diff --git a/bpf_progs/Android.bp b/bpf_progs/Android.bp
index b3f8ed6..cdf47e7 100644
--- a/bpf_progs/Android.bp
+++ b/bpf_progs/Android.bp
@@ -45,6 +45,7 @@
         "com.android.tethering",
     ],
     visibility: [
+        "//packages/modules/Connectivity/DnsResolver",
         "//packages/modules/Connectivity/netd",
         "//packages/modules/Connectivity/service",
         "//packages/modules/Connectivity/service/native/libs/libclat",
diff --git a/bpf_progs/block.c b/bpf_progs/block.c
index f2a3e62..0a2b0b8 100644
--- a/bpf_progs/block.c
+++ b/bpf_progs/block.c
@@ -19,13 +19,13 @@
 #include <netinet/in.h>
 #include <stdint.h>
 
-// The resulting .o needs to load on the Android T beta 3 bpfloader
-#define BPFLOADER_MIN_VER BPFLOADER_T_BETA3_VERSION
+// The resulting .o needs to load on the Android T bpfloader
+#define BPFLOADER_MIN_VER BPFLOADER_T_VERSION
 
 #include "bpf_helpers.h"
 
-#define ALLOW 1
-#define DISALLOW 0
+static const int ALLOW = 1;
+static const int DISALLOW = 0;
 
 DEFINE_BPF_MAP_GRW(blocked_ports_map, ARRAY, int, uint64_t,
         1024 /* 64K ports -> 1024 u64s */, AID_SYSTEM)
@@ -57,17 +57,22 @@
     return ALLOW;
 }
 
-DEFINE_BPF_PROG_KVER("bind4/block_port", AID_ROOT, AID_SYSTEM,
-                     bind4_block_port, KVER(5, 4, 0))
+// the program need to be accessible/loadable by netd (from netd updatable plugin)
+#define DEFINE_NETD_RO_BPF_PROG(SECTION_NAME, the_prog, min_kver) \
+    DEFINE_BPF_PROG_EXT(SECTION_NAME, AID_ROOT, AID_ROOT, the_prog, min_kver, KVER_INF,  \
+                        BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, MANDATORY, \
+                        "", "netd_readonly/", LOAD_ON_ENG, LOAD_ON_USER, LOAD_ON_USERDEBUG)
+
+DEFINE_NETD_RO_BPF_PROG("bind4/block_port", bind4_block_port, KVER_4_19)
 (struct bpf_sock_addr *ctx) {
     return block_port(ctx);
 }
 
-DEFINE_BPF_PROG_KVER("bind6/block_port", AID_ROOT, AID_SYSTEM,
-                     bind6_block_port, KVER(5, 4, 0))
+DEFINE_NETD_RO_BPF_PROG("bind6/block_port", bind6_block_port, KVER_4_19)
 (struct bpf_sock_addr *ctx) {
     return block_port(ctx);
 }
 
 LICENSE("Apache 2.0");
 CRITICAL("ConnectivityNative");
+DISABLE_BTF_ON_USER_BUILDS();
diff --git a/bpf_progs/bpf_net_helpers.h b/bpf_progs/bpf_net_helpers.h
index ed33cc9..f3c7de5 100644
--- a/bpf_progs/bpf_net_helpers.h
+++ b/bpf_progs/bpf_net_helpers.h
@@ -87,29 +87,18 @@
     if (skb->data_end - skb->data < len) bpf_skb_pull_data(skb, len);
 }
 
-// constants for passing in to 'bool egress'
-static const bool INGRESS = false;
-static const bool EGRESS = true;
+struct egress_bool { bool egress; };
+#define INGRESS ((struct egress_bool){ .egress = false })
+#define EGRESS ((struct egress_bool){ .egress = true })
 
-// constants for passing in to 'bool downstream'
-static const bool UPSTREAM = false;
-static const bool DOWNSTREAM = true;
+struct stream_bool { bool down; };
+#define UPSTREAM ((struct stream_bool){ .down = false })
+#define DOWNSTREAM ((struct stream_bool){ .down = true })
 
-// constants for passing in to 'bool is_ethernet'
-static const bool RAWIP = false;
-static const bool ETHER = true;
+struct rawip_bool { bool rawip; };
+#define ETHER ((struct rawip_bool){ .rawip = false })
+#define RAWIP ((struct rawip_bool){ .rawip = true })
 
-// constants for passing in to 'bool updatetime'
-static const bool NO_UPDATETIME = false;
-static const bool UPDATETIME = true;
-
-// constants for passing in to ignore_on_eng / ignore_on_user / ignore_on_userdebug
-// define's instead of static const due to tm-mainline-prod compiler static_assert limitations
-#define LOAD_ON_ENG false
-#define LOAD_ON_USER false
-#define LOAD_ON_USERDEBUG false
-#define IGNORE_ON_ENG true
-#define IGNORE_ON_USER true
-#define IGNORE_ON_USERDEBUG true
-
-#define KVER_4_14 KVER(4, 14, 0)
+struct updatetime_bool { bool updatetime; };
+#define NO_UPDATETIME ((struct updatetime_bool){ .updatetime = false })
+#define UPDATETIME ((struct updatetime_bool){ .updatetime = true })
diff --git a/bpf_progs/clatd.c b/bpf_progs/clatd.c
index f05b93e..addb02f 100644
--- a/bpf_progs/clatd.c
+++ b/bpf_progs/clatd.c
@@ -30,8 +30,8 @@
 #define __kernel_udphdr udphdr
 #include <linux/udp.h>
 
-// The resulting .o needs to load on the Android T beta 3 bpfloader
-#define BPFLOADER_MIN_VER BPFLOADER_T_BETA3_VERSION
+// The resulting .o needs to load on the Android T bpfloader
+#define BPFLOADER_MIN_VER BPFLOADER_T_VERSION
 
 #include "bpf_helpers.h"
 #include "bpf_net_helpers.h"
@@ -55,8 +55,10 @@
 DEFINE_BPF_MAP_GRW(clat_ingress6_map, HASH, ClatIngress6Key, ClatIngress6Value, 16, AID_SYSTEM)
 
 static inline __always_inline int nat64(struct __sk_buff* skb,
-                                        const bool is_ethernet,
-                                        const unsigned kver) {
+                                        const struct rawip_bool rawip,
+                                        const struct kver_uint kver) {
+    const bool is_ethernet = !rawip.rawip;
+
     // Require ethernet dst mac address to be our unicast address.
     if (is_ethernet && (skb->pkt_type != PACKET_HOST)) return TC_ACT_PIPE;
 
@@ -115,7 +117,7 @@
 
     if (proto == IPPROTO_FRAGMENT) {
         // Fragment handling requires bpf_skb_adjust_room which is 4.14+
-        if (kver < KVER_4_14) return TC_ACT_PIPE;
+        if (!KVER_IS_AT_LEAST(kver, 4, 14, 0)) return TC_ACT_PIPE;
 
         // Must have (ethernet and) ipv6 header and ipv6 fragment extension header
         if (data + l2_header_size + sizeof(*ip6) + sizeof(struct frag_hdr) > data_end)
@@ -138,10 +140,11 @@
     }
 
     switch (proto) {
-        case IPPROTO_TCP:  // For TCP & UDP the checksum neutrality of the chosen IPv6
-        case IPPROTO_UDP:  // address means there is no need to update their checksums.
-        case IPPROTO_GRE:  // We do not need to bother looking at GRE/ESP headers,
-        case IPPROTO_ESP:  // since there is never a checksum to update.
+        case IPPROTO_TCP:      // For TCP, UDP & UDPLITE the checksum neutrality of the chosen
+        case IPPROTO_UDP:      // IPv6 address means there is no need to update their checksums.
+        case IPPROTO_UDPLITE:  //
+        case IPPROTO_GRE:      // We do not need to bother looking at GRE/ESP headers,
+        case IPPROTO_ESP:      // since there is never a checksum to update.
             break;
 
         default:  // do not know how to handle anything else
@@ -232,13 +235,15 @@
     //
     // Note: we currently have no TreeHugger coverage for 4.9-T devices (there are no such
     // Pixel or cuttlefish devices), so likely you won't notice for months if this breaks...
-    if (kver >= KVER_4_14 && frag_off != htons(IP_DF)) {
+    if (KVER_IS_AT_LEAST(kver, 4, 14, 0) && frag_off != htons(IP_DF)) {
         // If we're converting an IPv6 Fragment, we need to trim off 8 more bytes
         // We're beyond recovery on error here... but hard to imagine how this could fail.
         if (bpf_skb_adjust_room(skb, -(__s32)sizeof(struct frag_hdr), BPF_ADJ_ROOM_NET, /*flags*/0))
             return TC_ACT_SHOT;
     }
 
+    try_make_writable(skb, l2_header_size + sizeof(struct iphdr));
+
     // bpf_skb_change_proto() invalidates all pointers - reload them.
     data = (void*)(long)skb->data;
     data_end = (void*)(long)skb->data_end;
@@ -328,12 +333,13 @@
     if (ip4->frag_off & ~htons(IP_DF)) return TC_ACT_PIPE;
 
     switch (ip4->protocol) {
-        case IPPROTO_TCP:  // For TCP & UDP the checksum neutrality of the chosen IPv6
-        case IPPROTO_GRE:  // address means there is no need to update their checksums.
-        case IPPROTO_ESP:  // We do not need to bother looking at GRE/ESP headers,
-            break;         // since there is never a checksum to update.
+        case IPPROTO_TCP:      // For TCP, UDP & UDPLITE the checksum neutrality of the chosen
+        case IPPROTO_UDPLITE:  // IPv6 address means there is no need to update their checksums.
+        case IPPROTO_GRE:      // We do not need to bother looking at GRE/ESP headers,
+        case IPPROTO_ESP:      // since there is never a checksum to update.
+            break;
 
-        case IPPROTO_UDP:  // See above comment, but must also have UDP header...
+        case IPPROTO_UDP:      // See above comment, but must also have UDP header...
             if (data + sizeof(*ip4) + sizeof(struct udphdr) > data_end) return TC_ACT_PIPE;
             const struct udphdr* uh = (const struct udphdr*)(ip4 + 1);
             // If IPv4/UDP checksum is 0 then fallback to clatd so it can calculate the
@@ -416,3 +422,4 @@
 
 LICENSE("Apache 2.0");
 CRITICAL("Connectivity");
+DISABLE_BTF_ON_USER_BUILDS();
diff --git a/bpf_progs/dscpPolicy.c b/bpf_progs/dscpPolicy.c
index 72f63c6..e845a69 100644
--- a/bpf_progs/dscpPolicy.c
+++ b/bpf_progs/dscpPolicy.c
@@ -27,8 +27,8 @@
 #include <stdint.h>
 #include <string.h>
 
-// The resulting .o needs to load on the Android T beta 3 bpfloader
-#define BPFLOADER_MIN_VER BPFLOADER_T_BETA3_VERSION
+// The resulting .o needs to load on the Android T bpfloader
+#define BPFLOADER_MIN_VER BPFLOADER_T_VERSION
 
 #include "bpf_helpers.h"
 #include "dscpPolicy.h"
@@ -222,7 +222,7 @@
 }
 
 DEFINE_BPF_PROG_KVER("schedcls/set_dscp_ether", AID_ROOT, AID_SYSTEM, schedcls_set_dscp_ether,
-                     KVER(5, 15, 0))
+                     KVER_5_15)
 (struct __sk_buff* skb) {
     if (skb->pkt_type != PACKET_HOST) return TC_ACT_PIPE;
 
@@ -238,3 +238,4 @@
 
 LICENSE("Apache 2.0");
 CRITICAL("Connectivity");
+DISABLE_BTF_ON_USER_BUILDS();
diff --git a/bpf_progs/netd.c b/bpf_progs/netd.c
index d1dfb7b..59bbfba 100644
--- a/bpf_progs/netd.c
+++ b/bpf_progs/netd.c
@@ -14,8 +14,8 @@
  * limitations under the License.
  */
 
-// The resulting .o needs to load on the Android T Beta 3 bpfloader
-#define BPFLOADER_MIN_VER BPFLOADER_T_BETA3_VERSION
+// The resulting .o needs to load on the Android T bpfloader
+#define BPFLOADER_MIN_VER BPFLOADER_T_VERSION
 
 #include <bpf_helpers.h>
 #include <linux/bpf.h>
@@ -56,19 +56,21 @@
 // see include/uapi/linux/tcp.h
 #define TCP_FLAG32_OFF 12
 
+#define TCP_FLAG8_OFF (TCP_FLAG32_OFF + 1)
+
 // For maps netd does not need to access
-#define DEFINE_BPF_MAP_NO_NETD(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries)      \
-    DEFINE_BPF_MAP_EXT(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries,              \
-                       AID_ROOT, AID_NET_BW_ACCT, 0060, "fs_bpf_net_shared", "", false, \
-                       BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, LOAD_ON_ENG,       \
-                       LOAD_ON_USER, LOAD_ON_USERDEBUG)
+#define DEFINE_BPF_MAP_NO_NETD(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries) \
+    DEFINE_BPF_MAP_EXT(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries,         \
+                       AID_ROOT, AID_NET_BW_ACCT, 0060, "fs_bpf_net_shared", "",   \
+                       PRIVATE, BPFLOADER_MIN_VER, BPFLOADER_MAX_VER,              \
+                       LOAD_ON_ENG, LOAD_ON_USER, LOAD_ON_USERDEBUG)
 
 // For maps netd only needs read only access to
-#define DEFINE_BPF_MAP_RO_NETD(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries)         \
-    DEFINE_BPF_MAP_EXT(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries,                 \
-                       AID_ROOT, AID_NET_BW_ACCT, 0460, "fs_bpf_netd_readonly", "", false, \
-                       BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, LOAD_ON_ENG,       \
-                       LOAD_ON_USER, LOAD_ON_USERDEBUG)
+#define DEFINE_BPF_MAP_RO_NETD(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries)  \
+    DEFINE_BPF_MAP_EXT(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries,          \
+                       AID_ROOT, AID_NET_BW_ACCT, 0460, "fs_bpf_netd_readonly", "", \
+                       PRIVATE, BPFLOADER_MIN_VER, BPFLOADER_MAX_VER,               \
+                       LOAD_ON_ENG, LOAD_ON_USER, LOAD_ON_USERDEBUG)
 
 // For maps netd needs to be able to read and write
 #define DEFINE_BPF_MAP_RW_NETD(the_map, TYPE, TypeOfKey, TypeOfValue, num_entries) \
@@ -87,27 +89,31 @@
 DEFINE_BPF_MAP_RW_NETD(cookie_tag_map, HASH, uint64_t, UidTagValue, COOKIE_UID_MAP_SIZE)
 DEFINE_BPF_MAP_NO_NETD(uid_counterset_map, HASH, uint32_t, uint8_t, UID_COUNTERSET_MAP_SIZE)
 DEFINE_BPF_MAP_NO_NETD(app_uid_stats_map, HASH, uint32_t, StatsValue, APP_STATS_MAP_SIZE)
-DEFINE_BPF_MAP_RW_NETD(stats_map_A, HASH, StatsKey, StatsValue, STATS_MAP_SIZE)
+DEFINE_BPF_MAP_RO_NETD(stats_map_A, HASH, StatsKey, StatsValue, STATS_MAP_SIZE)
 DEFINE_BPF_MAP_RO_NETD(stats_map_B, HASH, StatsKey, StatsValue, STATS_MAP_SIZE)
 DEFINE_BPF_MAP_NO_NETD(iface_stats_map, HASH, uint32_t, StatsValue, IFACE_STATS_MAP_SIZE)
-DEFINE_BPF_MAP_NO_NETD(uid_owner_map, HASH, uint32_t, UidOwnerValue, UID_OWNER_MAP_SIZE)
-DEFINE_BPF_MAP_RW_NETD(uid_permission_map, HASH, uint32_t, uint8_t, UID_OWNER_MAP_SIZE)
+DEFINE_BPF_MAP_RO_NETD(uid_owner_map, HASH, uint32_t, UidOwnerValue, UID_OWNER_MAP_SIZE)
+DEFINE_BPF_MAP_RO_NETD(uid_permission_map, HASH, uint32_t, uint8_t, UID_OWNER_MAP_SIZE)
+DEFINE_BPF_MAP_NO_NETD(ingress_discard_map, HASH, IngressDiscardKey, IngressDiscardValue,
+                       INGRESS_DISCARD_MAP_SIZE)
 
 /* never actually used from ebpf */
 DEFINE_BPF_MAP_NO_NETD(iface_index_name_map, HASH, uint32_t, IfaceValue, IFACE_INDEX_NAME_MAP_SIZE)
 
 // A single-element configuration array, packet tracing is enabled when 'true'.
 DEFINE_BPF_MAP_EXT(packet_trace_enabled_map, ARRAY, uint32_t, bool, 1,
-                   AID_ROOT, AID_SYSTEM, 0060, "fs_bpf_net_shared", "", false,
+                   AID_ROOT, AID_SYSTEM, 0060, "fs_bpf_net_shared", "", PRIVATE,
                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, LOAD_ON_ENG,
-                   IGNORE_ON_USER, LOAD_ON_USERDEBUG)
+                   LOAD_ON_USER, LOAD_ON_USERDEBUG)
 
-// A ring buffer on which packet information is pushed. This map will only be loaded
-// on eng and userdebug devices. User devices won't load this to save memory.
+// A ring buffer on which packet information is pushed.
 DEFINE_BPF_RINGBUF_EXT(packet_trace_ringbuf, PacketTrace, PACKET_TRACE_BUF_SIZE,
-                       AID_ROOT, AID_SYSTEM, 0060, "fs_bpf_net_shared", "", false,
+                       AID_ROOT, AID_SYSTEM, 0060, "fs_bpf_net_shared", "", PRIVATE,
                        BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, LOAD_ON_ENG,
-                       IGNORE_ON_USER, LOAD_ON_USERDEBUG);
+                       LOAD_ON_USER, LOAD_ON_USERDEBUG);
+
+DEFINE_BPF_MAP_RO_NETD(data_saver_enabled_map, ARRAY, uint32_t, bool,
+                       DATA_SAVER_ENABLED_MAP_SIZE)
 
 // iptables xt_bpf programs need to be usable by both netd and netutils_wrappers
 // selinux contexts, because even non-xt_bpf iptables mutations are implemented as
@@ -124,8 +130,8 @@
 // which is loaded into netd and thus runs as netd uid/gid/selinux context)
 #define DEFINE_NETD_BPF_PROG_KVER_RANGE(SECTION_NAME, prog_uid, prog_gid, the_prog, minKV, maxKV) \
     DEFINE_BPF_PROG_EXT(SECTION_NAME, prog_uid, prog_gid, the_prog,                               \
-                        minKV, maxKV, BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, false,                \
-                        "fs_bpf_netd_readonly", "", false, false, false)
+                        minKV, maxKV, BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, MANDATORY,            \
+                        "fs_bpf_netd_readonly", "", LOAD_ON_ENG, LOAD_ON_USER, LOAD_ON_USERDEBUG)
 
 #define DEFINE_NETD_BPF_PROG_KVER(SECTION_NAME, prog_uid, prog_gid, the_prog, min_kv) \
     DEFINE_NETD_BPF_PROG_KVER_RANGE(SECTION_NAME, prog_uid, prog_gid, the_prog, min_kv, KVER_INF)
@@ -136,14 +142,8 @@
 // programs that only need to be usable by the system server
 #define DEFINE_SYS_BPF_PROG(SECTION_NAME, prog_uid, prog_gid, the_prog) \
     DEFINE_BPF_PROG_EXT(SECTION_NAME, prog_uid, prog_gid, the_prog, KVER_NONE, KVER_INF,  \
-                        BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, false, "fs_bpf_net_shared", \
-                        "", false, false, false)
-
-static __always_inline int is_system_uid(uint32_t uid) {
-    // MIN_SYSTEM_UID is AID_ROOT == 0, so uint32_t is *always* >= 0
-    // MAX_SYSTEM_UID is AID_NOBODY == 9999, while AID_APP_START == 10000
-    return (uid < AID_APP_START);
-}
+                        BPFLOADER_MIN_VER, BPFLOADER_MAX_VER, MANDATORY, \
+                        "fs_bpf_net_shared", "", LOAD_ON_ENG, LOAD_ON_USER, LOAD_ON_USERDEBUG)
 
 /*
  * Note: this blindly assumes an MTU of 1500, and that packets > MTU are always TCP,
@@ -175,8 +175,8 @@
 #define DEFINE_UPDATE_STATS(the_stats_map, TypeOfKey)                                            \
     static __always_inline inline void update_##the_stats_map(const struct __sk_buff* const skb, \
                                                               const TypeOfKey* const key,        \
-                                                              const bool egress,                 \
-                                                              const unsigned kver) {             \
+                                                              const struct egress_bool egress,   \
+                                                              const struct kver_uint kver) {     \
         StatsValue* value = bpf_##the_stats_map##_lookup_elem(key);                              \
         if (!value) {                                                                            \
             StatsValue newValue = {};                                                            \
@@ -196,7 +196,7 @@
                 packets = (payload + mss - 1) / mss;                                             \
                 bytes = tcp_overhead * packets + payload;                                        \
             }                                                                                    \
-            if (egress) {                                                                        \
+            if (egress.egress) {                                                                 \
                 __sync_fetch_and_add(&value->txPackets, packets);                                \
                 __sync_fetch_and_add(&value->txBytes, bytes);                                    \
             } else {                                                                             \
@@ -216,7 +216,7 @@
                                                          const int L3_off,
                                                          void* const to,
                                                          const int len,
-                                                         const unsigned kver) {
+                                                         const struct kver_uint kver) {
     // 'kver' (here and throughout) is the compile time guaranteed minimum kernel version,
     // ie. we're building (a version of) the bpf program for kver (or newer!) kernels.
     //
@@ -233,16 +233,16 @@
     //
     // For similar reasons this will fail with non-offloaded VLAN tags on < 4.19 kernels,
     // since those extend the ethernet header from 14 to 18 bytes.
-    return kver >= KVER(4, 19, 0)
+    return KVER_IS_AT_LEAST(kver, 4, 19, 0)
         ? bpf_skb_load_bytes_relative(skb, L3_off, to, len, BPF_HDR_START_NET)
         : bpf_skb_load_bytes(skb, L3_off, to, len);
 }
 
 static __always_inline inline void do_packet_tracing(
-        const struct __sk_buff* const skb, const bool egress, const uint32_t uid,
-        const uint32_t tag, const bool enable_tracing, const unsigned kver) {
+        const struct __sk_buff* const skb, const struct egress_bool egress, const uint32_t uid,
+        const uint32_t tag, const bool enable_tracing, const struct kver_uint kver) {
     if (!enable_tracing) return;
-    if (kver < KVER(5, 8, 0)) return;
+    if (!KVER_IS_AT_LEAST(kver, 5, 8, 0)) return;
 
     uint32_t mapKey = 0;
     bool* traceConfig = bpf_packet_trace_enabled_map_lookup_elem(&mapKey);
@@ -268,17 +268,41 @@
         (void)bpf_skb_load_bytes_net(skb, IP6_OFFSET(nexthdr), &proto, sizeof(proto), kver);
         L4_off = sizeof(struct ipv6hdr);
         ipVersion = 6;
+        // skip over a *single* HOPOPTS or DSTOPTS extension header (if present)
+        if (proto == IPPROTO_HOPOPTS || proto == IPPROTO_DSTOPTS) {
+            struct {
+                uint8_t proto, len;
+            } ext_hdr;
+            if (!bpf_skb_load_bytes_net(skb, L4_off, &ext_hdr, sizeof(ext_hdr), kver)) {
+                proto = ext_hdr.proto;
+                L4_off += (ext_hdr.len + 1) * 8;
+            }
+        }
     }
 
     uint8_t flags = 0;
     __be16 sport = 0, dport = 0;
-    if (proto == IPPROTO_TCP && L4_off >= 20) {
-        (void)bpf_skb_load_bytes_net(skb, L4_off + TCP_FLAG32_OFF + 1, &flags, sizeof(flags), kver);
-        (void)bpf_skb_load_bytes_net(skb, L4_off + TCP_OFFSET(source), &sport, sizeof(sport), kver);
-        (void)bpf_skb_load_bytes_net(skb, L4_off + TCP_OFFSET(dest), &dport, sizeof(dport), kver);
-    } else if (proto == IPPROTO_UDP && L4_off >= 20) {
-        (void)bpf_skb_load_bytes_net(skb, L4_off + UDP_OFFSET(source), &sport, sizeof(sport), kver);
-        (void)bpf_skb_load_bytes_net(skb, L4_off + UDP_OFFSET(dest), &dport, sizeof(dport), kver);
+    if (L4_off >= 20) {
+      switch (proto) {
+        case IPPROTO_TCP:
+          (void)bpf_skb_load_bytes_net(skb, L4_off + TCP_FLAG8_OFF, &flags, sizeof(flags), kver);
+          // fallthrough
+        case IPPROTO_DCCP:
+        case IPPROTO_UDP:
+        case IPPROTO_UDPLITE:
+        case IPPROTO_SCTP:
+          // all of these L4 protocols start with be16 src & dst port
+          (void)bpf_skb_load_bytes_net(skb, L4_off + 0, &sport, sizeof(sport), kver);
+          (void)bpf_skb_load_bytes_net(skb, L4_off + 2, &dport, sizeof(dport), kver);
+          break;
+        case IPPROTO_ICMP:
+        case IPPROTO_ICMPV6:
+          // Both IPv4 and IPv6 icmp start with u8 type & code, which we store in the bottom
+          // (ie. second) byte of sport/dport (which are be16s), the top byte is already zero.
+          (void)bpf_skb_load_bytes_net(skb, L4_off + 0, (char *)&sport + 1, 1, kver); //type
+          (void)bpf_skb_load_bytes_net(skb, L4_off + 1, (char *)&dport + 1, 1, kver); //code
+          break;
+      }
     }
 
     pkt->timestampNs = bpf_ktime_get_boot_ns();
@@ -290,7 +314,8 @@
     pkt->sport = sport;
     pkt->dport = dport;
 
-    pkt->egress = egress;
+    pkt->egress = egress.egress;
+    pkt->wakeup = !egress.egress && (skb->mark & 0x80000000);  // Fwmark.ingress_cpu_wakeup
     pkt->ipProto = proto;
     pkt->tcpFlags = flags;
     pkt->ipVersion = ipVersion;
@@ -298,8 +323,9 @@
     bpf_packet_trace_ringbuf_submit(pkt);
 }
 
-static __always_inline inline bool skip_owner_match(struct __sk_buff* skb, bool egress,
-                                                    const unsigned kver) {
+static __always_inline inline bool skip_owner_match(struct __sk_buff* skb,
+                                                    const struct egress_bool egress,
+                                                    const struct kver_uint kver) {
     uint32_t flag = 0;
     if (skb->protocol == htons(ETH_P_IP)) {
         uint8_t proto;
@@ -330,7 +356,7 @@
         return false;
     }
     // Always allow RST's, and additionally allow ingress FINs
-    return flag & (TCP_FLAG_RST | (egress ? 0 : TCP_FLAG_FIN));  // false on read failure
+    return flag & (TCP_FLAG_RST | (egress.egress ? 0 : TCP_FLAG_FIN));  // false on read failure
 }
 
 static __always_inline inline BpfConfig getConfig(uint32_t configKey) {
@@ -343,10 +369,35 @@
     return *config;
 }
 
-#define FIREWALL_DROP_IF_SET (OEM_DENY_1_MATCH)
-#define FIREWALL_DROP_IF_UNSET (RESTRICTED_MATCH)
+static __always_inline inline bool ingress_should_discard(struct __sk_buff* skb,
+                                                          const struct kver_uint kver) {
+    // Require 4.19, since earlier kernels don't have bpf_skb_load_bytes_relative() which
+    // provides relative to L3 header reads.  Without that we could fetch the wrong bytes.
+    // Additionally earlier bpf verifiers are much harder to please.
+    if (!KVER_IS_AT_LEAST(kver, 4, 19, 0)) return false;
 
-// Must be __always_inline or the call from inet_socket_create will crash-reboot the system
+    IngressDiscardKey k = {};
+    if (skb->protocol == htons(ETH_P_IP)) {
+        k.daddr.s6_addr32[2] = htonl(0xFFFF);
+        (void)bpf_skb_load_bytes_net(skb, IP4_OFFSET(daddr), &k.daddr.s6_addr32[3], 4, kver);
+    } else if (skb->protocol == htons(ETH_P_IPV6)) {
+        (void)bpf_skb_load_bytes_net(skb, IP6_OFFSET(daddr), &k.daddr, sizeof(k.daddr), kver);
+    } else {
+        return false; // non IPv4/IPv6, so no IP to match on
+    }
+
+    // we didn't check for load success, because destination bytes will be zeroed if
+    // bpf_skb_load_bytes_net() fails, instead we rely on daddr of '::' and '::ffff:0.0.0.0'
+    // never being present in the map itself
+
+    IngressDiscardValue* v = bpf_ingress_discard_map_lookup_elem(&k);
+    if (!v) return false;  // lookup failure -> no protection in place -> allow
+    // if (skb->ifindex == 1) return false;  // allow 'lo', but can't happen - see callsite
+    if (skb->ifindex == v->iif[0]) return false;  // allowed interface
+    if (skb->ifindex == v->iif[1]) return false;  // allowed interface
+    return true;  // disallowed interface
+}
+
 static __always_inline inline int bpf_owner_firewall_match(uint32_t uid) {
     if (is_system_uid(uid)) return PASS;
 
@@ -362,13 +413,9 @@
     return PASS;
 }
 
-// DROP_IF_SET is set of rules that DROP if rule is globally enabled, and per-uid bit is set
-#define DROP_IF_SET (STANDBY_MATCH | OEM_DENY_1_MATCH | OEM_DENY_2_MATCH | OEM_DENY_3_MATCH)
-// DROP_IF_UNSET is set of rules that should DROP if globally enabled, and per-uid bit is NOT set
-#define DROP_IF_UNSET (DOZABLE_MATCH | POWERSAVE_MATCH | RESTRICTED_MATCH | LOW_POWER_STANDBY_MATCH)
-
 static __always_inline inline int bpf_owner_match(struct __sk_buff* skb, uint32_t uid,
-                                                  bool egress, const unsigned kver) {
+                                                  const struct egress_bool egress,
+                                                  const struct kver_uint kver) {
     if (is_system_uid(uid)) return PASS;
 
     if (skip_owner_match(skb, egress, kver)) return PASS;
@@ -379,14 +426,10 @@
     uint32_t uidRules = uidEntry ? uidEntry->rule : 0;
     uint32_t allowed_iif = uidEntry ? uidEntry->iif : 0;
 
-    // Warning: funky bit-wise arithmetic: in parallel, for all DROP_IF_SET/UNSET rules
-    // check whether the rules are globally enabled, and if so whether the rules are
-    // set/unset for the specific uid.  DROP if that is the case for ANY of the rules.
-    // We achieve this by masking out only the bits/rules we're interested in checking,
-    // and negating (via bit-wise xor) the bits/rules that should drop if unset.
-    if (enabledRules & (DROP_IF_SET | DROP_IF_UNSET) & (uidRules ^ DROP_IF_UNSET)) return DROP;
+    if (isBlockedByUidRules(enabledRules, uidRules)) return DROP;
 
-    if (!egress && skb->ifindex != 1) {
+    if (!egress.egress && skb->ifindex != 1) {
+        if (ingress_should_discard(skb, kver)) return DROP;
         if (uidRules & IIF_MATCH) {
             if (allowed_iif && skb->ifindex != allowed_iif) {
                 // Drops packets not coming from lo nor the allowed interface
@@ -405,8 +448,8 @@
 static __always_inline inline void update_stats_with_config(const uint32_t selectedMap,
                                                             const struct __sk_buff* const skb,
                                                             const StatsKey* const key,
-                                                            const bool egress,
-                                                            const unsigned kver) {
+                                                            const struct egress_bool egress,
+                                                            const struct kver_uint kver) {
     if (selectedMap == SELECT_MAP_A) {
         update_stats_map_A(skb, key, egress, kver);
     } else {
@@ -414,9 +457,10 @@
     }
 }
 
-static __always_inline inline int bpf_traffic_account(struct __sk_buff* skb, bool egress,
+static __always_inline inline int bpf_traffic_account(struct __sk_buff* skb,
+                                                      const struct egress_bool egress,
                                                       const bool enable_tracing,
-                                                      const unsigned kver) {
+                                                      const struct kver_uint kver) {
     uint32_t sock_uid = bpf_get_socket_uid(skb);
     uint64_t cookie = bpf_get_socket_cookie(skb);
     UidTagValue* utag = bpf_cookie_tag_map_lookup_elem(&cookie);
@@ -431,10 +475,9 @@
 
     // Always allow and never count clat traffic. Only the IPv4 traffic on the stacked
     // interface is accounted for and subject to usage restrictions.
-    // TODO: remove sock_uid check once Nat464Xlat javaland adds the socket tag AID_CLAT for clat.
-    if (sock_uid == AID_CLAT || uid == AID_CLAT) {
-        return PASS;
-    }
+    // CLAT IPv6 TX sockets are *always* tagged with CLAT uid, see tagSocketAsClat()
+    // CLAT daemon receives via an untagged AF_PACKET socket.
+    if (egress.egress && uid == AID_CLAT) return PASS;
 
     int match = bpf_owner_match(skb, sock_uid, egress, kver);
 
@@ -450,7 +493,7 @@
     }
 
     // If an outbound packet is going to be dropped, we do not count that traffic.
-    if (egress && (match == DROP)) return DROP;
+    if (egress.egress && (match == DROP)) return DROP;
 
     StatsKey key = {.uid = uid, .tag = tag, .counterSet = 0, .ifaceIndex = skb->ifindex};
 
@@ -460,57 +503,81 @@
     uint32_t mapSettingKey = CURRENT_STATS_MAP_CONFIGURATION_KEY;
     uint32_t* selectedMap = bpf_configuration_map_lookup_elem(&mapSettingKey);
 
-    // Use asm("%0 &= 1" : "+r"(match)) before return match,
-    // to help kernel's bpf verifier, so that it can be 100% certain
-    // that the returned value is always BPF_NOMATCH(0) or BPF_MATCH(1).
-    if (!selectedMap) {
-        asm("%0 &= 1" : "+r"(match));
-        return match;
-    }
+    if (!selectedMap) return PASS;  // cannot happen, needed to keep bpf verifier happy
 
     do_packet_tracing(skb, egress, uid, tag, enable_tracing, kver);
     update_stats_with_config(*selectedMap, skb, &key, egress, kver);
     update_app_uid_stats_map(skb, &uid, egress, kver);
+
+    // We've already handled DROP_UNLESS_DNS up above, thus when we reach here the only
+    // possible values of match are DROP(0) or PASS(1), however we need to use
+    // "match &= 1" before 'return match' to help the kernel's bpf verifier,
+    // so that it can be 100% certain that the returned value is always 0 or 1.
+    // We use assembly so that it cannot be optimized out by a too smart compiler.
     asm("%0 &= 1" : "+r"(match));
     return match;
 }
 
-DEFINE_BPF_PROG_EXT("cgroupskb/ingress/stats$trace", AID_ROOT, AID_SYSTEM,
-                    bpf_cgroup_ingress_trace, KVER(5, 8, 0), KVER_INF,
-                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, false,
-                    "fs_bpf_netd_readonly", "", false, true, false)
+// This program is optional, and enables tracing on Android U+, 5.8+ on user builds.
+DEFINE_BPF_PROG_EXT("cgroupskb/ingress/stats$trace_user", AID_ROOT, AID_SYSTEM,
+                    bpf_cgroup_ingress_trace_user, KVER_5_8, KVER_INF,
+                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, OPTIONAL,
+                    "fs_bpf_netd_readonly", "",
+                    IGNORE_ON_ENG, LOAD_ON_USER, IGNORE_ON_USERDEBUG)
 (struct __sk_buff* skb) {
-    return bpf_traffic_account(skb, INGRESS, TRACE_ON, KVER(5, 8, 0));
+    return bpf_traffic_account(skb, INGRESS, TRACE_ON, KVER_5_8);
+}
+
+// This program is required, and enables tracing on Android U+, 5.8+, userdebug/eng.
+DEFINE_BPF_PROG_EXT("cgroupskb/ingress/stats$trace", AID_ROOT, AID_SYSTEM,
+                    bpf_cgroup_ingress_trace, KVER_5_8, KVER_INF,
+                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, MANDATORY,
+                    "fs_bpf_netd_readonly", "",
+                    LOAD_ON_ENG, IGNORE_ON_USER, LOAD_ON_USERDEBUG)
+(struct __sk_buff* skb) {
+    return bpf_traffic_account(skb, INGRESS, TRACE_ON, KVER_5_8);
 }
 
 DEFINE_NETD_BPF_PROG_KVER_RANGE("cgroupskb/ingress/stats$4_19", AID_ROOT, AID_SYSTEM,
-                                bpf_cgroup_ingress_4_19, KVER(4, 19, 0), KVER_INF)
+                                bpf_cgroup_ingress_4_19, KVER_4_19, KVER_INF)
 (struct __sk_buff* skb) {
-    return bpf_traffic_account(skb, INGRESS, TRACE_OFF, KVER(4, 19, 0));
+    return bpf_traffic_account(skb, INGRESS, TRACE_OFF, KVER_4_19);
 }
 
 DEFINE_NETD_BPF_PROG_KVER_RANGE("cgroupskb/ingress/stats$4_14", AID_ROOT, AID_SYSTEM,
-                                bpf_cgroup_ingress_4_14, KVER_NONE, KVER(4, 19, 0))
+                                bpf_cgroup_ingress_4_14, KVER_NONE, KVER_4_19)
 (struct __sk_buff* skb) {
     return bpf_traffic_account(skb, INGRESS, TRACE_OFF, KVER_NONE);
 }
 
-DEFINE_BPF_PROG_EXT("cgroupskb/egress/stats$trace", AID_ROOT, AID_SYSTEM,
-                    bpf_cgroup_egress_trace, KVER(5, 8, 0), KVER_INF,
-                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, false,
-                    "fs_bpf_netd_readonly", "", false, true, false)
+// This program is optional, and enables tracing on Android U+, 5.8+ on user builds.
+DEFINE_BPF_PROG_EXT("cgroupskb/egress/stats$trace_user", AID_ROOT, AID_SYSTEM,
+                    bpf_cgroup_egress_trace_user, KVER_5_8, KVER_INF,
+                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, OPTIONAL,
+                    "fs_bpf_netd_readonly", "",
+                    LOAD_ON_ENG, IGNORE_ON_USER, LOAD_ON_USERDEBUG)
 (struct __sk_buff* skb) {
-    return bpf_traffic_account(skb, EGRESS, TRACE_ON, KVER(5, 8, 0));
+    return bpf_traffic_account(skb, EGRESS, TRACE_ON, KVER_5_8);
+}
+
+// This program is required, and enables tracing on Android U+, 5.8+, userdebug/eng.
+DEFINE_BPF_PROG_EXT("cgroupskb/egress/stats$trace", AID_ROOT, AID_SYSTEM,
+                    bpf_cgroup_egress_trace, KVER_5_8, KVER_INF,
+                    BPFLOADER_IGNORED_ON_VERSION, BPFLOADER_MAX_VER, MANDATORY,
+                    "fs_bpf_netd_readonly", "",
+                    LOAD_ON_ENG, IGNORE_ON_USER, LOAD_ON_USERDEBUG)
+(struct __sk_buff* skb) {
+    return bpf_traffic_account(skb, EGRESS, TRACE_ON, KVER_5_8);
 }
 
 DEFINE_NETD_BPF_PROG_KVER_RANGE("cgroupskb/egress/stats$4_19", AID_ROOT, AID_SYSTEM,
-                                bpf_cgroup_egress_4_19, KVER(4, 19, 0), KVER_INF)
+                                bpf_cgroup_egress_4_19, KVER_4_19, KVER_INF)
 (struct __sk_buff* skb) {
-    return bpf_traffic_account(skb, EGRESS, TRACE_OFF, KVER(4, 19, 0));
+    return bpf_traffic_account(skb, EGRESS, TRACE_OFF, KVER_4_19);
 }
 
 DEFINE_NETD_BPF_PROG_KVER_RANGE("cgroupskb/egress/stats$4_14", AID_ROOT, AID_SYSTEM,
-                                bpf_cgroup_egress_4_14, KVER_NONE, KVER(4, 19, 0))
+                                bpf_cgroup_egress_4_14, KVER_NONE, KVER_4_19)
 (struct __sk_buff* skb) {
     return bpf_traffic_account(skb, EGRESS, TRACE_OFF, KVER_NONE);
 }
@@ -521,9 +588,8 @@
     // Clat daemon does not generate new traffic, all its traffic is accounted for already
     // on the v4-* interfaces (except for the 20 (or 28) extra bytes of IPv6 vs IPv4 overhead,
     // but that can be corrected for later when merging v4-foo stats into interface foo's).
-    // TODO: remove sock_uid check once Nat464Xlat javaland adds the socket tag AID_CLAT for clat.
+    // CLAT sockets are created by system server and tagged as uid CLAT, see tagSocketAsClat()
     uint32_t sock_uid = bpf_get_socket_uid(skb);
-    if (sock_uid == AID_CLAT) return BPF_NOMATCH;
     if (sock_uid == AID_SYSTEM) {
         uint64_t cookie = bpf_get_socket_cookie(skb);
         UidTagValue* utag = bpf_cookie_tag_map_lookup_elem(&cookie);
@@ -586,10 +652,7 @@
     return BPF_NOMATCH;
 }
 
-DEFINE_NETD_BPF_PROG_KVER("cgroupsock/inet/create", AID_ROOT, AID_ROOT, inet_socket_create,
-                          KVER(4, 14, 0))
-(struct bpf_sock* sk) {
-    uint32_t uid = (bpf_get_current_uid_gid() & 0xffffffff);
+static __always_inline inline uint8_t get_app_permissions(uint32_t uid) {
     /*
      * A given app is guaranteed to have the same app ID in all the profiles in
      * which it is installed, and install permission is granted to app for all
@@ -598,13 +661,22 @@
      */
     uint32_t appId = uid % AID_USER_OFFSET;  // == PER_USER_RANGE == 100000
     uint8_t* permissions = bpf_uid_permission_map_lookup_elem(&appId);
-    // If UID is in map but missing BPF_PERMISSION_INTERNET, app has no INTERNET permission.
-    if (permissions && ((*permissions & BPF_PERMISSION_INTERNET) != BPF_PERMISSION_INTERNET)) {
-        return DROP;
+    // if UID not in map, then default to just INTERNET permission.
+    return permissions ? *permissions : BPF_PERMISSION_INTERNET;
+}
+
+DEFINE_NETD_BPF_PROG_KVER("cgroupsock/inet/create", AID_ROOT, AID_ROOT, inet_socket_create,
+                          KVER_4_14)
+(struct bpf_sock* sk) {
+    uint64_t uid = bpf_get_current_uid_gid() & 0xffffffff;
+    // A return value of 1 means allow, everything else means deny.
+    if (get_app_permissions(uid) & BPF_PERMISSION_INTERNET) {
+      return bpf_owner_firewall_match(uid) == PASS ? 1 : 0;
+    } else {
+      return 0;
     }
-    // Only allow if uid is not blocked by the user firewall.
-    return bpf_owner_firewall_match(uid);
 }
 
 LICENSE("Apache 2.0");
 CRITICAL("Connectivity and netd");
+DISABLE_BTF_ON_USER_BUILDS();
diff --git a/bpf_progs/netd.h b/bpf_progs/netd.h
index be604f9..681cb00 100644
--- a/bpf_progs/netd.h
+++ b/bpf_progs/netd.h
@@ -16,6 +16,7 @@
 
 #pragma once
 
+#include <cutils/android_filesystem_config.h>
 #include <linux/if.h>
 #include <linux/if_ether.h>
 #include <linux/in.h>
@@ -55,21 +56,22 @@
 } StatsValue;
 STRUCT_SIZE(StatsValue, 4 * 8);  // 32
 
+#ifdef __cplusplus
+static inline StatsValue& operator+=(StatsValue& lhs, const StatsValue& rhs) {
+    lhs.rxPackets += rhs.rxPackets;
+    lhs.rxBytes += rhs.rxBytes;
+    lhs.txPackets += rhs.txPackets;
+    lhs.txBytes += rhs.txBytes;
+    return lhs;
+}
+#endif
+
 typedef struct {
     char name[IFNAMSIZ];
 } IfaceValue;
 STRUCT_SIZE(IfaceValue, 16);
 
 typedef struct {
-    uint64_t rxBytes;
-    uint64_t rxPackets;
-    uint64_t txBytes;
-    uint64_t txPackets;
-    uint64_t tcpRxPackets;
-    uint64_t tcpTxPackets;
-} Stats;
-
-typedef struct {
   uint64_t timestampNs;
   uint32_t ifindex;
   uint32_t length;
@@ -80,7 +82,8 @@
   __be16 sport;
   __be16 dport;
 
-  bool egress;
+  bool egress:1,
+       wakeup:1;
   uint8_t ipProto;
   uint8_t tcpFlags;
   uint8_t ipVersion; // 4=IPv4, 6=IPv6, 0=unknown
@@ -121,7 +124,9 @@
 static const int IFACE_STATS_MAP_SIZE = 1000;
 static const int CONFIGURATION_MAP_SIZE = 2;
 static const int UID_OWNER_MAP_SIZE = 4000;
+static const int INGRESS_DISCARD_MAP_SIZE = 100;
 static const int PACKET_TRACE_BUF_SIZE = 32 * 1024;
+static const int DATA_SAVER_ENABLED_MAP_SIZE = 1;
 
 #ifdef __cplusplus
 
@@ -165,8 +170,10 @@
 #define CONFIGURATION_MAP_PATH BPF_NETD_PATH "map_netd_configuration_map"
 #define UID_OWNER_MAP_PATH BPF_NETD_PATH "map_netd_uid_owner_map"
 #define UID_PERMISSION_MAP_PATH BPF_NETD_PATH "map_netd_uid_permission_map"
+#define INGRESS_DISCARD_MAP_PATH BPF_NETD_PATH "map_netd_ingress_discard_map"
 #define PACKET_TRACE_RINGBUF_PATH BPF_NETD_PATH "map_netd_packet_trace_ringbuf"
 #define PACKET_TRACE_ENABLED_MAP_PATH BPF_NETD_PATH "map_netd_packet_trace_enabled_map"
+#define DATA_SAVER_ENABLED_MAP_PATH BPF_NETD_PATH "map_netd_data_saver_enabled_map"
 
 #endif // __cplusplus
 
@@ -185,8 +192,9 @@
     OEM_DENY_1_MATCH = (1 << 9),
     OEM_DENY_2_MATCH = (1 << 10),
     OEM_DENY_3_MATCH = (1 << 11),
+    BACKGROUND_MATCH = (1 << 12)
 };
-// LINT.ThenChange(packages/modules/Connectivity/service/src/com/android/server/BpfNetMaps.java)
+// LINT.ThenChange(../framework/src/android/net/BpfNetMapsConstants.java)
 
 enum BpfPermissionMatch {
     BPF_PERMISSION_INTERNET = 1 << 2,
@@ -213,9 +221,47 @@
 } UidOwnerValue;
 STRUCT_SIZE(UidOwnerValue, 2 * 4);  // 8
 
+typedef struct {
+    // The destination ip of the incoming packet.  IPv4 uses IPv4-mapped IPv6 address format.
+    struct in6_addr daddr;
+} IngressDiscardKey;
+STRUCT_SIZE(IngressDiscardKey, 16);  // 16
+
+typedef struct {
+    // Allowed interface indexes.  Use same value multiple times if you just want to match 1 value.
+    uint32_t iif[2];
+} IngressDiscardValue;
+STRUCT_SIZE(IngressDiscardValue, 2 * 4);  // 8
+
 // Entry in the configuration map that stores which UID rules are enabled.
 #define UID_RULES_CONFIGURATION_KEY 0
 // Entry in the configuration map that stores which stats map is currently in use.
 #define CURRENT_STATS_MAP_CONFIGURATION_KEY 1
+// Entry in the data saver enabled map that stores whether data saver is enabled or not.
+#define DATA_SAVER_ENABLED_KEY 0
 
 #undef STRUCT_SIZE
+
+// DROP_IF_SET is set of rules that DROP if rule is globally enabled, and per-uid bit is set
+#define DROP_IF_SET (STANDBY_MATCH | OEM_DENY_1_MATCH | OEM_DENY_2_MATCH | OEM_DENY_3_MATCH)
+// DROP_IF_UNSET is set of rules that should DROP if globally enabled, and per-uid bit is NOT set
+#define DROP_IF_UNSET (DOZABLE_MATCH | POWERSAVE_MATCH | RESTRICTED_MATCH \
+                        | LOW_POWER_STANDBY_MATCH | BACKGROUND_MATCH)
+
+#define FIREWALL_DROP_IF_SET (OEM_DENY_1_MATCH)
+#define FIREWALL_DROP_IF_UNSET (RESTRICTED_MATCH)
+
+// Warning: funky bit-wise arithmetic: in parallel, for all DROP_IF_SET/UNSET rules
+// check whether the rules are globally enabled, and if so whether the rules are
+// set/unset for the specific uid.  DROP if that is the case for ANY of the rules.
+// We achieve this by masking out only the bits/rules we're interested in checking,
+// and negating (via bit-wise xor) the bits/rules that should drop if unset.
+static inline bool isBlockedByUidRules(BpfConfig enabledRules, uint32_t uidRules) {
+    return enabledRules & (DROP_IF_SET | DROP_IF_UNSET) & (uidRules ^ DROP_IF_UNSET);
+}
+
+static inline bool is_system_uid(uint32_t uid) {
+    // MIN_SYSTEM_UID is AID_ROOT == 0, so uint32_t is *always* >= 0
+    // MAX_SYSTEM_UID is AID_NOBODY == 9999, while AID_APP_START == 10000
+    return (uid < AID_APP_START);
+}
diff --git a/bpf_progs/offload.c b/bpf_progs/offload.c
index 80d1a41..90f96a1 100644
--- a/bpf_progs/offload.c
+++ b/bpf_progs/offload.c
@@ -124,8 +124,12 @@
 DEFINE_BPF_MAP_GRW(tether_upstream6_map, HASH, TetherUpstream6Key, Tether6Value, 64,
                    TETHERING_GID)
 
-static inline __always_inline int do_forward6(struct __sk_buff* skb, const bool is_ethernet,
-        const bool downstream, const unsigned kver) {
+static inline __always_inline int do_forward6(struct __sk_buff* skb,
+                                              const struct rawip_bool rawip,
+                                              const struct stream_bool stream,
+                                              const struct kver_uint kver) {
+    const bool is_ethernet = !rawip.rawip;
+
     // Must be meta-ethernet IPv6 frame
     if (skb->protocol != htons(ETH_P_IPV6)) return TC_ACT_PIPE;
 
@@ -184,7 +188,7 @@
         TC_PUNT(NON_GLOBAL_DST);
 
     // In the upstream direction do not forward traffic within the same /64 subnet.
-    if (!downstream && (src32 == dst32) && (ip6->saddr.s6_addr32[1] == ip6->daddr.s6_addr32[1]))
+    if (!stream.down && (src32 == dst32) && (ip6->saddr.s6_addr32[1] == ip6->daddr.s6_addr32[1]))
         TC_PUNT(LOCAL_SRC_DST);
 
     TetherDownstream6Key kd = {
@@ -194,16 +198,18 @@
 
     TetherUpstream6Key ku = {
             .iif = skb->ifindex,
+            // Retrieve the first 64 bits of the source IPv6 address in network order
+            .src64 = *(uint64_t*)&(ip6->saddr.s6_addr32[0]),
     };
-    if (is_ethernet) __builtin_memcpy(downstream ? kd.dstMac : ku.dstMac, eth->h_dest, ETH_ALEN);
+    if (is_ethernet) __builtin_memcpy(stream.down ? kd.dstMac : ku.dstMac, eth->h_dest, ETH_ALEN);
 
-    Tether6Value* v = downstream ? bpf_tether_downstream6_map_lookup_elem(&kd)
-                                 : bpf_tether_upstream6_map_lookup_elem(&ku);
+    Tether6Value* v = stream.down ? bpf_tether_downstream6_map_lookup_elem(&kd)
+                                  : bpf_tether_upstream6_map_lookup_elem(&ku);
 
     // If we don't find any offload information then simply let the core stack handle it...
     if (!v) return TC_ACT_PIPE;
 
-    uint32_t stat_and_limit_k = downstream ? skb->ifindex : v->oif;
+    uint32_t stat_and_limit_k = stream.down ? skb->ifindex : v->oif;
 
     TetherStatsValue* stat_v = bpf_tether_stats_map_lookup_elem(&stat_and_limit_k);
 
@@ -248,7 +254,7 @@
         // We do this even if TX interface is RAWIP and thus does not need an ethernet header,
         // because this is easier and the kernel will strip extraneous ethernet header.
         if (bpf_skb_change_head(skb, sizeof(struct ethhdr), /*flags*/ 0)) {
-            __sync_fetch_and_add(downstream ? &stat_v->rxErrors : &stat_v->txErrors, 1);
+            __sync_fetch_and_add(stream.down ? &stat_v->rxErrors : &stat_v->txErrors, 1);
             TC_PUNT(CHANGE_HEAD_FAILED);
         }
 
@@ -260,7 +266,7 @@
 
         // I do not believe this can ever happen, but keep the verifier happy...
         if (data + sizeof(struct ethhdr) + sizeof(*ip6) > data_end) {
-            __sync_fetch_and_add(downstream ? &stat_v->rxErrors : &stat_v->txErrors, 1);
+            __sync_fetch_and_add(stream.down ? &stat_v->rxErrors : &stat_v->txErrors, 1);
             TC_DROP(TOO_SHORT);
         }
     };
@@ -280,8 +286,8 @@
     // (-ENOTSUPP) if it isn't.
     bpf_csum_update(skb, 0xFFFF - ntohs(old_hl) + ntohs(new_hl));
 
-    __sync_fetch_and_add(downstream ? &stat_v->rxPackets : &stat_v->txPackets, packets);
-    __sync_fetch_and_add(downstream ? &stat_v->rxBytes : &stat_v->txBytes, L3_bytes);
+    __sync_fetch_and_add(stream.down ? &stat_v->rxPackets : &stat_v->txPackets, packets);
+    __sync_fetch_and_add(stream.down ? &stat_v->rxBytes : &stat_v->txBytes, L3_bytes);
 
     // Overwrite any mac header with the new one
     // For a rawip tx interface it will simply be a bunch of zeroes and later stripped.
@@ -323,26 +329,26 @@
 //
 // Hence, these mandatory (must load successfully) implementations for 4.14+ kernels:
 DEFINE_BPF_PROG_KVER("schedcls/tether_downstream6_rawip$4_14", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_downstream6_rawip_4_14, KVER(4, 14, 0))
+                     sched_cls_tether_downstream6_rawip_4_14, KVER_4_14)
 (struct __sk_buff* skb) {
-    return do_forward6(skb, RAWIP, DOWNSTREAM, KVER(4, 14, 0));
+    return do_forward6(skb, RAWIP, DOWNSTREAM, KVER_4_14);
 }
 
 DEFINE_BPF_PROG_KVER("schedcls/tether_upstream6_rawip$4_14", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_upstream6_rawip_4_14, KVER(4, 14, 0))
+                     sched_cls_tether_upstream6_rawip_4_14, KVER_4_14)
 (struct __sk_buff* skb) {
-    return do_forward6(skb, RAWIP, UPSTREAM, KVER(4, 14, 0));
+    return do_forward6(skb, RAWIP, UPSTREAM, KVER_4_14);
 }
 
 // and define no-op stubs for pre-4.14 kernels.
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_downstream6_rawip$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_downstream6_rawip_stub, KVER_NONE, KVER(4, 14, 0))
+                           sched_cls_tether_downstream6_rawip_stub, KVER_NONE, KVER_4_14)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_upstream6_rawip$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_upstream6_rawip_stub, KVER_NONE, KVER(4, 14, 0))
+                           sched_cls_tether_upstream6_rawip_stub, KVER_NONE, KVER_4_14)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
@@ -355,9 +361,10 @@
 
 static inline __always_inline int do_forward4_bottom(struct __sk_buff* skb,
         const int l2_header_size, void* data, const void* data_end,
-        struct ethhdr* eth, struct iphdr* ip, const bool is_ethernet,
-        const bool downstream, const bool updatetime, const bool is_tcp,
-        const unsigned kver) {
+        struct ethhdr* eth, struct iphdr* ip, const struct rawip_bool rawip,
+        const struct stream_bool stream, const struct updatetime_bool updatetime,
+        const bool is_tcp, const struct kver_uint kver) {
+    const bool is_ethernet = !rawip.rawip;
     struct tcphdr* tcph = is_tcp ? (void*)(ip + 1) : NULL;
     struct udphdr* udph = is_tcp ? NULL : (void*)(ip + 1);
 
@@ -415,13 +422,13 @@
     };
     if (is_ethernet) __builtin_memcpy(k.dstMac, eth->h_dest, ETH_ALEN);
 
-    Tether4Value* v = downstream ? bpf_tether_downstream4_map_lookup_elem(&k)
-                                 : bpf_tether_upstream4_map_lookup_elem(&k);
+    Tether4Value* v = stream.down ? bpf_tether_downstream4_map_lookup_elem(&k)
+                                  : bpf_tether_upstream4_map_lookup_elem(&k);
 
     // If we don't find any offload information then simply let the core stack handle it...
     if (!v) return TC_ACT_PIPE;
 
-    uint32_t stat_and_limit_k = downstream ? skb->ifindex : v->oif;
+    uint32_t stat_and_limit_k = stream.down ? skb->ifindex : v->oif;
 
     TetherStatsValue* stat_v = bpf_tether_stats_map_lookup_elem(&stat_and_limit_k);
 
@@ -466,7 +473,7 @@
         // We do this even if TX interface is RAWIP and thus does not need an ethernet header,
         // because this is easier and the kernel will strip extraneous ethernet header.
         if (bpf_skb_change_head(skb, sizeof(struct ethhdr), /*flags*/ 0)) {
-            __sync_fetch_and_add(downstream ? &stat_v->rxErrors : &stat_v->txErrors, 1);
+            __sync_fetch_and_add(stream.down ? &stat_v->rxErrors : &stat_v->txErrors, 1);
             TC_PUNT(CHANGE_HEAD_FAILED);
         }
 
@@ -480,7 +487,7 @@
 
         // I do not believe this can ever happen, but keep the verifier happy...
         if (data + sizeof(struct ethhdr) + sizeof(*ip) + (is_tcp ? sizeof(*tcph) : sizeof(*udph)) > data_end) {
-            __sync_fetch_and_add(downstream ? &stat_v->rxErrors : &stat_v->txErrors, 1);
+            __sync_fetch_and_add(stream.down ? &stat_v->rxErrors : &stat_v->txErrors, 1);
             TC_DROP(TOO_SHORT);
         }
     };
@@ -532,10 +539,10 @@
 
     // This requires the bpf_ktime_get_boot_ns() helper which was added in 5.8,
     // and backported to all Android Common Kernel 4.14+ trees.
-    if (updatetime) v->last_used = bpf_ktime_get_boot_ns();
+    if (updatetime.updatetime) v->last_used = bpf_ktime_get_boot_ns();
 
-    __sync_fetch_and_add(downstream ? &stat_v->rxPackets : &stat_v->txPackets, packets);
-    __sync_fetch_and_add(downstream ? &stat_v->rxBytes : &stat_v->txBytes, L3_bytes);
+    __sync_fetch_and_add(stream.down ? &stat_v->rxPackets : &stat_v->txPackets, packets);
+    __sync_fetch_and_add(stream.down ? &stat_v->rxBytes : &stat_v->txBytes, L3_bytes);
 
     // Redirect to forwarded interface.
     //
@@ -546,8 +553,13 @@
     return bpf_redirect(v->oif, 0 /* this is effectively BPF_F_EGRESS */);
 }
 
-static inline __always_inline int do_forward4(struct __sk_buff* skb, const bool is_ethernet,
-        const bool downstream, const bool updatetime, const unsigned kver) {
+static inline __always_inline int do_forward4(struct __sk_buff* skb,
+                                              const struct rawip_bool rawip,
+                                              const struct stream_bool stream,
+                                              const struct updatetime_bool updatetime,
+                                              const struct kver_uint kver) {
+    const bool is_ethernet = !rawip.rawip;
+
     // Require ethernet dst mac address to be our unicast address.
     if (is_ethernet && (skb->pkt_type != PACKET_HOST)) return TC_ACT_PIPE;
 
@@ -605,16 +617,16 @@
     // in such a situation we can only support TCP.  This also has the added nice benefit of
     // using a separate error counter, and thus making it obvious which version of the program
     // is loaded.
-    if (!updatetime && ip->protocol != IPPROTO_TCP) TC_PUNT(NON_TCP);
+    if (!updatetime.updatetime && ip->protocol != IPPROTO_TCP) TC_PUNT(NON_TCP);
 
     // We do not support offloading anything besides IPv4 TCP and UDP, due to need for NAT,
     // but no need to check this if !updatetime due to check immediately above.
-    if (updatetime && (ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
+    if (updatetime.updatetime && (ip->protocol != IPPROTO_TCP) && (ip->protocol != IPPROTO_UDP))
         TC_PUNT(NON_TCP_UDP);
 
     // We want to make sure that the compiler will, in the !updatetime case, entirely optimize
     // out all the non-tcp logic.  Also note that at this point is_udp === !is_tcp.
-    const bool is_tcp = !updatetime || (ip->protocol == IPPROTO_TCP);
+    const bool is_tcp = !updatetime.updatetime || (ip->protocol == IPPROTO_TCP);
 
     // This is a bit of a hack to make things easier on the bpf verifier.
     // (In particular I believe the Linux 4.14 kernel's verifier can get confused later on about
@@ -635,37 +647,37 @@
     // if the underlying requisite kernel support (bpf_ktime_get_boot_ns) was backported.
     if (is_tcp) {
       return do_forward4_bottom(skb, l2_header_size, data, data_end, eth, ip,
-                                is_ethernet, downstream, updatetime, /* is_tcp */ true, kver);
+                                rawip, stream, updatetime, /* is_tcp */ true, kver);
     } else {
       return do_forward4_bottom(skb, l2_header_size, data, data_end, eth, ip,
-                                is_ethernet, downstream, updatetime, /* is_tcp */ false, kver);
+                                rawip, stream, updatetime, /* is_tcp */ false, kver);
     }
 }
 
 // Full featured (required) implementations for 5.8+ kernels (these are S+ by definition)
 
 DEFINE_BPF_PROG_KVER("schedcls/tether_downstream4_rawip$5_8", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_downstream4_rawip_5_8, KVER(5, 8, 0))
+                     sched_cls_tether_downstream4_rawip_5_8, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, DOWNSTREAM, UPDATETIME, KVER(5, 8, 0));
+    return do_forward4(skb, RAWIP, DOWNSTREAM, UPDATETIME, KVER_5_8);
 }
 
 DEFINE_BPF_PROG_KVER("schedcls/tether_upstream4_rawip$5_8", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_upstream4_rawip_5_8, KVER(5, 8, 0))
+                     sched_cls_tether_upstream4_rawip_5_8, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, UPSTREAM, UPDATETIME, KVER(5, 8, 0));
+    return do_forward4(skb, RAWIP, UPSTREAM, UPDATETIME, KVER_5_8);
 }
 
 DEFINE_BPF_PROG_KVER("schedcls/tether_downstream4_ether$5_8", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_downstream4_ether_5_8, KVER(5, 8, 0))
+                     sched_cls_tether_downstream4_ether_5_8, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, DOWNSTREAM, UPDATETIME, KVER(5, 8, 0));
+    return do_forward4(skb, ETHER, DOWNSTREAM, UPDATETIME, KVER_5_8);
 }
 
 DEFINE_BPF_PROG_KVER("schedcls/tether_upstream4_ether$5_8", TETHERING_UID, TETHERING_GID,
-                     sched_cls_tether_upstream4_ether_5_8, KVER(5, 8, 0))
+                     sched_cls_tether_upstream4_ether_5_8, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, UPSTREAM, UPDATETIME, KVER(5, 8, 0));
+    return do_forward4(skb, ETHER, UPSTREAM, UPDATETIME, KVER_5_8);
 }
 
 // Full featured (optional) implementations for 4.14-S, 4.19-S & 5.4-S kernels
@@ -674,33 +686,33 @@
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_rawip$opt",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_downstream4_rawip_opt,
-                                    KVER(4, 14, 0), KVER(5, 8, 0))
+                                    KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, DOWNSTREAM, UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, RAWIP, DOWNSTREAM, UPDATETIME, KVER_4_14);
 }
 
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_rawip$opt",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_upstream4_rawip_opt,
-                                    KVER(4, 14, 0), KVER(5, 8, 0))
+                                    KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, UPSTREAM, UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, RAWIP, UPSTREAM, UPDATETIME, KVER_4_14);
 }
 
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_ether$opt",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_downstream4_ether_opt,
-                                    KVER(4, 14, 0), KVER(5, 8, 0))
+                                    KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, DOWNSTREAM, UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, ETHER, DOWNSTREAM, UPDATETIME, KVER_4_14);
 }
 
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_ether$opt",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_upstream4_ether_opt,
-                                    KVER(4, 14, 0), KVER(5, 8, 0))
+                                    KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, UPSTREAM, UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, ETHER, UPSTREAM, UPDATETIME, KVER_4_14);
 }
 
 // Partial (TCP-only: will not update 'last_used' field) implementations for 4.14+ kernels.
@@ -718,15 +730,15 @@
 // RAWIP: Required for 5.4-R kernels -- which always support bpf_skb_change_head().
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_rawip$5_4", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_downstream4_rawip_5_4, KVER(5, 4, 0), KVER(5, 8, 0))
+                           sched_cls_tether_downstream4_rawip_5_4, KVER_5_4, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, DOWNSTREAM, NO_UPDATETIME, KVER(5, 4, 0));
+    return do_forward4(skb, RAWIP, DOWNSTREAM, NO_UPDATETIME, KVER_5_4);
 }
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_rawip$5_4", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_upstream4_rawip_5_4, KVER(5, 4, 0), KVER(5, 8, 0))
+                           sched_cls_tether_upstream4_rawip_5_4, KVER_5_4, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, UPSTREAM, NO_UPDATETIME, KVER(5, 4, 0));
+    return do_forward4(skb, RAWIP, UPSTREAM, NO_UPDATETIME, KVER_5_4);
 }
 
 // RAWIP: Optional for 4.14/4.19 (R) kernels -- which support bpf_skb_change_head().
@@ -735,31 +747,31 @@
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_rawip$4_14",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_downstream4_rawip_4_14,
-                                    KVER(4, 14, 0), KVER(5, 4, 0))
+                                    KVER_4_14, KVER_5_4)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, DOWNSTREAM, NO_UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, RAWIP, DOWNSTREAM, NO_UPDATETIME, KVER_4_14);
 }
 
 DEFINE_OPTIONAL_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_rawip$4_14",
                                     TETHERING_UID, TETHERING_GID,
                                     sched_cls_tether_upstream4_rawip_4_14,
-                                    KVER(4, 14, 0), KVER(5, 4, 0))
+                                    KVER_4_14, KVER_5_4)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, RAWIP, UPSTREAM, NO_UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, RAWIP, UPSTREAM, NO_UPDATETIME, KVER_4_14);
 }
 
 // ETHER: Required for 4.14-Q/R, 4.19-Q/R & 5.4-R kernels.
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_ether$4_14", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_downstream4_ether_4_14, KVER(4, 14, 0), KVER(5, 8, 0))
+                           sched_cls_tether_downstream4_ether_4_14, KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, DOWNSTREAM, NO_UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, ETHER, DOWNSTREAM, NO_UPDATETIME, KVER_4_14);
 }
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_ether$4_14", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_upstream4_ether_4_14, KVER(4, 14, 0), KVER(5, 8, 0))
+                           sched_cls_tether_upstream4_ether_4_14, KVER_4_14, KVER_5_8)
 (struct __sk_buff* skb) {
-    return do_forward4(skb, ETHER, UPSTREAM, NO_UPDATETIME, KVER(4, 14, 0));
+    return do_forward4(skb, ETHER, UPSTREAM, NO_UPDATETIME, KVER_4_14);
 }
 
 // Placeholder (no-op) implementations for older Q kernels
@@ -767,13 +779,13 @@
 // RAWIP: 4.9-P/Q, 4.14-P/Q & 4.19-Q kernels -- without bpf_skb_change_head() for tc programs
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_rawip$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_downstream4_rawip_stub, KVER_NONE, KVER(5, 4, 0))
+                           sched_cls_tether_downstream4_rawip_stub, KVER_NONE, KVER_5_4)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_rawip$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_upstream4_rawip_stub, KVER_NONE, KVER(5, 4, 0))
+                           sched_cls_tether_upstream4_rawip_stub, KVER_NONE, KVER_5_4)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
@@ -781,13 +793,13 @@
 // ETHER: 4.9-P/Q kernel
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_downstream4_ether$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_downstream4_ether_stub, KVER_NONE, KVER(4, 14, 0))
+                           sched_cls_tether_downstream4_ether_stub, KVER_NONE, KVER_4_14)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
 
 DEFINE_BPF_PROG_KVER_RANGE("schedcls/tether_upstream4_ether$stub", TETHERING_UID, TETHERING_GID,
-                           sched_cls_tether_upstream4_ether_stub, KVER_NONE, KVER(4, 14, 0))
+                           sched_cls_tether_upstream4_ether_stub, KVER_NONE, KVER_4_14)
 (struct __sk_buff* skb) {
     return TC_ACT_PIPE;
 }
@@ -796,17 +808,18 @@
 
 DEFINE_BPF_MAP_GRW(tether_dev_map, DEVMAP_HASH, uint32_t, uint32_t, 64, TETHERING_GID)
 
-static inline __always_inline int do_xdp_forward6(struct xdp_md *ctx, const bool is_ethernet,
-        const bool downstream) {
+static inline __always_inline int do_xdp_forward6(struct xdp_md *ctx, const struct rawip_bool rawip,
+        const struct stream_bool stream) {
     return XDP_PASS;
 }
 
-static inline __always_inline int do_xdp_forward4(struct xdp_md *ctx, const bool is_ethernet,
-        const bool downstream) {
+static inline __always_inline int do_xdp_forward4(struct xdp_md *ctx, const struct rawip_bool rawip,
+        const struct stream_bool stream) {
     return XDP_PASS;
 }
 
-static inline __always_inline int do_xdp_forward_ether(struct xdp_md *ctx, const bool downstream) {
+static inline __always_inline int do_xdp_forward_ether(struct xdp_md *ctx,
+                                                       const struct stream_bool stream) {
     const void* data = (void*)(long)ctx->data;
     const void* data_end = (void*)(long)ctx->data_end;
     const struct ethhdr* eth = data;
@@ -815,15 +828,16 @@
     if ((void*)(eth + 1) > data_end) return XDP_PASS;
 
     if (eth->h_proto == htons(ETH_P_IPV6))
-        return do_xdp_forward6(ctx, ETHER, downstream);
+        return do_xdp_forward6(ctx, ETHER, stream);
     if (eth->h_proto == htons(ETH_P_IP))
-        return do_xdp_forward4(ctx, ETHER, downstream);
+        return do_xdp_forward4(ctx, ETHER, stream);
 
     // Anything else we don't know how to handle...
     return XDP_PASS;
 }
 
-static inline __always_inline int do_xdp_forward_rawip(struct xdp_md *ctx, const bool downstream) {
+static inline __always_inline int do_xdp_forward_rawip(struct xdp_md *ctx,
+                                                       const struct stream_bool stream) {
     const void* data = (void*)(long)ctx->data;
     const void* data_end = (void*)(long)ctx->data_end;
 
@@ -831,15 +845,15 @@
     if (data_end - data < 1) return XDP_PASS;
     const uint8_t v = (*(uint8_t*)data) >> 4;
 
-    if (v == 6) return do_xdp_forward6(ctx, RAWIP, downstream);
-    if (v == 4) return do_xdp_forward4(ctx, RAWIP, downstream);
+    if (v == 6) return do_xdp_forward6(ctx, RAWIP, stream);
+    if (v == 4) return do_xdp_forward4(ctx, RAWIP, stream);
 
     // Anything else we don't know how to handle...
     return XDP_PASS;
 }
 
 #define DEFINE_XDP_PROG(str, func) \
-    DEFINE_BPF_PROG_KVER(str, TETHERING_UID, TETHERING_GID, func, KVER(5, 9, 0))(struct xdp_md *ctx)
+    DEFINE_BPF_PROG_KVER(str, TETHERING_UID, TETHERING_GID, func, KVER_5_9)(struct xdp_md *ctx)
 
 DEFINE_XDP_PROG("xdp/tether_downstream_ether",
                  xdp_tether_downstream_ether) {
@@ -863,3 +877,4 @@
 
 LICENSE("Apache 2.0");
 CRITICAL("Connectivity (Tethering)");
+DISABLE_BTF_ON_USER_BUILDS();
diff --git a/bpf_progs/offload.h b/bpf_progs/offload.h
index 9dae6c9..1e28f01 100644
--- a/bpf_progs/offload.h
+++ b/bpf_progs/offload.h
@@ -135,10 +135,10 @@
 typedef struct {
     uint32_t iif;              // The input interface index
     uint8_t dstMac[ETH_ALEN];  // destination ethernet mac address (zeroed iff rawip ingress)
-    uint8_t zero[2];           // zero pad for 8 byte alignment
-                               // TODO: extend this to include src ip /64 subnet
+    uint8_t zero[6];           // zero pad for 8 byte alignment
+    uint64_t src64;            // Top 64-bits of the src ip
 } TetherUpstream6Key;
-STRUCT_SIZE(TetherUpstream6Key, 12);
+STRUCT_SIZE(TetherUpstream6Key, 4 + 6 + 6 + 8);  // 24
 
 typedef struct {
     uint32_t iif;              // The input interface index
diff --git a/bpf_progs/test.c b/bpf_progs/test.c
index 091743c..70b08b7 100644
--- a/bpf_progs/test.c
+++ b/bpf_progs/test.c
@@ -49,7 +49,7 @@
 DEFINE_BPF_MAP_GRW(bitmap, ARRAY, int, uint64_t, 2, TETHERING_GID)
 
 DEFINE_BPF_PROG_KVER("xdp/drop_ipv4_udp_ether", TETHERING_UID, TETHERING_GID,
-                      xdp_test, KVER(5, 9, 0))
+                      xdp_test, KVER_5_9)
 (struct xdp_md *ctx) {
     void *data = (void *)(long)ctx->data;
     void *data_end = (void *)(long)ctx->data_end;
@@ -67,3 +67,4 @@
 }
 
 LICENSE("Apache 2.0");
+DISABLE_BTF_ON_USER_BUILDS();