Merge branch 'net-smc-add-event-based-framework-for-LLC-msgs'

Karsten Graul says:

====================
net/smc: add event-based framework for LLC msgs

These patches are the next step towards SMC-R link failover support. They add
a new framework to handle Link Layer Control (LLC) messages and adapt the
existing code to use the new framework.
====================
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

net/smc/af_smc.c

@@ -382,22 +382,24 @@ static int smcr_lgr_reg_rmbs(struct smc_link_group *lgr,
 static int smcr_clnt_conf_first_link(struct smc_sock *smc)
 {
 	struct smc_link *link = smc->conn.lnk;
-	int rest;
+	struct smc_llc_qentry *qentry;
 	int rc;
 
+	link->lgr->type = SMC_LGR_SINGLE;
+
 	/* receive CONFIRM LINK request from server over RoCE fabric */
-	rest = wait_for_completion_interruptible_timeout(
-		&link->llc_confirm,
-		SMC_LLC_WAIT_FIRST_TIME);
-	if (rest <= 0) {
+	qentry = smc_llc_wait(link->lgr, NULL, SMC_LLC_WAIT_TIME,
+			      SMC_LLC_CONFIRM_LINK);
+	if (!qentry) {
 		struct smc_clc_msg_decline dclc;
 
 		rc = smc_clc_wait_msg(smc, &dclc, sizeof(dclc),
 				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
 		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_CL : rc;
 	}
-
-	if (link->llc_confirm_rc)
+	rc = smc_llc_eval_conf_link(qentry, SMC_LLC_REQ);
+	smc_llc_flow_qentry_del(&link->lgr->llc_flow_lcl);
+	if (rc)
 		return SMC_CLC_DECL_RMBE_EC;
 
 	rc = smc_ib_modify_qp_rts(link);
@@ -409,31 +411,30 @@ static int smcr_clnt_conf_first_link(struct smc_sock *smc)
 	if (smcr_link_reg_rmb(link, smc->conn.rmb_desc, false))
 		return SMC_CLC_DECL_ERR_REGRMB;
 
+	/* confirm_rkey is implicit on 1st contact */
+	smc->conn.rmb_desc->is_conf_rkey = true;
+
 	/* send CONFIRM LINK response over RoCE fabric */
 	rc = smc_llc_send_confirm_link(link, SMC_LLC_RESP);
 	if (rc < 0)
 		return SMC_CLC_DECL_TIMEOUT_CL;
 
-	/* receive ADD LINK request from server over RoCE fabric */
-	rest = wait_for_completion_interruptible_timeout(&link->llc_add,
-							 SMC_LLC_WAIT_TIME);
-	if (rest <= 0) {
+	smc_llc_link_active(link);
+
+	/* optional 2nd link, receive ADD LINK request from server */
+	qentry = smc_llc_wait(link->lgr, NULL, SMC_LLC_WAIT_TIME,
+			      SMC_LLC_ADD_LINK);
+	if (!qentry) {
 		struct smc_clc_msg_decline dclc;
 
 		rc = smc_clc_wait_msg(smc, &dclc, sizeof(dclc),
 				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
-		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_AL : rc;
+		if (rc == -EAGAIN)
+			rc = 0; /* no DECLINE received, go with one link */
+		return rc;
 	}
-
-	/* send add link reject message, only one link supported for now */
-	rc = smc_llc_send_add_link(link,
-				   link->smcibdev->mac[link->ibport - 1],
-				   link->gid, SMC_LLC_RESP);
-	if (rc < 0)
-		return SMC_CLC_DECL_TIMEOUT_AL;
-
-	smc_llc_link_active(link);
-
+	smc_llc_flow_qentry_clr(&link->lgr->llc_flow_lcl);
+	/* tbd: call smc_llc_cli_add_link(link, qentry); */
 	return 0;
 }
 
@@ -613,8 +614,8 @@ static int smc_connect_rdma(struct smc_sock *smc,
 			    struct smc_clc_msg_accept_confirm *aclc,
 			    struct smc_init_info *ini)
 {
+	int i, reason_code = 0;
 	struct smc_link *link;
-	int reason_code = 0;
 
 	ini->is_smcd = false;
 	ini->ib_lcl = &aclc->lcl;
@@ -627,10 +628,28 @@ static int smc_connect_rdma(struct smc_sock *smc,
 		mutex_unlock(&smc_client_lgr_pending);
 		return reason_code;
 	}
-	link = smc->conn.lnk;
 
 	smc_conn_save_peer_info(smc, aclc);
 
+	if (ini->cln_first_contact == SMC_FIRST_CONTACT) {
+		link = smc->conn.lnk;
+	} else {
+		/* set link that was assigned by server */
+		link = NULL;
+		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
+			struct smc_link *l = &smc->conn.lgr->lnk[i];
+
+			if (l->peer_qpn == ntoh24(aclc->qpn)) {
+				link = l;
+				break;
+			}
+		}
+		if (!link)
+			return smc_connect_abort(smc, SMC_CLC_DECL_NOSRVLINK,
+						 ini->cln_first_contact);
+		smc->conn.lnk = link;
+	}
+
 	/* create send buffer and rmb */
 	if (smc_buf_create(smc, false))
 		return smc_connect_abort(smc, SMC_CLC_DECL_MEM,
@@ -666,7 +685,9 @@ static int smc_connect_rdma(struct smc_sock *smc,
 
 	if (ini->cln_first_contact == SMC_FIRST_CONTACT) {
 		/* QP confirmation over RoCE fabric */
+		smc_llc_flow_initiate(link->lgr, SMC_LLC_FLOW_ADD_LINK);
 		reason_code = smcr_clnt_conf_first_link(smc);
+		smc_llc_flow_stop(link->lgr, &link->lgr->llc_flow_lcl);
 		if (reason_code)
 			return smc_connect_abort(smc, reason_code,
 						 ini->cln_first_contact);
@@ -1019,9 +1040,11 @@ void smc_close_non_accepted(struct sock *sk)
 static int smcr_serv_conf_first_link(struct smc_sock *smc)
 {
 	struct smc_link *link = smc->conn.lnk;
-	int rest;
+	struct smc_llc_qentry *qentry;
 	int rc;
 
+	link->lgr->type = SMC_LGR_SINGLE;
+
 	if (smcr_link_reg_rmb(link, smc->conn.rmb_desc, false))
 		return SMC_CLC_DECL_ERR_REGRMB;
 
@@ -1031,40 +1054,27 @@ static int smcr_serv_conf_first_link(struct smc_sock *smc)
 		return SMC_CLC_DECL_TIMEOUT_CL;
 
 	/* receive CONFIRM LINK response from client over the RoCE fabric */
-	rest = wait_for_completion_interruptible_timeout(
-		&link->llc_confirm_resp,
-		SMC_LLC_WAIT_FIRST_TIME);
-	if (rest <= 0) {
+	qentry = smc_llc_wait(link->lgr, link, SMC_LLC_WAIT_TIME,
+			      SMC_LLC_CONFIRM_LINK);
+	if (!qentry) {
 		struct smc_clc_msg_decline dclc;
 
 		rc = smc_clc_wait_msg(smc, &dclc, sizeof(dclc),
 				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
 		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_CL : rc;
 	}
-
-	if (link->llc_confirm_resp_rc)
+	rc = smc_llc_eval_conf_link(qentry, SMC_LLC_RESP);
+	smc_llc_flow_qentry_del(&link->lgr->llc_flow_lcl);
+	if (rc)
 		return SMC_CLC_DECL_RMBE_EC;
 
-	/* send ADD LINK request to client over the RoCE fabric */
-	rc = smc_llc_send_add_link(link,
-				   link->smcibdev->mac[link->ibport - 1],
-				   link->gid, SMC_LLC_REQ);
-	if (rc < 0)
-		return SMC_CLC_DECL_TIMEOUT_AL;
-
-	/* receive ADD LINK response from client over the RoCE fabric */
-	rest = wait_for_completion_interruptible_timeout(&link->llc_add_resp,
-							 SMC_LLC_WAIT_TIME);
-	if (rest <= 0) {
-		struct smc_clc_msg_decline dclc;
-
-		rc = smc_clc_wait_msg(smc, &dclc, sizeof(dclc),
-				      SMC_CLC_DECLINE, CLC_WAIT_TIME_SHORT);
-		return rc == -EAGAIN ? SMC_CLC_DECL_TIMEOUT_AL : rc;
-	}
+	/* confirm_rkey is implicit on 1st contact */
+	smc->conn.rmb_desc->is_conf_rkey = true;
 
 	smc_llc_link_active(link);
 
+	/* initial contact - try to establish second link */
+	/* tbd: call smc_llc_srv_add_link(link); */
 	return 0;
 }
 
@@ -1240,7 +1250,9 @@ static int smc_listen_rdma_finish(struct smc_sock *new_smc,
 			goto decline;
 		}
 		/* QP confirmation over RoCE fabric */
+		smc_llc_flow_initiate(link->lgr, SMC_LLC_FLOW_ADD_LINK);
 		reason_code = smcr_serv_conf_first_link(new_smc);
+		smc_llc_flow_stop(link->lgr, &link->lgr->llc_flow_lcl);
 		if (reason_code)
 			goto decline;
 	}

net/smc/smc_clc.h

@@ -45,6 +45,7 @@
 #define SMC_CLC_DECL_GETVLANERR	0x03080000  /* err to get vlan id of ip device*/
 #define SMC_CLC_DECL_ISMVLANERR	0x03090000  /* err to reg vlan id on ism dev  */
 #define SMC_CLC_DECL_NOACTLINK	0x030a0000  /* no active smc-r link in lgr    */
+#define SMC_CLC_DECL_NOSRVLINK	0x030b0000  /* SMC-R link from srv not found  */
 #define SMC_CLC_DECL_SYNCERR	0x04000000  /* synchronization error          */
 #define SMC_CLC_DECL_PEERDECL	0x05000000  /* peer declined during handshake */
 #define SMC_CLC_DECL_INTERR	0x09990000  /* internal error		      */

net/smc/smc_core.c

@@ -200,7 +200,6 @@ static int smcr_link_send_delete(struct smc_link *lnk, bool orderly)
 {
 	if (lnk->state == SMC_LNK_ACTIVE &&
 	    !smc_llc_send_delete_link(lnk, SMC_LLC_REQ, orderly)) {
-		smc_llc_link_deleting(lnk);
 		return 0;
 	}
 	return -ENOTCONN;
@@ -263,6 +262,7 @@ static void smc_lgr_free_work(struct work_struct *work)
 			if (smc_link_usable(lnk))
 				lnk->state = SMC_LNK_INACTIVE;
 		}
+		wake_up_interruptible_all(&lgr->llc_waiter);
 	}
 	smc_lgr_free(lgr);
 }
@@ -445,13 +445,11 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini)
 }
 
 static void smcr_buf_unuse(struct smc_buf_desc *rmb_desc,
-			   struct smc_link *lnk)
+			   struct smc_link_group *lgr)
 {
-	struct smc_link_group *lgr = lnk->lgr;
-
 	if (rmb_desc->is_conf_rkey && !list_empty(&lgr->list)) {
 		/* unregister rmb with peer */
-		smc_llc_do_delete_rkey(lnk, rmb_desc);
+		smc_llc_do_delete_rkey(lgr, rmb_desc);
 		rmb_desc->is_conf_rkey = false;
 	}
 	if (rmb_desc->is_reg_err) {
@@ -474,7 +472,7 @@ static void smc_buf_unuse(struct smc_connection *conn,
 	if (conn->rmb_desc && lgr->is_smcd)
 		conn->rmb_desc->used = 0;
 	else if (conn->rmb_desc)
-		smcr_buf_unuse(conn->rmb_desc, conn->lnk);
+		smcr_buf_unuse(conn->rmb_desc, lgr);
 }
 
 /* remove a finished connection from its link group */
@@ -696,6 +694,7 @@ static void smc_lgr_cleanup(struct smc_link_group *lgr)
 			if (smc_link_usable(lnk))
 				lnk->state = SMC_LNK_INACTIVE;
 		}
+		wake_up_interruptible_all(&lgr->llc_waiter);
 	}
 }
 
@@ -767,8 +766,7 @@ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport)
 			continue;
 		/* tbd - terminate only when no more links are active */
 		for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) {
-			if (!smc_link_usable(&lgr->lnk[i]) ||
-			    lgr->lnk[i].state == SMC_LNK_DELETING)
+			if (!smc_link_usable(&lgr->lnk[i]))
 				continue;
 			if (lgr->lnk[i].smcibdev == smcibdev &&
 			    lgr->lnk[i].ibport == ibport) {
@@ -1167,7 +1165,6 @@ static int smcr_buf_map_usable_links(struct smc_link_group *lgr,
 		if (!smc_link_usable(lnk))
 			continue;
 		if (smcr_buf_map_link(buf_desc, is_rmb, lnk)) {
-			smcr_buf_unuse(buf_desc, lnk);
 			rc = -ENOMEM;
 			goto out;
 		}
@@ -1273,6 +1270,7 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_smcd, bool is_rmb)
 
 	if (!is_smcd) {
 		if (smcr_buf_map_usable_links(lgr, buf_desc, is_rmb)) {
+			smcr_buf_unuse(buf_desc, lgr);
 			return -ENOMEM;
 		}
 	}
@@ -1368,6 +1366,53 @@ static inline int smc_rmb_reserve_rtoken_idx(struct smc_link_group *lgr)
 	return -ENOSPC;
 }
 
+static int smc_rtoken_find_by_link(struct smc_link_group *lgr, int lnk_idx,
+				   u32 rkey)
+{
+	int i;
+
+	for (i = 0; i < SMC_RMBS_PER_LGR_MAX; i++) {
+		if (test_bit(i, lgr->rtokens_used_mask) &&
+		    lgr->rtokens[i][lnk_idx].rkey == rkey)
+			return i;
+	}
+	return -ENOENT;
+}
+
+/* set rtoken for a new link to an existing rmb */
+void smc_rtoken_set(struct smc_link_group *lgr, int link_idx, int link_idx_new,
+		    __be32 nw_rkey_known, __be64 nw_vaddr, __be32 nw_rkey)
+{
+	int rtok_idx;
+
+	rtok_idx = smc_rtoken_find_by_link(lgr, link_idx, ntohl(nw_rkey_known));
+	if (rtok_idx == -ENOENT)
+		return;
+	lgr->rtokens[rtok_idx][link_idx_new].rkey = ntohl(nw_rkey);
+	lgr->rtokens[rtok_idx][link_idx_new].dma_addr = be64_to_cpu(nw_vaddr);
+}
+
+/* set rtoken for a new link whose link_id is given */
+void smc_rtoken_set2(struct smc_link_group *lgr, int rtok_idx, int link_id,
+		     __be64 nw_vaddr, __be32 nw_rkey)
+{
+	u64 dma_addr = be64_to_cpu(nw_vaddr);
+	u32 rkey = ntohl(nw_rkey);
+	bool found = false;
+	int link_idx;
+
+	for (link_idx = 0; link_idx < SMC_LINKS_PER_LGR_MAX; link_idx++) {
+		if (lgr->lnk[link_idx].link_id == link_id) {
+			found = true;
+			break;
+		}
+	}
+	if (!found)
+		return;
+	lgr->rtokens[rtok_idx][link_idx].rkey = rkey;
+	lgr->rtokens[rtok_idx][link_idx].dma_addr = dma_addr;
+}
+
 /* add a new rtoken from peer */
 int smc_rtoken_add(struct smc_link *lnk, __be64 nw_vaddr, __be32 nw_rkey)
 {

net/smc/smc_core.h

@@ -36,7 +36,6 @@ enum smc_link_state {			/* possible states of a link */
 	SMC_LNK_INACTIVE,	/* link is inactive */
 	SMC_LNK_ACTIVATING,	/* link is being activated */
 	SMC_LNK_ACTIVE,		/* link is active */
-	SMC_LNK_DELETING,	/* link is being deleted */
 };
 
 #define SMC_WR_BUF_SIZE		48	/* size of work request buffer */
@@ -120,20 +119,9 @@ struct smc_link {
 	struct smc_link_group	*lgr;		/* parent link group */
 
 	enum smc_link_state	state;		/* state of link */
-	struct completion	llc_confirm;	/* wait for rx of conf link */
-	struct completion	llc_confirm_resp; /* wait 4 rx of cnf lnk rsp */
-	int			llc_confirm_rc; /* rc from confirm link msg */
-	int			llc_confirm_resp_rc; /* rc from conf_resp msg */
-	struct completion	llc_add;	/* wait for rx of add link */
-	struct completion	llc_add_resp;	/* wait for rx of add link rsp*/
 	struct delayed_work	llc_testlink_wrk; /* testlink worker */
 	struct completion	llc_testlink_resp; /* wait for rx of testlink */
 	int			llc_testlink_time; /* testlink interval */
-	struct completion	llc_confirm_rkey_resp; /* w4 rx of cnf rkey */
-	int			llc_confirm_rkey_resp_rc; /* rc from cnf rkey */
-	struct completion	llc_delete_rkey_resp; /* w4 rx of del rkey */
-	int			llc_delete_rkey_resp_rc; /* rc from del rkey */
-	struct mutex		llc_delete_rkey_mutex; /* serialize usage */
 };
 
 /* For now we just allow one parallel link per link group. The SMC protocol
@@ -197,6 +185,28 @@ struct smc_rtoken {				/* address/key of remote RMB */
 
 struct smcd_dev;
 
+enum smc_lgr_type {				/* redundancy state of lgr */
+	SMC_LGR_NONE,			/* no active links, lgr to be deleted */
+	SMC_LGR_SINGLE,			/* 1 active RNIC on each peer */
+	SMC_LGR_SYMMETRIC,		/* 2 active RNICs on each peer */
+	SMC_LGR_ASYMMETRIC_PEER,	/* local has 2, peer 1 active RNICs */
+	SMC_LGR_ASYMMETRIC_LOCAL,	/* local has 1, peer 2 active RNICs */
+};
+
+enum smc_llc_flowtype {
+	SMC_LLC_FLOW_NONE	= 0,
+	SMC_LLC_FLOW_ADD_LINK	= 2,
+	SMC_LLC_FLOW_DEL_LINK	= 4,
+	SMC_LLC_FLOW_RKEY	= 6,
+};
+
+struct smc_llc_qentry;
+
+struct smc_llc_flow {
+	enum smc_llc_flowtype type;
+	struct smc_llc_qentry *qentry;
+};
+
 struct smc_link_group {
 	struct list_head	list;
 	struct rb_root		conns_all;	/* connection tree */
@@ -232,12 +242,24 @@ struct smc_link_group {
 			DECLARE_BITMAP(rtokens_used_mask, SMC_RMBS_PER_LGR_MAX);
 						/* used rtoken elements */
 			u8			next_link_id;
+			enum smc_lgr_type	type;
+						/* redundancy state */
 			struct list_head	llc_event_q;
 						/* queue for llc events */
 			spinlock_t		llc_event_q_lock;
 						/* protects llc_event_q */
 			struct work_struct	llc_event_work;
 						/* llc event worker */
+			wait_queue_head_t	llc_waiter;
+						/* w4 next llc event */
+			struct smc_llc_flow	llc_flow_lcl;
+						/* llc local control field */
+			struct smc_llc_flow	llc_flow_rmt;
+						/* llc remote control field */
+			struct smc_llc_qentry	*delayed_event;
+						/* arrived when flow active */
+			spinlock_t		llc_flow_lock;
+						/* protects llc flow */
 			int			llc_testlink_time;
 						/* link keep alive time */
 		};
@@ -329,6 +351,10 @@ int smc_rmb_rtoken_handling(struct smc_connection *conn, struct smc_link *link,
 			    struct smc_clc_msg_accept_confirm *clc);
 int smc_rtoken_add(struct smc_link *lnk, __be64 nw_vaddr, __be32 nw_rkey);
 int smc_rtoken_delete(struct smc_link *lnk, __be32 nw_rkey);
+void smc_rtoken_set(struct smc_link_group *lgr, int link_idx, int link_idx_new,
+		    __be32 nw_rkey_known, __be64 nw_vaddr, __be32 nw_rkey);
+void smc_rtoken_set2(struct smc_link_group *lgr, int rtok_idx, int link_id,
+		     __be64 nw_vaddr, __be32 nw_rkey);
 void smc_sndbuf_sync_sg_for_cpu(struct smc_connection *conn);
 void smc_sndbuf_sync_sg_for_device(struct smc_connection *conn);
 void smc_rmb_sync_sg_for_cpu(struct smc_connection *conn);

net/smc/smc_llc.h

@@ -57,12 +57,21 @@ void smc_llc_lgr_init(struct smc_link_group *lgr, struct smc_sock *smc);
 void smc_llc_lgr_clear(struct smc_link_group *lgr);
 int smc_llc_link_init(struct smc_link *link);
 void smc_llc_link_active(struct smc_link *link);
-void smc_llc_link_deleting(struct smc_link *link);
 void smc_llc_link_clear(struct smc_link *link);
-int smc_llc_do_confirm_rkey(struct smc_link *link,
+int smc_llc_do_confirm_rkey(struct smc_link *send_link,
 			    struct smc_buf_desc *rmb_desc);
-int smc_llc_do_delete_rkey(struct smc_link *link,
+int smc_llc_do_delete_rkey(struct smc_link_group *lgr,
 			   struct smc_buf_desc *rmb_desc);
+int smc_llc_flow_initiate(struct smc_link_group *lgr,
+			  enum smc_llc_flowtype type);
+void smc_llc_flow_stop(struct smc_link_group *lgr, struct smc_llc_flow *flow);
+int smc_llc_eval_conf_link(struct smc_llc_qentry *qentry,
+			   enum smc_llc_reqresp type);
+struct smc_llc_qentry *smc_llc_wait(struct smc_link_group *lgr,
+				    struct smc_link *lnk,
+				    int time_out, u8 exp_msg);
+struct smc_llc_qentry *smc_llc_flow_qentry_clr(struct smc_llc_flow *flow);
+void smc_llc_flow_qentry_del(struct smc_llc_flow *flow);
 int smc_llc_init(void) __init;
 
 #endif /* SMC_LLC_H */