bpf: Pass packets to host via stack instead of redirecting

[ upstream commit 5f50d82 ]

Packets to a host IP are currently redirected via
cilium_host/cilium_net. The reason for this is mostly historic. For
other packets where routing by the kernel routing tables is desired,
packets are already passed on via TC_ACT_OK to the stack directly.

The two cases where this redirection is needed are:
* For proxy redirection due to a kernel limitation on passing the
  routing tables multiple times. This case is left untouched.
* For the HOST_REDIRECT_TO_INGRESS case, e.g. flannel integration. This
  case is left untouched. The IPv4 and IPv6 case is brought in line to
  not accidently lose this logic later on.

A side effect of this is that the skb gets scrubbed including the
skb->mark. The presence of the identity in the skb->mark is being relied
on in a follow-up fix however.

Therfore, pass packets via the stack using TC_ACT_OK. This is faster,
simpler, and allows for the identity to be carried in the mark.

Fixes: #9784

Signed-off-by: Thomas Graf <thomas@cilium.io>
Signed-off-by: Quentin Monnet <quentin@isovalent.com>

bpf/bpf_lxc.c

@@ -348,7 +348,7 @@ static inline int ipv6_l3_from_lxc(struct __sk_buff *skb,
 
 #ifdef ENABLE_ROUTING
 to_host:
-	if (is_defined(ENABLE_HOST_REDIRECT)) {
+	if (is_defined(HOST_REDIRECT_TO_INGRESS)) {
 		union macaddr host_mac = HOST_IFINDEX_MAC;
 
 		ret = ipv6_l3(skb, l3_off, (__u8 *) &router_mac.addr, (__u8 *) &host_mac.addr, METRIC_EGRESS);
@@ -359,7 +359,7 @@ static inline int ipv6_l3_from_lxc(struct __sk_buff *skb,
 				  HOST_IFINDEX, reason, monitor);
 
 		cilium_dbg_capture(skb, DBG_CAPTURE_DELIVERY, HOST_IFINDEX);
-		return redirect(HOST_IFINDEX, 0);
+		return redirect(HOST_IFINDEX, BPF_F_INGRESS);
 	}
 #endif
 
@@ -675,7 +675,7 @@ static inline int handle_ipv4_from_lxc(struct __sk_buff *skb, __u32 *dstID)
 
 #ifdef ENABLE_ROUTING
 to_host:
-	if (is_defined(ENABLE_HOST_REDIRECT)) {
+	if (is_defined(HOST_REDIRECT_TO_INGRESS)) {
 		union macaddr host_mac = HOST_IFINDEX_MAC;
 
 		ret = ipv4_l3(skb, l3_off, (__u8 *) &router_mac.addr, (__u8 *) &host_mac.addr, ip4);
@@ -686,11 +686,7 @@ static inline int handle_ipv4_from_lxc(struct __sk_buff *skb, __u32 *dstID)
 				  reason, monitor);
 
 		cilium_dbg_capture(skb, DBG_CAPTURE_DELIVERY, HOST_IFINDEX);
-#ifdef HOST_REDIRECT_TO_INGRESS
 		return redirect(HOST_IFINDEX, BPF_F_INGRESS);
-#else
-		return redirect(HOST_IFINDEX, 0);
-#endif
 	}
 #endif
 

daemon/bpf.sha

@@ -1,2 +1,2 @@
-GO_BINDATA_SHA1SUM=d03d8900221291eb9f1080707e28776960944277
+GO_BINDATA_SHA1SUM=09a05065f3000f3852a0fe6ee12071c8ee53b68d
 BPF_FILES=../bpf/COPYING ../bpf/Makefile ../bpf/Makefile.bpf ../bpf/bpf_alignchecker.c ../bpf/bpf_features.h ../bpf/bpf_hostdev_ingress.c ../bpf/bpf_ipsec.c ../bpf/bpf_lb.c ../bpf/bpf_lxc.c ../bpf/bpf_netdev.c ../bpf/bpf_network.c ../bpf/bpf_overlay.c ../bpf/bpf_sock.c ../bpf/bpf_xdp.c ../bpf/cilium-map-migrate.c ../bpf/filter_config.h ../bpf/include/bpf/api.h ../bpf/include/elf/elf.h ../bpf/include/elf/gelf.h ../bpf/include/elf/libelf.h ../bpf/include/iproute2/bpf_elf.h ../bpf/include/linux/bpf.h ../bpf/include/linux/bpf_common.h ../bpf/include/linux/byteorder.h ../bpf/include/linux/byteorder/big_endian.h ../bpf/include/linux/byteorder/little_endian.h ../bpf/include/linux/icmp.h ../bpf/include/linux/icmpv6.h ../bpf/include/linux/if_arp.h ../bpf/include/linux/if_ether.h ../bpf/include/linux/if_packet.h ../bpf/include/linux/in.h ../bpf/include/linux/in6.h ../bpf/include/linux/ioctl.h ../bpf/include/linux/ip.h ../bpf/include/linux/ipv6.h ../bpf/include/linux/perf_event.h ../bpf/include/linux/swab.h ../bpf/include/linux/tcp.h ../bpf/include/linux/type_mapper.h ../bpf/include/linux/udp.h ../bpf/init.sh ../bpf/lib/arp.h ../bpf/lib/common.h ../bpf/lib/config.h ../bpf/lib/conntrack.h ../bpf/lib/conntrack_map.h ../bpf/lib/csum.h ../bpf/lib/dbg.h ../bpf/lib/drop.h ../bpf/lib/encap.h ../bpf/lib/eps.h ../bpf/lib/eth.h ../bpf/lib/events.h ../bpf/lib/icmp6.h ../bpf/lib/ipv4.h ../bpf/lib/ipv6.h ../bpf/lib/l3.h ../bpf/lib/l4.h ../bpf/lib/lb.h ../bpf/lib/lxc.h ../bpf/lib/maps.h ../bpf/lib/metrics.h ../bpf/lib/nat.h ../bpf/lib/nat46.h ../bpf/lib/nodeport.h ../bpf/lib/policy.h ../bpf/lib/signal.h ../bpf/lib/tailcall.h ../bpf/lib/trace.h ../bpf/lib/utils.h ../bpf/lib/xdp.h ../bpf/lxc_config.h ../bpf/netdev_config.h ../bpf/node_config.h ../bpf/probes/raw_change_tail.t ../bpf/probes/raw_fib_lookup.t ../bpf/probes/raw_insn.h ../bpf/probes/raw_invalidate_hash.t ../bpf/probes/raw_lpm_map.t ../bpf/probes/raw_lru_map.t ../bpf/probes/raw_main.c ../bpf/probes/raw_map_val_adj.t ../bpf/probes/raw_mark_map_val.t ../bpf/probes/raw_max_insn.t ../bpf/probes/raw_sock_cookie.t ../bpf/run_probes.sh ../bpf/sockops/Makefile ../bpf/sockops/bpf_redir.c ../bpf/sockops/bpf_sockops.c ../bpf/sockops/bpf_sockops.h ../bpf/sockops/sockops_config.h