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