Back to home page

OSCL-LXR
 
 

     

0001 // SPDX-License-Identifier: GPL-2.0-or-later
0002 /*
0003  *  NET3:   Implementation of the ICMP protocol layer.
0004  *
0005  *      Alan Cox, <alan@lxorguk.ukuu.org.uk>
0006  *
0007  *  Some of the function names and the icmp unreach table for this
0008  *  module were derived from [icmp.c 1.0.11 06/02/93] by
0009  *  Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
0010  *  Other than that this module is a complete rewrite.
0011  *
0012  *  Fixes:
0013  *  Clemens Fruhwirth   :   introduce global icmp rate limiting
0014  *                  with icmp type masking ability instead
0015  *                  of broken per type icmp timeouts.
0016  *      Mike Shaver :   RFC1122 checks.
0017  *      Alan Cox    :   Multicast ping reply as self.
0018  *      Alan Cox    :   Fix atomicity lockup in ip_build_xmit
0019  *                  call.
0020  *      Alan Cox    :   Added 216,128 byte paths to the MTU
0021  *                  code.
0022  *      Martin Mares    :   RFC1812 checks.
0023  *      Martin Mares    :   Can be configured to follow redirects
0024  *                  if acting as a router _without_ a
0025  *                  routing protocol (RFC 1812).
0026  *      Martin Mares    :   Echo requests may be configured to
0027  *                  be ignored (RFC 1812).
0028  *      Martin Mares    :   Limitation of ICMP error message
0029  *                  transmit rate (RFC 1812).
0030  *      Martin Mares    :   TOS and Precedence set correctly
0031  *                  (RFC 1812).
0032  *      Martin Mares    :   Now copying as much data from the
0033  *                  original packet as we can without
0034  *                  exceeding 576 bytes (RFC 1812).
0035  *  Willy Konynenberg   :   Transparent proxying support.
0036  *      Keith Owens :   RFC1191 correction for 4.2BSD based
0037  *                  path MTU bug.
0038  *      Thomas Quinot   :   ICMP Dest Unreach codes up to 15 are
0039  *                  valid (RFC 1812).
0040  *      Andi Kleen  :   Check all packet lengths properly
0041  *                  and moved all kfree_skb() up to
0042  *                  icmp_rcv.
0043  *      Andi Kleen  :   Move the rate limit bookkeeping
0044  *                  into the dest entry and use a token
0045  *                  bucket filter (thanks to ANK). Make
0046  *                  the rates sysctl configurable.
0047  *      Yu Tianli   :   Fixed two ugly bugs in icmp_send
0048  *                  - IP option length was accounted wrongly
0049  *                  - ICMP header length was not accounted
0050  *                    at all.
0051  *              Tristan Greaves :       Added sysctl option to ignore bogus
0052  *                          broadcast responses from broken routers.
0053  *
0054  * To Fix:
0055  *
0056  *  - Should use skb_pull() instead of all the manual checking.
0057  *    This would also greatly simply some upper layer error handlers. --AK
0058  */
0059 
0060 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
0061 
0062 #include <linux/module.h>
0063 #include <linux/types.h>
0064 #include <linux/jiffies.h>
0065 #include <linux/kernel.h>
0066 #include <linux/fcntl.h>
0067 #include <linux/socket.h>
0068 #include <linux/in.h>
0069 #include <linux/inet.h>
0070 #include <linux/inetdevice.h>
0071 #include <linux/netdevice.h>
0072 #include <linux/string.h>
0073 #include <linux/netfilter_ipv4.h>
0074 #include <linux/slab.h>
0075 #include <net/snmp.h>
0076 #include <net/ip.h>
0077 #include <net/route.h>
0078 #include <net/protocol.h>
0079 #include <net/icmp.h>
0080 #include <net/tcp.h>
0081 #include <net/udp.h>
0082 #include <net/raw.h>
0083 #include <net/ping.h>
0084 #include <linux/skbuff.h>
0085 #include <net/sock.h>
0086 #include <linux/errno.h>
0087 #include <linux/timer.h>
0088 #include <linux/init.h>
0089 #include <linux/uaccess.h>
0090 #include <net/checksum.h>
0091 #include <net/xfrm.h>
0092 #include <net/inet_common.h>
0093 #include <net/ip_fib.h>
0094 #include <net/l3mdev.h>
0095 
0096 /*
0097  *  Build xmit assembly blocks
0098  */
0099 
0100 struct icmp_bxm {
0101     struct sk_buff *skb;
0102     int offset;
0103     int data_len;
0104 
0105     struct {
0106         struct icmphdr icmph;
0107         __be32         times[3];
0108     } data;
0109     int head_len;
0110     struct ip_options_data replyopts;
0111 };
0112 
0113 /* An array of errno for error messages from dest unreach. */
0114 /* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOST_UNREACH and SR_FAILED MUST be considered 'transient errs'. */
0115 
0116 const struct icmp_err icmp_err_convert[] = {
0117     {
0118         .errno = ENETUNREACH,   /* ICMP_NET_UNREACH */
0119         .fatal = 0,
0120     },
0121     {
0122         .errno = EHOSTUNREACH,  /* ICMP_HOST_UNREACH */
0123         .fatal = 0,
0124     },
0125     {
0126         .errno = ENOPROTOOPT    /* ICMP_PROT_UNREACH */,
0127         .fatal = 1,
0128     },
0129     {
0130         .errno = ECONNREFUSED,  /* ICMP_PORT_UNREACH */
0131         .fatal = 1,
0132     },
0133     {
0134         .errno = EMSGSIZE,  /* ICMP_FRAG_NEEDED */
0135         .fatal = 0,
0136     },
0137     {
0138         .errno = EOPNOTSUPP,    /* ICMP_SR_FAILED */
0139         .fatal = 0,
0140     },
0141     {
0142         .errno = ENETUNREACH,   /* ICMP_NET_UNKNOWN */
0143         .fatal = 1,
0144     },
0145     {
0146         .errno = EHOSTDOWN, /* ICMP_HOST_UNKNOWN */
0147         .fatal = 1,
0148     },
0149     {
0150         .errno = ENONET,    /* ICMP_HOST_ISOLATED */
0151         .fatal = 1,
0152     },
0153     {
0154         .errno = ENETUNREACH,   /* ICMP_NET_ANO */
0155         .fatal = 1,
0156     },
0157     {
0158         .errno = EHOSTUNREACH,  /* ICMP_HOST_ANO */
0159         .fatal = 1,
0160     },
0161     {
0162         .errno = ENETUNREACH,   /* ICMP_NET_UNR_TOS */
0163         .fatal = 0,
0164     },
0165     {
0166         .errno = EHOSTUNREACH,  /* ICMP_HOST_UNR_TOS */
0167         .fatal = 0,
0168     },
0169     {
0170         .errno = EHOSTUNREACH,  /* ICMP_PKT_FILTERED */
0171         .fatal = 1,
0172     },
0173     {
0174         .errno = EHOSTUNREACH,  /* ICMP_PREC_VIOLATION */
0175         .fatal = 1,
0176     },
0177     {
0178         .errno = EHOSTUNREACH,  /* ICMP_PREC_CUTOFF */
0179         .fatal = 1,
0180     },
0181 };
0182 EXPORT_SYMBOL(icmp_err_convert);
0183 
0184 /*
0185  *  ICMP control array. This specifies what to do with each ICMP.
0186  */
0187 
0188 struct icmp_control {
0189     enum skb_drop_reason (*handler)(struct sk_buff *skb);
0190     short   error;      /* This ICMP is classed as an error message */
0191 };
0192 
0193 static const struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];
0194 
0195 static DEFINE_PER_CPU(struct sock *, ipv4_icmp_sk);
0196 
0197 /* Called with BH disabled */
0198 static inline struct sock *icmp_xmit_lock(struct net *net)
0199 {
0200     struct sock *sk;
0201 
0202     sk = this_cpu_read(ipv4_icmp_sk);
0203 
0204     if (unlikely(!spin_trylock(&sk->sk_lock.slock))) {
0205         /* This can happen if the output path signals a
0206          * dst_link_failure() for an outgoing ICMP packet.
0207          */
0208         return NULL;
0209     }
0210     sock_net_set(sk, net);
0211     return sk;
0212 }
0213 
0214 static inline void icmp_xmit_unlock(struct sock *sk)
0215 {
0216     sock_net_set(sk, &init_net);
0217     spin_unlock(&sk->sk_lock.slock);
0218 }
0219 
0220 int sysctl_icmp_msgs_per_sec __read_mostly = 1000;
0221 int sysctl_icmp_msgs_burst __read_mostly = 50;
0222 
0223 static struct {
0224     spinlock_t  lock;
0225     u32     credit;
0226     u32     stamp;
0227 } icmp_global = {
0228     .lock       = __SPIN_LOCK_UNLOCKED(icmp_global.lock),
0229 };
0230 
0231 /**
0232  * icmp_global_allow - Are we allowed to send one more ICMP message ?
0233  *
0234  * Uses a token bucket to limit our ICMP messages to ~sysctl_icmp_msgs_per_sec.
0235  * Returns false if we reached the limit and can not send another packet.
0236  * Note: called with BH disabled
0237  */
0238 bool icmp_global_allow(void)
0239 {
0240     u32 credit, delta, incr = 0, now = (u32)jiffies;
0241     bool rc = false;
0242 
0243     /* Check if token bucket is empty and cannot be refilled
0244      * without taking the spinlock. The READ_ONCE() are paired
0245      * with the following WRITE_ONCE() in this same function.
0246      */
0247     if (!READ_ONCE(icmp_global.credit)) {
0248         delta = min_t(u32, now - READ_ONCE(icmp_global.stamp), HZ);
0249         if (delta < HZ / 50)
0250             return false;
0251     }
0252 
0253     spin_lock(&icmp_global.lock);
0254     delta = min_t(u32, now - icmp_global.stamp, HZ);
0255     if (delta >= HZ / 50) {
0256         incr = READ_ONCE(sysctl_icmp_msgs_per_sec) * delta / HZ;
0257         if (incr)
0258             WRITE_ONCE(icmp_global.stamp, now);
0259     }
0260     credit = min_t(u32, icmp_global.credit + incr,
0261                READ_ONCE(sysctl_icmp_msgs_burst));
0262     if (credit) {
0263         /* We want to use a credit of one in average, but need to randomize
0264          * it for security reasons.
0265          */
0266         credit = max_t(int, credit - prandom_u32_max(3), 0);
0267         rc = true;
0268     }
0269     WRITE_ONCE(icmp_global.credit, credit);
0270     spin_unlock(&icmp_global.lock);
0271     return rc;
0272 }
0273 EXPORT_SYMBOL(icmp_global_allow);
0274 
0275 static bool icmpv4_mask_allow(struct net *net, int type, int code)
0276 {
0277     if (type > NR_ICMP_TYPES)
0278         return true;
0279 
0280     /* Don't limit PMTU discovery. */
0281     if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
0282         return true;
0283 
0284     /* Limit if icmp type is enabled in ratemask. */
0285     if (!((1 << type) & READ_ONCE(net->ipv4.sysctl_icmp_ratemask)))
0286         return true;
0287 
0288     return false;
0289 }
0290 
0291 static bool icmpv4_global_allow(struct net *net, int type, int code)
0292 {
0293     if (icmpv4_mask_allow(net, type, code))
0294         return true;
0295 
0296     if (icmp_global_allow())
0297         return true;
0298 
0299     return false;
0300 }
0301 
0302 /*
0303  *  Send an ICMP frame.
0304  */
0305 
0306 static bool icmpv4_xrlim_allow(struct net *net, struct rtable *rt,
0307                    struct flowi4 *fl4, int type, int code)
0308 {
0309     struct dst_entry *dst = &rt->dst;
0310     struct inet_peer *peer;
0311     bool rc = true;
0312     int vif;
0313 
0314     if (icmpv4_mask_allow(net, type, code))
0315         goto out;
0316 
0317     /* No rate limit on loopback */
0318     if (dst->dev && (dst->dev->flags&IFF_LOOPBACK))
0319         goto out;
0320 
0321     vif = l3mdev_master_ifindex(dst->dev);
0322     peer = inet_getpeer_v4(net->ipv4.peers, fl4->daddr, vif, 1);
0323     rc = inet_peer_xrlim_allow(peer,
0324                    READ_ONCE(net->ipv4.sysctl_icmp_ratelimit));
0325     if (peer)
0326         inet_putpeer(peer);
0327 out:
0328     return rc;
0329 }
0330 
0331 /*
0332  *  Maintain the counters used in the SNMP statistics for outgoing ICMP
0333  */
0334 void icmp_out_count(struct net *net, unsigned char type)
0335 {
0336     ICMPMSGOUT_INC_STATS(net, type);
0337     ICMP_INC_STATS(net, ICMP_MIB_OUTMSGS);
0338 }
0339 
0340 /*
0341  *  Checksum each fragment, and on the first include the headers and final
0342  *  checksum.
0343  */
0344 static int icmp_glue_bits(void *from, char *to, int offset, int len, int odd,
0345               struct sk_buff *skb)
0346 {
0347     struct icmp_bxm *icmp_param = from;
0348     __wsum csum;
0349 
0350     csum = skb_copy_and_csum_bits(icmp_param->skb,
0351                       icmp_param->offset + offset,
0352                       to, len);
0353 
0354     skb->csum = csum_block_add(skb->csum, csum, odd);
0355     if (icmp_pointers[icmp_param->data.icmph.type].error)
0356         nf_ct_attach(skb, icmp_param->skb);
0357     return 0;
0358 }
0359 
0360 static void icmp_push_reply(struct sock *sk,
0361                 struct icmp_bxm *icmp_param,
0362                 struct flowi4 *fl4,
0363                 struct ipcm_cookie *ipc, struct rtable **rt)
0364 {
0365     struct sk_buff *skb;
0366 
0367     if (ip_append_data(sk, fl4, icmp_glue_bits, icmp_param,
0368                icmp_param->data_len+icmp_param->head_len,
0369                icmp_param->head_len,
0370                ipc, rt, MSG_DONTWAIT) < 0) {
0371         __ICMP_INC_STATS(sock_net(sk), ICMP_MIB_OUTERRORS);
0372         ip_flush_pending_frames(sk);
0373     } else if ((skb = skb_peek(&sk->sk_write_queue)) != NULL) {
0374         struct icmphdr *icmph = icmp_hdr(skb);
0375         __wsum csum;
0376         struct sk_buff *skb1;
0377 
0378         csum = csum_partial_copy_nocheck((void *)&icmp_param->data,
0379                          (char *)icmph,
0380                          icmp_param->head_len);
0381         skb_queue_walk(&sk->sk_write_queue, skb1) {
0382             csum = csum_add(csum, skb1->csum);
0383         }
0384         icmph->checksum = csum_fold(csum);
0385         skb->ip_summed = CHECKSUM_NONE;
0386         ip_push_pending_frames(sk, fl4);
0387     }
0388 }
0389 
0390 /*
0391  *  Driving logic for building and sending ICMP messages.
0392  */
0393 
0394 static void icmp_reply(struct icmp_bxm *icmp_param, struct sk_buff *skb)
0395 {
0396     struct ipcm_cookie ipc;
0397     struct rtable *rt = skb_rtable(skb);
0398     struct net *net = dev_net(rt->dst.dev);
0399     struct flowi4 fl4;
0400     struct sock *sk;
0401     struct inet_sock *inet;
0402     __be32 daddr, saddr;
0403     u32 mark = IP4_REPLY_MARK(net, skb->mark);
0404     int type = icmp_param->data.icmph.type;
0405     int code = icmp_param->data.icmph.code;
0406 
0407     if (ip_options_echo(net, &icmp_param->replyopts.opt.opt, skb))
0408         return;
0409 
0410     /* Needed by both icmp_global_allow and icmp_xmit_lock */
0411     local_bh_disable();
0412 
0413     /* global icmp_msgs_per_sec */
0414     if (!icmpv4_global_allow(net, type, code))
0415         goto out_bh_enable;
0416 
0417     sk = icmp_xmit_lock(net);
0418     if (!sk)
0419         goto out_bh_enable;
0420     inet = inet_sk(sk);
0421 
0422     icmp_param->data.icmph.checksum = 0;
0423 
0424     ipcm_init(&ipc);
0425     inet->tos = ip_hdr(skb)->tos;
0426     ipc.sockc.mark = mark;
0427     daddr = ipc.addr = ip_hdr(skb)->saddr;
0428     saddr = fib_compute_spec_dst(skb);
0429 
0430     if (icmp_param->replyopts.opt.opt.optlen) {
0431         ipc.opt = &icmp_param->replyopts.opt;
0432         if (ipc.opt->opt.srr)
0433             daddr = icmp_param->replyopts.opt.opt.faddr;
0434     }
0435     memset(&fl4, 0, sizeof(fl4));
0436     fl4.daddr = daddr;
0437     fl4.saddr = saddr;
0438     fl4.flowi4_mark = mark;
0439     fl4.flowi4_uid = sock_net_uid(net, NULL);
0440     fl4.flowi4_tos = RT_TOS(ip_hdr(skb)->tos);
0441     fl4.flowi4_proto = IPPROTO_ICMP;
0442     fl4.flowi4_oif = l3mdev_master_ifindex(skb->dev);
0443     security_skb_classify_flow(skb, flowi4_to_flowi_common(&fl4));
0444     rt = ip_route_output_key(net, &fl4);
0445     if (IS_ERR(rt))
0446         goto out_unlock;
0447     if (icmpv4_xrlim_allow(net, rt, &fl4, type, code))
0448         icmp_push_reply(sk, icmp_param, &fl4, &ipc, &rt);
0449     ip_rt_put(rt);
0450 out_unlock:
0451     icmp_xmit_unlock(sk);
0452 out_bh_enable:
0453     local_bh_enable();
0454 }
0455 
0456 /*
0457  * The device used for looking up which routing table to use for sending an ICMP
0458  * error is preferably the source whenever it is set, which should ensure the
0459  * icmp error can be sent to the source host, else lookup using the routing
0460  * table of the destination device, else use the main routing table (index 0).
0461  */
0462 static struct net_device *icmp_get_route_lookup_dev(struct sk_buff *skb)
0463 {
0464     struct net_device *route_lookup_dev = NULL;
0465 
0466     if (skb->dev)
0467         route_lookup_dev = skb->dev;
0468     else if (skb_dst(skb))
0469         route_lookup_dev = skb_dst(skb)->dev;
0470     return route_lookup_dev;
0471 }
0472 
0473 static struct rtable *icmp_route_lookup(struct net *net,
0474                     struct flowi4 *fl4,
0475                     struct sk_buff *skb_in,
0476                     const struct iphdr *iph,
0477                     __be32 saddr, u8 tos, u32 mark,
0478                     int type, int code,
0479                     struct icmp_bxm *param)
0480 {
0481     struct net_device *route_lookup_dev;
0482     struct rtable *rt, *rt2;
0483     struct flowi4 fl4_dec;
0484     int err;
0485 
0486     memset(fl4, 0, sizeof(*fl4));
0487     fl4->daddr = (param->replyopts.opt.opt.srr ?
0488               param->replyopts.opt.opt.faddr : iph->saddr);
0489     fl4->saddr = saddr;
0490     fl4->flowi4_mark = mark;
0491     fl4->flowi4_uid = sock_net_uid(net, NULL);
0492     fl4->flowi4_tos = RT_TOS(tos);
0493     fl4->flowi4_proto = IPPROTO_ICMP;
0494     fl4->fl4_icmp_type = type;
0495     fl4->fl4_icmp_code = code;
0496     route_lookup_dev = icmp_get_route_lookup_dev(skb_in);
0497     fl4->flowi4_oif = l3mdev_master_ifindex(route_lookup_dev);
0498 
0499     security_skb_classify_flow(skb_in, flowi4_to_flowi_common(fl4));
0500     rt = ip_route_output_key_hash(net, fl4, skb_in);
0501     if (IS_ERR(rt))
0502         return rt;
0503 
0504     /* No need to clone since we're just using its address. */
0505     rt2 = rt;
0506 
0507     rt = (struct rtable *) xfrm_lookup(net, &rt->dst,
0508                        flowi4_to_flowi(fl4), NULL, 0);
0509     if (!IS_ERR(rt)) {
0510         if (rt != rt2)
0511             return rt;
0512     } else if (PTR_ERR(rt) == -EPERM) {
0513         rt = NULL;
0514     } else
0515         return rt;
0516 
0517     err = xfrm_decode_session_reverse(skb_in, flowi4_to_flowi(&fl4_dec), AF_INET);
0518     if (err)
0519         goto relookup_failed;
0520 
0521     if (inet_addr_type_dev_table(net, route_lookup_dev,
0522                      fl4_dec.saddr) == RTN_LOCAL) {
0523         rt2 = __ip_route_output_key(net, &fl4_dec);
0524         if (IS_ERR(rt2))
0525             err = PTR_ERR(rt2);
0526     } else {
0527         struct flowi4 fl4_2 = {};
0528         unsigned long orefdst;
0529 
0530         fl4_2.daddr = fl4_dec.saddr;
0531         rt2 = ip_route_output_key(net, &fl4_2);
0532         if (IS_ERR(rt2)) {
0533             err = PTR_ERR(rt2);
0534             goto relookup_failed;
0535         }
0536         /* Ugh! */
0537         orefdst = skb_in->_skb_refdst; /* save old refdst */
0538         skb_dst_set(skb_in, NULL);
0539         err = ip_route_input(skb_in, fl4_dec.daddr, fl4_dec.saddr,
0540                      RT_TOS(tos), rt2->dst.dev);
0541 
0542         dst_release(&rt2->dst);
0543         rt2 = skb_rtable(skb_in);
0544         skb_in->_skb_refdst = orefdst; /* restore old refdst */
0545     }
0546 
0547     if (err)
0548         goto relookup_failed;
0549 
0550     rt2 = (struct rtable *) xfrm_lookup(net, &rt2->dst,
0551                         flowi4_to_flowi(&fl4_dec), NULL,
0552                         XFRM_LOOKUP_ICMP);
0553     if (!IS_ERR(rt2)) {
0554         dst_release(&rt->dst);
0555         memcpy(fl4, &fl4_dec, sizeof(*fl4));
0556         rt = rt2;
0557     } else if (PTR_ERR(rt2) == -EPERM) {
0558         if (rt)
0559             dst_release(&rt->dst);
0560         return rt2;
0561     } else {
0562         err = PTR_ERR(rt2);
0563         goto relookup_failed;
0564     }
0565     return rt;
0566 
0567 relookup_failed:
0568     if (rt)
0569         return rt;
0570     return ERR_PTR(err);
0571 }
0572 
0573 /*
0574  *  Send an ICMP message in response to a situation
0575  *
0576  *  RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header.
0577  *        MAY send more (we do).
0578  *          MUST NOT change this header information.
0579  *          MUST NOT reply to a multicast/broadcast IP address.
0580  *          MUST NOT reply to a multicast/broadcast MAC address.
0581  *          MUST reply to only the first fragment.
0582  */
0583 
0584 void __icmp_send(struct sk_buff *skb_in, int type, int code, __be32 info,
0585          const struct ip_options *opt)
0586 {
0587     struct iphdr *iph;
0588     int room;
0589     struct icmp_bxm icmp_param;
0590     struct rtable *rt = skb_rtable(skb_in);
0591     struct ipcm_cookie ipc;
0592     struct flowi4 fl4;
0593     __be32 saddr;
0594     u8  tos;
0595     u32 mark;
0596     struct net *net;
0597     struct sock *sk;
0598 
0599     if (!rt)
0600         goto out;
0601 
0602     if (rt->dst.dev)
0603         net = dev_net(rt->dst.dev);
0604     else if (skb_in->dev)
0605         net = dev_net(skb_in->dev);
0606     else
0607         goto out;
0608 
0609     /*
0610      *  Find the original header. It is expected to be valid, of course.
0611      *  Check this, icmp_send is called from the most obscure devices
0612      *  sometimes.
0613      */
0614     iph = ip_hdr(skb_in);
0615 
0616     if ((u8 *)iph < skb_in->head ||
0617         (skb_network_header(skb_in) + sizeof(*iph)) >
0618         skb_tail_pointer(skb_in))
0619         goto out;
0620 
0621     /*
0622      *  No replies to physical multicast/broadcast
0623      */
0624     if (skb_in->pkt_type != PACKET_HOST)
0625         goto out;
0626 
0627     /*
0628      *  Now check at the protocol level
0629      */
0630     if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
0631         goto out;
0632 
0633     /*
0634      *  Only reply to fragment 0. We byte re-order the constant
0635      *  mask for efficiency.
0636      */
0637     if (iph->frag_off & htons(IP_OFFSET))
0638         goto out;
0639 
0640     /*
0641      *  If we send an ICMP error to an ICMP error a mess would result..
0642      */
0643     if (icmp_pointers[type].error) {
0644         /*
0645          *  We are an error, check if we are replying to an
0646          *  ICMP error
0647          */
0648         if (iph->protocol == IPPROTO_ICMP) {
0649             u8 _inner_type, *itp;
0650 
0651             itp = skb_header_pointer(skb_in,
0652                          skb_network_header(skb_in) +
0653                          (iph->ihl << 2) +
0654                          offsetof(struct icmphdr,
0655                               type) -
0656                          skb_in->data,
0657                          sizeof(_inner_type),
0658                          &_inner_type);
0659             if (!itp)
0660                 goto out;
0661 
0662             /*
0663              *  Assume any unknown ICMP type is an error. This
0664              *  isn't specified by the RFC, but think about it..
0665              */
0666             if (*itp > NR_ICMP_TYPES ||
0667                 icmp_pointers[*itp].error)
0668                 goto out;
0669         }
0670     }
0671 
0672     /* Needed by both icmp_global_allow and icmp_xmit_lock */
0673     local_bh_disable();
0674 
0675     /* Check global sysctl_icmp_msgs_per_sec ratelimit, unless
0676      * incoming dev is loopback.  If outgoing dev change to not be
0677      * loopback, then peer ratelimit still work (in icmpv4_xrlim_allow)
0678      */
0679     if (!(skb_in->dev && (skb_in->dev->flags&IFF_LOOPBACK)) &&
0680           !icmpv4_global_allow(net, type, code))
0681         goto out_bh_enable;
0682 
0683     sk = icmp_xmit_lock(net);
0684     if (!sk)
0685         goto out_bh_enable;
0686 
0687     /*
0688      *  Construct source address and options.
0689      */
0690 
0691     saddr = iph->daddr;
0692     if (!(rt->rt_flags & RTCF_LOCAL)) {
0693         struct net_device *dev = NULL;
0694 
0695         rcu_read_lock();
0696         if (rt_is_input_route(rt) &&
0697             READ_ONCE(net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr))
0698             dev = dev_get_by_index_rcu(net, inet_iif(skb_in));
0699 
0700         if (dev)
0701             saddr = inet_select_addr(dev, iph->saddr,
0702                          RT_SCOPE_LINK);
0703         else
0704             saddr = 0;
0705         rcu_read_unlock();
0706     }
0707 
0708     tos = icmp_pointers[type].error ? (RT_TOS(iph->tos) |
0709                        IPTOS_PREC_INTERNETCONTROL) :
0710                        iph->tos;
0711     mark = IP4_REPLY_MARK(net, skb_in->mark);
0712 
0713     if (__ip_options_echo(net, &icmp_param.replyopts.opt.opt, skb_in, opt))
0714         goto out_unlock;
0715 
0716 
0717     /*
0718      *  Prepare data for ICMP header.
0719      */
0720 
0721     icmp_param.data.icmph.type   = type;
0722     icmp_param.data.icmph.code   = code;
0723     icmp_param.data.icmph.un.gateway = info;
0724     icmp_param.data.icmph.checksum   = 0;
0725     icmp_param.skb    = skb_in;
0726     icmp_param.offset = skb_network_offset(skb_in);
0727     inet_sk(sk)->tos = tos;
0728     ipcm_init(&ipc);
0729     ipc.addr = iph->saddr;
0730     ipc.opt = &icmp_param.replyopts.opt;
0731     ipc.sockc.mark = mark;
0732 
0733     rt = icmp_route_lookup(net, &fl4, skb_in, iph, saddr, tos, mark,
0734                    type, code, &icmp_param);
0735     if (IS_ERR(rt))
0736         goto out_unlock;
0737 
0738     /* peer icmp_ratelimit */
0739     if (!icmpv4_xrlim_allow(net, rt, &fl4, type, code))
0740         goto ende;
0741 
0742     /* RFC says return as much as we can without exceeding 576 bytes. */
0743 
0744     room = dst_mtu(&rt->dst);
0745     if (room > 576)
0746         room = 576;
0747     room -= sizeof(struct iphdr) + icmp_param.replyopts.opt.opt.optlen;
0748     room -= sizeof(struct icmphdr);
0749 
0750     icmp_param.data_len = skb_in->len - icmp_param.offset;
0751     if (icmp_param.data_len > room)
0752         icmp_param.data_len = room;
0753     icmp_param.head_len = sizeof(struct icmphdr);
0754 
0755     /* if we don't have a source address at this point, fall back to the
0756      * dummy address instead of sending out a packet with a source address
0757      * of 0.0.0.0
0758      */
0759     if (!fl4.saddr)
0760         fl4.saddr = htonl(INADDR_DUMMY);
0761 
0762     icmp_push_reply(sk, &icmp_param, &fl4, &ipc, &rt);
0763 ende:
0764     ip_rt_put(rt);
0765 out_unlock:
0766     icmp_xmit_unlock(sk);
0767 out_bh_enable:
0768     local_bh_enable();
0769 out:;
0770 }
0771 EXPORT_SYMBOL(__icmp_send);
0772 
0773 #if IS_ENABLED(CONFIG_NF_NAT)
0774 #include <net/netfilter/nf_conntrack.h>
0775 void icmp_ndo_send(struct sk_buff *skb_in, int type, int code, __be32 info)
0776 {
0777     struct sk_buff *cloned_skb = NULL;
0778     struct ip_options opts = { 0 };
0779     enum ip_conntrack_info ctinfo;
0780     struct nf_conn *ct;
0781     __be32 orig_ip;
0782 
0783     ct = nf_ct_get(skb_in, &ctinfo);
0784     if (!ct || !(ct->status & IPS_SRC_NAT)) {
0785         __icmp_send(skb_in, type, code, info, &opts);
0786         return;
0787     }
0788 
0789     if (skb_shared(skb_in))
0790         skb_in = cloned_skb = skb_clone(skb_in, GFP_ATOMIC);
0791 
0792     if (unlikely(!skb_in || skb_network_header(skb_in) < skb_in->head ||
0793         (skb_network_header(skb_in) + sizeof(struct iphdr)) >
0794         skb_tail_pointer(skb_in) || skb_ensure_writable(skb_in,
0795         skb_network_offset(skb_in) + sizeof(struct iphdr))))
0796         goto out;
0797 
0798     orig_ip = ip_hdr(skb_in)->saddr;
0799     ip_hdr(skb_in)->saddr = ct->tuplehash[0].tuple.src.u3.ip;
0800     __icmp_send(skb_in, type, code, info, &opts);
0801     ip_hdr(skb_in)->saddr = orig_ip;
0802 out:
0803     consume_skb(cloned_skb);
0804 }
0805 EXPORT_SYMBOL(icmp_ndo_send);
0806 #endif
0807 
0808 static void icmp_socket_deliver(struct sk_buff *skb, u32 info)
0809 {
0810     const struct iphdr *iph = (const struct iphdr *)skb->data;
0811     const struct net_protocol *ipprot;
0812     int protocol = iph->protocol;
0813 
0814     /* Checkin full IP header plus 8 bytes of protocol to
0815      * avoid additional coding at protocol handlers.
0816      */
0817     if (!pskb_may_pull(skb, iph->ihl * 4 + 8)) {
0818         __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS);
0819         return;
0820     }
0821 
0822     raw_icmp_error(skb, protocol, info);
0823 
0824     ipprot = rcu_dereference(inet_protos[protocol]);
0825     if (ipprot && ipprot->err_handler)
0826         ipprot->err_handler(skb, info);
0827 }
0828 
0829 static bool icmp_tag_validation(int proto)
0830 {
0831     bool ok;
0832 
0833     rcu_read_lock();
0834     ok = rcu_dereference(inet_protos[proto])->icmp_strict_tag_validation;
0835     rcu_read_unlock();
0836     return ok;
0837 }
0838 
0839 /*
0840  *  Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEEDED, ICMP_QUENCH, and
0841  *  ICMP_PARAMETERPROB.
0842  */
0843 
0844 static enum skb_drop_reason icmp_unreach(struct sk_buff *skb)
0845 {
0846     enum skb_drop_reason reason = SKB_NOT_DROPPED_YET;
0847     const struct iphdr *iph;
0848     struct icmphdr *icmph;
0849     struct net *net;
0850     u32 info = 0;
0851 
0852     net = dev_net(skb_dst(skb)->dev);
0853 
0854     /*
0855      *  Incomplete header ?
0856      *  Only checks for the IP header, there should be an
0857      *  additional check for longer headers in upper levels.
0858      */
0859 
0860     if (!pskb_may_pull(skb, sizeof(struct iphdr)))
0861         goto out_err;
0862 
0863     icmph = icmp_hdr(skb);
0864     iph   = (const struct iphdr *)skb->data;
0865 
0866     if (iph->ihl < 5)  { /* Mangled header, drop. */
0867         reason = SKB_DROP_REASON_IP_INHDR;
0868         goto out_err;
0869     }
0870 
0871     switch (icmph->type) {
0872     case ICMP_DEST_UNREACH:
0873         switch (icmph->code & 15) {
0874         case ICMP_NET_UNREACH:
0875         case ICMP_HOST_UNREACH:
0876         case ICMP_PROT_UNREACH:
0877         case ICMP_PORT_UNREACH:
0878             break;
0879         case ICMP_FRAG_NEEDED:
0880             /* for documentation of the ip_no_pmtu_disc
0881              * values please see
0882              * Documentation/networking/ip-sysctl.rst
0883              */
0884             switch (READ_ONCE(net->ipv4.sysctl_ip_no_pmtu_disc)) {
0885             default:
0886                 net_dbg_ratelimited("%pI4: fragmentation needed and DF set\n",
0887                             &iph->daddr);
0888                 break;
0889             case 2:
0890                 goto out;
0891             case 3:
0892                 if (!icmp_tag_validation(iph->protocol))
0893                     goto out;
0894                 fallthrough;
0895             case 0:
0896                 info = ntohs(icmph->un.frag.mtu);
0897             }
0898             break;
0899         case ICMP_SR_FAILED:
0900             net_dbg_ratelimited("%pI4: Source Route Failed\n",
0901                         &iph->daddr);
0902             break;
0903         default:
0904             break;
0905         }
0906         if (icmph->code > NR_ICMP_UNREACH)
0907             goto out;
0908         break;
0909     case ICMP_PARAMETERPROB:
0910         info = ntohl(icmph->un.gateway) >> 24;
0911         break;
0912     case ICMP_TIME_EXCEEDED:
0913         __ICMP_INC_STATS(net, ICMP_MIB_INTIMEEXCDS);
0914         if (icmph->code == ICMP_EXC_FRAGTIME)
0915             goto out;
0916         break;
0917     }
0918 
0919     /*
0920      *  Throw it at our lower layers
0921      *
0922      *  RFC 1122: 3.2.2 MUST extract the protocol ID from the passed
0923      *        header.
0924      *  RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the
0925      *        transport layer.
0926      *  RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to
0927      *        transport layer.
0928      */
0929 
0930     /*
0931      *  Check the other end isn't violating RFC 1122. Some routers send
0932      *  bogus responses to broadcast frames. If you see this message
0933      *  first check your netmask matches at both ends, if it does then
0934      *  get the other vendor to fix their kit.
0935      */
0936 
0937     if (!READ_ONCE(net->ipv4.sysctl_icmp_ignore_bogus_error_responses) &&
0938         inet_addr_type_dev_table(net, skb->dev, iph->daddr) == RTN_BROADCAST) {
0939         net_warn_ratelimited("%pI4 sent an invalid ICMP type %u, code %u error to a broadcast: %pI4 on %s\n",
0940                      &ip_hdr(skb)->saddr,
0941                      icmph->type, icmph->code,
0942                      &iph->daddr, skb->dev->name);
0943         goto out;
0944     }
0945 
0946     icmp_socket_deliver(skb, info);
0947 
0948 out:
0949     return reason;
0950 out_err:
0951     __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
0952     return reason ?: SKB_DROP_REASON_NOT_SPECIFIED;
0953 }
0954 
0955 
0956 /*
0957  *  Handle ICMP_REDIRECT.
0958  */
0959 
0960 static enum skb_drop_reason icmp_redirect(struct sk_buff *skb)
0961 {
0962     if (skb->len < sizeof(struct iphdr)) {
0963         __ICMP_INC_STATS(dev_net(skb->dev), ICMP_MIB_INERRORS);
0964         return SKB_DROP_REASON_PKT_TOO_SMALL;
0965     }
0966 
0967     if (!pskb_may_pull(skb, sizeof(struct iphdr))) {
0968         /* there aught to be a stat */
0969         return SKB_DROP_REASON_NOMEM;
0970     }
0971 
0972     icmp_socket_deliver(skb, ntohl(icmp_hdr(skb)->un.gateway));
0973     return SKB_NOT_DROPPED_YET;
0974 }
0975 
0976 /*
0977  *  Handle ICMP_ECHO ("ping") and ICMP_EXT_ECHO ("PROBE") requests.
0978  *
0979  *  RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo
0980  *        requests.
0981  *  RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be
0982  *        included in the reply.
0983  *  RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring
0984  *        echo requests, MUST have default=NOT.
0985  *  RFC 8335: 8 MUST have a config option to enable/disable ICMP
0986  *        Extended Echo Functionality, MUST be disabled by default
0987  *  See also WRT handling of options once they are done and working.
0988  */
0989 
0990 static enum skb_drop_reason icmp_echo(struct sk_buff *skb)
0991 {
0992     struct icmp_bxm icmp_param;
0993     struct net *net;
0994 
0995     net = dev_net(skb_dst(skb)->dev);
0996     /* should there be an ICMP stat for ignored echos? */
0997     if (READ_ONCE(net->ipv4.sysctl_icmp_echo_ignore_all))
0998         return SKB_NOT_DROPPED_YET;
0999 
1000     icmp_param.data.icmph      = *icmp_hdr(skb);
1001     icmp_param.skb         = skb;
1002     icmp_param.offset      = 0;
1003     icmp_param.data_len    = skb->len;
1004     icmp_param.head_len    = sizeof(struct icmphdr);
1005 
1006     if (icmp_param.data.icmph.type == ICMP_ECHO)
1007         icmp_param.data.icmph.type = ICMP_ECHOREPLY;
1008     else if (!icmp_build_probe(skb, &icmp_param.data.icmph))
1009         return SKB_NOT_DROPPED_YET;
1010 
1011     icmp_reply(&icmp_param, skb);
1012     return SKB_NOT_DROPPED_YET;
1013 }
1014 
1015 /*  Helper for icmp_echo and icmpv6_echo_reply.
1016  *  Searches for net_device that matches PROBE interface identifier
1017  *      and builds PROBE reply message in icmphdr.
1018  *
1019  *  Returns false if PROBE responses are disabled via sysctl
1020  */
1021 
1022 bool icmp_build_probe(struct sk_buff *skb, struct icmphdr *icmphdr)
1023 {
1024     struct icmp_ext_hdr *ext_hdr, _ext_hdr;
1025     struct icmp_ext_echo_iio *iio, _iio;
1026     struct net *net = dev_net(skb->dev);
1027     struct net_device *dev;
1028     char buff[IFNAMSIZ];
1029     u16 ident_len;
1030     u8 status;
1031 
1032     if (!READ_ONCE(net->ipv4.sysctl_icmp_echo_enable_probe))
1033         return false;
1034 
1035     /* We currently only support probing interfaces on the proxy node
1036      * Check to ensure L-bit is set
1037      */
1038     if (!(ntohs(icmphdr->un.echo.sequence) & 1))
1039         return false;
1040     /* Clear status bits in reply message */
1041     icmphdr->un.echo.sequence &= htons(0xFF00);
1042     if (icmphdr->type == ICMP_EXT_ECHO)
1043         icmphdr->type = ICMP_EXT_ECHOREPLY;
1044     else
1045         icmphdr->type = ICMPV6_EXT_ECHO_REPLY;
1046     ext_hdr = skb_header_pointer(skb, 0, sizeof(_ext_hdr), &_ext_hdr);
1047     /* Size of iio is class_type dependent.
1048      * Only check header here and assign length based on ctype in the switch statement
1049      */
1050     iio = skb_header_pointer(skb, sizeof(_ext_hdr), sizeof(iio->extobj_hdr), &_iio);
1051     if (!ext_hdr || !iio)
1052         goto send_mal_query;
1053     if (ntohs(iio->extobj_hdr.length) <= sizeof(iio->extobj_hdr) ||
1054         ntohs(iio->extobj_hdr.length) > sizeof(_iio))
1055         goto send_mal_query;
1056     ident_len = ntohs(iio->extobj_hdr.length) - sizeof(iio->extobj_hdr);
1057     iio = skb_header_pointer(skb, sizeof(_ext_hdr),
1058                  sizeof(iio->extobj_hdr) + ident_len, &_iio);
1059     if (!iio)
1060         goto send_mal_query;
1061 
1062     status = 0;
1063     dev = NULL;
1064     switch (iio->extobj_hdr.class_type) {
1065     case ICMP_EXT_ECHO_CTYPE_NAME:
1066         if (ident_len >= IFNAMSIZ)
1067             goto send_mal_query;
1068         memset(buff, 0, sizeof(buff));
1069         memcpy(buff, &iio->ident.name, ident_len);
1070         dev = dev_get_by_name(net, buff);
1071         break;
1072     case ICMP_EXT_ECHO_CTYPE_INDEX:
1073         if (ident_len != sizeof(iio->ident.ifindex))
1074             goto send_mal_query;
1075         dev = dev_get_by_index(net, ntohl(iio->ident.ifindex));
1076         break;
1077     case ICMP_EXT_ECHO_CTYPE_ADDR:
1078         if (ident_len < sizeof(iio->ident.addr.ctype3_hdr) ||
1079             ident_len != sizeof(iio->ident.addr.ctype3_hdr) +
1080                  iio->ident.addr.ctype3_hdr.addrlen)
1081             goto send_mal_query;
1082         switch (ntohs(iio->ident.addr.ctype3_hdr.afi)) {
1083         case ICMP_AFI_IP:
1084             if (iio->ident.addr.ctype3_hdr.addrlen != sizeof(struct in_addr))
1085                 goto send_mal_query;
1086             dev = ip_dev_find(net, iio->ident.addr.ip_addr.ipv4_addr);
1087             break;
1088 #if IS_ENABLED(CONFIG_IPV6)
1089         case ICMP_AFI_IP6:
1090             if (iio->ident.addr.ctype3_hdr.addrlen != sizeof(struct in6_addr))
1091                 goto send_mal_query;
1092             dev = ipv6_stub->ipv6_dev_find(net, &iio->ident.addr.ip_addr.ipv6_addr, dev);
1093             dev_hold(dev);
1094             break;
1095 #endif
1096         default:
1097             goto send_mal_query;
1098         }
1099         break;
1100     default:
1101         goto send_mal_query;
1102     }
1103     if (!dev) {
1104         icmphdr->code = ICMP_EXT_CODE_NO_IF;
1105         return true;
1106     }
1107     /* Fill bits in reply message */
1108     if (dev->flags & IFF_UP)
1109         status |= ICMP_EXT_ECHOREPLY_ACTIVE;
1110     if (__in_dev_get_rcu(dev) && __in_dev_get_rcu(dev)->ifa_list)
1111         status |= ICMP_EXT_ECHOREPLY_IPV4;
1112     if (!list_empty(&rcu_dereference(dev->ip6_ptr)->addr_list))
1113         status |= ICMP_EXT_ECHOREPLY_IPV6;
1114     dev_put(dev);
1115     icmphdr->un.echo.sequence |= htons(status);
1116     return true;
1117 send_mal_query:
1118     icmphdr->code = ICMP_EXT_CODE_MAL_QUERY;
1119     return true;
1120 }
1121 EXPORT_SYMBOL_GPL(icmp_build_probe);
1122 
1123 /*
1124  *  Handle ICMP Timestamp requests.
1125  *  RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
1126  *        SHOULD be in the kernel for minimum random latency.
1127  *        MUST be accurate to a few minutes.
1128  *        MUST be updated at least at 15Hz.
1129  */
1130 static enum skb_drop_reason icmp_timestamp(struct sk_buff *skb)
1131 {
1132     struct icmp_bxm icmp_param;
1133     /*
1134      *  Too short.
1135      */
1136     if (skb->len < 4)
1137         goto out_err;
1138 
1139     /*
1140      *  Fill in the current time as ms since midnight UT:
1141      */
1142     icmp_param.data.times[1] = inet_current_timestamp();
1143     icmp_param.data.times[2] = icmp_param.data.times[1];
1144 
1145     BUG_ON(skb_copy_bits(skb, 0, &icmp_param.data.times[0], 4));
1146 
1147     icmp_param.data.icmph      = *icmp_hdr(skb);
1148     icmp_param.data.icmph.type = ICMP_TIMESTAMPREPLY;
1149     icmp_param.data.icmph.code = 0;
1150     icmp_param.skb         = skb;
1151     icmp_param.offset      = 0;
1152     icmp_param.data_len    = 0;
1153     icmp_param.head_len    = sizeof(struct icmphdr) + 12;
1154     icmp_reply(&icmp_param, skb);
1155     return SKB_NOT_DROPPED_YET;
1156 
1157 out_err:
1158     __ICMP_INC_STATS(dev_net(skb_dst(skb)->dev), ICMP_MIB_INERRORS);
1159     return SKB_DROP_REASON_PKT_TOO_SMALL;
1160 }
1161 
1162 static enum skb_drop_reason icmp_discard(struct sk_buff *skb)
1163 {
1164     /* pretend it was a success */
1165     return SKB_NOT_DROPPED_YET;
1166 }
1167 
1168 /*
1169  *  Deal with incoming ICMP packets.
1170  */
1171 int icmp_rcv(struct sk_buff *skb)
1172 {
1173     enum skb_drop_reason reason = SKB_DROP_REASON_NOT_SPECIFIED;
1174     struct rtable *rt = skb_rtable(skb);
1175     struct net *net = dev_net(rt->dst.dev);
1176     struct icmphdr *icmph;
1177 
1178     if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
1179         struct sec_path *sp = skb_sec_path(skb);
1180         int nh;
1181 
1182         if (!(sp && sp->xvec[sp->len - 1]->props.flags &
1183                  XFRM_STATE_ICMP)) {
1184             reason = SKB_DROP_REASON_XFRM_POLICY;
1185             goto drop;
1186         }
1187 
1188         if (!pskb_may_pull(skb, sizeof(*icmph) + sizeof(struct iphdr)))
1189             goto drop;
1190 
1191         nh = skb_network_offset(skb);
1192         skb_set_network_header(skb, sizeof(*icmph));
1193 
1194         if (!xfrm4_policy_check_reverse(NULL, XFRM_POLICY_IN,
1195                         skb)) {
1196             reason = SKB_DROP_REASON_XFRM_POLICY;
1197             goto drop;
1198         }
1199 
1200         skb_set_network_header(skb, nh);
1201     }
1202 
1203     __ICMP_INC_STATS(net, ICMP_MIB_INMSGS);
1204 
1205     if (skb_checksum_simple_validate(skb))
1206         goto csum_error;
1207 
1208     if (!pskb_pull(skb, sizeof(*icmph)))
1209         goto error;
1210 
1211     icmph = icmp_hdr(skb);
1212 
1213     ICMPMSGIN_INC_STATS(net, icmph->type);
1214 
1215     /* Check for ICMP Extended Echo (PROBE) messages */
1216     if (icmph->type == ICMP_EXT_ECHO) {
1217         /* We can't use icmp_pointers[].handler() because it is an array of
1218          * size NR_ICMP_TYPES + 1 (19 elements) and PROBE has code 42.
1219          */
1220         reason = icmp_echo(skb);
1221         goto reason_check;
1222     }
1223 
1224     if (icmph->type == ICMP_EXT_ECHOREPLY) {
1225         reason = ping_rcv(skb);
1226         goto reason_check;
1227     }
1228 
1229     /*
1230      *  18 is the highest 'known' ICMP type. Anything else is a mystery
1231      *
1232      *  RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently
1233      *        discarded.
1234      */
1235     if (icmph->type > NR_ICMP_TYPES) {
1236         reason = SKB_DROP_REASON_UNHANDLED_PROTO;
1237         goto error;
1238     }
1239 
1240     /*
1241      *  Parse the ICMP message
1242      */
1243 
1244     if (rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) {
1245         /*
1246          *  RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be
1247          *    silently ignored (we let user decide with a sysctl).
1248          *  RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently
1249          *    discarded if to broadcast/multicast.
1250          */
1251         if ((icmph->type == ICMP_ECHO ||
1252              icmph->type == ICMP_TIMESTAMP) &&
1253             READ_ONCE(net->ipv4.sysctl_icmp_echo_ignore_broadcasts)) {
1254             reason = SKB_DROP_REASON_INVALID_PROTO;
1255             goto error;
1256         }
1257         if (icmph->type != ICMP_ECHO &&
1258             icmph->type != ICMP_TIMESTAMP &&
1259             icmph->type != ICMP_ADDRESS &&
1260             icmph->type != ICMP_ADDRESSREPLY) {
1261             reason = SKB_DROP_REASON_INVALID_PROTO;
1262             goto error;
1263         }
1264     }
1265 
1266     reason = icmp_pointers[icmph->type].handler(skb);
1267 reason_check:
1268     if (!reason)  {
1269         consume_skb(skb);
1270         return NET_RX_SUCCESS;
1271     }
1272 
1273 drop:
1274     kfree_skb_reason(skb, reason);
1275     return NET_RX_DROP;
1276 csum_error:
1277     reason = SKB_DROP_REASON_ICMP_CSUM;
1278     __ICMP_INC_STATS(net, ICMP_MIB_CSUMERRORS);
1279 error:
1280     __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
1281     goto drop;
1282 }
1283 
1284 static bool ip_icmp_error_rfc4884_validate(const struct sk_buff *skb, int off)
1285 {
1286     struct icmp_extobj_hdr *objh, _objh;
1287     struct icmp_ext_hdr *exth, _exth;
1288     u16 olen;
1289 
1290     exth = skb_header_pointer(skb, off, sizeof(_exth), &_exth);
1291     if (!exth)
1292         return false;
1293     if (exth->version != 2)
1294         return true;
1295 
1296     if (exth->checksum &&
1297         csum_fold(skb_checksum(skb, off, skb->len - off, 0)))
1298         return false;
1299 
1300     off += sizeof(_exth);
1301     while (off < skb->len) {
1302         objh = skb_header_pointer(skb, off, sizeof(_objh), &_objh);
1303         if (!objh)
1304             return false;
1305 
1306         olen = ntohs(objh->length);
1307         if (olen < sizeof(_objh))
1308             return false;
1309 
1310         off += olen;
1311         if (off > skb->len)
1312             return false;
1313     }
1314 
1315     return true;
1316 }
1317 
1318 void ip_icmp_error_rfc4884(const struct sk_buff *skb,
1319                struct sock_ee_data_rfc4884 *out,
1320                int thlen, int off)
1321 {
1322     int hlen;
1323 
1324     /* original datagram headers: end of icmph to payload (skb->data) */
1325     hlen = -skb_transport_offset(skb) - thlen;
1326 
1327     /* per rfc 4884: minimal datagram length of 128 bytes */
1328     if (off < 128 || off < hlen)
1329         return;
1330 
1331     /* kernel has stripped headers: return payload offset in bytes */
1332     off -= hlen;
1333     if (off + sizeof(struct icmp_ext_hdr) > skb->len)
1334         return;
1335 
1336     out->len = off;
1337 
1338     if (!ip_icmp_error_rfc4884_validate(skb, off))
1339         out->flags |= SO_EE_RFC4884_FLAG_INVALID;
1340 }
1341 EXPORT_SYMBOL_GPL(ip_icmp_error_rfc4884);
1342 
1343 int icmp_err(struct sk_buff *skb, u32 info)
1344 {
1345     struct iphdr *iph = (struct iphdr *)skb->data;
1346     int offset = iph->ihl<<2;
1347     struct icmphdr *icmph = (struct icmphdr *)(skb->data + offset);
1348     int type = icmp_hdr(skb)->type;
1349     int code = icmp_hdr(skb)->code;
1350     struct net *net = dev_net(skb->dev);
1351 
1352     /*
1353      * Use ping_err to handle all icmp errors except those
1354      * triggered by ICMP_ECHOREPLY which sent from kernel.
1355      */
1356     if (icmph->type != ICMP_ECHOREPLY) {
1357         ping_err(skb, offset, info);
1358         return 0;
1359     }
1360 
1361     if (type == ICMP_DEST_UNREACH && code == ICMP_FRAG_NEEDED)
1362         ipv4_update_pmtu(skb, net, info, 0, IPPROTO_ICMP);
1363     else if (type == ICMP_REDIRECT)
1364         ipv4_redirect(skb, net, 0, IPPROTO_ICMP);
1365 
1366     return 0;
1367 }
1368 
1369 /*
1370  *  This table is the definition of how we handle ICMP.
1371  */
1372 static const struct icmp_control icmp_pointers[NR_ICMP_TYPES + 1] = {
1373     [ICMP_ECHOREPLY] = {
1374         .handler = ping_rcv,
1375     },
1376     [1] = {
1377         .handler = icmp_discard,
1378         .error = 1,
1379     },
1380     [2] = {
1381         .handler = icmp_discard,
1382         .error = 1,
1383     },
1384     [ICMP_DEST_UNREACH] = {
1385         .handler = icmp_unreach,
1386         .error = 1,
1387     },
1388     [ICMP_SOURCE_QUENCH] = {
1389         .handler = icmp_unreach,
1390         .error = 1,
1391     },
1392     [ICMP_REDIRECT] = {
1393         .handler = icmp_redirect,
1394         .error = 1,
1395     },
1396     [6] = {
1397         .handler = icmp_discard,
1398         .error = 1,
1399     },
1400     [7] = {
1401         .handler = icmp_discard,
1402         .error = 1,
1403     },
1404     [ICMP_ECHO] = {
1405         .handler = icmp_echo,
1406     },
1407     [9] = {
1408         .handler = icmp_discard,
1409         .error = 1,
1410     },
1411     [10] = {
1412         .handler = icmp_discard,
1413         .error = 1,
1414     },
1415     [ICMP_TIME_EXCEEDED] = {
1416         .handler = icmp_unreach,
1417         .error = 1,
1418     },
1419     [ICMP_PARAMETERPROB] = {
1420         .handler = icmp_unreach,
1421         .error = 1,
1422     },
1423     [ICMP_TIMESTAMP] = {
1424         .handler = icmp_timestamp,
1425     },
1426     [ICMP_TIMESTAMPREPLY] = {
1427         .handler = icmp_discard,
1428     },
1429     [ICMP_INFO_REQUEST] = {
1430         .handler = icmp_discard,
1431     },
1432     [ICMP_INFO_REPLY] = {
1433         .handler = icmp_discard,
1434     },
1435     [ICMP_ADDRESS] = {
1436         .handler = icmp_discard,
1437     },
1438     [ICMP_ADDRESSREPLY] = {
1439         .handler = icmp_discard,
1440     },
1441 };
1442 
1443 static int __net_init icmp_sk_init(struct net *net)
1444 {
1445     /* Control parameters for ECHO replies. */
1446     net->ipv4.sysctl_icmp_echo_ignore_all = 0;
1447     net->ipv4.sysctl_icmp_echo_enable_probe = 0;
1448     net->ipv4.sysctl_icmp_echo_ignore_broadcasts = 1;
1449 
1450     /* Control parameter - ignore bogus broadcast responses? */
1451     net->ipv4.sysctl_icmp_ignore_bogus_error_responses = 1;
1452 
1453     /*
1454      *  Configurable global rate limit.
1455      *
1456      *  ratelimit defines tokens/packet consumed for dst->rate_token
1457      *  bucket ratemask defines which icmp types are ratelimited by
1458      *  setting it's bit position.
1459      *
1460      *  default:
1461      *  dest unreachable (3), source quench (4),
1462      *  time exceeded (11), parameter problem (12)
1463      */
1464 
1465     net->ipv4.sysctl_icmp_ratelimit = 1 * HZ;
1466     net->ipv4.sysctl_icmp_ratemask = 0x1818;
1467     net->ipv4.sysctl_icmp_errors_use_inbound_ifaddr = 0;
1468 
1469     return 0;
1470 }
1471 
1472 static struct pernet_operations __net_initdata icmp_sk_ops = {
1473        .init = icmp_sk_init,
1474 };
1475 
1476 int __init icmp_init(void)
1477 {
1478     int err, i;
1479 
1480     for_each_possible_cpu(i) {
1481         struct sock *sk;
1482 
1483         err = inet_ctl_sock_create(&sk, PF_INET,
1484                        SOCK_RAW, IPPROTO_ICMP, &init_net);
1485         if (err < 0)
1486             return err;
1487 
1488         per_cpu(ipv4_icmp_sk, i) = sk;
1489 
1490         /* Enough space for 2 64K ICMP packets, including
1491          * sk_buff/skb_shared_info struct overhead.
1492          */
1493         sk->sk_sndbuf = 2 * SKB_TRUESIZE(64 * 1024);
1494 
1495         /*
1496          * Speedup sock_wfree()
1497          */
1498         sock_set_flag(sk, SOCK_USE_WRITE_QUEUE);
1499         inet_sk(sk)->pmtudisc = IP_PMTUDISC_DONT;
1500     }
1501     return register_pernet_subsys(&icmp_sk_ops);
1502 }
[ Source navigation ] [ Diff markup ] [ Identifier search ] [ general search ]

This page was automatically generated by the 2.1.0 LXR engine.

The LXR team
Valid CSS 2.1! Valid HTML 4.01!