| 1 |
// SPDX-License-Identifier: GPL-2.0-or-later |
| 2 |
/* |
| 3 |
* ip6_flowlabel.c IPv6 flowlabel manager. |
| 4 |
* |
| 5 |
* Authors: Alexey Kuznetsov, <[email protected]> |
| 6 |
*/ |
| 7 |
|
| 8 |
#include <linux/capability.h> |
| 9 |
#include <linux/errno.h> |
| 10 |
#include <linux/types.h> |
| 11 |
#include <linux/socket.h> |
| 12 |
#include <linux/net.h> |
| 13 |
#include <linux/netdevice.h> |
| 14 |
#include <linux/in6.h> |
| 15 |
#include <linux/proc_fs.h> |
| 16 |
#include <linux/seq_file.h> |
| 17 |
#include <linux/slab.h> |
| 18 |
#include <linux/export.h> |
| 19 |
#include <linux/pid_namespace.h> |
| 20 |
#include <linux/jump_label_ratelimit.h> |
| 21 |
|
| 22 |
#include <net/net_namespace.h> |
| 23 |
#include <net/sock.h> |
| 24 |
|
| 25 |
#include <net/ipv6.h> |
| 26 |
#include <net/rawv6.h> |
| 27 |
#include <net/transp_v6.h> |
| 28 |
|
| 29 |
#include <linux/uaccess.h> |
| 30 |
|
| 31 |
#define FL_MIN_LINGER 6 /* Minimal linger. It is set to 6sec specified |
| 32 |
in old IPv6 RFC. Well, it was reasonable value. |
| 33 |
*/ |
| 34 |
#define FL_MAX_LINGER 150 /* Maximal linger timeout */ |
| 35 |
|
| 36 |
/* FL hash table */ |
| 37 |
|
| 38 |
#define FL_MAX_PER_SOCK 32 |
| 39 |
#define FL_MAX_SIZE 4096 |
| 40 |
#define FL_HASH_MASK 255 |
| 41 |
#define FL_HASH(l) (ntohl(l)&FL_HASH_MASK) |
| 42 |
|
| 43 |
static atomic_t fl_size = ATOMIC_INIT(0); |
| 44 |
static struct ip6_flowlabel __rcu *fl_ht[FL_HASH_MASK+1]; |
| 45 |
|
| 46 |
static void ip6_fl_gc(struct timer_list *unused); |
| 47 |
static DEFINE_TIMER(ip6_fl_gc_timer, ip6_fl_gc); |
| 48 |
|
| 49 |
/* FL hash table lock: it protects only of GC */ |
| 50 |
|
| 51 |
static DEFINE_SPINLOCK(ip6_fl_lock); |
| 52 |
|
| 53 |
/* Big socket sock */ |
| 54 |
|
| 55 |
static DEFINE_SPINLOCK(ip6_sk_fl_lock); |
| 56 |
|
| 57 |
DEFINE_STATIC_KEY_DEFERRED_FALSE(ipv6_flowlabel_exclusive, HZ); |
| 58 |
EXPORT_SYMBOL(ipv6_flowlabel_exclusive); |
| 59 |
|
| 60 |
#define for_each_fl_rcu(hash, fl) \ |
| 61 |
for (fl = rcu_dereference(fl_ht[(hash)]); \ |
| 62 |
fl != NULL; \ |
| 63 |
fl = rcu_dereference(fl->next)) |
| 64 |
#define for_each_fl_continue_rcu(fl) \ |
| 65 |
for (fl = rcu_dereference(fl->next); \ |
| 66 |
fl != NULL; \ |
| 67 |
fl = rcu_dereference(fl->next)) |
| 68 |
|
| 69 |
#define for_each_sk_fl_rcu(sk, sfl) \ |
| 70 |
for (sfl = rcu_dereference(inet_sk(sk)->ipv6_fl_list); \ |
| 71 |
sfl != NULL; \ |
| 72 |
sfl = rcu_dereference(sfl->next)) |
| 73 |
|
| 74 |
static inline struct ip6_flowlabel *__fl_lookup(struct net *net, __be32 label) |
| 75 |
{ |
| 76 |
struct ip6_flowlabel *fl; |
| 77 |
|
| 78 |
for_each_fl_rcu(FL_HASH(label), fl) { |
| 79 |
if (fl->label == label && net_eq(fl->fl_net, net)) |
| 80 |
return fl; |
| 81 |
} |
| 82 |
return NULL; |
| 83 |
} |
| 84 |
|
| 85 |
static struct ip6_flowlabel *fl_lookup(struct net *net, __be32 label) |
| 86 |
{ |
| 87 |
struct ip6_flowlabel *fl; |
| 88 |
|
| 89 |
rcu_read_lock(); |
| 90 |
fl = __fl_lookup(net, label); |
| 91 |
if (fl && !atomic_inc_not_zero(&fl->users)) |
| 92 |
fl = NULL; |
| 93 |
rcu_read_unlock(); |
| 94 |
return fl; |
| 95 |
} |
| 96 |
|
| 97 |
static bool fl_shared_exclusive(struct ip6_flowlabel *fl) |
| 98 |
{ |
| 99 |
return fl->share == IPV6_FL_S_EXCL || |
| 100 |
fl->share == IPV6_FL_S_PROCESS || |
| 101 |
fl->share == IPV6_FL_S_USER; |
| 102 |
} |
| 103 |
|
| 104 |
static void fl_free_rcu(struct rcu_head *head) |
| 105 |
{ |
| 106 |
struct ip6_flowlabel *fl = container_of(head, struct ip6_flowlabel, rcu); |
| 107 |
|
| 108 |
if (fl->share == IPV6_FL_S_PROCESS) |
| 109 |
put_pid(fl->owner.pid); |
| 110 |
kfree(fl->opt); |
| 111 |
kfree(fl); |
| 112 |
} |
| 113 |
|
| 114 |
|
| 115 |
static void fl_free(struct ip6_flowlabel *fl) |
| 116 |
{ |
| 117 |
if (!fl) |
| 118 |
return; |
| 119 |
|
| 120 |
if (fl_shared_exclusive(fl) || fl->opt) |
| 121 |
static_branch_slow_dec_deferred(&ipv6_flowlabel_exclusive); |
| 122 |
|
| 123 |
call_rcu(&fl->rcu, fl_free_rcu); |
| 124 |
} |
| 125 |
|
| 126 |
static void fl_release(struct ip6_flowlabel *fl) |
| 127 |
{ |
| 128 |
spin_lock_bh(&ip6_fl_lock); |
| 129 |
|
| 130 |
fl->lastuse = jiffies; |
| 131 |
if (atomic_dec_and_test(&fl->users)) { |
| 132 |
unsigned long ttd = fl->lastuse + fl->linger; |
| 133 |
if (time_after(ttd, fl->expires)) |
| 134 |
fl->expires = ttd; |
| 135 |
ttd = fl->expires; |
| 136 |
if (!timer_pending(&ip6_fl_gc_timer) || |
| 137 |
time_after(ip6_fl_gc_timer.expires, ttd)) |
| 138 |
mod_timer(&ip6_fl_gc_timer, ttd); |
| 139 |
} |
| 140 |
spin_unlock_bh(&ip6_fl_lock); |
| 141 |
} |
| 142 |
|
| 143 |
static void ip6_fl_gc(struct timer_list *unused) |
| 144 |
{ |
| 145 |
int i; |
| 146 |
unsigned long now = jiffies; |
| 147 |
unsigned long sched = 0; |
| 148 |
|
| 149 |
spin_lock(&ip6_fl_lock); |
| 150 |
|
| 151 |
for (i = 0; i <= FL_HASH_MASK; i++) { |
| 152 |
struct ip6_flowlabel *fl; |
| 153 |
struct ip6_flowlabel __rcu **flp; |
| 154 |
|
| 155 |
flp = &fl_ht[i]; |
| 156 |
while ((fl = rcu_dereference_protected(*flp, |
| 157 |
lockdep_is_held(&ip6_fl_lock))) != NULL) { |
| 158 |
if (atomic_read(&fl->users) == 0) { |
| 159 |
unsigned long ttd = fl->lastuse + fl->linger; |
| 160 |
if (time_after(ttd, fl->expires)) |
| 161 |
fl->expires = ttd; |
| 162 |
ttd = fl->expires; |
| 163 |
if (time_after_eq(now, ttd)) { |
| 164 |
*flp = fl->next; |
| 165 |
fl_free(fl); |
| 166 |
atomic_dec(&fl_size); |
| 167 |
continue; |
| 168 |
} |
| 169 |
if (!sched || time_before(ttd, sched)) |
| 170 |
sched = ttd; |
| 171 |
} |
| 172 |
flp = &fl->next; |
| 173 |
} |
| 174 |
} |
| 175 |
if (!sched && atomic_read(&fl_size)) |
| 176 |
sched = now + FL_MAX_LINGER; |
| 177 |
if (sched) { |
| 178 |
mod_timer(&ip6_fl_gc_timer, sched); |
| 179 |
} |
| 180 |
spin_unlock(&ip6_fl_lock); |
| 181 |
} |
| 182 |
|
| 183 |
static void __net_exit ip6_fl_purge(struct net *net) |
| 184 |
{ |
| 185 |
int i; |
| 186 |
|
| 187 |
spin_lock_bh(&ip6_fl_lock); |
| 188 |
for (i = 0; i <= FL_HASH_MASK; i++) { |
| 189 |
struct ip6_flowlabel *fl; |
| 190 |
struct ip6_flowlabel __rcu **flp; |
| 191 |
|
| 192 |
flp = &fl_ht[i]; |
| 193 |
while ((fl = rcu_dereference_protected(*flp, |
| 194 |
lockdep_is_held(&ip6_fl_lock))) != NULL) { |
| 195 |
if (net_eq(fl->fl_net, net) && |
| 196 |
atomic_read(&fl->users) == 0) { |
| 197 |
*flp = fl->next; |
| 198 |
fl_free(fl); |
| 199 |
atomic_dec(&fl_size); |
| 200 |
continue; |
| 201 |
} |
| 202 |
flp = &fl->next; |
| 203 |
} |
| 204 |
} |
| 205 |
spin_unlock_bh(&ip6_fl_lock); |
| 206 |
} |
| 207 |
|
| 208 |
static struct ip6_flowlabel *fl_intern(struct net *net, |
| 209 |
struct ip6_flowlabel *fl, __be32 label) |
| 210 |
{ |
| 211 |
struct ip6_flowlabel *lfl; |
| 212 |
|
| 213 |
fl->label = label & IPV6_FLOWLABEL_MASK; |
| 214 |
|
| 215 |
rcu_read_lock(); |
| 216 |
spin_lock_bh(&ip6_fl_lock); |
| 217 |
if (label == 0) { |
| 218 |
for (;;) { |
| 219 |
fl->label = htonl(get_random_u32())&IPV6_FLOWLABEL_MASK; |
| 220 |
if (fl->label) { |
| 221 |
lfl = __fl_lookup(net, fl->label); |
| 222 |
if (!lfl) |
| 223 |
break; |
| 224 |
} |
| 225 |
} |
| 226 |
} else { |
| 227 |
/* |
| 228 |
* we dropper the ip6_fl_lock, so this entry could reappear |
| 229 |
* and we need to recheck with it. |
| 230 |
* |
| 231 |
* OTOH no need to search the active socket first, like it is |
| 232 |
* done in ipv6_flowlabel_opt - sock is locked, so new entry |
| 233 |
* with the same label can only appear on another sock |
| 234 |
*/ |
| 235 |
lfl = __fl_lookup(net, fl->label); |
| 236 |
if (lfl) { |
| 237 |
atomic_inc(&lfl->users); |
| 238 |
spin_unlock_bh(&ip6_fl_lock); |
| 239 |
rcu_read_unlock(); |
| 240 |
return lfl; |
| 241 |
} |
| 242 |
} |
| 243 |
|
| 244 |
fl->lastuse = jiffies; |
| 245 |
fl->next = fl_ht[FL_HASH(fl->label)]; |
| 246 |
rcu_assign_pointer(fl_ht[FL_HASH(fl->label)], fl); |
| 247 |
atomic_inc(&fl_size); |
| 248 |
spin_unlock_bh(&ip6_fl_lock); |
| 249 |
rcu_read_unlock(); |
| 250 |
return NULL; |
| 251 |
} |
| 252 |
|
| 253 |
|
| 254 |
|
| 255 |
/* Socket flowlabel lists */ |
| 256 |
|
| 257 |
struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label) |
| 258 |
{ |
| 259 |
struct ipv6_fl_socklist *sfl; |
| 260 |
|
| 261 |
label &= IPV6_FLOWLABEL_MASK; |
| 262 |
|
| 263 |
rcu_read_lock(); |
| 264 |
for_each_sk_fl_rcu(sk, sfl) { |
| 265 |
struct ip6_flowlabel *fl = sfl->fl; |
| 266 |
|
| 267 |
if (fl->label == label && atomic_inc_not_zero(&fl->users)) { |
| 268 |
fl->lastuse = jiffies; |
| 269 |
rcu_read_unlock(); |
| 270 |
return fl; |
| 271 |
} |
| 272 |
} |
| 273 |
rcu_read_unlock(); |
| 274 |
return NULL; |
| 275 |
} |
| 276 |
EXPORT_SYMBOL_GPL(__fl6_sock_lookup); |
| 277 |
|
| 278 |
void fl6_free_socklist(struct sock *sk) |
| 279 |
{ |
| 280 |
struct inet_sock *inet = inet_sk(sk); |
| 281 |
struct ipv6_fl_socklist *sfl; |
| 282 |
|
| 283 |
if (!rcu_access_pointer(inet->ipv6_fl_list)) |
| 284 |
return; |
| 285 |
|
| 286 |
spin_lock_bh(&ip6_sk_fl_lock); |
| 287 |
while ((sfl = rcu_dereference_protected(inet->ipv6_fl_list, |
| 288 |
lockdep_is_held(&ip6_sk_fl_lock))) != NULL) { |
| 289 |
inet->ipv6_fl_list = sfl->next; |
| 290 |
spin_unlock_bh(&ip6_sk_fl_lock); |
| 291 |
|
| 292 |
fl_release(sfl->fl); |
| 293 |
kfree_rcu(sfl, rcu); |
| 294 |
|
| 295 |
spin_lock_bh(&ip6_sk_fl_lock); |
| 296 |
} |
| 297 |
spin_unlock_bh(&ip6_sk_fl_lock); |
| 298 |
} |
| 299 |
|
| 300 |
/* Service routines */ |
| 301 |
|
| 302 |
|
| 303 |
/* |
| 304 |
It is the only difficult place. flowlabel enforces equal headers |
| 305 |
before and including routing header, however user may supply options |
| 306 |
following rthdr. |
| 307 |
*/ |
| 308 |
|
| 309 |
struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space, |
| 310 |
struct ip6_flowlabel *fl, |
| 311 |
struct ipv6_txoptions *fopt) |
| 312 |
{ |
| 313 |
struct ipv6_txoptions *fl_opt = fl->opt; |
| 314 |
|
| 315 |
if (!fopt || fopt->opt_flen == 0) |
| 316 |
return fl_opt; |
| 317 |
|
| 318 |
if (fl_opt) { |
| 319 |
opt_space->hopopt = fl_opt->hopopt; |
| 320 |
opt_space->dst0opt = fl_opt->dst0opt; |
| 321 |
opt_space->srcrt = fl_opt->srcrt; |
| 322 |
opt_space->opt_nflen = fl_opt->opt_nflen; |
| 323 |
} else { |
| 324 |
if (fopt->opt_nflen == 0) |
| 325 |
return fopt; |
| 326 |
opt_space->hopopt = NULL; |
| 327 |
opt_space->dst0opt = NULL; |
| 328 |
opt_space->srcrt = NULL; |
| 329 |
opt_space->opt_nflen = 0; |
| 330 |
} |
| 331 |
opt_space->dst1opt = fopt->dst1opt; |
| 332 |
opt_space->opt_flen = fopt->opt_flen; |
| 333 |
opt_space->tot_len = fopt->tot_len; |
| 334 |
return opt_space; |
| 335 |
} |
| 336 |
EXPORT_SYMBOL_GPL(fl6_merge_options); |
| 337 |
|
| 338 |
static unsigned long check_linger(unsigned long ttl) |
| 339 |
{ |
| 340 |
if (ttl < FL_MIN_LINGER) |
| 341 |
return FL_MIN_LINGER*HZ; |
| 342 |
if (ttl > FL_MAX_LINGER && !capable(CAP_NET_ADMIN)) |
| 343 |
return 0; |
| 344 |
return ttl*HZ; |
| 345 |
} |
| 346 |
|
| 347 |
static int fl6_renew(struct ip6_flowlabel *fl, unsigned long linger, unsigned long expires) |
| 348 |
{ |
| 349 |
linger = check_linger(linger); |
| 350 |
if (!linger) |
| 351 |
return -EPERM; |
| 352 |
expires = check_linger(expires); |
| 353 |
if (!expires) |
| 354 |
return -EPERM; |
| 355 |
|
| 356 |
spin_lock_bh(&ip6_fl_lock); |
| 357 |
fl->lastuse = jiffies; |
| 358 |
if (time_before(fl->linger, linger)) |
| 359 |
fl->linger = linger; |
| 360 |
if (time_before(expires, fl->linger)) |
| 361 |
expires = fl->linger; |
| 362 |
if (time_before(fl->expires, fl->lastuse + expires)) |
| 363 |
fl->expires = fl->lastuse + expires; |
| 364 |
spin_unlock_bh(&ip6_fl_lock); |
| 365 |
|
| 366 |
return 0; |
| 367 |
} |
| 368 |
|
| 369 |
static struct ip6_flowlabel * |
| 370 |
fl_create(struct net *net, struct sock *sk, struct in6_flowlabel_req *freq, |
| 371 |
sockptr_t optval, int optlen, int *err_p) |
| 372 |
{ |
| 373 |
struct ip6_flowlabel *fl = NULL; |
| 374 |
int olen; |
| 375 |
int addr_type; |
| 376 |
int err; |
| 377 |
|
| 378 |
olen = optlen - CMSG_ALIGN(sizeof(*freq)); |
| 379 |
err = -EINVAL; |
| 380 |
if (olen > 64 * 1024) |
| 381 |
goto done; |
| 382 |
|
| 383 |
err = -ENOMEM; |
| 384 |
fl = kzalloc_obj(*fl); |
| 385 |
if (!fl) |
| 386 |
goto done; |
| 387 |
|
| 388 |
if (olen > 0) { |
| 389 |
struct msghdr msg; |
| 390 |
struct flowi6 flowi6; |
| 391 |
struct ipcm6_cookie ipc6; |
| 392 |
|
| 393 |
err = -ENOMEM; |
| 394 |
fl->opt = kmalloc(sizeof(*fl->opt) + olen, GFP_KERNEL); |
| 395 |
if (!fl->opt) |
| 396 |
goto done; |
| 397 |
|
| 398 |
memset(fl->opt, 0, sizeof(*fl->opt)); |
| 399 |
fl->opt->tot_len = sizeof(*fl->opt) + olen; |
| 400 |
err = -EFAULT; |
| 401 |
if (copy_from_sockptr_offset(fl->opt + 1, optval, |
| 402 |
CMSG_ALIGN(sizeof(*freq)), olen)) |
| 403 |
goto done; |
| 404 |
|
| 405 |
msg.msg_controllen = olen; |
| 406 |
msg.msg_control = (void *)(fl->opt+1); |
| 407 |
memset(&flowi6, 0, sizeof(flowi6)); |
| 408 |
|
| 409 |
ipc6.opt = fl->opt; |
| 410 |
err = ip6_datagram_send_ctl(net, sk, &msg, &flowi6, &ipc6); |
| 411 |
if (err) |
| 412 |
goto done; |
| 413 |
err = -EINVAL; |
| 414 |
if (fl->opt->opt_flen) |
| 415 |
goto done; |
| 416 |
if (fl->opt->opt_nflen == 0) { |
| 417 |
kfree(fl->opt); |
| 418 |
fl->opt = NULL; |
| 419 |
} |
| 420 |
} |
| 421 |
|
| 422 |
fl->fl_net = net; |
| 423 |
fl->expires = jiffies; |
| 424 |
err = fl6_renew(fl, freq->flr_linger, freq->flr_expires); |
| 425 |
if (err) |
| 426 |
goto done; |
| 427 |
fl->share = freq->flr_share; |
| 428 |
addr_type = ipv6_addr_type(&freq->flr_dst); |
| 429 |
if ((addr_type & IPV6_ADDR_MAPPED) || |
| 430 |
addr_type == IPV6_ADDR_ANY) { |
| 431 |
err = -EINVAL; |
| 432 |
goto done; |
| 433 |
} |
| 434 |
fl->dst = freq->flr_dst; |
| 435 |
atomic_set(&fl->users, 1); |
| 436 |
switch (fl->share) { |
| 437 |
case IPV6_FL_S_EXCL: |
| 438 |
case IPV6_FL_S_ANY: |
| 439 |
break; |
| 440 |
case IPV6_FL_S_PROCESS: |
| 441 |
fl->owner.pid = get_task_pid(current, PIDTYPE_PID); |
| 442 |
break; |
| 443 |
case IPV6_FL_S_USER: |
| 444 |
fl->owner.uid = current_euid(); |
| 445 |
break; |
| 446 |
default: |
| 447 |
err = -EINVAL; |
| 448 |
goto done; |
| 449 |
} |
| 450 |
if (fl_shared_exclusive(fl) || fl->opt) { |
| 451 |
WRITE_ONCE(sock_net(sk)->ipv6.flowlabel_has_excl, 1); |
| 452 |
static_branch_deferred_inc(&ipv6_flowlabel_exclusive); |
| 453 |
} |
| 454 |
return fl; |
| 455 |
|
| 456 |
done: |
| 457 |
if (fl) { |
| 458 |
kfree(fl->opt); |
| 459 |
kfree(fl); |
| 460 |
} |
| 461 |
*err_p = err; |
| 462 |
return NULL; |
| 463 |
} |
| 464 |
|
| 465 |
static int mem_check(struct sock *sk) |
| 466 |
{ |
| 467 |
int room = FL_MAX_SIZE - atomic_read(&fl_size); |
| 468 |
struct ipv6_fl_socklist *sfl; |
| 469 |
int count = 0; |
| 470 |
|
| 471 |
if (room > FL_MAX_SIZE - FL_MAX_PER_SOCK) |
| 472 |
return 0; |
| 473 |
|
| 474 |
rcu_read_lock(); |
| 475 |
for_each_sk_fl_rcu(sk, sfl) |
| 476 |
count++; |
| 477 |
rcu_read_unlock(); |
| 478 |
|
| 479 |
if (room <= 0 || |
| 480 |
((count >= FL_MAX_PER_SOCK || |
| 481 |
(count > 0 && room < FL_MAX_SIZE/2) || room < FL_MAX_SIZE/4) && |
| 482 |
!capable(CAP_NET_ADMIN))) |
| 483 |
return -ENOBUFS; |
| 484 |
|
| 485 |
return 0; |
| 486 |
} |
| 487 |
|
| 488 |
static inline void fl_link(struct sock *sk, struct ipv6_fl_socklist *sfl, |
| 489 |
struct ip6_flowlabel *fl) |
| 490 |
{ |
| 491 |
struct inet_sock *inet = inet_sk(sk); |
| 492 |
|
| 493 |
spin_lock_bh(&ip6_sk_fl_lock); |
| 494 |
sfl->fl = fl; |
| 495 |
sfl->next = inet->ipv6_fl_list; |
| 496 |
rcu_assign_pointer(inet->ipv6_fl_list, sfl); |
| 497 |
spin_unlock_bh(&ip6_sk_fl_lock); |
| 498 |
} |
| 499 |
|
| 500 |
int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq, |
| 501 |
int flags) |
| 502 |
{ |
| 503 |
struct ipv6_pinfo *np = inet6_sk(sk); |
| 504 |
struct ipv6_fl_socklist *sfl; |
| 505 |
|
| 506 |
if (flags & IPV6_FL_F_REMOTE) { |
| 507 |
freq->flr_label = np->rcv_flowinfo & IPV6_FLOWLABEL_MASK; |
| 508 |
return 0; |
| 509 |
} |
| 510 |
|
| 511 |
if (inet6_test_bit(REPFLOW, sk)) { |
| 512 |
freq->flr_label = np->flow_label; |
| 513 |
return 0; |
| 514 |
} |
| 515 |
|
| 516 |
rcu_read_lock(); |
| 517 |
|
| 518 |
for_each_sk_fl_rcu(sk, sfl) { |
| 519 |
if (sfl->fl->label == (np->flow_label & IPV6_FLOWLABEL_MASK)) { |
| 520 |
spin_lock_bh(&ip6_fl_lock); |
| 521 |
freq->flr_label = sfl->fl->label; |
| 522 |
freq->flr_dst = sfl->fl->dst; |
| 523 |
freq->flr_share = sfl->fl->share; |
| 524 |
freq->flr_expires = (sfl->fl->expires - jiffies) / HZ; |
| 525 |
freq->flr_linger = sfl->fl->linger / HZ; |
| 526 |
|
| 527 |
spin_unlock_bh(&ip6_fl_lock); |
| 528 |
rcu_read_unlock(); |
| 529 |
return 0; |
| 530 |
} |
| 531 |
} |
| 532 |
rcu_read_unlock(); |
| 533 |
|
| 534 |
return -ENOENT; |
| 535 |
} |
| 536 |
|
| 537 |
#define socklist_dereference(__sflp) \ |
| 538 |
rcu_dereference_protected(__sflp, lockdep_is_held(&ip6_sk_fl_lock)) |
| 539 |
|
| 540 |
static int ipv6_flowlabel_put(struct sock *sk, struct in6_flowlabel_req *freq) |
| 541 |
{ |
| 542 |
struct ipv6_pinfo *np = inet6_sk(sk); |
| 543 |
struct ipv6_fl_socklist __rcu **sflp; |
| 544 |
struct ipv6_fl_socklist *sfl; |
| 545 |
|
| 546 |
if (freq->flr_flags & IPV6_FL_F_REFLECT) { |
| 547 |
if (sk->sk_protocol != IPPROTO_TCP) |
| 548 |
return -ENOPROTOOPT; |
| 549 |
if (!inet6_test_bit(REPFLOW, sk)) |
| 550 |
return -ESRCH; |
| 551 |
np->flow_label = 0; |
| 552 |
inet6_clear_bit(REPFLOW, sk); |
| 553 |
return 0; |
| 554 |
} |
| 555 |
|
| 556 |
spin_lock_bh(&ip6_sk_fl_lock); |
| 557 |
for (sflp = &inet_sk(sk)->ipv6_fl_list; |
| 558 |
(sfl = socklist_dereference(*sflp)) != NULL; |
| 559 |
sflp = &sfl->next) { |
| 560 |
if (sfl->fl->label == freq->flr_label) |
| 561 |
goto found; |
| 562 |
} |
| 563 |
spin_unlock_bh(&ip6_sk_fl_lock); |
| 564 |
return -ESRCH; |
| 565 |
found: |
| 566 |
if (freq->flr_label == (np->flow_label & IPV6_FLOWLABEL_MASK)) |
| 567 |
np->flow_label &= ~IPV6_FLOWLABEL_MASK; |
| 568 |
*sflp = sfl->next; |
| 569 |
spin_unlock_bh(&ip6_sk_fl_lock); |
| 570 |
fl_release(sfl->fl); |
| 571 |
kfree_rcu(sfl, rcu); |
| 572 |
return 0; |
| 573 |
} |
| 574 |
|
| 575 |
static int ipv6_flowlabel_renew(struct sock *sk, struct in6_flowlabel_req *freq) |
| 576 |
{ |
| 577 |
struct net *net = sock_net(sk); |
| 578 |
struct ipv6_fl_socklist *sfl; |
| 579 |
int err; |
| 580 |
|
| 581 |
rcu_read_lock(); |
| 582 |
for_each_sk_fl_rcu(sk, sfl) { |
| 583 |
if (sfl->fl->label == freq->flr_label) { |
| 584 |
err = fl6_renew(sfl->fl, freq->flr_linger, |
| 585 |
freq->flr_expires); |
| 586 |
rcu_read_unlock(); |
| 587 |
return err; |
| 588 |
} |
| 589 |
} |
| 590 |
rcu_read_unlock(); |
| 591 |
|
| 592 |
if (freq->flr_share == IPV6_FL_S_NONE && |
| 593 |
ns_capable(net->user_ns, CAP_NET_ADMIN)) { |
| 594 |
struct ip6_flowlabel *fl = fl_lookup(net, freq->flr_label); |
| 595 |
|
| 596 |
if (fl) { |
| 597 |
err = fl6_renew(fl, freq->flr_linger, |
| 598 |
freq->flr_expires); |
| 599 |
fl_release(fl); |
| 600 |
return err; |
| 601 |
} |
| 602 |
} |
| 603 |
return -ESRCH; |
| 604 |
} |
| 605 |
|
| 606 |
static int ipv6_flowlabel_get(struct sock *sk, struct in6_flowlabel_req *freq, |
| 607 |
sockptr_t optval, int optlen) |
| 608 |
{ |
| 609 |
struct ipv6_fl_socklist *sfl, *sfl1 = NULL; |
| 610 |
struct ip6_flowlabel *fl, *fl1 = NULL; |
| 611 |
struct net *net = sock_net(sk); |
| 612 |
int err; |
| 613 |
|
| 614 |
if (freq->flr_flags & IPV6_FL_F_REFLECT) { |
| 615 |
if (net->ipv6.sysctl.flowlabel_consistency) { |
| 616 |
net_info_ratelimited("Can not set IPV6_FL_F_REFLECT if flowlabel_consistency sysctl is enable\n"); |
| 617 |
return -EPERM; |
| 618 |
} |
| 619 |
|
| 620 |
if (sk->sk_protocol != IPPROTO_TCP) |
| 621 |
return -ENOPROTOOPT; |
| 622 |
inet6_set_bit(REPFLOW, sk); |
| 623 |
return 0; |
| 624 |
} |
| 625 |
|
| 626 |
if (freq->flr_label & ~IPV6_FLOWLABEL_MASK) |
| 627 |
return -EINVAL; |
| 628 |
if (net->ipv6.sysctl.flowlabel_state_ranges && |
| 629 |
(freq->flr_label & IPV6_FLOWLABEL_STATELESS_FLAG)) |
| 630 |
return -ERANGE; |
| 631 |
|
| 632 |
fl = fl_create(net, sk, freq, optval, optlen, &err); |
| 633 |
if (!fl) |
| 634 |
return err; |
| 635 |
|
| 636 |
sfl1 = kmalloc_obj(*sfl1); |
| 637 |
|
| 638 |
if (freq->flr_label) { |
| 639 |
err = -EEXIST; |
| 640 |
rcu_read_lock(); |
| 641 |
for_each_sk_fl_rcu(sk, sfl) { |
| 642 |
if (sfl->fl->label == freq->flr_label) { |
| 643 |
if (freq->flr_flags & IPV6_FL_F_EXCL) { |
| 644 |
rcu_read_unlock(); |
| 645 |
goto done; |
| 646 |
} |
| 647 |
fl1 = sfl->fl; |
| 648 |
if (!atomic_inc_not_zero(&fl1->users)) |
| 649 |
fl1 = NULL; |
| 650 |
break; |
| 651 |
} |
| 652 |
} |
| 653 |
rcu_read_unlock(); |
| 654 |
|
| 655 |
if (!fl1) |
| 656 |
fl1 = fl_lookup(net, freq->flr_label); |
| 657 |
if (fl1) { |
| 658 |
recheck: |
| 659 |
err = -EEXIST; |
| 660 |
if (freq->flr_flags&IPV6_FL_F_EXCL) |
| 661 |
goto release; |
| 662 |
err = -EPERM; |
| 663 |
if (fl1->share == IPV6_FL_S_EXCL || |
| 664 |
fl1->share != fl->share || |
| 665 |
((fl1->share == IPV6_FL_S_PROCESS) && |
| 666 |
(fl1->owner.pid != fl->owner.pid)) || |
| 667 |
((fl1->share == IPV6_FL_S_USER) && |
| 668 |
!uid_eq(fl1->owner.uid, fl->owner.uid))) |
| 669 |
goto release; |
| 670 |
|
| 671 |
err = -ENOMEM; |
| 672 |
if (!sfl1) |
| 673 |
goto release; |
| 674 |
if (fl->linger > fl1->linger) |
| 675 |
fl1->linger = fl->linger; |
| 676 |
if ((long)(fl->expires - fl1->expires) > 0) |
| 677 |
fl1->expires = fl->expires; |
| 678 |
fl_link(sk, sfl1, fl1); |
| 679 |
fl_free(fl); |
| 680 |
return 0; |
| 681 |
|
| 682 |
release: |
| 683 |
fl_release(fl1); |
| 684 |
goto done; |
| 685 |
} |
| 686 |
} |
| 687 |
err = -ENOENT; |
| 688 |
if (!(freq->flr_flags & IPV6_FL_F_CREATE)) |
| 689 |
goto done; |
| 690 |
|
| 691 |
err = -ENOMEM; |
| 692 |
if (!sfl1) |
| 693 |
goto done; |
| 694 |
|
| 695 |
err = mem_check(sk); |
| 696 |
if (err != 0) |
| 697 |
goto done; |
| 698 |
|
| 699 |
fl1 = fl_intern(net, fl, freq->flr_label); |
| 700 |
if (fl1) |
| 701 |
goto recheck; |
| 702 |
|
| 703 |
if (!freq->flr_label) { |
| 704 |
size_t offset = offsetof(struct in6_flowlabel_req, flr_label); |
| 705 |
|
| 706 |
if (copy_to_sockptr_offset(optval, offset, &fl->label, |
| 707 |
sizeof(fl->label))) { |
| 708 |
/* Intentionally ignore fault. */ |
| 709 |
} |
| 710 |
} |
| 711 |
|
| 712 |
fl_link(sk, sfl1, fl); |
| 713 |
return 0; |
| 714 |
done: |
| 715 |
fl_free(fl); |
| 716 |
kfree(sfl1); |
| 717 |
return err; |
| 718 |
} |
| 719 |
|
| 720 |
int ipv6_flowlabel_opt(struct sock *sk, sockptr_t optval, int optlen) |
| 721 |
{ |
| 722 |
struct in6_flowlabel_req freq; |
| 723 |
|
| 724 |
if (optlen < sizeof(freq)) |
| 725 |
return -EINVAL; |
| 726 |
if (copy_from_sockptr(&freq, optval, sizeof(freq))) |
| 727 |
return -EFAULT; |
| 728 |
|
| 729 |
switch (freq.flr_action) { |
| 730 |
case IPV6_FL_A_PUT: |
| 731 |
return ipv6_flowlabel_put(sk, &freq); |
| 732 |
case IPV6_FL_A_RENEW: |
| 733 |
return ipv6_flowlabel_renew(sk, &freq); |
| 734 |
case IPV6_FL_A_GET: |
| 735 |
return ipv6_flowlabel_get(sk, &freq, optval, optlen); |
| 736 |
default: |
| 737 |
return -EINVAL; |
| 738 |
} |
| 739 |
} |
| 740 |
|
| 741 |
#ifdef CONFIG_PROC_FS |
| 742 |
|
| 743 |
struct ip6fl_iter_state { |
| 744 |
struct seq_net_private p; |
| 745 |
struct pid_namespace *pid_ns; |
| 746 |
int bucket; |
| 747 |
}; |
| 748 |
|
| 749 |
#define ip6fl_seq_private(seq) ((struct ip6fl_iter_state *)(seq)->private) |
| 750 |
|
| 751 |
static struct ip6_flowlabel *ip6fl_get_first(struct seq_file *seq) |
| 752 |
{ |
| 753 |
struct ip6_flowlabel *fl = NULL; |
| 754 |
struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 755 |
struct net *net = seq_file_net(seq); |
| 756 |
|
| 757 |
for (state->bucket = 0; state->bucket <= FL_HASH_MASK; ++state->bucket) { |
| 758 |
for_each_fl_rcu(state->bucket, fl) { |
| 759 |
if (net_eq(fl->fl_net, net)) |
| 760 |
goto out; |
| 761 |
} |
| 762 |
} |
| 763 |
fl = NULL; |
| 764 |
out: |
| 765 |
return fl; |
| 766 |
} |
| 767 |
|
| 768 |
static struct ip6_flowlabel *ip6fl_get_next(struct seq_file *seq, struct ip6_flowlabel *fl) |
| 769 |
{ |
| 770 |
struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 771 |
struct net *net = seq_file_net(seq); |
| 772 |
|
| 773 |
for_each_fl_continue_rcu(fl) { |
| 774 |
if (net_eq(fl->fl_net, net)) |
| 775 |
goto out; |
| 776 |
} |
| 777 |
|
| 778 |
try_again: |
| 779 |
if (++state->bucket <= FL_HASH_MASK) { |
| 780 |
for_each_fl_rcu(state->bucket, fl) { |
| 781 |
if (net_eq(fl->fl_net, net)) |
| 782 |
goto out; |
| 783 |
} |
| 784 |
goto try_again; |
| 785 |
} |
| 786 |
fl = NULL; |
| 787 |
|
| 788 |
out: |
| 789 |
return fl; |
| 790 |
} |
| 791 |
|
| 792 |
static struct ip6_flowlabel *ip6fl_get_idx(struct seq_file *seq, loff_t pos) |
| 793 |
{ |
| 794 |
struct ip6_flowlabel *fl = ip6fl_get_first(seq); |
| 795 |
if (fl) |
| 796 |
while (pos && (fl = ip6fl_get_next(seq, fl)) != NULL) |
| 797 |
--pos; |
| 798 |
return pos ? NULL : fl; |
| 799 |
} |
| 800 |
|
| 801 |
static void *ip6fl_seq_start(struct seq_file *seq, loff_t *pos) |
| 802 |
__acquires(RCU) |
| 803 |
{ |
| 804 |
struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 805 |
|
| 806 |
state->pid_ns = proc_pid_ns(file_inode(seq->file)->i_sb); |
| 807 |
|
| 808 |
rcu_read_lock(); |
| 809 |
return *pos ? ip6fl_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; |
| 810 |
} |
| 811 |
|
| 812 |
static void *ip6fl_seq_next(struct seq_file *seq, void *v, loff_t *pos) |
| 813 |
{ |
| 814 |
struct ip6_flowlabel *fl; |
| 815 |
|
| 816 |
if (v == SEQ_START_TOKEN) |
| 817 |
fl = ip6fl_get_first(seq); |
| 818 |
else |
| 819 |
fl = ip6fl_get_next(seq, v); |
| 820 |
++*pos; |
| 821 |
return fl; |
| 822 |
} |
| 823 |
|
| 824 |
static void ip6fl_seq_stop(struct seq_file *seq, void *v) |
| 825 |
__releases(RCU) |
| 826 |
{ |
| 827 |
rcu_read_unlock(); |
| 828 |
} |
| 829 |
|
| 830 |
static int ip6fl_seq_show(struct seq_file *seq, void *v) |
| 831 |
{ |
| 832 |
struct ip6fl_iter_state *state = ip6fl_seq_private(seq); |
| 833 |
if (v == SEQ_START_TOKEN) { |
| 834 |
seq_puts(seq, "Label S Owner Users Linger Expires Dst Opt\n"); |
| 835 |
} else { |
| 836 |
struct ip6_flowlabel *fl = v; |
| 837 |
seq_printf(seq, |
| 838 |
"%05X %-1d %-6d %-6d %-6ld %-8ld %pi6 %-4d\n", |
| 839 |
(unsigned int)ntohl(fl->label), |
| 840 |
fl->share, |
| 841 |
((fl->share == IPV6_FL_S_PROCESS) ? |
| 842 |
pid_nr_ns(fl->owner.pid, state->pid_ns) : |
| 843 |
((fl->share == IPV6_FL_S_USER) ? |
| 844 |
from_kuid_munged(seq_user_ns(seq), fl->owner.uid) : |
| 845 |
0)), |
| 846 |
atomic_read(&fl->users), |
| 847 |
fl->linger/HZ, |
| 848 |
(long)(fl->expires - jiffies)/HZ, |
| 849 |
&fl->dst, |
| 850 |
fl->opt ? fl->opt->opt_nflen : 0); |
| 851 |
} |
| 852 |
return 0; |
| 853 |
} |
| 854 |
|
| 855 |
static const struct seq_operations ip6fl_seq_ops = { |
| 856 |
.start = ip6fl_seq_start, |
| 857 |
.next = ip6fl_seq_next, |
| 858 |
.stop = ip6fl_seq_stop, |
| 859 |
.show = ip6fl_seq_show, |
| 860 |
}; |
| 861 |
|
| 862 |
static int __net_init ip6_flowlabel_proc_init(struct net *net) |
| 863 |
{ |
| 864 |
if (!proc_create_net("ip6_flowlabel", 0444, net->proc_net, |
| 865 |
&ip6fl_seq_ops, sizeof(struct ip6fl_iter_state))) |
| 866 |
return -ENOMEM; |
| 867 |
return 0; |
| 868 |
} |
| 869 |
|
| 870 |
static void __net_exit ip6_flowlabel_proc_fini(struct net *net) |
| 871 |
{ |
| 872 |
remove_proc_entry("ip6_flowlabel", net->proc_net); |
| 873 |
} |
| 874 |
#else |
| 875 |
static inline int ip6_flowlabel_proc_init(struct net *net) |
| 876 |
{ |
| 877 |
return 0; |
| 878 |
} |
| 879 |
static inline void ip6_flowlabel_proc_fini(struct net *net) |
| 880 |
{ |
| 881 |
} |
| 882 |
#endif |
| 883 |
|
| 884 |
static void __net_exit ip6_flowlabel_net_exit(struct net *net) |
| 885 |
{ |
| 886 |
ip6_fl_purge(net); |
| 887 |
ip6_flowlabel_proc_fini(net); |
| 888 |
} |
| 889 |
|
| 890 |
static struct pernet_operations ip6_flowlabel_net_ops = { |
| 891 |
.init = ip6_flowlabel_proc_init, |
| 892 |
.exit = ip6_flowlabel_net_exit, |
| 893 |
}; |
| 894 |
|
| 895 |
int ip6_flowlabel_init(void) |
| 896 |
{ |
| 897 |
return register_pernet_subsys(&ip6_flowlabel_net_ops); |
| 898 |
} |
| 899 |
|
| 900 |
void ip6_flowlabel_cleanup(void) |
| 901 |
{ |
| 902 |
static_key_deferred_flush(&ipv6_flowlabel_exclusive); |
| 903 |
timer_delete(&ip6_fl_gc_timer); |
| 904 |
unregister_pernet_subsys(&ip6_flowlabel_net_ops); |
| 905 |
} |