linux/net/ipv4/ping.c
<<
>>
Prefs
   1/*
   2 * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3 *              operating system.  INET is implemented using the  BSD Socket
   4 *              interface as the means of communication with the user level.
   5 *
   6 *              "Ping" sockets
   7 *
   8 *              This program is free software; you can redistribute it and/or
   9 *              modify it under the terms of the GNU General Public License
  10 *              as published by the Free Software Foundation; either version
  11 *              2 of the License, or (at your option) any later version.
  12 *
  13 * Based on ipv4/udp.c code.
  14 *
  15 * Authors:     Vasiliy Kulikov / Openwall (for Linux 2.6),
  16 *              Pavel Kankovsky (for Linux 2.4.32)
  17 *
  18 * Pavel gave all rights to bugs to Vasiliy,
  19 * none of the bugs are Pavel's now.
  20 *
  21 */
  22
  23#include <linux/uaccess.h>
  24#include <linux/types.h>
  25#include <linux/fcntl.h>
  26#include <linux/socket.h>
  27#include <linux/sockios.h>
  28#include <linux/in.h>
  29#include <linux/errno.h>
  30#include <linux/timer.h>
  31#include <linux/mm.h>
  32#include <linux/inet.h>
  33#include <linux/netdevice.h>
  34#include <net/snmp.h>
  35#include <net/ip.h>
  36#include <net/icmp.h>
  37#include <net/protocol.h>
  38#include <linux/skbuff.h>
  39#include <linux/proc_fs.h>
  40#include <linux/export.h>
  41#include <net/sock.h>
  42#include <net/ping.h>
  43#include <net/udp.h>
  44#include <net/route.h>
  45#include <net/inet_common.h>
  46#include <net/checksum.h>
  47
  48#if IS_ENABLED(CONFIG_IPV6)
  49#include <linux/in6.h>
  50#include <linux/icmpv6.h>
  51#include <net/addrconf.h>
  52#include <net/ipv6.h>
  53#include <net/transp_v6.h>
  54#endif
  55
  56struct ping_table {
  57        struct hlist_nulls_head hash[PING_HTABLE_SIZE];
  58        rwlock_t                lock;
  59};
  60
  61static struct ping_table ping_table;
  62struct pingv6_ops pingv6_ops;
  63EXPORT_SYMBOL_GPL(pingv6_ops);
  64
  65static u16 ping_port_rover;
  66
  67static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
  68{
  69        u32 res = (num + net_hash_mix(net)) & mask;
  70
  71        pr_debug("hash(%u) = %u\n", num, res);
  72        return res;
  73}
  74EXPORT_SYMBOL_GPL(ping_hash);
  75
  76static inline struct hlist_nulls_head *ping_hashslot(struct ping_table *table,
  77                                             struct net *net, unsigned int num)
  78{
  79        return &table->hash[ping_hashfn(net, num, PING_HTABLE_MASK)];
  80}
  81
  82int ping_get_port(struct sock *sk, unsigned short ident)
  83{
  84        struct hlist_nulls_node *node;
  85        struct hlist_nulls_head *hlist;
  86        struct inet_sock *isk, *isk2;
  87        struct sock *sk2 = NULL;
  88
  89        isk = inet_sk(sk);
  90        write_lock_bh(&ping_table.lock);
  91        if (ident == 0) {
  92                u32 i;
  93                u16 result = ping_port_rover + 1;
  94
  95                for (i = 0; i < (1L << 16); i++, result++) {
  96                        if (!result)
  97                                result++; /* avoid zero */
  98                        hlist = ping_hashslot(&ping_table, sock_net(sk),
  99                                            result);
 100                        ping_portaddr_for_each_entry(sk2, node, hlist) {
 101                                isk2 = inet_sk(sk2);
 102
 103                                if (isk2->inet_num == result)
 104                                        goto next_port;
 105                        }
 106
 107                        /* found */
 108                        ping_port_rover = ident = result;
 109                        break;
 110next_port:
 111                        ;
 112                }
 113                if (i >= (1L << 16))
 114                        goto fail;
 115        } else {
 116                hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
 117                ping_portaddr_for_each_entry(sk2, node, hlist) {
 118                        isk2 = inet_sk(sk2);
 119
 120                        /* BUG? Why is this reuse and not reuseaddr? ping.c
 121                         * doesn't turn off SO_REUSEADDR, and it doesn't expect
 122                         * that other ping processes can steal its packets.
 123                         */
 124                        if ((isk2->inet_num == ident) &&
 125                            (sk2 != sk) &&
 126                            (!sk2->sk_reuse || !sk->sk_reuse))
 127                                goto fail;
 128                }
 129        }
 130
 131        pr_debug("found port/ident = %d\n", ident);
 132        isk->inet_num = ident;
 133        if (sk_unhashed(sk)) {
 134                pr_debug("was not hashed\n");
 135                sock_hold(sk);
 136                hlist_nulls_add_head(&sk->sk_nulls_node, hlist);
 137                sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
 138        }
 139        write_unlock_bh(&ping_table.lock);
 140        return 0;
 141
 142fail:
 143        write_unlock_bh(&ping_table.lock);
 144        return 1;
 145}
 146EXPORT_SYMBOL_GPL(ping_get_port);
 147
 148int ping_hash(struct sock *sk)
 149{
 150        pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
 151        BUG(); /* "Please do not press this button again." */
 152
 153        return 0;
 154}
 155
 156void ping_unhash(struct sock *sk)
 157{
 158        struct inet_sock *isk = inet_sk(sk);
 159
 160        pr_debug("ping_unhash(isk=%p,isk->num=%u)\n", isk, isk->inet_num);
 161        write_lock_bh(&ping_table.lock);
 162        if (sk_hashed(sk)) {
 163                hlist_nulls_del(&sk->sk_nulls_node);
 164                sk_nulls_node_init(&sk->sk_nulls_node);
 165                sock_put(sk);
 166                isk->inet_num = 0;
 167                isk->inet_sport = 0;
 168                sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
 169        }
 170        write_unlock_bh(&ping_table.lock);
 171}
 172EXPORT_SYMBOL_GPL(ping_unhash);
 173
 174static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 175{
 176        struct hlist_nulls_head *hslot = ping_hashslot(&ping_table, net, ident);
 177        struct sock *sk = NULL;
 178        struct inet_sock *isk;
 179        struct hlist_nulls_node *hnode;
 180        int dif = skb->dev->ifindex;
 181
 182        if (skb->protocol == htons(ETH_P_IP)) {
 183                pr_debug("try to find: num = %d, daddr = %pI4, dif = %d\n",
 184                         (int)ident, &ip_hdr(skb)->daddr, dif);
 185#if IS_ENABLED(CONFIG_IPV6)
 186        } else if (skb->protocol == htons(ETH_P_IPV6)) {
 187                pr_debug("try to find: num = %d, daddr = %pI6c, dif = %d\n",
 188                         (int)ident, &ipv6_hdr(skb)->daddr, dif);
 189#endif
 190        }
 191
 192        read_lock_bh(&ping_table.lock);
 193
 194        ping_portaddr_for_each_entry(sk, hnode, hslot) {
 195                isk = inet_sk(sk);
 196
 197                pr_debug("iterate\n");
 198                if (isk->inet_num != ident)
 199                        continue;
 200
 201                if (skb->protocol == htons(ETH_P_IP) &&
 202                    sk->sk_family == AF_INET) {
 203                        pr_debug("found: %p: num=%d, daddr=%pI4, dif=%d\n", sk,
 204                                 (int) isk->inet_num, &isk->inet_rcv_saddr,
 205                                 sk->sk_bound_dev_if);
 206
 207                        if (isk->inet_rcv_saddr &&
 208                            isk->inet_rcv_saddr != ip_hdr(skb)->daddr)
 209                                continue;
 210#if IS_ENABLED(CONFIG_IPV6)
 211                } else if (skb->protocol == htons(ETH_P_IPV6) &&
 212                           sk->sk_family == AF_INET6) {
 213
 214                        pr_debug("found: %p: num=%d, daddr=%pI6c, dif=%d\n", sk,
 215                                 (int) isk->inet_num,
 216                                 &sk->sk_v6_rcv_saddr,
 217                                 sk->sk_bound_dev_if);
 218
 219                        if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr) &&
 220                            !ipv6_addr_equal(&sk->sk_v6_rcv_saddr,
 221                                             &ipv6_hdr(skb)->daddr))
 222                                continue;
 223#endif
 224                } else {
 225                        continue;
 226                }
 227
 228                if (sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)
 229                        continue;
 230
 231                sock_hold(sk);
 232                goto exit;
 233        }
 234
 235        sk = NULL;
 236exit:
 237        read_unlock_bh(&ping_table.lock);
 238
 239        return sk;
 240}
 241
 242static void inet_get_ping_group_range_net(struct net *net, kgid_t *low,
 243                                          kgid_t *high)
 244{
 245        kgid_t *data = net->ipv4.ping_group_range.range;
 246        unsigned int seq;
 247
 248        do {
 249                seq = read_seqbegin(&net->ipv4.ping_group_range.lock);
 250
 251                *low = data[0];
 252                *high = data[1];
 253        } while (read_seqretry(&net->ipv4.ping_group_range.lock, seq));
 254}
 255
 256
 257int ping_init_sock(struct sock *sk)
 258{
 259        struct net *net = sock_net(sk);
 260        kgid_t group = current_egid();
 261        struct group_info *group_info;
 262        int i;
 263        kgid_t low, high;
 264        int ret = 0;
 265
 266        if (sk->sk_family == AF_INET6)
 267                sk->sk_ipv6only = 1;
 268
 269        inet_get_ping_group_range_net(net, &low, &high);
 270        if (gid_lte(low, group) && gid_lte(group, high))
 271                return 0;
 272
 273        group_info = get_current_groups();
 274        for (i = 0; i < group_info->ngroups; i++) {
 275                kgid_t gid = group_info->gid[i];
 276
 277                if (gid_lte(low, gid) && gid_lte(gid, high))
 278                        goto out_release_group;
 279        }
 280
 281        ret = -EACCES;
 282
 283out_release_group:
 284        put_group_info(group_info);
 285        return ret;
 286}
 287EXPORT_SYMBOL_GPL(ping_init_sock);
 288
 289void ping_close(struct sock *sk, long timeout)
 290{
 291        pr_debug("ping_close(sk=%p,sk->num=%u)\n",
 292                 inet_sk(sk), inet_sk(sk)->inet_num);
 293        pr_debug("isk->refcnt = %d\n", refcount_read(&sk->sk_refcnt));
 294
 295        sk_common_release(sk);
 296}
 297EXPORT_SYMBOL_GPL(ping_close);
 298
 299/* Checks the bind address and possibly modifies sk->sk_bound_dev_if. */
 300static int ping_check_bind_addr(struct sock *sk, struct inet_sock *isk,
 301                                struct sockaddr *uaddr, int addr_len) {
 302        struct net *net = sock_net(sk);
 303        if (sk->sk_family == AF_INET) {
 304                struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
 305                int chk_addr_ret;
 306
 307                if (addr_len < sizeof(*addr))
 308                        return -EINVAL;
 309
 310                if (addr->sin_family != AF_INET &&
 311                    !(addr->sin_family == AF_UNSPEC &&
 312                      addr->sin_addr.s_addr == htonl(INADDR_ANY)))
 313                        return -EAFNOSUPPORT;
 314
 315                pr_debug("ping_check_bind_addr(sk=%p,addr=%pI4,port=%d)\n",
 316                         sk, &addr->sin_addr.s_addr, ntohs(addr->sin_port));
 317
 318                chk_addr_ret = inet_addr_type(net, addr->sin_addr.s_addr);
 319
 320                if (addr->sin_addr.s_addr == htonl(INADDR_ANY))
 321                        chk_addr_ret = RTN_LOCAL;
 322
 323                if ((net->ipv4.sysctl_ip_nonlocal_bind == 0 &&
 324                    isk->freebind == 0 && isk->transparent == 0 &&
 325                     chk_addr_ret != RTN_LOCAL) ||
 326                    chk_addr_ret == RTN_MULTICAST ||
 327                    chk_addr_ret == RTN_BROADCAST)
 328                        return -EADDRNOTAVAIL;
 329
 330#if IS_ENABLED(CONFIG_IPV6)
 331        } else if (sk->sk_family == AF_INET6) {
 332                struct sockaddr_in6 *addr = (struct sockaddr_in6 *) uaddr;
 333                int addr_type, scoped, has_addr;
 334                struct net_device *dev = NULL;
 335
 336                if (addr_len < sizeof(*addr))
 337                        return -EINVAL;
 338
 339                if (addr->sin6_family != AF_INET6)
 340                        return -EAFNOSUPPORT;
 341
 342                pr_debug("ping_check_bind_addr(sk=%p,addr=%pI6c,port=%d)\n",
 343                         sk, addr->sin6_addr.s6_addr, ntohs(addr->sin6_port));
 344
 345                addr_type = ipv6_addr_type(&addr->sin6_addr);
 346                scoped = __ipv6_addr_needs_scope_id(addr_type);
 347                if ((addr_type != IPV6_ADDR_ANY &&
 348                     !(addr_type & IPV6_ADDR_UNICAST)) ||
 349                    (scoped && !addr->sin6_scope_id))
 350                        return -EINVAL;
 351
 352                rcu_read_lock();
 353                if (addr->sin6_scope_id) {
 354                        dev = dev_get_by_index_rcu(net, addr->sin6_scope_id);
 355                        if (!dev) {
 356                                rcu_read_unlock();
 357                                return -ENODEV;
 358                        }
 359                }
 360                has_addr = pingv6_ops.ipv6_chk_addr(net, &addr->sin6_addr, dev,
 361                                                    scoped);
 362                rcu_read_unlock();
 363
 364                if (!(net->ipv6.sysctl.ip_nonlocal_bind ||
 365                      isk->freebind || isk->transparent || has_addr ||
 366                      addr_type == IPV6_ADDR_ANY))
 367                        return -EADDRNOTAVAIL;
 368
 369                if (scoped)
 370                        sk->sk_bound_dev_if = addr->sin6_scope_id;
 371#endif
 372        } else {
 373                return -EAFNOSUPPORT;
 374        }
 375        return 0;
 376}
 377
 378static void ping_set_saddr(struct sock *sk, struct sockaddr *saddr)
 379{
 380        if (saddr->sa_family == AF_INET) {
 381                struct inet_sock *isk = inet_sk(sk);
 382                struct sockaddr_in *addr = (struct sockaddr_in *) saddr;
 383                isk->inet_rcv_saddr = isk->inet_saddr = addr->sin_addr.s_addr;
 384#if IS_ENABLED(CONFIG_IPV6)
 385        } else if (saddr->sa_family == AF_INET6) {
 386                struct sockaddr_in6 *addr = (struct sockaddr_in6 *) saddr;
 387                struct ipv6_pinfo *np = inet6_sk(sk);
 388                sk->sk_v6_rcv_saddr = np->saddr = addr->sin6_addr;
 389#endif
 390        }
 391}
 392
 393static void ping_clear_saddr(struct sock *sk, int dif)
 394{
 395        sk->sk_bound_dev_if = dif;
 396        if (sk->sk_family == AF_INET) {
 397                struct inet_sock *isk = inet_sk(sk);
 398                isk->inet_rcv_saddr = isk->inet_saddr = 0;
 399#if IS_ENABLED(CONFIG_IPV6)
 400        } else if (sk->sk_family == AF_INET6) {
 401                struct ipv6_pinfo *np = inet6_sk(sk);
 402                memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
 403                memset(&np->saddr, 0, sizeof(np->saddr));
 404#endif
 405        }
 406}
 407/*
 408 * We need our own bind because there are no privileged id's == local ports.
 409 * Moreover, we don't allow binding to multi- and broadcast addresses.
 410 */
 411
 412int ping_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
 413{
 414        struct inet_sock *isk = inet_sk(sk);
 415        unsigned short snum;
 416        int err;
 417        int dif = sk->sk_bound_dev_if;
 418
 419        err = ping_check_bind_addr(sk, isk, uaddr, addr_len);
 420        if (err)
 421                return err;
 422
 423        lock_sock(sk);
 424
 425        err = -EINVAL;
 426        if (isk->inet_num != 0)
 427                goto out;
 428
 429        err = -EADDRINUSE;
 430        ping_set_saddr(sk, uaddr);
 431        snum = ntohs(((struct sockaddr_in *)uaddr)->sin_port);
 432        if (ping_get_port(sk, snum) != 0) {
 433                ping_clear_saddr(sk, dif);
 434                goto out;
 435        }
 436
 437        pr_debug("after bind(): num = %hu, dif = %d\n",
 438                 isk->inet_num,
 439                 sk->sk_bound_dev_if);
 440
 441        err = 0;
 442        if (sk->sk_family == AF_INET && isk->inet_rcv_saddr)
 443                sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 444#if IS_ENABLED(CONFIG_IPV6)
 445        if (sk->sk_family == AF_INET6 && !ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 446                sk->sk_userlocks |= SOCK_BINDADDR_LOCK;
 447#endif
 448
 449        if (snum)
 450                sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
 451        isk->inet_sport = htons(isk->inet_num);
 452        isk->inet_daddr = 0;
 453        isk->inet_dport = 0;
 454
 455#if IS_ENABLED(CONFIG_IPV6)
 456        if (sk->sk_family == AF_INET6)
 457                memset(&sk->sk_v6_daddr, 0, sizeof(sk->sk_v6_daddr));
 458#endif
 459
 460        sk_dst_reset(sk);
 461out:
 462        release_sock(sk);
 463        pr_debug("ping_v4_bind -> %d\n", err);
 464        return err;
 465}
 466EXPORT_SYMBOL_GPL(ping_bind);
 467
 468/*
 469 * Is this a supported type of ICMP message?
 470 */
 471
 472static inline int ping_supported(int family, int type, int code)
 473{
 474        return (family == AF_INET && type == ICMP_ECHO && code == 0) ||
 475               (family == AF_INET6 && type == ICMPV6_ECHO_REQUEST && code == 0);
 476}
 477
 478/*
 479 * This routine is called by the ICMP module when it gets some
 480 * sort of error condition.
 481 */
 482
 483void ping_err(struct sk_buff *skb, int offset, u32 info)
 484{
 485        int family;
 486        struct icmphdr *icmph;
 487        struct inet_sock *inet_sock;
 488        int type;
 489        int code;
 490        struct net *net = dev_net(skb->dev);
 491        struct sock *sk;
 492        int harderr;
 493        int err;
 494
 495        if (skb->protocol == htons(ETH_P_IP)) {
 496                family = AF_INET;
 497                type = icmp_hdr(skb)->type;
 498                code = icmp_hdr(skb)->code;
 499                icmph = (struct icmphdr *)(skb->data + offset);
 500        } else if (skb->protocol == htons(ETH_P_IPV6)) {
 501                family = AF_INET6;
 502                type = icmp6_hdr(skb)->icmp6_type;
 503                code = icmp6_hdr(skb)->icmp6_code;
 504                icmph = (struct icmphdr *) (skb->data + offset);
 505        } else {
 506                BUG();
 507        }
 508
 509        /* We assume the packet has already been checked by icmp_unreach */
 510
 511        if (!ping_supported(family, icmph->type, icmph->code))
 512                return;
 513
 514        pr_debug("ping_err(proto=0x%x,type=%d,code=%d,id=%04x,seq=%04x)\n",
 515                 skb->protocol, type, code, ntohs(icmph->un.echo.id),
 516                 ntohs(icmph->un.echo.sequence));
 517
 518        sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 519        if (!sk) {
 520                pr_debug("no socket, dropping\n");
 521                return; /* No socket for error */
 522        }
 523        pr_debug("err on socket %p\n", sk);
 524
 525        err = 0;
 526        harderr = 0;
 527        inet_sock = inet_sk(sk);
 528
 529        if (skb->protocol == htons(ETH_P_IP)) {
 530                switch (type) {
 531                default:
 532                case ICMP_TIME_EXCEEDED:
 533                        err = EHOSTUNREACH;
 534                        break;
 535                case ICMP_SOURCE_QUENCH:
 536                        /* This is not a real error but ping wants to see it.
 537                         * Report it with some fake errno.
 538                         */
 539                        err = EREMOTEIO;
 540                        break;
 541                case ICMP_PARAMETERPROB:
 542                        err = EPROTO;
 543                        harderr = 1;
 544                        break;
 545                case ICMP_DEST_UNREACH:
 546                        if (code == ICMP_FRAG_NEEDED) { /* Path MTU discovery */
 547                                ipv4_sk_update_pmtu(skb, sk, info);
 548                                if (inet_sock->pmtudisc != IP_PMTUDISC_DONT) {
 549                                        err = EMSGSIZE;
 550                                        harderr = 1;
 551                                        break;
 552                                }
 553                                goto out;
 554                        }
 555                        err = EHOSTUNREACH;
 556                        if (code <= NR_ICMP_UNREACH) {
 557                                harderr = icmp_err_convert[code].fatal;
 558                                err = icmp_err_convert[code].errno;
 559                        }
 560                        break;
 561                case ICMP_REDIRECT:
 562                        /* See ICMP_SOURCE_QUENCH */
 563                        ipv4_sk_redirect(skb, sk);
 564                        err = EREMOTEIO;
 565                        break;
 566                }
 567#if IS_ENABLED(CONFIG_IPV6)
 568        } else if (skb->protocol == htons(ETH_P_IPV6)) {
 569                harderr = pingv6_ops.icmpv6_err_convert(type, code, &err);
 570#endif
 571        }
 572
 573        /*
 574         *      RFC1122: OK.  Passes ICMP errors back to application, as per
 575         *      4.1.3.3.
 576         */
 577        if ((family == AF_INET && !inet_sock->recverr) ||
 578            (family == AF_INET6 && !inet6_sk(sk)->recverr)) {
 579                if (!harderr || sk->sk_state != TCP_ESTABLISHED)
 580                        goto out;
 581        } else {
 582                if (family == AF_INET) {
 583                        ip_icmp_error(sk, skb, err, 0 /* no remote port */,
 584                                      info, (u8 *)icmph);
 585#if IS_ENABLED(CONFIG_IPV6)
 586                } else if (family == AF_INET6) {
 587                        pingv6_ops.ipv6_icmp_error(sk, skb, err, 0,
 588                                                   info, (u8 *)icmph);
 589#endif
 590                }
 591        }
 592        sk->sk_err = err;
 593        sk->sk_error_report(sk);
 594out:
 595        sock_put(sk);
 596}
 597EXPORT_SYMBOL_GPL(ping_err);
 598
 599/*
 600 *      Copy and checksum an ICMP Echo packet from user space into a buffer
 601 *      starting from the payload.
 602 */
 603
 604int ping_getfrag(void *from, char *to,
 605                 int offset, int fraglen, int odd, struct sk_buff *skb)
 606{
 607        struct pingfakehdr *pfh = (struct pingfakehdr *)from;
 608
 609        if (offset == 0) {
 610                fraglen -= sizeof(struct icmphdr);
 611                if (fraglen < 0)
 612                        BUG();
 613                if (!csum_and_copy_from_iter_full(to + sizeof(struct icmphdr),
 614                            fraglen, &pfh->wcheck,
 615                            &pfh->msg->msg_iter))
 616                        return -EFAULT;
 617        } else if (offset < sizeof(struct icmphdr)) {
 618                        BUG();
 619        } else {
 620                if (!csum_and_copy_from_iter_full(to, fraglen, &pfh->wcheck,
 621                                            &pfh->msg->msg_iter))
 622                        return -EFAULT;
 623        }
 624
 625#if IS_ENABLED(CONFIG_IPV6)
 626        /* For IPv6, checksum each skb as we go along, as expected by
 627         * icmpv6_push_pending_frames. For IPv4, accumulate the checksum in
 628         * wcheck, it will be finalized in ping_v4_push_pending_frames.
 629         */
 630        if (pfh->family == AF_INET6) {
 631                skb->csum = pfh->wcheck;
 632                skb->ip_summed = CHECKSUM_NONE;
 633                pfh->wcheck = 0;
 634        }
 635#endif
 636
 637        return 0;
 638}
 639EXPORT_SYMBOL_GPL(ping_getfrag);
 640
 641static int ping_v4_push_pending_frames(struct sock *sk, struct pingfakehdr *pfh,
 642                                       struct flowi4 *fl4)
 643{
 644        struct sk_buff *skb = skb_peek(&sk->sk_write_queue);
 645
 646        if (!skb)
 647                return 0;
 648        pfh->wcheck = csum_partial((char *)&pfh->icmph,
 649                sizeof(struct icmphdr), pfh->wcheck);
 650        pfh->icmph.checksum = csum_fold(pfh->wcheck);
 651        memcpy(icmp_hdr(skb), &pfh->icmph, sizeof(struct icmphdr));
 652        skb->ip_summed = CHECKSUM_NONE;
 653        return ip_push_pending_frames(sk, fl4);
 654}
 655
 656int ping_common_sendmsg(int family, struct msghdr *msg, size_t len,
 657                        void *user_icmph, size_t icmph_len) {
 658        u8 type, code;
 659
 660        if (len > 0xFFFF)
 661                return -EMSGSIZE;
 662
 663        /* Must have at least a full ICMP header. */
 664        if (len < icmph_len)
 665                return -EINVAL;
 666
 667        /*
 668         *      Check the flags.
 669         */
 670
 671        /* Mirror BSD error message compatibility */
 672        if (msg->msg_flags & MSG_OOB)
 673                return -EOPNOTSUPP;
 674
 675        /*
 676         *      Fetch the ICMP header provided by the userland.
 677         *      iovec is modified! The ICMP header is consumed.
 678         */
 679        if (memcpy_from_msg(user_icmph, msg, icmph_len))
 680                return -EFAULT;
 681
 682        if (family == AF_INET) {
 683                type = ((struct icmphdr *) user_icmph)->type;
 684                code = ((struct icmphdr *) user_icmph)->code;
 685#if IS_ENABLED(CONFIG_IPV6)
 686        } else if (family == AF_INET6) {
 687                type = ((struct icmp6hdr *) user_icmph)->icmp6_type;
 688                code = ((struct icmp6hdr *) user_icmph)->icmp6_code;
 689#endif
 690        } else {
 691                BUG();
 692        }
 693
 694        if (!ping_supported(family, type, code))
 695                return -EINVAL;
 696
 697        return 0;
 698}
 699EXPORT_SYMBOL_GPL(ping_common_sendmsg);
 700
 701static int ping_v4_sendmsg(struct sock *sk, struct msghdr *msg, size_t len)
 702{
 703        struct net *net = sock_net(sk);
 704        struct flowi4 fl4;
 705        struct inet_sock *inet = inet_sk(sk);
 706        struct ipcm_cookie ipc;
 707        struct icmphdr user_icmph;
 708        struct pingfakehdr pfh;
 709        struct rtable *rt = NULL;
 710        struct ip_options_data opt_copy;
 711        int free = 0;
 712        __be32 saddr, daddr, faddr;
 713        u8  tos;
 714        int err;
 715
 716        pr_debug("ping_v4_sendmsg(sk=%p,sk->num=%u)\n", inet, inet->inet_num);
 717
 718        err = ping_common_sendmsg(AF_INET, msg, len, &user_icmph,
 719                                  sizeof(user_icmph));
 720        if (err)
 721                return err;
 722
 723        /*
 724         *      Get and verify the address.
 725         */
 726
 727        if (msg->msg_name) {
 728                DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
 729                if (msg->msg_namelen < sizeof(*usin))
 730                        return -EINVAL;
 731                if (usin->sin_family != AF_INET)
 732                        return -EAFNOSUPPORT;
 733                daddr = usin->sin_addr.s_addr;
 734                /* no remote port */
 735        } else {
 736                if (sk->sk_state != TCP_ESTABLISHED)
 737                        return -EDESTADDRREQ;
 738                daddr = inet->inet_daddr;
 739                /* no remote port */
 740        }
 741
 742        ipc.sockc.tsflags = sk->sk_tsflags;
 743        ipc.addr = inet->inet_saddr;
 744        ipc.opt = NULL;
 745        ipc.oif = sk->sk_bound_dev_if;
 746        ipc.tx_flags = 0;
 747        ipc.ttl = 0;
 748        ipc.tos = -1;
 749
 750        if (msg->msg_controllen) {
 751                err = ip_cmsg_send(sk, msg, &ipc, false);
 752                if (unlikely(err)) {
 753                        kfree(ipc.opt);
 754                        return err;
 755                }
 756                if (ipc.opt)
 757                        free = 1;
 758        }
 759        if (!ipc.opt) {
 760                struct ip_options_rcu *inet_opt;
 761
 762                rcu_read_lock();
 763                inet_opt = rcu_dereference(inet->inet_opt);
 764                if (inet_opt) {
 765                        memcpy(&opt_copy, inet_opt,
 766                               sizeof(*inet_opt) + inet_opt->opt.optlen);
 767                        ipc.opt = &opt_copy.opt;
 768                }
 769                rcu_read_unlock();
 770        }
 771
 772        sock_tx_timestamp(sk, ipc.sockc.tsflags, &ipc.tx_flags);
 773
 774        saddr = ipc.addr;
 775        ipc.addr = faddr = daddr;
 776
 777        if (ipc.opt && ipc.opt->opt.srr) {
 778                if (!daddr)
 779                        return -EINVAL;
 780                faddr = ipc.opt->opt.faddr;
 781        }
 782        tos = get_rttos(&ipc, inet);
 783        if (sock_flag(sk, SOCK_LOCALROUTE) ||
 784            (msg->msg_flags & MSG_DONTROUTE) ||
 785            (ipc.opt && ipc.opt->opt.is_strictroute)) {
 786                tos |= RTO_ONLINK;
 787        }
 788
 789        if (ipv4_is_multicast(daddr)) {
 790                if (!ipc.oif)
 791                        ipc.oif = inet->mc_index;
 792                if (!saddr)
 793                        saddr = inet->mc_addr;
 794        } else if (!ipc.oif)
 795                ipc.oif = inet->uc_index;
 796
 797        flowi4_init_output(&fl4, ipc.oif, sk->sk_mark, tos,
 798                           RT_SCOPE_UNIVERSE, sk->sk_protocol,
 799                           inet_sk_flowi_flags(sk), faddr, saddr, 0, 0,
 800                           sk->sk_uid);
 801
 802        security_sk_classify_flow(sk, flowi4_to_flowi(&fl4));
 803        rt = ip_route_output_flow(net, &fl4, sk);
 804        if (IS_ERR(rt)) {
 805                err = PTR_ERR(rt);
 806                rt = NULL;
 807                if (err == -ENETUNREACH)
 808                        IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
 809                goto out;
 810        }
 811
 812        err = -EACCES;
 813        if ((rt->rt_flags & RTCF_BROADCAST) &&
 814            !sock_flag(sk, SOCK_BROADCAST))
 815                goto out;
 816
 817        if (msg->msg_flags & MSG_CONFIRM)
 818                goto do_confirm;
 819back_from_confirm:
 820
 821        if (!ipc.addr)
 822                ipc.addr = fl4.daddr;
 823
 824        lock_sock(sk);
 825
 826        pfh.icmph.type = user_icmph.type; /* already checked */
 827        pfh.icmph.code = user_icmph.code; /* ditto */
 828        pfh.icmph.checksum = 0;
 829        pfh.icmph.un.echo.id = inet->inet_sport;
 830        pfh.icmph.un.echo.sequence = user_icmph.un.echo.sequence;
 831        pfh.msg = msg;
 832        pfh.wcheck = 0;
 833        pfh.family = AF_INET;
 834
 835        err = ip_append_data(sk, &fl4, ping_getfrag, &pfh, len,
 836                        0, &ipc, &rt, msg->msg_flags);
 837        if (err)
 838                ip_flush_pending_frames(sk);
 839        else
 840                err = ping_v4_push_pending_frames(sk, &pfh, &fl4);
 841        release_sock(sk);
 842
 843out:
 844        ip_rt_put(rt);
 845        if (free)
 846                kfree(ipc.opt);
 847        if (!err) {
 848                icmp_out_count(sock_net(sk), user_icmph.type);
 849                return len;
 850        }
 851        return err;
 852
 853do_confirm:
 854        if (msg->msg_flags & MSG_PROBE)
 855                dst_confirm_neigh(&rt->dst, &fl4.daddr);
 856        if (!(msg->msg_flags & MSG_PROBE) || len)
 857                goto back_from_confirm;
 858        err = 0;
 859        goto out;
 860}
 861
 862int ping_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int noblock,
 863                 int flags, int *addr_len)
 864{
 865        struct inet_sock *isk = inet_sk(sk);
 866        int family = sk->sk_family;
 867        struct sk_buff *skb;
 868        int copied, err;
 869
 870        pr_debug("ping_recvmsg(sk=%p,sk->num=%u)\n", isk, isk->inet_num);
 871
 872        err = -EOPNOTSUPP;
 873        if (flags & MSG_OOB)
 874                goto out;
 875
 876        if (flags & MSG_ERRQUEUE)
 877                return inet_recv_error(sk, msg, len, addr_len);
 878
 879        skb = skb_recv_datagram(sk, flags, noblock, &err);
 880        if (!skb)
 881                goto out;
 882
 883        copied = skb->len;
 884        if (copied > len) {
 885                msg->msg_flags |= MSG_TRUNC;
 886                copied = len;
 887        }
 888
 889        /* Don't bother checking the checksum */
 890        err = skb_copy_datagram_msg(skb, 0, msg, copied);
 891        if (err)
 892                goto done;
 893
 894        sock_recv_timestamp(msg, sk, skb);
 895
 896        /* Copy the address and add cmsg data. */
 897        if (family == AF_INET) {
 898                DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name);
 899
 900                if (sin) {
 901                        sin->sin_family = AF_INET;
 902                        sin->sin_port = 0 /* skb->h.uh->source */;
 903                        sin->sin_addr.s_addr = ip_hdr(skb)->saddr;
 904                        memset(sin->sin_zero, 0, sizeof(sin->sin_zero));
 905                        *addr_len = sizeof(*sin);
 906                }
 907
 908                if (isk->cmsg_flags)
 909                        ip_cmsg_recv(msg, skb);
 910
 911#if IS_ENABLED(CONFIG_IPV6)
 912        } else if (family == AF_INET6) {
 913                struct ipv6_pinfo *np = inet6_sk(sk);
 914                struct ipv6hdr *ip6 = ipv6_hdr(skb);
 915                DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
 916
 917                if (sin6) {
 918                        sin6->sin6_family = AF_INET6;
 919                        sin6->sin6_port = 0;
 920                        sin6->sin6_addr = ip6->saddr;
 921                        sin6->sin6_flowinfo = 0;
 922                        if (np->sndflow)
 923                                sin6->sin6_flowinfo = ip6_flowinfo(ip6);
 924                        sin6->sin6_scope_id =
 925                                ipv6_iface_scope_id(&sin6->sin6_addr,
 926                                                    inet6_iif(skb));
 927                        *addr_len = sizeof(*sin6);
 928                }
 929
 930                if (inet6_sk(sk)->rxopt.all)
 931                        pingv6_ops.ip6_datagram_recv_common_ctl(sk, msg, skb);
 932                if (skb->protocol == htons(ETH_P_IPV6) &&
 933                    inet6_sk(sk)->rxopt.all)
 934                        pingv6_ops.ip6_datagram_recv_specific_ctl(sk, msg, skb);
 935                else if (skb->protocol == htons(ETH_P_IP) && isk->cmsg_flags)
 936                        ip_cmsg_recv(msg, skb);
 937#endif
 938        } else {
 939                BUG();
 940        }
 941
 942        err = copied;
 943
 944done:
 945        skb_free_datagram(sk, skb);
 946out:
 947        pr_debug("ping_recvmsg -> %d\n", err);
 948        return err;
 949}
 950EXPORT_SYMBOL_GPL(ping_recvmsg);
 951
 952int ping_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 953{
 954        pr_debug("ping_queue_rcv_skb(sk=%p,sk->num=%d,skb=%p)\n",
 955                 inet_sk(sk), inet_sk(sk)->inet_num, skb);
 956        if (sock_queue_rcv_skb(sk, skb) < 0) {
 957                kfree_skb(skb);
 958                pr_debug("ping_queue_rcv_skb -> failed\n");
 959                return -1;
 960        }
 961        return 0;
 962}
 963EXPORT_SYMBOL_GPL(ping_queue_rcv_skb);
 964
 965
 966/*
 967 *      All we need to do is get the socket.
 968 */
 969
 970bool ping_rcv(struct sk_buff *skb)
 971{
 972        struct sock *sk;
 973        struct net *net = dev_net(skb->dev);
 974        struct icmphdr *icmph = icmp_hdr(skb);
 975
 976        /* We assume the packet has already been checked by icmp_rcv */
 977
 978        pr_debug("ping_rcv(skb=%p,id=%04x,seq=%04x)\n",
 979                 skb, ntohs(icmph->un.echo.id), ntohs(icmph->un.echo.sequence));
 980
 981        /* Push ICMP header back */
 982        skb_push(skb, skb->data - (u8 *)icmph);
 983
 984        sk = ping_lookup(net, skb, ntohs(icmph->un.echo.id));
 985        if (sk) {
 986                struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 987
 988                pr_debug("rcv on socket %p\n", sk);
 989                if (skb2)
 990                        ping_queue_rcv_skb(sk, skb2);
 991                sock_put(sk);
 992                return true;
 993        }
 994        pr_debug("no socket, dropping\n");
 995
 996        return false;
 997}
 998EXPORT_SYMBOL_GPL(ping_rcv);
 999
1000struct proto ping_prot = {
1001        .name =         "PING",
1002        .owner =        THIS_MODULE,
1003        .init =         ping_init_sock,
1004        .close =        ping_close,
1005        .connect =      ip4_datagram_connect,
1006        .disconnect =   __udp_disconnect,
1007        .setsockopt =   ip_setsockopt,
1008        .getsockopt =   ip_getsockopt,
1009        .sendmsg =      ping_v4_sendmsg,
1010        .recvmsg =      ping_recvmsg,
1011        .bind =         ping_bind,
1012        .backlog_rcv =  ping_queue_rcv_skb,
1013        .release_cb =   ip4_datagram_release_cb,
1014        .hash =         ping_hash,
1015        .unhash =       ping_unhash,
1016        .get_port =     ping_get_port,
1017        .obj_size =     sizeof(struct inet_sock),
1018};
1019EXPORT_SYMBOL(ping_prot);
1020
1021#ifdef CONFIG_PROC_FS
1022
1023static struct sock *ping_get_first(struct seq_file *seq, int start)
1024{
1025        struct sock *sk;
1026        struct ping_iter_state *state = seq->private;
1027        struct net *net = seq_file_net(seq);
1028
1029        for (state->bucket = start; state->bucket < PING_HTABLE_SIZE;
1030             ++state->bucket) {
1031                struct hlist_nulls_node *node;
1032                struct hlist_nulls_head *hslot;
1033
1034                hslot = &ping_table.hash[state->bucket];
1035
1036                if (hlist_nulls_empty(hslot))
1037                        continue;
1038
1039                sk_nulls_for_each(sk, node, hslot) {
1040                        if (net_eq(sock_net(sk), net) &&
1041                            sk->sk_family == state->family)
1042                                goto found;
1043                }
1044        }
1045        sk = NULL;
1046found:
1047        return sk;
1048}
1049
1050static struct sock *ping_get_next(struct seq_file *seq, struct sock *sk)
1051{
1052        struct ping_iter_state *state = seq->private;
1053        struct net *net = seq_file_net(seq);
1054
1055        do {
1056                sk = sk_nulls_next(sk);
1057        } while (sk && (!net_eq(sock_net(sk), net)));
1058
1059        if (!sk)
1060                return ping_get_first(seq, state->bucket + 1);
1061        return sk;
1062}
1063
1064static struct sock *ping_get_idx(struct seq_file *seq, loff_t pos)
1065{
1066        struct sock *sk = ping_get_first(seq, 0);
1067
1068        if (sk)
1069                while (pos && (sk = ping_get_next(seq, sk)) != NULL)
1070                        --pos;
1071        return pos ? NULL : sk;
1072}
1073
1074void *ping_seq_start(struct seq_file *seq, loff_t *pos, sa_family_t family)
1075        __acquires(ping_table.lock)
1076{
1077        struct ping_iter_state *state = seq->private;
1078        state->bucket = 0;
1079        state->family = family;
1080
1081        read_lock_bh(&ping_table.lock);
1082
1083        return *pos ? ping_get_idx(seq, *pos-1) : SEQ_START_TOKEN;
1084}
1085EXPORT_SYMBOL_GPL(ping_seq_start);
1086
1087static void *ping_v4_seq_start(struct seq_file *seq, loff_t *pos)
1088{
1089        return ping_seq_start(seq, pos, AF_INET);
1090}
1091
1092void *ping_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1093{
1094        struct sock *sk;
1095
1096        if (v == SEQ_START_TOKEN)
1097                sk = ping_get_idx(seq, 0);
1098        else
1099                sk = ping_get_next(seq, v);
1100
1101        ++*pos;
1102        return sk;
1103}
1104EXPORT_SYMBOL_GPL(ping_seq_next);
1105
1106void ping_seq_stop(struct seq_file *seq, void *v)
1107        __releases(ping_table.lock)
1108{
1109        read_unlock_bh(&ping_table.lock);
1110}
1111EXPORT_SYMBOL_GPL(ping_seq_stop);
1112
1113static void ping_v4_format_sock(struct sock *sp, struct seq_file *f,
1114                int bucket)
1115{
1116        struct inet_sock *inet = inet_sk(sp);
1117        __be32 dest = inet->inet_daddr;
1118        __be32 src = inet->inet_rcv_saddr;
1119        __u16 destp = ntohs(inet->inet_dport);
1120        __u16 srcp = ntohs(inet->inet_sport);
1121
1122        seq_printf(f, "%5d: %08X:%04X %08X:%04X"
1123                " %02X %08X:%08X %02X:%08lX %08X %5u %8d %lu %d %pK %d",
1124                bucket, src, srcp, dest, destp, sp->sk_state,
1125                sk_wmem_alloc_get(sp),
1126                sk_rmem_alloc_get(sp),
1127                0, 0L, 0,
1128                from_kuid_munged(seq_user_ns(f), sock_i_uid(sp)),
1129                0, sock_i_ino(sp),
1130                refcount_read(&sp->sk_refcnt), sp,
1131                atomic_read(&sp->sk_drops));
1132}
1133
1134static int ping_v4_seq_show(struct seq_file *seq, void *v)
1135{
1136        seq_setwidth(seq, 127);
1137        if (v == SEQ_START_TOKEN)
1138                seq_puts(seq, "  sl  local_address rem_address   st tx_queue "
1139                           "rx_queue tr tm->when retrnsmt   uid  timeout "
1140                           "inode ref pointer drops");
1141        else {
1142                struct ping_iter_state *state = seq->private;
1143
1144                ping_v4_format_sock(v, seq, state->bucket);
1145        }
1146        seq_pad(seq, '\n');
1147        return 0;
1148}
1149
1150static int ping_seq_open(struct inode *inode, struct file *file)
1151{
1152        struct ping_seq_afinfo *afinfo = PDE_DATA(inode);
1153        return seq_open_net(inode, file, &afinfo->seq_ops,
1154                           sizeof(struct ping_iter_state));
1155}
1156
1157const struct file_operations ping_seq_fops = {
1158        .open           = ping_seq_open,
1159        .read           = seq_read,
1160        .llseek         = seq_lseek,
1161        .release        = seq_release_net,
1162};
1163EXPORT_SYMBOL_GPL(ping_seq_fops);
1164
1165static struct ping_seq_afinfo ping_v4_seq_afinfo = {
1166        .name           = "icmp",
1167        .family         = AF_INET,
1168        .seq_fops       = &ping_seq_fops,
1169        .seq_ops        = {
1170                .start          = ping_v4_seq_start,
1171                .show           = ping_v4_seq_show,
1172                .next           = ping_seq_next,
1173                .stop           = ping_seq_stop,
1174        },
1175};
1176
1177int ping_proc_register(struct net *net, struct ping_seq_afinfo *afinfo)
1178{
1179        struct proc_dir_entry *p;
1180        p = proc_create_data(afinfo->name, S_IRUGO, net->proc_net,
1181                             afinfo->seq_fops, afinfo);
1182        if (!p)
1183                return -ENOMEM;
1184        return 0;
1185}
1186EXPORT_SYMBOL_GPL(ping_proc_register);
1187
1188void ping_proc_unregister(struct net *net, struct ping_seq_afinfo *afinfo)
1189{
1190        remove_proc_entry(afinfo->name, net->proc_net);
1191}
1192EXPORT_SYMBOL_GPL(ping_proc_unregister);
1193
1194static int __net_init ping_v4_proc_init_net(struct net *net)
1195{
1196        return ping_proc_register(net, &ping_v4_seq_afinfo);
1197}
1198
1199static void __net_exit ping_v4_proc_exit_net(struct net *net)
1200{
1201        ping_proc_unregister(net, &ping_v4_seq_afinfo);
1202}
1203
1204static struct pernet_operations ping_v4_net_ops = {
1205        .init = ping_v4_proc_init_net,
1206        .exit = ping_v4_proc_exit_net,
1207};
1208
1209int __init ping_proc_init(void)
1210{
1211        return register_pernet_subsys(&ping_v4_net_ops);
1212}
1213
1214void ping_proc_exit(void)
1215{
1216        unregister_pernet_subsys(&ping_v4_net_ops);
1217}
1218
1219#endif
1220
1221void __init ping_init(void)
1222{
1223        int i;
1224
1225        for (i = 0; i < PING_HTABLE_SIZE; i++)
1226                INIT_HLIST_NULLS_HEAD(&ping_table.hash[i], i);
1227        rwlock_init(&ping_table.lock);
1228}
1229