From f126e746fc01c75bc99329d10ce9127b26b23814 Mon Sep 17 00:00:00 2001 From: Klement Sekera Date: Thu, 10 Oct 2019 09:46:06 +0000 Subject: nat: use SVR Remove NAT's implementation of shallow virtual reassembly with corresponding CLIs, APIs & tests. Replace with standalone shallow virtual reassembly provided by ipX-sv-reass* nodes. Type: refactor Change-Id: I7e6c7487a5a500d591f6871474a359e0993e59b6 Signed-off-by: Klement Sekera --- src/plugins/nat/nat64_in2out.c | 748 +++++++++++------------------------------ 1 file changed, 190 insertions(+), 558 deletions(-) (limited to 'src/plugins/nat/nat64_in2out.c') diff --git a/src/plugins/nat/nat64_in2out.c b/src/plugins/nat/nat64_in2out.c index 8d1d734c459..8d4b1a89cad 100644 --- a/src/plugins/nat/nat64_in2out.c +++ b/src/plugins/nat/nat64_in2out.c @@ -18,7 +18,6 @@ */ #include -#include #include #include #include @@ -47,38 +46,12 @@ format_nat64_in2out_trace (u8 * s, va_list * args) return s; } -typedef struct -{ - u32 sw_if_index; - u32 next_index; - u8 cached; -} nat64_in2out_reass_trace_t; - -static u8 * -format_nat64_in2out_reass_trace (u8 * s, va_list * args) -{ - CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); - CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); - nat64_in2out_reass_trace_t *t = - va_arg (*args, nat64_in2out_reass_trace_t *); - - s = - format (s, "NAT64-in2out-reass: sw_if_index %d, next index %d, status %s", - t->sw_if_index, t->next_index, - t->cached ? "cached" : "translated"); - - return s; -} - - #define foreach_nat64_in2out_error \ _(UNSUPPORTED_PROTOCOL, "unsupported protocol") \ _(IN2OUT_PACKETS, "good in2out packets processed") \ _(NO_TRANSLATION, "no translation") \ _(UNKNOWN, "unknown") \ _(DROP_FRAGMENT, "drop fragment") \ -_(MAX_REASS, "maximum reassemblies exceeded") \ -_(MAX_FRAG, "maximum fragments per reassembly exceeded") \ _(TCP_PACKETS, "TCP packets") \ _(UDP_PACKETS, "UDP packets") \ _(ICMP_PACKETS, "ICMP packets") \ @@ -108,7 +81,6 @@ typedef enum NAT64_IN2OUT_NEXT_IP6_LOOKUP, NAT64_IN2OUT_NEXT_DROP, NAT64_IN2OUT_NEXT_SLOWPATH, - NAT64_IN2OUT_NEXT_REASS, NAT64_IN2OUT_N_NEXT, } nat64_in2out_next_t; @@ -165,32 +137,75 @@ is_hairpinning (ip6_address_t * dst_addr) } static int -nat64_in2out_tcp_udp_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, - void *arg) +nat64_in2out_tcp_udp (vlib_main_t * vm, vlib_buffer_t * p, u16 l4_offset, + u16 frag_hdr_offset, nat64_in2out_set_ctx_t * ctx) { + ip6_header_t *ip6; + ip_csum_t csum = 0; + ip4_header_t *ip4; + u16 fragment_id; + u8 frag_more; + u16 frag_offset; nat64_main_t *nm = &nat64_main; - nat64_in2out_set_ctx_t *ctx = arg; nat64_db_bib_entry_t *bibe; nat64_db_st_entry_t *ste; - ip46_address_t saddr, daddr; + ip46_address_t old_saddr, old_daddr; + ip4_address_t new_daddr; u32 sw_if_index, fib_index; - udp_header_t *udp = ip6_next_header (ip6); - u8 proto = ip6->protocol; - u16 sport = udp->src_port; - u16 dport = udp->dst_port; + u8 proto = vnet_buffer (p)->ip.reass.ip_proto; + u16 sport = vnet_buffer (p)->ip.reass.l4_src_port; + u16 dport = vnet_buffer (p)->ip.reass.l4_dst_port; nat64_db_t *db = &nm->db[ctx->thread_index]; + ip6 = vlib_buffer_get_current (p); + + vlib_buffer_advance (p, l4_offset - sizeof (*ip4)); + ip4 = vlib_buffer_get_current (p); + + u32 ip_version_traffic_class_and_flow_label = + ip6->ip_version_traffic_class_and_flow_label; + u16 payload_length = ip6->payload_length; + u8 hop_limit = ip6->hop_limit; + + old_saddr.as_u64[0] = ip6->src_address.as_u64[0]; + old_saddr.as_u64[1] = ip6->src_address.as_u64[1]; + old_daddr.as_u64[0] = ip6->dst_address.as_u64[0]; + old_daddr.as_u64[1] = ip6->dst_address.as_u64[1]; + + if (PREDICT_FALSE (frag_hdr_offset)) + { + //Only the first fragment + ip6_frag_hdr_t *hdr = + (ip6_frag_hdr_t *) u8_ptr_add (ip6, frag_hdr_offset); + fragment_id = frag_id_6to4 (hdr->identification); + frag_more = ip6_frag_hdr_more (hdr); + frag_offset = ip6_frag_hdr_offset (hdr); + } + else + { + fragment_id = 0; + frag_offset = 0; + frag_more = 0; + } + + ip4->ip_version_and_header_length = + IP4_VERSION_AND_HEADER_LENGTH_NO_OPTIONS; + ip4->tos = ip6_translate_tos (ip_version_traffic_class_and_flow_label); + ip4->length = + u16_net_add (payload_length, sizeof (*ip4) + sizeof (*ip6) - l4_offset); + ip4->fragment_id = fragment_id; + ip4->flags_and_fragment_offset = + clib_host_to_net_u16 (frag_offset | + (frag_more ? IP4_HEADER_FLAG_MORE_FRAGMENTS : 0)); + ip4->ttl = hop_limit; + ip4->protocol = (proto == IP_PROTOCOL_ICMP6) ? IP_PROTOCOL_ICMP : proto; + sw_if_index = vnet_buffer (ctx->b)->sw_if_index[VLIB_RX]; fib_index = fib_table_get_index_for_sw_if_index (FIB_PROTOCOL_IP6, sw_if_index); - saddr.as_u64[0] = ip6->src_address.as_u64[0]; - saddr.as_u64[1] = ip6->src_address.as_u64[1]; - daddr.as_u64[0] = ip6->dst_address.as_u64[0]; - daddr.as_u64[1] = ip6->dst_address.as_u64[1]; - ste = - nat64_db_st_entry_find (db, &saddr, &daddr, sport, dport, proto, + nat64_db_st_entry_find (db, &old_saddr, &old_daddr, sport, dport, proto, fib_index, 1); if (ste) @@ -201,7 +216,8 @@ nat64_in2out_tcp_udp_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, } else { - bibe = nat64_db_bib_entry_find (db, &saddr, sport, proto, fib_index, 1); + bibe = + nat64_db_bib_entry_find (db, &old_saddr, sport, proto, fib_index, 1); if (!bibe) { @@ -214,7 +230,7 @@ nat64_in2out_tcp_udp_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, bibe = nat64_db_bib_entry_create (ctx->thread_index, db, - &ip6->src_address, &out_addr, sport, + &old_saddr.ip6, &out_addr, sport, out_port, fib_index, proto, 0); if (!bibe) return -1; @@ -223,10 +239,10 @@ nat64_in2out_tcp_udp_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, db->bib.bib_entries_num); } - nat64_extract_ip4 (&ip6->dst_address, &daddr.ip4, fib_index); + nat64_extract_ip4 (&old_daddr.ip6, &new_daddr, fib_index); ste = nat64_db_st_entry_create (ctx->thread_index, db, bibe, - &ip6->dst_address, &daddr.ip4, dport); + &old_daddr.ip6, &new_daddr, dport); if (!ste) return -1; @@ -235,22 +251,36 @@ nat64_in2out_tcp_udp_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, } ip4->src_address.as_u32 = bibe->out_addr.as_u32; - udp->src_port = bibe->out_port; - ip4->dst_address.as_u32 = ste->out_r_addr.as_u32; - if (proto == IP_PROTOCOL_TCP) + ip4->checksum = ip4_header_checksum (ip4); + + if (!vnet_buffer (p)->ip.reass.is_non_first_fragment) { - u16 *checksum; - ip_csum_t csum; - tcp_header_t *tcp = ip6_next_header (ip6); + udp_header_t *udp = (udp_header_t *) (ip4 + 1); + udp->src_port = bibe->out_port; - nat64_tcp_session_set_state (ste, tcp, 1); - checksum = &tcp->checksum; - csum = ip_csum_sub_even (*checksum, sport); - csum = ip_csum_add_even (csum, udp->src_port); - mss_clamping (nm->sm, tcp, &csum); - *checksum = ip_csum_fold (csum); + //UDP checksum is optional over IPv4 + if (proto == IP_PROTOCOL_UDP) + { + udp->checksum = 0; + } + else + { + tcp_header_t *tcp = (tcp_header_t *) (ip4 + 1); + csum = ip_csum_sub_even (tcp->checksum, old_saddr.as_u64[0]); + csum = ip_csum_sub_even (csum, old_saddr.as_u64[1]); + csum = ip_csum_sub_even (csum, old_daddr.as_u64[0]); + csum = ip_csum_sub_even (csum, old_daddr.as_u64[1]); + csum = ip_csum_add_even (csum, ip4->dst_address.as_u32); + csum = ip_csum_add_even (csum, ip4->src_address.as_u32); + csum = ip_csum_sub_even (csum, sport); + csum = ip_csum_add_even (csum, udp->src_port); + mss_clamping (nm->sm, tcp, &csum); + tcp->checksum = ip_csum_fold (csum); + + nat64_tcp_session_set_state (ste, tcp, 1); + } } nat64_session_reset_timeout (ste, ctx->vm); @@ -480,16 +510,43 @@ unk_proto_st_walk (nat64_db_st_entry_t * ste, void *arg) } static int -nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, - void *arg) +nat64_in2out_unk_proto (vlib_main_t * vm, vlib_buffer_t * p, u8 l4_protocol, + u16 l4_offset, u16 frag_hdr_offset, + nat64_in2out_set_ctx_t * s_ctx) { + ip6_header_t *ip6; + ip4_header_t *ip4; + u16 fragment_id; + u16 frag_offset; + u8 frag_more; + + ip6 = vlib_buffer_get_current (p); + + ip4 = (ip4_header_t *) u8_ptr_add (ip6, l4_offset - sizeof (*ip4)); + + vlib_buffer_advance (p, l4_offset - sizeof (*ip4)); + + if (PREDICT_FALSE (frag_hdr_offset)) + { + //Only the first fragment + ip6_frag_hdr_t *hdr = + (ip6_frag_hdr_t *) u8_ptr_add (ip6, frag_hdr_offset); + fragment_id = frag_id_6to4 (hdr->identification); + frag_offset = ip6_frag_hdr_offset (hdr); + frag_more = ip6_frag_hdr_more (hdr); + } + else + { + fragment_id = 0; + frag_offset = 0; + frag_more = 0; + } + nat64_main_t *nm = &nat64_main; - nat64_in2out_set_ctx_t *s_ctx = arg; nat64_db_bib_entry_t *bibe; nat64_db_st_entry_t *ste; ip46_address_t saddr, daddr, addr; u32 sw_if_index, fib_index; - u8 proto = ip6->protocol; int i; nat64_db_t *db = &nm->db[s_ctx->thread_index]; @@ -503,17 +560,19 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, daddr.as_u64[1] = ip6->dst_address.as_u64[1]; ste = - nat64_db_st_entry_find (db, &saddr, &daddr, 0, 0, proto, fib_index, 1); + nat64_db_st_entry_find (db, &saddr, &daddr, 0, 0, l4_protocol, fib_index, + 1); if (ste) { - bibe = nat64_db_bib_entry_by_index (db, proto, ste->bibe_index); + bibe = nat64_db_bib_entry_by_index (db, l4_protocol, ste->bibe_index); if (!bibe) return -1; } else { - bibe = nat64_db_bib_entry_find (db, &saddr, 0, proto, fib_index, 1); + bibe = + nat64_db_bib_entry_find (db, &saddr, 0, l4_protocol, fib_index, 1); if (!bibe) { @@ -525,7 +584,7 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, .dst_addr.as_u64[1] = ip6->dst_address.as_u64[1], .out_addr.as_u32 = 0, .fib_index = fib_index, - .proto = proto, + .proto = l4_protocol, .thread_index = s_ctx->thread_index, }; @@ -537,7 +596,7 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, /* Verify if out address is not already in use for protocol */ clib_memset (&addr, 0, sizeof (addr)); addr.ip4.as_u32 = ctx.out_addr.as_u32; - if (nat64_db_bib_entry_find (db, &addr, 0, proto, 0, 0)) + if (nat64_db_bib_entry_find (db, &addr, 0, l4_protocol, 0, 0)) ctx.out_addr.as_u32 = 0; if (!ctx.out_addr.as_u32) @@ -545,7 +604,8 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, for (i = 0; i < vec_len (nm->addr_pool); i++) { addr.ip4.as_u32 = nm->addr_pool[i].addr.as_u32; - if (!nat64_db_bib_entry_find (db, &addr, 0, proto, 0, 0)) + if (!nat64_db_bib_entry_find + (db, &addr, 0, l4_protocol, 0, 0)) break; } } @@ -556,7 +616,7 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, bibe = nat64_db_bib_entry_create (s_ctx->thread_index, db, &ip6->src_address, &ctx.out_addr, - 0, 0, fib_index, proto, 0); + 0, 0, fib_index, l4_protocol, 0); if (!bibe) return -1; @@ -580,27 +640,39 @@ nat64_in2out_unk_proto_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, ip4->src_address.as_u32 = bibe->out_addr.as_u32; ip4->dst_address.as_u32 = ste->out_r_addr.as_u32; + ip4->ip_version_and_header_length = + IP4_VERSION_AND_HEADER_LENGTH_NO_OPTIONS; + ip4->tos = ip6_translate_tos (ip6->ip_version_traffic_class_and_flow_label); + ip4->length = u16_net_add (ip6->payload_length, + sizeof (*ip4) + sizeof (*ip6) - l4_offset); + ip4->fragment_id = fragment_id; + ip4->flags_and_fragment_offset = + clib_host_to_net_u16 (frag_offset | + (frag_more ? IP4_HEADER_FLAG_MORE_FRAGMENTS : 0)); + ip4->ttl = ip6->hop_limit; + ip4->protocol = l4_protocol; + ip4->checksum = ip4_header_checksum (ip4); + return 0; } - - static int nat64_in2out_tcp_udp_hairpinning (vlib_main_t * vm, vlib_buffer_t * b, - ip6_header_t * ip6, u32 thread_index) + ip6_header_t * ip6, u32 l4_offset, + u32 thread_index) { nat64_main_t *nm = &nat64_main; nat64_db_bib_entry_t *bibe; nat64_db_st_entry_t *ste; ip46_address_t saddr, daddr; u32 sw_if_index, fib_index; - udp_header_t *udp = ip6_next_header (ip6); - tcp_header_t *tcp = ip6_next_header (ip6); - u8 proto = ip6->protocol; - u16 sport = udp->src_port; - u16 dport = udp->dst_port; - u16 *checksum; - ip_csum_t csum; + udp_header_t *udp = (udp_header_t *) u8_ptr_add (ip6, l4_offset); + tcp_header_t *tcp = (tcp_header_t *) u8_ptr_add (ip6, l4_offset); + u8 proto = vnet_buffer (b)->ip.reass.ip_proto; + u16 sport = vnet_buffer (b)->ip.reass.l4_src_port; + u16 dport = vnet_buffer (b)->ip.reass.l4_dst_port; + u16 *checksum = NULL; + ip_csum_t csum = 0; nat64_db_t *db = &nm->db[thread_index]; sw_if_index = vnet_buffer (b)->sw_if_index[VLIB_RX]; @@ -612,17 +684,17 @@ nat64_in2out_tcp_udp_hairpinning (vlib_main_t * vm, vlib_buffer_t * b, daddr.as_u64[0] = ip6->dst_address.as_u64[0]; daddr.as_u64[1] = ip6->dst_address.as_u64[1]; - if (proto == IP_PROTOCOL_UDP) - checksum = &udp->checksum; - else - checksum = &tcp->checksum; - - csum = ip_csum_sub_even (*checksum, ip6->src_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->src_address.as_u64[1]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[1]); - csum = ip_csum_sub_even (csum, sport); - csum = ip_csum_sub_even (csum, dport); + if (!vnet_buffer (b)->ip.reass.is_non_first_fragment) + { + if (proto == IP_PROTOCOL_UDP) + checksum = &udp->checksum; + else + checksum = &tcp->checksum; + csum = ip_csum_sub_even (*checksum, ip6->src_address.as_u64[0]); + csum = ip_csum_sub_even (csum, ip6->src_address.as_u64[1]); + csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[0]); + csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[1]); + } ste = nat64_db_st_entry_find (db, &saddr, &daddr, sport, dport, proto, @@ -674,7 +746,11 @@ nat64_in2out_tcp_udp_hairpinning (vlib_main_t * vm, vlib_buffer_t * b, nat64_session_reset_timeout (ste, vm); - sport = udp->src_port = bibe->out_port; + if (!vnet_buffer (b)->ip.reass.is_non_first_fragment) + { + udp->src_port = bibe->out_port; + } + nat64_compose_ip6 (&ip6->src_address, &bibe->out_addr, fib_index); clib_memset (&daddr, 0, sizeof (daddr)); @@ -696,15 +772,20 @@ nat64_in2out_tcp_udp_hairpinning (vlib_main_t * vm, vlib_buffer_t * b, ip6->dst_address.as_u64[0] = bibe->in_addr.as_u64[0]; ip6->dst_address.as_u64[1] = bibe->in_addr.as_u64[1]; - udp->dst_port = bibe->in_port; - csum = ip_csum_add_even (csum, ip6->src_address.as_u64[0]); - csum = ip_csum_add_even (csum, ip6->src_address.as_u64[1]); - csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[0]); - csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[1]); - csum = ip_csum_add_even (csum, udp->src_port); - csum = ip_csum_add_even (csum, udp->dst_port); - *checksum = ip_csum_fold (csum); + if (!vnet_buffer (b)->ip.reass.is_non_first_fragment) + { + csum = ip_csum_add_even (csum, ip6->src_address.as_u64[0]); + csum = ip_csum_add_even (csum, ip6->src_address.as_u64[1]); + csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[0]); + csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[1]); + csum = ip_csum_sub_even (csum, sport); + csum = ip_csum_sub_even (csum, dport); + udp->dst_port = bibe->in_port; + csum = ip_csum_add_even (csum, udp->src_port); + csum = ip_csum_add_even (csum, udp->dst_port); + *checksum = ip_csum_fold (csum); + } return 0; } @@ -990,7 +1071,7 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_buffer_t *b0; u32 next0; ip6_header_t *ip60; - u16 l4_offset0, frag_offset0; + u16 l4_offset0, frag_hdr_offset0; u8 l4_protocol0; u32 proto0; nat64_in2out_set_ctx_t ctx0; @@ -1015,8 +1096,8 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, if (PREDICT_FALSE (ip6_parse - (ip60, b0->current_length, &l4_protocol0, &l4_offset0, - &frag_offset0))) + (vm, b0, ip60, b0->current_length, &l4_protocol0, &l4_offset0, + &frag_hdr_offset0))) { next0 = NAT64_IN2OUT_NEXT_DROP; b0->error = node->errors[NAT64_IN2OUT_ERROR_UNKNOWN]; @@ -1051,7 +1132,9 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, goto trace0; } - if (ip6_to_ip4 (b0, nat64_in2out_unk_proto_set_cb, &ctx0)) + if (nat64_in2out_unk_proto + (vm, b0, l4_protocol0, l4_offset0, frag_hdr_offset0, + &ctx0)) { next0 = NAT64_IN2OUT_NEXT_DROP; b0->error = @@ -1070,14 +1153,6 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, } } - if (PREDICT_FALSE - (ip60->protocol == IP_PROTOCOL_IPV6_FRAGMENTATION)) - { - next0 = NAT64_IN2OUT_NEXT_REASS; - fragments++; - goto trace0; - } - if (proto0 == SNAT_PROTOCOL_ICMP) { icmp_packets++; @@ -1095,7 +1170,7 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, } if (icmp6_to_icmp - (b0, nat64_in2out_icmp_set_cb, &ctx0, + (vm, b0, nat64_in2out_icmp_set_cb, &ctx0, nat64_in2out_inner_icmp_set_cb, &ctx0)) { next0 = NAT64_IN2OUT_NEXT_DROP; @@ -1114,7 +1189,7 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, { next0 = NAT64_IN2OUT_NEXT_IP6_LOOKUP; if (nat64_in2out_tcp_udp_hairpinning - (vm, b0, ip60, thread_index)) + (vm, b0, ip60, l4_offset0, thread_index)) { next0 = NAT64_IN2OUT_NEXT_DROP; b0->error = @@ -1123,8 +1198,8 @@ nat64_in2out_node_fn_inline (vlib_main_t * vm, vlib_node_runtime_t * node, goto trace0; } - if (ip6_to_ip4_tcp_udp - (b0, nat64_in2out_tcp_udp_set_cb, &ctx0, 0)) + if (nat64_in2out_tcp_udp + (vm, b0, l4_offset0, frag_hdr_offset0, &ctx0)) { next0 = NAT64_IN2OUT_NEXT_DROP; b0->error = node->errors[NAT64_IN2OUT_ERROR_NO_TRANSLATION]; @@ -1191,7 +1266,6 @@ VLIB_REGISTER_NODE (nat64_in2out_node) = { [NAT64_IN2OUT_NEXT_IP4_LOOKUP] = "ip4-lookup", [NAT64_IN2OUT_NEXT_IP6_LOOKUP] = "ip6-lookup", [NAT64_IN2OUT_NEXT_SLOWPATH] = "nat64-in2out-slowpath", - [NAT64_IN2OUT_NEXT_REASS] = "nat64-in2out-reass", }, }; /* *INDENT-ON* */ @@ -1218,7 +1292,6 @@ VLIB_REGISTER_NODE (nat64_in2out_slowpath_node) = { [NAT64_IN2OUT_NEXT_IP4_LOOKUP] = "ip4-lookup", [NAT64_IN2OUT_NEXT_IP6_LOOKUP] = "ip6-lookup", [NAT64_IN2OUT_NEXT_SLOWPATH] = "nat64-in2out-slowpath", - [NAT64_IN2OUT_NEXT_REASS] = "nat64-in2out-reass", }, }; /* *INDENT-ON* */ @@ -1233,447 +1306,6 @@ typedef struct nat64_in2out_frag_set_ctx_t_ u8 first_frag; } nat64_in2out_frag_set_ctx_t; -static int -nat64_in2out_frag_set_cb (ip6_header_t * ip6, ip4_header_t * ip4, void *arg) -{ - nat64_main_t *nm = &nat64_main; - nat64_in2out_frag_set_ctx_t *ctx = arg; - nat64_db_st_entry_t *ste; - nat64_db_bib_entry_t *bibe; - udp_header_t *udp; - nat64_db_t *db = &nm->db[ctx->thread_index]; - - ste = nat64_db_st_entry_by_index (db, ctx->proto, ctx->sess_index); - if (!ste) - return -1; - - bibe = nat64_db_bib_entry_by_index (db, ctx->proto, ste->bibe_index); - if (!bibe) - return -1; - - nat64_session_reset_timeout (ste, ctx->vm); - - if (ctx->first_frag) - { - udp = (udp_header_t *) u8_ptr_add (ip6, ctx->l4_offset); - - if (ctx->proto == IP_PROTOCOL_TCP) - { - u16 *checksum; - ip_csum_t csum; - tcp_header_t *tcp = (tcp_header_t *) udp; - - nat64_tcp_session_set_state (ste, tcp, 1); - checksum = &tcp->checksum; - csum = ip_csum_sub_even (*checksum, tcp->src_port); - csum = ip_csum_sub_even (csum, ip6->src_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->src_address.as_u64[1]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[1]); - csum = ip_csum_add_even (csum, bibe->out_port); - csum = ip_csum_add_even (csum, bibe->out_addr.as_u32); - csum = ip_csum_add_even (csum, ste->out_r_addr.as_u32); - *checksum = ip_csum_fold (csum); - } - - udp->src_port = bibe->out_port; - } - - ip4->src_address.as_u32 = bibe->out_addr.as_u32; - ip4->dst_address.as_u32 = ste->out_r_addr.as_u32; - - return 0; -} - -static int -nat64_in2out_frag_hairpinning (vlib_buffer_t * b, ip6_header_t * ip6, - nat64_in2out_frag_set_ctx_t * ctx) -{ - nat64_main_t *nm = &nat64_main; - nat64_db_st_entry_t *ste; - nat64_db_bib_entry_t *bibe; - udp_header_t *udp = (udp_header_t *) u8_ptr_add (ip6, ctx->l4_offset); - tcp_header_t *tcp = (tcp_header_t *) udp; - u16 sport = udp->src_port; - u16 dport = udp->dst_port; - u16 *checksum; - ip_csum_t csum; - ip46_address_t daddr; - nat64_db_t *db = &nm->db[ctx->thread_index]; - - if (ctx->first_frag) - { - if (ctx->proto == IP_PROTOCOL_UDP) - checksum = &udp->checksum; - else - checksum = &tcp->checksum; - - csum = ip_csum_sub_even (*checksum, ip6->src_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->src_address.as_u64[1]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[0]); - csum = ip_csum_sub_even (csum, ip6->dst_address.as_u64[1]); - csum = ip_csum_sub_even (csum, sport); - csum = ip_csum_sub_even (csum, dport); - } - - ste = nat64_db_st_entry_by_index (db, ctx->proto, ctx->sess_index); - if (!ste) - return -1; - - bibe = nat64_db_bib_entry_by_index (db, ctx->proto, ste->bibe_index); - if (!bibe) - return -1; - - if (ctx->proto == IP_PROTOCOL_TCP) - nat64_tcp_session_set_state (ste, tcp, 1); - - nat64_session_reset_timeout (ste, ctx->vm); - - sport = bibe->out_port; - dport = ste->r_port; - - nat64_compose_ip6 (&ip6->src_address, &bibe->out_addr, bibe->fib_index); - - clib_memset (&daddr, 0, sizeof (daddr)); - daddr.ip4.as_u32 = ste->out_r_addr.as_u32; - - bibe = 0; - /* *INDENT-OFF* */ - vec_foreach (db, nm->db) - { - bibe = nat64_db_bib_entry_find (db, &daddr, dport, ctx->proto, 0, 0); - - if (bibe) - break; - } - /* *INDENT-ON* */ - - if (!bibe) - return -1; - - ip6->dst_address.as_u64[0] = bibe->in_addr.as_u64[0]; - ip6->dst_address.as_u64[1] = bibe->in_addr.as_u64[1]; - - if (ctx->first_frag) - { - udp->dst_port = bibe->in_port; - udp->src_port = sport; - csum = ip_csum_add_even (csum, ip6->src_address.as_u64[0]); - csum = ip_csum_add_even (csum, ip6->src_address.as_u64[1]); - csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[0]); - csum = ip_csum_add_even (csum, ip6->dst_address.as_u64[1]); - csum = ip_csum_add_even (csum, udp->src_port); - csum = ip_csum_add_even (csum, udp->dst_port); - *checksum = ip_csum_fold (csum); - } - - return 0; -} - -VLIB_NODE_FN (nat64_in2out_reass_node) (vlib_main_t * vm, - vlib_node_runtime_t * node, - vlib_frame_t * frame) -{ - u32 n_left_from, *from, *to_next; - nat64_in2out_next_t next_index; - u32 pkts_processed = 0, cached_fragments = 0; - u32 *fragments_to_drop = 0; - u32 *fragments_to_loopback = 0; - nat64_main_t *nm = &nat64_main; - u32 thread_index = vm->thread_index; - - from = vlib_frame_vector_args (frame); - n_left_from = frame->n_vectors; - next_index = node->cached_next_index; - - while (n_left_from > 0) - { - u32 n_left_to_next; - - vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); - - while (n_left_from > 0 && n_left_to_next > 0) - { - u32 bi0; - vlib_buffer_t *b0; - u32 next0; - u8 cached0 = 0; - ip6_header_t *ip60; - u16 l4_offset0, frag_offset0; - u8 l4_protocol0; - nat_reass_ip6_t *reass0; - ip6_frag_hdr_t *frag0; - nat64_db_bib_entry_t *bibe0; - nat64_db_st_entry_t *ste0; - udp_header_t *udp0; - snat_protocol_t proto0; - u32 sw_if_index0, fib_index0; - ip46_address_t saddr0, daddr0; - nat64_in2out_frag_set_ctx_t ctx0; - nat64_db_t *db = &nm->db[thread_index]; - - /* speculatively enqueue b0 to the current next frame */ - bi0 = from[0]; - to_next[0] = bi0; - from += 1; - to_next += 1; - n_left_from -= 1; - n_left_to_next -= 1; - - b0 = vlib_get_buffer (vm, bi0); - next0 = NAT64_IN2OUT_NEXT_IP4_LOOKUP; - - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; - fib_index0 = - fib_table_get_index_for_sw_if_index (FIB_PROTOCOL_IP6, - sw_if_index0); - - ctx0.thread_index = thread_index; - - if (PREDICT_FALSE (nat_reass_is_drop_frag (1))) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_DROP_FRAGMENT]; - goto trace0; - } - - ip60 = (ip6_header_t *) vlib_buffer_get_current (b0); - - if (PREDICT_FALSE - (ip6_parse - (ip60, b0->current_length, &l4_protocol0, &l4_offset0, - &frag_offset0))) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_UNKNOWN]; - goto trace0; - } - - if (PREDICT_FALSE - (!(l4_protocol0 == IP_PROTOCOL_TCP - || l4_protocol0 == IP_PROTOCOL_UDP))) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_DROP_FRAGMENT]; - goto trace0; - } - - udp0 = (udp_header_t *) u8_ptr_add (ip60, l4_offset0); - frag0 = (ip6_frag_hdr_t *) u8_ptr_add (ip60, frag_offset0); - proto0 = ip_proto_to_snat_proto (l4_protocol0); - - reass0 = nat_ip6_reass_find_or_create (ip60->src_address, - ip60->dst_address, - frag0->identification, - l4_protocol0, - 1, &fragments_to_drop); - - if (PREDICT_FALSE (!reass0)) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_MAX_REASS]; - goto trace0; - } - - if (PREDICT_TRUE (ip6_frag_hdr_offset (frag0))) - { - ctx0.first_frag = 0; - if (PREDICT_FALSE (reass0->sess_index == (u32) ~ 0)) - { - if (nat_ip6_reass_add_fragment - (thread_index, reass0, bi0, &fragments_to_drop)) - { - b0->error = node->errors[NAT64_IN2OUT_ERROR_MAX_FRAG]; - next0 = NAT64_IN2OUT_NEXT_DROP; - goto trace0; - } - cached0 = 1; - goto trace0; - } - } - else - { - ctx0.first_frag = 1; - - saddr0.as_u64[0] = ip60->src_address.as_u64[0]; - saddr0.as_u64[1] = ip60->src_address.as_u64[1]; - daddr0.as_u64[0] = ip60->dst_address.as_u64[0]; - daddr0.as_u64[1] = ip60->dst_address.as_u64[1]; - - ste0 = - nat64_db_st_entry_find (db, &saddr0, &daddr0, - udp0->src_port, udp0->dst_port, - l4_protocol0, fib_index0, 1); - if (!ste0) - { - bibe0 = - nat64_db_bib_entry_find (db, &saddr0, udp0->src_port, - l4_protocol0, fib_index0, 1); - if (!bibe0) - { - u16 out_port0; - ip4_address_t out_addr0; - if (nat64_alloc_out_addr_and_port - (fib_index0, proto0, &out_addr0, &out_port0, - thread_index)) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = - node->errors[NAT64_IN2OUT_ERROR_NO_TRANSLATION]; - goto trace0; - } - - bibe0 = - nat64_db_bib_entry_create (thread_index, db, - &ip60->src_address, - &out_addr0, udp0->src_port, - out_port0, fib_index0, - l4_protocol0, 0); - if (!bibe0) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = - node->errors[NAT64_IN2OUT_ERROR_NO_TRANSLATION]; - goto trace0; - } - vlib_set_simple_counter (&nm->total_bibs, thread_index, - 0, db->bib.bib_entries_num); - } - nat64_extract_ip4 (&ip60->dst_address, &daddr0.ip4, - fib_index0); - ste0 = - nat64_db_st_entry_create (thread_index, db, bibe0, - &ip60->dst_address, &daddr0.ip4, - udp0->dst_port); - if (!ste0) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = - node->errors[NAT64_IN2OUT_ERROR_NO_TRANSLATION]; - goto trace0; - } - - vlib_set_simple_counter (&nm->total_sessions, thread_index, - 0, db->st.st_entries_num); - } - reass0->sess_index = nat64_db_st_entry_get_index (db, ste0); - - nat_ip6_reass_get_frags (reass0, &fragments_to_loopback); - } - - ctx0.sess_index = reass0->sess_index; - ctx0.proto = l4_protocol0; - ctx0.vm = vm; - ctx0.l4_offset = l4_offset0; - - if (PREDICT_FALSE (is_hairpinning (&ip60->dst_address))) - { - next0 = NAT64_IN2OUT_NEXT_IP6_LOOKUP; - if (nat64_in2out_frag_hairpinning (b0, ip60, &ctx0)) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_NO_TRANSLATION]; - } - goto trace0; - } - else - { - if (ip6_to_ip4_fragmented (b0, nat64_in2out_frag_set_cb, &ctx0)) - { - next0 = NAT64_IN2OUT_NEXT_DROP; - b0->error = node->errors[NAT64_IN2OUT_ERROR_UNKNOWN]; - goto trace0; - } - } - - trace0: - if (PREDICT_FALSE - ((node->flags & VLIB_NODE_FLAG_TRACE) - && (b0->flags & VLIB_BUFFER_IS_TRACED))) - { - nat64_in2out_reass_trace_t *t = - vlib_add_trace (vm, node, b0, sizeof (*t)); - t->cached = cached0; - t->sw_if_index = sw_if_index0; - t->next_index = next0; - } - - if (cached0) - { - n_left_to_next++; - to_next--; - cached_fragments++; - } - else - { - pkts_processed += next0 != NAT64_IN2OUT_NEXT_DROP; - - /* verify speculative enqueue, maybe switch current next frame */ - vlib_validate_buffer_enqueue_x1 (vm, node, next_index, - to_next, n_left_to_next, - bi0, next0); - } - - if (n_left_from == 0 && vec_len (fragments_to_loopback)) - { - from = vlib_frame_vector_args (frame); - u32 len = vec_len (fragments_to_loopback); - if (len <= VLIB_FRAME_SIZE) - { - clib_memcpy_fast (from, fragments_to_loopback, - sizeof (u32) * len); - n_left_from = len; - vec_reset_length (fragments_to_loopback); - } - else - { - clib_memcpy_fast (from, fragments_to_loopback + - (len - VLIB_FRAME_SIZE), - sizeof (u32) * VLIB_FRAME_SIZE); - n_left_from = VLIB_FRAME_SIZE; - _vec_len (fragments_to_loopback) = len - VLIB_FRAME_SIZE; - } - } - } - - vlib_put_next_frame (vm, node, next_index, n_left_to_next); - } - - vlib_node_increment_counter (vm, nm->in2out_reass_node_index, - NAT64_IN2OUT_ERROR_PROCESSED_FRAGMENTS, - pkts_processed); - vlib_node_increment_counter (vm, nm->in2out_reass_node_index, - NAT64_IN2OUT_ERROR_CACHED_FRAGMENTS, - cached_fragments); - - nat_send_all_to_node (vm, fragments_to_drop, node, - &node->errors[NAT64_IN2OUT_ERROR_DROP_FRAGMENT], - NAT64_IN2OUT_NEXT_DROP); - - vec_free (fragments_to_drop); - vec_free (fragments_to_loopback); - return frame->n_vectors; -} - -/* *INDENT-OFF* */ -VLIB_REGISTER_NODE (nat64_in2out_reass_node) = { - .name = "nat64-in2out-reass", - .vector_size = sizeof (u32), - .format_trace = format_nat64_in2out_reass_trace, - .type = VLIB_NODE_TYPE_INTERNAL, - .n_errors = ARRAY_LEN (nat64_in2out_error_strings), - .error_strings = nat64_in2out_error_strings, - .n_next_nodes = NAT64_IN2OUT_N_NEXT, - /* edit / add dispositions here */ - .next_nodes = { - [NAT64_IN2OUT_NEXT_DROP] = "error-drop", - [NAT64_IN2OUT_NEXT_IP4_LOOKUP] = "ip4-lookup", - [NAT64_IN2OUT_NEXT_IP6_LOOKUP] = "ip6-lookup", - [NAT64_IN2OUT_NEXT_SLOWPATH] = "nat64-in2out-slowpath", - [NAT64_IN2OUT_NEXT_REASS] = "nat64-in2out-reass", - }, -}; -/* *INDENT-ON* */ #define foreach_nat64_in2out_handoff_error \ _(CONGESTION_DROP, "congestion drop") \ -- cgit 1.2.3-korg