Merge branch 'Do not limit cb_flags when creating child sk'

Martin KaFai says: ==================== This set fixes an issue that the bpf_skops_init_child() unnecessarily limited the child sk from inheriting all bpf_sock_ops_cb_flags of the listen sk. It also adds a test to check that. ==================== Tested-by: Stanislav Fomichev <sdf@google.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org>

Author: Alexei Starovoitov

include/net/tcp.h

@@ -2228,34 +2228,6 @@ int __tcp_bpf_recvmsg(struct sock *sk, struct sk_psock *psock,
 #endif /* CONFIG_NET_SOCK_MSG */
 
 #ifdef CONFIG_CGROUP_BPF
-/* Copy the listen sk's HDR_OPT_CB flags to its child.
- *
- * During 3-Way-HandShake, the synack is usually sent from
- * the listen sk with the HDR_OPT_CB flags set so that
- * bpf-prog will be called to write the BPF hdr option.
- *
- * In fastopen, the child sk is used to send synack instead
- * of the listen sk. Thus, inheriting the HDR_OPT_CB flags
- * from the listen sk gives the bpf-prog a chance to write
- * BPF hdr option in the synack pkt during fastopen.
- *
- * Both fastopen and non-fastopen child will inherit the
- * HDR_OPT_CB flags to keep the bpf-prog having a consistent
- * behavior when deciding to clear this cb flags (or not)
- * during the PASSIVE_ESTABLISHED_CB.
- *
- * In the future, other cb flags could be inherited here also.
- */
-static inline void bpf_skops_init_child(const struct sock *sk,
-struct sock *child)
-{
-tcp_sk(child)->bpf_sock_ops_cb_flags =
-tcp_sk(sk)->bpf_sock_ops_cb_flags &
-(BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG |
- BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG |
- BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG);
-}
-
 static inline void bpf_skops_init_skb(struct bpf_sock_ops_kern *skops,
  struct sk_buff *skb,
  unsigned int end_offset)
@@ -2264,11 +2236,6 @@ static inline void bpf_skops_init_skb(struct bpf_sock_ops_kern *skops,
 skops->skb_data_end = skb->data + end_offset;
 }
 #else
-static inline void bpf_skops_init_child(const struct sock *sk,
-struct sock *child)
-{
-}
-
 static inline void bpf_skops_init_skb(struct bpf_sock_ops_kern *skops,
  struct sk_buff *skb,
  unsigned int end_offset)

net/ipv4/tcp_minisocks.c

@@ -548,7 +548,6 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
 newtp->fastopen_req = NULL;
 RCU_INIT_POINTER(newtp->fastopen_rsk, NULL);
 
-bpf_skops_init_child(sk, newsk);
 tcp_bpf_clone(sk, newsk);
 
 __TCP_INC_STATS(sock_net(sk), TCP_MIB_PASSIVEOPENS);

tools/testing/selftests/bpf/prog_tests/tcp_hdr_options.c

@@ -264,9 +264,19 @@ static int check_error_linum(const struct sk_fds *sk_fds)
 
 static void check_hdr_and_close_fds(struct sk_fds *sk_fds)
 {
+const __u32 expected_inherit_cb_flags =
+BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG |
+BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG |
+BPF_SOCK_OPS_STATE_CB_FLAG;
+
 if (sk_fds_shutdown(sk_fds))
 goto check_linum;
 
+if (CHECK(expected_inherit_cb_flags != skel->bss->inherit_cb_flags,
+ "Unexpected inherit_cb_flags", "0x%x != 0x%x\n",
+ skel->bss->inherit_cb_flags, expected_inherit_cb_flags))
+goto check_linum;
+
 if (check_hdr_stg(&exp_passive_hdr_stg, sk_fds->passive_fd,
  "passive_hdr_stg"))
 goto check_linum;
@@ -321,6 +331,8 @@ static void reset_test(void)
 memset(&skel->bss->active_estab_in, 0, optsize);
 memset(&skel->bss->active_fin_in, 0, optsize);
 
+skel->bss->inherit_cb_flags = 0;
+
 skel->data->test_kind = TCPOPT_EXP;
 skel->data->test_magic = 0xeB9F;
 

tools/testing/selftests/bpf/progs/test_misc_tcp_hdr_options.c

@@ -304,10 +304,10 @@ int misc_estab(struct bpf_sock_ops *skops)
 passive_lport_n = __bpf_htons(passive_lport_h);
 bpf_setsockopt(skops, SOL_TCP, TCP_SAVE_SYN,
  &true_val, sizeof(true_val));
-set_hdr_cb_flags(skops);
+set_hdr_cb_flags(skops, 0);
 break;
 case BPF_SOCK_OPS_TCP_CONNECT_CB:
-set_hdr_cb_flags(skops);
+set_hdr_cb_flags(skops, 0);
 break;
 case BPF_SOCK_OPS_PARSE_HDR_OPT_CB:
 return handle_parse_hdr(skops);

tools/testing/selftests/bpf/progs/test_tcp_hdr_options.c

@@ -21,6 +21,7 @@
 
 __u8 test_kind = TCPOPT_EXP;
 __u16 test_magic = 0xeB9F;
+__u32 inherit_cb_flags = 0;
 
 struct bpf_test_option passive_synack_out = {};
 struct bpf_test_option passive_fin_out= {};
@@ -467,6 +468,8 @@ static int handle_passive_estab(struct bpf_sock_ops *skops)
 struct tcphdr *th;
 int err;
 
+inherit_cb_flags = skops->bpf_sock_ops_cb_flags;
+
 err = load_option(skops, &passive_estab_in, true);
 if (err == -ENOENT) {
 /* saved_syn is not found. It was in syncookie mode.
@@ -600,10 +603,10 @@ int estab(struct bpf_sock_ops *skops)
 case BPF_SOCK_OPS_TCP_LISTEN_CB:
 bpf_setsockopt(skops, SOL_TCP, TCP_SAVE_SYN,
  &true_val, sizeof(true_val));
-set_hdr_cb_flags(skops);
+set_hdr_cb_flags(skops, BPF_SOCK_OPS_STATE_CB_FLAG);
 break;
 case BPF_SOCK_OPS_TCP_CONNECT_CB:
-set_hdr_cb_flags(skops);
+set_hdr_cb_flags(skops, 0);
 break;
 case BPF_SOCK_OPS_PARSE_HDR_OPT_CB:
 return handle_parse_hdr(skops);

tools/testing/selftests/bpf/test_tcp_hdr_options.h

@@ -110,12 +110,13 @@ static inline void clear_hdr_cb_flags(struct bpf_sock_ops *skops)
  BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG));
 }
 
-static inline void set_hdr_cb_flags(struct bpf_sock_ops *skops)
+static inline void set_hdr_cb_flags(struct bpf_sock_ops *skops, __u32 extra)
 {
 bpf_sock_ops_cb_flags_set(skops,
  skops->bpf_sock_ops_cb_flags |
  BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG |
- BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG);
+ BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG |
+ extra);
 }
 static inline void
 clear_parse_all_hdr_cb_flags(struct bpf_sock_ops *skops)