bitkeeper revision 1.1159.1.161 (4149eda74Qj-LHfB3E2bfFqo4OxI5g)
authorcl349@freefall.cl.cam.ac.uk <cl349@freefall.cl.cam.ac.uk>
Thu, 16 Sep 2004 19:46:47 +0000 (19:46 +0000)
committercl349@freefall.cl.cam.ac.uk <cl349@freefall.cl.cam.ac.uk>
Thu, 16 Sep 2004 19:46:47 +0000 (19:46 +0000)
The icmp_filter for raw ip sockets can't handle the skbs we build.  It wants
to look at the ICMP header and expects it to be located at skb->data past the
IP header.  Our skbs have all the data past the IP header in fragment buffers.

Pull the icmphdr into skb->data before looking at the icmp header.

.rootkeys
linux-2.6.8.1-xen-sparse/net/ipv4/raw.c [new file with mode: 0644]

index 8725479d091f0adf78ab72f9dae8999f53ba0e63..045d64edb33f05a3455964fbbb6b0cd5f0555837 100644 (file)
--- a/.rootkeys
+++ b/.rootkeys
 40f56a0ddHCSs3501MY4hRf22tctOw linux-2.6.8.1-xen-sparse/mkbuildtree
 412f46c0LJuKAgSPGoC0Z1DEkLfuLA linux-2.6.8.1-xen-sparse/mm/memory.c
 410a94a4KT6I6X0LVc7djB39tRDp4g linux-2.6.8.1-xen-sparse/mm/page_alloc.c
+4149ec79wMpIHdvbntxqVGLRZZjPxw linux-2.6.8.1-xen-sparse/net/ipv4/raw.c
 413cb1e4zst25MDYjg63Y-NGC5_pLg netbsd-2.0-xen-sparse/Makefile
 413cb1e5c_Mkxf_X0zimEhTKI_l4DA netbsd-2.0-xen-sparse/mkbuildtree
 413cb1e5kY_Zil7-b0kI6hvCIxBEYg netbsd-2.0-xen-sparse/nbconfig-xen
diff --git a/linux-2.6.8.1-xen-sparse/net/ipv4/raw.c b/linux-2.6.8.1-xen-sparse/net/ipv4/raw.c
new file mode 100644 (file)
index 0000000..8cc0ea6
--- /dev/null
@@ -0,0 +1,837 @@
+/*
+ * INET                An implementation of the TCP/IP protocol suite for the LINUX
+ *             operating system.  INET is implemented using the  BSD Socket
+ *             interface as the means of communication with the user level.
+ *
+ *             RAW - implementation of IP "raw" sockets.
+ *
+ * Version:    $Id: raw.c,v 1.64 2002/02/01 22:01:04 davem Exp $
+ *
+ * Authors:    Ross Biro, <bir7@leland.Stanford.Edu>
+ *             Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
+ *
+ * Fixes:
+ *             Alan Cox        :       verify_area() fixed up
+ *             Alan Cox        :       ICMP error handling
+ *             Alan Cox        :       EMSGSIZE if you send too big a packet
+ *             Alan Cox        :       Now uses generic datagrams and shared
+ *                                     skbuff library. No more peek crashes,
+ *                                     no more backlogs
+ *             Alan Cox        :       Checks sk->broadcast.
+ *             Alan Cox        :       Uses skb_free_datagram/skb_copy_datagram
+ *             Alan Cox        :       Raw passes ip options too
+ *             Alan Cox        :       Setsocketopt added
+ *             Alan Cox        :       Fixed error return for broadcasts
+ *             Alan Cox        :       Removed wake_up calls
+ *             Alan Cox        :       Use ttl/tos
+ *             Alan Cox        :       Cleaned up old debugging
+ *             Alan Cox        :       Use new kernel side addresses
+ *     Arnt Gulbrandsen        :       Fixed MSG_DONTROUTE in raw sockets.
+ *             Alan Cox        :       BSD style RAW socket demultiplexing.
+ *             Alan Cox        :       Beginnings of mrouted support.
+ *             Alan Cox        :       Added IP_HDRINCL option.
+ *             Alan Cox        :       Skip broadcast check if BSDism set.
+ *             David S. Miller :       New socket lookup architecture.
+ *
+ *             This program is free software; you can redistribute it and/or
+ *             modify it under the terms of the GNU General Public License
+ *             as published by the Free Software Foundation; either version
+ *             2 of the License, or (at your option) any later version.
+ */
+#include <linux/config.h> 
+#include <asm/atomic.h>
+#include <asm/byteorder.h>
+#include <asm/current.h>
+#include <asm/uaccess.h>
+#include <asm/ioctls.h>
+#include <linux/types.h>
+#include <linux/stddef.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <linux/aio.h>
+#include <linux/kernel.h>
+#include <linux/spinlock.h>
+#include <linux/sockios.h>
+#include <linux/socket.h>
+#include <linux/in.h>
+#include <linux/mroute.h>
+#include <linux/netdevice.h>
+#include <linux/in_route.h>
+#include <linux/route.h>
+#include <linux/tcp.h>
+#include <linux/skbuff.h>
+#include <net/dst.h>
+#include <net/sock.h>
+#include <linux/gfp.h>
+#include <linux/ip.h>
+#include <linux/net.h>
+#include <net/ip.h>
+#include <net/icmp.h>
+#include <net/udp.h>
+#include <net/raw.h>
+#include <net/snmp.h>
+#include <net/inet_common.h>
+#include <net/checksum.h>
+#include <net/xfrm.h>
+#include <linux/rtnetlink.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter_ipv4.h>
+
+struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE];
+rwlock_t raw_v4_lock = RW_LOCK_UNLOCKED;
+
+static void raw_v4_hash(struct sock *sk)
+{
+       struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num &
+                                                (RAWV4_HTABLE_SIZE - 1)];
+
+       write_lock_bh(&raw_v4_lock);
+       sk_add_node(sk, head);
+       sock_prot_inc_use(sk->sk_prot);
+       write_unlock_bh(&raw_v4_lock);
+}
+
+static void raw_v4_unhash(struct sock *sk)
+{
+       write_lock_bh(&raw_v4_lock);
+       if (sk_del_node_init(sk))
+               sock_prot_dec_use(sk->sk_prot);
+       write_unlock_bh(&raw_v4_lock);
+}
+
+struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num,
+                            unsigned long raddr, unsigned long laddr,
+                            int dif)
+{
+       struct hlist_node *node;
+
+       sk_for_each_from(sk, node) {
+               struct inet_opt *inet = inet_sk(sk);
+
+               if (inet->num == num                                    &&
+                   !(inet->daddr && inet->daddr != raddr)              &&
+                   !(inet->rcv_saddr && inet->rcv_saddr != laddr)      &&
+                   !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif))
+                       goto found; /* gotcha */
+       }
+       sk = NULL;
+found:
+       return sk;
+}
+
+/*
+ *     0 - deliver
+ *     1 - block
+ */
+static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb)
+{
+       int type;
+
+       if (!pskb_may_pull(skb, sizeof(struct icmphdr)))
+               return 1;
+
+       type = skb->h.icmph->type;
+       if (type < 32) {
+               __u32 data = raw4_sk(sk)->filter.data;
+
+               return ((1 << type) & data) != 0;
+       }
+
+       /* Do not block unknown ICMP types */
+       return 0;
+}
+
+/* IP input processing comes here for RAW socket delivery.
+ * Caller owns SKB, so we must make clones.
+ *
+ * RFC 1122: SHOULD pass TOS value up to the transport layer.
+ * -> It does. And not only TOS, but all IP header.
+ */
+void raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash)
+{
+       struct sock *sk;
+       struct hlist_head *head;
+
+       read_lock(&raw_v4_lock);
+       head = &raw_v4_htable[hash];
+       if (hlist_empty(head))
+               goto out;
+       sk = __raw_v4_lookup(__sk_head(head), iph->protocol,
+                            iph->saddr, iph->daddr,
+                            skb->dev->ifindex);
+
+       while (sk) {
+               if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) {
+                       struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC);
+
+                       /* Not releasing hash table! */
+                       if (clone)
+                               raw_rcv(sk, clone);
+               }
+               sk = __raw_v4_lookup(sk_next(sk), iph->protocol,
+                                    iph->saddr, iph->daddr,
+                                    skb->dev->ifindex);
+       }
+out:
+       read_unlock(&raw_v4_lock);
+}
+
+void raw_err (struct sock *sk, struct sk_buff *skb, u32 info)
+{
+       struct inet_opt *inet = inet_sk(sk);
+       int type = skb->h.icmph->type;
+       int code = skb->h.icmph->code;
+       int err = 0;
+       int harderr = 0;
+
+       /* Report error on raw socket, if:
+          1. User requested ip_recverr.
+          2. Socket is connected (otherwise the error indication
+             is useless without ip_recverr and error is hard.
+        */
+       if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED)
+               return;
+
+       switch (type) {
+       default:
+       case ICMP_TIME_EXCEEDED:
+               err = EHOSTUNREACH;
+               break;
+       case ICMP_SOURCE_QUENCH:
+               return;
+       case ICMP_PARAMETERPROB:
+               err = EPROTO;
+               harderr = 1;
+               break;
+       case ICMP_DEST_UNREACH:
+               err = EHOSTUNREACH;
+               if (code > NR_ICMP_UNREACH)
+                       break;
+               err = icmp_err_convert[code].errno;
+               harderr = icmp_err_convert[code].fatal;
+               if (code == ICMP_FRAG_NEEDED) {
+                       harderr = inet->pmtudisc != IP_PMTUDISC_DONT;
+                       err = EMSGSIZE;
+               }
+       }
+
+       if (inet->recverr) {
+               struct iphdr *iph = (struct iphdr*)skb->data;
+               u8 *payload = skb->data + (iph->ihl << 2);
+
+               if (inet->hdrincl)
+                       payload = skb->data;
+               ip_icmp_error(sk, skb, err, 0, info, payload);
+       }
+
+       if (inet->recverr || harderr) {
+               sk->sk_err = err;
+               sk->sk_error_report(sk);
+       }
+}
+
+static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb)
+{
+       /* Charge it to the socket. */
+       
+       if (sock_queue_rcv_skb(sk, skb) < 0) {
+               /* FIXME: increment a raw drops counter here */
+               kfree_skb(skb);
+               return NET_RX_DROP;
+       }
+
+       return NET_RX_SUCCESS;
+}
+
+int raw_rcv(struct sock *sk, struct sk_buff *skb)
+{
+       if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) {
+               kfree_skb(skb);
+               return NET_RX_DROP;
+       }
+
+       skb_push(skb, skb->data - skb->nh.raw);
+
+       raw_rcv_skb(sk, skb);
+       return 0;
+}
+
+static int raw_send_hdrinc(struct sock *sk, void *from, int length,
+                       struct rtable *rt, 
+                       unsigned int flags)
+{
+       struct inet_opt *inet = inet_sk(sk);
+       int hh_len;
+       struct iphdr *iph;
+       struct sk_buff *skb;
+       int err;
+
+       if (length > rt->u.dst.dev->mtu) {
+               ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport,
+                              rt->u.dst.dev->mtu);
+               return -EMSGSIZE;
+       }
+       if (flags&MSG_PROBE)
+               goto out;
+
+       hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
+
+       skb = sock_alloc_send_skb(sk, length+hh_len+15,
+                                 flags&MSG_DONTWAIT, &err);
+       if (skb == NULL)
+               goto error; 
+       skb_reserve(skb, hh_len);
+
+       skb->priority = sk->sk_priority;
+       skb->dst = dst_clone(&rt->u.dst);
+
+       skb->nh.iph = iph = (struct iphdr *)skb_put(skb, length);
+
+       skb->ip_summed = CHECKSUM_NONE;
+
+       skb->h.raw = skb->nh.raw;
+       err = memcpy_fromiovecend((void *)iph, from, 0, length);
+       if (err)
+               goto error_fault;
+
+       /* We don't modify invalid header */
+       if (length >= sizeof(*iph) && iph->ihl * 4 <= length) {
+               if (!iph->saddr)
+                       iph->saddr = rt->rt_src;
+               iph->check   = 0;
+               iph->tot_len = htons(length);
+               if (!iph->id)
+                       ip_select_ident(iph, &rt->u.dst, NULL);
+
+               iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
+       }
+
+       err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev,
+                     dst_output);
+       if (err > 0)
+               err = inet->recverr ? net_xmit_errno(err) : 0;
+       if (err)
+               goto error;
+out:
+       return 0;
+
+error_fault:
+       err = -EFAULT;
+       kfree_skb(skb);
+error:
+       IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS);
+       return err; 
+}
+
+static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
+                      size_t len)
+{
+       struct inet_opt *inet = inet_sk(sk);
+       struct ipcm_cookie ipc;
+       struct rtable *rt = NULL;
+       int free = 0;
+       u32 daddr;
+       u32 saddr;
+       u8  tos;
+       int err;
+
+       err = -EMSGSIZE;
+       if (len < 0 || len > 0xFFFF)
+               goto out;
+
+       /*
+        *      Check the flags.
+        */
+
+       err = -EOPNOTSUPP;
+       if (msg->msg_flags & MSG_OOB)   /* Mirror BSD error message */
+               goto out;               /* compatibility */
+                        
+       /*
+        *      Get and verify the address. 
+        */
+
+       if (msg->msg_namelen) {
+               struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name;
+               err = -EINVAL;
+               if (msg->msg_namelen < sizeof(*usin))
+                       goto out;
+               if (usin->sin_family != AF_INET) {
+                       static int complained;
+                       if (!complained++)
+                               printk(KERN_INFO "%s forgot to set AF_INET in "
+                                                "raw sendmsg. Fix it!\n",
+                                                current->comm);
+                       err = -EINVAL;
+                       if (usin->sin_family)
+                               goto out;
+               }
+               daddr = usin->sin_addr.s_addr;
+               /* ANK: I did not forget to get protocol from port field.
+                * I just do not know, who uses this weirdness.
+                * IP_HDRINCL is much more convenient.
+                */
+       } else {
+               err = -EDESTADDRREQ;
+               if (sk->sk_state != TCP_ESTABLISHED) 
+                       goto out;
+               daddr = inet->daddr;
+       }
+
+       ipc.addr = inet->saddr;
+       ipc.opt = NULL;
+       ipc.oif = sk->sk_bound_dev_if;
+
+       if (msg->msg_controllen) {
+               err = ip_cmsg_send(msg, &ipc);
+               if (err)
+                       goto out;
+               if (ipc.opt)
+                       free = 1;
+       }
+
+       saddr = ipc.addr;
+       ipc.addr = daddr;
+
+       if (!ipc.opt)
+               ipc.opt = inet->opt;
+
+       if (ipc.opt) {
+               err = -EINVAL;
+               /* Linux does not mangle headers on raw sockets,
+                * so that IP options + IP_HDRINCL is non-sense.
+                */
+               if (inet->hdrincl)
+                       goto done;
+               if (ipc.opt->srr) {
+                       if (!daddr)
+                               goto done;
+                       daddr = ipc.opt->faddr;
+               }
+       }
+       tos = RT_TOS(inet->tos) | sk->sk_localroute;
+       if (msg->msg_flags & MSG_DONTROUTE)
+               tos |= RTO_ONLINK;
+
+       if (MULTICAST(daddr)) {
+               if (!ipc.oif)
+                       ipc.oif = inet->mc_index;
+               if (!saddr)
+                       saddr = inet->mc_addr;
+       }
+
+       {
+               struct flowi fl = { .oif = ipc.oif,
+                                   .nl_u = { .ip4_u =
+                                             { .daddr = daddr,
+                                               .saddr = saddr,
+                                               .tos = tos } },
+                                   .proto = inet->hdrincl ? IPPROTO_RAW :
+                                                            sk->sk_protocol,
+                                 };
+               err = ip_route_output_flow(&rt, &fl, sk, !(msg->msg_flags&MSG_DONTWAIT));
+       }
+       if (err)
+               goto done;
+
+       err = -EACCES;
+       if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST))
+               goto done;
+
+       if (msg->msg_flags & MSG_CONFIRM)
+               goto do_confirm;
+back_from_confirm:
+
+       if (inet->hdrincl)
+               err = raw_send_hdrinc(sk, msg->msg_iov, len, 
+                                       rt, msg->msg_flags);
+       
+        else {
+               if (!ipc.addr)
+                       ipc.addr = rt->rt_dst;
+               lock_sock(sk);
+               err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0,
+                                       &ipc, rt, msg->msg_flags);
+               if (err)
+                       ip_flush_pending_frames(sk);
+               else if (!(msg->msg_flags & MSG_MORE))
+                       err = ip_push_pending_frames(sk);
+               release_sock(sk);
+       }
+done:
+       if (free)
+               kfree(ipc.opt);
+       ip_rt_put(rt);
+
+out:   return err < 0 ? err : len;
+
+do_confirm:
+       dst_confirm(&rt->u.dst);
+       if (!(msg->msg_flags & MSG_PROBE) || len)
+               goto back_from_confirm;
+       err = 0;
+       goto done;
+}
+
+static void raw_close(struct sock *sk, long timeout)
+{
+        /*
+        * Raw sockets may have direct kernel refereneces. Kill them.
+        */
+       ip_ra_control(sk, 0, NULL);
+
+       sk_common_release(sk);
+}
+
+/* This gets rid of all the nasties in af_inet. -DaveM */
+static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len)
+{
+       struct inet_opt *inet = inet_sk(sk);
+       struct sockaddr_in *addr = (struct sockaddr_in *) uaddr;
+       int ret = -EINVAL;
+       int chk_addr_ret;
+
+       if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in))
+               goto out;
+       chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr);
+       ret = -EADDRNOTAVAIL;
+       if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL &&
+           chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST)
+               goto out;
+       inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr;
+       if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)
+               inet->saddr = 0;  /* Use device */
+       sk_dst_reset(sk);
+       ret = 0;
+out:   return ret;
+}
+
+/*
+ *     This should be easy, if there is something there
+ *     we return it, otherwise we block.
+ */
+
+int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg,
+               size_t len, int noblock, int flags, int *addr_len)
+{
+       struct inet_opt *inet = inet_sk(sk);
+       size_t copied = 0;
+       int err = -EOPNOTSUPP;
+       struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name;
+       struct sk_buff *skb;
+
+       if (flags & MSG_OOB)
+               goto out;
+
+       if (addr_len)
+               *addr_len = sizeof(*sin);
+
+       if (flags & MSG_ERRQUEUE) {
+               err = ip_recv_error(sk, msg, len);
+               goto out;
+       }
+
+       skb = skb_recv_datagram(sk, flags, noblock, &err);
+       if (!skb)
+               goto out;
+
+       copied = skb->len;
+       if (len < copied) {
+               msg->msg_flags |= MSG_TRUNC;
+               copied = len;
+       }
+
+       err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
+       if (err)
+               goto done;
+
+       sock_recv_timestamp(msg, sk, skb);
+
+       /* Copy the address. */
+       if (sin) {
+               sin->sin_family = AF_INET;
+               sin->sin_addr.s_addr = skb->nh.iph->saddr;
+               memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
+       }
+       if (inet->cmsg_flags)
+               ip_cmsg_recv(msg, skb);
+       if (flags & MSG_TRUNC)
+               copied = skb->len;
+done:
+       skb_free_datagram(sk, skb);
+out:   return err ? err : copied;
+}
+
+static int raw_init(struct sock *sk)
+{
+       struct raw_opt *tp = raw4_sk(sk);
+       if (inet_sk(sk)->num == IPPROTO_ICMP)
+               memset(&tp->filter, 0, sizeof(tp->filter));
+       return 0;
+}
+
+static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen)
+{
+       if (optlen > sizeof(struct icmp_filter))
+               optlen = sizeof(struct icmp_filter);
+       if (copy_from_user(&raw4_sk(sk)->filter, optval, optlen))
+               return -EFAULT;
+       return 0;
+}
+
+static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen)
+{
+       int len, ret = -EFAULT;
+
+       if (get_user(len, optlen))
+               goto out;
+       ret = -EINVAL;
+       if (len < 0)
+               goto out;
+       if (len > sizeof(struct icmp_filter))
+               len = sizeof(struct icmp_filter);
+       ret = -EFAULT;
+       if (put_user(len, optlen) ||
+           copy_to_user(optval, &raw4_sk(sk)->filter, len))
+               goto out;
+       ret = 0;
+out:   return ret;
+}
+
+static int raw_setsockopt(struct sock *sk, int level, int optname, 
+                         char __user *optval, int optlen)
+{
+       if (level != SOL_RAW)
+               return ip_setsockopt(sk, level, optname, optval, optlen);
+
+       if (optname == ICMP_FILTER) {
+               if (inet_sk(sk)->num != IPPROTO_ICMP)
+                       return -EOPNOTSUPP;
+               else
+                       return raw_seticmpfilter(sk, optval, optlen);
+       }
+       return -ENOPROTOOPT;
+}
+
+static int raw_getsockopt(struct sock *sk, int level, int optname, 
+                         char __user *optval, int __user *optlen)
+{
+       if (level != SOL_RAW)
+               return ip_getsockopt(sk, level, optname, optval, optlen);
+
+       if (optname == ICMP_FILTER) {
+               if (inet_sk(sk)->num != IPPROTO_ICMP)
+                       return -EOPNOTSUPP;
+               else
+                       return raw_geticmpfilter(sk, optval, optlen);
+       }
+       return -ENOPROTOOPT;
+}
+
+static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg)
+{
+       switch (cmd) {
+               case SIOCOUTQ: {
+                       int amount = atomic_read(&sk->sk_wmem_alloc);
+                       return put_user(amount, (int __user *)arg);
+               }
+               case SIOCINQ: {
+                       struct sk_buff *skb;
+                       int amount = 0;
+
+                       spin_lock_irq(&sk->sk_receive_queue.lock);
+                       skb = skb_peek(&sk->sk_receive_queue);
+                       if (skb != NULL)
+                               amount = skb->len;
+                       spin_unlock_irq(&sk->sk_receive_queue.lock);
+                       return put_user(amount, (int __user *)arg);
+               }
+
+               default:
+#ifdef CONFIG_IP_MROUTE
+                       return ipmr_ioctl(sk, cmd, (void __user *)arg);
+#else
+                       return -ENOIOCTLCMD;
+#endif
+       }
+}
+
+struct proto raw_prot = {
+       .name =         "RAW",
+       .close =        raw_close,
+       .connect =      ip4_datagram_connect,
+       .disconnect =   udp_disconnect,
+       .ioctl =        raw_ioctl,
+       .init =         raw_init,
+       .setsockopt =   raw_setsockopt,
+       .getsockopt =   raw_getsockopt,
+       .sendmsg =      raw_sendmsg,
+       .recvmsg =      raw_recvmsg,
+       .bind =         raw_bind,
+       .backlog_rcv =  raw_rcv_skb,
+       .hash =         raw_v4_hash,
+       .unhash =       raw_v4_unhash,
+};
+
+#ifdef CONFIG_PROC_FS
+struct raw_iter_state {
+       int bucket;
+};
+
+#define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private)
+
+static struct sock *raw_get_first(struct seq_file *seq)
+{
+       struct sock *sk;
+       struct raw_iter_state* state = raw_seq_private(seq);
+
+       for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) {
+               struct hlist_node *node;
+
+               sk_for_each(sk, node, &raw_v4_htable[state->bucket])
+                       if (sk->sk_family == PF_INET)
+                               goto found;
+       }
+       sk = NULL;
+found:
+       return sk;
+}
+
+static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk)
+{
+       struct raw_iter_state* state = raw_seq_private(seq);
+
+       do {
+               sk = sk_next(sk);
+try_again:
+               ;
+       } while (sk && sk->sk_family != PF_INET);
+
+       if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) {
+               sk = sk_head(&raw_v4_htable[state->bucket]);
+               goto try_again;
+       }
+       return sk;
+}
+
+static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos)
+{
+       struct sock *sk = raw_get_first(seq);
+
+       if (sk)
+               while (pos && (sk = raw_get_next(seq, sk)) != NULL)
+                       --pos;
+       return pos ? NULL : sk;
+}
+
+static void *raw_seq_start(struct seq_file *seq, loff_t *pos)
+{
+       read_lock(&raw_v4_lock);
+       return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN;
+}
+
+static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+       struct sock *sk;
+
+       if (v == SEQ_START_TOKEN)
+               sk = raw_get_first(seq);
+       else
+               sk = raw_get_next(seq, v);
+       ++*pos;
+       return sk;
+}
+
+static void raw_seq_stop(struct seq_file *seq, void *v)
+{
+       read_unlock(&raw_v4_lock);
+}
+
+static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i)
+{
+       struct inet_opt *inet = inet_sk(sp);
+       unsigned int dest = inet->daddr,
+                    src = inet->rcv_saddr;
+       __u16 destp = 0,
+             srcp  = inet->num;
+
+       sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X"
+               " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p",
+               i, src, srcp, dest, destp, sp->sk_state, 
+               atomic_read(&sp->sk_wmem_alloc),
+               atomic_read(&sp->sk_rmem_alloc),
+               0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp),
+               atomic_read(&sp->sk_refcnt), sp);
+       return tmpbuf;
+}
+
+static int raw_seq_show(struct seq_file *seq, void *v)
+{
+       char tmpbuf[129];
+
+       if (v == SEQ_START_TOKEN)
+               seq_printf(seq, "%-127s\n",
+                              "  sl  local_address rem_address   st tx_queue "
+                              "rx_queue tr tm->when retrnsmt   uid  timeout "
+                              "inode");
+       else {
+               struct raw_iter_state *state = raw_seq_private(seq);
+
+               seq_printf(seq, "%-127s\n",
+                          get_raw_sock(v, tmpbuf, state->bucket));
+       }
+       return 0;
+}
+
+static struct seq_operations raw_seq_ops = {
+       .start = raw_seq_start,
+       .next  = raw_seq_next,
+       .stop  = raw_seq_stop,
+       .show  = raw_seq_show,
+};
+
+static int raw_seq_open(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq;
+       int rc = -ENOMEM;
+       struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL);
+
+       if (!s)
+               goto out;
+       rc = seq_open(file, &raw_seq_ops);
+       if (rc)
+               goto out_kfree;
+
+       seq = file->private_data;
+       seq->private = s;
+       memset(s, 0, sizeof(*s));
+out:
+       return rc;
+out_kfree:
+       kfree(s);
+       goto out;
+}
+
+static struct file_operations raw_seq_fops = {
+       .owner   = THIS_MODULE,
+       .open    = raw_seq_open,
+       .read    = seq_read,
+       .llseek  = seq_lseek,
+       .release = seq_release_private,
+};
+
+int __init raw_proc_init(void)
+{
+       if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops))
+               return -ENOMEM;
+       return 0;
+}
+
+void __init raw_proc_exit(void)
+{
+       proc_net_remove("raw");
+}
+#endif /* CONFIG_PROC_FS */