
| 1362 |
return __ip6_ins_rt(rt, &info, NULL); |
| 1363 |
} |
| 1364 |
|
| 1365 |
static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res, |
| 1366 |
const struct in6_addr *daddr, |
| 1367 |
const struct in6_addr *saddr) |
| 1368 |
{ |
| 1369 |
struct fib6_info *f6i = res->f6i; |
| 1370 |
struct net_device *dev; |
| 1371 |
struct rt6_info *rt; |
| 1372 |
|
| 1373 |
/* |
| 1374 |
* Clone the route. |
| 1375 |
*/ |
| 1376 |
|
| 1377 |
if (!fib6_info_hold_safe(f6i)) |
| 1378 |
return NULL; |
| 1379 |
|
| 1380 |
dev = ip6_rt_get_dev_rcu(res); |
| 1381 |
rt = ip6_dst_alloc(dev_net(dev), dev, 0); |
| 1382 |
if (!rt) { |
| 1383 |
fib6_info_release(f6i); |
| 1384 |
return NULL; |
| 1385 |
} |
| 1386 |
|
| 1387 |
ip6_rt_copy_init(rt, res); |
| 1388 |
rt->rt6i_flags |= RTF_CACHE; |
| 1389 |
rt->rt6i_dst.addr = *daddr; |
| 1390 |
rt->rt6i_dst.plen = 128; |
| 1391 |
|
| 1392 |
if (!rt6_is_gw_or_nonexthop(res)) { |
| 1393 |
if (f6i->fib6_dst.plen != 128 && |
| 1394 |
ipv6_addr_equal(&f6i->fib6_dst.addr, daddr)) |
| 1395 |
rt->rt6i_flags |= RTF_ANYCAST; |
| 1396 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1397 |
if (rt->rt6i_src.plen && saddr) { |
| 1398 |
rt->rt6i_src.addr = *saddr; |
| 1399 |
rt->rt6i_src.plen = 128; |
| 1400 |
} |
| 1401 |
#endif |
| 1402 |
} |
| 1403 |
|
| 1404 |
return rt; |
| 1405 |
} |
| 1406 |
|
| 1407 |
static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res) |
| 1408 |
{ |
| 1409 |
struct fib6_info *f6i = res->f6i; |
| 1410 |
unsigned short flags = fib6_info_dst_flags(f6i); |
| 1411 |
struct net_device *dev; |
| 1412 |
struct rt6_info *pcpu_rt; |
| 1413 |
|
| 1414 |
if (!fib6_info_hold_safe(f6i)) |
| 1415 |
return NULL; |
| 1416 |
|
| 1417 |
rcu_read_lock(); |
| 1418 |
dev = ip6_rt_get_dev_rcu(res); |
| 1419 |
pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags | DST_NOCOUNT); |
| 1420 |
rcu_read_unlock(); |
| 1421 |
if (!pcpu_rt) { |
| 1422 |
fib6_info_release(f6i); |
| 1423 |
return NULL; |
| 1424 |
} |
| 1425 |
ip6_rt_copy_init(pcpu_rt, res); |
| 1426 |
pcpu_rt->rt6i_flags |= RTF_PCPU; |
| 1427 |
|
| 1428 |
if (f6i->nh) |
| 1429 |
pcpu_rt->sernum = rt_genid_ipv6(dev_net(dev)); |
| 1430 |
|
| 1431 |
return pcpu_rt; |
| 1432 |
} |
| 1433 |
|
| 1434 |
static bool rt6_is_valid(const struct rt6_info *rt6) |
| 1435 |
{ |
| 1436 |
return rt6->sernum == rt_genid_ipv6(dev_net(rt6->dst.dev)); |
| 1437 |
} |
| 1438 |
|
| 1439 |
/* It should be called with rcu_read_lock() acquired */ |
| 1440 |
static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res) |
| 1441 |
{ |
| 1442 |
struct rt6_info *pcpu_rt; |
| 1443 |
|
| 1444 |
pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu); |
| 1445 |
|
| 1446 |
if (pcpu_rt && pcpu_rt->sernum && !rt6_is_valid(pcpu_rt)) { |
| 1447 |
struct rt6_info *prev, **p; |
| 1448 |
|
| 1449 |
p = this_cpu_ptr(res->nh->rt6i_pcpu); |
| 1450 |
/* Paired with READ_ONCE() in __fib6_drop_pcpu_from() */ |
| 1451 |
prev = xchg(p, NULL); |
| 1452 |
if (prev) { |
| 1453 |
dst_dev_put(&prev->dst); |
| 1454 |
dst_release(&prev->dst); |
| 1455 |
} |
| 1456 |
|
| 1457 |
pcpu_rt = NULL; |
| 1458 |
} |
| 1459 |
|
| 1460 |
return pcpu_rt; |
| 1461 |
} |
| 1462 |
|
| 1463 |
static struct rt6_info *rt6_make_pcpu_route(struct net *net, |
| 1464 |
const struct fib6_result *res) |
| 1465 |
{ |
| 1466 |
struct rt6_info *pcpu_rt, *prev, **p; |
| 1467 |
|
| 1468 |
pcpu_rt = ip6_rt_pcpu_alloc(res); |
| 1469 |
if (!pcpu_rt) |
| 1470 |
return NULL; |
| 1471 |
|
| 1472 |
p = this_cpu_ptr(res->nh->rt6i_pcpu); |
| 1473 |
prev = cmpxchg(p, NULL, pcpu_rt); |
| 1474 |
if (unlikely(prev)) { |
| 1475 |
/* |
| 1476 |
* Another task on this CPU already installed a pcpu_rt. |
| 1477 |
* This can happen on PREEMPT_RT where preemption is possible. |
| 1478 |
* Free our allocation and return the existing one. |
| 1479 |
*/ |
| 1480 |
WARN_ON_ONCE(!IS_ENABLED(CONFIG_PREEMPT_RT)); |
| 1481 |
|
| 1482 |
dst_dev_put(&pcpu_rt->dst); |
| 1483 |
dst_release(&pcpu_rt->dst); |
| 1484 |
return prev; |
| 1485 |
} |
| 1486 |
|
| 1487 |
if (res->f6i->fib6_destroying) { |
| 1488 |
struct fib6_info *from; |
| 1489 |
|
| 1490 |
from = unrcu_pointer(xchg(&pcpu_rt->from, NULL)); |
| 1491 |
fib6_info_release(from); |
| 1492 |
} |
| 1493 |
|
| 1494 |
return pcpu_rt; |
| 1495 |
} |
| 1496 |
|
| 1497 |
/* exception hash table implementation |
| 1498 |
*/ |
| 1499 |
static DEFINE_SPINLOCK(rt6_exception_lock); |
| 1500 |
|
| 1501 |
/* Remove rt6_ex from hash table and free the memory |
| 1502 |
* Caller must hold rt6_exception_lock |
| 1503 |
*/ |
| 1504 |
static void rt6_remove_exception(struct rt6_exception_bucket *bucket, |
| 1505 |
struct rt6_exception *rt6_ex) |
| 1506 |
{ |
| 1507 |
struct net *net; |
| 1508 |
|
| 1509 |
if (!bucket || !rt6_ex) |
| 1510 |
return; |
| 1511 |
|
| 1512 |
net = dev_net(rt6_ex->rt6i->dst.dev); |
| 1513 |
net->ipv6.rt6_stats->fib_rt_cache--; |
| 1514 |
|
| 1515 |
/* purge completely the exception to allow releasing the held resources: |
| 1516 |
* some [sk] cache may keep the dst around for unlimited time |
| 1517 |
*/ |
| 1518 |
dst_dev_put(&rt6_ex->rt6i->dst); |
| 1519 |
|
| 1520 |
hlist_del_rcu(&rt6_ex->hlist); |
| 1521 |
dst_release(&rt6_ex->rt6i->dst); |
| 1522 |
kfree_rcu(rt6_ex, rcu); |
| 1523 |
WARN_ON_ONCE(!bucket->depth); |
| 1524 |
bucket->depth--; |
| 1525 |
} |
| 1526 |
|
| 1527 |
/* Remove oldest rt6_ex in bucket and free the memory |
| 1528 |
* Caller must hold rt6_exception_lock |
| 1529 |
*/ |
| 1530 |
static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket) |
| 1531 |
{ |
| 1532 |
struct rt6_exception *rt6_ex, *oldest = NULL; |
| 1533 |
|
| 1534 |
if (!bucket) |
| 1535 |
return; |
| 1536 |
|
| 1537 |
hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) { |
| 1538 |
if (!oldest || time_before(rt6_ex->stamp, oldest->stamp)) |
| 1539 |
oldest = rt6_ex; |
| 1540 |
} |
| 1541 |
rt6_remove_exception(bucket, oldest); |
| 1542 |
} |
| 1543 |
|
| 1544 |
static u32 rt6_exception_hash(const struct in6_addr *dst, |
| 1545 |
const struct in6_addr *src) |
| 1546 |
{ |
| 1547 |
static siphash_aligned_key_t rt6_exception_key; |
| 1548 |
struct { |
| 1549 |
struct in6_addr dst; |
| 1550 |
struct in6_addr src; |
| 1551 |
} __aligned(SIPHASH_ALIGNMENT) combined = { |
| 1552 |
.dst = *dst, |
| 1553 |
}; |
| 1554 |
u64 val; |
| 1555 |
|
| 1556 |
net_get_random_once(&rt6_exception_key, sizeof(rt6_exception_key)); |
| 1557 |
|
| 1558 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1559 |
if (src) |
| 1560 |
combined.src = *src; |
| 1561 |
#endif |
| 1562 |
val = siphash(&combined, sizeof(combined), &rt6_exception_key); |
| 1563 |
|
| 1564 |
return hash_64(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT); |
| 1565 |
} |
| 1566 |
|
| 1567 |
/* Helper function to find the cached rt in the hash table |
| 1568 |
* and update bucket pointer to point to the bucket for this |
| 1569 |
* (daddr, saddr) pair |
| 1570 |
* Caller must hold rt6_exception_lock |
| 1571 |
*/ |
| 1572 |
static struct rt6_exception * |
| 1573 |
__rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket, |
| 1574 |
const struct in6_addr *daddr, |
| 1575 |
const struct in6_addr *saddr) |
| 1576 |
{ |
| 1577 |
struct rt6_exception *rt6_ex; |
| 1578 |
u32 hval; |
| 1579 |
|
| 1580 |
if (!(*bucket) || !daddr) |
| 1581 |
return NULL; |
| 1582 |
|
| 1583 |
hval = rt6_exception_hash(daddr, saddr); |
| 1584 |
*bucket += hval; |
| 1585 |
|
| 1586 |
hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) { |
| 1587 |
struct rt6_info *rt6 = rt6_ex->rt6i; |
| 1588 |
bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr); |
| 1589 |
|
| 1590 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1591 |
if (matched && saddr) |
| 1592 |
matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr); |
| 1593 |
#endif |
| 1594 |
if (matched) |
| 1595 |
return rt6_ex; |
| 1596 |
} |
| 1597 |
return NULL; |
| 1598 |
} |
| 1599 |
|
| 1600 |
/* Helper function to find the cached rt in the hash table |
| 1601 |
* and update bucket pointer to point to the bucket for this |
| 1602 |
* (daddr, saddr) pair |
| 1603 |
* Caller must hold rcu_read_lock() |
| 1604 |
*/ |
| 1605 |
static struct rt6_exception * |
| 1606 |
__rt6_find_exception_rcu(struct rt6_exception_bucket **bucket, |
| 1607 |
const struct in6_addr *daddr, |
| 1608 |
const struct in6_addr *saddr) |
| 1609 |
{ |
| 1610 |
struct rt6_exception *rt6_ex; |
| 1611 |
u32 hval; |
| 1612 |
|
| 1613 |
WARN_ON_ONCE(!rcu_read_lock_held()); |
| 1614 |
|
| 1615 |
if (!(*bucket) || !daddr) |
| 1616 |
return NULL; |
| 1617 |
|
| 1618 |
hval = rt6_exception_hash(daddr, saddr); |
| 1619 |
*bucket += hval; |
| 1620 |
|
| 1621 |
hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) { |
| 1622 |
struct rt6_info *rt6 = rt6_ex->rt6i; |
| 1623 |
bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr); |
| 1624 |
|
| 1625 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1626 |
if (matched && saddr) |
| 1627 |
matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr); |
| 1628 |
#endif |
| 1629 |
if (matched) |
| 1630 |
return rt6_ex; |
| 1631 |
} |
| 1632 |
return NULL; |
| 1633 |
} |
| 1634 |
|
| 1635 |
static unsigned int fib6_mtu(const struct fib6_result *res) |
| 1636 |
{ |
| 1637 |
const struct fib6_nh *nh = res->nh; |
| 1638 |
unsigned int mtu; |
| 1639 |
|
| 1640 |
if (res->f6i->fib6_pmtu) { |
| 1641 |
mtu = res->f6i->fib6_pmtu; |
| 1642 |
} else { |
| 1643 |
struct net_device *dev = nh->fib_nh_dev; |
| 1644 |
struct inet6_dev *idev; |
| 1645 |
|
| 1646 |
rcu_read_lock(); |
| 1647 |
idev = __in6_dev_get(dev); |
| 1648 |
mtu = READ_ONCE(idev->cnf.mtu6); |
| 1649 |
rcu_read_unlock(); |
| 1650 |
} |
| 1651 |
|
| 1652 |
mtu = min_t(unsigned int, mtu, IP6_MAX_MTU); |
| 1653 |
|
| 1654 |
return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu); |
| 1655 |
} |
| 1656 |
|
| 1657 |
#define FIB6_EXCEPTION_BUCKET_FLUSHED 0x1UL |
| 1658 |
|
| 1659 |
/* used when the flushed bit is not relevant, only access to the bucket |
| 1660 |
* (ie., all bucket users except rt6_insert_exception); |
| 1661 |
* |
| 1662 |
* called under rcu lock; sometimes called with rt6_exception_lock held |
| 1663 |
*/ |
| 1664 |
static |
| 1665 |
struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh, |
| 1666 |
spinlock_t *lock) |
| 1667 |
{ |
| 1668 |
struct rt6_exception_bucket *bucket; |
| 1669 |
|
| 1670 |
if (lock) |
| 1671 |
bucket = rcu_dereference_protected(nh->rt6i_exception_bucket, |
| 1672 |
lockdep_is_held(lock)); |
| 1673 |
else |
| 1674 |
bucket = rcu_dereference(nh->rt6i_exception_bucket); |
| 1675 |
|
| 1676 |
/* remove bucket flushed bit if set */ |
| 1677 |
if (bucket) { |
| 1678 |
unsigned long p = (unsigned long)bucket; |
| 1679 |
|
| 1680 |
p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED; |
| 1681 |
bucket = (struct rt6_exception_bucket *)p; |
| 1682 |
} |
| 1683 |
|
| 1684 |
return bucket; |
| 1685 |
} |
| 1686 |
|
| 1687 |
static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket) |
| 1688 |
{ |
| 1689 |
unsigned long p = (unsigned long)bucket; |
| 1690 |
|
| 1691 |
return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED); |
| 1692 |
} |
| 1693 |
|
| 1694 |
/* called with rt6_exception_lock held */ |
| 1695 |
static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh, |
| 1696 |
spinlock_t *lock) |
| 1697 |
{ |
| 1698 |
struct rt6_exception_bucket *bucket; |
| 1699 |
unsigned long p; |
| 1700 |
|
| 1701 |
bucket = rcu_dereference_protected(nh->rt6i_exception_bucket, |
| 1702 |
lockdep_is_held(lock)); |
| 1703 |
|
| 1704 |
p = (unsigned long)bucket; |
| 1705 |
p |= FIB6_EXCEPTION_BUCKET_FLUSHED; |
| 1706 |
bucket = (struct rt6_exception_bucket *)p; |
| 1707 |
rcu_assign_pointer(nh->rt6i_exception_bucket, bucket); |
| 1708 |
} |
| 1709 |
|
| 1710 |
static int rt6_insert_exception(struct rt6_info *nrt, |
| 1711 |
const struct fib6_result *res) |
| 1712 |
{ |
| 1713 |
struct net *net = dev_net(nrt->dst.dev); |
| 1714 |
struct rt6_exception_bucket *bucket; |
| 1715 |
struct fib6_info *f6i = res->f6i; |
| 1716 |
struct in6_addr *src_key = NULL; |
| 1717 |
struct rt6_exception *rt6_ex; |
| 1718 |
struct fib6_nh *nh = res->nh; |
| 1719 |
int max_depth; |
| 1720 |
int err = 0; |
| 1721 |
|
| 1722 |
spin_lock_bh(&rt6_exception_lock); |
| 1723 |
|
| 1724 |
bucket = rcu_dereference_protected(nh->rt6i_exception_bucket, |
| 1725 |
lockdep_is_held(&rt6_exception_lock)); |
| 1726 |
if (!bucket) { |
| 1727 |
bucket = kzalloc_objs(*bucket, FIB6_EXCEPTION_BUCKET_SIZE, |
| 1728 |
GFP_ATOMIC); |
| 1729 |
if (!bucket) { |
| 1730 |
err = -ENOMEM; |
| 1731 |
goto out; |
| 1732 |
} |
| 1733 |
rcu_assign_pointer(nh->rt6i_exception_bucket, bucket); |
| 1734 |
} else if (fib6_nh_excptn_bucket_flushed(bucket)) { |
| 1735 |
err = -EINVAL; |
| 1736 |
goto out; |
| 1737 |
} |
| 1738 |
|
| 1739 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1740 |
/* fib6_src.plen != 0 indicates f6i is in subtree |
| 1741 |
* and exception table is indexed by a hash of |
| 1742 |
* both fib6_dst and fib6_src. |
| 1743 |
* Otherwise, the exception table is indexed by |
| 1744 |
* a hash of only fib6_dst. |
| 1745 |
*/ |
| 1746 |
if (f6i->fib6_src.plen) |
| 1747 |
src_key = &nrt->rt6i_src.addr; |
| 1748 |
#endif |
| 1749 |
/* rt6_mtu_change() might lower mtu on f6i. |
| 1750 |
* Only insert this exception route if its mtu |
| 1751 |
* is less than f6i's mtu value. |
| 1752 |
*/ |
| 1753 |
if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) { |
| 1754 |
err = -EINVAL; |
| 1755 |
goto out; |
| 1756 |
} |
| 1757 |
|
| 1758 |
rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr, |
| 1759 |
src_key); |
| 1760 |
if (rt6_ex) |
| 1761 |
rt6_remove_exception(bucket, rt6_ex); |
| 1762 |
|
| 1763 |
rt6_ex = kzalloc_obj(*rt6_ex, GFP_ATOMIC); |
| 1764 |
if (!rt6_ex) { |
| 1765 |
err = -ENOMEM; |
| 1766 |
goto out; |
| 1767 |
} |
| 1768 |
rt6_ex->rt6i = nrt; |
| 1769 |
rt6_ex->stamp = jiffies; |
| 1770 |
hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain); |
| 1771 |
bucket->depth++; |
| 1772 |
net->ipv6.rt6_stats->fib_rt_cache++; |
| 1773 |
|
| 1774 |
/* Randomize max depth to avoid some side channels attacks. */ |
| 1775 |
max_depth = FIB6_MAX_DEPTH + get_random_u32_below(FIB6_MAX_DEPTH); |
| 1776 |
while (bucket->depth > max_depth) |
| 1777 |
rt6_exception_remove_oldest(bucket); |
| 1778 |
|
| 1779 |
out: |
| 1780 |
spin_unlock_bh(&rt6_exception_lock); |
| 1781 |
|
| 1782 |
/* Update fn->fn_sernum to invalidate all cached dst */ |
| 1783 |
if (!err) { |
| 1784 |
spin_lock_bh(&f6i->fib6_table->tb6_lock); |
| 1785 |
fib6_update_sernum(net, f6i); |
| 1786 |
fib6_add_gc_list(f6i); |
| 1787 |
spin_unlock_bh(&f6i->fib6_table->tb6_lock); |
| 1788 |
fib6_force_start_gc(net); |
| 1789 |
} |
| 1790 |
|
| 1791 |
return err; |
| 1792 |
} |
| 1793 |
|
| 1794 |
static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from) |
| 1795 |
{ |
| 1796 |
struct rt6_exception_bucket *bucket; |
| 1797 |
struct rt6_exception *rt6_ex; |
| 1798 |
struct hlist_node *tmp; |
| 1799 |
int i; |
| 1800 |
|
| 1801 |
spin_lock_bh(&rt6_exception_lock); |
| 1802 |
|
| 1803 |
bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock); |
| 1804 |
if (!bucket) |
| 1805 |
goto out; |
| 1806 |
|
| 1807 |
/* Prevent rt6_insert_exception() to recreate the bucket list */ |
| 1808 |
if (!from) |
| 1809 |
fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock); |
| 1810 |
|
| 1811 |
for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) { |
| 1812 |
hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) { |
| 1813 |
if (!from || |
| 1814 |
rcu_access_pointer(rt6_ex->rt6i->from) == from) |
| 1815 |
rt6_remove_exception(bucket, rt6_ex); |
| 1816 |
} |
| 1817 |
WARN_ON_ONCE(!from && bucket->depth); |
| 1818 |
bucket++; |
| 1819 |
} |
| 1820 |
out: |
| 1821 |
spin_unlock_bh(&rt6_exception_lock); |
| 1822 |
} |
| 1823 |
|
| 1824 |
static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg) |
| 1825 |
{ |
| 1826 |
struct fib6_info *f6i = arg; |
| 1827 |
|
| 1828 |
fib6_nh_flush_exceptions(nh, f6i); |
| 1829 |
|
| 1830 |
return 0; |
| 1831 |
} |
| 1832 |
|
| 1833 |
void rt6_flush_exceptions(struct fib6_info *f6i) |
| 1834 |
{ |
| 1835 |
if (f6i->nh) { |
| 1836 |
rcu_read_lock(); |
| 1837 |
nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions, f6i); |
| 1838 |
rcu_read_unlock(); |
| 1839 |
} else { |
| 1840 |
fib6_nh_flush_exceptions(f6i->fib6_nh, f6i); |
| 1841 |
} |
| 1842 |
} |
| 1843 |
|
| 1844 |
/* Find cached rt in the hash table inside passed in rt |
| 1845 |
* Caller has to hold rcu_read_lock() |
| 1846 |
*/ |
| 1847 |
static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res, |
| 1848 |
const struct in6_addr *daddr, |
| 1849 |
const struct in6_addr *saddr) |
| 1850 |
{ |
| 1851 |
const struct in6_addr *src_key = NULL; |
| 1852 |
struct rt6_exception_bucket *bucket; |
| 1853 |
struct rt6_exception *rt6_ex; |
| 1854 |
struct rt6_info *ret = NULL; |
| 1855 |
|
| 1856 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1857 |
/* fib6i_src.plen != 0 indicates f6i is in subtree |
| 1858 |
* and exception table is indexed by a hash of |
| 1859 |
* both fib6_dst and fib6_src. |
| 1860 |
* However, the src addr used to create the hash |
| 1861 |
* might not be exactly the passed in saddr which |
| 1862 |
* is a /128 addr from the flow. |
| 1863 |
* So we need to use f6i->fib6_src to redo lookup |
| 1864 |
* if the passed in saddr does not find anything. |
| 1865 |
* (See the logic in ip6_rt_cache_alloc() on how |
| 1866 |
* rt->rt6i_src is updated.) |
| 1867 |
*/ |
| 1868 |
if (res->f6i->fib6_src.plen) |
| 1869 |
src_key = saddr; |
| 1870 |
find_ex: |
| 1871 |
#endif |
| 1872 |
bucket = fib6_nh_get_excptn_bucket(res->nh, NULL); |
| 1873 |
rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key); |
| 1874 |
|
| 1875 |
if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i)) |
| 1876 |
ret = rt6_ex->rt6i; |
| 1877 |
|
| 1878 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1879 |
/* Use fib6_src as src_key and redo lookup */ |
| 1880 |
if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) { |
| 1881 |
src_key = &res->f6i->fib6_src.addr; |
| 1882 |
goto find_ex; |
| 1883 |
} |
| 1884 |
#endif |
| 1885 |
|
| 1886 |
return ret; |
| 1887 |
} |
| 1888 |
|
| 1889 |
/* Remove the passed in cached rt from the hash table that contains it */ |
| 1890 |
static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen, |
| 1891 |
const struct rt6_info *rt) |
| 1892 |
{ |
| 1893 |
const struct in6_addr *src_key = NULL; |
| 1894 |
struct rt6_exception_bucket *bucket; |
| 1895 |
struct rt6_exception *rt6_ex; |
| 1896 |
int err; |
| 1897 |
|
| 1898 |
if (!rcu_access_pointer(nh->rt6i_exception_bucket)) |
| 1899 |
return -ENOENT; |
| 1900 |
|
| 1901 |
spin_lock_bh(&rt6_exception_lock); |
| 1902 |
bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock); |
| 1903 |
|
| 1904 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1905 |
/* rt6i_src.plen != 0 indicates 'from' is in subtree |
| 1906 |
* and exception table is indexed by a hash of |
| 1907 |
* both rt6i_dst and rt6i_src. |
| 1908 |
* Otherwise, the exception table is indexed by |
| 1909 |
* a hash of only rt6i_dst. |
| 1910 |
*/ |
| 1911 |
if (plen) |
| 1912 |
src_key = &rt->rt6i_src.addr; |
| 1913 |
#endif |
| 1914 |
rt6_ex = __rt6_find_exception_spinlock(&bucket, |
| 1915 |
&rt->rt6i_dst.addr, |
| 1916 |
src_key); |
| 1917 |
if (rt6_ex) { |
| 1918 |
rt6_remove_exception(bucket, rt6_ex); |
| 1919 |
err = 0; |
| 1920 |
} else { |
| 1921 |
err = -ENOENT; |
| 1922 |
} |
| 1923 |
|
| 1924 |
spin_unlock_bh(&rt6_exception_lock); |
| 1925 |
return err; |
| 1926 |
} |
| 1927 |
|
| 1928 |
struct fib6_nh_excptn_arg { |
| 1929 |
struct rt6_info *rt; |
| 1930 |
int plen; |
| 1931 |
}; |
| 1932 |
|
| 1933 |
static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg) |
| 1934 |
{ |
| 1935 |
struct fib6_nh_excptn_arg *arg = _arg; |
| 1936 |
int err; |
| 1937 |
|
| 1938 |
err = fib6_nh_remove_exception(nh, arg->plen, arg->rt); |
| 1939 |
if (err == 0) |
| 1940 |
return 1; |
| 1941 |
|
| 1942 |
return 0; |
| 1943 |
} |
| 1944 |
|
| 1945 |
static int rt6_remove_exception_rt(struct rt6_info *rt) |
| 1946 |
{ |
| 1947 |
struct fib6_info *from; |
| 1948 |
|
| 1949 |
from = rcu_dereference(rt->from); |
| 1950 |
if (!from || !(rt->rt6i_flags & RTF_CACHE)) |
| 1951 |
return -EINVAL; |
| 1952 |
|
| 1953 |
if (from->nh) { |
| 1954 |
struct fib6_nh_excptn_arg arg = { |
| 1955 |
.rt = rt, |
| 1956 |
.plen = from->fib6_src.plen |
| 1957 |
}; |
| 1958 |
int rc; |
| 1959 |
|
| 1960 |
/* rc = 1 means an entry was found */ |
| 1961 |
rc = nexthop_for_each_fib6_nh(from->nh, |
| 1962 |
rt6_nh_remove_exception_rt, |
| 1963 |
&arg); |
| 1964 |
return rc ? 0 : -ENOENT; |
| 1965 |
} |
| 1966 |
|
| 1967 |
return fib6_nh_remove_exception(from->fib6_nh, |
| 1968 |
from->fib6_src.plen, rt); |
| 1969 |
} |
| 1970 |
|
| 1971 |
/* Find rt6_ex which contains the passed in rt cache and |
| 1972 |
* refresh its stamp |
| 1973 |
*/ |
| 1974 |
static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen, |
| 1975 |
const struct rt6_info *rt) |
| 1976 |
{ |
| 1977 |
const struct in6_addr *src_key = NULL; |
| 1978 |
struct rt6_exception_bucket *bucket; |
| 1979 |
struct rt6_exception *rt6_ex; |
| 1980 |
|
| 1981 |
bucket = fib6_nh_get_excptn_bucket(nh, NULL); |
| 1982 |
#ifdef CONFIG_IPV6_SUBTREES |
| 1983 |
/* rt6i_src.plen != 0 indicates 'from' is in subtree |
| 1984 |
* and exception table is indexed by a hash of |
| 1985 |
* both rt6i_dst and rt6i_src. |
| 1986 |
* Otherwise, the exception table is indexed by |
| 1987 |
* a hash of only rt6i_dst. |
| 1988 |
*/ |
| 1989 |
if (plen) |
| 1990 |
src_key = &rt->rt6i_src.addr; |
| 1991 |
#endif |
| 1992 |
rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key); |
| 1993 |
if (rt6_ex) |
| 1994 |
rt6_ex->stamp = jiffies; |
| 1995 |
} |
| 1996 |
|
| 1997 |
struct fib6_nh_match_arg { |
| 1998 |
const struct net_device *dev; |
| 1999 |
const struct in6_addr *gw; |
| 2000 |
struct fib6_nh *match; |
| 2001 |
}; |
| 2002 |
|
| 2003 |
/* determine if fib6_nh has given device and gateway */ |
| 2004 |
static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg) |
| 2005 |
{ |
| 2006 |
struct fib6_nh_match_arg *arg = _arg; |
| 2007 |
|
| 2008 |
if (arg->dev != nh->fib_nh_dev || |
| 2009 |
(arg->gw && !nh->fib_nh_gw_family) || |
| 2010 |
(!arg->gw && nh->fib_nh_gw_family) || |
| 2011 |
(arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6))) |
| 2012 |
return 0; |
| 2013 |
|
| 2014 |
arg->match = nh; |
| 2015 |
|
| 2016 |
/* found a match, break the loop */ |
| 2017 |
return 1; |
| 2018 |
} |
| 2019 |
|
| 2020 |
static void rt6_update_exception_stamp_rt(struct rt6_info *rt) |
| 2021 |
{ |
| 2022 |
struct fib6_info *from; |
| 2023 |
struct fib6_nh *fib6_nh; |
| 2024 |
|
| 2025 |
rcu_read_lock(); |
| 2026 |
|
| 2027 |
from = rcu_dereference(rt->from); |
| 2028 |
if (!from || !(rt->rt6i_flags & RTF_CACHE)) |
| 2029 |
goto unlock; |
| 2030 |
|
| 2031 |
if (from->nh) { |
| 2032 |
struct fib6_nh_match_arg arg = { |
| 2033 |
.dev = rt->dst.dev, |
| 2034 |
.gw = &rt->rt6i_gateway, |
| 2035 |
}; |
| 2036 |
|
| 2037 |
nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg); |
| 2038 |
|
| 2039 |
if (!arg.match) |
| 2040 |
goto unlock; |
| 2041 |
fib6_nh = arg.match; |
| 2042 |
} else { |
| 2043 |
fib6_nh = from->fib6_nh; |
| 2044 |
} |
| 2045 |
fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt); |
| 2046 |
unlock: |
| 2047 |
rcu_read_unlock(); |
| 2048 |
} |
| 2049 |
|
| 2050 |
static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev, |
| 2051 |
struct rt6_info *rt, int mtu) |
| 2052 |
{ |
| 2053 |
u32 dmtu = dst6_mtu(&rt->dst); |
| 2054 |
|
| 2055 |
/* If the new MTU is lower than the route PMTU, this new MTU will be the |
| 2056 |
* lowest MTU in the path: always allow updating the route PMTU to |
| 2057 |
* reflect PMTU decreases. |
| 2058 |
* |
| 2059 |
* If the new MTU is higher, and the route PMTU is equal to the local |
| 2060 |
* MTU, this means the old MTU is the lowest in the path, so allow |
| 2061 |
* updating it: if other nodes now have lower MTUs, PMTU discovery will |
| 2062 |
* handle this. |
| 2063 |
*/ |
| 2064 |
|
| 2065 |
if (dmtu >= mtu) |
| 2066 |
return true; |
| 2067 |
|
| 2068 |
if (dmtu == idev->cnf.mtu6) |
| 2069 |
return true; |
| 2070 |
|
| 2071 |
return false; |
| 2072 |
} |
| 2073 |
|
| 2074 |
static void rt6_exceptions_update_pmtu(struct inet6_dev *idev, |
| 2075 |
const struct fib6_nh *nh, int mtu) |
| 2076 |
{ |
| 2077 |
struct rt6_exception_bucket *bucket; |
| 2078 |
struct rt6_exception *rt6_ex; |
| 2079 |
int i; |
| 2080 |
|
| 2081 |
bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock); |
| 2082 |
if (!bucket) |
| 2083 |
return; |
| 2084 |
|
| 2085 |
for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) { |
| 2086 |
hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) { |
| 2087 |
struct rt6_info *entry = rt6_ex->rt6i; |
| 2088 |
|
| 2089 |
/* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected |
| 2090 |
* route), the metrics of its rt->from have already |
| 2091 |
* been updated. |
| 2092 |
*/ |
| 2093 |
if (dst_metric_raw(&entry->dst, RTAX_MTU) && |
| 2094 |
rt6_mtu_change_route_allowed(idev, entry, mtu)) |
| 2095 |
dst_metric_set(&entry->dst, RTAX_MTU, mtu); |
| 2096 |
} |
| 2097 |
bucket++; |
| 2098 |
} |
| 2099 |
} |
| 2100 |
|
| 2101 |
#define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE) |
| 2102 |
|
| 2103 |
static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh, |
| 2104 |
const struct in6_addr *gateway) |
| 2105 |
{ |
| 2106 |
struct rt6_exception_bucket *bucket; |
| 2107 |
struct rt6_exception *rt6_ex; |
| 2108 |
struct hlist_node *tmp; |
| 2109 |
int i; |
| 2110 |
|
| 2111 |
if (!rcu_access_pointer(nh->rt6i_exception_bucket)) |
| 2112 |
return; |
| 2113 |
|
| 2114 |
spin_lock_bh(&rt6_exception_lock); |
| 2115 |
bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock); |
| 2116 |
if (bucket) { |
| 2117 |
for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) { |
| 2118 |
hlist_for_each_entry_safe(rt6_ex, tmp, |
| 2119 |
&bucket->chain, hlist) { |
| 2120 |
struct rt6_info *entry = rt6_ex->rt6i; |
| 2121 |
|
| 2122 |
if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) == |
| 2123 |
RTF_CACHE_GATEWAY && |
| 2124 |
ipv6_addr_equal(gateway, |
| 2125 |
&entry->rt6i_gateway)) { |
| 2126 |
rt6_remove_exception(bucket, rt6_ex); |
| 2127 |
} |
| 2128 |
} |
| 2129 |
bucket++; |
| 2130 |
} |
| 2131 |
} |
| 2132 |
|
| 2133 |
spin_unlock_bh(&rt6_exception_lock); |
| 2134 |
} |
| 2135 |
|
| 2136 |
static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket, |
| 2137 |
struct rt6_exception *rt6_ex, |
| 2138 |
struct fib6_gc_args *gc_args, |
| 2139 |
unsigned long now) |
| 2140 |
{ |
| 2141 |
struct rt6_info *rt = rt6_ex->rt6i; |
| 2142 |
|
| 2143 |
/* we are pruning and obsoleting aged-out and non gateway exceptions |
| 2144 |
* even if others have still references to them, so that on next |
| 2145 |
* dst_check() such references can be dropped. |
| 2146 |
* EXPIRES exceptions - e.g. pmtu-generated ones are pruned when |
| 2147 |
* expired, independently from their aging, as per RFC 8201 section 4 |
| 2148 |
*/ |
| 2149 |
if (!(rt->rt6i_flags & RTF_EXPIRES)) { |
| 2150 |
if (time_after_eq(now, READ_ONCE(rt->dst.lastuse) + |
| 2151 |
gc_args->timeout)) { |
| 2152 |
pr_debug("aging clone %p\n", rt); |
| 2153 |
rt6_remove_exception(bucket, rt6_ex); |
| 2154 |
return; |
| 2155 |
} |
| 2156 |
} else if (time_after(jiffies, READ_ONCE(rt->dst.expires))) { |
| 2157 |
pr_debug("purging expired route %p\n", rt); |
| 2158 |
rt6_remove_exception(bucket, rt6_ex); |
| 2159 |
return; |
| 2160 |
} |
| 2161 |
|
| 2162 |
if (rt->rt6i_flags & RTF_GATEWAY) { |
| 2163 |
struct neighbour *neigh; |
| 2164 |
|
| 2165 |
neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway); |
| 2166 |
|
| 2167 |
if (!(neigh && (neigh->flags & NTF_ROUTER))) { |
| 2168 |
pr_debug("purging route %p via non-router but gateway\n", |
| 2169 |
rt); |
| 2170 |
rt6_remove_exception(bucket, rt6_ex); |
| 2171 |
return; |
| 2172 |
} |
| 2173 |
} |
| 2174 |
|
| 2175 |
gc_args->more++; |
| 2176 |
} |
| 2177 |
|
| 2178 |
static void fib6_nh_age_exceptions(const struct fib6_nh *nh, |
| 2179 |
struct fib6_gc_args *gc_args, |
| 2180 |
unsigned long now) |
| 2181 |
{ |
| 2182 |
struct rt6_exception_bucket *bucket; |
| 2183 |
struct rt6_exception *rt6_ex; |
| 2184 |
struct hlist_node *tmp; |
| 2185 |
int i; |
| 2186 |
|
| 2187 |
if (!rcu_access_pointer(nh->rt6i_exception_bucket)) |
| 2188 |
return; |
| 2189 |
|
| 2190 |
rcu_read_lock_bh(); |
| 2191 |
spin_lock(&rt6_exception_lock); |
| 2192 |
bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock); |
| 2193 |
if (bucket) { |
| 2194 |
for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) { |
| 2195 |
hlist_for_each_entry_safe(rt6_ex, tmp, |
| 2196 |
&bucket->chain, hlist) { |
| 2197 |
rt6_age_examine_exception(bucket, rt6_ex, |
| 2198 |
gc_args, now); |
| 2199 |
} |
| 2200 |
bucket++; |
| 2201 |
} |
| 2202 |
} |
| 2203 |
spin_unlock(&rt6_exception_lock); |
| 2204 |
rcu_read_unlock_bh(); |
| 2205 |
} |
| 2206 |
|
| 2207 |
struct fib6_nh_age_excptn_arg { |
| 2208 |
struct fib6_gc_args *gc_args; |
| 2209 |
unsigned long now; |
| 2210 |
}; |
| 2211 |
|
| 2212 |
static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg) |
| 2213 |
{ |
| 2214 |
struct fib6_nh_age_excptn_arg *arg = _arg; |
| 2215 |
|
| 2216 |
fib6_nh_age_exceptions(nh, arg->gc_args, arg->now); |
| 2217 |
return 0; |
| 2218 |
} |
| 2219 |
|
| 2220 |
void rt6_age_exceptions(struct fib6_info *f6i, |
| 2221 |
struct fib6_gc_args *gc_args, |
| 2222 |
unsigned long now) |
| 2223 |
{ |
| 2224 |
if (f6i->nh) { |
| 2225 |
struct fib6_nh_age_excptn_arg arg = { |
| 2226 |
.gc_args = gc_args, |
| 2227 |
.now = now |
| 2228 |
}; |
| 2229 |
|
| 2230 |
nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions, |
| 2231 |
&arg); |
| 2232 |
} else { |
| 2233 |
fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now); |
| 2234 |
} |
| 2235 |
} |
| 2236 |
|
| 2237 |
/* must be called with rcu lock held */ |
| 2238 |
int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif, |
| 2239 |
struct flowi6 *fl6, struct fib6_result *res, int strict) |
| 2240 |
{ |
| 2241 |
struct fib6_node *fn, *saved_fn; |
| 2242 |
|
| 2243 |
fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr); |
| 2244 |
saved_fn = fn; |
| 2245 |
|
| 2246 |
redo_rt6_select: |
| 2247 |
rt6_select(net, fn, oif, res, strict); |
| 2248 |
if (res->f6i == net->ipv6.fib6_null_entry) { |
| 2249 |
fn = fib6_backtrack(fn, &fl6->saddr); |
| 2250 |
if (fn) |
| 2251 |
goto redo_rt6_select; |
| 2252 |
else if (strict & RT6_LOOKUP_F_REACHABLE) { |
| 2253 |
/* also consider unreachable route */ |
| 2254 |
strict &= ~RT6_LOOKUP_F_REACHABLE; |
| 2255 |
fn = saved_fn; |
| 2256 |
goto redo_rt6_select; |
| 2257 |
} |
| 2258 |
} |
| 2259 |
|
| 2260 |
trace_fib6_table_lookup(net, res, table, fl6); |
| 2261 |
|
| 2262 |
return 0; |
| 2263 |
} |
| 2264 |
|
| 2265 |
struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, |
| 2266 |
int oif, struct flowi6 *fl6, |
| 2267 |
const struct sk_buff *skb, int flags) |
| 2268 |
{ |
| 2269 |
struct fib6_result res = {}; |
| 2270 |
struct rt6_info *rt = NULL; |
| 2271 |
int strict = 0; |
| 2272 |
|
| 2273 |
WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) && |
| 2274 |
!rcu_read_lock_held()); |
| 2275 |
|
| 2276 |
strict |= flags & RT6_LOOKUP_F_IFACE; |
| 2277 |
strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE; |
| 2278 |
if (READ_ONCE(net->ipv6.devconf_all->forwarding) == 0) |
| 2279 |
strict |= RT6_LOOKUP_F_REACHABLE; |
| 2280 |
|
| 2281 |
rcu_read_lock(); |
| 2282 |
|
| 2283 |
fib6_table_lookup(net, table, oif, fl6, &res, strict); |
| 2284 |
if (res.f6i == net->ipv6.fib6_null_entry) |
| 2285 |
goto out; |
| 2286 |
|
| 2287 |
fib6_select_path(net, &res, fl6, oif, false, skb, strict); |
| 2288 |
|
| 2289 |
/*Search through exception table */ |
| 2290 |
rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr); |
| 2291 |
if (rt) { |
| 2292 |
goto out; |
| 2293 |
} else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) && |
| 2294 |
!res.nh->fib_nh_gw_family)) { |
| 2295 |
/* Create a RTF_CACHE clone which will not be |
| 2296 |
* owned by the fib6 tree. It is for the special case where |
| 2297 |
* the daddr in the skb during the neighbor look-up is different |
| 2298 |
* from the fl6->daddr used to look-up route here. |
| 2299 |
*/ |
| 2300 |
rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL); |
| 2301 |
|
| 2302 |
if (rt) { |
| 2303 |
/* 1 refcnt is taken during ip6_rt_cache_alloc(). |
| 2304 |
* As rt6_uncached_list_add() does not consume refcnt, |
| 2305 |
* this refcnt is always returned to the caller even |
| 2306 |
* if caller sets RT6_LOOKUP_F_DST_NOREF flag. |
| 2307 |
*/ |
| 2308 |
rt6_uncached_list_add(rt); |
| 2309 |
rcu_read_unlock(); |
| 2310 |
|
| 2311 |
return rt; |
| 2312 |
} |
| 2313 |
} else { |
| 2314 |
/* Get a percpu copy */ |
| 2315 |
local_bh_disable(); |
| 2316 |
rt = rt6_get_pcpu_route(&res); |
| 2317 |
|
| 2318 |
if (!rt) |
| 2319 |
rt = rt6_make_pcpu_route(net, &res); |
| 2320 |
|
| 2321 |
local_bh_enable(); |
| 2322 |
} |
| 2323 |
out: |
| 2324 |
if (!rt) |
| 2325 |
rt = net->ipv6.ip6_null_entry; |
| 2326 |
if (!(flags & RT6_LOOKUP_F_DST_NOREF)) |
| 2327 |
ip6_hold_safe(net, &rt); |
| 2328 |
rcu_read_unlock(); |
| 2329 |
|
| 2330 |
return rt; |
| 2331 |
} |
| 2332 |
EXPORT_SYMBOL_GPL(ip6_pol_route); |
| 2333 |
|
| 2334 |
INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_input(struct net *net, |
| 2335 |
struct fib6_table *table, |
| 2336 |
struct flowi6 *fl6, |
| 2337 |
const struct sk_buff *skb, |
| 2338 |
int flags) |
| 2339 |
{ |
| 2340 |
return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags); |
| 2341 |
} |
| 2342 |
|
| 2343 |
struct dst_entry *ip6_route_input_lookup(struct net *net, |
| 2344 |
struct net_device *dev, |
| 2345 |
struct flowi6 *fl6, |
| 2346 |
const struct sk_buff *skb, |
| 2347 |
int flags) |
| 2348 |
{ |
| 2349 |
if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG) |
| 2350 |
flags |= RT6_LOOKUP_F_IFACE; |
| 2351 |
|
| 2352 |
return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input); |
| 2353 |
} |
| 2354 |
EXPORT_SYMBOL_GPL(ip6_route_input_lookup); |
| 2355 |
|
| 2356 |
static void ip6_multipath_l3_keys(const struct sk_buff *skb, |
| 2357 |
struct flow_keys *keys, |
| 2358 |
struct flow_keys *flkeys) |
| 2359 |
{ |
| 2360 |
const struct ipv6hdr *outer_iph = ipv6_hdr(skb); |
| 2361 |
const struct ipv6hdr *key_iph = outer_iph; |
| 2362 |
struct flow_keys *_flkeys = flkeys; |
| 2363 |
const struct ipv6hdr *inner_iph; |
| 2364 |
const struct icmp6hdr *icmph; |
| 2365 |
struct ipv6hdr _inner_iph; |
| 2366 |
struct icmp6hdr _icmph; |
| 2367 |
|
| 2368 |
if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6)) |
| 2369 |
goto out; |
| 2370 |
|
| 2371 |
icmph = skb_header_pointer(skb, skb_transport_offset(skb), |
| 2372 |
sizeof(_icmph), &_icmph); |
| 2373 |
if (!icmph) |
| 2374 |
goto out; |
| 2375 |
|
| 2376 |
if (!icmpv6_is_err(icmph->icmp6_type)) |
| 2377 |
goto out; |
| 2378 |
|
| 2379 |
inner_iph = skb_header_pointer(skb, |
| 2380 |
skb_transport_offset(skb) + sizeof(*icmph), |
| 2381 |
sizeof(_inner_iph), &_inner_iph); |
| 2382 |
if (!inner_iph) |
| 2383 |
goto out; |
| 2384 |
|
| 2385 |
key_iph = inner_iph; |
| 2386 |
_flkeys = NULL; |
| 2387 |
out: |
| 2388 |
if (_flkeys) { |
| 2389 |
keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src; |
| 2390 |
keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst; |
| 2391 |
keys->tags.flow_label = _flkeys->tags.flow_label; |
| 2392 |
keys->basic.ip_proto = _flkeys->basic.ip_proto; |
| 2393 |
} else { |
| 2394 |
keys->addrs.v6addrs.src = key_iph->saddr; |
| 2395 |
keys->addrs.v6addrs.dst = key_iph->daddr; |
| 2396 |
keys->tags.flow_label = ip6_flowlabel(key_iph); |
| 2397 |
keys->basic.ip_proto = key_iph->nexthdr; |
| 2398 |
} |
| 2399 |
} |
| 2400 |
|
| 2401 |
static u32 rt6_multipath_custom_hash_outer(const struct net *net, |
| 2402 |
const struct sk_buff *skb, |
| 2403 |
bool *p_has_inner) |
| 2404 |
{ |
| 2405 |
u32 hash_fields = ip6_multipath_hash_fields(net); |
| 2406 |
struct flow_keys keys, hash_keys; |
| 2407 |
|
| 2408 |
if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK)) |
| 2409 |
return 0; |
| 2410 |
|
| 2411 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2412 |
skb_flow_dissect_flow_keys(skb, &keys, FLOW_DISSECTOR_F_STOP_AT_ENCAP); |
| 2413 |
|
| 2414 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2415 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP) |
| 2416 |
hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src; |
| 2417 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP) |
| 2418 |
hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst; |
| 2419 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO) |
| 2420 |
hash_keys.basic.ip_proto = keys.basic.ip_proto; |
| 2421 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL) |
| 2422 |
hash_keys.tags.flow_label = keys.tags.flow_label; |
| 2423 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT) |
| 2424 |
hash_keys.ports.src = keys.ports.src; |
| 2425 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT) |
| 2426 |
hash_keys.ports.dst = keys.ports.dst; |
| 2427 |
|
| 2428 |
*p_has_inner = !!(keys.control.flags & FLOW_DIS_ENCAPSULATION); |
| 2429 |
return fib_multipath_hash_from_keys(net, &hash_keys); |
| 2430 |
} |
| 2431 |
|
| 2432 |
static u32 rt6_multipath_custom_hash_inner(const struct net *net, |
| 2433 |
const struct sk_buff *skb, |
| 2434 |
bool has_inner) |
| 2435 |
{ |
| 2436 |
u32 hash_fields = ip6_multipath_hash_fields(net); |
| 2437 |
struct flow_keys keys, hash_keys; |
| 2438 |
|
| 2439 |
/* We assume the packet carries an encapsulation, but if none was |
| 2440 |
* encountered during dissection of the outer flow, then there is no |
| 2441 |
* point in calling the flow dissector again. |
| 2442 |
*/ |
| 2443 |
if (!has_inner) |
| 2444 |
return 0; |
| 2445 |
|
| 2446 |
if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_MASK)) |
| 2447 |
return 0; |
| 2448 |
|
| 2449 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2450 |
skb_flow_dissect_flow_keys(skb, &keys, 0); |
| 2451 |
|
| 2452 |
if (!(keys.control.flags & FLOW_DIS_ENCAPSULATION)) |
| 2453 |
return 0; |
| 2454 |
|
| 2455 |
if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) { |
| 2456 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; |
| 2457 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP) |
| 2458 |
hash_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src; |
| 2459 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP) |
| 2460 |
hash_keys.addrs.v4addrs.dst = keys.addrs.v4addrs.dst; |
| 2461 |
} else if (keys.control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) { |
| 2462 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2463 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_IP) |
| 2464 |
hash_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src; |
| 2465 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_IP) |
| 2466 |
hash_keys.addrs.v6addrs.dst = keys.addrs.v6addrs.dst; |
| 2467 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_FLOWLABEL) |
| 2468 |
hash_keys.tags.flow_label = keys.tags.flow_label; |
| 2469 |
} |
| 2470 |
|
| 2471 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_IP_PROTO) |
| 2472 |
hash_keys.basic.ip_proto = keys.basic.ip_proto; |
| 2473 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_SRC_PORT) |
| 2474 |
hash_keys.ports.src = keys.ports.src; |
| 2475 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_INNER_DST_PORT) |
| 2476 |
hash_keys.ports.dst = keys.ports.dst; |
| 2477 |
|
| 2478 |
return fib_multipath_hash_from_keys(net, &hash_keys); |
| 2479 |
} |
| 2480 |
|
| 2481 |
static u32 rt6_multipath_custom_hash_skb(const struct net *net, |
| 2482 |
const struct sk_buff *skb) |
| 2483 |
{ |
| 2484 |
u32 mhash, mhash_inner; |
| 2485 |
bool has_inner = true; |
| 2486 |
|
| 2487 |
mhash = rt6_multipath_custom_hash_outer(net, skb, &has_inner); |
| 2488 |
mhash_inner = rt6_multipath_custom_hash_inner(net, skb, has_inner); |
| 2489 |
|
| 2490 |
return jhash_2words(mhash, mhash_inner, 0); |
| 2491 |
} |
| 2492 |
|
| 2493 |
static u32 rt6_multipath_custom_hash_fl6(const struct net *net, |
| 2494 |
const struct flowi6 *fl6) |
| 2495 |
{ |
| 2496 |
u32 hash_fields = ip6_multipath_hash_fields(net); |
| 2497 |
struct flow_keys hash_keys; |
| 2498 |
|
| 2499 |
if (!(hash_fields & FIB_MULTIPATH_HASH_FIELD_OUTER_MASK)) |
| 2500 |
return 0; |
| 2501 |
|
| 2502 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2503 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2504 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_IP) |
| 2505 |
hash_keys.addrs.v6addrs.src = fl6->saddr; |
| 2506 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_IP) |
| 2507 |
hash_keys.addrs.v6addrs.dst = fl6->daddr; |
| 2508 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_IP_PROTO) |
| 2509 |
hash_keys.basic.ip_proto = fl6->flowi6_proto; |
| 2510 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_FLOWLABEL) |
| 2511 |
hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); |
| 2512 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_SRC_PORT) { |
| 2513 |
if (fl6->flowi6_flags & FLOWI_FLAG_ANY_SPORT) |
| 2514 |
hash_keys.ports.src = (__force __be16)get_random_u16(); |
| 2515 |
else |
| 2516 |
hash_keys.ports.src = fl6->fl6_sport; |
| 2517 |
} |
| 2518 |
if (hash_fields & FIB_MULTIPATH_HASH_FIELD_DST_PORT) |
| 2519 |
hash_keys.ports.dst = fl6->fl6_dport; |
| 2520 |
|
| 2521 |
return fib_multipath_hash_from_keys(net, &hash_keys); |
| 2522 |
} |
| 2523 |
|
| 2524 |
/* if skb is set it will be used and fl6 can be NULL */ |
| 2525 |
u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6, |
| 2526 |
const struct sk_buff *skb, struct flow_keys *flkeys) |
| 2527 |
{ |
| 2528 |
struct flow_keys hash_keys; |
| 2529 |
u32 mhash = 0; |
| 2530 |
|
| 2531 |
switch (ip6_multipath_hash_policy(net)) { |
| 2532 |
case 0: |
| 2533 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2534 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2535 |
if (skb) { |
| 2536 |
ip6_multipath_l3_keys(skb, &hash_keys, flkeys); |
| 2537 |
} else { |
| 2538 |
hash_keys.addrs.v6addrs.src = fl6->saddr; |
| 2539 |
hash_keys.addrs.v6addrs.dst = fl6->daddr; |
| 2540 |
hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); |
| 2541 |
hash_keys.basic.ip_proto = fl6->flowi6_proto; |
| 2542 |
} |
| 2543 |
mhash = fib_multipath_hash_from_keys(net, &hash_keys); |
| 2544 |
break; |
| 2545 |
case 1: |
| 2546 |
if (skb) { |
| 2547 |
unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP; |
| 2548 |
struct flow_keys keys; |
| 2549 |
|
| 2550 |
/* short-circuit if we already have L4 hash present */ |
| 2551 |
if (skb->l4_hash) |
| 2552 |
return skb_get_hash_raw(skb) >> 1; |
| 2553 |
|
| 2554 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2555 |
|
| 2556 |
if (!flkeys) { |
| 2557 |
skb_flow_dissect_flow_keys(skb, &keys, flag); |
| 2558 |
flkeys = &keys; |
| 2559 |
} |
| 2560 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2561 |
hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src; |
| 2562 |
hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst; |
| 2563 |
hash_keys.ports.src = flkeys->ports.src; |
| 2564 |
hash_keys.ports.dst = flkeys->ports.dst; |
| 2565 |
hash_keys.basic.ip_proto = flkeys->basic.ip_proto; |
| 2566 |
} else { |
| 2567 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2568 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2569 |
hash_keys.addrs.v6addrs.src = fl6->saddr; |
| 2570 |
hash_keys.addrs.v6addrs.dst = fl6->daddr; |
| 2571 |
if (fl6->flowi6_flags & FLOWI_FLAG_ANY_SPORT) |
| 2572 |
hash_keys.ports.src = (__force __be16)get_random_u16(); |
| 2573 |
else |
| 2574 |
hash_keys.ports.src = fl6->fl6_sport; |
| 2575 |
hash_keys.ports.dst = fl6->fl6_dport; |
| 2576 |
hash_keys.basic.ip_proto = fl6->flowi6_proto; |
| 2577 |
} |
| 2578 |
mhash = fib_multipath_hash_from_keys(net, &hash_keys); |
| 2579 |
break; |
| 2580 |
case 2: |
| 2581 |
memset(&hash_keys, 0, sizeof(hash_keys)); |
| 2582 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2583 |
if (skb) { |
| 2584 |
struct flow_keys keys; |
| 2585 |
|
| 2586 |
if (!flkeys) { |
| 2587 |
skb_flow_dissect_flow_keys(skb, &keys, 0); |
| 2588 |
flkeys = &keys; |
| 2589 |
} |
| 2590 |
|
| 2591 |
/* Inner can be v4 or v6 */ |
| 2592 |
if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) { |
| 2593 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS; |
| 2594 |
hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src; |
| 2595 |
hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst; |
| 2596 |
} else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) { |
| 2597 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2598 |
hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src; |
| 2599 |
hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst; |
| 2600 |
hash_keys.tags.flow_label = flkeys->tags.flow_label; |
| 2601 |
hash_keys.basic.ip_proto = flkeys->basic.ip_proto; |
| 2602 |
} else { |
| 2603 |
/* Same as case 0 */ |
| 2604 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2605 |
ip6_multipath_l3_keys(skb, &hash_keys, flkeys); |
| 2606 |
} |
| 2607 |
} else { |
| 2608 |
/* Same as case 0 */ |
| 2609 |
hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; |
| 2610 |
hash_keys.addrs.v6addrs.src = fl6->saddr; |
| 2611 |
hash_keys.addrs.v6addrs.dst = fl6->daddr; |
| 2612 |
hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6); |
| 2613 |
hash_keys.basic.ip_proto = fl6->flowi6_proto; |
| 2614 |
} |
| 2615 |
mhash = fib_multipath_hash_from_keys(net, &hash_keys); |
| 2616 |
break; |
| 2617 |
case 3: |
| 2618 |
if (skb) |
| 2619 |
mhash = rt6_multipath_custom_hash_skb(net, skb); |
| 2620 |
else |
| 2621 |
mhash = rt6_multipath_custom_hash_fl6(net, fl6); |
| 2622 |
break; |
| 2623 |
} |
| 2624 |
|
| 2625 |
return mhash >> 1; |
| 2626 |
} |
| 2627 |
|
| 2628 |
/* Called with rcu held */ |
| 2629 |
void ip6_route_input(struct sk_buff *skb) |
| 2630 |
{ |
| 2631 |
const struct ipv6hdr *iph = ipv6_hdr(skb); |
| 2632 |
struct net *net = dev_net(skb->dev); |
| 2633 |
int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF; |
| 2634 |
struct ip_tunnel_info *tun_info; |
| 2635 |
struct flowi6 fl6 = { |
| 2636 |
.flowi6_iif = skb->dev->ifindex, |
| 2637 |
.daddr = iph->daddr, |
| 2638 |
.saddr = iph->saddr, |
| 2639 |
.flowlabel = ip6_flowinfo(iph), |
| 2640 |
.flowi6_mark = skb->mark, |
| 2641 |
.flowi6_proto = iph->nexthdr, |
| 2642 |
}; |
| 2643 |
struct flow_keys *flkeys = NULL, _flkeys; |
| 2644 |
|
| 2645 |
tun_info = skb_tunnel_info(skb); |
| 2646 |
if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX)) |
| 2647 |
fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id; |
| 2648 |
|
| 2649 |
if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys)) |
| 2650 |
flkeys = &_flkeys; |
| 2651 |
|
| 2652 |
if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6)) |
| 2653 |
fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys); |
| 2654 |
skb_dst_drop(skb); |
| 2655 |
skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev, |
| 2656 |
&fl6, skb, flags)); |
| 2657 |
} |
| 2658 |
|
| 2659 |
INDIRECT_CALLABLE_SCOPE struct rt6_info *ip6_pol_route_output(struct net *net, |
| 2660 |
struct fib6_table *table, |
| 2661 |
struct flowi6 *fl6, |
| 2662 |
const struct sk_buff *skb, |
| 2663 |
int flags) |
| 2664 |
{ |
| 2665 |
return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags); |
| 2666 |
} |
| 2667 |
|
| 2668 |
static struct dst_entry *ip6_route_output_flags_noref(struct net *net, |
| 2669 |
const struct sock *sk, |
| 2670 |
struct flowi6 *fl6, |
| 2671 |
int flags) |
| 2672 |
{ |
| 2673 |
bool any_src; |
| 2674 |
|
| 2675 |
if (ipv6_addr_type(&fl6->daddr) & |
| 2676 |
(IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) { |
| 2677 |
struct dst_entry *dst; |
| 2678 |
|
| 2679 |
/* This function does not take refcnt on the dst */ |
| 2680 |
dst = l3mdev_link_scope_lookup(net, fl6); |
| 2681 |
if (dst) |
| 2682 |
return dst; |
| 2683 |
} |
| 2684 |
|
| 2685 |
fl6->flowi6_iif = LOOPBACK_IFINDEX; |
| 2686 |
|
| 2687 |
flags |= RT6_LOOKUP_F_DST_NOREF; |
| 2688 |
any_src = ipv6_addr_any(&fl6->saddr); |
| 2689 |
if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) || |
| 2690 |
(fl6->flowi6_oif && any_src)) |
| 2691 |
flags |= RT6_LOOKUP_F_IFACE; |
| 2692 |
|
| 2693 |
if (!any_src) |
| 2694 |
flags |= RT6_LOOKUP_F_HAS_SADDR; |
| 2695 |
else if (sk) |
| 2696 |
flags |= rt6_srcprefs2flags(READ_ONCE(inet6_sk(sk)->srcprefs)); |
| 2697 |
|
| 2698 |
return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output); |
| 2699 |
} |
| 2700 |
|
| 2701 |
struct dst_entry *ip6_route_output_flags(struct net *net, |
| 2702 |
const struct sock *sk, |
| 2703 |
struct flowi6 *fl6, |
| 2704 |
int flags) |
| 2705 |
{ |
| 2706 |
struct dst_entry *dst; |
| 2707 |
struct rt6_info *rt6; |
| 2708 |
|
| 2709 |
rcu_read_lock(); |
| 2710 |
dst = ip6_route_output_flags_noref(net, sk, fl6, flags); |
| 2711 |
rt6 = dst_rt6_info(dst); |
| 2712 |
/* For dst cached in uncached_list, refcnt is already taken. */ |
| 2713 |
if (list_empty(&rt6->dst.rt_uncached) && !dst_hold_safe(dst)) { |
| 2714 |
dst = &net->ipv6.ip6_null_entry->dst; |
| 2715 |
dst_hold(dst); |
| 2716 |
} |
| 2717 |
rcu_read_unlock(); |
| 2718 |
|
| 2719 |
return dst; |
| 2720 |
} |
| 2721 |
EXPORT_SYMBOL_GPL(ip6_route_output_flags); |
| 2722 |
|
| 2723 |
struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig) |
| 2724 |
{ |
| 2725 |
struct rt6_info *rt, *ort = dst_rt6_info(dst_orig); |
| 2726 |
struct net_device *loopback_dev = net->loopback_dev; |
| 2727 |
struct dst_entry *new = NULL; |
| 2728 |
|
| 2729 |
rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, |
| 2730 |
DST_OBSOLETE_DEAD, 0); |
| 2731 |
if (rt) { |
| 2732 |
rt6_info_init(rt); |
| 2733 |
atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc); |
| 2734 |
|
| 2735 |
new = &rt->dst; |
| 2736 |
new->__use = 1; |
| 2737 |
new->input = dst_discard; |
| 2738 |
new->output = dst_discard_out; |
| 2739 |
|
| 2740 |
dst_copy_metrics(new, &ort->dst); |
| 2741 |
|
| 2742 |
rt->rt6i_idev = in6_dev_get(loopback_dev); |
| 2743 |
rt->rt6i_gateway = ort->rt6i_gateway; |
| 2744 |
rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU; |
| 2745 |
|
| 2746 |
memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key)); |
| 2747 |
#ifdef CONFIG_IPV6_SUBTREES |
| 2748 |
memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key)); |
| 2749 |
#endif |
| 2750 |
} |
| 2751 |
|
| 2752 |
dst_release(dst_orig); |
| 2753 |
return new ? new : ERR_PTR(-ENOMEM); |
| 2754 |
} |
| 2755 |
|
| 2756 |
/* |
| 2757 |
* Destination cache support functions |
| 2758 |
*/ |
| 2759 |
|
| 2760 |
static bool fib6_check(struct fib6_info *f6i, u32 cookie) |
| 2761 |
{ |
| 2762 |
u32 rt_cookie = 0; |
| 2763 |
|
| 2764 |
if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie) |
| 2765 |
return false; |
| 2766 |
|
| 2767 |
if (fib6_check_expired(f6i)) |
| 2768 |
return false; |
| 2769 |
|
| 2770 |
return true; |
| 2771 |
} |
| 2772 |
|
| 2773 |
static struct dst_entry *rt6_check(struct rt6_info *rt, |
| 2774 |
struct fib6_info *from, |
| 2775 |
u32 cookie) |
| 2776 |
{ |
| 2777 |
u32 rt_cookie = 0; |
| 2778 |
|
| 2779 |
if (!from || !fib6_get_cookie_safe(from, &rt_cookie) || |
| 2780 |
rt_cookie != cookie) |
| 2781 |
return NULL; |
| 2782 |
|
| 2783 |
if (rt6_check_expired(rt)) |
| 2784 |
return NULL; |
| 2785 |
|
| 2786 |
return &rt->dst; |
| 2787 |
} |
| 2788 |
|
| 2789 |
static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, |
| 2790 |
struct fib6_info *from, |
| 2791 |
u32 cookie) |
| 2792 |
{ |
| 2793 |
if (!__rt6_check_expired(rt) && |
| 2794 |
READ_ONCE(rt->dst.obsolete) == DST_OBSOLETE_FORCE_CHK && |
| 2795 |
fib6_check(from, cookie)) |
| 2796 |
return &rt->dst; |
| 2797 |
return NULL; |
| 2798 |
} |
| 2799 |
|
| 2800 |
INDIRECT_CALLABLE_SCOPE struct dst_entry *ip6_dst_check(struct dst_entry *dst, |
| 2801 |
u32 cookie) |
| 2802 |
{ |
| 2803 |
struct dst_entry *dst_ret; |
| 2804 |
struct fib6_info *from; |
| 2805 |
struct rt6_info *rt; |
| 2806 |
|
| 2807 |
rt = dst_rt6_info(dst); |
| 2808 |
|
| 2809 |
if (rt->sernum) |
| 2810 |
return rt6_is_valid(rt) ? dst : NULL; |
| 2811 |
|
| 2812 |
rcu_read_lock(); |
| 2813 |
|
| 2814 |
/* All IPV6 dsts are created with ->obsolete set to the value |
| 2815 |
* DST_OBSOLETE_FORCE_CHK which forces validation calls down |
| 2816 |
* into this function always. |
| 2817 |
*/ |
| 2818 |
|
| 2819 |
from = rcu_dereference(rt->from); |
| 2820 |
|
| 2821 |
if (from && (rt->rt6i_flags & RTF_PCPU || |
| 2822 |
unlikely(!list_empty(&rt->dst.rt_uncached)))) |
| 2823 |
dst_ret = rt6_dst_from_check(rt, from, cookie); |
| 2824 |
else |
| 2825 |
dst_ret = rt6_check(rt, from, cookie); |
| 2826 |
|
| 2827 |
rcu_read_unlock(); |
| 2828 |
|
| 2829 |
return dst_ret; |
| 2830 |
} |
| 2831 |
EXPORT_INDIRECT_CALLABLE(ip6_dst_check); |
| 2832 |
|
| 2833 |
static void ip6_negative_advice(struct sock *sk, |
| 2834 |
struct dst_entry *dst) |
| 2835 |
{ |
| 2836 |
struct rt6_info *rt = dst_rt6_info(dst); |
| 2837 |
|
| 2838 |
if (rt->rt6i_flags & RTF_CACHE) { |
| 2839 |
rcu_read_lock(); |
| 2840 |
if (rt6_check_expired(rt)) { |
| 2841 |
/* rt/dst can not be destroyed yet, |
| 2842 |
* because of rcu_read_lock() |
| 2843 |
*/ |
| 2844 |
sk_dst_reset(sk); |
| 2845 |
rt6_remove_exception_rt(rt); |
| 2846 |
} |
| 2847 |
rcu_read_unlock(); |
| 2848 |
return; |
| 2849 |
} |
| 2850 |
sk_dst_reset(sk); |
| 2851 |
} |
| 2852 |
|
| 2853 |
static void ip6_link_failure(struct sk_buff *skb) |
| 2854 |
{ |
| 2855 |
struct rt6_info *rt; |
| 2856 |
|
| 2857 |
icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0); |
| 2858 |
|
| 2859 |
rt = dst_rt6_info(skb_dst(skb)); |
| 2860 |
if (rt) { |
| 2861 |
rcu_read_lock(); |
| 2862 |
if (rt->rt6i_flags & RTF_CACHE) { |
| 2863 |
rt6_remove_exception_rt(rt); |
| 2864 |
} else { |
| 2865 |
struct fib6_info *from; |
| 2866 |
struct fib6_node *fn; |
| 2867 |
|
| 2868 |
from = rcu_dereference(rt->from); |
| 2869 |
if (from) { |
| 2870 |
fn = rcu_dereference(from->fib6_node); |
| 2871 |
if (fn && (rt->rt6i_flags & RTF_DEFAULT)) |
| 2872 |
WRITE_ONCE(fn->fn_sernum, -1); |
| 2873 |
} |
| 2874 |
} |
| 2875 |
rcu_read_unlock(); |
| 2876 |
} |
| 2877 |
} |
| 2878 |
|
| 2879 |
static void rt6_update_expires(struct rt6_info *rt0, int timeout) |
| 2880 |
{ |
| 2881 |
if (!(rt0->rt6i_flags & RTF_EXPIRES)) { |
| 2882 |
struct fib6_info *from; |
| 2883 |
|
| 2884 |
rcu_read_lock(); |
| 2885 |
from = rcu_dereference(rt0->from); |
| 2886 |
if (from) |
| 2887 |
WRITE_ONCE(rt0->dst.expires, from->expires); |
| 2888 |
rcu_read_unlock(); |
| 2889 |
} |
| 2890 |
|
| 2891 |
dst_set_expires(&rt0->dst, timeout); |
| 2892 |
rt0->rt6i_flags |= RTF_EXPIRES; |
| 2893 |
} |
| 2894 |
|
| 2895 |
static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu) |
| 2896 |
{ |
| 2897 |
struct net *net = dev_net(rt->dst.dev); |
| 2898 |
|
| 2899 |
dst_metric_set(&rt->dst, RTAX_MTU, mtu); |
| 2900 |
rt->rt6i_flags |= RTF_MODIFIED; |
| 2901 |
rt6_update_expires(rt, READ_ONCE(net->ipv6.sysctl.ip6_rt_mtu_expires)); |
| 2902 |
} |
| 2903 |
|
| 2904 |
static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt) |
| 2905 |
{ |
| 2906 |
return !(rt->rt6i_flags & RTF_CACHE) && |
| 2907 |
(rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from)); |
| 2908 |
} |
| 2909 |
|
| 2910 |
static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk, |
| 2911 |
const struct ipv6hdr *iph, u32 mtu, |
| 2912 |
bool confirm_neigh) |
| 2913 |
{ |
| 2914 |
const struct in6_addr *daddr, *saddr; |
| 2915 |
struct rt6_info *rt6 = dst_rt6_info(dst); |
| 2916 |
|
| 2917 |
/* Note: do *NOT* check dst_metric_locked(dst, RTAX_MTU) |
| 2918 |
* IPv6 pmtu discovery isn't optional, so 'mtu lock' cannot disable it. |
| 2919 |
* [see also comment in rt6_mtu_change_route()] |
| 2920 |
*/ |
| 2921 |
|
| 2922 |
if (iph) { |
| 2923 |
daddr = &iph->daddr; |
| 2924 |
saddr = &iph->saddr; |
| 2925 |
} else if (sk) { |
| 2926 |
daddr = &sk->sk_v6_daddr; |
| 2927 |
saddr = &inet6_sk(sk)->saddr; |
| 2928 |
} else { |
| 2929 |
daddr = NULL; |
| 2930 |
saddr = NULL; |
| 2931 |
} |
| 2932 |
|
| 2933 |
if (confirm_neigh) |
| 2934 |
dst_confirm_neigh(dst, daddr); |
| 2935 |
|
| 2936 |
if (mtu < IPV6_MIN_MTU) |
| 2937 |
return; |
| 2938 |
if (mtu >= dst6_mtu(dst)) |
| 2939 |
return; |
| 2940 |
|
| 2941 |
if (!rt6_cache_allowed_for_pmtu(rt6)) { |
| 2942 |
rt6_do_update_pmtu(rt6, mtu); |
| 2943 |
/* update rt6_ex->stamp for cache */ |
| 2944 |
if (rt6->rt6i_flags & RTF_CACHE) |
| 2945 |
rt6_update_exception_stamp_rt(rt6); |
| 2946 |
} else if (daddr) { |
| 2947 |
struct fib6_result res = {}; |
| 2948 |
struct rt6_info *nrt6; |
| 2949 |
|
| 2950 |
rcu_read_lock(); |
| 2951 |
res.f6i = rcu_dereference(rt6->from); |
| 2952 |
if (!res.f6i) |
| 2953 |
goto out_unlock; |
| 2954 |
|
| 2955 |
res.fib6_flags = res.f6i->fib6_flags; |
| 2956 |
res.fib6_type = res.f6i->fib6_type; |
| 2957 |
|
| 2958 |
if (res.f6i->nh) { |
| 2959 |
struct fib6_nh_match_arg arg = { |
| 2960 |
.dev = dst_dev_rcu(dst), |
| 2961 |
.gw = &rt6->rt6i_gateway, |
| 2962 |
}; |
| 2963 |
|
| 2964 |
nexthop_for_each_fib6_nh(res.f6i->nh, |
| 2965 |
fib6_nh_find_match, &arg); |
| 2966 |
|
| 2967 |
/* fib6_info uses a nexthop that does not have fib6_nh |
| 2968 |
* using the dst->dev + gw. Should be impossible. |
| 2969 |
*/ |
| 2970 |
if (!arg.match) |
| 2971 |
goto out_unlock; |
| 2972 |
|
| 2973 |
res.nh = arg.match; |
| 2974 |
} else { |
| 2975 |
res.nh = res.f6i->fib6_nh; |
| 2976 |
} |
| 2977 |
|
| 2978 |
nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr); |
| 2979 |
if (nrt6) { |
| 2980 |
rt6_do_update_pmtu(nrt6, mtu); |
| 2981 |
if (rt6_insert_exception(nrt6, &res)) |
| 2982 |
dst_release_immediate(&nrt6->dst); |
| 2983 |
} |
| 2984 |
out_unlock: |
| 2985 |
rcu_read_unlock(); |
| 2986 |
} |
| 2987 |
} |
| 2988 |
|
| 2989 |
static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk, |
| 2990 |
struct sk_buff *skb, u32 mtu, |
| 2991 |
bool confirm_neigh) |
| 2992 |
{ |
| 2993 |
__ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu, |
| 2994 |
confirm_neigh); |
| 2995 |
} |
| 2996 |
|
| 2997 |
void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu, |
| 2998 |
int oif, u32 mark, kuid_t uid) |
| 2999 |
{ |
| 3000 |
const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data; |
| 3001 |
struct dst_entry *dst; |
| 3002 |
struct flowi6 fl6 = { |
| 3003 |
.flowi6_oif = oif, |
| 3004 |
.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark), |
| 3005 |
.daddr = iph->daddr, |
| 3006 |
.saddr = iph->saddr, |
| 3007 |
.flowlabel = ip6_flowinfo(iph), |
| 3008 |
.flowi6_uid = uid, |
| 3009 |
}; |
| 3010 |
|
| 3011 |
dst = ip6_route_output(net, NULL, &fl6); |
| 3012 |
if (!dst->error) |
| 3013 |
__ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu), true); |
| 3014 |
dst_release(dst); |
| 3015 |
} |
| 3016 |
EXPORT_SYMBOL_GPL(ip6_update_pmtu); |
| 3017 |
|
| 3018 |
void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu) |
| 3019 |
{ |
| 3020 |
int oif = sk->sk_bound_dev_if; |
| 3021 |
struct dst_entry *dst; |
| 3022 |
|
| 3023 |
if (!oif && skb->dev) |
| 3024 |
oif = l3mdev_master_ifindex(skb->dev); |
| 3025 |
|
| 3026 |
ip6_update_pmtu(skb, sock_net(sk), mtu, oif, READ_ONCE(sk->sk_mark), |
| 3027 |
sk_uid(sk)); |
| 3028 |
|
| 3029 |
dst = __sk_dst_get(sk); |
| 3030 |
if (!dst || !READ_ONCE(dst->obsolete) || |
| 3031 |
dst->ops->check(dst, inet6_sk(sk)->dst_cookie)) |
| 3032 |
return; |
| 3033 |
|
| 3034 |
bh_lock_sock(sk); |
| 3035 |
if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) |
| 3036 |
ip6_datagram_dst_update(sk, false); |
| 3037 |
bh_unlock_sock(sk); |
| 3038 |
} |
| 3039 |
EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu); |
| 3040 |
|
| 3041 |
void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst, |
| 3042 |
const struct flowi6 *fl6) |
| 3043 |
{ |
| 3044 |
#ifdef CONFIG_IPV6_SUBTREES |
| 3045 |
struct ipv6_pinfo *np = inet6_sk(sk); |
| 3046 |
#endif |
| 3047 |
|
| 3048 |
ip6_dst_store(sk, dst, |
| 3049 |
ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr), |
| 3050 |
#ifdef CONFIG_IPV6_SUBTREES |
| 3051 |
ipv6_addr_equal(&fl6->saddr, &np->saddr) ? |
| 3052 |
true : |
| 3053 |
#endif |
| 3054 |
false); |
| 3055 |
} |
| 3056 |
|
| 3057 |
static bool ip6_redirect_nh_match(const struct fib6_result *res, |
| 3058 |
struct flowi6 *fl6, |
| 3059 |
const struct in6_addr *gw, |
| 3060 |
struct rt6_info **ret) |
| 3061 |
{ |
| 3062 |
const struct fib6_nh *nh = res->nh; |
| 3063 |
|
| 3064 |
if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family || |
| 3065 |
fl6->flowi6_oif != nh->fib_nh_dev->ifindex) |
| 3066 |
return false; |
| 3067 |
|
| 3068 |
/* rt_cache's gateway might be different from its 'parent' |
| 3069 |
* in the case of an ip redirect. |
| 3070 |
* So we keep searching in the exception table if the gateway |
| 3071 |
* is different. |
| 3072 |
*/ |
| 3073 |
if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) { |
| 3074 |
struct rt6_info *rt_cache; |
| 3075 |
|
| 3076 |
rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr); |
| 3077 |
if (rt_cache && |
| 3078 |
ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) { |
| 3079 |
*ret = rt_cache; |
| 3080 |
return true; |
| 3081 |
} |
| 3082 |
return false; |
| 3083 |
} |
| 3084 |
return true; |
| 3085 |
} |
| 3086 |
|
| 3087 |
struct fib6_nh_rd_arg { |
| 3088 |
struct fib6_result *res; |
| 3089 |
struct flowi6 *fl6; |
| 3090 |
const struct in6_addr *gw; |
| 3091 |
struct rt6_info **ret; |
| 3092 |
}; |
| 3093 |
|
| 3094 |
static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg) |
| 3095 |
{ |
| 3096 |
struct fib6_nh_rd_arg *arg = _arg; |
| 3097 |
|
| 3098 |
arg->res->nh = nh; |
| 3099 |
return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret); |
| 3100 |
} |
| 3101 |
|
| 3102 |
/* Handle redirects */ |
| 3103 |
struct ip6rd_flowi { |
| 3104 |
struct flowi6 fl6; |
| 3105 |
struct in6_addr gateway; |
| 3106 |
}; |
| 3107 |
|
| 3108 |
INDIRECT_CALLABLE_SCOPE struct rt6_info *__ip6_route_redirect(struct net *net, |
| 3109 |
struct fib6_table *table, |
| 3110 |
struct flowi6 *fl6, |
| 3111 |
const struct sk_buff *skb, |
| 3112 |
int flags) |
| 3113 |
{ |
| 3114 |
struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6; |
| 3115 |
struct rt6_info *ret = NULL; |
| 3116 |
struct fib6_result res = {}; |
| 3117 |
struct fib6_nh_rd_arg arg = { |
| 3118 |
.res = &res, |
| 3119 |
.fl6 = fl6, |
| 3120 |
.gw = &rdfl->gateway, |
| 3121 |
.ret = &ret |
| 3122 |
}; |
| 3123 |
struct fib6_info *rt; |
| 3124 |
struct fib6_node *fn; |
| 3125 |
|
| 3126 |
/* Get the "current" route for this destination and |
| 3127 |
* check if the redirect has come from appropriate router. |
| 3128 |
* |
| 3129 |
* RFC 4861 specifies that redirects should only be |
| 3130 |
* accepted if they come from the nexthop to the target. |
| 3131 |
* Due to the way the routes are chosen, this notion |
| 3132 |
* is a bit fuzzy and one might need to check all possible |
| 3133 |
* routes. |
| 3134 |
*/ |
| 3135 |
|
| 3136 |
rcu_read_lock(); |
| 3137 |
fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr); |
| 3138 |
restart: |
| 3139 |
for_each_fib6_node_rt_rcu(fn) { |
| 3140 |
res.f6i = rt; |
| 3141 |
if (fib6_check_expired(rt)) |
| 3142 |
continue; |
| 3143 |
if (rt->fib6_flags & RTF_REJECT) |
| 3144 |
break; |
| 3145 |
if (unlikely(rt->nh)) { |
| 3146 |
if (nexthop_is_blackhole(rt->nh)) |
| 3147 |
continue; |
| 3148 |
/* on match, res->nh is filled in and potentially ret */ |
| 3149 |
if (nexthop_for_each_fib6_nh(rt->nh, |
| 3150 |
fib6_nh_redirect_match, |
| 3151 |
&arg)) |
| 3152 |
goto out; |
| 3153 |
} else { |
| 3154 |
res.nh = rt->fib6_nh; |
| 3155 |
if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway, |
| 3156 |
&ret)) |
| 3157 |
goto out; |
| 3158 |
} |
| 3159 |
} |
| 3160 |
|
| 3161 |
if (!rt) |
| 3162 |
rt = net->ipv6.fib6_null_entry; |
| 3163 |
else if (rt->fib6_flags & RTF_REJECT) { |
| 3164 |
ret = net->ipv6.ip6_null_entry; |
| 3165 |
goto out; |
| 3166 |
} |
| 3167 |
|
| 3168 |
if (rt == net->ipv6.fib6_null_entry) { |
| 3169 |
fn = fib6_backtrack(fn, &fl6->saddr); |
| 3170 |
if (fn) |
| 3171 |
goto restart; |
| 3172 |
} |
| 3173 |
|
| 3174 |
res.f6i = rt; |
| 3175 |
res.nh = rt->fib6_nh; |
| 3176 |
out: |
| 3177 |
if (ret) { |
| 3178 |
ip6_hold_safe(net, &ret); |
| 3179 |
} else { |
| 3180 |
res.fib6_flags = res.f6i->fib6_flags; |
| 3181 |
res.fib6_type = res.f6i->fib6_type; |
| 3182 |
ret = ip6_create_rt_rcu(&res); |
| 3183 |
} |
| 3184 |
|
| 3185 |
rcu_read_unlock(); |
| 3186 |
|
| 3187 |
trace_fib6_table_lookup(net, &res, table, fl6); |
| 3188 |
return ret; |
| 3189 |
}; |
| 3190 |
|
| 3191 |
static struct dst_entry *ip6_route_redirect(struct net *net, |
| 3192 |
const struct flowi6 *fl6, |
| 3193 |
const struct sk_buff *skb, |
| 3194 |
const struct in6_addr *gateway) |
| 3195 |
{ |
| 3196 |
int flags = RT6_LOOKUP_F_HAS_SADDR; |
| 3197 |
struct ip6rd_flowi rdfl; |
| 3198 |
|
| 3199 |
rdfl.fl6 = *fl6; |
| 3200 |
rdfl.gateway = *gateway; |
| 3201 |
|
| 3202 |
return fib6_rule_lookup(net, &rdfl.fl6, skb, |
| 3203 |
flags, __ip6_route_redirect); |
| 3204 |
} |
| 3205 |
|
| 3206 |
void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark, |
| 3207 |
kuid_t uid) |
| 3208 |
{ |
| 3209 |
const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data; |
| 3210 |
struct dst_entry *dst; |
| 3211 |
struct flowi6 fl6 = { |
| 3212 |
.flowi6_iif = LOOPBACK_IFINDEX, |
| 3213 |
.flowi6_oif = oif, |
| 3214 |
.flowi6_mark = mark, |
| 3215 |
.daddr = iph->daddr, |
| 3216 |
.saddr = iph->saddr, |
| 3217 |
.flowlabel = ip6_flowinfo(iph), |
| 3218 |
.flowi6_uid = uid, |
| 3219 |
}; |
| 3220 |
|
| 3221 |
dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr); |
| 3222 |
rt6_do_redirect(dst, NULL, skb); |
| 3223 |
dst_release(dst); |
| 3224 |
} |
| 3225 |
EXPORT_SYMBOL_GPL(ip6_redirect); |
| 3226 |
|
| 3227 |
void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif) |
| 3228 |
{ |
| 3229 |
const struct ipv6hdr *iph = ipv6_hdr(skb); |
| 3230 |
const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb); |
| 3231 |
struct dst_entry *dst; |
| 3232 |
struct flowi6 fl6 = { |
| 3233 |
.flowi6_iif = LOOPBACK_IFINDEX, |
| 3234 |
.flowi6_oif = oif, |
| 3235 |
.daddr = msg->dest, |
| 3236 |
.saddr = iph->daddr, |
| 3237 |
.flowi6_uid = sock_net_uid(net, NULL), |
| 3238 |
}; |
| 3239 |
|
| 3240 |
dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr); |
| 3241 |
rt6_do_redirect(dst, NULL, skb); |
| 3242 |
dst_release(dst); |
| 3243 |
} |
| 3244 |
|
| 3245 |
void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk) |
| 3246 |
{ |
| 3247 |
ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, |
| 3248 |
READ_ONCE(sk->sk_mark), sk_uid(sk)); |
| 3249 |
} |
| 3250 |
EXPORT_SYMBOL_GPL(ip6_sk_redirect); |
| 3251 |
|
| 3252 |
static unsigned int ip6_default_advmss(const struct dst_entry *dst) |
| 3253 |
{ |
| 3254 |
unsigned int mtu = dst6_mtu(dst); |
| 3255 |
struct net *net; |
| 3256 |
|
| 3257 |
mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr); |
| 3258 |
|
| 3259 |
rcu_read_lock(); |
| 3260 |
|
| 3261 |
net = dst_dev_net_rcu(dst); |
| 3262 |
mtu = max_t(unsigned int, mtu, |
| 3263 |
READ_ONCE(net->ipv6.sysctl.ip6_rt_min_advmss)); |
| 3264 |
|
| 3265 |
rcu_read_unlock(); |
| 3266 |
|
| 3267 |
/* |
| 3268 |
* Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and |
| 3269 |
* corresponding MSS is IPV6_MAXPLEN - tcp_header_size. |
| 3270 |
* IPV6_MAXPLEN is also valid and means: "any MSS, |
| 3271 |
* rely only on pmtu discovery" |
| 3272 |
*/ |
| 3273 |
if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr)) |
| 3274 |
mtu = IPV6_MAXPLEN; |
| 3275 |
return mtu; |
| 3276 |
} |
| 3277 |
|
| 3278 |
INDIRECT_CALLABLE_SCOPE unsigned int ip6_mtu(const struct dst_entry *dst) |
| 3279 |
{ |
| 3280 |
return ip6_dst_mtu_maybe_forward(dst, false); |
| 3281 |
} |
| 3282 |
EXPORT_INDIRECT_CALLABLE(ip6_mtu); |
| 3283 |
|
| 3284 |
/* MTU selection: |
| 3285 |
* 1. mtu on route is locked - use it |
| 3286 |
* 2. mtu from nexthop exception |
| 3287 |
* 3. mtu from egress device |
| 3288 |
* |
| 3289 |
* based on ip6_dst_mtu_forward and exception logic of |
| 3290 |
* rt6_find_cached_rt; called with rcu_read_lock |
| 3291 |
*/ |
| 3292 |
u32 ip6_mtu_from_fib6(const struct fib6_result *res, |
| 3293 |
const struct in6_addr *daddr, |
| 3294 |
const struct in6_addr *saddr) |
| 3295 |
{ |
| 3296 |
const struct fib6_nh *nh = res->nh; |
| 3297 |
struct fib6_info *f6i = res->f6i; |
| 3298 |
struct inet6_dev *idev; |
| 3299 |
struct rt6_info *rt; |
| 3300 |
u32 mtu = 0; |
| 3301 |
|
| 3302 |
if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) { |
| 3303 |
mtu = f6i->fib6_pmtu; |
| 3304 |
if (mtu) |
| 3305 |
goto out; |
| 3306 |
} |
| 3307 |
|
| 3308 |
rt = rt6_find_cached_rt(res, daddr, saddr); |
| 3309 |
if (unlikely(rt)) { |
| 3310 |
mtu = dst_metric_raw(&rt->dst, RTAX_MTU); |
| 3311 |
} else { |
| 3312 |
struct net_device *dev = nh->fib_nh_dev; |
| 3313 |
|
| 3314 |
mtu = IPV6_MIN_MTU; |
| 3315 |
idev = __in6_dev_get(dev); |
| 3316 |
if (idev) |
| 3317 |
mtu = max_t(u32, mtu, READ_ONCE(idev->cnf.mtu6)); |
| 3318 |
} |
| 3319 |
|
| 3320 |
mtu = min_t(unsigned int, mtu, IP6_MAX_MTU); |
| 3321 |
out: |
| 3322 |
return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu); |
| 3323 |
} |
| 3324 |
|
| 3325 |
struct dst_entry *icmp6_dst_alloc(struct net_device *dev, |
| 3326 |
struct flowi6 *fl6) |
| 3327 |
{ |
| 3328 |
struct dst_entry *dst; |
| 3329 |
struct rt6_info *rt; |
| 3330 |
struct inet6_dev *idev = in6_dev_get(dev); |
| 3331 |
struct net *net = dev_net(dev); |
| 3332 |
|
| 3333 |
if (unlikely(!idev)) |
| 3334 |
return ERR_PTR(-ENODEV); |
| 3335 |
|
| 3336 |
rt = ip6_dst_alloc(net, dev, 0); |
| 3337 |
if (unlikely(!rt)) { |
| 3338 |
in6_dev_put(idev); |
| 3339 |
dst = ERR_PTR(-ENOMEM); |
| 3340 |
goto out; |
| 3341 |
} |
| 3342 |
|
| 3343 |
rt->dst.input = ip6_input; |
| 3344 |
rt->dst.output = ip6_output; |
| 3345 |
rt->rt6i_gateway = fl6->daddr; |
| 3346 |
rt->rt6i_dst.addr = fl6->daddr; |
| 3347 |
rt->rt6i_dst.plen = 128; |
| 3348 |
rt->rt6i_idev = idev; |
| 3349 |
dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0); |
| 3350 |
|
| 3351 |
/* Add this dst into uncached_list so that rt6_disable_ip() can |
| 3352 |
* do proper release of the net_device |
| 3353 |
*/ |
| 3354 |
rt6_uncached_list_add(rt); |
| 3355 |
|
| 3356 |
dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0); |
| 3357 |
|
| 3358 |
out: |
| 3359 |
return dst; |
| 3360 |
} |
| 3361 |
|
| 3362 |
static void ip6_dst_gc(struct dst_ops *ops) |
| 3363 |
{ |
| 3364 |
struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops); |
| 3365 |
int rt_min_interval = READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_min_interval); |
| 3366 |
int rt_elasticity = READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_elasticity); |
| 3367 |
int rt_gc_timeout = READ_ONCE(net->ipv6.sysctl.ip6_rt_gc_timeout); |
| 3368 |
unsigned long rt_last_gc = READ_ONCE(net->ipv6.ip6_rt_last_gc); |
| 3369 |
unsigned int val; |
| 3370 |
int entries; |
| 3371 |
|
| 3372 |
if (time_after(rt_last_gc + rt_min_interval, jiffies)) |
| 3373 |
goto out; |
| 3374 |
|
| 3375 |
fib6_run_gc(atomic_inc_return(&net->ipv6.ip6_rt_gc_expire), net, true); |
| 3376 |
entries = dst_entries_get_slow(ops); |
| 3377 |
if (entries < ops->gc_thresh) |
| 3378 |
atomic_set(&net->ipv6.ip6_rt_gc_expire, rt_gc_timeout >> 1); |
| 3379 |
out: |
| 3380 |
val = atomic_read(&net->ipv6.ip6_rt_gc_expire); |
| 3381 |
atomic_set(&net->ipv6.ip6_rt_gc_expire, val - (val >> rt_elasticity)); |
| 3382 |
} |
| 3383 |
|
| 3384 |
static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg, |
| 3385 |
const struct in6_addr *gw_addr, u32 tbid, |
| 3386 |
int flags, struct fib6_result *res) |
| 3387 |
{ |
| 3388 |
struct flowi6 fl6 = { |
| 3389 |
.flowi6_oif = cfg->fc_ifindex, |
| 3390 |
.daddr = *gw_addr, |
| 3391 |
.saddr = cfg->fc_prefsrc, |
| 3392 |
}; |
| 3393 |
struct fib6_table *table; |
| 3394 |
int err; |
| 3395 |
|
| 3396 |
table = fib6_get_table(net, tbid); |
| 3397 |
if (!table) |
| 3398 |
return -EINVAL; |
| 3399 |
|
| 3400 |
if (!ipv6_addr_any(&cfg->fc_prefsrc)) |
| 3401 |
flags |= RT6_LOOKUP_F_HAS_SADDR; |
| 3402 |
|
| 3403 |
flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE; |
| 3404 |
|
| 3405 |
err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags); |
| 3406 |
if (!err && res->f6i != net->ipv6.fib6_null_entry) |
| 3407 |
fib6_select_path(net, res, &fl6, cfg->fc_ifindex, |
| 3408 |
cfg->fc_ifindex != 0, NULL, flags); |
| 3409 |
|
| 3410 |
return err; |
| 3411 |
} |
| 3412 |
|
| 3413 |
static int ip6_route_check_nh_onlink(struct net *net, |
| 3414 |
struct fib6_config *cfg, |
| 3415 |
const struct net_device *dev, |
| 3416 |
struct netlink_ext_ack *extack) |
| 3417 |
{ |
| 3418 |
u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN; |
| 3419 |
const struct in6_addr *gw_addr = &cfg->fc_gateway; |
| 3420 |
struct fib6_result res = {}; |
| 3421 |
int err; |
| 3422 |
|
| 3423 |
err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res); |
| 3424 |
if (!err && !(res.fib6_flags & RTF_REJECT) && |
| 3425 |
res.fib6_type != RTN_UNICAST) { |
| 3426 |
NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway"); |
| 3427 |
err = -EINVAL; |
| 3428 |
} |
| 3429 |
|
| 3430 |
return err; |
| 3431 |
} |
| 3432 |
|
| 3433 |
static int ip6_route_check_nh(struct net *net, |
| 3434 |
struct fib6_config *cfg, |
| 3435 |
struct net_device **_dev, |
| 3436 |
netdevice_tracker *dev_tracker, |
| 3437 |
struct inet6_dev **idev) |
| 3438 |
{ |
| 3439 |
const struct in6_addr *gw_addr = &cfg->fc_gateway; |
| 3440 |
struct net_device *dev = _dev ? *_dev : NULL; |
| 3441 |
int flags = RT6_LOOKUP_F_IFACE; |
| 3442 |
struct fib6_result res = {}; |
| 3443 |
int err = -EHOSTUNREACH; |
| 3444 |
|
| 3445 |
if (cfg->fc_table) { |
| 3446 |
err = ip6_nh_lookup_table(net, cfg, gw_addr, |
| 3447 |
cfg->fc_table, flags, &res); |
| 3448 |
/* gw_addr can not require a gateway or resolve to a reject |
| 3449 |
* route. If a device is given, it must match the result. |
| 3450 |
*/ |
| 3451 |
if (err || res.fib6_flags & RTF_REJECT || |
| 3452 |
res.nh->fib_nh_gw_family || |
| 3453 |
(dev && dev != res.nh->fib_nh_dev)) |
| 3454 |
err = -EHOSTUNREACH; |
| 3455 |
} |
| 3456 |
|
| 3457 |
if (err < 0) { |
| 3458 |
struct flowi6 fl6 = { |
| 3459 |
.flowi6_oif = cfg->fc_ifindex, |
| 3460 |
.daddr = *gw_addr, |
| 3461 |
}; |
| 3462 |
|
| 3463 |
err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags); |
| 3464 |
if (err || res.fib6_flags & RTF_REJECT || |
| 3465 |
res.nh->fib_nh_gw_family) |
| 3466 |
err = -EHOSTUNREACH; |
| 3467 |
|
| 3468 |
if (err) |
| 3469 |
return err; |
| 3470 |
|
| 3471 |
fib6_select_path(net, &res, &fl6, cfg->fc_ifindex, |
| 3472 |
cfg->fc_ifindex != 0, NULL, flags); |
| 3473 |
} |
| 3474 |
|
| 3475 |
err = 0; |
| 3476 |
if (dev) { |
| 3477 |
if (dev != res.nh->fib_nh_dev) |
| 3478 |
err = -EHOSTUNREACH; |
| 3479 |
} else { |
| 3480 |
*_dev = dev = res.nh->fib_nh_dev; |
| 3481 |
netdev_hold(dev, dev_tracker, GFP_ATOMIC); |
| 3482 |
*idev = in6_dev_get(dev); |
| 3483 |
} |
| 3484 |
|
| 3485 |
return err; |
| 3486 |
} |
| 3487 |
|
| 3488 |
static int ip6_validate_gw(struct net *net, struct fib6_config *cfg, |
| 3489 |
struct net_device **_dev, |
| 3490 |
netdevice_tracker *dev_tracker, |
| 3491 |
struct inet6_dev **idev, |
| 3492 |
struct netlink_ext_ack *extack) |
| 3493 |
{ |
| 3494 |
const struct in6_addr *gw_addr = &cfg->fc_gateway; |
| 3495 |
int gwa_type = ipv6_addr_type(gw_addr); |
| 3496 |
bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true; |
| 3497 |
const struct net_device *dev = *_dev; |
| 3498 |
bool need_addr_check = !dev; |
| 3499 |
int err = -EINVAL; |
| 3500 |
|
| 3501 |
/* if gw_addr is local we will fail to detect this in case |
| 3502 |
* address is still TENTATIVE (DAD in progress). rt6_lookup() |
| 3503 |
* will return already-added prefix route via interface that |
| 3504 |
* prefix route was assigned to, which might be non-loopback. |
| 3505 |
*/ |
| 3506 |
if (dev && |
| 3507 |
ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) { |
| 3508 |
NL_SET_ERR_MSG(extack, "Gateway can not be a local address"); |
| 3509 |
goto out; |
| 3510 |
} |
| 3511 |
|
| 3512 |
if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) { |
| 3513 |
/* IPv6 strictly inhibits using not link-local |
| 3514 |
* addresses as nexthop address. |
| 3515 |
* Otherwise, router will not able to send redirects. |
| 3516 |
* It is very good, but in some (rare!) circumstances |
| 3517 |
* (SIT, PtP, NBMA NOARP links) it is handy to allow |
| 3518 |
* some exceptions. --ANK |
| 3519 |
* We allow IPv4-mapped nexthops to support RFC4798-type |
| 3520 |
* addressing |
| 3521 |
*/ |
| 3522 |
if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) { |
| 3523 |
NL_SET_ERR_MSG(extack, "Invalid gateway address"); |
| 3524 |
goto out; |
| 3525 |
} |
| 3526 |
|
| 3527 |
rcu_read_lock(); |
| 3528 |
|
| 3529 |
if (cfg->fc_flags & RTNH_F_ONLINK) |
| 3530 |
err = ip6_route_check_nh_onlink(net, cfg, dev, extack); |
| 3531 |
else |
| 3532 |
err = ip6_route_check_nh(net, cfg, _dev, dev_tracker, |
| 3533 |
idev); |
| 3534 |
|
| 3535 |
rcu_read_unlock(); |
| 3536 |
|
| 3537 |
if (err) |
| 3538 |
goto out; |
| 3539 |
} |
| 3540 |
|
| 3541 |
/* reload in case device was changed */ |
| 3542 |
dev = *_dev; |
| 3543 |
|
| 3544 |
err = -EINVAL; |
| 3545 |
if (!dev) { |
| 3546 |
NL_SET_ERR_MSG(extack, "Egress device not specified"); |
| 3547 |
goto out; |
| 3548 |
} else if (dev->flags & IFF_LOOPBACK) { |
| 3549 |
NL_SET_ERR_MSG(extack, |
| 3550 |
"Egress device can not be loopback device for this route"); |
| 3551 |
goto out; |
| 3552 |
} |
| 3553 |
|
| 3554 |
/* if we did not check gw_addr above, do so now that the |
| 3555 |
* egress device has been resolved. |
| 3556 |
*/ |
| 3557 |
if (need_addr_check && |
| 3558 |
ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) { |
| 3559 |
NL_SET_ERR_MSG(extack, "Gateway can not be a local address"); |
| 3560 |
goto out; |
| 3561 |
} |
| 3562 |
|
| 3563 |
err = 0; |
| 3564 |
out: |
| 3565 |
return err; |
| 3566 |
} |
| 3567 |
|
| 3568 |
static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type) |
| 3569 |
{ |
| 3570 |
if ((flags & RTF_REJECT) || |
| 3571 |
(dev && (dev->flags & IFF_LOOPBACK) && |
| 3572 |
!(addr_type & IPV6_ADDR_LOOPBACK) && |
| 3573 |
!(flags & (RTF_ANYCAST | RTF_LOCAL)))) |
| 3574 |
return true; |
| 3575 |
|
| 3576 |
return false; |
| 3577 |
} |
| 3578 |
|
| 3579 |
int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh, |
| 3580 |
struct fib6_config *cfg, gfp_t gfp_flags, |
| 3581 |
struct netlink_ext_ack *extack) |
| 3582 |
{ |
| 3583 |
netdevice_tracker *dev_tracker = &fib6_nh->fib_nh_dev_tracker; |
| 3584 |
struct net_device *dev = NULL; |
| 3585 |
struct inet6_dev *idev = NULL; |
| 3586 |
int err; |
| 3587 |
|
| 3588 |
fib6_nh->fib_nh_family = AF_INET6; |
| 3589 |
#ifdef CONFIG_IPV6_ROUTER_PREF |
| 3590 |
fib6_nh->last_probe = jiffies; |
| 3591 |
#endif |
| 3592 |
if (cfg->fc_is_fdb) { |
| 3593 |
fib6_nh->fib_nh_gw6 = cfg->fc_gateway; |
| 3594 |
fib6_nh->fib_nh_gw_family = AF_INET6; |
| 3595 |
return 0; |
| 3596 |
} |
| 3597 |
|
| 3598 |
err = -ENODEV; |
| 3599 |
if (cfg->fc_ifindex) { |
| 3600 |
dev = netdev_get_by_index(net, cfg->fc_ifindex, |
| 3601 |
dev_tracker, gfp_flags); |
| 3602 |
if (!dev) |
| 3603 |
goto out; |
| 3604 |
idev = in6_dev_get(dev); |
| 3605 |
if (!idev) |
| 3606 |
goto out; |
| 3607 |
} |
| 3608 |
|
| 3609 |
if (cfg->fc_flags & RTNH_F_ONLINK) { |
| 3610 |
if (!dev) { |
| 3611 |
NL_SET_ERR_MSG(extack, |
| 3612 |
"Nexthop device required for onlink"); |
| 3613 |
goto out; |
| 3614 |
} |
| 3615 |
|
| 3616 |
if (!(dev->flags & IFF_UP)) { |
| 3617 |
NL_SET_ERR_MSG(extack, "Nexthop device is not up"); |
| 3618 |
err = -ENETDOWN; |
| 3619 |
goto out; |
| 3620 |
} |
| 3621 |
|
| 3622 |
fib6_nh->fib_nh_flags |= RTNH_F_ONLINK; |
| 3623 |
} |
| 3624 |
|
| 3625 |
fib6_nh->fib_nh_weight = 1; |
| 3626 |
|
| 3627 |
/* Reset the nexthop device to the loopback device in case of reject |
| 3628 |
* routes. |
| 3629 |
*/ |
| 3630 |
if (cfg->fc_flags & RTF_REJECT) { |
| 3631 |
/* hold loopback dev/idev if we haven't done so. */ |
| 3632 |
if (dev != net->loopback_dev) { |
| 3633 |
if (dev) { |
| 3634 |
netdev_put(dev, dev_tracker); |
| 3635 |
in6_dev_put(idev); |
| 3636 |
} |
| 3637 |
dev = net->loopback_dev; |
| 3638 |
netdev_hold(dev, dev_tracker, gfp_flags); |
| 3639 |
idev = in6_dev_get(dev); |
| 3640 |
if (!idev) { |
| 3641 |
err = -ENODEV; |
| 3642 |
goto out; |
| 3643 |
} |
| 3644 |
} |
| 3645 |
goto pcpu_alloc; |
| 3646 |
} |
| 3647 |
|
| 3648 |
if (cfg->fc_flags & RTF_GATEWAY) { |
| 3649 |
err = ip6_validate_gw(net, cfg, &dev, dev_tracker, |
| 3650 |
&idev, extack); |
| 3651 |
if (err) |
| 3652 |
goto out; |
| 3653 |
|
| 3654 |
fib6_nh->fib_nh_gw6 = cfg->fc_gateway; |
| 3655 |
fib6_nh->fib_nh_gw_family = AF_INET6; |
| 3656 |
} |
| 3657 |
|
| 3658 |
err = -ENODEV; |
| 3659 |
if (!dev) |
| 3660 |
goto out; |
| 3661 |
|
| 3662 |
if (!idev || idev->cnf.disable_ipv6) { |
| 3663 |
NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device"); |
| 3664 |
err = -EACCES; |
| 3665 |
goto out; |
| 3666 |
} |
| 3667 |
|
| 3668 |
if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) { |
| 3669 |
NL_SET_ERR_MSG(extack, "Nexthop device is not up"); |
| 3670 |
err = -ENETDOWN; |
| 3671 |
goto out; |
| 3672 |
} |
| 3673 |
|
| 3674 |
if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) && |
| 3675 |
!netif_carrier_ok(dev)) |
| 3676 |
fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN; |
| 3677 |
|
| 3678 |
err = fib_nh_common_init(net, &fib6_nh->nh_common, cfg->fc_encap, |
| 3679 |
cfg->fc_encap_type, cfg, gfp_flags, extack); |
| 3680 |
if (err) |
| 3681 |
goto out; |
| 3682 |
|
| 3683 |
pcpu_alloc: |
| 3684 |
fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags); |
| 3685 |
if (!fib6_nh->rt6i_pcpu) { |
| 3686 |
err = -ENOMEM; |
| 3687 |
goto out; |
| 3688 |
} |
| 3689 |
|
| 3690 |
fib6_nh->fib_nh_dev = dev; |
| 3691 |
fib6_nh->fib_nh_oif = dev->ifindex; |
| 3692 |
err = 0; |
| 3693 |
out: |
| 3694 |
if (idev) |
| 3695 |
in6_dev_put(idev); |
| 3696 |
|
| 3697 |
if (err) { |
| 3698 |
fib_nh_common_release(&fib6_nh->nh_common); |
| 3699 |
fib6_nh->nh_common.nhc_pcpu_rth_output = NULL; |
| 3700 |
fib6_nh->fib_nh_lws = NULL; |
| 3701 |
netdev_put(dev, dev_tracker); |
| 3702 |
} |
| 3703 |
|
| 3704 |
return err; |
| 3705 |
} |
| 3706 |
|
| 3707 |
void fib6_nh_release(struct fib6_nh *fib6_nh) |
| 3708 |
{ |
| 3709 |
struct rt6_exception_bucket *bucket; |
| 3710 |
|
| 3711 |
rcu_read_lock(); |
| 3712 |
|
| 3713 |
fib6_nh_flush_exceptions(fib6_nh, NULL); |
| 3714 |
bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL); |
| 3715 |
if (bucket) { |
| 3716 |
rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL); |
| 3717 |
kfree(bucket); |
| 3718 |
} |
| 3719 |
|
| 3720 |
rcu_read_unlock(); |
| 3721 |
|
| 3722 |
fib6_nh_release_dsts(fib6_nh); |
| 3723 |
free_percpu(fib6_nh->rt6i_pcpu); |
| 3724 |
|
| 3725 |
fib_nh_common_release(&fib6_nh->nh_common); |
| 3726 |
} |
| 3727 |
|
| 3728 |
void fib6_nh_release_dsts(struct fib6_nh *fib6_nh) |
| 3729 |
{ |
| 3730 |
int cpu; |
| 3731 |
|
| 3732 |
if (!fib6_nh->rt6i_pcpu) |
| 3733 |
return; |
| 3734 |
|
| 3735 |
for_each_possible_cpu(cpu) { |
| 3736 |
struct rt6_info *pcpu_rt, **ppcpu_rt; |
| 3737 |
|
| 3738 |
ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu); |
| 3739 |
pcpu_rt = xchg(ppcpu_rt, NULL); |
| 3740 |
if (pcpu_rt) { |
| 3741 |
dst_dev_put(&pcpu_rt->dst); |
| 3742 |
dst_release(&pcpu_rt->dst); |
| 3743 |
} |
| 3744 |
} |
| 3745 |
} |
| 3746 |
|
| 3747 |
static int fib6_config_validate(struct fib6_config *cfg, |
| 3748 |
struct netlink_ext_ack *extack) |
| 3749 |
{ |
| 3750 |
/* RTF_PCPU is an internal flag; can not be set by userspace */ |
| 3751 |
if (cfg->fc_flags & RTF_PCPU) { |
| 3752 |
NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU"); |
| 3753 |
goto errout; |
| 3754 |
} |
| 3755 |
|
| 3756 |
/* RTF_CACHE is an internal flag; can not be set by userspace */ |
| 3757 |
if (cfg->fc_flags & RTF_CACHE) { |
| 3758 |
NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE"); |
| 3759 |
goto errout; |
| 3760 |
} |
| 3761 |
|
| 3762 |
if (cfg->fc_type > RTN_MAX) { |
| 3763 |
NL_SET_ERR_MSG(extack, "Invalid route type"); |
| 3764 |
goto errout; |
| 3765 |
} |
| 3766 |
|
| 3767 |
if (cfg->fc_dst_len > 128) { |
| 3768 |
NL_SET_ERR_MSG(extack, "Invalid prefix length"); |
| 3769 |
goto errout; |
| 3770 |
} |
| 3771 |
|
| 3772 |
#ifdef CONFIG_IPV6_SUBTREES |
| 3773 |
if (cfg->fc_src_len > 128) { |
| 3774 |
NL_SET_ERR_MSG(extack, "Invalid source address length"); |
| 3775 |
goto errout; |
| 3776 |
} |
| 3777 |
|
| 3778 |
if (cfg->fc_nh_id && cfg->fc_src_len) { |
| 3779 |
NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing"); |
| 3780 |
goto errout; |
| 3781 |
} |
| 3782 |
#else |
| 3783 |
if (cfg->fc_src_len) { |
| 3784 |
NL_SET_ERR_MSG(extack, |
| 3785 |
"Specifying source address requires IPV6_SUBTREES to be enabled"); |
| 3786 |
goto errout; |
| 3787 |
} |
| 3788 |
#endif |
| 3789 |
return 0; |
| 3790 |
errout: |
| 3791 |
return -EINVAL; |
| 3792 |
} |
| 3793 |
|
| 3794 |
static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg, |
| 3795 |
gfp_t gfp_flags, |
| 3796 |
struct netlink_ext_ack *extack) |
| 3797 |
{ |
| 3798 |
struct net *net = cfg->fc_nlinfo.nl_net; |
| 3799 |
struct fib6_table *table; |
| 3800 |
struct fib6_info *rt; |
| 3801 |
int err; |
| 3802 |
|
| 3803 |
if (cfg->fc_nlinfo.nlh && |
| 3804 |
!(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) { |
| 3805 |
table = fib6_get_table(net, cfg->fc_table); |
| 3806 |
if (!table) { |
| 3807 |
pr_warn("NLM_F_CREATE should be specified when creating new route\n"); |
| 3808 |
table = fib6_new_table(net, cfg->fc_table); |
| 3809 |
} |
| 3810 |
} else { |
| 3811 |
table = fib6_new_table(net, cfg->fc_table); |
| 3812 |
} |
| 3813 |
if (!table) { |
| 3814 |
err = -ENOBUFS; |
| 3815 |
goto err; |
| 3816 |
} |
| 3817 |
|
| 3818 |
rt = fib6_info_alloc(gfp_flags, !cfg->fc_nh_id); |
| 3819 |
if (!rt) { |
| 3820 |
err = -ENOMEM; |
| 3821 |
goto err; |
| 3822 |
} |
| 3823 |
|
| 3824 |
rt->fib6_metrics = ip_fib_metrics_init(cfg->fc_mx, cfg->fc_mx_len, |
| 3825 |
extack); |
| 3826 |
if (IS_ERR(rt->fib6_metrics)) { |
| 3827 |
err = PTR_ERR(rt->fib6_metrics); |
| 3828 |
goto free; |
| 3829 |
} |
| 3830 |
|
| 3831 |
if (cfg->fc_flags & RTF_ADDRCONF) |
| 3832 |
rt->dst_nocount = true; |
| 3833 |
|
| 3834 |
if (cfg->fc_flags & RTF_EXPIRES) |
| 3835 |
fib6_set_expires(rt, jiffies + |
| 3836 |
clock_t_to_jiffies(cfg->fc_expires)); |
| 3837 |
|
| 3838 |
if (cfg->fc_protocol == RTPROT_UNSPEC) |
| 3839 |
cfg->fc_protocol = RTPROT_BOOT; |
| 3840 |
|
| 3841 |
rt->fib6_protocol = cfg->fc_protocol; |
| 3842 |
rt->fib6_table = table; |
| 3843 |
rt->fib6_metric = cfg->fc_metric; |
| 3844 |
rt->fib6_type = cfg->fc_type ? : RTN_UNICAST; |
| 3845 |
rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY; |
| 3846 |
|
| 3847 |
ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len); |
| 3848 |
rt->fib6_dst.plen = cfg->fc_dst_len; |
| 3849 |
|
| 3850 |
#ifdef CONFIG_IPV6_SUBTREES |
| 3851 |
ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len); |
| 3852 |
rt->fib6_src.plen = cfg->fc_src_len; |
| 3853 |
#endif |
| 3854 |
return rt; |
| 3855 |
free: |
| 3856 |
kfree(rt); |
| 3857 |
err: |
| 3858 |
return ERR_PTR(err); |
| 3859 |
} |
| 3860 |
|
| 3861 |
static int ip6_route_info_create_nh(struct fib6_info *rt, |
| 3862 |
struct fib6_config *cfg, |
| 3863 |
gfp_t gfp_flags, |
| 3864 |
struct netlink_ext_ack *extack) |
| 3865 |
{ |
| 3866 |
struct net *net = cfg->fc_nlinfo.nl_net; |
| 3867 |
struct fib6_nh *fib6_nh; |
| 3868 |
int err; |
| 3869 |
|
| 3870 |
if (cfg->fc_nh_id) { |
| 3871 |
struct nexthop *nh; |
| 3872 |
|
| 3873 |
rcu_read_lock(); |
| 3874 |
|
| 3875 |
nh = nexthop_find_by_id(net, cfg->fc_nh_id); |
| 3876 |
if (!nh) { |
| 3877 |
err = -EINVAL; |
| 3878 |
NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); |
| 3879 |
goto out_free; |
| 3880 |
} |
| 3881 |
|
| 3882 |
err = fib6_check_nexthop(nh, cfg, extack); |
| 3883 |
if (err) |
| 3884 |
goto out_free; |
| 3885 |
|
| 3886 |
if (!nexthop_get(nh)) { |
| 3887 |
NL_SET_ERR_MSG(extack, "Nexthop has been deleted"); |
| 3888 |
err = -ENOENT; |
| 3889 |
goto out_free; |
| 3890 |
} |
| 3891 |
|
| 3892 |
rt->nh = nh; |
| 3893 |
fib6_nh = nexthop_fib6_nh(rt->nh); |
| 3894 |
|
| 3895 |
rcu_read_unlock(); |
| 3896 |
} else { |
| 3897 |
int addr_type; |
| 3898 |
|
| 3899 |
err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack); |
| 3900 |
if (err) |
| 3901 |
goto out_release; |
| 3902 |
|
| 3903 |
fib6_nh = rt->fib6_nh; |
| 3904 |
|
| 3905 |
/* We cannot add true routes via loopback here, they would |
| 3906 |
* result in kernel looping; promote them to reject routes |
| 3907 |
*/ |
| 3908 |
addr_type = ipv6_addr_type(&cfg->fc_dst); |
| 3909 |
if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev, |
| 3910 |
addr_type)) |
| 3911 |
rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP; |
| 3912 |
} |
| 3913 |
|
| 3914 |
if (!ipv6_addr_any(&cfg->fc_prefsrc)) { |
| 3915 |
struct net_device *dev = fib6_nh->fib_nh_dev; |
| 3916 |
|
| 3917 |
if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) { |
| 3918 |
NL_SET_ERR_MSG(extack, "Invalid source address"); |
| 3919 |
err = -EINVAL; |
| 3920 |
goto out_release; |
| 3921 |
} |
| 3922 |
rt->fib6_prefsrc.addr = cfg->fc_prefsrc; |
| 3923 |
rt->fib6_prefsrc.plen = 128; |
| 3924 |
} |
| 3925 |
|
| 3926 |
return 0; |
| 3927 |
out_release: |
| 3928 |
fib6_info_release(rt); |
| 3929 |
return err; |
| 3930 |
out_free: |
| 3931 |
rcu_read_unlock(); |
| 3932 |
ip_fib_metrics_put(rt->fib6_metrics); |
| 3933 |
kfree(rt); |
| 3934 |
return err; |
| 3935 |
} |
| 3936 |
|
| 3937 |
int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags, |
| 3938 |
struct netlink_ext_ack *extack) |
| 3939 |
{ |
| 3940 |
struct fib6_info *rt; |
| 3941 |
int err; |
| 3942 |
|
| 3943 |
err = fib6_config_validate(cfg, extack); |
| 3944 |
if (err) |
| 3945 |
return err; |
| 3946 |
|
| 3947 |
rt = ip6_route_info_create(cfg, gfp_flags, extack); |
| 3948 |
if (IS_ERR(rt)) |
| 3949 |
return PTR_ERR(rt); |
| 3950 |
|
| 3951 |
err = ip6_route_info_create_nh(rt, cfg, gfp_flags, extack); |
| 3952 |
if (err) |
| 3953 |
return err; |
| 3954 |
|
| 3955 |
err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack); |
| 3956 |
fib6_info_release(rt); |
| 3957 |
|
| 3958 |
return err; |
| 3959 |
} |
| 3960 |
|
| 3961 |
static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info) |
| 3962 |
{ |
| 3963 |
struct net *net = info->nl_net; |
| 3964 |
struct fib6_table *table; |
| 3965 |
int err; |
| 3966 |
|
| 3967 |
if (rt == net->ipv6.fib6_null_entry) { |
| 3968 |
err = -ENOENT; |
| 3969 |
goto out; |
| 3970 |
} |
| 3971 |
|
| 3972 |
table = rt->fib6_table; |
| 3973 |
spin_lock_bh(&table->tb6_lock); |
| 3974 |
err = fib6_del(rt, info); |
| 3975 |
spin_unlock_bh(&table->tb6_lock); |
| 3976 |
|
| 3977 |
out: |
| 3978 |
fib6_info_release(rt); |
| 3979 |
return err; |
| 3980 |
} |
| 3981 |
|
| 3982 |
int ip6_del_rt(struct net *net, struct fib6_info *rt, bool skip_notify) |
| 3983 |
{ |
| 3984 |
struct nl_info info = { |
| 3985 |
.nl_net = net, |
| 3986 |
.skip_notify = skip_notify |
| 3987 |
}; |
| 3988 |
|
| 3989 |
return __ip6_del_rt(rt, &info); |
| 3990 |
} |
| 3991 |
|
| 3992 |
static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg) |
| 3993 |
{ |
| 3994 |
struct nl_info *info = &cfg->fc_nlinfo; |
| 3995 |
struct net *net = info->nl_net; |
| 3996 |
struct sk_buff *skb = NULL; |
| 3997 |
struct fib6_table *table; |
| 3998 |
int err = -ENOENT; |
| 3999 |
|
| 4000 |
if (rt == net->ipv6.fib6_null_entry) |
| 4001 |
goto out_put; |
| 4002 |
table = rt->fib6_table; |
| 4003 |
spin_lock_bh(&table->tb6_lock); |
| 4004 |
|
| 4005 |
if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) { |
| 4006 |
struct fib6_info *sibling, *next_sibling; |
| 4007 |
struct fib6_node *fn; |
| 4008 |
|
| 4009 |
/* prefer to send a single notification with all hops */ |
| 4010 |
skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any()); |
| 4011 |
if (skb) { |
| 4012 |
u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
| 4013 |
|
| 4014 |
if (rt6_fill_node(net, skb, rt, NULL, |
| 4015 |
NULL, NULL, 0, RTM_DELROUTE, |
| 4016 |
info->portid, seq, 0) < 0) { |
| 4017 |
kfree_skb(skb); |
| 4018 |
skb = NULL; |
| 4019 |
} else |
| 4020 |
info->skip_notify = 1; |
| 4021 |
} |
| 4022 |
|
| 4023 |
/* 'rt' points to the first sibling route. If it is not the |
| 4024 |
* leaf, then we do not need to send a notification. Otherwise, |
| 4025 |
* we need to check if the last sibling has a next route or not |
| 4026 |
* and emit a replace or delete notification, respectively. |
| 4027 |
*/ |
| 4028 |
info->skip_notify_kernel = 1; |
| 4029 |
fn = rcu_dereference_protected(rt->fib6_node, |
| 4030 |
lockdep_is_held(&table->tb6_lock)); |
| 4031 |
if (rcu_access_pointer(fn->leaf) == rt) { |
| 4032 |
struct fib6_info *last_sibling, *replace_rt; |
| 4033 |
|
| 4034 |
last_sibling = list_last_entry(&rt->fib6_siblings, |
| 4035 |
struct fib6_info, |
| 4036 |
fib6_siblings); |
| 4037 |
replace_rt = rcu_dereference_protected( |
| 4038 |
last_sibling->fib6_next, |
| 4039 |
lockdep_is_held(&table->tb6_lock)); |
| 4040 |
if (replace_rt) |
| 4041 |
call_fib6_entry_notifiers_replace(net, |
| 4042 |
replace_rt); |
| 4043 |
else |
| 4044 |
call_fib6_multipath_entry_notifiers(net, |
| 4045 |
FIB_EVENT_ENTRY_DEL, |
| 4046 |
rt, rt->fib6_nsiblings, |
| 4047 |
NULL); |
| 4048 |
} |
| 4049 |
list_for_each_entry_safe(sibling, next_sibling, |
| 4050 |
&rt->fib6_siblings, |
| 4051 |
fib6_siblings) { |
| 4052 |
err = fib6_del(sibling, info); |
| 4053 |
if (err) |
| 4054 |
goto out_unlock; |
| 4055 |
} |
| 4056 |
} |
| 4057 |
|
| 4058 |
err = fib6_del(rt, info); |
| 4059 |
out_unlock: |
| 4060 |
spin_unlock_bh(&table->tb6_lock); |
| 4061 |
out_put: |
| 4062 |
fib6_info_release(rt); |
| 4063 |
|
| 4064 |
if (skb) { |
| 4065 |
rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE, |
| 4066 |
info->nlh, gfp_any()); |
| 4067 |
} |
| 4068 |
return err; |
| 4069 |
} |
| 4070 |
|
| 4071 |
static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg) |
| 4072 |
{ |
| 4073 |
int rc = -ESRCH; |
| 4074 |
|
| 4075 |
if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex) |
| 4076 |
goto out; |
| 4077 |
|
| 4078 |
if (cfg->fc_flags & RTF_GATEWAY && |
| 4079 |
!ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway)) |
| 4080 |
goto out; |
| 4081 |
|
| 4082 |
rc = rt6_remove_exception_rt(rt); |
| 4083 |
out: |
| 4084 |
return rc; |
| 4085 |
} |
| 4086 |
|
| 4087 |
static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt, |
| 4088 |
struct fib6_nh *nh) |
| 4089 |
{ |
| 4090 |
struct fib6_result res = { |
| 4091 |
.f6i = rt, |
| 4092 |
.nh = nh, |
| 4093 |
}; |
| 4094 |
struct rt6_info *rt_cache; |
| 4095 |
|
| 4096 |
rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src); |
| 4097 |
if (rt_cache) |
| 4098 |
return __ip6_del_cached_rt(rt_cache, cfg); |
| 4099 |
|
| 4100 |
return 0; |
| 4101 |
} |
| 4102 |
|
| 4103 |
struct fib6_nh_del_cached_rt_arg { |
| 4104 |
struct fib6_config *cfg; |
| 4105 |
struct fib6_info *f6i; |
| 4106 |
}; |
| 4107 |
|
| 4108 |
static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg) |
| 4109 |
{ |
| 4110 |
struct fib6_nh_del_cached_rt_arg *arg = _arg; |
| 4111 |
int rc; |
| 4112 |
|
| 4113 |
rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh); |
| 4114 |
return rc != -ESRCH ? rc : 0; |
| 4115 |
} |
| 4116 |
|
| 4117 |
static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i) |
| 4118 |
{ |
| 4119 |
struct fib6_nh_del_cached_rt_arg arg = { |
| 4120 |
.cfg = cfg, |
| 4121 |
.f6i = f6i |
| 4122 |
}; |
| 4123 |
|
| 4124 |
return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg); |
| 4125 |
} |
| 4126 |
|
| 4127 |
static int ip6_route_del(struct fib6_config *cfg, |
| 4128 |
struct netlink_ext_ack *extack) |
| 4129 |
{ |
| 4130 |
struct fib6_table *table; |
| 4131 |
struct fib6_info *rt; |
| 4132 |
struct fib6_node *fn; |
| 4133 |
int err = -ESRCH; |
| 4134 |
|
| 4135 |
table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table); |
| 4136 |
if (!table) { |
| 4137 |
NL_SET_ERR_MSG(extack, "FIB table does not exist"); |
| 4138 |
return err; |
| 4139 |
} |
| 4140 |
|
| 4141 |
rcu_read_lock(); |
| 4142 |
|
| 4143 |
fn = fib6_locate(&table->tb6_root, |
| 4144 |
&cfg->fc_dst, cfg->fc_dst_len, |
| 4145 |
&cfg->fc_src, cfg->fc_src_len, |
| 4146 |
!(cfg->fc_flags & RTF_CACHE)); |
| 4147 |
|
| 4148 |
if (fn) { |
| 4149 |
for_each_fib6_node_rt_rcu(fn) { |
| 4150 |
struct fib6_nh *nh; |
| 4151 |
|
| 4152 |
if (rt->nh && cfg->fc_nh_id && |
| 4153 |
rt->nh->id != cfg->fc_nh_id) |
| 4154 |
continue; |
| 4155 |
|
| 4156 |
if (cfg->fc_flags & RTF_CACHE) { |
| 4157 |
int rc = 0; |
| 4158 |
|
| 4159 |
if (rt->nh) { |
| 4160 |
rc = ip6_del_cached_rt_nh(cfg, rt); |
| 4161 |
} else if (cfg->fc_nh_id) { |
| 4162 |
continue; |
| 4163 |
} else { |
| 4164 |
nh = rt->fib6_nh; |
| 4165 |
rc = ip6_del_cached_rt(cfg, rt, nh); |
| 4166 |
} |
| 4167 |
if (rc != -ESRCH) { |
| 4168 |
rcu_read_unlock(); |
| 4169 |
return rc; |
| 4170 |
} |
| 4171 |
continue; |
| 4172 |
} |
| 4173 |
|
| 4174 |
if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric) |
| 4175 |
continue; |
| 4176 |
if (cfg->fc_protocol && |
| 4177 |
cfg->fc_protocol != rt->fib6_protocol) |
| 4178 |
continue; |
| 4179 |
|
| 4180 |
if (rt->nh) { |
| 4181 |
if (!fib6_info_hold_safe(rt)) |
| 4182 |
continue; |
| 4183 |
|
| 4184 |
err = __ip6_del_rt(rt, &cfg->fc_nlinfo); |
| 4185 |
break; |
| 4186 |
} |
| 4187 |
if (cfg->fc_nh_id) |
| 4188 |
continue; |
| 4189 |
|
| 4190 |
nh = rt->fib6_nh; |
| 4191 |
if (cfg->fc_ifindex && |
| 4192 |
(!nh->fib_nh_dev || |
| 4193 |
nh->fib_nh_dev->ifindex != cfg->fc_ifindex)) |
| 4194 |
continue; |
| 4195 |
if (cfg->fc_flags & RTF_GATEWAY && |
| 4196 |
!ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6)) |
| 4197 |
continue; |
| 4198 |
if (!fib6_info_hold_safe(rt)) |
| 4199 |
continue; |
| 4200 |
|
| 4201 |
/* if gateway was specified only delete the one hop */ |
| 4202 |
if (cfg->fc_flags & RTF_GATEWAY) |
| 4203 |
err = __ip6_del_rt(rt, &cfg->fc_nlinfo); |
| 4204 |
else |
| 4205 |
err = __ip6_del_rt_siblings(rt, cfg); |
| 4206 |
break; |
| 4207 |
} |
| 4208 |
} |
| 4209 |
rcu_read_unlock(); |
| 4210 |
|
| 4211 |
return err; |
| 4212 |
} |
| 4213 |
|
| 4214 |
static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb) |
| 4215 |
{ |
| 4216 |
struct netevent_redirect netevent; |
| 4217 |
struct rt6_info *rt, *nrt = NULL; |
| 4218 |
struct fib6_result res = {}; |
| 4219 |
struct ndisc_options ndopts; |
| 4220 |
struct inet6_dev *in6_dev; |
| 4221 |
struct neighbour *neigh; |
| 4222 |
struct rd_msg *msg; |
| 4223 |
int optlen, on_link; |
| 4224 |
u8 *lladdr; |
| 4225 |
|
| 4226 |
optlen = skb_tail_pointer(skb) - skb_transport_header(skb); |
| 4227 |
optlen -= sizeof(*msg); |
| 4228 |
|
| 4229 |
if (optlen < 0) { |
| 4230 |
net_dbg_ratelimited("rt6_do_redirect: packet too short\n"); |
| 4231 |
return; |
| 4232 |
} |
| 4233 |
|
| 4234 |
msg = (struct rd_msg *)icmp6_hdr(skb); |
| 4235 |
|
| 4236 |
if (ipv6_addr_is_multicast(&msg->dest)) { |
| 4237 |
net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n"); |
| 4238 |
return; |
| 4239 |
} |
| 4240 |
|
| 4241 |
on_link = 0; |
| 4242 |
if (ipv6_addr_equal(&msg->dest, &msg->target)) { |
| 4243 |
on_link = 1; |
| 4244 |
} else if (ipv6_addr_type(&msg->target) != |
| 4245 |
(IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) { |
| 4246 |
net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n"); |
| 4247 |
return; |
| 4248 |
} |
| 4249 |
|
| 4250 |
in6_dev = __in6_dev_get(skb->dev); |
| 4251 |
if (!in6_dev) |
| 4252 |
return; |
| 4253 |
if (READ_ONCE(in6_dev->cnf.forwarding) || |
| 4254 |
!READ_ONCE(in6_dev->cnf.accept_redirects)) |
| 4255 |
return; |
| 4256 |
|
| 4257 |
/* RFC2461 8.1: |
| 4258 |
* The IP source address of the Redirect MUST be the same as the current |
| 4259 |
* first-hop router for the specified ICMP Destination Address. |
| 4260 |
*/ |
| 4261 |
|
| 4262 |
if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) { |
| 4263 |
net_dbg_ratelimited("rt6_redirect: invalid ND options\n"); |
| 4264 |
return; |
| 4265 |
} |
| 4266 |
|
| 4267 |
lladdr = NULL; |
| 4268 |
if (ndopts.nd_opts_tgt_lladdr) { |
| 4269 |
lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, |
| 4270 |
skb->dev); |
| 4271 |
if (!lladdr) { |
| 4272 |
net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n"); |
| 4273 |
return; |
| 4274 |
} |
| 4275 |
} |
| 4276 |
|
| 4277 |
rt = dst_rt6_info(dst); |
| 4278 |
if (rt->rt6i_flags & RTF_REJECT) { |
| 4279 |
net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n"); |
| 4280 |
return; |
| 4281 |
} |
| 4282 |
|
| 4283 |
/* Redirect received -> path was valid. |
| 4284 |
* Look, redirects are sent only in response to data packets, |
| 4285 |
* so that this nexthop apparently is reachable. --ANK |
| 4286 |
*/ |
| 4287 |
dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr); |
| 4288 |
|
| 4289 |
neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1); |
| 4290 |
if (!neigh) |
| 4291 |
return; |
| 4292 |
|
| 4293 |
/* |
| 4294 |
* We have finally decided to accept it. |
| 4295 |
*/ |
| 4296 |
|
| 4297 |
ndisc_update(skb->dev, neigh, lladdr, NUD_STALE, |
| 4298 |
NEIGH_UPDATE_F_WEAK_OVERRIDE| |
| 4299 |
NEIGH_UPDATE_F_OVERRIDE| |
| 4300 |
(on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER| |
| 4301 |
NEIGH_UPDATE_F_ISROUTER)), |
| 4302 |
NDISC_REDIRECT, &ndopts); |
| 4303 |
|
| 4304 |
rcu_read_lock(); |
| 4305 |
res.f6i = rcu_dereference(rt->from); |
| 4306 |
if (!res.f6i) |
| 4307 |
goto out; |
| 4308 |
|
| 4309 |
if (res.f6i->nh) { |
| 4310 |
struct fib6_nh_match_arg arg = { |
| 4311 |
.dev = dst_dev_rcu(dst), |
| 4312 |
.gw = &rt->rt6i_gateway, |
| 4313 |
}; |
| 4314 |
|
| 4315 |
nexthop_for_each_fib6_nh(res.f6i->nh, |
| 4316 |
fib6_nh_find_match, &arg); |
| 4317 |
|
| 4318 |
/* fib6_info uses a nexthop that does not have fib6_nh |
| 4319 |
* using the dst->dev. Should be impossible |
| 4320 |
*/ |
| 4321 |
if (!arg.match) |
| 4322 |
goto out; |
| 4323 |
res.nh = arg.match; |
| 4324 |
} else { |
| 4325 |
res.nh = res.f6i->fib6_nh; |
| 4326 |
} |
| 4327 |
|
| 4328 |
res.fib6_flags = res.f6i->fib6_flags; |
| 4329 |
res.fib6_type = res.f6i->fib6_type; |
| 4330 |
nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL); |
| 4331 |
if (!nrt) |
| 4332 |
goto out; |
| 4333 |
|
| 4334 |
nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE; |
| 4335 |
if (on_link) |
| 4336 |
nrt->rt6i_flags &= ~RTF_GATEWAY; |
| 4337 |
|
| 4338 |
nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key; |
| 4339 |
|
| 4340 |
/* rt6_insert_exception() will take care of duplicated exceptions */ |
| 4341 |
if (rt6_insert_exception(nrt, &res)) { |
| 4342 |
dst_release_immediate(&nrt->dst); |
| 4343 |
goto out; |
| 4344 |
} |
| 4345 |
|
| 4346 |
netevent.old = &rt->dst; |
| 4347 |
netevent.new = &nrt->dst; |
| 4348 |
netevent.daddr = &msg->dest; |
| 4349 |
netevent.neigh = neigh; |
| 4350 |
call_netevent_notifiers(NETEVENT_REDIRECT, &netevent); |
| 4351 |
|
| 4352 |
out: |
| 4353 |
rcu_read_unlock(); |
| 4354 |
neigh_release(neigh); |
| 4355 |
} |
| 4356 |
|
| 4357 |
#ifdef CONFIG_IPV6_ROUTE_INFO |
| 4358 |
static struct fib6_info *rt6_get_route_info(struct net *net, |
| 4359 |
const struct in6_addr *prefix, int prefixlen, |
| 4360 |
const struct in6_addr *gwaddr, |
| 4361 |
struct net_device *dev) |
| 4362 |
{ |
| 4363 |
u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO; |
| 4364 |
int ifindex = dev->ifindex; |
| 4365 |
struct fib6_node *fn; |
| 4366 |
struct fib6_info *rt = NULL; |
| 4367 |
struct fib6_table *table; |
| 4368 |
|
| 4369 |
table = fib6_get_table(net, tb_id); |
| 4370 |
if (!table) |
| 4371 |
return NULL; |
| 4372 |
|
| 4373 |
rcu_read_lock(); |
| 4374 |
fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true); |
| 4375 |
if (!fn) |
| 4376 |
goto out; |
| 4377 |
|
| 4378 |
for_each_fib6_node_rt_rcu(fn) { |
| 4379 |
/* these routes do not use nexthops */ |
| 4380 |
if (rt->nh) |
| 4381 |
continue; |
| 4382 |
if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex) |
| 4383 |
continue; |
| 4384 |
if (!(rt->fib6_flags & RTF_ROUTEINFO) || |
| 4385 |
!rt->fib6_nh->fib_nh_gw_family) |
| 4386 |
continue; |
| 4387 |
if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr)) |
| 4388 |
continue; |
| 4389 |
if (!fib6_info_hold_safe(rt)) |
| 4390 |
continue; |
| 4391 |
break; |
| 4392 |
} |
| 4393 |
out: |
| 4394 |
rcu_read_unlock(); |
| 4395 |
return rt; |
| 4396 |
} |
| 4397 |
|
| 4398 |
static struct fib6_info *rt6_add_route_info(struct net *net, |
| 4399 |
const struct in6_addr *prefix, int prefixlen, |
| 4400 |
const struct in6_addr *gwaddr, |
| 4401 |
struct net_device *dev, |
| 4402 |
unsigned int pref) |
| 4403 |
{ |
| 4404 |
struct fib6_config cfg = { |
| 4405 |
.fc_metric = IP6_RT_PRIO_USER, |
| 4406 |
.fc_ifindex = dev->ifindex, |
| 4407 |
.fc_dst_len = prefixlen, |
| 4408 |
.fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO | |
| 4409 |
RTF_UP | RTF_PREF(pref), |
| 4410 |
.fc_protocol = RTPROT_RA, |
| 4411 |
.fc_type = RTN_UNICAST, |
| 4412 |
.fc_nlinfo.portid = 0, |
| 4413 |
.fc_nlinfo.nlh = NULL, |
| 4414 |
.fc_nlinfo.nl_net = net, |
| 4415 |
}; |
| 4416 |
|
| 4417 |
cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO; |
| 4418 |
cfg.fc_dst = *prefix; |
| 4419 |
cfg.fc_gateway = *gwaddr; |
| 4420 |
|
| 4421 |
/* We should treat it as a default route if prefix length is 0. */ |
| 4422 |
if (!prefixlen) |
| 4423 |
cfg.fc_flags |= RTF_DEFAULT; |
| 4424 |
|
| 4425 |
ip6_route_add(&cfg, GFP_ATOMIC, NULL); |
| 4426 |
|
| 4427 |
return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev); |
| 4428 |
} |
| 4429 |
#endif |
| 4430 |
|
| 4431 |
struct fib6_info *rt6_get_dflt_router(struct net *net, |
| 4432 |
const struct in6_addr *addr, |
| 4433 |
struct net_device *dev) |
| 4434 |
{ |
| 4435 |
u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT; |
| 4436 |
struct fib6_info *rt; |
| 4437 |
struct fib6_table *table; |
| 4438 |
|
| 4439 |
table = fib6_get_table(net, tb_id); |
| 4440 |
if (!table) |
| 4441 |
return NULL; |
| 4442 |
|
| 4443 |
rcu_read_lock(); |
| 4444 |
for_each_fib6_node_rt_rcu(&table->tb6_root) { |
| 4445 |
struct fib6_nh *nh; |
| 4446 |
|
| 4447 |
/* RA routes do not use nexthops */ |
| 4448 |
if (rt->nh) |
| 4449 |
continue; |
| 4450 |
|
| 4451 |
nh = rt->fib6_nh; |
| 4452 |
if (dev == nh->fib_nh_dev && |
| 4453 |
((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) && |
| 4454 |
ipv6_addr_equal(&nh->fib_nh_gw6, addr)) |
| 4455 |
break; |
| 4456 |
} |
| 4457 |
if (rt && !fib6_info_hold_safe(rt)) |
| 4458 |
rt = NULL; |
| 4459 |
rcu_read_unlock(); |
| 4460 |
return rt; |
| 4461 |
} |
| 4462 |
|
| 4463 |
struct fib6_info *rt6_add_dflt_router(struct net *net, |
| 4464 |
const struct in6_addr *gwaddr, |
| 4465 |
struct net_device *dev, |
| 4466 |
unsigned int pref, |
| 4467 |
u32 defrtr_usr_metric, |
| 4468 |
int lifetime) |
| 4469 |
{ |
| 4470 |
struct fib6_config cfg = { |
| 4471 |
.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT, |
| 4472 |
.fc_metric = defrtr_usr_metric, |
| 4473 |
.fc_ifindex = dev->ifindex, |
| 4474 |
.fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT | |
| 4475 |
RTF_UP | RTF_EXPIRES | RTF_PREF(pref), |
| 4476 |
.fc_protocol = RTPROT_RA, |
| 4477 |
.fc_type = RTN_UNICAST, |
| 4478 |
.fc_nlinfo.portid = 0, |
| 4479 |
.fc_nlinfo.nlh = NULL, |
| 4480 |
.fc_nlinfo.nl_net = net, |
| 4481 |
.fc_expires = jiffies_to_clock_t(lifetime * HZ), |
| 4482 |
}; |
| 4483 |
|
| 4484 |
cfg.fc_gateway = *gwaddr; |
| 4485 |
|
| 4486 |
if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) { |
| 4487 |
struct fib6_table *table; |
| 4488 |
|
| 4489 |
table = fib6_get_table(dev_net(dev), cfg.fc_table); |
| 4490 |
if (table) |
| 4491 |
table->flags |= RT6_TABLE_HAS_DFLT_ROUTER; |
| 4492 |
} |
| 4493 |
|
| 4494 |
return rt6_get_dflt_router(net, gwaddr, dev); |
| 4495 |
} |
| 4496 |
|
| 4497 |
static void __rt6_purge_dflt_routers(struct net *net, |
| 4498 |
struct fib6_table *table) |
| 4499 |
{ |
| 4500 |
struct fib6_info *rt; |
| 4501 |
|
| 4502 |
restart: |
| 4503 |
rcu_read_lock(); |
| 4504 |
for_each_fib6_node_rt_rcu(&table->tb6_root) { |
| 4505 |
struct net_device *dev = fib6_info_nh_dev(rt); |
| 4506 |
struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL; |
| 4507 |
|
| 4508 |
if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) && |
| 4509 |
(!idev || idev->cnf.accept_ra != 2) && |
| 4510 |
fib6_info_hold_safe(rt)) { |
| 4511 |
rcu_read_unlock(); |
| 4512 |
ip6_del_rt(net, rt, false); |
| 4513 |
goto restart; |
| 4514 |
} |
| 4515 |
} |
| 4516 |
rcu_read_unlock(); |
| 4517 |
|
| 4518 |
table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER; |
| 4519 |
} |
| 4520 |
|
| 4521 |
void rt6_purge_dflt_routers(struct net *net) |
| 4522 |
{ |
| 4523 |
struct fib6_table *table; |
| 4524 |
struct hlist_head *head; |
| 4525 |
unsigned int h; |
| 4526 |
|
| 4527 |
rcu_read_lock(); |
| 4528 |
|
| 4529 |
for (h = 0; h < FIB6_TABLE_HASHSZ; h++) { |
| 4530 |
head = &net->ipv6.fib_table_hash[h]; |
| 4531 |
hlist_for_each_entry_rcu(table, head, tb6_hlist) { |
| 4532 |
if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER) |
| 4533 |
__rt6_purge_dflt_routers(net, table); |
| 4534 |
} |
| 4535 |
} |
| 4536 |
|
| 4537 |
rcu_read_unlock(); |
| 4538 |
} |
| 4539 |
|
| 4540 |
static void rtmsg_to_fib6_config(struct net *net, |
| 4541 |
struct in6_rtmsg *rtmsg, |
| 4542 |
struct fib6_config *cfg) |
| 4543 |
{ |
| 4544 |
*cfg = (struct fib6_config){ |
| 4545 |
.fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ? |
| 4546 |
: RT6_TABLE_MAIN, |
| 4547 |
.fc_ifindex = rtmsg->rtmsg_ifindex, |
| 4548 |
.fc_metric = rtmsg->rtmsg_metric, |
| 4549 |
.fc_expires = rtmsg->rtmsg_info, |
| 4550 |
.fc_dst_len = rtmsg->rtmsg_dst_len, |
| 4551 |
.fc_src_len = rtmsg->rtmsg_src_len, |
| 4552 |
.fc_flags = rtmsg->rtmsg_flags, |
| 4553 |
.fc_type = rtmsg->rtmsg_type, |
| 4554 |
|
| 4555 |
.fc_nlinfo.nl_net = net, |
| 4556 |
|
| 4557 |
.fc_dst = rtmsg->rtmsg_dst, |
| 4558 |
.fc_src = rtmsg->rtmsg_src, |
| 4559 |
.fc_gateway = rtmsg->rtmsg_gateway, |
| 4560 |
}; |
| 4561 |
} |
| 4562 |
|
| 4563 |
int ipv6_route_ioctl(struct net *net, unsigned int cmd, struct in6_rtmsg *rtmsg) |
| 4564 |
{ |
| 4565 |
struct fib6_config cfg; |
| 4566 |
int err; |
| 4567 |
|
| 4568 |
if (cmd != SIOCADDRT && cmd != SIOCDELRT) |
| 4569 |
return -EINVAL; |
| 4570 |
if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) |
| 4571 |
return -EPERM; |
| 4572 |
|
| 4573 |
rtmsg_to_fib6_config(net, rtmsg, &cfg); |
| 4574 |
|
| 4575 |
switch (cmd) { |
| 4576 |
case SIOCADDRT: |
| 4577 |
/* Only do the default setting of fc_metric in route adding */ |
| 4578 |
if (cfg.fc_metric == 0) |
| 4579 |
cfg.fc_metric = IP6_RT_PRIO_USER; |
| 4580 |
err = ip6_route_add(&cfg, GFP_KERNEL, NULL); |
| 4581 |
break; |
| 4582 |
case SIOCDELRT: |
| 4583 |
err = ip6_route_del(&cfg, NULL); |
| 4584 |
break; |
| 4585 |
} |
| 4586 |
|
| 4587 |
return err; |
| 4588 |
} |
| 4589 |
|
| 4590 |
/* |
| 4591 |
* Drop the packet on the floor |
| 4592 |
*/ |
| 4593 |
|
| 4594 |
static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes) |
| 4595 |
{ |
| 4596 |
struct dst_entry *dst = skb_dst(skb); |
| 4597 |
struct net_device *dev = dst_dev(dst); |
| 4598 |
struct net *net = dev_net(dev); |
| 4599 |
struct inet6_dev *idev; |
| 4600 |
SKB_DR(reason); |
| 4601 |
int type; |
| 4602 |
|
| 4603 |
if (netif_is_l3_master(skb->dev) || |
| 4604 |
dev == net->loopback_dev) |
| 4605 |
idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif)); |
| 4606 |
else |
| 4607 |
idev = ip6_dst_idev(dst); |
| 4608 |
|
| 4609 |
switch (ipstats_mib_noroutes) { |
| 4610 |
case IPSTATS_MIB_INNOROUTES: |
| 4611 |
type = ipv6_addr_type(&ipv6_hdr(skb)->daddr); |
| 4612 |
if (type == IPV6_ADDR_ANY) { |
| 4613 |
SKB_DR_SET(reason, IP_INADDRERRORS); |
| 4614 |
IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS); |
| 4615 |
break; |
| 4616 |
} |
| 4617 |
SKB_DR_SET(reason, IP_INNOROUTES); |
| 4618 |
fallthrough; |
| 4619 |
case IPSTATS_MIB_OUTNOROUTES: |
| 4620 |
SKB_DR_OR(reason, IP_OUTNOROUTES); |
| 4621 |
IP6_INC_STATS(net, idev, ipstats_mib_noroutes); |
| 4622 |
break; |
| 4623 |
} |
| 4624 |
|
| 4625 |
/* Start over by dropping the dst for l3mdev case */ |
| 4626 |
if (netif_is_l3_master(skb->dev)) |
| 4627 |
skb_dst_drop(skb); |
| 4628 |
|
| 4629 |
icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0); |
| 4630 |
kfree_skb_reason(skb, reason); |
| 4631 |
return 0; |
| 4632 |
} |
| 4633 |
|
| 4634 |
static int ip6_pkt_discard(struct sk_buff *skb) |
| 4635 |
{ |
| 4636 |
return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES); |
| 4637 |
} |
| 4638 |
|
| 4639 |
static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 4640 |
{ |
| 4641 |
skb->dev = skb_dst_dev(skb); |
| 4642 |
return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES); |
| 4643 |
} |
| 4644 |
|
| 4645 |
static int ip6_pkt_prohibit(struct sk_buff *skb) |
| 4646 |
{ |
| 4647 |
return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES); |
| 4648 |
} |
| 4649 |
|
| 4650 |
static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb) |
| 4651 |
{ |
| 4652 |
skb->dev = skb_dst_dev(skb); |
| 4653 |
return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES); |
| 4654 |
} |
| 4655 |
|
| 4656 |
/* |
| 4657 |
* Allocate a dst for local (unicast / anycast) address. |
| 4658 |
*/ |
| 4659 |
|
| 4660 |
struct fib6_info *addrconf_f6i_alloc(struct net *net, |
| 4661 |
struct inet6_dev *idev, |
| 4662 |
const struct in6_addr *addr, |
| 4663 |
bool anycast, gfp_t gfp_flags, |
| 4664 |
struct netlink_ext_ack *extack) |
| 4665 |
{ |
| 4666 |
struct fib6_config cfg = { |
| 4667 |
.fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL, |
| 4668 |
.fc_ifindex = idev->dev->ifindex, |
| 4669 |
.fc_flags = RTF_UP | RTF_NONEXTHOP, |
| 4670 |
.fc_dst = *addr, |
| 4671 |
.fc_dst_len = 128, |
| 4672 |
.fc_protocol = RTPROT_KERNEL, |
| 4673 |
.fc_nlinfo.nl_net = net, |
| 4674 |
.fc_ignore_dev_down = true, |
| 4675 |
}; |
| 4676 |
struct fib6_info *f6i; |
| 4677 |
int err; |
| 4678 |
|
| 4679 |
if (anycast) { |
| 4680 |
cfg.fc_type = RTN_ANYCAST; |
| 4681 |
cfg.fc_flags |= RTF_ANYCAST; |
| 4682 |
} else { |
| 4683 |
cfg.fc_type = RTN_LOCAL; |
| 4684 |
cfg.fc_flags |= RTF_LOCAL; |
| 4685 |
} |
| 4686 |
|
| 4687 |
f6i = ip6_route_info_create(&cfg, gfp_flags, extack); |
| 4688 |
if (IS_ERR(f6i)) |
| 4689 |
return f6i; |
| 4690 |
|
| 4691 |
err = ip6_route_info_create_nh(f6i, &cfg, gfp_flags, extack); |
| 4692 |
if (err) |
| 4693 |
return ERR_PTR(err); |
| 4694 |
|
| 4695 |
f6i->dst_nocount = true; |
| 4696 |
|
| 4697 |
if (!anycast && |
| 4698 |
(READ_ONCE(net->ipv6.devconf_all->disable_policy) || |
| 4699 |
READ_ONCE(idev->cnf.disable_policy))) |
| 4700 |
f6i->dst_nopolicy = true; |
| 4701 |
|
| 4702 |
return f6i; |
| 4703 |
} |
| 4704 |
|
| 4705 |
/* remove deleted ip from prefsrc entries */ |
| 4706 |
struct arg_dev_net_ip { |
| 4707 |
struct net *net; |
| 4708 |
struct in6_addr *addr; |
| 4709 |
}; |
| 4710 |
|
| 4711 |
static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg) |
| 4712 |
{ |
| 4713 |
struct net *net = ((struct arg_dev_net_ip *)arg)->net; |
| 4714 |
struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr; |
| 4715 |
|
| 4716 |
if (!rt->nh && |
| 4717 |
rt != net->ipv6.fib6_null_entry && |
| 4718 |
ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr) && |
| 4719 |
!ipv6_chk_addr(net, addr, rt->fib6_nh->fib_nh_dev, 0)) { |
| 4720 |
spin_lock_bh(&rt6_exception_lock); |
| 4721 |
/* remove prefsrc entry */ |
| 4722 |
rt->fib6_prefsrc.plen = 0; |
| 4723 |
spin_unlock_bh(&rt6_exception_lock); |
| 4724 |
} |
| 4725 |
return 0; |
| 4726 |
} |
| 4727 |
|
| 4728 |
void rt6_remove_prefsrc(struct inet6_ifaddr *ifp) |
| 4729 |
{ |
| 4730 |
struct net *net = dev_net(ifp->idev->dev); |
| 4731 |
struct arg_dev_net_ip adni = { |
| 4732 |
.net = net, |
| 4733 |
.addr = &ifp->addr, |
| 4734 |
}; |
| 4735 |
fib6_clean_all(net, fib6_remove_prefsrc, &adni); |
| 4736 |
} |
| 4737 |
|
| 4738 |
#define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT) |
| 4739 |
|
| 4740 |
/* Remove routers and update dst entries when gateway turn into host. */ |
| 4741 |
static int fib6_clean_tohost(struct fib6_info *rt, void *arg) |
| 4742 |
{ |
| 4743 |
struct in6_addr *gateway = (struct in6_addr *)arg; |
| 4744 |
struct fib6_nh *nh; |
| 4745 |
|
| 4746 |
/* RA routes do not use nexthops */ |
| 4747 |
if (rt->nh) |
| 4748 |
return 0; |
| 4749 |
|
| 4750 |
nh = rt->fib6_nh; |
| 4751 |
if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) && |
| 4752 |
nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6)) |
| 4753 |
return -1; |
| 4754 |
|
| 4755 |
/* Further clean up cached routes in exception table. |
| 4756 |
* This is needed because cached route may have a different |
| 4757 |
* gateway than its 'parent' in the case of an ip redirect. |
| 4758 |
*/ |
| 4759 |
fib6_nh_exceptions_clean_tohost(nh, gateway); |
| 4760 |
|
| 4761 |
return 0; |
| 4762 |
} |
| 4763 |
|
| 4764 |
void rt6_clean_tohost(struct net *net, struct in6_addr *gateway) |
| 4765 |
{ |
| 4766 |
fib6_clean_all(net, fib6_clean_tohost, gateway); |
| 4767 |
} |
| 4768 |
|
| 4769 |
struct arg_netdev_event { |
| 4770 |
const struct net_device *dev; |
| 4771 |
union { |
| 4772 |
unsigned char nh_flags; |
| 4773 |
unsigned long event; |
| 4774 |
}; |
| 4775 |
}; |
| 4776 |
|
| 4777 |
static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt) |
| 4778 |
{ |
| 4779 |
struct fib6_info *iter; |
| 4780 |
struct fib6_node *fn; |
| 4781 |
|
| 4782 |
fn = rcu_dereference_protected(rt->fib6_node, |
| 4783 |
lockdep_is_held(&rt->fib6_table->tb6_lock)); |
| 4784 |
iter = rcu_dereference_protected(fn->leaf, |
| 4785 |
lockdep_is_held(&rt->fib6_table->tb6_lock)); |
| 4786 |
while (iter) { |
| 4787 |
if (iter->fib6_metric == rt->fib6_metric && |
| 4788 |
rt6_qualify_for_ecmp(iter)) |
| 4789 |
return iter; |
| 4790 |
iter = rcu_dereference_protected(iter->fib6_next, |
| 4791 |
lockdep_is_held(&rt->fib6_table->tb6_lock)); |
| 4792 |
} |
| 4793 |
|
| 4794 |
return NULL; |
| 4795 |
} |
| 4796 |
|
| 4797 |
/* only called for fib entries with builtin fib6_nh */ |
| 4798 |
static bool rt6_is_dead(const struct fib6_info *rt) |
| 4799 |
{ |
| 4800 |
if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD || |
| 4801 |
(rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN && |
| 4802 |
ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev))) |
| 4803 |
return true; |
| 4804 |
|
| 4805 |
return false; |
| 4806 |
} |
| 4807 |
|
| 4808 |
static int rt6_multipath_total_weight(const struct fib6_info *rt) |
| 4809 |
{ |
| 4810 |
struct fib6_info *iter; |
| 4811 |
int total = 0; |
| 4812 |
|
| 4813 |
if (!rt6_is_dead(rt)) |
| 4814 |
total += rt->fib6_nh->fib_nh_weight; |
| 4815 |
|
| 4816 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) { |
| 4817 |
if (!rt6_is_dead(iter)) |
| 4818 |
total += iter->fib6_nh->fib_nh_weight; |
| 4819 |
} |
| 4820 |
|
| 4821 |
return total; |
| 4822 |
} |
| 4823 |
|
| 4824 |
static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total) |
| 4825 |
{ |
| 4826 |
int upper_bound = -1; |
| 4827 |
|
| 4828 |
if (!rt6_is_dead(rt)) { |
| 4829 |
*weight += rt->fib6_nh->fib_nh_weight; |
| 4830 |
upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31, |
| 4831 |
total) - 1; |
| 4832 |
} |
| 4833 |
atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound); |
| 4834 |
} |
| 4835 |
|
| 4836 |
static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total) |
| 4837 |
{ |
| 4838 |
struct fib6_info *iter; |
| 4839 |
int weight = 0; |
| 4840 |
|
| 4841 |
rt6_upper_bound_set(rt, &weight, total); |
| 4842 |
|
| 4843 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) |
| 4844 |
rt6_upper_bound_set(iter, &weight, total); |
| 4845 |
} |
| 4846 |
|
| 4847 |
void rt6_multipath_rebalance(struct fib6_info *rt) |
| 4848 |
{ |
| 4849 |
struct fib6_info *first; |
| 4850 |
int total; |
| 4851 |
|
| 4852 |
/* In case the entire multipath route was marked for flushing, |
| 4853 |
* then there is no need to rebalance upon the removal of every |
| 4854 |
* sibling route. |
| 4855 |
*/ |
| 4856 |
if (!rt->fib6_nsiblings || rt->should_flush) |
| 4857 |
return; |
| 4858 |
|
| 4859 |
/* During lookup routes are evaluated in order, so we need to |
| 4860 |
* make sure upper bounds are assigned from the first sibling |
| 4861 |
* onwards. |
| 4862 |
*/ |
| 4863 |
first = rt6_multipath_first_sibling(rt); |
| 4864 |
if (WARN_ON_ONCE(!first)) |
| 4865 |
return; |
| 4866 |
|
| 4867 |
total = rt6_multipath_total_weight(first); |
| 4868 |
rt6_multipath_upper_bound_set(first, total); |
| 4869 |
} |
| 4870 |
|
| 4871 |
static int fib6_ifup(struct fib6_info *rt, void *p_arg) |
| 4872 |
{ |
| 4873 |
const struct arg_netdev_event *arg = p_arg; |
| 4874 |
struct net *net = dev_net(arg->dev); |
| 4875 |
|
| 4876 |
if (rt != net->ipv6.fib6_null_entry && !rt->nh && |
| 4877 |
rt->fib6_nh->fib_nh_dev == arg->dev) { |
| 4878 |
rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags; |
| 4879 |
fib6_update_sernum_upto_root(net, rt); |
| 4880 |
rt6_multipath_rebalance(rt); |
| 4881 |
} |
| 4882 |
|
| 4883 |
return 0; |
| 4884 |
} |
| 4885 |
|
| 4886 |
void rt6_sync_up(struct net_device *dev, unsigned char nh_flags) |
| 4887 |
{ |
| 4888 |
struct arg_netdev_event arg = { |
| 4889 |
.dev = dev, |
| 4890 |
{ |
| 4891 |
.nh_flags = nh_flags, |
| 4892 |
}, |
| 4893 |
}; |
| 4894 |
|
| 4895 |
if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev)) |
| 4896 |
arg.nh_flags |= RTNH_F_LINKDOWN; |
| 4897 |
|
| 4898 |
fib6_clean_all(dev_net(dev), fib6_ifup, &arg); |
| 4899 |
} |
| 4900 |
|
| 4901 |
/* only called for fib entries with inline fib6_nh */ |
| 4902 |
static bool rt6_multipath_uses_dev(const struct fib6_info *rt, |
| 4903 |
const struct net_device *dev) |
| 4904 |
{ |
| 4905 |
struct fib6_info *iter; |
| 4906 |
|
| 4907 |
if (rt->fib6_nh->fib_nh_dev == dev) |
| 4908 |
return true; |
| 4909 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) |
| 4910 |
if (iter->fib6_nh->fib_nh_dev == dev) |
| 4911 |
return true; |
| 4912 |
|
| 4913 |
return false; |
| 4914 |
} |
| 4915 |
|
| 4916 |
static void rt6_multipath_flush(struct fib6_info *rt) |
| 4917 |
{ |
| 4918 |
struct fib6_info *iter; |
| 4919 |
|
| 4920 |
rt->should_flush = 1; |
| 4921 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) |
| 4922 |
iter->should_flush = 1; |
| 4923 |
} |
| 4924 |
|
| 4925 |
static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt, |
| 4926 |
const struct net_device *down_dev) |
| 4927 |
{ |
| 4928 |
struct fib6_info *iter; |
| 4929 |
unsigned int dead = 0; |
| 4930 |
|
| 4931 |
if (rt->fib6_nh->fib_nh_dev == down_dev || |
| 4932 |
rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD) |
| 4933 |
dead++; |
| 4934 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) |
| 4935 |
if (iter->fib6_nh->fib_nh_dev == down_dev || |
| 4936 |
iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD) |
| 4937 |
dead++; |
| 4938 |
|
| 4939 |
return dead; |
| 4940 |
} |
| 4941 |
|
| 4942 |
static void rt6_multipath_nh_flags_set(struct fib6_info *rt, |
| 4943 |
const struct net_device *dev, |
| 4944 |
unsigned char nh_flags) |
| 4945 |
{ |
| 4946 |
struct fib6_info *iter; |
| 4947 |
|
| 4948 |
if (rt->fib6_nh->fib_nh_dev == dev) |
| 4949 |
rt->fib6_nh->fib_nh_flags |= nh_flags; |
| 4950 |
list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) |
| 4951 |
if (iter->fib6_nh->fib_nh_dev == dev) |
| 4952 |
iter->fib6_nh->fib_nh_flags |= nh_flags; |
| 4953 |
} |
| 4954 |
|
| 4955 |
/* called with write lock held for table with rt */ |
| 4956 |
static int fib6_ifdown(struct fib6_info *rt, void *p_arg) |
| 4957 |
{ |
| 4958 |
const struct arg_netdev_event *arg = p_arg; |
| 4959 |
const struct net_device *dev = arg->dev; |
| 4960 |
struct net *net = dev_net(dev); |
| 4961 |
|
| 4962 |
if (rt == net->ipv6.fib6_null_entry || rt->nh) |
| 4963 |
return 0; |
| 4964 |
|
| 4965 |
switch (arg->event) { |
| 4966 |
case NETDEV_UNREGISTER: |
| 4967 |
return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0; |
| 4968 |
case NETDEV_DOWN: |
| 4969 |
if (rt->should_flush) |
| 4970 |
return -1; |
| 4971 |
if (!rt->fib6_nsiblings) |
| 4972 |
return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0; |
| 4973 |
if (rt6_multipath_uses_dev(rt, dev)) { |
| 4974 |
unsigned int count; |
| 4975 |
|
| 4976 |
count = rt6_multipath_dead_count(rt, dev); |
| 4977 |
if (rt->fib6_nsiblings + 1 == count) { |
| 4978 |
rt6_multipath_flush(rt); |
| 4979 |
return -1; |
| 4980 |
} |
| 4981 |
rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD | |
| 4982 |
RTNH_F_LINKDOWN); |
| 4983 |
fib6_update_sernum(net, rt); |
| 4984 |
rt6_multipath_rebalance(rt); |
| 4985 |
} |
| 4986 |
return -2; |
| 4987 |
case NETDEV_CHANGE: |
| 4988 |
if (rt->fib6_nh->fib_nh_dev != dev || |
| 4989 |
rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) |
| 4990 |
break; |
| 4991 |
rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN; |
| 4992 |
rt6_multipath_rebalance(rt); |
| 4993 |
break; |
| 4994 |
} |
| 4995 |
|
| 4996 |
return 0; |
| 4997 |
} |
| 4998 |
|
| 4999 |
void rt6_sync_down_dev(struct net_device *dev, unsigned long event) |
| 5000 |
{ |
| 5001 |
struct arg_netdev_event arg = { |
| 5002 |
.dev = dev, |
| 5003 |
{ |
| 5004 |
.event = event, |
| 5005 |
}, |
| 5006 |
}; |
| 5007 |
struct net *net = dev_net(dev); |
| 5008 |
|
| 5009 |
if (READ_ONCE(net->ipv6.sysctl.skip_notify_on_dev_down)) |
| 5010 |
fib6_clean_all_skip_notify(net, fib6_ifdown, &arg); |
| 5011 |
else |
| 5012 |
fib6_clean_all(net, fib6_ifdown, &arg); |
| 5013 |
} |
| 5014 |
|
| 5015 |
void rt6_disable_ip(struct net_device *dev, unsigned long event) |
| 5016 |
{ |
| 5017 |
rt6_sync_down_dev(dev, event); |
| 5018 |
rt6_uncached_list_flush_dev(dev); |
| 5019 |
neigh_ifdown(&nd_tbl, dev); |
| 5020 |
} |
| 5021 |
|
| 5022 |
struct rt6_mtu_change_arg { |
| 5023 |
struct net_device *dev; |
| 5024 |
unsigned int mtu; |
| 5025 |
struct fib6_info *f6i; |
| 5026 |
}; |
| 5027 |
|
| 5028 |
static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg) |
| 5029 |
{ |
| 5030 |
struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg; |
| 5031 |
struct fib6_info *f6i = arg->f6i; |
| 5032 |
|
| 5033 |
/* For administrative MTU increase, there is no way to discover |
| 5034 |
* IPv6 PMTU increase, so PMTU increase should be updated here. |
| 5035 |
* Since RFC 1981 doesn't include administrative MTU increase |
| 5036 |
* update PMTU increase is a MUST. (i.e. jumbo frame) |
| 5037 |
*/ |
| 5038 |
if (nh->fib_nh_dev == arg->dev) { |
| 5039 |
struct inet6_dev *idev = __in6_dev_get(arg->dev); |
| 5040 |
u32 mtu = f6i->fib6_pmtu; |
| 5041 |
|
| 5042 |
if (mtu >= arg->mtu || |
| 5043 |
(mtu < arg->mtu && mtu == idev->cnf.mtu6)) |
| 5044 |
fib6_metric_set(f6i, RTAX_MTU, arg->mtu); |
| 5045 |
|
| 5046 |
spin_lock_bh(&rt6_exception_lock); |
| 5047 |
rt6_exceptions_update_pmtu(idev, nh, arg->mtu); |
| 5048 |
spin_unlock_bh(&rt6_exception_lock); |
| 5049 |
} |
| 5050 |
|
| 5051 |
return 0; |
| 5052 |
} |
| 5053 |
|
| 5054 |
static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg) |
| 5055 |
{ |
| 5056 |
struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg; |
| 5057 |
struct inet6_dev *idev; |
| 5058 |
|
| 5059 |
/* In IPv6 pmtu discovery is not optional, |
| 5060 |
so that RTAX_MTU lock cannot disable it. |
| 5061 |
We still use this lock to block changes |
| 5062 |
caused by addrconf/ndisc. |
| 5063 |
*/ |
| 5064 |
|
| 5065 |
idev = __in6_dev_get(arg->dev); |
| 5066 |
if (!idev) |
| 5067 |
return 0; |
| 5068 |
|
| 5069 |
if (fib6_metric_locked(f6i, RTAX_MTU)) |
| 5070 |
return 0; |
| 5071 |
|
| 5072 |
arg->f6i = f6i; |
| 5073 |
if (f6i->nh) { |
| 5074 |
/* fib6_nh_mtu_change only returns 0, so this is safe */ |
| 5075 |
return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change, |
| 5076 |
arg); |
| 5077 |
} |
| 5078 |
|
| 5079 |
return fib6_nh_mtu_change(f6i->fib6_nh, arg); |
| 5080 |
} |
| 5081 |
|
| 5082 |
void rt6_mtu_change(struct net_device *dev, unsigned int mtu) |
| 5083 |
{ |
| 5084 |
struct rt6_mtu_change_arg arg = { |
| 5085 |
.dev = dev, |
| 5086 |
.mtu = mtu, |
| 5087 |
}; |
| 5088 |
|
| 5089 |
fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg); |
| 5090 |
} |
| 5091 |
|
| 5092 |
static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = { |
| 5093 |
[RTA_UNSPEC] = { .strict_start_type = RTA_DPORT + 1 }, |
| 5094 |
[RTA_GATEWAY] = { .len = sizeof(struct in6_addr) }, |
| 5095 |
[RTA_PREFSRC] = { .len = sizeof(struct in6_addr) }, |
| 5096 |
[RTA_OIF] = { .type = NLA_U32 }, |
| 5097 |
[RTA_IIF] = { .type = NLA_U32 }, |
| 5098 |
[RTA_PRIORITY] = { .type = NLA_U32 }, |
| 5099 |
[RTA_METRICS] = { .type = NLA_NESTED }, |
| 5100 |
[RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) }, |
| 5101 |
[RTA_PREF] = { .type = NLA_U8 }, |
| 5102 |
[RTA_ENCAP_TYPE] = { .type = NLA_U16 }, |
| 5103 |
[RTA_ENCAP] = { .type = NLA_NESTED }, |
| 5104 |
[RTA_EXPIRES] = { .type = NLA_U32 }, |
| 5105 |
[RTA_UID] = { .type = NLA_U32 }, |
| 5106 |
[RTA_MARK] = { .type = NLA_U32 }, |
| 5107 |
[RTA_TABLE] = { .type = NLA_U32 }, |
| 5108 |
[RTA_IP_PROTO] = { .type = NLA_U8 }, |
| 5109 |
[RTA_SPORT] = { .type = NLA_U16 }, |
| 5110 |
[RTA_DPORT] = { .type = NLA_U16 }, |
| 5111 |
[RTA_NH_ID] = { .type = NLA_U32 }, |
| 5112 |
[RTA_FLOWLABEL] = { .type = NLA_BE32 }, |
| 5113 |
}; |
| 5114 |
|
| 5115 |
static int rtm_to_fib6_multipath_config(struct fib6_config *cfg, |
| 5116 |
struct netlink_ext_ack *extack, |
| 5117 |
bool newroute) |
| 5118 |
{ |
| 5119 |
struct rtnexthop *rtnh; |
| 5120 |
int remaining; |
| 5121 |
|
| 5122 |
remaining = cfg->fc_mp_len; |
| 5123 |
rtnh = (struct rtnexthop *)cfg->fc_mp; |
| 5124 |
|
| 5125 |
if (!rtnh_ok(rtnh, remaining)) { |
| 5126 |
NL_SET_ERR_MSG(extack, "Invalid nexthop configuration - no valid nexthops"); |
| 5127 |
return -EINVAL; |
| 5128 |
} |
| 5129 |
|
| 5130 |
do { |
| 5131 |
bool has_gateway = cfg->fc_flags & RTF_GATEWAY; |
| 5132 |
int attrlen = rtnh_attrlen(rtnh); |
| 5133 |
|
| 5134 |
if (attrlen > 0) { |
| 5135 |
struct nlattr *nla, *attrs; |
| 5136 |
|
| 5137 |
attrs = rtnh_attrs(rtnh); |
| 5138 |
nla = nla_find(attrs, attrlen, RTA_GATEWAY); |
| 5139 |
if (nla) { |
| 5140 |
if (nla_len(nla) < sizeof(cfg->fc_gateway)) { |
| 5141 |
NL_SET_ERR_MSG(extack, |
| 5142 |
"Invalid IPv6 address in RTA_GATEWAY"); |
| 5143 |
return -EINVAL; |
| 5144 |
} |
| 5145 |
|
| 5146 |
has_gateway = true; |
| 5147 |
} |
| 5148 |
} |
| 5149 |
|
| 5150 |
if (newroute && (cfg->fc_nh_id || !has_gateway)) { |
| 5151 |
NL_SET_ERR_MSG(extack, |
| 5152 |
"Device only routes can not be added for IPv6 using the multipath API."); |
| 5153 |
return -EINVAL; |
| 5154 |
} |
| 5155 |
|
| 5156 |
rtnh = rtnh_next(rtnh, &remaining); |
| 5157 |
} while (rtnh_ok(rtnh, remaining)); |
| 5158 |
|
| 5159 |
return lwtunnel_valid_encap_type_attr(cfg->fc_mp, cfg->fc_mp_len, extack); |
| 5160 |
} |
| 5161 |
|
| 5162 |
static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh, |
| 5163 |
struct fib6_config *cfg, |
| 5164 |
struct netlink_ext_ack *extack) |
| 5165 |
{ |
| 5166 |
bool newroute = nlh->nlmsg_type == RTM_NEWROUTE; |
| 5167 |
struct nlattr *tb[RTA_MAX+1]; |
| 5168 |
struct rtmsg *rtm; |
| 5169 |
unsigned int pref; |
| 5170 |
int err; |
| 5171 |
|
| 5172 |
err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, |
| 5173 |
rtm_ipv6_policy, extack); |
| 5174 |
if (err < 0) |
| 5175 |
goto errout; |
| 5176 |
|
| 5177 |
err = -EINVAL; |
| 5178 |
rtm = nlmsg_data(nlh); |
| 5179 |
|
| 5180 |
if (rtm->rtm_tos) { |
| 5181 |
NL_SET_ERR_MSG(extack, |
| 5182 |
"Invalid dsfield (tos): option not available for IPv6"); |
| 5183 |
goto errout; |
| 5184 |
} |
| 5185 |
|
| 5186 |
if (tb[RTA_FLOWLABEL]) { |
| 5187 |
NL_SET_ERR_MSG_ATTR(extack, tb[RTA_FLOWLABEL], |
| 5188 |
"Flow label cannot be specified for this operation"); |
| 5189 |
goto errout; |
| 5190 |
} |
| 5191 |
|
| 5192 |
*cfg = (struct fib6_config){ |
| 5193 |
.fc_table = rtm->rtm_table, |
| 5194 |
.fc_dst_len = rtm->rtm_dst_len, |
| 5195 |
.fc_src_len = rtm->rtm_src_len, |
| 5196 |
.fc_flags = RTF_UP, |
| 5197 |
.fc_protocol = rtm->rtm_protocol, |
| 5198 |
.fc_type = rtm->rtm_type, |
| 5199 |
|
| 5200 |
.fc_nlinfo.portid = NETLINK_CB(skb).portid, |
| 5201 |
.fc_nlinfo.nlh = nlh, |
| 5202 |
.fc_nlinfo.nl_net = sock_net(skb->sk), |
| 5203 |
}; |
| 5204 |
|
| 5205 |
if (rtm->rtm_type == RTN_UNREACHABLE || |
| 5206 |
rtm->rtm_type == RTN_BLACKHOLE || |
| 5207 |
rtm->rtm_type == RTN_PROHIBIT || |
| 5208 |
rtm->rtm_type == RTN_THROW) |
| 5209 |
cfg->fc_flags |= RTF_REJECT; |
| 5210 |
|
| 5211 |
if (rtm->rtm_type == RTN_LOCAL) |
| 5212 |
cfg->fc_flags |= RTF_LOCAL; |
| 5213 |
|
| 5214 |
if (rtm->rtm_flags & RTM_F_CLONED) |
| 5215 |
cfg->fc_flags |= RTF_CACHE; |
| 5216 |
|
| 5217 |
cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK); |
| 5218 |
|
| 5219 |
if (tb[RTA_NH_ID]) { |
| 5220 |
if (tb[RTA_GATEWAY] || tb[RTA_OIF] || |
| 5221 |
tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) { |
| 5222 |
NL_SET_ERR_MSG(extack, |
| 5223 |
"Nexthop specification and nexthop id are mutually exclusive"); |
| 5224 |
goto errout; |
| 5225 |
} |
| 5226 |
cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]); |
| 5227 |
} |
| 5228 |
|
| 5229 |
if (tb[RTA_GATEWAY]) { |
| 5230 |
cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]); |
| 5231 |
cfg->fc_flags |= RTF_GATEWAY; |
| 5232 |
} |
| 5233 |
if (tb[RTA_VIA]) { |
| 5234 |
NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute"); |
| 5235 |
goto errout; |
| 5236 |
} |
| 5237 |
|
| 5238 |
if (tb[RTA_DST]) { |
| 5239 |
int plen = (rtm->rtm_dst_len + 7) >> 3; |
| 5240 |
|
| 5241 |
if (nla_len(tb[RTA_DST]) < plen) |
| 5242 |
goto errout; |
| 5243 |
|
| 5244 |
nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen); |
| 5245 |
} |
| 5246 |
|
| 5247 |
if (tb[RTA_SRC]) { |
| 5248 |
int plen = (rtm->rtm_src_len + 7) >> 3; |
| 5249 |
|
| 5250 |
if (nla_len(tb[RTA_SRC]) < plen) |
| 5251 |
goto errout; |
| 5252 |
|
| 5253 |
nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen); |
| 5254 |
} |
| 5255 |
|
| 5256 |
if (tb[RTA_PREFSRC]) |
| 5257 |
cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]); |
| 5258 |
|
| 5259 |
if (tb[RTA_OIF]) |
| 5260 |
cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]); |
| 5261 |
|
| 5262 |
if (tb[RTA_PRIORITY]) |
| 5263 |
cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]); |
| 5264 |
|
| 5265 |
if (tb[RTA_METRICS]) { |
| 5266 |
cfg->fc_mx = nla_data(tb[RTA_METRICS]); |
| 5267 |
cfg->fc_mx_len = nla_len(tb[RTA_METRICS]); |
| 5268 |
} |
| 5269 |
|
| 5270 |
if (tb[RTA_TABLE]) |
| 5271 |
cfg->fc_table = nla_get_u32(tb[RTA_TABLE]); |
| 5272 |
|
| 5273 |
if (tb[RTA_MULTIPATH]) { |
| 5274 |
cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]); |
| 5275 |
cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]); |
| 5276 |
|
| 5277 |
err = rtm_to_fib6_multipath_config(cfg, extack, newroute); |
| 5278 |
if (err < 0) |
| 5279 |
goto errout; |
| 5280 |
} |
| 5281 |
|
| 5282 |
if (tb[RTA_PREF]) { |
| 5283 |
pref = nla_get_u8(tb[RTA_PREF]); |
| 5284 |
if (pref != ICMPV6_ROUTER_PREF_LOW && |
| 5285 |
pref != ICMPV6_ROUTER_PREF_HIGH) |
| 5286 |
pref = ICMPV6_ROUTER_PREF_MEDIUM; |
| 5287 |
cfg->fc_flags |= RTF_PREF(pref); |
| 5288 |
} |
| 5289 |
|
| 5290 |
if (tb[RTA_ENCAP]) |
| 5291 |
cfg->fc_encap = tb[RTA_ENCAP]; |
| 5292 |
|
| 5293 |
if (tb[RTA_ENCAP_TYPE]) { |
| 5294 |
cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]); |
| 5295 |
|
| 5296 |
err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack); |
| 5297 |
if (err < 0) |
| 5298 |
goto errout; |
| 5299 |
} |
| 5300 |
|
| 5301 |
if (tb[RTA_EXPIRES]) { |
| 5302 |
unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ); |
| 5303 |
|
| 5304 |
if (addrconf_finite_timeout(timeout)) { |
| 5305 |
cfg->fc_expires = jiffies_to_clock_t(timeout * HZ); |
| 5306 |
cfg->fc_flags |= RTF_EXPIRES; |
| 5307 |
} |
| 5308 |
} |
| 5309 |
|
| 5310 |
err = 0; |
| 5311 |
errout: |
| 5312 |
return err; |
| 5313 |
} |
| 5314 |
|
| 5315 |
struct rt6_nh { |
| 5316 |
struct fib6_info *fib6_info; |
| 5317 |
struct fib6_config r_cfg; |
| 5318 |
struct list_head list; |
| 5319 |
}; |
| 5320 |
|
| 5321 |
static int ip6_route_info_append(struct list_head *rt6_nh_list, |
| 5322 |
struct fib6_info *rt, |
| 5323 |
struct fib6_config *r_cfg) |
| 5324 |
{ |
| 5325 |
struct rt6_nh *nh; |
| 5326 |
|
| 5327 |
list_for_each_entry(nh, rt6_nh_list, list) { |
| 5328 |
/* check if fib6_info already exists */ |
| 5329 |
if (rt6_duplicate_nexthop(nh->fib6_info, rt)) |
| 5330 |
return -EEXIST; |
| 5331 |
} |
| 5332 |
|
| 5333 |
nh = kzalloc_obj(*nh); |
| 5334 |
if (!nh) |
| 5335 |
return -ENOMEM; |
| 5336 |
|
| 5337 |
nh->fib6_info = rt; |
| 5338 |
memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg)); |
| 5339 |
list_add_tail(&nh->list, rt6_nh_list); |
| 5340 |
|
| 5341 |
return 0; |
| 5342 |
} |
| 5343 |
|
| 5344 |
static void ip6_route_mpath_notify(struct fib6_info *rt, |
| 5345 |
struct fib6_info *rt_last, |
| 5346 |
struct nl_info *info, |
| 5347 |
__u16 nlflags) |
| 5348 |
{ |
| 5349 |
/* if this is an APPEND route, then rt points to the first route |
| 5350 |
* inserted and rt_last points to last route inserted. Userspace |
| 5351 |
* wants a consistent dump of the route which starts at the first |
| 5352 |
* nexthop. Since sibling routes are always added at the end of |
| 5353 |
* the list, find the first sibling of the last route appended |
| 5354 |
*/ |
| 5355 |
rcu_read_lock(); |
| 5356 |
|
| 5357 |
if ((nlflags & NLM_F_APPEND) && rt_last && |
| 5358 |
READ_ONCE(rt_last->fib6_nsiblings)) { |
| 5359 |
rt = list_first_or_null_rcu(&rt_last->fib6_siblings, |
| 5360 |
struct fib6_info, |
| 5361 |
fib6_siblings); |
| 5362 |
} |
| 5363 |
|
| 5364 |
if (rt) |
| 5365 |
inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags); |
| 5366 |
|
| 5367 |
rcu_read_unlock(); |
| 5368 |
} |
| 5369 |
|
| 5370 |
static bool ip6_route_mpath_should_notify(const struct fib6_info *rt) |
| 5371 |
{ |
| 5372 |
bool rt_can_ecmp = rt6_qualify_for_ecmp(rt); |
| 5373 |
bool should_notify = false; |
| 5374 |
struct fib6_info *leaf; |
| 5375 |
struct fib6_node *fn; |
| 5376 |
|
| 5377 |
rcu_read_lock(); |
| 5378 |
fn = rcu_dereference(rt->fib6_node); |
| 5379 |
if (!fn) |
| 5380 |
goto out; |
| 5381 |
|
| 5382 |
leaf = rcu_dereference(fn->leaf); |
| 5383 |
if (!leaf) |
| 5384 |
goto out; |
| 5385 |
|
| 5386 |
if (rt == leaf || |
| 5387 |
(rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric && |
| 5388 |
rt6_qualify_for_ecmp(leaf))) |
| 5389 |
should_notify = true; |
| 5390 |
out: |
| 5391 |
rcu_read_unlock(); |
| 5392 |
|
| 5393 |
return should_notify; |
| 5394 |
} |
| 5395 |
|
| 5396 |
static int ip6_route_multipath_add(struct fib6_config *cfg, |
| 5397 |
struct netlink_ext_ack *extack) |
| 5398 |
{ |
| 5399 |
struct fib6_info *rt_notif = NULL, *rt_last = NULL; |
| 5400 |
struct nl_info *info = &cfg->fc_nlinfo; |
| 5401 |
struct rt6_nh *nh, *nh_safe; |
| 5402 |
struct fib6_config r_cfg; |
| 5403 |
struct rtnexthop *rtnh; |
| 5404 |
LIST_HEAD(rt6_nh_list); |
| 5405 |
struct rt6_nh *err_nh; |
| 5406 |
struct fib6_info *rt; |
| 5407 |
__u16 nlflags; |
| 5408 |
int remaining; |
| 5409 |
int attrlen; |
| 5410 |
int replace; |
| 5411 |
int nhn = 0; |
| 5412 |
int err; |
| 5413 |
|
| 5414 |
err = fib6_config_validate(cfg, extack); |
| 5415 |
if (err) |
| 5416 |
return err; |
| 5417 |
|
| 5418 |
replace = (cfg->fc_nlinfo.nlh && |
| 5419 |
(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE)); |
| 5420 |
|
| 5421 |
nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE; |
| 5422 |
if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND) |
| 5423 |
nlflags |= NLM_F_APPEND; |
| 5424 |
|
| 5425 |
remaining = cfg->fc_mp_len; |
| 5426 |
rtnh = (struct rtnexthop *)cfg->fc_mp; |
| 5427 |
|
| 5428 |
/* Parse a Multipath Entry and build a list (rt6_nh_list) of |
| 5429 |
* fib6_info structs per nexthop |
| 5430 |
*/ |
| 5431 |
while (rtnh_ok(rtnh, remaining)) { |
| 5432 |
memcpy(&r_cfg, cfg, sizeof(*cfg)); |
| 5433 |
if (rtnh->rtnh_ifindex) |
| 5434 |
r_cfg.fc_ifindex = rtnh->rtnh_ifindex; |
| 5435 |
|
| 5436 |
attrlen = rtnh_attrlen(rtnh); |
| 5437 |
if (attrlen > 0) { |
| 5438 |
struct nlattr *nla, *attrs = rtnh_attrs(rtnh); |
| 5439 |
|
| 5440 |
nla = nla_find(attrs, attrlen, RTA_GATEWAY); |
| 5441 |
if (nla) { |
| 5442 |
r_cfg.fc_gateway = nla_get_in6_addr(nla); |
| 5443 |
r_cfg.fc_flags |= RTF_GATEWAY; |
| 5444 |
} |
| 5445 |
|
| 5446 |
r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP); |
| 5447 |
nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE); |
| 5448 |
if (nla) |
| 5449 |
r_cfg.fc_encap_type = nla_get_u16(nla); |
| 5450 |
} |
| 5451 |
|
| 5452 |
r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK); |
| 5453 |
rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack); |
| 5454 |
if (IS_ERR(rt)) { |
| 5455 |
err = PTR_ERR(rt); |
| 5456 |
rt = NULL; |
| 5457 |
goto cleanup; |
| 5458 |
} |
| 5459 |
|
| 5460 |
err = ip6_route_info_create_nh(rt, &r_cfg, GFP_KERNEL, extack); |
| 5461 |
if (err) { |
| 5462 |
rt = NULL; |
| 5463 |
goto cleanup; |
| 5464 |
} |
| 5465 |
|
| 5466 |
rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1; |
| 5467 |
|
| 5468 |
err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg); |
| 5469 |
if (err) { |
| 5470 |
fib6_info_release(rt); |
| 5471 |
goto cleanup; |
| 5472 |
} |
| 5473 |
|
| 5474 |
rtnh = rtnh_next(rtnh, &remaining); |
| 5475 |
} |
| 5476 |
|
| 5477 |
/* for add and replace send one notification with all nexthops. |
| 5478 |
* Skip the notification in fib6_add_rt2node and send one with |
| 5479 |
* the full route when done |
| 5480 |
*/ |
| 5481 |
info->skip_notify = 1; |
| 5482 |
|
| 5483 |
/* For add and replace, send one notification with all nexthops. For |
| 5484 |
* append, send one notification with all appended nexthops. |
| 5485 |
*/ |
| 5486 |
info->skip_notify_kernel = 1; |
| 5487 |
|
| 5488 |
err_nh = NULL; |
| 5489 |
list_for_each_entry(nh, &rt6_nh_list, list) { |
| 5490 |
err = __ip6_ins_rt(nh->fib6_info, info, extack); |
| 5491 |
|
| 5492 |
if (err) { |
| 5493 |
if (replace && nhn) |
| 5494 |
NL_SET_ERR_MSG_MOD(extack, |
| 5495 |
"multipath route replace failed (check consistency of installed routes)"); |
| 5496 |
err_nh = nh; |
| 5497 |
goto add_errout; |
| 5498 |
} |
| 5499 |
/* save reference to last route successfully inserted */ |
| 5500 |
rt_last = nh->fib6_info; |
| 5501 |
|
| 5502 |
/* save reference to first route for notification */ |
| 5503 |
if (!rt_notif) |
| 5504 |
rt_notif = nh->fib6_info; |
| 5505 |
|
| 5506 |
/* Because each route is added like a single route we remove |
| 5507 |
* these flags after the first nexthop: if there is a collision, |
| 5508 |
* we have already failed to add the first nexthop: |
| 5509 |
* fib6_add_rt2node() has rejected it; when replacing, old |
| 5510 |
* nexthops have been replaced by first new, the rest should |
| 5511 |
* be added to it. |
| 5512 |
*/ |
| 5513 |
if (cfg->fc_nlinfo.nlh) { |
| 5514 |
cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL | |
| 5515 |
NLM_F_REPLACE); |
| 5516 |
cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE; |
| 5517 |
} |
| 5518 |
nhn++; |
| 5519 |
} |
| 5520 |
|
| 5521 |
/* An in-kernel notification should only be sent in case the new |
| 5522 |
* multipath route is added as the first route in the node, or if |
| 5523 |
* it was appended to it. We pass 'rt_notif' since it is the first |
| 5524 |
* sibling and might allow us to skip some checks in the replace case. |
| 5525 |
*/ |
| 5526 |
if (ip6_route_mpath_should_notify(rt_notif)) { |
| 5527 |
enum fib_event_type fib_event; |
| 5528 |
|
| 5529 |
if (rt_notif->fib6_nsiblings != nhn - 1) |
| 5530 |
fib_event = FIB_EVENT_ENTRY_APPEND; |
| 5531 |
else |
| 5532 |
fib_event = FIB_EVENT_ENTRY_REPLACE; |
| 5533 |
|
| 5534 |
err = call_fib6_multipath_entry_notifiers(info->nl_net, |
| 5535 |
fib_event, rt_notif, |
| 5536 |
nhn - 1, extack); |
| 5537 |
if (err) { |
| 5538 |
/* Delete all the siblings that were just added */ |
| 5539 |
err_nh = NULL; |
| 5540 |
goto add_errout; |
| 5541 |
} |
| 5542 |
} |
| 5543 |
|
| 5544 |
/* success ... tell user about new route */ |
| 5545 |
ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags); |
| 5546 |
goto cleanup; |
| 5547 |
|
| 5548 |
add_errout: |
| 5549 |
/* send notification for routes that were added so that |
| 5550 |
* the delete notifications sent by ip6_route_del are |
| 5551 |
* coherent |
| 5552 |
*/ |
| 5553 |
if (rt_notif) |
| 5554 |
ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags); |
| 5555 |
|
| 5556 |
/* Delete routes that were already added */ |
| 5557 |
list_for_each_entry(nh, &rt6_nh_list, list) { |
| 5558 |
if (err_nh == nh) |
| 5559 |
break; |
| 5560 |
ip6_route_del(&nh->r_cfg, extack); |
| 5561 |
} |
| 5562 |
|
| 5563 |
cleanup: |
| 5564 |
list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, list) { |
| 5565 |
fib6_info_release(nh->fib6_info); |
| 5566 |
list_del(&nh->list); |
| 5567 |
kfree(nh); |
| 5568 |
} |
| 5569 |
|
| 5570 |
return err; |
| 5571 |
} |
| 5572 |
|
| 5573 |
static int ip6_route_multipath_del(struct fib6_config *cfg, |
| 5574 |
struct netlink_ext_ack *extack) |
| 5575 |
{ |
| 5576 |
struct fib6_config r_cfg; |
| 5577 |
struct rtnexthop *rtnh; |
| 5578 |
int last_err = 0; |
| 5579 |
int remaining; |
| 5580 |
int attrlen; |
| 5581 |
int err; |
| 5582 |
|
| 5583 |
remaining = cfg->fc_mp_len; |
| 5584 |
rtnh = (struct rtnexthop *)cfg->fc_mp; |
| 5585 |
|
| 5586 |
/* Parse a Multipath Entry */ |
| 5587 |
while (rtnh_ok(rtnh, remaining)) { |
| 5588 |
memcpy(&r_cfg, cfg, sizeof(*cfg)); |
| 5589 |
if (rtnh->rtnh_ifindex) |
| 5590 |
r_cfg.fc_ifindex = rtnh->rtnh_ifindex; |
| 5591 |
|
| 5592 |
attrlen = rtnh_attrlen(rtnh); |
| 5593 |
if (attrlen > 0) { |
| 5594 |
struct nlattr *nla, *attrs = rtnh_attrs(rtnh); |
| 5595 |
|
| 5596 |
nla = nla_find(attrs, attrlen, RTA_GATEWAY); |
| 5597 |
if (nla) { |
| 5598 |
r_cfg.fc_gateway = nla_get_in6_addr(nla); |
| 5599 |
r_cfg.fc_flags |= RTF_GATEWAY; |
| 5600 |
} |
| 5601 |
} |
| 5602 |
|
| 5603 |
err = ip6_route_del(&r_cfg, extack); |
| 5604 |
if (err) |
| 5605 |
last_err = err; |
| 5606 |
|
| 5607 |
rtnh = rtnh_next(rtnh, &remaining); |
| 5608 |
} |
| 5609 |
|
| 5610 |
return last_err; |
| 5611 |
} |
| 5612 |
|
| 5613 |
static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
| 5614 |
struct netlink_ext_ack *extack) |
| 5615 |
{ |
| 5616 |
struct fib6_config cfg; |
| 5617 |
int err; |
| 5618 |
|
| 5619 |
err = rtm_to_fib6_config(skb, nlh, &cfg, extack); |
| 5620 |
if (err < 0) |
| 5621 |
return err; |
| 5622 |
|
| 5623 |
if (cfg.fc_nh_id) { |
| 5624 |
rcu_read_lock(); |
| 5625 |
err = !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id); |
| 5626 |
rcu_read_unlock(); |
| 5627 |
|
| 5628 |
if (err) { |
| 5629 |
NL_SET_ERR_MSG(extack, "Nexthop id does not exist"); |
| 5630 |
return -EINVAL; |
| 5631 |
} |
| 5632 |
} |
| 5633 |
|
| 5634 |
if (cfg.fc_mp) { |
| 5635 |
return ip6_route_multipath_del(&cfg, extack); |
| 5636 |
} else { |
| 5637 |
cfg.fc_delete_all_nh = 1; |
| 5638 |
return ip6_route_del(&cfg, extack); |
| 5639 |
} |
| 5640 |
} |
| 5641 |
|
| 5642 |
static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh, |
| 5643 |
struct netlink_ext_ack *extack) |
| 5644 |
{ |
| 5645 |
struct fib6_config cfg; |
| 5646 |
int err; |
| 5647 |
|
| 5648 |
err = rtm_to_fib6_config(skb, nlh, &cfg, extack); |
| 5649 |
if (err < 0) |
| 5650 |
return err; |
| 5651 |
|
| 5652 |
if (cfg.fc_metric == 0) |
| 5653 |
cfg.fc_metric = IP6_RT_PRIO_USER; |
| 5654 |
|
| 5655 |
if (cfg.fc_mp) |
| 5656 |
return ip6_route_multipath_add(&cfg, extack); |
| 5657 |
else |
| 5658 |
return ip6_route_add(&cfg, GFP_KERNEL, extack); |
| 5659 |
} |
| 5660 |
|
| 5661 |
/* add the overhead of this fib6_nh to nexthop_len */ |
| 5662 |
static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg) |
| 5663 |
{ |
| 5664 |
int *nexthop_len = arg; |
| 5665 |
|
| 5666 |
*nexthop_len += nla_total_size(0) /* RTA_MULTIPATH */ |
| 5667 |
+ NLA_ALIGN(sizeof(struct rtnexthop)) |
| 5668 |
+ nla_total_size(16); /* RTA_GATEWAY */ |
| 5669 |
|
| 5670 |
if (nh->fib_nh_lws) { |
| 5671 |
/* RTA_ENCAP_TYPE */ |
| 5672 |
*nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws); |
| 5673 |
/* RTA_ENCAP */ |
| 5674 |
*nexthop_len += nla_total_size(2); |
| 5675 |
} |
| 5676 |
|
| 5677 |
return 0; |
| 5678 |
} |
| 5679 |
|
| 5680 |
static size_t rt6_nlmsg_size(struct fib6_info *f6i) |
| 5681 |
{ |
| 5682 |
struct fib6_info *sibling; |
| 5683 |
struct fib6_nh *nh; |
| 5684 |
int nexthop_len; |
| 5685 |
|
| 5686 |
if (f6i->nh) { |
| 5687 |
nexthop_len = nla_total_size(4); /* RTA_NH_ID */ |
| 5688 |
nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size, |
| 5689 |
&nexthop_len); |
| 5690 |
goto common; |
| 5691 |
} |
| 5692 |
|
| 5693 |
rcu_read_lock(); |
| 5694 |
retry: |
| 5695 |
nh = f6i->fib6_nh; |
| 5696 |
nexthop_len = 0; |
| 5697 |
if (READ_ONCE(f6i->fib6_nsiblings)) { |
| 5698 |
rt6_nh_nlmsg_size(nh, &nexthop_len); |
| 5699 |
|
| 5700 |
list_for_each_entry_rcu(sibling, &f6i->fib6_siblings, |
| 5701 |
fib6_siblings) { |
| 5702 |
rt6_nh_nlmsg_size(sibling->fib6_nh, &nexthop_len); |
| 5703 |
if (!READ_ONCE(f6i->fib6_nsiblings)) |
| 5704 |
goto retry; |
| 5705 |
} |
| 5706 |
} |
| 5707 |
rcu_read_unlock(); |
| 5708 |
nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws); |
| 5709 |
common: |
| 5710 |
return NLMSG_ALIGN(sizeof(struct rtmsg)) |
| 5711 |
+ nla_total_size(16) /* RTA_SRC */ |
| 5712 |
+ nla_total_size(16) /* RTA_DST */ |
| 5713 |
+ nla_total_size(16) /* RTA_GATEWAY */ |
| 5714 |
+ nla_total_size(16) /* RTA_PREFSRC */ |
| 5715 |
+ nla_total_size(4) /* RTA_TABLE */ |
| 5716 |
+ nla_total_size(4) /* RTA_IIF */ |
| 5717 |
+ nla_total_size(4) /* RTA_OIF */ |
| 5718 |
+ nla_total_size(4) /* RTA_PRIORITY */ |
| 5719 |
+ RTAX_MAX * nla_total_size(4) /* RTA_METRICS */ |
| 5720 |
+ nla_total_size(sizeof(struct rta_cacheinfo)) |
| 5721 |
+ nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */ |
| 5722 |
+ nla_total_size(1) /* RTA_PREF */ |
| 5723 |
+ nexthop_len; |
| 5724 |
} |
| 5725 |
|
| 5726 |
static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh, |
| 5727 |
unsigned char *flags) |
| 5728 |
{ |
| 5729 |
if (nexthop_is_multipath(nh)) { |
| 5730 |
struct nlattr *mp; |
| 5731 |
|
| 5732 |
mp = nla_nest_start_noflag(skb, RTA_MULTIPATH); |
| 5733 |
if (!mp) |
| 5734 |
goto nla_put_failure; |
| 5735 |
|
| 5736 |
if (nexthop_mpath_fill_node(skb, nh, AF_INET6)) |
| 5737 |
goto nla_put_failure; |
| 5738 |
|
| 5739 |
nla_nest_end(skb, mp); |
| 5740 |
} else { |
| 5741 |
struct fib6_nh *fib6_nh; |
| 5742 |
|
| 5743 |
fib6_nh = nexthop_fib6_nh(nh); |
| 5744 |
if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6, |
| 5745 |
flags, false) < 0) |
| 5746 |
goto nla_put_failure; |
| 5747 |
} |
| 5748 |
|
| 5749 |
return 0; |
| 5750 |
|
| 5751 |
nla_put_failure: |
| 5752 |
return -EMSGSIZE; |
| 5753 |
} |
| 5754 |
|
| 5755 |
static int rt6_fill_node(struct net *net, struct sk_buff *skb, |
| 5756 |
struct fib6_info *rt, struct dst_entry *dst, |
| 5757 |
struct in6_addr *dest, struct in6_addr *src, |
| 5758 |
int iif, int type, u32 portid, u32 seq, |
| 5759 |
unsigned int flags) |
| 5760 |
{ |
| 5761 |
struct rt6_info *rt6 = dst_rt6_info(dst); |
| 5762 |
struct rt6key *rt6_dst, *rt6_src; |
| 5763 |
u32 *pmetrics, table, rt6_flags; |
| 5764 |
unsigned char nh_flags = 0; |
| 5765 |
struct nlmsghdr *nlh; |
| 5766 |
struct rtmsg *rtm; |
| 5767 |
long expires = 0; |
| 5768 |
|
| 5769 |
nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags); |
| 5770 |
if (!nlh) |
| 5771 |
return -EMSGSIZE; |
| 5772 |
|
| 5773 |
if (rt6) { |
| 5774 |
rt6_dst = &rt6->rt6i_dst; |
| 5775 |
rt6_src = &rt6->rt6i_src; |
| 5776 |
rt6_flags = rt6->rt6i_flags; |
| 5777 |
} else { |
| 5778 |
rt6_dst = &rt->fib6_dst; |
| 5779 |
rt6_src = &rt->fib6_src; |
| 5780 |
rt6_flags = rt->fib6_flags; |
| 5781 |
} |
| 5782 |
|
| 5783 |
rtm = nlmsg_data(nlh); |
| 5784 |
rtm->rtm_family = AF_INET6; |
| 5785 |
rtm->rtm_dst_len = rt6_dst->plen; |
| 5786 |
rtm->rtm_src_len = rt6_src->plen; |
| 5787 |
rtm->rtm_tos = 0; |
| 5788 |
if (rt->fib6_table) |
| 5789 |
table = rt->fib6_table->tb6_id; |
| 5790 |
else |
| 5791 |
table = RT6_TABLE_UNSPEC; |
| 5792 |
rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT; |
| 5793 |
if (nla_put_u32(skb, RTA_TABLE, table)) |
| 5794 |
goto nla_put_failure; |
| 5795 |
|
| 5796 |
rtm->rtm_type = rt->fib6_type; |
| 5797 |
rtm->rtm_flags = 0; |
| 5798 |
rtm->rtm_scope = RT_SCOPE_UNIVERSE; |
| 5799 |
rtm->rtm_protocol = rt->fib6_protocol; |
| 5800 |
|
| 5801 |
if (rt6_flags & RTF_CACHE) |
| 5802 |
rtm->rtm_flags |= RTM_F_CLONED; |
| 5803 |
|
| 5804 |
if (dest) { |
| 5805 |
if (nla_put_in6_addr(skb, RTA_DST, dest)) |
| 5806 |
goto nla_put_failure; |
| 5807 |
rtm->rtm_dst_len = 128; |
| 5808 |
} else if (rtm->rtm_dst_len) |
| 5809 |
if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr)) |
| 5810 |
goto nla_put_failure; |
| 5811 |
#ifdef CONFIG_IPV6_SUBTREES |
| 5812 |
if (src) { |
| 5813 |
if (nla_put_in6_addr(skb, RTA_SRC, src)) |
| 5814 |
goto nla_put_failure; |
| 5815 |
rtm->rtm_src_len = 128; |
| 5816 |
} else if (rtm->rtm_src_len && |
| 5817 |
nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr)) |
| 5818 |
goto nla_put_failure; |
| 5819 |
#endif |
| 5820 |
if (iif) { |
| 5821 |
#ifdef CONFIG_IPV6_MROUTE |
| 5822 |
if (ipv6_addr_is_multicast(&rt6_dst->addr)) { |
| 5823 |
int err = ip6mr_get_route(net, skb, rtm, portid); |
| 5824 |
|
| 5825 |
if (err == 0) |
| 5826 |
return 0; |
| 5827 |
if (err < 0) |
| 5828 |
goto nla_put_failure; |
| 5829 |
} else |
| 5830 |
#endif |
| 5831 |
if (nla_put_u32(skb, RTA_IIF, iif)) |
| 5832 |
goto nla_put_failure; |
| 5833 |
} else if (dest) { |
| 5834 |
struct in6_addr saddr_buf; |
| 5835 |
if (ip6_route_get_saddr(net, rt, dest, 0, 0, &saddr_buf) == 0 && |
| 5836 |
nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf)) |
| 5837 |
goto nla_put_failure; |
| 5838 |
} |
| 5839 |
|
| 5840 |
if (rt->fib6_prefsrc.plen) { |
| 5841 |
struct in6_addr saddr_buf; |
| 5842 |
saddr_buf = rt->fib6_prefsrc.addr; |
| 5843 |
if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf)) |
| 5844 |
goto nla_put_failure; |
| 5845 |
} |
| 5846 |
|
| 5847 |
pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics; |
| 5848 |
if (rtnetlink_put_metrics(skb, pmetrics) < 0) |
| 5849 |
goto nla_put_failure; |
| 5850 |
|
| 5851 |
if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric)) |
| 5852 |
goto nla_put_failure; |
| 5853 |
|
| 5854 |
/* For multipath routes, walk the siblings list and add |
| 5855 |
* each as a nexthop within RTA_MULTIPATH. |
| 5856 |
*/ |
| 5857 |
if (rt6) { |
| 5858 |
struct net_device *dev; |
| 5859 |
|
| 5860 |
if (rt6_flags & RTF_GATEWAY && |
| 5861 |
nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway)) |
| 5862 |
goto nla_put_failure; |
| 5863 |
|
| 5864 |
dev = dst_dev(dst); |
| 5865 |
if (dev && nla_put_u32(skb, RTA_OIF, dev->ifindex)) |
| 5866 |
goto nla_put_failure; |
| 5867 |
|
| 5868 |
if (lwtunnel_fill_encap(skb, dst->lwtstate, RTA_ENCAP, RTA_ENCAP_TYPE) < 0) |
| 5869 |
goto nla_put_failure; |
| 5870 |
} else if (READ_ONCE(rt->fib6_nsiblings)) { |
| 5871 |
struct fib6_info *sibling; |
| 5872 |
struct nlattr *mp; |
| 5873 |
|
| 5874 |
mp = nla_nest_start_noflag(skb, RTA_MULTIPATH); |
| 5875 |
if (!mp) |
| 5876 |
goto nla_put_failure; |
| 5877 |
|
| 5878 |
if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common, |
| 5879 |
rt->fib6_nh->fib_nh_weight, AF_INET6, |
| 5880 |
0) < 0) |
| 5881 |
goto nla_put_failure; |
| 5882 |
|
| 5883 |
rcu_read_lock(); |
| 5884 |
|
| 5885 |
list_for_each_entry_rcu(sibling, &rt->fib6_siblings, |
| 5886 |
fib6_siblings) { |
| 5887 |
if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common, |
| 5888 |
sibling->fib6_nh->fib_nh_weight, |
| 5889 |
AF_INET6, 0) < 0) { |
| 5890 |
rcu_read_unlock(); |
| 5891 |
|
| 5892 |
goto nla_put_failure; |
| 5893 |
} |
| 5894 |
} |
| 5895 |
|
| 5896 |
rcu_read_unlock(); |
| 5897 |
|
| 5898 |
nla_nest_end(skb, mp); |
| 5899 |
} else if (rt->nh) { |
| 5900 |
if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id)) |
| 5901 |
goto nla_put_failure; |
| 5902 |
|
| 5903 |
if (nexthop_is_blackhole(rt->nh)) |
| 5904 |
rtm->rtm_type = RTN_BLACKHOLE; |
| 5905 |
|
| 5906 |
if (READ_ONCE(net->ipv4.sysctl_nexthop_compat_mode) && |
| 5907 |
rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0) |
| 5908 |
goto nla_put_failure; |
| 5909 |
|
| 5910 |
rtm->rtm_flags |= nh_flags; |
| 5911 |
} else { |
| 5912 |
if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6, |
| 5913 |
&nh_flags, false) < 0) |
| 5914 |
goto nla_put_failure; |
| 5915 |
|
| 5916 |
rtm->rtm_flags |= nh_flags; |
| 5917 |
} |
| 5918 |
|
| 5919 |
if (rt6_flags & RTF_EXPIRES) { |
| 5920 |
expires = dst ? READ_ONCE(dst->expires) : rt->expires; |
| 5921 |
expires -= jiffies; |
| 5922 |
} |
| 5923 |
|
| 5924 |
if (!dst) { |
| 5925 |
if (READ_ONCE(rt->offload)) |
| 5926 |
rtm->rtm_flags |= RTM_F_OFFLOAD; |
| 5927 |
if (READ_ONCE(rt->trap)) |
| 5928 |
rtm->rtm_flags |= RTM_F_TRAP; |
| 5929 |
if (READ_ONCE(rt->offload_failed)) |
| 5930 |
rtm->rtm_flags |= RTM_F_OFFLOAD_FAILED; |
| 5931 |
} |
| 5932 |
|
| 5933 |
if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0) |
| 5934 |
goto nla_put_failure; |
| 5935 |
|
| 5936 |
if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags))) |
| 5937 |
goto nla_put_failure; |
| 5938 |
|
| 5939 |
|
| 5940 |
nlmsg_end(skb, nlh); |
| 5941 |
return 0; |
| 5942 |
|
| 5943 |
nla_put_failure: |
| 5944 |
nlmsg_cancel(skb, nlh); |
| 5945 |
return -EMSGSIZE; |
| 5946 |
} |
| 5947 |
|
| 5948 |
static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg) |
| 5949 |
{ |
| 5950 |
const struct net_device *dev = arg; |
| 5951 |
|
| 5952 |
if (nh->fib_nh_dev == dev) |
| 5953 |
return 1; |
| 5954 |
|
| 5955 |
return 0; |
| 5956 |
} |
| 5957 |
|
| 5958 |
static bool fib6_info_uses_dev(const struct fib6_info *f6i, |
| 5959 |
const struct net_device *dev) |
| 5960 |
{ |
| 5961 |
if (f6i->nh) { |
| 5962 |
struct net_device *_dev = (struct net_device *)dev; |
| 5963 |
|
| 5964 |
return !!nexthop_for_each_fib6_nh(f6i->nh, |
| 5965 |
fib6_info_nh_uses_dev, |
| 5966 |
_dev); |
| 5967 |
} |
| 5968 |
|
| 5969 |
if (f6i->fib6_nh->fib_nh_dev == dev) |
| 5970 |
return true; |
| 5971 |
|
| 5972 |
if (READ_ONCE(f6i->fib6_nsiblings)) { |
| 5973 |
const struct fib6_info *sibling; |
| 5974 |
|
| 5975 |
rcu_read_lock(); |
| 5976 |
list_for_each_entry_rcu(sibling, &f6i->fib6_siblings, |
| 5977 |
fib6_siblings) { |
| 5978 |
if (sibling->fib6_nh->fib_nh_dev == dev) { |
| 5979 |
rcu_read_unlock(); |
| 5980 |
return true; |
| 5981 |
} |
| 5982 |
if (!READ_ONCE(f6i->fib6_nsiblings)) |
| 5983 |
break; |
| 5984 |
} |
| 5985 |
rcu_read_unlock(); |
| 5986 |
} |
| 5987 |
return false; |
| 5988 |
} |
| 5989 |
|
| 5990 |
struct fib6_nh_exception_dump_walker { |
| 5991 |
struct rt6_rtnl_dump_arg *dump; |
| 5992 |
struct fib6_info *rt; |
| 5993 |
unsigned int flags; |
| 5994 |
unsigned int skip; |
| 5995 |
unsigned int count; |
| 5996 |
}; |
| 5997 |
|
| 5998 |
static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg) |
| 5999 |
{ |
| 6000 |
struct fib6_nh_exception_dump_walker *w = arg; |
| 6001 |
struct rt6_rtnl_dump_arg *dump = w->dump; |
| 6002 |
struct rt6_exception_bucket *bucket; |
| 6003 |
struct rt6_exception *rt6_ex; |
| 6004 |
int i, err; |
| 6005 |
|
| 6006 |
bucket = fib6_nh_get_excptn_bucket(nh, NULL); |
| 6007 |
if (!bucket) |
| 6008 |
return 0; |
| 6009 |
|
| 6010 |
for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) { |
| 6011 |
hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) { |
| 6012 |
if (w->skip) { |
| 6013 |
w->skip--; |
| 6014 |
continue; |
| 6015 |
} |
| 6016 |
|
| 6017 |
/* Expiration of entries doesn't bump sernum, insertion |
| 6018 |
* does. Removal is triggered by insertion, so we can |
| 6019 |
* rely on the fact that if entries change between two |
| 6020 |
* partial dumps, this node is scanned again completely, |
| 6021 |
* see rt6_insert_exception() and fib6_dump_table(). |
| 6022 |
* |
| 6023 |
* Count expired entries we go through as handled |
| 6024 |
* entries that we'll skip next time, in case of partial |
| 6025 |
* node dump. Otherwise, if entries expire meanwhile, |
| 6026 |
* we'll skip the wrong amount. |
| 6027 |
*/ |
| 6028 |
if (rt6_check_expired(rt6_ex->rt6i)) { |
| 6029 |
w->count++; |
| 6030 |
continue; |
| 6031 |
} |
| 6032 |
|
| 6033 |
err = rt6_fill_node(dump->net, dump->skb, w->rt, |
| 6034 |
&rt6_ex->rt6i->dst, NULL, NULL, 0, |
| 6035 |
RTM_NEWROUTE, |
| 6036 |
NETLINK_CB(dump->cb->skb).portid, |
| 6037 |
dump->cb->nlh->nlmsg_seq, w->flags); |
| 6038 |
if (err) |
| 6039 |
return err; |
| 6040 |
|
| 6041 |
w->count++; |
| 6042 |
} |
| 6043 |
bucket++; |
| 6044 |
} |
| 6045 |
|
| 6046 |
return 0; |
| 6047 |
} |
| 6048 |
|
| 6049 |
/* Return -1 if done with node, number of handled routes on partial dump */ |
| 6050 |
int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip) |
| 6051 |
{ |
| 6052 |
struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg; |
| 6053 |
struct fib_dump_filter *filter = &arg->filter; |
| 6054 |
unsigned int flags = NLM_F_MULTI; |
| 6055 |
struct net *net = arg->net; |
| 6056 |
int count = 0; |
| 6057 |
|
| 6058 |
if (rt == net->ipv6.fib6_null_entry) |
| 6059 |
return -1; |
| 6060 |
|
| 6061 |
if ((filter->flags & RTM_F_PREFIX) && |
| 6062 |
!(rt->fib6_flags & RTF_PREFIX_RT)) { |
| 6063 |
/* success since this is not a prefix route */ |
| 6064 |
return -1; |
| 6065 |
} |
| 6066 |
if (filter->filter_set && |
| 6067 |
((filter->rt_type && rt->fib6_type != filter->rt_type) || |
| 6068 |
(filter->dev && !fib6_info_uses_dev(rt, filter->dev)) || |
| 6069 |
(filter->protocol && rt->fib6_protocol != filter->protocol))) { |
| 6070 |
return -1; |
| 6071 |
} |
| 6072 |
|
| 6073 |
if (filter->filter_set || |
| 6074 |
!filter->dump_routes || !filter->dump_exceptions) { |
| 6075 |
flags |= NLM_F_DUMP_FILTERED; |
| 6076 |
} |
| 6077 |
|
| 6078 |
if (filter->dump_routes) { |
| 6079 |
if (skip) { |
| 6080 |
skip--; |
| 6081 |
} else { |
| 6082 |
if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL, |
| 6083 |
0, RTM_NEWROUTE, |
| 6084 |
NETLINK_CB(arg->cb->skb).portid, |
| 6085 |
arg->cb->nlh->nlmsg_seq, flags)) { |
| 6086 |
return 0; |
| 6087 |
} |
| 6088 |
count++; |
| 6089 |
} |
| 6090 |
} |
| 6091 |
|
| 6092 |
if (filter->dump_exceptions) { |
| 6093 |
struct fib6_nh_exception_dump_walker w = { .dump = arg, |
| 6094 |
.rt = rt, |
| 6095 |
.flags = flags, |
| 6096 |
.skip = skip, |
| 6097 |
.count = 0 }; |
| 6098 |
int err; |
| 6099 |
|
| 6100 |
rcu_read_lock(); |
| 6101 |
if (rt->nh) { |
| 6102 |
err = nexthop_for_each_fib6_nh(rt->nh, |
| 6103 |
rt6_nh_dump_exceptions, |
| 6104 |
&w); |
| 6105 |
} else { |
| 6106 |
err = rt6_nh_dump_exceptions(rt->fib6_nh, &w); |
| 6107 |
} |
| 6108 |
rcu_read_unlock(); |
| 6109 |
|
| 6110 |
if (err) |
| 6111 |
return count + w.count; |
| 6112 |
} |
| 6113 |
|
| 6114 |
return -1; |
| 6115 |
} |
| 6116 |
|
| 6117 |
static int inet6_rtm_valid_getroute_req(struct sk_buff *skb, |
| 6118 |
const struct nlmsghdr *nlh, |
| 6119 |
struct nlattr **tb, |
| 6120 |
struct netlink_ext_ack *extack) |
| 6121 |
{ |
| 6122 |
struct rtmsg *rtm; |
| 6123 |
int i, err; |
| 6124 |
|
| 6125 |
rtm = nlmsg_payload(nlh, sizeof(*rtm)); |
| 6126 |
if (!rtm) { |
| 6127 |
NL_SET_ERR_MSG_MOD(extack, |
| 6128 |
"Invalid header for get route request"); |
| 6129 |
return -EINVAL; |
| 6130 |
} |
| 6131 |
|
| 6132 |
if (!netlink_strict_get_check(skb)) |
| 6133 |
return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX, |
| 6134 |
rtm_ipv6_policy, extack); |
| 6135 |
|
| 6136 |
if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) || |
| 6137 |
(rtm->rtm_dst_len && rtm->rtm_dst_len != 128) || |
| 6138 |
rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope || |
| 6139 |
rtm->rtm_type) { |
| 6140 |
NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request"); |
| 6141 |
return -EINVAL; |
| 6142 |
} |
| 6143 |
if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) { |
| 6144 |
NL_SET_ERR_MSG_MOD(extack, |
| 6145 |
"Invalid flags for get route request"); |
| 6146 |
return -EINVAL; |
| 6147 |
} |
| 6148 |
|
| 6149 |
err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX, |
| 6150 |
rtm_ipv6_policy, extack); |
| 6151 |
if (err) |
| 6152 |
return err; |
| 6153 |
|
| 6154 |
if ((tb[RTA_SRC] && !rtm->rtm_src_len) || |
| 6155 |
(tb[RTA_DST] && !rtm->rtm_dst_len)) { |
| 6156 |
NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6"); |
| 6157 |
return -EINVAL; |
| 6158 |
} |
| 6159 |
|
| 6160 |
if (tb[RTA_FLOWLABEL] && |
| 6161 |
(nla_get_be32(tb[RTA_FLOWLABEL]) & ~IPV6_FLOWLABEL_MASK)) { |
| 6162 |
NL_SET_ERR_MSG_ATTR(extack, tb[RTA_FLOWLABEL], |
| 6163 |
"Invalid flow label"); |
| 6164 |
return -EINVAL; |
| 6165 |
} |
| 6166 |
|
| 6167 |
for (i = 0; i <= RTA_MAX; i++) { |
| 6168 |
if (!tb[i]) |
| 6169 |
continue; |
| 6170 |
|
| 6171 |
switch (i) { |
| 6172 |
case RTA_SRC: |
| 6173 |
case RTA_DST: |
| 6174 |
case RTA_IIF: |
| 6175 |
case RTA_OIF: |
| 6176 |
case RTA_MARK: |
| 6177 |
case RTA_UID: |
| 6178 |
case RTA_SPORT: |
| 6179 |
case RTA_DPORT: |
| 6180 |
case RTA_IP_PROTO: |
| 6181 |
case RTA_FLOWLABEL: |
| 6182 |
break; |
| 6183 |
default: |
| 6184 |
NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request"); |
| 6185 |
return -EINVAL; |
| 6186 |
} |
| 6187 |
} |
| 6188 |
|
| 6189 |
return 0; |
| 6190 |
} |
| 6191 |
|
| 6192 |
static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh, |
| 6193 |
struct netlink_ext_ack *extack) |
| 6194 |
{ |
| 6195 |
struct net *net = sock_net(in_skb->sk); |
| 6196 |
struct nlattr *tb[RTA_MAX+1]; |
| 6197 |
int err, iif = 0, oif = 0; |
| 6198 |
struct fib6_info *from; |
| 6199 |
struct dst_entry *dst; |
| 6200 |
struct rt6_info *rt; |
| 6201 |
struct sk_buff *skb; |
| 6202 |
struct rtmsg *rtm; |
| 6203 |
struct flowi6 fl6 = {}; |
| 6204 |
__be32 flowlabel; |
| 6205 |
bool fibmatch; |
| 6206 |
|
| 6207 |
err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack); |
| 6208 |
if (err < 0) |
| 6209 |
goto errout; |
| 6210 |
|
| 6211 |
err = -EINVAL; |
| 6212 |
rtm = nlmsg_data(nlh); |
| 6213 |
fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH); |
| 6214 |
|
| 6215 |
if (tb[RTA_SRC]) { |
| 6216 |
if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr)) |
| 6217 |
goto errout; |
| 6218 |
|
| 6219 |
fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]); |
| 6220 |
} |
| 6221 |
|
| 6222 |
if (tb[RTA_DST]) { |
| 6223 |
if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr)) |
| 6224 |
goto errout; |
| 6225 |
|
| 6226 |
fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]); |
| 6227 |
} |
| 6228 |
|
| 6229 |
if (tb[RTA_IIF]) |
| 6230 |
iif = nla_get_u32(tb[RTA_IIF]); |
| 6231 |
|
| 6232 |
if (tb[RTA_OIF]) |
| 6233 |
oif = nla_get_u32(tb[RTA_OIF]); |
| 6234 |
|
| 6235 |
if (tb[RTA_MARK]) |
| 6236 |
fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]); |
| 6237 |
|
| 6238 |
if (tb[RTA_UID]) |
| 6239 |
fl6.flowi6_uid = make_kuid(current_user_ns(), |
| 6240 |
nla_get_u32(tb[RTA_UID])); |
| 6241 |
else |
| 6242 |
fl6.flowi6_uid = iif ? INVALID_UID : current_uid(); |
| 6243 |
|
| 6244 |
if (tb[RTA_SPORT]) |
| 6245 |
fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]); |
| 6246 |
|
| 6247 |
if (tb[RTA_DPORT]) |
| 6248 |
fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]); |
| 6249 |
|
| 6250 |
if (tb[RTA_IP_PROTO]) { |
| 6251 |
err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO], |
| 6252 |
&fl6.flowi6_proto, AF_INET6, |
| 6253 |
extack); |
| 6254 |
if (err) |
| 6255 |
goto errout; |
| 6256 |
} |
| 6257 |
|
| 6258 |
flowlabel = nla_get_be32_default(tb[RTA_FLOWLABEL], 0); |
| 6259 |
fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, flowlabel); |
| 6260 |
|
| 6261 |
if (iif) { |
| 6262 |
struct net_device *dev; |
| 6263 |
int flags = 0; |
| 6264 |
|
| 6265 |
rcu_read_lock(); |
| 6266 |
|
| 6267 |
dev = dev_get_by_index_rcu(net, iif); |
| 6268 |
if (!dev) { |
| 6269 |
rcu_read_unlock(); |
| 6270 |
err = -ENODEV; |
| 6271 |
goto errout; |
| 6272 |
} |
| 6273 |
|
| 6274 |
fl6.flowi6_iif = iif; |
| 6275 |
|
| 6276 |
if (!ipv6_addr_any(&fl6.saddr)) |
| 6277 |
flags |= RT6_LOOKUP_F_HAS_SADDR; |
| 6278 |
|
| 6279 |
dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags); |
| 6280 |
|
| 6281 |
rcu_read_unlock(); |
| 6282 |
} else { |
| 6283 |
fl6.flowi6_oif = oif; |
| 6284 |
|
| 6285 |
dst = ip6_route_output(net, NULL, &fl6); |
| 6286 |
} |
| 6287 |
|
| 6288 |
|
| 6289 |
rt = dst_rt6_info(dst); |
| 6290 |
if (rt->dst.error) { |
| 6291 |
err = rt->dst.error; |
| 6292 |
ip6_rt_put(rt); |
| 6293 |
goto errout; |
| 6294 |
} |
| 6295 |
|
| 6296 |
if (rt == net->ipv6.ip6_null_entry) { |
| 6297 |
err = rt->dst.error; |
| 6298 |
ip6_rt_put(rt); |
| 6299 |
goto errout; |
| 6300 |
} |
| 6301 |
|
| 6302 |
skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL); |
| 6303 |
if (!skb) { |
| 6304 |
ip6_rt_put(rt); |
| 6305 |
err = -ENOBUFS; |
| 6306 |
goto errout; |
| 6307 |
} |
| 6308 |
|
| 6309 |
skb_dst_set(skb, &rt->dst); |
| 6310 |
|
| 6311 |
rcu_read_lock(); |
| 6312 |
from = rcu_dereference(rt->from); |
| 6313 |
if (from) { |
| 6314 |
if (fibmatch) |
| 6315 |
err = rt6_fill_node(net, skb, from, NULL, NULL, NULL, |
| 6316 |
iif, RTM_NEWROUTE, |
| 6317 |
NETLINK_CB(in_skb).portid, |
| 6318 |
nlh->nlmsg_seq, 0); |
| 6319 |
else |
| 6320 |
err = rt6_fill_node(net, skb, from, dst, &fl6.daddr, |
| 6321 |
&fl6.saddr, iif, RTM_NEWROUTE, |
| 6322 |
NETLINK_CB(in_skb).portid, |
| 6323 |
nlh->nlmsg_seq, 0); |
| 6324 |
} else { |
| 6325 |
err = -ENETUNREACH; |
| 6326 |
} |
| 6327 |
rcu_read_unlock(); |
| 6328 |
|
| 6329 |
if (err < 0) { |
| 6330 |
kfree_skb(skb); |
| 6331 |
goto errout; |
| 6332 |
} |
| 6333 |
|
| 6334 |
err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); |
| 6335 |
errout: |
| 6336 |
return err; |
| 6337 |
} |
| 6338 |
|
| 6339 |
void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info, |
| 6340 |
unsigned int nlm_flags) |
| 6341 |
{ |
| 6342 |
struct net *net = info->nl_net; |
| 6343 |
struct sk_buff *skb; |
| 6344 |
size_t sz; |
| 6345 |
u32 seq; |
| 6346 |
int err; |
| 6347 |
|
| 6348 |
err = -ENOBUFS; |
| 6349 |
seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
| 6350 |
|
| 6351 |
rcu_read_lock(); |
| 6352 |
sz = rt6_nlmsg_size(rt); |
| 6353 |
retry: |
| 6354 |
skb = nlmsg_new(sz, GFP_ATOMIC); |
| 6355 |
if (!skb) |
| 6356 |
goto errout; |
| 6357 |
|
| 6358 |
err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0, |
| 6359 |
event, info->portid, seq, nlm_flags); |
| 6360 |
if (err < 0) { |
| 6361 |
kfree_skb(skb); |
| 6362 |
/* -EMSGSIZE implies needed space grew under us. */ |
| 6363 |
if (err == -EMSGSIZE) { |
| 6364 |
sz = max(rt6_nlmsg_size(rt), sz << 1); |
| 6365 |
goto retry; |
| 6366 |
} |
| 6367 |
goto errout; |
| 6368 |
} |
| 6369 |
|
| 6370 |
rcu_read_unlock(); |
| 6371 |
|
| 6372 |
rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE, |
| 6373 |
info->nlh, GFP_ATOMIC); |
| 6374 |
return; |
| 6375 |
errout: |
| 6376 |
rcu_read_unlock(); |
| 6377 |
rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err); |
| 6378 |
} |
| 6379 |
|
| 6380 |
void fib6_rt_update(struct net *net, struct fib6_info *rt, |
| 6381 |
struct nl_info *info) |
| 6382 |
{ |
| 6383 |
u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0; |
| 6384 |
struct sk_buff *skb; |
| 6385 |
int err = -ENOBUFS; |
| 6386 |
|
| 6387 |
skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any()); |
| 6388 |
if (!skb) |
| 6389 |
goto errout; |
| 6390 |
|
| 6391 |
err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0, |
| 6392 |
RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE); |
| 6393 |
if (err < 0) { |
| 6394 |
/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */ |
| 6395 |
WARN_ON(err == -EMSGSIZE); |
| 6396 |
kfree_skb(skb); |
| 6397 |
goto errout; |
| 6398 |
} |
| 6399 |
rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE, |
| 6400 |
info->nlh, gfp_any()); |
| 6401 |
return; |
| 6402 |
errout: |
| 6403 |
rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err); |
| 6404 |
} |
| 6405 |
|
| 6406 |
void fib6_info_hw_flags_set(struct net *net, struct fib6_info *f6i, |
| 6407 |
bool offload, bool trap, bool offload_failed) |
| 6408 |
{ |
| 6409 |
u8 fib_notify_on_flag_change; |
| 6410 |
struct sk_buff *skb; |
| 6411 |
int err; |
| 6412 |
|
| 6413 |
if (READ_ONCE(f6i->offload) == offload && |
| 6414 |
READ_ONCE(f6i->trap) == trap && |
| 6415 |
READ_ONCE(f6i->offload_failed) == offload_failed) |
| 6416 |
return; |
| 6417 |
|
| 6418 |
WRITE_ONCE(f6i->offload, offload); |
| 6419 |
WRITE_ONCE(f6i->trap, trap); |
| 6420 |
|
| 6421 |
fib_notify_on_flag_change = READ_ONCE(net->ipv6.sysctl.fib_notify_on_flag_change); |
| 6422 |
/* 2 means send notifications only if offload_failed was changed. */ |
| 6423 |
if (fib_notify_on_flag_change == 2 && |
| 6424 |
READ_ONCE(f6i->offload_failed) == offload_failed) |
| 6425 |
return; |
| 6426 |
|
| 6427 |
WRITE_ONCE(f6i->offload_failed, offload_failed); |
| 6428 |
|
| 6429 |
if (!rcu_access_pointer(f6i->fib6_node)) |
| 6430 |
/* The route was removed from the tree, do not send |
| 6431 |
* notification. |
| 6432 |
*/ |
| 6433 |
return; |
| 6434 |
|
| 6435 |
if (!fib_notify_on_flag_change) |
| 6436 |
return; |
| 6437 |
|
| 6438 |
skb = nlmsg_new(rt6_nlmsg_size(f6i), GFP_KERNEL); |
| 6439 |
if (!skb) { |
| 6440 |
err = -ENOBUFS; |
| 6441 |
goto errout; |
| 6442 |
} |
| 6443 |
|
| 6444 |
err = rt6_fill_node(net, skb, f6i, NULL, NULL, NULL, 0, RTM_NEWROUTE, 0, |
| 6445 |
0, 0); |
| 6446 |
if (err < 0) { |
| 6447 |
/* -EMSGSIZE implies BUG in rt6_nlmsg_size() */ |
| 6448 |
WARN_ON(err == -EMSGSIZE); |
| 6449 |
kfree_skb(skb); |
| 6450 |
goto errout; |
| 6451 |
} |
| 6452 |
|
| 6453 |
rtnl_notify(skb, net, 0, RTNLGRP_IPV6_ROUTE, NULL, GFP_KERNEL); |
| 6454 |
return; |
| 6455 |
|
| 6456 |
errout: |
| 6457 |
rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err); |
| 6458 |
} |
| 6459 |
EXPORT_SYMBOL(fib6_info_hw_flags_set); |
| 6460 |
|
| 6461 |
static int ip6_route_dev_notify(struct notifier_block *this, |
| 6462 |
unsigned long event, void *ptr) |
| 6463 |
{ |
| 6464 |
struct net_device *dev = netdev_notifier_info_to_dev(ptr); |
| 6465 |
struct net *net = dev_net(dev); |
| 6466 |
|
| 6467 |
if (!(dev->flags & IFF_LOOPBACK)) |
| 6468 |
return NOTIFY_OK; |
| 6469 |
|
| 6470 |
if (event == NETDEV_REGISTER) { |
| 6471 |
net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev; |
| 6472 |
net->ipv6.ip6_null_entry->dst.dev = dev; |
| 6473 |
net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev); |
| 6474 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6475 |
net->ipv6.ip6_prohibit_entry->dst.dev = dev; |
| 6476 |
net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev); |
| 6477 |
net->ipv6.ip6_blk_hole_entry->dst.dev = dev; |
| 6478 |
net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev); |
| 6479 |
#endif |
| 6480 |
} else if (event == NETDEV_UNREGISTER && |
| 6481 |
dev->reg_state != NETREG_UNREGISTERED) { |
| 6482 |
/* NETDEV_UNREGISTER could be fired for multiple times by |
| 6483 |
* netdev_wait_allrefs(). Make sure we only call this once. |
| 6484 |
*/ |
| 6485 |
in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev); |
| 6486 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6487 |
in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev); |
| 6488 |
in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev); |
| 6489 |
#endif |
| 6490 |
} |
| 6491 |
|
| 6492 |
return NOTIFY_OK; |
| 6493 |
} |
| 6494 |
|
| 6495 |
/* |
| 6496 |
* /proc |
| 6497 |
*/ |
| 6498 |
|
| 6499 |
#ifdef CONFIG_PROC_FS |
| 6500 |
static int rt6_stats_seq_show(struct seq_file *seq, void *v) |
| 6501 |
{ |
| 6502 |
struct net *net = (struct net *)seq->private; |
| 6503 |
seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n", |
| 6504 |
net->ipv6.rt6_stats->fib_nodes, |
| 6505 |
net->ipv6.rt6_stats->fib_route_nodes, |
| 6506 |
atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc), |
| 6507 |
net->ipv6.rt6_stats->fib_rt_entries, |
| 6508 |
net->ipv6.rt6_stats->fib_rt_cache, |
| 6509 |
dst_entries_get_slow(&net->ipv6.ip6_dst_ops), |
| 6510 |
net->ipv6.rt6_stats->fib_discarded_routes); |
| 6511 |
|
| 6512 |
return 0; |
| 6513 |
} |
| 6514 |
#endif /* CONFIG_PROC_FS */ |
| 6515 |
|
| 6516 |
#ifdef CONFIG_SYSCTL |
| 6517 |
|
| 6518 |
static int ipv6_sysctl_rtcache_flush(const struct ctl_table *ctl, int write, |
| 6519 |
void *buffer, size_t *lenp, loff_t *ppos) |
| 6520 |
{ |
| 6521 |
struct net *net; |
| 6522 |
int delay; |
| 6523 |
int ret; |
| 6524 |
if (!write) |
| 6525 |
return -EINVAL; |
| 6526 |
|
| 6527 |
ret = proc_dointvec(ctl, write, buffer, lenp, ppos); |
| 6528 |
if (ret) |
| 6529 |
return ret; |
| 6530 |
|
| 6531 |
net = (struct net *)ctl->extra1; |
| 6532 |
delay = READ_ONCE(net->ipv6.sysctl.flush_delay); |
| 6533 |
fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0); |
| 6534 |
return 0; |
| 6535 |
} |
| 6536 |
|
| 6537 |
static struct ctl_table ipv6_route_table_template[] = { |
| 6538 |
{ |
| 6539 |
.procname = "max_size", |
| 6540 |
.data = &init_net.ipv6.sysctl.ip6_rt_max_size, |
| 6541 |
.maxlen = sizeof(int), |
| 6542 |
.mode = 0644, |
| 6543 |
.proc_handler = proc_dointvec, |
| 6544 |
}, |
| 6545 |
{ |
| 6546 |
.procname = "gc_thresh", |
| 6547 |
.data = &ip6_dst_ops_template.gc_thresh, |
| 6548 |
.maxlen = sizeof(int), |
| 6549 |
.mode = 0644, |
| 6550 |
.proc_handler = proc_dointvec, |
| 6551 |
}, |
| 6552 |
{ |
| 6553 |
.procname = "flush", |
| 6554 |
.data = &init_net.ipv6.sysctl.flush_delay, |
| 6555 |
.maxlen = sizeof(int), |
| 6556 |
.mode = 0200, |
| 6557 |
.proc_handler = ipv6_sysctl_rtcache_flush |
| 6558 |
}, |
| 6559 |
{ |
| 6560 |
.procname = "gc_min_interval", |
| 6561 |
.data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval, |
| 6562 |
.maxlen = sizeof(int), |
| 6563 |
.mode = 0644, |
| 6564 |
.proc_handler = proc_dointvec_jiffies, |
| 6565 |
}, |
| 6566 |
{ |
| 6567 |
.procname = "gc_timeout", |
| 6568 |
.data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout, |
| 6569 |
.maxlen = sizeof(int), |
| 6570 |
.mode = 0644, |
| 6571 |
.proc_handler = proc_dointvec_jiffies, |
| 6572 |
}, |
| 6573 |
{ |
| 6574 |
.procname = "gc_interval", |
| 6575 |
.data = &init_net.ipv6.sysctl.ip6_rt_gc_interval, |
| 6576 |
.maxlen = sizeof(int), |
| 6577 |
.mode = 0644, |
| 6578 |
.proc_handler = proc_dointvec_jiffies, |
| 6579 |
}, |
| 6580 |
{ |
| 6581 |
.procname = "gc_elasticity", |
| 6582 |
.data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity, |
| 6583 |
.maxlen = sizeof(int), |
| 6584 |
.mode = 0644, |
| 6585 |
.proc_handler = proc_dointvec, |
| 6586 |
}, |
| 6587 |
{ |
| 6588 |
.procname = "mtu_expires", |
| 6589 |
.data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires, |
| 6590 |
.maxlen = sizeof(int), |
| 6591 |
.mode = 0644, |
| 6592 |
.proc_handler = proc_dointvec_jiffies, |
| 6593 |
}, |
| 6594 |
{ |
| 6595 |
.procname = "min_adv_mss", |
| 6596 |
.data = &init_net.ipv6.sysctl.ip6_rt_min_advmss, |
| 6597 |
.maxlen = sizeof(int), |
| 6598 |
.mode = 0644, |
| 6599 |
.proc_handler = proc_dointvec, |
| 6600 |
}, |
| 6601 |
{ |
| 6602 |
.procname = "gc_min_interval_ms", |
| 6603 |
.data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval, |
| 6604 |
.maxlen = sizeof(int), |
| 6605 |
.mode = 0644, |
| 6606 |
.proc_handler = proc_dointvec_ms_jiffies, |
| 6607 |
}, |
| 6608 |
{ |
| 6609 |
.procname = "skip_notify_on_dev_down", |
| 6610 |
.data = &init_net.ipv6.sysctl.skip_notify_on_dev_down, |
| 6611 |
.maxlen = sizeof(u8), |
| 6612 |
.mode = 0644, |
| 6613 |
.proc_handler = proc_dou8vec_minmax, |
| 6614 |
.extra1 = SYSCTL_ZERO, |
| 6615 |
.extra2 = SYSCTL_ONE, |
| 6616 |
}, |
| 6617 |
}; |
| 6618 |
|
| 6619 |
struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net) |
| 6620 |
{ |
| 6621 |
struct ctl_table *table; |
| 6622 |
|
| 6623 |
table = kmemdup(ipv6_route_table_template, |
| 6624 |
sizeof(ipv6_route_table_template), |
| 6625 |
GFP_KERNEL); |
| 6626 |
|
| 6627 |
if (table) { |
| 6628 |
table[0].data = &net->ipv6.sysctl.ip6_rt_max_size; |
| 6629 |
table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh; |
| 6630 |
table[2].data = &net->ipv6.sysctl.flush_delay; |
| 6631 |
table[2].extra1 = net; |
| 6632 |
table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval; |
| 6633 |
table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout; |
| 6634 |
table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval; |
| 6635 |
table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity; |
| 6636 |
table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires; |
| 6637 |
table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss; |
| 6638 |
table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval; |
| 6639 |
table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down; |
| 6640 |
} |
| 6641 |
|
| 6642 |
return table; |
| 6643 |
} |
| 6644 |
|
| 6645 |
size_t ipv6_route_sysctl_table_size(struct net *net) |
| 6646 |
{ |
| 6647 |
/* Don't export sysctls to unprivileged users */ |
| 6648 |
if (net->user_ns != &init_user_ns) |
| 6649 |
return 1; |
| 6650 |
|
| 6651 |
return ARRAY_SIZE(ipv6_route_table_template); |
| 6652 |
} |
| 6653 |
#endif |
| 6654 |
|
| 6655 |
static int __net_init ip6_route_net_init(struct net *net) |
| 6656 |
{ |
| 6657 |
int ret = -ENOMEM; |
| 6658 |
|
| 6659 |
memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template, |
| 6660 |
sizeof(net->ipv6.ip6_dst_ops)); |
| 6661 |
|
| 6662 |
if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0) |
| 6663 |
goto out_ip6_dst_ops; |
| 6664 |
|
| 6665 |
net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true); |
| 6666 |
if (!net->ipv6.fib6_null_entry) |
| 6667 |
goto out_ip6_dst_entries; |
| 6668 |
memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template, |
| 6669 |
sizeof(*net->ipv6.fib6_null_entry)); |
| 6670 |
|
| 6671 |
net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template, |
| 6672 |
sizeof(*net->ipv6.ip6_null_entry), |
| 6673 |
GFP_KERNEL); |
| 6674 |
if (!net->ipv6.ip6_null_entry) |
| 6675 |
goto out_fib6_null_entry; |
| 6676 |
net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops; |
| 6677 |
dst_init_metrics(&net->ipv6.ip6_null_entry->dst, |
| 6678 |
ip6_template_metrics, true); |
| 6679 |
INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->dst.rt_uncached); |
| 6680 |
|
| 6681 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6682 |
net->ipv6.fib6_has_custom_rules = false; |
| 6683 |
net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template, |
| 6684 |
sizeof(*net->ipv6.ip6_prohibit_entry), |
| 6685 |
GFP_KERNEL); |
| 6686 |
if (!net->ipv6.ip6_prohibit_entry) |
| 6687 |
goto out_ip6_null_entry; |
| 6688 |
net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops; |
| 6689 |
dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst, |
| 6690 |
ip6_template_metrics, true); |
| 6691 |
INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->dst.rt_uncached); |
| 6692 |
|
| 6693 |
net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template, |
| 6694 |
sizeof(*net->ipv6.ip6_blk_hole_entry), |
| 6695 |
GFP_KERNEL); |
| 6696 |
if (!net->ipv6.ip6_blk_hole_entry) |
| 6697 |
goto out_ip6_prohibit_entry; |
| 6698 |
net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops; |
| 6699 |
dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst, |
| 6700 |
ip6_template_metrics, true); |
| 6701 |
INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->dst.rt_uncached); |
| 6702 |
#ifdef CONFIG_IPV6_SUBTREES |
| 6703 |
net->ipv6.fib6_routes_require_src = 0; |
| 6704 |
#endif |
| 6705 |
#endif |
| 6706 |
|
| 6707 |
net->ipv6.sysctl.flush_delay = 0; |
| 6708 |
net->ipv6.sysctl.ip6_rt_max_size = INT_MAX; |
| 6709 |
net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2; |
| 6710 |
net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ; |
| 6711 |
net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ; |
| 6712 |
net->ipv6.sysctl.ip6_rt_gc_elasticity = 9; |
| 6713 |
net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ; |
| 6714 |
net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40; |
| 6715 |
net->ipv6.sysctl.skip_notify_on_dev_down = 0; |
| 6716 |
|
| 6717 |
atomic_set(&net->ipv6.ip6_rt_gc_expire, 30*HZ); |
| 6718 |
|
| 6719 |
ret = 0; |
| 6720 |
out: |
| 6721 |
return ret; |
| 6722 |
|
| 6723 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6724 |
out_ip6_prohibit_entry: |
| 6725 |
kfree(net->ipv6.ip6_prohibit_entry); |
| 6726 |
out_ip6_null_entry: |
| 6727 |
kfree(net->ipv6.ip6_null_entry); |
| 6728 |
#endif |
| 6729 |
out_fib6_null_entry: |
| 6730 |
kfree(net->ipv6.fib6_null_entry); |
| 6731 |
out_ip6_dst_entries: |
| 6732 |
dst_entries_destroy(&net->ipv6.ip6_dst_ops); |
| 6733 |
out_ip6_dst_ops: |
| 6734 |
goto out; |
| 6735 |
} |
| 6736 |
|
| 6737 |
static void __net_exit ip6_route_net_exit(struct net *net) |
| 6738 |
{ |
| 6739 |
kfree(net->ipv6.fib6_null_entry); |
| 6740 |
kfree(net->ipv6.ip6_null_entry); |
| 6741 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6742 |
kfree(net->ipv6.ip6_prohibit_entry); |
| 6743 |
kfree(net->ipv6.ip6_blk_hole_entry); |
| 6744 |
#endif |
| 6745 |
dst_entries_destroy(&net->ipv6.ip6_dst_ops); |
| 6746 |
} |
| 6747 |
|
| 6748 |
static int __net_init ip6_route_net_init_late(struct net *net) |
| 6749 |
{ |
| 6750 |
#ifdef CONFIG_PROC_FS |
| 6751 |
if (!proc_create_net("ipv6_route", 0, net->proc_net, |
| 6752 |
&ipv6_route_seq_ops, |
| 6753 |
sizeof(struct ipv6_route_iter))) |
| 6754 |
return -ENOMEM; |
| 6755 |
|
| 6756 |
if (!proc_create_net_single("rt6_stats", 0444, net->proc_net, |
| 6757 |
rt6_stats_seq_show, NULL)) { |
| 6758 |
remove_proc_entry("ipv6_route", net->proc_net); |
| 6759 |
return -ENOMEM; |
| 6760 |
} |
| 6761 |
#endif |
| 6762 |
return 0; |
| 6763 |
} |
| 6764 |
|
| 6765 |
static void __net_exit ip6_route_net_exit_late(struct net *net) |
| 6766 |
{ |
| 6767 |
#ifdef CONFIG_PROC_FS |
| 6768 |
remove_proc_entry("ipv6_route", net->proc_net); |
| 6769 |
remove_proc_entry("rt6_stats", net->proc_net); |
| 6770 |
#endif |
| 6771 |
} |
| 6772 |
|
| 6773 |
static struct pernet_operations ip6_route_net_ops = { |
| 6774 |
.init = ip6_route_net_init, |
| 6775 |
.exit = ip6_route_net_exit, |
| 6776 |
}; |
| 6777 |
|
| 6778 |
static int __net_init ipv6_inetpeer_init(struct net *net) |
| 6779 |
{ |
| 6780 |
struct inet_peer_base *bp = kmalloc_obj(*bp); |
| 6781 |
|
| 6782 |
if (!bp) |
| 6783 |
return -ENOMEM; |
| 6784 |
inet_peer_base_init(bp); |
| 6785 |
net->ipv6.peers = bp; |
| 6786 |
return 0; |
| 6787 |
} |
| 6788 |
|
| 6789 |
static void __net_exit ipv6_inetpeer_exit(struct net *net) |
| 6790 |
{ |
| 6791 |
struct inet_peer_base *bp = net->ipv6.peers; |
| 6792 |
|
| 6793 |
net->ipv6.peers = NULL; |
| 6794 |
inetpeer_invalidate_tree(bp); |
| 6795 |
kfree(bp); |
| 6796 |
} |
| 6797 |
|
| 6798 |
static struct pernet_operations ipv6_inetpeer_ops = { |
| 6799 |
.init = ipv6_inetpeer_init, |
| 6800 |
.exit = ipv6_inetpeer_exit, |
| 6801 |
}; |
| 6802 |
|
| 6803 |
static struct pernet_operations ip6_route_net_late_ops = { |
| 6804 |
.init = ip6_route_net_init_late, |
| 6805 |
.exit = ip6_route_net_exit_late, |
| 6806 |
}; |
| 6807 |
|
| 6808 |
static struct notifier_block ip6_route_dev_notifier = { |
| 6809 |
.notifier_call = ip6_route_dev_notify, |
| 6810 |
.priority = ADDRCONF_NOTIFY_PRIORITY - 10, |
| 6811 |
}; |
| 6812 |
|
| 6813 |
void __init ip6_route_init_special_entries(void) |
| 6814 |
{ |
| 6815 |
/* Registering of the loopback is done before this portion of code, |
| 6816 |
* the loopback reference in rt6_info will not be taken, do it |
| 6817 |
* manually for init_net */ |
| 6818 |
init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev; |
| 6819 |
init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev; |
| 6820 |
init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev); |
| 6821 |
#ifdef CONFIG_IPV6_MULTIPLE_TABLES |
| 6822 |
init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev; |
| 6823 |
init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev); |
| 6824 |
init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev; |
| 6825 |
init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev); |
| 6826 |
#endif |
| 6827 |
} |
| 6828 |
|
| 6829 |
#if IS_BUILTIN(CONFIG_IPV6) |
| 6830 |
#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS) |
| 6831 |
DEFINE_BPF_ITER_FUNC(ipv6_route, struct bpf_iter_meta *meta, struct fib6_info *rt) |
| 6832 |
|
| 6833 |
BTF_ID_LIST_SINGLE(btf_fib6_info_id, struct, fib6_info) |
| 6834 |
|
| 6835 |
static const struct bpf_iter_seq_info ipv6_route_seq_info = { |
| 6836 |
.seq_ops = &ipv6_route_seq_ops, |
| 6837 |
.init_seq_private = bpf_iter_init_seq_net, |
| 6838 |
.fini_seq_private = bpf_iter_fini_seq_net, |
| 6839 |
.seq_priv_size = sizeof(struct ipv6_route_iter), |
| 6840 |
}; |
| 6841 |
|
| 6842 |
static struct bpf_iter_reg ipv6_route_reg_info = { |
| 6843 |
.target = "ipv6_route", |
| 6844 |
.ctx_arg_info_size = 1, |
| 6845 |
.ctx_arg_info = { |
| 6846 |
{ offsetof(struct bpf_iter__ipv6_route, rt), |
| 6847 |
PTR_TO_BTF_ID_OR_NULL }, |
| 6848 |
}, |
| 6849 |
.seq_info = &ipv6_route_seq_info, |
| 6850 |
}; |
| 6851 |
|
| 6852 |
static int __init bpf_iter_register(void) |
| 6853 |
{ |
| 6854 |
ipv6_route_reg_info.ctx_arg_info[0].btf_id = *btf_fib6_info_id; |
| 6855 |
return bpf_iter_reg_target(&ipv6_route_reg_info); |
| 6856 |
} |
| 6857 |
|
| 6858 |
static void bpf_iter_unregister(void) |
| 6859 |
{ |
| 6860 |
bpf_iter_unreg_target(&ipv6_route_reg_info); |
| 6861 |
} |
| 6862 |
#endif |
| 6863 |
#endif |
| 6864 |
|
| 6865 |
static const struct rtnl_msg_handler ip6_route_rtnl_msg_handlers[] __initconst_or_module = { |
| 6866 |
{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWROUTE, |
| 6867 |
.doit = inet6_rtm_newroute, .flags = RTNL_FLAG_DOIT_UNLOCKED}, |
| 6868 |
{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELROUTE, |
| 6869 |
.doit = inet6_rtm_delroute, .flags = RTNL_FLAG_DOIT_UNLOCKED}, |
| 6870 |
{.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETROUTE, |
| 6871 |
.doit = inet6_rtm_getroute, .flags = RTNL_FLAG_DOIT_UNLOCKED}, |
| 6872 |
}; |
| 6873 |
|
| 6874 |
int __init ip6_route_init(void) |
| 6875 |
{ |
| 6876 |
int ret; |
| 6877 |
int cpu; |
| 6878 |
|
| 6879 |
ret = -ENOMEM; |
| 6880 |
ip6_dst_ops_template.kmem_cachep = |
| 6881 |
kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0, |
| 6882 |
SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT, NULL); |
| 6883 |
if (!ip6_dst_ops_template.kmem_cachep) |
| 6884 |
goto out; |
| 6885 |
|
| 6886 |
ret = dst_entries_init(&ip6_dst_blackhole_ops); |
| 6887 |
if (ret) |
| 6888 |
goto out_kmem_cache; |
| 6889 |
|
| 6890 |
ret = register_pernet_subsys(&ipv6_inetpeer_ops); |
| 6891 |
if (ret) |
| 6892 |
goto out_dst_entries; |
| 6893 |
|
| 6894 |
ret = register_pernet_subsys(&ip6_route_net_ops); |
| 6895 |
if (ret) |
| 6896 |
goto out_register_inetpeer; |
| 6897 |
|
| 6898 |
ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep; |
| 6899 |
|
| 6900 |
ret = fib6_init(); |
| 6901 |
if (ret) |
| 6902 |
goto out_register_subsys; |
| 6903 |
|
| 6904 |
ret = xfrm6_init(); |
| 6905 |
if (ret) |
| 6906 |
goto out_fib6_init; |
| 6907 |
|
| 6908 |
ret = fib6_rules_init(); |
| 6909 |
if (ret) |
| 6910 |
goto xfrm6_init; |
| 6911 |
|
| 6912 |
ret = register_pernet_subsys(&ip6_route_net_late_ops); |
| 6913 |
if (ret) |
| 6914 |
goto fib6_rules_init; |
| 6915 |
|
| 6916 |
ret = rtnl_register_many(ip6_route_rtnl_msg_handlers); |
| 6917 |
if (ret < 0) |
| 6918 |
goto out_register_late_subsys; |
| 6919 |
|
| 6920 |
ret = register_netdevice_notifier(&ip6_route_dev_notifier); |
| 6921 |
if (ret) |
| 6922 |
goto out_register_late_subsys; |
| 6923 |
|
| 6924 |
#if IS_BUILTIN(CONFIG_IPV6) |
| 6925 |
#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS) |
| 6926 |
ret = bpf_iter_register(); |
| 6927 |
if (ret) |
| 6928 |
goto out_register_late_subsys; |
| 6929 |
#endif |
| 6930 |
#endif |
| 6931 |
|
| 6932 |
for_each_possible_cpu(cpu) { |
| 6933 |
struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu); |
| 6934 |
|
| 6935 |
INIT_LIST_HEAD(&ul->head); |
| 6936 |
spin_lock_init(&ul->lock); |
| 6937 |
} |
| 6938 |
|
| 6939 |
out: |
| 6940 |
return ret; |
| 6941 |
|
| 6942 |
out_register_late_subsys: |
| 6943 |
rtnl_unregister_all(PF_INET6); |
| 6944 |
unregister_pernet_subsys(&ip6_route_net_late_ops); |
| 6945 |
fib6_rules_init: |
| 6946 |
fib6_rules_cleanup(); |
| 6947 |
xfrm6_init: |
| 6948 |
xfrm6_fini(); |
| 6949 |
out_fib6_init: |
| 6950 |
fib6_gc_cleanup(); |
| 6951 |
out_register_subsys: |
| 6952 |
unregister_pernet_subsys(&ip6_route_net_ops); |
| 6953 |
out_register_inetpeer: |
| 6954 |
unregister_pernet_subsys(&ipv6_inetpeer_ops); |
| 6955 |
out_dst_entries: |
| 6956 |
dst_entries_destroy(&ip6_dst_blackhole_ops); |
| 6957 |
out_kmem_cache: |
| 6958 |
kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep); |
| 6959 |
goto out; |
| 6960 |
} |
| 6961 |
|
| 6962 |
void ip6_route_cleanup(void) |
| 6963 |
{ |
| 6964 |
#if IS_BUILTIN(CONFIG_IPV6) |
| 6965 |
#if defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_PROC_FS) |
| 6966 |
bpf_iter_unregister(); |
| 6967 |
#endif |
| 6968 |
#endif |
| 6969 |
unregister_netdevice_notifier(&ip6_route_dev_notifier); |
| 6970 |
unregister_pernet_subsys(&ip6_route_net_late_ops); |
| 6971 |
fib6_rules_cleanup(); |
| 6972 |
xfrm6_fini(); |
| 6973 |
fib6_gc_cleanup(); |
| 6974 |
unregister_pernet_subsys(&ipv6_inetpeer_ops); |
| 6975 |
unregister_pernet_subsys(&ip6_route_net_ops); |
| 6976 |
dst_entries_destroy(&ip6_dst_blackhole_ops); |
| 6977 |
kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep); |
| 6978 |
} |