selftests/bpf: verifier/sock converted to inline assembly
authorEduard Zingerman <eddyz87@gmail.com>
Fri, 21 Apr 2023 17:42:29 +0000 (20:42 +0300)
committerAlexei Starovoitov <ast@kernel.org>
Fri, 21 Apr 2023 19:25:19 +0000 (12:25 -0700)
Test verifier/sock automatically converted to use inline assembly.

Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20230421174234.2391278-20-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
tools/testing/selftests/bpf/prog_tests/verifier.c
tools/testing/selftests/bpf/progs/verifier_sock.c [new file with mode: 0644]
tools/testing/selftests/bpf/verifier/sock.c [deleted file]

index fd4a6e7b41de24a92acf83d149c0949d63cb56ae..906f1c532cb9d5674dc1216fca727e5ea4bce427 100644 (file)
@@ -50,6 +50,7 @@
 #include "verifier_ringbuf.skel.h"
 #include "verifier_runtime_jit.skel.h"
 #include "verifier_search_pruning.skel.h"
+#include "verifier_sock.skel.h"
 #include "verifier_spill_fill.skel.h"
 #include "verifier_stack_ptr.skel.h"
 #include "verifier_uninit.skel.h"
@@ -141,6 +142,7 @@ void test_verifier_regalloc(void)             { RUN(verifier_regalloc); }
 void test_verifier_ringbuf(void)              { RUN(verifier_ringbuf); }
 void test_verifier_runtime_jit(void)          { RUN(verifier_runtime_jit); }
 void test_verifier_search_pruning(void)       { RUN(verifier_search_pruning); }
+void test_verifier_sock(void)                 { RUN(verifier_sock); }
 void test_verifier_spill_fill(void)           { RUN(verifier_spill_fill); }
 void test_verifier_stack_ptr(void)            { RUN(verifier_stack_ptr); }
 void test_verifier_uninit(void)               { RUN(verifier_uninit); }
diff --git a/tools/testing/selftests/bpf/progs/verifier_sock.c b/tools/testing/selftests/bpf/progs/verifier_sock.c
new file mode 100644 (file)
index 0000000..ee76b51
--- /dev/null
@@ -0,0 +1,980 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Converted from tools/testing/selftests/bpf/verifier/sock.c */
+
+#include <linux/bpf.h>
+#include <bpf/bpf_helpers.h>
+#include "bpf_misc.h"
+
+#define sizeof_field(TYPE, MEMBER) sizeof((((TYPE *)0)->MEMBER))
+#define offsetofend(TYPE, MEMBER) \
+       (offsetof(TYPE, MEMBER) + sizeof_field(TYPE, MEMBER))
+
+struct {
+       __uint(type, BPF_MAP_TYPE_REUSEPORT_SOCKARRAY);
+       __uint(max_entries, 1);
+       __type(key, __u32);
+       __type(value, __u64);
+} map_reuseport_array SEC(".maps");
+
+struct {
+       __uint(type, BPF_MAP_TYPE_SOCKHASH);
+       __uint(max_entries, 1);
+       __type(key, int);
+       __type(value, int);
+} map_sockhash SEC(".maps");
+
+struct {
+       __uint(type, BPF_MAP_TYPE_SOCKMAP);
+       __uint(max_entries, 1);
+       __type(key, int);
+       __type(value, int);
+} map_sockmap SEC(".maps");
+
+struct {
+       __uint(type, BPF_MAP_TYPE_XSKMAP);
+       __uint(max_entries, 1);
+       __type(key, int);
+       __type(value, int);
+} map_xskmap SEC(".maps");
+
+struct val {
+       int cnt;
+       struct bpf_spin_lock l;
+};
+
+struct {
+       __uint(type, BPF_MAP_TYPE_SK_STORAGE);
+       __uint(max_entries, 0);
+       __type(key, int);
+       __type(value, struct val);
+       __uint(map_flags, BPF_F_NO_PREALLOC);
+} sk_storage_map SEC(".maps");
+
+SEC("cgroup/skb")
+__description("skb->sk: no NULL check")
+__failure __msg("invalid mem access 'sock_common_or_null'")
+__failure_unpriv
+__naked void skb_sk_no_null_check(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       r0 = *(u32*)(r1 + 0);                           \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("skb->sk: sk->family [non fullsock field]")
+__success __success_unpriv __retval(0)
+__naked void sk_family_non_fullsock_field_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: r0 = *(u32*)(r1 + %[bpf_sock_family]);          \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_family, offsetof(struct bpf_sock, family))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("skb->sk: sk->type [fullsock field]")
+__failure __msg("invalid sock_common access")
+__failure_unpriv
+__naked void sk_sk_type_fullsock_field_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: r0 = *(u32*)(r1 + %[bpf_sock_type]);            \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_sk_fullsock(skb->sk): no !skb->sk check")
+__failure __msg("type=sock_common_or_null expected=sock_common")
+__failure_unpriv
+__naked void sk_no_skb_sk_check_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       call %[bpf_sk_fullsock];                        \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): no NULL check on ret")
+__failure __msg("invalid mem access 'sock_or_null'")
+__failure_unpriv
+__naked void no_null_check_on_ret_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       r0 = *(u32*)(r0 + %[bpf_sock_type]);            \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->type [fullsock field]")
+__success __success_unpriv __retval(0)
+__naked void sk_sk_type_fullsock_field_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u32*)(r0 + %[bpf_sock_type]);            \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->family [non fullsock field]")
+__success __success_unpriv __retval(0)
+__naked void sk_family_non_fullsock_field_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r0 = *(u32*)(r0 + %[bpf_sock_family]);          \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_family, offsetof(struct bpf_sock, family))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->state [narrow load]")
+__success __success_unpriv __retval(0)
+__naked void sk_sk_state_narrow_load(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r0 + %[bpf_sock_state]);            \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_state, offsetof(struct bpf_sock, state))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_port [word load] (backward compatibility)")
+__success __success_unpriv __retval(0)
+__naked void port_word_load_backward_compatibility(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u32*)(r0 + %[bpf_sock_dst_port]);        \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_dst_port, offsetof(struct bpf_sock, dst_port))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_port [half load]")
+__success __success_unpriv __retval(0)
+__naked void sk_dst_port_half_load(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u16*)(r0 + %[bpf_sock_dst_port]);        \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_dst_port, offsetof(struct bpf_sock, dst_port))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_port [half load] (invalid)")
+__failure __msg("invalid sock access")
+__failure_unpriv
+__naked void dst_port_half_load_invalid_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u16*)(r0 + %[__imm_0]);                  \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__imm_0, offsetof(struct bpf_sock, dst_port) + 2),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_port [byte load]")
+__success __success_unpriv __retval(0)
+__naked void sk_dst_port_byte_load(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r2 = *(u8*)(r0 + %[bpf_sock_dst_port]);         \
+       r2 = *(u8*)(r0 + %[__imm_0]);                   \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__imm_0, offsetof(struct bpf_sock, dst_port) + 1),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_dst_port, offsetof(struct bpf_sock, dst_port))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_port [byte load] (invalid)")
+__failure __msg("invalid sock access")
+__failure_unpriv
+__naked void dst_port_byte_load_invalid(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r0 + %[__imm_0]);                   \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__imm_0, offsetof(struct bpf_sock, dst_port) + 2),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): past sk->dst_port [half load] (invalid)")
+__failure __msg("invalid sock access")
+__failure_unpriv
+__naked void dst_port_half_load_invalid_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u16*)(r0 + %[bpf_sock_dst_port__end]);   \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_dst_port__end, offsetofend(struct bpf_sock, dst_port))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->dst_ip6 [load 2nd byte]")
+__success __success_unpriv __retval(0)
+__naked void dst_ip6_load_2nd_byte(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r0 + %[__imm_0]);                   \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__imm_0, offsetof(struct bpf_sock, dst_ip6[0]) + 1),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->type [narrow load]")
+__success __success_unpriv __retval(0)
+__naked void sk_sk_type_narrow_load(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r0 + %[bpf_sock_type]);             \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): sk->protocol [narrow load]")
+__success __success_unpriv __retval(0)
+__naked void sk_sk_protocol_narrow_load(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r0 + %[bpf_sock_protocol]);         \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_protocol, offsetof(struct bpf_sock, protocol))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("sk_fullsock(skb->sk): beyond last field")
+__failure __msg("invalid sock access")
+__failure_unpriv
+__naked void skb_sk_beyond_last_field_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u32*)(r0 + %[bpf_sock_rx_queue_mapping__end]);\
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_sock_rx_queue_mapping__end, offsetofend(struct bpf_sock, rx_queue_mapping))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(skb->sk): no !skb->sk check")
+__failure __msg("type=sock_common_or_null expected=sock_common")
+__failure_unpriv
+__naked void sk_no_skb_sk_check_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       call %[bpf_tcp_sock];                           \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(skb->sk): no NULL check on ret")
+__failure __msg("invalid mem access 'tcp_sock_or_null'")
+__failure_unpriv
+__naked void no_null_check_on_ret_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_tcp_sock];                           \
+       r0 = *(u32*)(r0 + %[bpf_tcp_sock_snd_cwnd]);    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_tcp_sock_snd_cwnd, offsetof(struct bpf_tcp_sock, snd_cwnd))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(skb->sk): tp->snd_cwnd")
+__success __success_unpriv __retval(0)
+__naked void skb_sk_tp_snd_cwnd_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_tcp_sock];                           \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r0 = *(u32*)(r0 + %[bpf_tcp_sock_snd_cwnd]);    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_tcp_sock_snd_cwnd, offsetof(struct bpf_tcp_sock, snd_cwnd))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(skb->sk): tp->bytes_acked")
+__success __success_unpriv __retval(0)
+__naked void skb_sk_tp_bytes_acked(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_tcp_sock];                           \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r0 = *(u64*)(r0 + %[bpf_tcp_sock_bytes_acked]); \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_tcp_sock_bytes_acked, offsetof(struct bpf_tcp_sock, bytes_acked))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(skb->sk): beyond last field")
+__failure __msg("invalid tcp_sock access")
+__failure_unpriv
+__naked void skb_sk_beyond_last_field_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_tcp_sock];                           \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r0 = *(u64*)(r0 + %[bpf_tcp_sock_bytes_acked__end]);\
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_tcp_sock_bytes_acked__end, offsetofend(struct bpf_tcp_sock, bytes_acked))
+       : __clobber_all);
+}
+
+SEC("cgroup/skb")
+__description("bpf_tcp_sock(bpf_sk_fullsock(skb->sk)): tp->snd_cwnd")
+__success __success_unpriv __retval(0)
+__naked void skb_sk_tp_snd_cwnd_2(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r1 = r0;                                        \
+       call %[bpf_tcp_sock];                           \
+       if r0 != 0 goto l2_%=;                          \
+       exit;                                           \
+l2_%=: r0 = *(u32*)(r0 + %[bpf_tcp_sock_snd_cwnd]);    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk)),
+         __imm_const(bpf_tcp_sock_snd_cwnd, offsetof(struct bpf_tcp_sock, snd_cwnd))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("bpf_sk_release(skb->sk)")
+__failure __msg("R1 must be referenced when passed to release function")
+__naked void bpf_sk_release_skb_sk(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 == 0 goto l0_%=;                          \
+       call %[bpf_sk_release];                         \
+l0_%=: r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_release),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("bpf_sk_release(bpf_sk_fullsock(skb->sk))")
+__failure __msg("R1 must be referenced when passed to release function")
+__naked void bpf_sk_fullsock_skb_sk(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r1 = r0;                                        \
+       call %[bpf_sk_release];                         \
+       r0 = 1;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm(bpf_sk_release),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("bpf_sk_release(bpf_tcp_sock(skb->sk))")
+__failure __msg("R1 must be referenced when passed to release function")
+__naked void bpf_tcp_sock_skb_sk(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_tcp_sock];                           \
+       if r0 != 0 goto l1_%=;                          \
+       exit;                                           \
+l1_%=: r1 = r0;                                        \
+       call %[bpf_sk_release];                         \
+       r0 = 1;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_release),
+         __imm(bpf_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("sk_storage_get(map, skb->sk, NULL, 0): value == NULL")
+__success __retval(0)
+__naked void sk_null_0_value_null(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r4 = 0;                                         \
+       r3 = 0;                                         \
+       r2 = r0;                                        \
+       r1 = %[sk_storage_map] ll;                      \
+       call %[bpf_sk_storage_get];                     \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm(bpf_sk_storage_get),
+         __imm_addr(sk_storage_map),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("sk_storage_get(map, skb->sk, 1, 1): value == 1")
+__failure __msg("R3 type=scalar expected=fp")
+__naked void sk_1_1_value_1(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r4 = 1;                                         \
+       r3 = 1;                                         \
+       r2 = r0;                                        \
+       r1 = %[sk_storage_map] ll;                      \
+       call %[bpf_sk_storage_get];                     \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm(bpf_sk_storage_get),
+         __imm_addr(sk_storage_map),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("sk_storage_get(map, skb->sk, &stack_value, 1): stack_value")
+__success __retval(0)
+__naked void stack_value_1_stack_value(void)
+{
+       asm volatile ("                                 \
+       r2 = 0;                                         \
+       *(u64*)(r10 - 8) = r2;                          \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: call %[bpf_sk_fullsock];                        \
+       if r0 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r4 = 1;                                         \
+       r3 = r10;                                       \
+       r3 += -8;                                       \
+       r2 = r0;                                        \
+       r1 = %[sk_storage_map] ll;                      \
+       call %[bpf_sk_storage_get];                     \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_fullsock),
+         __imm(bpf_sk_storage_get),
+         __imm_addr(sk_storage_map),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("bpf_map_lookup_elem(smap, &key)")
+__failure __msg("cannot pass map_type 24 into func bpf_map_lookup_elem")
+__naked void map_lookup_elem_smap_key(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 4) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -4;                                       \
+       r1 = %[sk_storage_map] ll;                      \
+       call %[bpf_map_lookup_elem];                    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm_addr(sk_storage_map)
+       : __clobber_all);
+}
+
+SEC("xdp")
+__description("bpf_map_lookup_elem(xskmap, &key); xs->queue_id")
+__success __retval(0)
+__naked void xskmap_key_xs_queue_id(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 8) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -8;                                       \
+       r1 = %[map_xskmap] ll;                          \
+       call %[bpf_map_lookup_elem];                    \
+       if r0 != 0 goto l0_%=;                          \
+       exit;                                           \
+l0_%=: r0 = *(u32*)(r0 + %[bpf_xdp_sock_queue_id]);    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm_addr(map_xskmap),
+         __imm_const(bpf_xdp_sock_queue_id, offsetof(struct bpf_xdp_sock, queue_id))
+       : __clobber_all);
+}
+
+SEC("sk_skb")
+__description("bpf_map_lookup_elem(sockmap, &key)")
+__failure __msg("Unreleased reference id=2 alloc_insn=6")
+__naked void map_lookup_elem_sockmap_key(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 4) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -4;                                       \
+       r1 = %[map_sockmap] ll;                         \
+       call %[bpf_map_lookup_elem];                    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm_addr(map_sockmap)
+       : __clobber_all);
+}
+
+SEC("sk_skb")
+__description("bpf_map_lookup_elem(sockhash, &key)")
+__failure __msg("Unreleased reference id=2 alloc_insn=6")
+__naked void map_lookup_elem_sockhash_key(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 4) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -4;                                       \
+       r1 = %[map_sockhash] ll;                        \
+       call %[bpf_map_lookup_elem];                    \
+       r0 = 0;                                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm_addr(map_sockhash)
+       : __clobber_all);
+}
+
+SEC("sk_skb")
+__description("bpf_map_lookup_elem(sockmap, &key); sk->type [fullsock field]; bpf_sk_release(sk)")
+__success
+__naked void field_bpf_sk_release_sk_1(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 4) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -4;                                       \
+       r1 = %[map_sockmap] ll;                         \
+       call %[bpf_map_lookup_elem];                    \
+       if r0 != 0 goto l0_%=;                          \
+       exit;                                           \
+l0_%=: r1 = r0;                                        \
+       r0 = *(u32*)(r0 + %[bpf_sock_type]);            \
+       call %[bpf_sk_release];                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm(bpf_sk_release),
+         __imm_addr(map_sockmap),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("sk_skb")
+__description("bpf_map_lookup_elem(sockhash, &key); sk->type [fullsock field]; bpf_sk_release(sk)")
+__success
+__naked void field_bpf_sk_release_sk_2(void)
+{
+       asm volatile ("                                 \
+       r1 = 0;                                         \
+       *(u32*)(r10 - 4) = r1;                          \
+       r2 = r10;                                       \
+       r2 += -4;                                       \
+       r1 = %[map_sockhash] ll;                        \
+       call %[bpf_map_lookup_elem];                    \
+       if r0 != 0 goto l0_%=;                          \
+       exit;                                           \
+l0_%=: r1 = r0;                                        \
+       r0 = *(u32*)(r0 + %[bpf_sock_type]);            \
+       call %[bpf_sk_release];                         \
+       exit;                                           \
+"      :
+       : __imm(bpf_map_lookup_elem),
+         __imm(bpf_sk_release),
+         __imm_addr(map_sockhash),
+         __imm_const(bpf_sock_type, offsetof(struct bpf_sock, type))
+       : __clobber_all);
+}
+
+SEC("sk_reuseport")
+__description("bpf_sk_select_reuseport(ctx, reuseport_array, &key, flags)")
+__success
+__naked void ctx_reuseport_array_key_flags(void)
+{
+       asm volatile ("                                 \
+       r4 = 0;                                         \
+       r2 = 0;                                         \
+       *(u32*)(r10 - 4) = r2;                          \
+       r3 = r10;                                       \
+       r3 += -4;                                       \
+       r2 = %[map_reuseport_array] ll;                 \
+       call %[bpf_sk_select_reuseport];                \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_select_reuseport),
+         __imm_addr(map_reuseport_array)
+       : __clobber_all);
+}
+
+SEC("sk_reuseport")
+__description("bpf_sk_select_reuseport(ctx, sockmap, &key, flags)")
+__success
+__naked void reuseport_ctx_sockmap_key_flags(void)
+{
+       asm volatile ("                                 \
+       r4 = 0;                                         \
+       r2 = 0;                                         \
+       *(u32*)(r10 - 4) = r2;                          \
+       r3 = r10;                                       \
+       r3 += -4;                                       \
+       r2 = %[map_sockmap] ll;                         \
+       call %[bpf_sk_select_reuseport];                \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_select_reuseport),
+         __imm_addr(map_sockmap)
+       : __clobber_all);
+}
+
+SEC("sk_reuseport")
+__description("bpf_sk_select_reuseport(ctx, sockhash, &key, flags)")
+__success
+__naked void reuseport_ctx_sockhash_key_flags(void)
+{
+       asm volatile ("                                 \
+       r4 = 0;                                         \
+       r2 = 0;                                         \
+       *(u32*)(r10 - 4) = r2;                          \
+       r3 = r10;                                       \
+       r3 += -4;                                       \
+       r2 = %[map_sockmap] ll;                         \
+       call %[bpf_sk_select_reuseport];                \
+       exit;                                           \
+"      :
+       : __imm(bpf_sk_select_reuseport),
+         __imm_addr(map_sockmap)
+       : __clobber_all);
+}
+
+SEC("tc")
+__description("mark null check on return value of bpf_skc_to helpers")
+__failure __msg("invalid mem access")
+__naked void of_bpf_skc_to_helpers(void)
+{
+       asm volatile ("                                 \
+       r1 = *(u64*)(r1 + %[__sk_buff_sk]);             \
+       if r1 != 0 goto l0_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l0_%=: r6 = r1;                                        \
+       call %[bpf_skc_to_tcp_sock];                    \
+       r7 = r0;                                        \
+       r1 = r6;                                        \
+       call %[bpf_skc_to_tcp_request_sock];            \
+       r8 = r0;                                        \
+       if r8 != 0 goto l1_%=;                          \
+       r0 = 0;                                         \
+       exit;                                           \
+l1_%=: r0 = *(u8*)(r7 + 0);                            \
+       exit;                                           \
+"      :
+       : __imm(bpf_skc_to_tcp_request_sock),
+         __imm(bpf_skc_to_tcp_sock),
+         __imm_const(__sk_buff_sk, offsetof(struct __sk_buff, sk))
+       : __clobber_all);
+}
+
+char _license[] SEC("license") = "GPL";
diff --git a/tools/testing/selftests/bpf/verifier/sock.c b/tools/testing/selftests/bpf/verifier/sock.c
deleted file mode 100644 (file)
index 108dd3e..0000000
+++ /dev/null
@@ -1,706 +0,0 @@
-{
-       "skb->sk: no NULL check",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_1, 0),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid mem access 'sock_common_or_null'",
-},
-{
-       "skb->sk: sk->family [non fullsock field]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_1, offsetof(struct bpf_sock, family)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "skb->sk: sk->type [fullsock field]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_1, offsetof(struct bpf_sock, type)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid sock_common access",
-},
-{
-       "bpf_sk_fullsock(skb->sk): no !skb->sk check",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "type=sock_common_or_null expected=sock_common",
-},
-{
-       "sk_fullsock(skb->sk): no NULL check on ret",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, type)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid mem access 'sock_or_null'",
-},
-{
-       "sk_fullsock(skb->sk): sk->type [fullsock field]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, type)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->family [non fullsock field]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, family)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->state [narrow load]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, state)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_port [word load] (backward compatibility)",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, dst_port)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_port [half load]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_H, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, dst_port)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_port [half load] (invalid)",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_H, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, dst_port) + 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid sock access",
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_port [byte load]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_2, BPF_REG_0, offsetof(struct bpf_sock, dst_port)),
-       BPF_LDX_MEM(BPF_B, BPF_REG_2, BPF_REG_0, offsetof(struct bpf_sock, dst_port) + 1),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_port [byte load] (invalid)",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, dst_port) + 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid sock access",
-},
-{
-       "sk_fullsock(skb->sk): past sk->dst_port [half load] (invalid)",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_H, BPF_REG_0, BPF_REG_0, offsetofend(struct bpf_sock, dst_port)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid sock access",
-},
-{
-       "sk_fullsock(skb->sk): sk->dst_ip6 [load 2nd byte]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, dst_ip6[0]) + 1),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->type [narrow load]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, type)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): sk->protocol [narrow load]",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, protocol)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "sk_fullsock(skb->sk): beyond last field",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetofend(struct bpf_sock, rx_queue_mapping)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid sock access",
-},
-{
-       "bpf_tcp_sock(skb->sk): no !skb->sk check",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "type=sock_common_or_null expected=sock_common",
-},
-{
-       "bpf_tcp_sock(skb->sk): no NULL check on ret",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_tcp_sock, snd_cwnd)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid mem access 'tcp_sock_or_null'",
-},
-{
-       "bpf_tcp_sock(skb->sk): tp->snd_cwnd",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_tcp_sock, snd_cwnd)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "bpf_tcp_sock(skb->sk): tp->bytes_acked",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_tcp_sock, bytes_acked)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "bpf_tcp_sock(skb->sk): beyond last field",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_DW, BPF_REG_0, BPF_REG_0, offsetofend(struct bpf_tcp_sock, bytes_acked)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = REJECT,
-       .errstr = "invalid tcp_sock access",
-},
-{
-       "bpf_tcp_sock(bpf_sk_fullsock(skb->sk)): tp->snd_cwnd",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_tcp_sock, snd_cwnd)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_CGROUP_SKB,
-       .result = ACCEPT,
-},
-{
-       "bpf_sk_release(skb->sk)",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 1),
-       BPF_EMIT_CALL(BPF_FUNC_sk_release),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "R1 must be referenced when passed to release function",
-},
-{
-       "bpf_sk_release(bpf_sk_fullsock(skb->sk))",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_release),
-       BPF_MOV64_IMM(BPF_REG_0, 1),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "R1 must be referenced when passed to release function",
-},
-{
-       "bpf_sk_release(bpf_tcp_sock(skb->sk))",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_tcp_sock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_release),
-       BPF_MOV64_IMM(BPF_REG_0, 1),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "R1 must be referenced when passed to release function",
-},
-{
-       "sk_storage_get(map, skb->sk, NULL, 0): value == NULL",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_IMM(BPF_REG_4, 0),
-       BPF_MOV64_IMM(BPF_REG_3, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_storage_get),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_sk_storage_map = { 11 },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = ACCEPT,
-},
-{
-       "sk_storage_get(map, skb->sk, 1, 1): value == 1",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_IMM(BPF_REG_4, 1),
-       BPF_MOV64_IMM(BPF_REG_3, 1),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_storage_get),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_sk_storage_map = { 11 },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "R3 type=scalar expected=fp",
-},
-{
-       "sk_storage_get(map, skb->sk, &stack_value, 1): stack_value",
-       .insns = {
-       BPF_MOV64_IMM(BPF_REG_2, 0),
-       BPF_STX_MEM(BPF_DW, BPF_REG_10, BPF_REG_2, -8),
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_EMIT_CALL(BPF_FUNC_sk_fullsock),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_IMM(BPF_REG_4, 1),
-       BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_3, -8),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_0),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_storage_get),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_sk_storage_map = { 14 },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = ACCEPT,
-},
-{
-       "bpf_map_lookup_elem(smap, &key)",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_sk_storage_map = { 3 },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "cannot pass map_type 24 into func bpf_map_lookup_elem",
-},
-{
-       "bpf_map_lookup_elem(xskmap, &key); xs->queue_id",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -8, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_xdp_sock, queue_id)),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_xskmap = { 3 },
-       .prog_type = BPF_PROG_TYPE_XDP,
-       .result = ACCEPT,
-},
-{
-       "bpf_map_lookup_elem(sockmap, &key)",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockmap = { 3 },
-       .prog_type = BPF_PROG_TYPE_SK_SKB,
-       .result = REJECT,
-       .errstr = "Unreleased reference id=2 alloc_insn=5",
-},
-{
-       "bpf_map_lookup_elem(sockhash, &key)",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockhash = { 3 },
-       .prog_type = BPF_PROG_TYPE_SK_SKB,
-       .result = REJECT,
-       .errstr = "Unreleased reference id=2 alloc_insn=5",
-},
-{
-       "bpf_map_lookup_elem(sockmap, &key); sk->type [fullsock field]; bpf_sk_release(sk)",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, type)),
-       BPF_EMIT_CALL(BPF_FUNC_sk_release),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockmap = { 3 },
-       .prog_type = BPF_PROG_TYPE_SK_SKB,
-       .result = ACCEPT,
-},
-{
-       "bpf_map_lookup_elem(sockhash, &key); sk->type [fullsock field]; bpf_sk_release(sk)",
-       .insns = {
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-       BPF_LD_MAP_FD(BPF_REG_1, 0),
-       BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_0, 0, 1),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct bpf_sock, type)),
-       BPF_EMIT_CALL(BPF_FUNC_sk_release),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockhash = { 3 },
-       .prog_type = BPF_PROG_TYPE_SK_SKB,
-       .result = ACCEPT,
-},
-{
-       "bpf_sk_select_reuseport(ctx, reuseport_array, &key, flags)",
-       .insns = {
-       BPF_MOV64_IMM(BPF_REG_4, 0),
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_3, -4),
-       BPF_LD_MAP_FD(BPF_REG_2, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_select_reuseport),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_reuseport_array = { 4 },
-       .prog_type = BPF_PROG_TYPE_SK_REUSEPORT,
-       .result = ACCEPT,
-},
-{
-       "bpf_sk_select_reuseport(ctx, sockmap, &key, flags)",
-       .insns = {
-       BPF_MOV64_IMM(BPF_REG_4, 0),
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_3, -4),
-       BPF_LD_MAP_FD(BPF_REG_2, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_select_reuseport),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockmap = { 4 },
-       .prog_type = BPF_PROG_TYPE_SK_REUSEPORT,
-       .result = ACCEPT,
-},
-{
-       "bpf_sk_select_reuseport(ctx, sockhash, &key, flags)",
-       .insns = {
-       BPF_MOV64_IMM(BPF_REG_4, 0),
-       BPF_ST_MEM(BPF_W, BPF_REG_10, -4, 0),
-       BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
-       BPF_ALU64_IMM(BPF_ADD, BPF_REG_3, -4),
-       BPF_LD_MAP_FD(BPF_REG_2, 0),
-       BPF_EMIT_CALL(BPF_FUNC_sk_select_reuseport),
-       BPF_EXIT_INSN(),
-       },
-       .fixup_map_sockmap = { 4 },
-       .prog_type = BPF_PROG_TYPE_SK_REUSEPORT,
-       .result = ACCEPT,
-},
-{
-       "mark null check on return value of bpf_skc_to helpers",
-       .insns = {
-       BPF_LDX_MEM(BPF_DW, BPF_REG_1, BPF_REG_1, offsetof(struct __sk_buff, sk)),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_1, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_MOV64_REG(BPF_REG_6, BPF_REG_1),
-       BPF_EMIT_CALL(BPF_FUNC_skc_to_tcp_sock),
-       BPF_MOV64_REG(BPF_REG_7, BPF_REG_0),
-       BPF_MOV64_REG(BPF_REG_1, BPF_REG_6),
-       BPF_EMIT_CALL(BPF_FUNC_skc_to_tcp_request_sock),
-       BPF_MOV64_REG(BPF_REG_8, BPF_REG_0),
-       BPF_JMP_IMM(BPF_JNE, BPF_REG_8, 0, 2),
-       BPF_MOV64_IMM(BPF_REG_0, 0),
-       BPF_EXIT_INSN(),
-       BPF_LDX_MEM(BPF_B, BPF_REG_0, BPF_REG_7, 0),
-       BPF_EXIT_INSN(),
-       },
-       .prog_type = BPF_PROG_TYPE_SCHED_CLS,
-       .result = REJECT,
-       .errstr = "invalid mem access",
-       .result_unpriv = REJECT,
-       .errstr_unpriv = "unknown func",
-},