| #include <linux/version.h> |
| #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) |
| # define __irq_h 1 /* stop it being included! */ |
| #else |
| # define _LINUX_TCP_H |
| #endif |
| #include <net/ip.h> |
| |
| #include "ipf-linux.h" |
| |
| #include <net/checksum.h> |
| #include <net/route.h> |
| |
| #include <linux/random.h> |
| #include <asm/ioctls.h> |
| |
| extern int sysctl_ip_default_ttl; |
| |
| static int frzerostats __P((caddr_t)); |
| static int ipf_send_ip __P((fr_info_t *, struct sk_buff *, struct sk_buff **)); |
| |
| ipfmutex_t ipl_mutex, ipf_auth_mx, ipf_rw, ipf_stinsert; |
| ipfmutex_t ipf_nat_new, ipf_natio, ipf_timeoutlock; |
| ipfrwlock_t ipf_mutex, ipf_global, ipf_ipidfrag, ipf_frcache; |
| ipfrwlock_t ipf_frag, ipf_state, ipf_nat, ipf_natfrag, ipf_authlk; |
| |
| static u_int ipf_linux_inout __P((u_int, struct sk_buff **, const struct net_device *, const struct net_device *, int (*okfn)(struct sk_buff *))); |
| |
| #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) |
| static struct nf_hook_ops ipf_hooks[] = { |
| { |
| .hook = ipf_linux_inout, |
| .owner = THIS_MODULE, |
| .pf = PF_INET, |
| .hooknum = NF_IP_PRE_ROUTING, |
| .priority = 200, |
| }, |
| { |
| .hook = ipf_linux_inout, |
| .owner = THIS_MODULE, |
| .pf = PF_INET, |
| .hooknum = NF_IP_POST_ROUTING, |
| .priority = 200, |
| }, |
| # ifdef USE_INET6 |
| { |
| .hook = ipf_linux_inout, |
| .owner = THIS_MODULE, |
| .pf = PF_INET6, |
| .hooknum = NF_IP_PRE_ROUTING, |
| .priority = 200, |
| }, |
| { |
| .hook = ipf_linux_inout, |
| .owner = THIS_MODULE, |
| .pf = PF_INET6, |
| .hooknum = NF_IP_POST_ROUTING, |
| .priority = 200, |
| } |
| # endif |
| }; |
| #else |
| static struct nf_hook_ops ipf_hooks[] = { |
| { |
| { NULL, NULL }, /* list */ |
| ipf_linux_inout, /* hook */ |
| PF_INET, /* pf */ |
| NF_IP_PRE_ROUTING, /* hooknum */ |
| 200 /* priority */ |
| }, |
| { |
| { NULL, NULL}, /* list */ |
| ipf_linux_inout, /* hook */ |
| PF_INET, /* pf */ |
| NF_IP_POST_ROUTING, /* hooknum */ |
| 200 /* priority */ |
| }, |
| # ifdef USE_INET6 |
| { |
| { NULL, NULL }, /* list */ |
| ipf_linux_inout, /* hook */ |
| PF_INET6, /* pf */ |
| NF_IP_PRE_ROUTING, /* hooknum */ |
| 200 /* priority */ |
| }, |
| { |
| { NULL, NULL}, /* list */ |
| ipf_linux_inout, /* hook */ |
| PF_INET6, /* pf */ |
| NF_IP_POST_ROUTING, /* hooknum */ |
| 200 /* priority */ |
| } |
| # endif |
| }; |
| #endif |
| |
| |
| /* |
| * Filter ioctl interface. |
| */ |
| int |
| ipf_ioctl(struct inode *in, struct file *fp, u_int cmd, u_long arg) |
| { |
| int error = 0, unit = 0; |
| caddr_t data; |
| mode_t mode; |
| |
| unit = MINOR(in->i_rdev); |
| if (unit < 0 || unit > IPL_LOGMAX) |
| return -ENXIO; |
| |
| if (ipf_running <= 0) { |
| if (unit != IPL_LOGIPF) |
| return -EIO; |
| if (cmd != SIOCIPFGETNEXT && cmd != SIOCIPFGET && |
| cmd != SIOCIPFSET && cmd != SIOCFRENB && |
| cmd != SIOCGETFS && cmd != SIOCGETFF) |
| return -EIO; |
| } |
| |
| mode = fp->f_mode; |
| data = (caddr_t)arg; |
| |
| error = ipf_ioctlswitch(unit, data, cmd, mode, fp->f_uid, fp); |
| if (error != -1) { |
| SPL_X(s); |
| if (error > 0) |
| error = -error; |
| return error; |
| } |
| SPL_X(s); |
| |
| if (error > 0) |
| error = -error; |
| return error; |
| } |
| |
| |
| u_32_t |
| ipf_newisn(fr_info_t *fin) |
| { |
| u_32_t isn; |
| |
| isn = secure_tcp_sequence_number(fin->fin_daddr, fin->fin_saddr, |
| fin->fin_dport, fin->fin_sport); |
| return isn; |
| } |
| |
| |
| int |
| ipf_send_reset(fr_info_t *fin) |
| { |
| tcphdr_t *tcp, *tcp2; |
| int tlen, hlen; |
| #ifdef USE_INET6 |
| ip6_t *ip6; |
| #endif |
| ip_t *ip; |
| mb_t *m; |
| |
| tcp = fin->fin_dp; |
| if (tcp->th_flags & TH_RST) |
| return -1; |
| |
| if (ipf_checkl4sum(fin) == -1) |
| return -1; |
| |
| m = skb_copy(fin->fin_m, GFP_ATOMIC); |
| if (m == NULL) |
| return -1; |
| |
| tlen = (tcp->th_flags & (TH_SYN|TH_FIN)) ? 1 : 0; |
| #ifdef USE_INET6 |
| if (fin->fin_v == 6) |
| hlen = sizeof(ip6_t); |
| else |
| #endif |
| hlen = sizeof(ip_t); |
| hlen += sizeof(*tcp2); |
| skb_trim(m, hlen); |
| |
| bzero(MTOD(m, char *), hlen); |
| ip = MTOD(m, ip_t *); |
| bzero((char *)ip, hlen); |
| ip->ip_v = fin->fin_v; |
| tcp2 = (tcphdr_t *)((char *)ip + hlen - sizeof(*tcp2)); |
| tcp2->th_dport = tcp->th_sport; |
| tcp2->th_sport = tcp->th_dport; |
| if (tcp->th_flags & TH_ACK) { |
| tcp2->th_seq = tcp->th_ack; |
| tcp2->th_flags = TH_RST; |
| } else { |
| tcp2->th_ack = ntohl(tcp->th_seq); |
| tcp2->th_ack += tlen; |
| tcp2->th_ack = htonl(tcp2->th_ack); |
| tcp2->th_flags = TH_RST|TH_ACK; |
| } |
| tcp2->th_off = sizeof(*tcp2) >> 2; |
| |
| #ifdef USE_INET6 |
| if (fin->fin_v == 6) { |
| ip6 = (ip6_t *)ip; |
| ip6->ip6_src = fin->fin_dst6; |
| ip6->ip6_dst = fin->fin_src6; |
| ip6->ip6_plen = htons(sizeof(*tcp)); |
| ip6->ip6_nxt = IPPROTO_TCP; |
| } else |
| #endif |
| { |
| ip->ip_hl = sizeof(*ip) >> 2; |
| ip->ip_src.s_addr = fin->fin_daddr; |
| ip->ip_dst.s_addr = fin->fin_saddr; |
| ip->ip_p = IPPROTO_TCP; |
| ip->ip_len = htons(sizeof(*ip) + sizeof(*tcp)); |
| tcp2->th_sum = ipf_cksum(m, ip, IPPROTO_TCP, tcp2, |
| ntohs(ip->ip_len)); |
| } |
| return ipf_send_ip(fin, m, &m); |
| } |
| |
| |
| /* |
| * On input, ip_len is in network byte order |
| */ |
| static int |
| ipf_send_ip(fr_info_t *fin, struct sk_buff *sk, struct sk_buff **skp) |
| { |
| fr_info_t fnew; |
| ip_t *ip, *oip; |
| int hlen; |
| |
| ip = MTOD(sk, ip_t *); |
| bzero((char *)&fnew, sizeof(fnew)); |
| oip = fin->fin_ip; |
| |
| switch (fin->fin_v) |
| { |
| case 4 : |
| fnew.fin_v = 4; |
| ip->ip_hl = sizeof(*oip) >> 2; |
| ip->ip_tos = oip->ip_tos; |
| ip->ip_id = 0; |
| ip->ip_ttl = sysctl_ip_default_ttl; |
| ip->ip_sum = 0; |
| ip->ip_off = 0x4000; /* IP_DF */ |
| hlen = sizeof(*ip); |
| break; |
| default : |
| return EINVAL; |
| } |
| |
| fnew.fin_ifp = fin->fin_ifp; |
| fnew.fin_flx = FI_NOCKSUM; |
| fnew.fin_m = sk; |
| fnew.fin_ip = ip; |
| fnew.fin_mp = skp; |
| fnew.fin_hlen = hlen; |
| fnew.fin_dp = (char *)ip + hlen; |
| (void) ipf_makefrip(hlen, ip, &fnew); |
| |
| return ipf_fastroute(sk, skp, &fnew, NULL); |
| } |
| |
| |
| int |
| ipf_send_icmp_err(int type, fr_info_t *fin, int isdst) |
| { |
| int hlen, code, leader, dlen; |
| struct net_device *ifp; |
| struct in_addr dst4; |
| struct icmp *icmp; |
| i6addr_t dst6; |
| u_short sz; |
| #ifdef USE_INET6 |
| ip6_t *ip6; |
| mb_t *mb; |
| #endif |
| ip_t *ip; |
| mb_t *m, *m0; |
| |
| if ((type < 0) || (type > ICMP_MAXTYPE)) |
| return -1; |
| |
| code = fin->fin_icode; |
| |
| #ifdef USE_INET6 |
| if ((code < 0) || (code > sizeof(icmptoicmp6unreach)/sizeof(int))) |
| return -1; |
| #endif |
| |
| if (ipf_checkl4sum(fin) == -1) |
| return -1; |
| |
| m0 = fin->fin_m; |
| #ifdef USE_INET6 |
| if (fin->fin_v == 6) { |
| sz = sizeof(ip6_t); |
| dlen = MIN(M_LEN(m0), 512); |
| sz += dlen; |
| hlen = sizeof(ip6_t); |
| type = icmptoicmp6types[type]; |
| if (type == ICMP6_DST_UNREACH) |
| code = icmptoicmp6unreach[code]; |
| } else |
| #endif |
| { |
| if ((fin->fin_p == IPPROTO_ICMP) && !(fin->fin_flx & FI_SHORT)) |
| switch (ntohs(fin->fin_data[0]) >> 8) |
| { |
| case ICMP_ECHO : |
| case ICMP_TSTAMP : |
| case ICMP_IREQ : |
| case ICMP_MASKREQ : |
| break; |
| default : |
| return 0; |
| } |
| |
| sz = sizeof(ip_t) * 2 + 4; |
| dlen = 8; /* 64 bits of data */ |
| sz += dlen; |
| hlen = sizeof(ip_t); |
| } |
| |
| leader = m0->data - m0->head; |
| if ((leader & 15) != 0) |
| leader += 16 - (leader & 15); |
| m = alloc_skb(sz + leader, GFP_ATOMIC); |
| if (m == NULL) |
| return -1; |
| |
| /* Set the data pointer */ |
| skb_reserve(m, leader); |
| |
| bzero(MTOD(m, char *), (size_t)sz); |
| |
| m->nh.iph = (struct iphdr *)skb_put(m, hlen); |
| ip = (ip_t *)m->nh.iph; |
| ip->ip_v = fin->fin_v; |
| |
| m->h.icmph = (struct icmphdr *)skb_put(m, hlen + 4 + dlen); |
| icmp = (icmphdr_t *)m->h.icmph; |
| icmp->icmp_type = type & 0xff; |
| icmp->icmp_code = code & 0xff; |
| #ifdef icmp_nextmtu |
| ifp = fin->fin_ifp; |
| if (type == ICMP_UNREACH && fin->fin_icode == ICMP_UNREACH_NEEDFRAG) { |
| if (fin->fin_mtu != 0) { |
| icmp->icmp_nextmtu = htons(fin->fin_mtu); |
| |
| } else if (ifp->mtu != 0) { |
| icmp->icmp_nextmtu = htons(ifp->mtu); |
| |
| } else { |
| icmp->icmp_nextmtu = htons(fin->fin_plen - 20); |
| } |
| } |
| #endif |
| |
| #ifdef USE_INET6 |
| if (fin->fin_v == 6) { |
| int csz; |
| |
| if (isdst == 0) { |
| if (ipf_ifpaddr(6, FRI_NORMAL, qif->qf_ill, |
| &dst6, NULL) == -1) { |
| FREE_MB_T(m); |
| return -1; |
| } |
| } else |
| dst6 = fin->fin_dst6; |
| |
| csz = sz; |
| sz -= sizeof(ip6_t); |
| ip6 = (ip6_t *)ip; |
| ip6->ip6_plen = htons((u_short)sz); |
| ip6->ip6_nxt = IPPROTO_ICMPV6; |
| ip6->ip6_src = dst6; |
| ip6->ip6_dst = fin->fin_src6; |
| sz -= offsetof(struct icmp, icmp_ip); |
| bcopy((char *)mb->b_rptr, (char *)&icmp->icmp_ip, sz); |
| icmp->icmp_cksum = csz - sizeof(ip6_t); |
| } else |
| #endif |
| { |
| ip = MTOD(m, ip_t *); |
| ip->ip_hl = sizeof(*ip) >> 2; |
| ip->ip_p = IPPROTO_ICMP; |
| ip->ip_len = (u_short)sz; |
| if (isdst == 0) { |
| if (ipf_ifpaddr(4, FRI_NORMAL, fin->fin_ifp, |
| &dst6, NULL) == -1) { |
| FREE_MB_T(m); |
| return -1; |
| } |
| dst4 = dst6.in4; |
| } else |
| dst4 = fin->fin_dst; |
| ip->ip_src = dst4; |
| ip->ip_dst = fin->fin_src; |
| bcopy((char *)fin->fin_ip, (char *)&icmp->icmp_ip, |
| sizeof(*fin->fin_ip)); |
| bcopy((char *)fin->fin_ip + fin->fin_hlen, |
| (char *)&icmp->icmp_ip + sizeof(*fin->fin_ip), 8); |
| icmp->icmp_cksum = ip_compute_csum((u_char *)icmp, |
| sz - sizeof(ip_t)); |
| ip->ip_len = htons(sz); |
| } |
| |
| /* |
| * Need to exit out of these so we don't recursively call rw_enter |
| * from fr_qout. |
| */ |
| return ipf_send_ip(fin, m, &m); |
| } |
| |
| |
| int |
| ipf_verifysrc(fr_info_t *fin) |
| { |
| return 0; |
| } |
| |
| |
| void |
| ipf_checkv4sum(fr_info_t *fin) |
| { |
| /* |
| * Linux 2.4.20-8smp (RedHat 9) |
| * Because ip_input() on linux clears the checksum flag in the sk_buff |
| * before calling the netfilter hook, it is not possible to take |
| * advantage of the work already done by the hardware. |
| */ |
| #ifdef IPFILTER_CKSUM |
| if (ipf_checkl4sum(fin) == -1) |
| fin->fin_flx |= FI_BAD; |
| #endif |
| } |
| |
| |
| u_short |
| ipf_nextipid(fr_info_t *fin) |
| { |
| #if 1 |
| static u_short ipid = 0; |
| |
| return ipid++; |
| #else |
| ip_t ip; |
| |
| __ip_select_ident(&ip, NULL); |
| return ip.ip_id; |
| #endif |
| } |
| |
| |
| /* |
| * min - pointer to mbuf where the IP packet starts |
| * mpp - pointer to the mbuf pointer that is the start of the mbuf chain |
| */ |
| /*ARGSUSED*/ |
| int |
| ipf_fastroute(mb_t *min, mb_t **mp, fr_info_t *fin, frdest_t *fdp) |
| { |
| struct net_device *ifp, *sifp; |
| struct in_addr dip; |
| struct rtable *rt; |
| frentry_t *fr; |
| int err, sout; |
| ip_t *ip; |
| |
| rt = NULL; |
| fr = fin->fin_fr; |
| ip = MTOD(min, ip_t *); |
| dip = ip->ip_dst; |
| |
| if (fdp != NULL) |
| ifp = fdp->fd_ifp; |
| else |
| ifp = fin->fin_ifp; |
| |
| if ((ifp == NULL) && ((fr == NULL) || !(fr->fr_flags & FR_FASTROUTE))) { |
| err = ENETUNREACH; |
| goto bad; |
| } |
| |
| if ((fdp != NULL) && (fdp->fd_ip.s_addr)) |
| dip = fdp->fd_ip; |
| |
| switch (fin->fin_v) |
| { |
| case 4 : |
| #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,0) |
| err = ip_route_output(&rt, dip.s_addr, 0, |
| RT_TOS(ip->ip_tos) | RTO_CONN, 0); |
| #else |
| err = 1; |
| #endif |
| if (err != 0 || rt == NULL) |
| goto bad; |
| |
| if (rt->u.dst.dev == NULL) { |
| err = EHOSTUNREACH; |
| goto bad; |
| } |
| break; |
| default : |
| err = EINVAL; |
| goto bad; |
| } |
| |
| if (fin->fin_out == 0) { |
| sifp = fin->fin_ifp; |
| sout = fin->fin_out; |
| fin->fin_ifp = ifp; |
| fin->fin_out = 1; |
| (void) ipf_acctpkt(fin, NULL); |
| fin->fin_fr = NULL; |
| if (!fr || !(fr->fr_flags & FR_RETMASK)) { |
| u_32_t pass; |
| |
| if (ipf_state_check(fin, &pass) != NULL) |
| ipf_state_deref((ipstate_t **)&fin->fin_state); |
| } |
| |
| switch (ipf_nat_checkout(fin, NULL)) |
| { |
| case 0 : |
| break; |
| case 1 : |
| ipf_nat_deref((nat_t **)&fin->fin_nat); |
| ip->ip_sum = 0; |
| break; |
| case -1 : |
| err = EINVAL; |
| goto bad; |
| break; |
| } |
| |
| fin->fin_ifp = sifp; |
| fin->fin_out = sout; |
| } |
| ip->ip_sum = 0; |
| |
| |
| if (min->dst != NULL) { |
| dst_release(min->dst); |
| min->dst = NULL; |
| } |
| |
| min->dst = &rt->u.dst; |
| #if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,21) |
| if (min->len > min->dst->pmtu) { |
| err = EMSGSIZE; |
| goto bad; |
| } |
| #endif |
| |
| switch (fin->fin_v) |
| { |
| case 4 : |
| ip->ip_sum = ip_fast_csum((u_char *)ip, ip->ip_hl); |
| |
| /*dumpskbuff(min);*/ |
| NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, min, NULL, |
| ifp, ip_finish_output); |
| err = 0; |
| break; |
| |
| default : |
| err = EINVAL; |
| break; |
| } |
| |
| if (err == 0) |
| return 0; |
| bad: |
| if (min != NULL) |
| kfree_skb(min); |
| return err; |
| } |
| |
| |
| int |
| ipf_ifpaddr(int v, int atype, void *ifptr, i6addr_t *inp, i6addr_t *inpmask) |
| { |
| struct sockaddr_in sin, sinmask; |
| struct net_device *dev; |
| struct in_ifaddr *ifa; |
| struct in_device *ifp; |
| |
| if ((ifptr == NULL) || (ifptr == (void *)-1)) |
| return -1; |
| |
| dev = ifptr; |
| ifp = __in_dev_get(dev); |
| |
| if (v == 4) |
| inp->in4.s_addr = 0; |
| #ifdef USE_INET6 |
| else if (v == 6) |
| return -1; |
| #endif |
| |
| ifa = ifp->ifa_list; |
| while (ifa != NULL) { |
| if (ifa->ifa_flags & IFA_F_SECONDARY) |
| continue; |
| break; |
| } |
| |
| if (ifa == NULL) |
| return -1; |
| |
| sin.sin_family = AF_INET; |
| sinmask.sin_addr.s_addr = ifa->ifa_mask; |
| if (atype == FRI_BROADCAST) |
| sin.sin_addr.s_addr = ifa->ifa_broadcast; |
| else if (atype == FRI_PEERADDR) |
| sin.sin_addr.s_addr = ifa->ifa_address; |
| else |
| sin.sin_addr.s_addr = ifa->ifa_local; |
| |
| return ipf_ifpfillv4addr(atype, (struct sockaddr_in *)&sin, |
| (struct sockaddr_in *)&sinmask, |
| &inp->in4, &inpmask->in4); |
| } |
| |
| |
| void |
| m_copydata(mb_t *m, int off, int len, caddr_t cp) |
| { |
| bcopy(MTOD(m, char *) + off, cp, len); |
| } |
| |
| |
| static int |
| ipf_zerostats(caddr_t data) |
| { |
| friostat_t fio; |
| int error; |
| |
| ipf_getstat(&fio); |
| error = copyoutptr(&fio, data, sizeof(fio)); |
| if (error) |
| return EFAULT; |
| |
| bzero((char *)frstats, sizeof(*frstats) * 2); |
| |
| return 0; |
| } |
| |
| |
| int |
| iplattach() |
| { |
| int err, i; |
| |
| SPL_NET(s); |
| if (ipf_running > 0) { |
| SPL_X(s); |
| return -EBUSY; |
| } |
| |
| bzero((char *)ipf_cache, sizeof(ipf_cache)); |
| MUTEX_INIT(&ipf_rw, "ipf rw mutex"); |
| MUTEX_INIT(&ipl_mutex, "ipf log mutex"); |
| MUTEX_INIT(&ipf_timeoutlock, "ipf timeout lock mutex"); |
| RWLOCK_INIT(&ipf_ipidfrag, "ipf IP NAT-Frag rwlock"); |
| |
| for (i = 0; i < sizeof(ipf_hooks)/sizeof(ipf_hooks[0]); i++) { |
| err = nf_register_hook(&ipf_hooks[i]); |
| if (err != 0) |
| return err; |
| } |
| |
| if (ipf_initialise() < 0) { |
| for (i = 0; i < sizeof(ipf_hooks)/sizeof(ipf_hooks[0]); i++) |
| nf_unregister_hook(&ipf_hooks[i]); |
| SPL_X(s); |
| return EIO; |
| } |
| |
| #ifdef notyet |
| if (ipf_control_forwarding & 1) |
| ipv4_devconf.forwarding = 1; |
| #endif |
| |
| SPL_X(s); |
| /* timeout(ipf_slowtimer, NULL, (hz / IPF_HZ_DIVIDE) * IPF_HZ_MULT); */ |
| return 0; |
| } |
| |
| |
| int |
| ipldetach() |
| { |
| int i; |
| |
| SPL_NET(s); |
| |
| for (i = 0; i < sizeof(ipf_hooks)/sizeof(ipf_hooks[0]); i++) |
| nf_unregister_hook(&ipf_hooks[i]); |
| /* untimeout(ipf_slowtimer, NULL); */ |
| |
| #ifdef notyet |
| if (ipf_control_forwarding & 2) |
| ipv4_devconf.forwarding = 0; |
| #endif |
| |
| ipf_deinitialise(); |
| |
| (void) frflush(IPL_LOGIPF, FR_INQUE|FR_OUTQUE|FR_INACTIVE); |
| (void) frflush(IPL_LOGIPF, FR_INQUE|FR_OUTQUE); |
| |
| MUTEX_DESTROY(&ipf_timeoutlock); |
| MUTEX_DESTROY(&ipl_mutex); |
| MUTEX_DESTROY(&ipf_rw); |
| RW_DESTROY(&ipf_ipidfrag); |
| |
| SPL_X(s); |
| |
| return 0; |
| } |
| |
| |
| static u_int |
| ipf_linux_inout(hooknum, skbp, inifp, outifp, okfn) |
| u_int hooknum; |
| struct sk_buff **skbp; |
| const struct net_device *inifp, *outifp; |
| int (*okfn)(struct sk_buff *); |
| { |
| int result, hlen, dir; |
| void *ifp; |
| ip_t *ip; |
| mb_t *sk; |
| |
| if (inifp == NULL && outifp != NULL) { |
| dir = IPF_OUT; |
| ifp = (void *)outifp; |
| } else if (inifp != NULL && outifp == NULL) { |
| dir = IPF_IN; |
| ifp = (void *)inifp; |
| } else |
| return NF_DROP; |
| |
| sk = *skbp; |
| ip = MTOD(sk, ip_t *); |
| if (ip->ip_v == 4) { |
| hlen = ip->ip_hl << 2; |
| #ifdef USE_INET6 |
| } else if (ip->ip_v == 6) { |
| hlen = sizeof(ip6_t); |
| #endif |
| } else |
| return NF_DROP; |
| result = ipf_check(ip, hlen, (struct net_device *)ifp, dir, skbp); |
| |
| /* |
| * This is kind of not always right...*skbp == NULL might really be |
| * a drop but Linux expects *skbp != NULL for NF_DROP. |
| */ |
| if (*skbp == NULL) |
| return NF_STOLEN; |
| |
| if (result != 0) |
| return NF_DROP; |
| |
| return NF_ACCEPT; |
| } |
| |
| |
| INLINE void |
| ipf_read_enter(ipfrwlock_t *rwlk) |
| { |
| #if defined(IPFDEBUG) && !defined(_KERNEL) |
| if (rwlk->ipf_magic != 0x97dd8b3a) { |
| printk("ipf_read_enter:rwlk %p ipf_magic 0x%x\n", |
| rwlk, rwlk->ipf_magic); |
| rwlk->ipf_magic = 0; |
| *((int *)rwlk->ipf_magic) = 1; |
| } |
| #endif |
| read_lock(&rwlk->ipf_lk); |
| ATOMIC_INC32(rwlk->ipf_isr); |
| } |
| |
| |
| INLINE void |
| ipf_write_enter(ipfrwlock_t *rwlk) |
| { |
| #if defined(IPFDEBUG) && !defined(_KERNEL) |
| if (rwlk->ipf_magic != 0x97dd8b3a) { |
| printk("ipf_write_enter:rwlk %p ipf_magic 0x%x\n", |
| rwlk, rwlk->ipf_magic); |
| rwlk->ipf_magic = 0; |
| *((int *)rwlk->ipf_magic) = 1; |
| } |
| #endif |
| write_lock(&rwlk->ipf_lk); |
| rwlk->ipf_isw = 1; |
| } |
| |
| |
| INLINE void |
| ipf_rw_exit(ipfrwlock_t *rwlk) |
| { |
| #if defined(IPFDEBUG) && !defined(_KERNEL) |
| if (rwlk->ipf_magic != 0x97dd8b3a) { |
| printk("ipf_rw_exit:rwlk %p ipf_magic 0x%x\n", |
| rwlk, rwlk->ipf_magic); |
| rwlk->ipf_magic = 0; |
| *((int *)rwlk->ipf_magic) = 1; |
| } |
| #endif |
| if (rwlk->ipf_isw > 0) { |
| rwlk->ipf_isw = 0; |
| write_unlock(&rwlk->ipf_lk); |
| } else if (rwlk->ipf_isr > 0) { |
| ATOMIC_DEC32(rwlk->ipf_isr); |
| read_unlock(&rwlk->ipf_lk); |
| } else { |
| panic("rwlk->ipf_isw %d isr %d rwlk %p name [%s]\n", |
| rwlk->ipf_isw, rwlk->ipf_isr, rwlk, rwlk->ipf_lname); |
| } |
| } |
| |
| |
| /* |
| * This is not a perfect solution for a downgrade because we can lose the lock |
| * on the object of desire. |
| */ |
| INLINE void |
| ipf_rw_downgrade(ipfrwlock_t *rwlk) |
| { |
| ipf_rw_exit(rwlk); |
| ipf_read_enter(rwlk); |
| } |
| |
| |
| #if 0 |
| void dumpskbuff(sk) |
| struct sk_buff *sk; |
| { |
| char line[80], *t; |
| u_char *s1, *s2; |
| int len, i; |
| u_char c; |
| |
| while (sk != NULL) { |
| len = sk->end - sk->data; |
| |
| for (s1 = MTOD(sk, char *); len > 0; ) { |
| t = line; |
| s2 = s1; |
| for (i = 0; i < len && i < 16; i++) { |
| c = *s2++; |
| sprintf(t, "%02x", c); |
| t += 2; |
| if (i &1) |
| *t++ = ' '; |
| } |
| *t++ = ' '; |
| *t++ = ' '; |
| for (i = 0; i < len && i < 16; i++) { |
| c = *s1++; |
| *t++ = (c >= 32 && c < 127) ? c : '.'; |
| } |
| *t = '\0'; |
| printf("%p@%03d %s\n", sk, s1 - MTOD(sk, u_char *), line); |
| if (len > 16) { |
| len -= 16; |
| s1 += 16; |
| } else |
| len = 0; |
| } |
| sk = sk->next; |
| } |
| } |
| #endif |
| |
| |
| mb_t * |
| m_pullup(mb_t *m, int len) |
| { |
| if (len <= M_LEN(m)) |
| return m; |
| kfree_skb(m); |
| return NULL; |
| } |
| |
| |
| /* ------------------------------------------------------------------------ */ |
| /* Function: ipf_pullup */ |
| /* Returns: NULL == pullup failed, else pointer to protocol header */ |
| /* Parameters: m(I) - pointer to buffer where data packet starts */ |
| /* fin(I) - pointer to packet information */ |
| /* len(I) - number of bytes to pullup */ |
| /* */ |
| /* Attempt to move at least len bytes (from the start of the buffer) into a */ |
| /* single buffer for ease of access. Operating system native functions are */ |
| /* used to manage buffers - if necessary. If the entire packet ends up in */ |
| /* a single buffer, set the FI_COALESCE flag even though ipf_coalesce() has */ |
| /* not been called. Both fin_ip and fin_dp are updated before exiting _IF_ */ |
| /* and ONLY if the pullup succeeds. */ |
| /* */ |
| /* We assume that 'min' is a pointer to a buffer that is part of the chain */ |
| /* of buffers that starts at *fin->fin_mp. */ |
| /* ------------------------------------------------------------------------ */ |
| void * |
| ipf_pullup(mb_t *min, fr_info_t *fin, int len) |
| { |
| int out = fin->fin_out, dpoff, ipoff; |
| mb_t *m = min; |
| char *ip; |
| |
| if (m == NULL) |
| return NULL; |
| |
| ip = (char *)fin->fin_ip; |
| if ((fin->fin_flx & FI_COALESCE) != 0) |
| return ip; |
| |
| ipoff = fin->fin_ipoff; |
| if (fin->fin_dp != NULL) |
| dpoff = (char *)fin->fin_dp - (char *)ip; |
| else |
| dpoff = 0; |
| |
| if (M_LEN(m) < len) { |
| m = m_pullup(m, len); |
| *fin->fin_mp = m; |
| fin->fin_m = m; |
| if (m == NULL) { |
| ATOMIC_INCL(frstats[out].fr_pull[1]); |
| return NULL; |
| } |
| ip = MTOD(m, char *) + ipoff; |
| } |
| |
| ATOMIC_INCL(frstats[out].fr_pull[0]); |
| fin->fin_ip = (ip_t *)ip; |
| if (fin->fin_dp != NULL) |
| fin->fin_dp = (char *)fin->fin_ip + dpoff; |
| |
| if (len == fin->fin_plen) |
| fin->fin_flx |= FI_COALESCE; |
| return ip; |
| } |
| |
| |
| /* |
| * In the face of no kernel random function, this is implemented...it is |
| * not meant to be random, just a fill in. |
| */ |
| int |
| ipf_random(int range) |
| { |
| static int last = 0; |
| static int calls = 0; |
| struct timeval tv; |
| int number; |
| |
| GETKTIME(&tv); |
| last *= tv.tv_usec + calls++; |
| last += (int)&range * ipf_ticks; |
| number = last + tv.tv_sec; |
| number %= range; |
| return number; |
| } |