Merge branch 'tcp-smc-rendezvous'

Ursula Braun says:

====================
TCP experimental option for SMC rendezvous

SMC-capability is to be negotiated with a TCP experimental option.
As requested during code review of our previous approach using
netfilter hooks, here's a new version. It touches tcp-code in the
first patch and exploits the new tcp flag in the smc-code.

Changelog:

V3:
* move include for linux/unaligned/access_ok.h to tcp_input.c

V2:
* switch to current jump labels API
* remove static key checking in smc_set_capability()
  (comment from Eric Dumazet)
* use inet_request_sock parameter for smc_set_option_cond()
* smc_listen_work(): replace local variable lgr_lock_taken by new labels
                     and separate this change into a prerequisite first
                     patch
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/tcp.h

@@ -98,7 +98,8 @@ struct tcp_options_received {
 		tstamp_ok : 1,	/* TIMESTAMP seen on SYN packet		*/
 		dsack : 1,	/* D-SACK is scheduled			*/
 		wscale_ok : 1,	/* Wscale seen on SYN packet		*/
-		sack_ok : 4,	/* SACK seen on SYN packet		*/
+		sack_ok : 3,	/* SACK seen on SYN packet		*/
+		smc_ok : 1,	/* SMC seen on SYN packet		*/
 		snd_wscale : 4,	/* Window scaling received from sender	*/
 		rcv_wscale : 4;	/* Window scaling to send to receiver	*/
 	u8	num_sacks;	/* Number of SACK blocks		*/
@@ -110,6 +111,9 @@ static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
 {
 	rx_opt->tstamp_ok = rx_opt->sack_ok = 0;
 	rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
+#if IS_ENABLED(CONFIG_SMC)
+	rx_opt->smc_ok = 0;
+#endif
 }
 
 /* This is the max number of SACKS that we'll generate and process. It's safe
@@ -229,7 +233,8 @@ struct tcp_sock {
 		syn_fastopen_ch:1, /* Active TFO re-enabling probe */
 		syn_data_acked:1,/* data in SYN is acked by SYN-ACK */
 		save_syn:1,	/* Save headers of SYN packet */
-		is_cwnd_limited:1;/* forward progress limited by snd_cwnd? */
+		is_cwnd_limited:1,/* forward progress limited by snd_cwnd? */
+		syn_smc:1;	/* SYN includes SMC */
 	u32	tlp_high_seq;	/* snd_nxt at the time of TLP retransmit. */
 
 /* RTT measurement */

include/net/inet_sock.h

@@ -92,7 +92,8 @@ struct inet_request_sock {
 				wscale_ok  : 1,
 				ecn_ok	   : 1,
 				acked	   : 1,
-				no_srccheck: 1;
+				no_srccheck: 1,
+				smc_ok	   : 1;
 	kmemcheck_bitfield_end(flags);
 	u32                     ir_mark;
 	union {

include/net/tcp.h

@@ -191,6 +191,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
  * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
  */
 #define TCPOPT_FASTOPEN_MAGIC	0xF989
+#define TCPOPT_SMC_MAGIC	0xE2D4C3D9
 
 /*
  *     TCP option lengths
@@ -203,6 +204,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 #define TCPOLEN_MD5SIG         18
 #define TCPOLEN_FASTOPEN_BASE  2
 #define TCPOLEN_EXP_FASTOPEN_BASE  4
+#define TCPOLEN_EXP_SMC_BASE   6
 
 /* But this is what stacks really send out. */
 #define TCPOLEN_TSTAMP_ALIGNED		12
@@ -213,6 +215,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 #define TCPOLEN_SACK_PERBLOCK		8
 #define TCPOLEN_MD5SIG_ALIGNED		20
 #define TCPOLEN_MSS_ALIGNED		4
+#define TCPOLEN_EXP_SMC_BASE_ALIGNED	8
 
 /* Flags in tp->nonagle */
 #define TCP_NAGLE_OFF		1	/* Nagle's algo is disabled */
@@ -2108,4 +2111,8 @@ static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk)
 {
 	return (tcp_call_bpf(sk, BPF_SOCK_OPS_NEEDS_ECN) == 1);
 }
+
+#if IS_ENABLED(CONFIG_SMC)
+extern struct static_key_false tcp_have_smc;
+#endif
 #endif	/* _TCP_H */

net/ipv4/tcp.c

@@ -270,6 +270,7 @@
 #include <linux/time.h>
 #include <linux/slab.h>
 #include <linux/errqueue.h>
+#include <linux/static_key.h>
 
 #include <net/icmp.h>
 #include <net/inet_common.h>
@@ -302,6 +303,11 @@ EXPORT_SYMBOL(sysctl_tcp_wmem);
 atomic_long_t tcp_memory_allocated;	/* Current allocated memory. */
 EXPORT_SYMBOL(tcp_memory_allocated);
 
+#if IS_ENABLED(CONFIG_SMC)
+DEFINE_STATIC_KEY_FALSE(tcp_have_smc);
+EXPORT_SYMBOL(tcp_have_smc);
+#endif
+
 /*
  * Current number of TCP sockets.
  */

net/ipv4/tcp_input.c

@@ -76,6 +76,8 @@
 #include <asm/unaligned.h>
 #include <linux/errqueue.h>
 #include <trace/events/tcp.h>
+#include <linux/unaligned/access_ok.h>
+#include <linux/static_key.h>
 
 int sysctl_tcp_fack __read_mostly;
 int sysctl_tcp_max_reordering __read_mostly = 300;
@@ -3737,6 +3739,21 @@ static void tcp_parse_fastopen_option(int len, const unsigned char *cookie,
 	foc->exp = exp_opt;
 }
 
+static void smc_parse_options(const struct tcphdr *th,
+			      struct tcp_options_received *opt_rx,
+			      const unsigned char *ptr,
+			      int opsize)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		if (th->syn && !(opsize & 1) &&
+		    opsize >= TCPOLEN_EXP_SMC_BASE &&
+		    get_unaligned_be32(ptr) == TCPOPT_SMC_MAGIC)
+			opt_rx->smc_ok = 1;
+	}
+#endif
+}
+
 /* Look for tcp options. Normally only called on SYN and SYNACK packets.
  * But, this can also be called on packets in the established flow when
  * the fast version below fails.
@@ -3844,6 +3861,9 @@ void tcp_parse_options(const struct net *net,
 					tcp_parse_fastopen_option(opsize -
 						TCPOLEN_EXP_FASTOPEN_BASE,
 						ptr + 2, th->syn, foc, true);
+				else
+					smc_parse_options(th, opt_rx, ptr,
+							  opsize);
 				break;
 
 			}
@@ -5598,6 +5618,16 @@ static bool tcp_rcv_fastopen_synack(struct sock *sk, struct sk_buff *synack,
 	return false;
 }
 
+static void smc_check_reset_syn(struct tcp_sock *tp)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		if (tp->syn_smc && !tp->rx_opt.smc_ok)
+			tp->syn_smc = 0;
+	}
+#endif
+}
+
 static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 					 const struct tcphdr *th)
 {
@@ -5704,6 +5734,8 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		 * is initialized. */
 		tp->copied_seq = tp->rcv_nxt;
 
+		smc_check_reset_syn(tp);
+
 		smp_mb();
 
 		tcp_finish_connect(sk, skb);
@@ -6157,6 +6189,9 @@ static void tcp_openreq_init(struct request_sock *req,
 	ireq->ir_rmt_port = tcp_hdr(skb)->source;
 	ireq->ir_num = ntohs(tcp_hdr(skb)->dest);
 	ireq->ir_mark = inet_request_mark(sk, skb);
+#if IS_ENABLED(CONFIG_SMC)
+	ireq->smc_ok = rx_opt->smc_ok;
+#endif
 }
 
 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,

net/ipv4/tcp_minisocks.c

@@ -23,6 +23,7 @@
 #include <linux/slab.h>
 #include <linux/sysctl.h>
 #include <linux/workqueue.h>
+#include <linux/static_key.h>
 #include <net/tcp.h>
 #include <net/inet_common.h>
 #include <net/xfrm.h>
@@ -416,6 +417,21 @@ void tcp_ca_openreq_child(struct sock *sk, const struct dst_entry *dst)
 }
 EXPORT_SYMBOL_GPL(tcp_ca_openreq_child);
 
+static void smc_check_reset_syn_req(struct tcp_sock *oldtp,
+				    struct request_sock *req,
+				    struct tcp_sock *newtp)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	struct inet_request_sock *ireq;
+
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		ireq = inet_rsk(req);
+		if (oldtp->syn_smc && !ireq->smc_ok)
+			newtp->syn_smc = 0;
+	}
+#endif
+}
+
 /* This is not only more efficient than what we used to do, it eliminates
  * a lot of code duplication between IPv4/IPv6 SYN recv processing. -DaveM
  *
@@ -433,6 +449,9 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
 		struct tcp_request_sock *treq = tcp_rsk(req);
 		struct inet_connection_sock *newicsk = inet_csk(newsk);
 		struct tcp_sock *newtp = tcp_sk(newsk);
+		struct tcp_sock *oldtp = tcp_sk(sk);
+
+		smc_check_reset_syn_req(oldtp, req, newtp);
 
 		/* Now setup tcp_sock */
 		newtp->pred_flags = 0;

net/ipv4/tcp_output.c

@@ -41,6 +41,7 @@
 #include <linux/compiler.h>
 #include <linux/gfp.h>
 #include <linux/module.h>
+#include <linux/static_key.h>
 
 #include <trace/events/tcp.h>
 
@@ -422,6 +423,22 @@ static inline bool tcp_urg_mode(const struct tcp_sock *tp)
 #define OPTION_MD5		(1 << 2)
 #define OPTION_WSCALE		(1 << 3)
 #define OPTION_FAST_OPEN_COOKIE	(1 << 8)
+#define OPTION_SMC		(1 << 9)
+
+static void smc_options_write(__be32 *ptr, u16 *options)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		if (unlikely(OPTION_SMC & *options)) {
+			*ptr++ = htonl((TCPOPT_NOP  << 24) |
+				       (TCPOPT_NOP  << 16) |
+				       (TCPOPT_EXP <<  8) |
+				       (TCPOLEN_EXP_SMC_BASE));
+			*ptr++ = htonl(TCPOPT_SMC_MAGIC);
+		}
+	}
+#endif
+}
 
 struct tcp_out_options {
 	u16 options;		/* bit field of OPTION_* */
@@ -540,6 +557,41 @@ static void tcp_options_write(__be32 *ptr, struct tcp_sock *tp,
 		}
 		ptr += (len + 3) >> 2;
 	}
+
+	smc_options_write(ptr, &options);
+}
+
+static void smc_set_option(const struct tcp_sock *tp,
+			   struct tcp_out_options *opts,
+			   unsigned int *remaining)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		if (tp->syn_smc) {
+			if (*remaining >= TCPOLEN_EXP_SMC_BASE_ALIGNED) {
+				opts->options |= OPTION_SMC;
+				*remaining -= TCPOLEN_EXP_SMC_BASE_ALIGNED;
+			}
+		}
+	}
+#endif
+}
+
+static void smc_set_option_cond(const struct tcp_sock *tp,
+				const struct inet_request_sock *ireq,
+				struct tcp_out_options *opts,
+				unsigned int *remaining)
+{
+#if IS_ENABLED(CONFIG_SMC)
+	if (static_branch_unlikely(&tcp_have_smc)) {
+		if (tp->syn_smc && ireq->smc_ok) {
+			if (*remaining >= TCPOLEN_EXP_SMC_BASE_ALIGNED) {
+				opts->options |= OPTION_SMC;
+				*remaining -= TCPOLEN_EXP_SMC_BASE_ALIGNED;
+			}
+		}
+	}
+#endif
 }
 
 /* Compute TCP options for SYN packets. This is not the final
@@ -607,11 +659,14 @@ static unsigned int tcp_syn_options(struct sock *sk, struct sk_buff *skb,
 		}
 	}
 
+	smc_set_option(tp, opts, &remaining);
+
 	return MAX_TCP_OPTION_SPACE - remaining;
 }
 
 /* Set up TCP options for SYN-ACKs. */
-static unsigned int tcp_synack_options(struct request_sock *req,
+static unsigned int tcp_synack_options(const struct sock *sk,
+				       struct request_sock *req,
 				       unsigned int mss, struct sk_buff *skb,
 				       struct tcp_out_options *opts,
 				       const struct tcp_md5sig_key *md5,
@@ -667,6 +722,8 @@ static unsigned int tcp_synack_options(struct request_sock *req,
 		}
 	}
 
+	smc_set_option_cond(tcp_sk(sk), ireq, opts, &remaining);
+
 	return MAX_TCP_OPTION_SPACE - remaining;
 }
 
@@ -3195,8 +3252,8 @@ struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,
 	md5 = tcp_rsk(req)->af_specific->req_md5_lookup(sk, req_to_sk(req));
 #endif
 	skb_set_hash(skb, tcp_rsk(req)->txhash, PKT_HASH_TYPE_L4);
-	tcp_header_size = tcp_synack_options(req, mss, skb, &opts, md5, foc) +
-			  sizeof(*th);
+	tcp_header_size = tcp_synack_options(sk, req, mss, skb, &opts, md5,
+					     foc) + sizeof(*th);
 
 	skb_push(skb, tcp_header_size);
 	skb_reset_transport_header(skb);

net/smc/af_smc.c

@@ -390,6 +390,12 @@ static int smc_connect_rdma(struct smc_sock *smc)
 	int rc = 0;
 	u8 ibport;
 
+	if (!tcp_sk(smc->clcsock->sk)->syn_smc) {
+		/* peer has not signalled SMC-capability */
+		smc->use_fallback = true;
+		goto out_connected;
+	}
+
 	/* IPSec connections opt out of SMC-R optimizations */
 	if (using_ipsec(smc)) {
 		reason_code = SMC_CLC_DECL_IPSEC;
@@ -555,6 +561,7 @@ static int smc_connect(struct socket *sock, struct sockaddr *addr,
 	}
 
 	smc_copy_sock_settings_to_clc(smc);
+	tcp_sk(smc->clcsock->sk)->syn_smc = 1;
 	rc = kernel_connect(smc->clcsock, addr, alen, flags);
 	if (rc)
 		goto out;
@@ -759,6 +766,12 @@ static void smc_listen_work(struct work_struct *work)
 	u8 prefix_len;
 	u8 ibport;
 
+	/* check if peer is smc capable */
+	if (!tcp_sk(newclcsock->sk)->syn_smc) {
+		new_smc->use_fallback = true;
+		goto out_connected;
+	}
+
 	/* do inband token exchange -
 	 *wait for and receive SMC Proposal CLC message
 	 */
@@ -808,7 +821,7 @@ static void smc_listen_work(struct work_struct *work)
 		rc = local_contact;
 		if (rc == -ENOMEM)
 			reason_code = SMC_CLC_DECL_MEM;/* insufficient memory*/
-		goto decline_rdma;
+		goto decline_rdma_unlock;
 	}
 	link = &new_smc->conn.lgr->lnk[SMC_SINGLE_LINK];
 
@@ -816,7 +829,7 @@ static void smc_listen_work(struct work_struct *work)
 	rc = smc_buf_create(new_smc);
 	if (rc) {
 		reason_code = SMC_CLC_DECL_MEM;
-		goto decline_rdma;
+		goto decline_rdma_unlock;
 	}
 
 	smc_close_init(new_smc);
@@ -831,7 +844,7 @@ static void smc_listen_work(struct work_struct *work)
 					     buf_desc->mr_rx[SMC_SINGLE_LINK]);
 			if (rc) {
 				reason_code = SMC_CLC_DECL_INTERR;
-				goto decline_rdma;
+				goto decline_rdma_unlock;
 			}
 		}
 	}
@@ -839,15 +852,15 @@ static void smc_listen_work(struct work_struct *work)
 
 	rc = smc_clc_send_accept(new_smc, local_contact);
 	if (rc)
-		goto out_err;
+		goto out_err_unlock;
 
 	/* receive SMC Confirm CLC message */
 	reason_code = smc_clc_wait_msg(new_smc, &cclc, sizeof(cclc),
 				       SMC_CLC_CONFIRM);
 	if (reason_code < 0)
-		goto out_err;
+		goto out_err_unlock;
 	if (reason_code > 0)
-		goto decline_rdma;
+		goto decline_rdma_unlock;
 	smc_conn_save_peer_info(new_smc, &cclc);
 	if (local_contact == SMC_FIRST_CONTACT)
 		smc_link_save_peer_info(link, &cclc);
@@ -855,34 +868,34 @@ static void smc_listen_work(struct work_struct *work)
 	rc = smc_rmb_rtoken_handling(&new_smc->conn, &cclc);
 	if (rc) {
 		reason_code = SMC_CLC_DECL_INTERR;
-		goto decline_rdma;
+		goto decline_rdma_unlock;
 	}
 
 	if (local_contact == SMC_FIRST_CONTACT) {
 		rc = smc_ib_ready_link(link);
 		if (rc) {
 			reason_code = SMC_CLC_DECL_INTERR;
-			goto decline_rdma;
+			goto decline_rdma_unlock;
 		}
 		/* QP confirmation over RoCE fabric */
 		reason_code = smc_serv_conf_first_link(new_smc);
 		if (reason_code < 0) {
 			/* peer is not aware of a problem */
 			rc = reason_code;
-			goto out_err;
+			goto out_err_unlock;
 		}
 		if (reason_code > 0)
-			goto decline_rdma;
+			goto decline_rdma_unlock;
 	}
 
 	smc_tx_init(new_smc);
+	mutex_unlock(&smc_create_lgr_pending);
 
 out_connected:
 	sk_refcnt_debug_inc(newsmcsk);
 	if (newsmcsk->sk_state == SMC_INIT)
 		newsmcsk->sk_state = SMC_ACTIVE;
 enqueue:
-	mutex_unlock(&smc_create_lgr_pending);
 	lock_sock_nested(&lsmc->sk, SINGLE_DEPTH_NESTING);
 	if (lsmc->sk.sk_state == SMC_LISTEN) {
 		smc_accept_enqueue(&lsmc->sk, newsmcsk);
@@ -896,6 +909,8 @@ static void smc_listen_work(struct work_struct *work)
 	sock_put(&lsmc->sk); /* sock_hold in smc_tcp_listen_work */
 	return;
 
+decline_rdma_unlock:
+	mutex_unlock(&smc_create_lgr_pending);
 decline_rdma:
 	/* RDMA setup failed, switch back to TCP */
 	smc_conn_free(&new_smc->conn);
@@ -907,6 +922,8 @@ static void smc_listen_work(struct work_struct *work)
 	}
 	goto out_connected;
 
+out_err_unlock:
+	mutex_unlock(&smc_create_lgr_pending);
 out_err:
 	newsmcsk->sk_state = SMC_CLOSED;
 	smc_conn_free(&new_smc->conn);
@@ -963,6 +980,7 @@ static int smc_listen(struct socket *sock, int backlog)
 	 * them to the clc socket -- copy smc socket options to clc socket
 	 */
 	smc_copy_sock_settings_to_clc(smc);
+	tcp_sk(smc->clcsock->sk)->syn_smc = 1;
 
 	rc = kernel_listen(smc->clcsock, backlog);
 	if (rc)
@@ -1405,6 +1423,7 @@ static int __init smc_init(void)
 		goto out_sock;
 	}
 
+	static_branch_enable(&tcp_have_smc);
 	return 0;
 
 out_sock:
@@ -1429,6 +1448,7 @@ static void __exit smc_exit(void)
 		list_del_init(&lgr->list);
 		smc_lgr_free(lgr); /* free link group */
 	}
+	static_branch_disable(&tcp_have_smc);
 	smc_ib_unregister_client();
 	sock_unregister(PF_SMC);
 	proto_unregister(&smc_proto);