net/smc: add base infrastructure for SMC-D and ISM

SMC supports two variants: SMC-R and SMC-D. For data transport, SMC-R
uses RDMA devices, SMC-D uses so-called Internal Shared Memory (ISM)
devices. An ISM device only allows shared memory communication between
SMC instances on the same machine. For example, this allows virtual
machines on the same host to communicate via SMC without RDMA devices.

This patch adds the base infrastructure for SMC-D and ISM devices to
the existing SMC code. It contains the following:

* ISM driver interface:
  This interface allows an ISM driver to register ISM devices in SMC. In
  the process, the driver provides a set of device ops for each device.
  SMC uses these ops to execute SMC specific operations on or transfer
  data over the device.

* Core SMC-D link group, connection, and buffer support:
  Link groups, SMC connections and SMC buffers (in smc_core) are
  extended to support SMC-D.

* SMC type checks:
  Some type checks are added to prevent using SMC-R specific code for
  SMC-D and vice versa.

To actually use SMC-D, additional changes to pnetid, CLC, CDC, etc. are
required. These are added in follow-up patches.

Signed-off-by: Hans Wippel <hwippel@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Suggested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/smc.h

@@ -20,4 +20,66 @@ struct smc_hashinfo {
 
 int smc_hash_sk(struct sock *sk);
 void smc_unhash_sk(struct sock *sk);
+
+/* SMCD/ISM device driver interface */
+struct smcd_dmb {
+	u64 dmb_tok;
+	u64 rgid;
+	u32 dmb_len;
+	u32 sba_idx;
+	u32 vlan_valid;
+	u32 vlan_id;
+	void *cpu_addr;
+	dma_addr_t dma_addr;
+};
+
+#define ISM_EVENT_DMB	0
+#define ISM_EVENT_GID	1
+#define ISM_EVENT_SWR	2
+
+struct smcd_event {
+	u32 type;
+	u32 code;
+	u64 tok;
+	u64 time;
+	u64 info;
+};
+
+struct smcd_dev;
+
+struct smcd_ops {
+	int (*query_remote_gid)(struct smcd_dev *dev, u64 rgid, u32 vid_valid,
+				u32 vid);
+	int (*register_dmb)(struct smcd_dev *dev, struct smcd_dmb *dmb);
+	int (*unregister_dmb)(struct smcd_dev *dev, struct smcd_dmb *dmb);
+	int (*add_vlan_id)(struct smcd_dev *dev, u64 vlan_id);
+	int (*del_vlan_id)(struct smcd_dev *dev, u64 vlan_id);
+	int (*set_vlan_required)(struct smcd_dev *dev);
+	int (*reset_vlan_required)(struct smcd_dev *dev);
+	int (*signal_event)(struct smcd_dev *dev, u64 rgid, u32 trigger_irq,
+			    u32 event_code, u64 info);
+	int (*move_data)(struct smcd_dev *dev, u64 dmb_tok, unsigned int idx,
+			 bool sf, unsigned int offset, void *data,
+			 unsigned int size);
+};
+
+struct smcd_dev {
+	const struct smcd_ops *ops;
+	struct device dev;
+	void *priv;
+	u64 local_gid;
+	struct list_head list;
+	spinlock_t lock;
+	struct smc_connection **conn;
+	struct list_head vlan;
+	struct workqueue_struct *event_wq;
+};
+
+struct smcd_dev *smcd_alloc_dev(struct device *parent, const char *name,
+				const struct smcd_ops *ops, int max_dmbs);
+int smcd_register_dev(struct smcd_dev *smcd);
+void smcd_unregister_dev(struct smcd_dev *smcd);
+void smcd_free_dev(struct smcd_dev *smcd);
+void smcd_handle_event(struct smcd_dev *dev, struct smcd_event *event);
+void smcd_handle_irq(struct smcd_dev *dev, unsigned int bit);
 #endif	/* _SMC_H */

net/smc/Makefile

@@ -1,4 +1,4 @@
 obj-$(CONFIG_SMC)	+= smc.o
 obj-$(CONFIG_SMC_DIAG)	+= smc_diag.o
 smc-y := af_smc.o smc_pnet.o smc_ib.o smc_clc.o smc_core.o smc_wr.o smc_llc.o
-smc-y += smc_cdc.o smc_tx.o smc_rx.o smc_close.o
+smc-y += smc_cdc.o smc_tx.o smc_rx.o smc_close.o smc_ism.o

net/smc/af_smc.c

@@ -475,8 +475,8 @@ static int smc_connect_rdma(struct smc_sock *smc,
 	int reason_code = 0;
 
 	mutex_lock(&smc_create_lgr_pending);
-	local_contact = smc_conn_create(smc, ibdev, ibport, &aclc->lcl,
-					aclc->hdr.flag);
+	local_contact = smc_conn_create(smc, false, aclc->hdr.flag, ibdev,
+					ibport, &aclc->lcl, NULL, 0);
 	if (local_contact < 0) {
 		if (local_contact == -ENOMEM)
 			reason_code = SMC_CLC_DECL_MEM;/* insufficient memory*/
@@ -491,7 +491,7 @@ static int smc_connect_rdma(struct smc_sock *smc,
 	smc_conn_save_peer_info(smc, aclc);
 
 	/* create send buffer and rmb */
-	if (smc_buf_create(smc))
+	if (smc_buf_create(smc, false))
 		return smc_connect_abort(smc, SMC_CLC_DECL_MEM, local_contact);
 
 	if (local_contact == SMC_FIRST_CONTACT)
@@ -894,7 +894,8 @@ static int smc_listen_rdma_init(struct smc_sock *new_smc,
 				int *local_contact)
 {
 	/* allocate connection / link group */
-	*local_contact = smc_conn_create(new_smc, ibdev, ibport, &pclc->lcl, 0);
+	*local_contact = smc_conn_create(new_smc, false, 0, ibdev, ibport,
+					 &pclc->lcl, NULL, 0);
 	if (*local_contact < 0) {
 		if (*local_contact == -ENOMEM)
 			return SMC_CLC_DECL_MEM;/* insufficient memory*/
@@ -902,7 +903,7 @@ static int smc_listen_rdma_init(struct smc_sock *new_smc,
 	}
 
 	/* create send buffer and rmb */
-	if (smc_buf_create(new_smc))
+	if (smc_buf_create(new_smc, false))
 		return SMC_CLC_DECL_MEM;
 
 	return 0;

net/smc/smc_core.c

@@ -25,6 +25,7 @@
 #include "smc_llc.h"
 #include "smc_cdc.h"
 #include "smc_close.h"
+#include "smc_ism.h"
 
 #define SMC_LGR_NUM_INCR		256
 #define SMC_LGR_FREE_DELAY_SERV		(600 * HZ)
@@ -46,8 +47,8 @@ static void smc_lgr_schedule_free_work(struct smc_link_group *lgr)
 	 * otherwise there is a risk of out-of-sync link groups.
 	 */
 	mod_delayed_work(system_wq, &lgr->free_work,
-			 lgr->role == SMC_CLNT ? SMC_LGR_FREE_DELAY_CLNT :
-						 SMC_LGR_FREE_DELAY_SERV);
+			 (!lgr->is_smcd && lgr->role == SMC_CLNT) ?
+			 SMC_LGR_FREE_DELAY_CLNT : SMC_LGR_FREE_DELAY_SERV);
 }
 
 /* Register connection's alert token in our lookup structure.
@@ -153,16 +154,18 @@ static void smc_lgr_free_work(struct work_struct *work)
 free:
 	spin_unlock_bh(&smc_lgr_list.lock);
 	if (!delayed_work_pending(&lgr->free_work)) {
-		if (lgr->lnk[SMC_SINGLE_LINK].state != SMC_LNK_INACTIVE)
+		if (!lgr->is_smcd &&
+		    lgr->lnk[SMC_SINGLE_LINK].state != SMC_LNK_INACTIVE)
 			smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
 		smc_lgr_free(lgr);
 	}
 }
 
 /* create a new SMC link group */
-static int smc_lgr_create(struct smc_sock *smc,
+static int smc_lgr_create(struct smc_sock *smc, bool is_smcd,
 			  struct smc_ib_device *smcibdev, u8 ibport,
-			  char *peer_systemid, unsigned short vlan_id)
+			  char *peer_systemid, unsigned short vlan_id,
+			  struct smcd_dev *smcismdev, u64 peer_gid)
 {
 	struct smc_link_group *lgr;
 	struct smc_link *lnk;
@@ -170,17 +173,23 @@ static int smc_lgr_create(struct smc_sock *smc,
 	int rc = 0;
 	int i;
 
+	if (is_smcd && vlan_id) {
+		rc = smc_ism_get_vlan(smcismdev, vlan_id);
+		if (rc)
+			goto out;
+	}
+
 	lgr = kzalloc(sizeof(*lgr), GFP_KERNEL);
 	if (!lgr) {
 		rc = -ENOMEM;
 		goto out;
 	}
-	lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT;
+	lgr->is_smcd = is_smcd;
 	lgr->sync_err = 0;
-	memcpy(lgr->peer_systemid, peer_systemid, SMC_SYSTEMID_LEN);
 	lgr->vlan_id = vlan_id;
 	rwlock_init(&lgr->sndbufs_lock);
 	rwlock_init(&lgr->rmbs_lock);
+	rwlock_init(&lgr->conns_lock);
 	for (i = 0; i < SMC_RMBE_SIZES; i++) {
 		INIT_LIST_HEAD(&lgr->sndbufs[i]);
 		INIT_LIST_HEAD(&lgr->rmbs[i]);
@@ -189,36 +198,44 @@ static int smc_lgr_create(struct smc_sock *smc,
 	memcpy(&lgr->id, (u8 *)&smc_lgr_list.num, SMC_LGR_ID_SIZE);
 	INIT_DELAYED_WORK(&lgr->free_work, smc_lgr_free_work);
 	lgr->conns_all = RB_ROOT;
-
-	lnk = &lgr->lnk[SMC_SINGLE_LINK];
-	/* initialize link */
-	lnk->state = SMC_LNK_ACTIVATING;
-	lnk->link_id = SMC_SINGLE_LINK;
-	lnk->smcibdev = smcibdev;
-	lnk->ibport = ibport;
-	lnk->path_mtu = smcibdev->pattr[ibport - 1].active_mtu;
-	if (!smcibdev->initialized)
-		smc_ib_setup_per_ibdev(smcibdev);
-	get_random_bytes(rndvec, sizeof(rndvec));
-	lnk->psn_initial = rndvec[0] + (rndvec[1] << 8) + (rndvec[2] << 16);
-	rc = smc_llc_link_init(lnk);
-	if (rc)
-		goto free_lgr;
-	rc = smc_wr_alloc_link_mem(lnk);
-	if (rc)
-		goto clear_llc_lnk;
-	rc = smc_ib_create_protection_domain(lnk);
-	if (rc)
-		goto free_link_mem;
-	rc = smc_ib_create_queue_pair(lnk);
-	if (rc)
-		goto dealloc_pd;
-	rc = smc_wr_create_link(lnk);
-	if (rc)
-		goto destroy_qp;
-
+	if (is_smcd) {
+		/* SMC-D specific settings */
+		lgr->peer_gid = peer_gid;
+		lgr->smcd = smcismdev;
+	} else {
+		/* SMC-R specific settings */
+		lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT;
+		memcpy(lgr->peer_systemid, peer_systemid, SMC_SYSTEMID_LEN);
+
+		lnk = &lgr->lnk[SMC_SINGLE_LINK];
+		/* initialize link */
+		lnk->state = SMC_LNK_ACTIVATING;
+		lnk->link_id = SMC_SINGLE_LINK;
+		lnk->smcibdev = smcibdev;
+		lnk->ibport = ibport;
+		lnk->path_mtu = smcibdev->pattr[ibport - 1].active_mtu;
+		if (!smcibdev->initialized)
+			smc_ib_setup_per_ibdev(smcibdev);
+		get_random_bytes(rndvec, sizeof(rndvec));
+		lnk->psn_initial = rndvec[0] + (rndvec[1] << 8) +
+			(rndvec[2] << 16);
+		rc = smc_llc_link_init(lnk);
+		if (rc)
+			goto free_lgr;
+		rc = smc_wr_alloc_link_mem(lnk);
+		if (rc)
+			goto clear_llc_lnk;
+		rc = smc_ib_create_protection_domain(lnk);
+		if (rc)
+			goto free_link_mem;
+		rc = smc_ib_create_queue_pair(lnk);
+		if (rc)
+			goto dealloc_pd;
+		rc = smc_wr_create_link(lnk);
+		if (rc)
+			goto destroy_qp;
+	}
 	smc->conn.lgr = lgr;
-	rwlock_init(&lgr->conns_lock);
 	spin_lock_bh(&smc_lgr_list.lock);
 	list_add(&lgr->list, &smc_lgr_list.list);
 	spin_unlock_bh(&smc_lgr_list.lock);
@@ -264,7 +281,10 @@ void smc_conn_free(struct smc_connection *conn)
 {
 	if (!conn->lgr)
 		return;
-	smc_cdc_tx_dismiss_slots(conn);
+	if (conn->lgr->is_smcd)
+		smc_ism_unset_conn(conn);
+	else
+		smc_cdc_tx_dismiss_slots(conn);
 	smc_lgr_unregister_conn(conn);
 	smc_buf_unuse(conn);
 }
@@ -280,8 +300,8 @@ static void smc_link_clear(struct smc_link *lnk)
 	smc_wr_free_link_mem(lnk);
 }
 
-static void smc_buf_free(struct smc_link_group *lgr, bool is_rmb,
-			 struct smc_buf_desc *buf_desc)
+static void smcr_buf_free(struct smc_link_group *lgr, bool is_rmb,
+			  struct smc_buf_desc *buf_desc)
 {
 	struct smc_link *lnk = &lgr->lnk[SMC_SINGLE_LINK];
 
@@ -301,6 +321,25 @@ static void smc_buf_free(struct smc_link_group *lgr, bool is_rmb,
 	kfree(buf_desc);
 }
 
+static void smcd_buf_free(struct smc_link_group *lgr, bool is_dmb,
+			  struct smc_buf_desc *buf_desc)
+{
+	if (is_dmb)
+		smc_ism_unregister_dmb(lgr->smcd, buf_desc);
+	else
+		kfree(buf_desc->cpu_addr);
+	kfree(buf_desc);
+}
+
+static void smc_buf_free(struct smc_link_group *lgr, bool is_rmb,
+			 struct smc_buf_desc *buf_desc)
+{
+	if (lgr->is_smcd)
+		smcd_buf_free(lgr, is_rmb, buf_desc);
+	else
+		smcr_buf_free(lgr, is_rmb, buf_desc);
+}
+
 static void __smc_lgr_free_bufs(struct smc_link_group *lgr, bool is_rmb)
 {
 	struct smc_buf_desc *buf_desc, *bf_desc;
@@ -332,7 +371,10 @@ static void smc_lgr_free_bufs(struct smc_link_group *lgr)
 void smc_lgr_free(struct smc_link_group *lgr)
 {
 	smc_lgr_free_bufs(lgr);
-	smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]);
+	if (lgr->is_smcd)
+		smc_ism_put_vlan(lgr->smcd, lgr->vlan_id);
+	else
+		smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]);
 	kfree(lgr);
 }
 
@@ -357,7 +399,8 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr)
 	lgr->terminating = 1;
 	if (!list_empty(&lgr->list)) /* forget lgr */
 		list_del_init(&lgr->list);
-	smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
+	if (!lgr->is_smcd)
+		smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
 
 	write_lock_bh(&lgr->conns_lock);
 	node = rb_first(&lgr->conns_all);
@@ -374,7 +417,8 @@ static void __smc_lgr_terminate(struct smc_link_group *lgr)
 		node = rb_first(&lgr->conns_all);
 	}
 	write_unlock_bh(&lgr->conns_lock);
-	wake_up(&lgr->lnk[SMC_SINGLE_LINK].wr_reg_wait);
+	if (!lgr->is_smcd)
+		wake_up(&lgr->lnk[SMC_SINGLE_LINK].wr_reg_wait);
 	smc_lgr_schedule_free_work(lgr);
 }
 
@@ -392,13 +436,40 @@ void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport)
 
 	spin_lock_bh(&smc_lgr_list.lock);
 	list_for_each_entry_safe(lgr, l, &smc_lgr_list.list, list) {
-		if (lgr->lnk[SMC_SINGLE_LINK].smcibdev == smcibdev &&
+		if (!lgr->is_smcd &&
+		    lgr->lnk[SMC_SINGLE_LINK].smcibdev == smcibdev &&
 		    lgr->lnk[SMC_SINGLE_LINK].ibport == ibport)
 			__smc_lgr_terminate(lgr);
 	}
 	spin_unlock_bh(&smc_lgr_list.lock);
 }
 
+/* Called when SMC-D device is terminated or peer is lost */
+void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid)
+{
+	struct smc_link_group *lgr, *l;
+	LIST_HEAD(lgr_free_list);
+
+	/* run common cleanup function and build free list */
+	spin_lock_bh(&smc_lgr_list.lock);
+	list_for_each_entry_safe(lgr, l, &smc_lgr_list.list, list) {
+		if (lgr->is_smcd && lgr->smcd == dev &&
+		    (!peer_gid || lgr->peer_gid == peer_gid) &&
+		    !list_empty(&lgr->list)) {
+			__smc_lgr_terminate(lgr);
+			list_move(&lgr->list, &lgr_free_list);
+		}
+	}
+	spin_unlock_bh(&smc_lgr_list.lock);
+
+	/* cancel the regular free workers and actually free lgrs */
+	list_for_each_entry_safe(lgr, l, &lgr_free_list, list) {
+		list_del_init(&lgr->list);
+		cancel_delayed_work_sync(&lgr->free_work);
+		smc_lgr_free(lgr);
+	}
+}
+
 /* Determine vlan of internal TCP socket.
  * @vlan_id: address to store the determined vlan id into
  */
@@ -477,10 +548,30 @@ static int smc_link_determine_gid(struct smc_link_group *lgr)
 	return -ENODEV;
 }
 
+static bool smcr_lgr_match(struct smc_link_group *lgr,
+			   struct smc_clc_msg_local *lcl,
+			   enum smc_lgr_role role)
+{
+	return !memcmp(lgr->peer_systemid, lcl->id_for_peer,
+		       SMC_SYSTEMID_LEN) &&
+		!memcmp(lgr->lnk[SMC_SINGLE_LINK].peer_gid, &lcl->gid,
+			SMC_GID_SIZE) &&
+		!memcmp(lgr->lnk[SMC_SINGLE_LINK].peer_mac, lcl->mac,
+			sizeof(lcl->mac)) &&
+		lgr->role == role;
+}
+
+static bool smcd_lgr_match(struct smc_link_group *lgr,
+			   struct smcd_dev *smcismdev, u64 peer_gid)
+{
+	return lgr->peer_gid == peer_gid && lgr->smcd == smcismdev;
+}
+
 /* create a new SMC connection (and a new link group if necessary) */
-int smc_conn_create(struct smc_sock *smc,
+int smc_conn_create(struct smc_sock *smc, bool is_smcd, int srv_first_contact,
 		    struct smc_ib_device *smcibdev, u8 ibport,
-		    struct smc_clc_msg_local *lcl, int srv_first_contact)
+		    struct smc_clc_msg_local *lcl, struct smcd_dev *smcd,
+		    u64 peer_gid)
 {
 	struct smc_connection *conn = &smc->conn;
 	int local_contact = SMC_FIRST_CONTACT;
@@ -502,17 +593,12 @@ int smc_conn_create(struct smc_sock *smc,
 	spin_lock_bh(&smc_lgr_list.lock);
 	list_for_each_entry(lgr, &smc_lgr_list.list, list) {
 		write_lock_bh(&lgr->conns_lock);
-		if (!memcmp(lgr->peer_systemid, lcl->id_for_peer,
-			    SMC_SYSTEMID_LEN) &&
-		    !memcmp(lgr->lnk[SMC_SINGLE_LINK].peer_gid, &lcl->gid,
-			    SMC_GID_SIZE) &&
-		    !memcmp(lgr->lnk[SMC_SINGLE_LINK].peer_mac, lcl->mac,
-			    sizeof(lcl->mac)) &&
+		if ((is_smcd ? smcd_lgr_match(lgr, smcd, peer_gid) :
+		     smcr_lgr_match(lgr, lcl, role)) &&
 		    !lgr->sync_err &&
-		    (lgr->role == role) &&
-		    (lgr->vlan_id == vlan_id) &&
-		    ((role == SMC_CLNT) ||
-		     (lgr->conns_num < SMC_RMBS_PER_LGR_MAX))) {
+		    lgr->vlan_id == vlan_id &&
+		    (role == SMC_CLNT ||
+		     lgr->conns_num < SMC_RMBS_PER_LGR_MAX)) {
 			/* link group found */
 			local_contact = SMC_REUSE_CONTACT;
 			conn->lgr = lgr;
@@ -535,12 +621,13 @@ int smc_conn_create(struct smc_sock *smc,
 
 create:
 	if (local_contact == SMC_FIRST_CONTACT) {
-		rc = smc_lgr_create(smc, smcibdev, ibport,
-				    lcl->id_for_peer, vlan_id);
+		rc = smc_lgr_create(smc, is_smcd, smcibdev, ibport,
+				    lcl->id_for_peer, vlan_id, smcd, peer_gid);
 		if (rc)
 			goto out;
 		smc_lgr_register_conn(conn); /* add smc conn to lgr */
-		rc = smc_link_determine_gid(conn->lgr);
+		if (!is_smcd)
+			rc = smc_link_determine_gid(conn->lgr);
 	}
 	conn->local_tx_ctrl.common.type = SMC_CDC_MSG_TYPE;
 	conn->local_tx_ctrl.len = SMC_WR_TX_SIZE;
@@ -609,8 +696,8 @@ static inline int smc_rmb_wnd_update_limit(int rmbe_size)
 	return min_t(int, rmbe_size / 10, SOCK_MIN_SNDBUF / 2);
 }
 
-static struct smc_buf_desc *smc_new_buf_create(struct smc_link_group *lgr,
-					       bool is_rmb, int bufsize)
+static struct smc_buf_desc *smcr_new_buf_create(struct smc_link_group *lgr,
+						bool is_rmb, int bufsize)
 {
 	struct smc_buf_desc *buf_desc;
 	struct smc_link *lnk;
@@ -668,7 +755,43 @@ static struct smc_buf_desc *smc_new_buf_create(struct smc_link_group *lgr,
 	return buf_desc;
 }
 
-static int __smc_buf_create(struct smc_sock *smc, bool is_rmb)
+#define SMCD_DMBE_SIZES		7 /* 0 -> 16KB, 1 -> 32KB, .. 6 -> 1MB */
+
+static struct smc_buf_desc *smcd_new_buf_create(struct smc_link_group *lgr,
+						bool is_dmb, int bufsize)
+{
+	struct smc_buf_desc *buf_desc;
+	int rc;
+
+	if (smc_compress_bufsize(bufsize) > SMCD_DMBE_SIZES)
+		return ERR_PTR(-EAGAIN);
+
+	/* try to alloc a new DMB */
+	buf_desc = kzalloc(sizeof(*buf_desc), GFP_KERNEL);
+	if (!buf_desc)
+		return ERR_PTR(-ENOMEM);
+	if (is_dmb) {
+		rc = smc_ism_register_dmb(lgr, bufsize, buf_desc);
+		if (rc) {
+			kfree(buf_desc);
+			return ERR_PTR(-EAGAIN);
+		}
+		memset(buf_desc->cpu_addr, 0, bufsize);
+		buf_desc->len = bufsize;
+	} else {
+		buf_desc->cpu_addr = kzalloc(bufsize, GFP_KERNEL |
+					     __GFP_NOWARN | __GFP_NORETRY |
+					     __GFP_NOMEMALLOC);
+		if (!buf_desc->cpu_addr) {
+			kfree(buf_desc);
+			return ERR_PTR(-EAGAIN);
+		}
+		buf_desc->len = bufsize;
+	}
+	return buf_desc;
+}
+
+static int __smc_buf_create(struct smc_sock *smc, bool is_smcd, bool is_rmb)
 {
 	struct smc_buf_desc *buf_desc = ERR_PTR(-ENOMEM);
 	struct smc_connection *conn = &smc->conn;
@@ -706,7 +829,11 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_rmb)
 			break; /* found reusable slot */
 		}
 
-		buf_desc = smc_new_buf_create(lgr, is_rmb, bufsize);
+		if (is_smcd)
+			buf_desc = smcd_new_buf_create(lgr, is_rmb, bufsize);
+		else
+			buf_desc = smcr_new_buf_create(lgr, is_rmb, bufsize);
+
 		if (PTR_ERR(buf_desc) == -ENOMEM)
 			break;
 		if (IS_ERR(buf_desc))
@@ -728,6 +855,8 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_rmb)
 		smc->sk.sk_rcvbuf = bufsize * 2;
 		atomic_set(&conn->bytes_to_rcv, 0);
 		conn->rmbe_update_limit = smc_rmb_wnd_update_limit(bufsize);
+		if (is_smcd)
+			smc_ism_set_conn(conn); /* map RMB/smcd_dev to conn */
 	} else {
 		conn->sndbuf_desc = buf_desc;
 		smc->sk.sk_sndbuf = bufsize * 2;
@@ -740,6 +869,8 @@ void smc_sndbuf_sync_sg_for_cpu(struct smc_connection *conn)
 {
 	struct smc_link_group *lgr = conn->lgr;
 
+	if (!conn->lgr || conn->lgr->is_smcd)
+		return;
 	smc_ib_sync_sg_for_cpu(lgr->lnk[SMC_SINGLE_LINK].smcibdev,
 			       conn->sndbuf_desc, DMA_TO_DEVICE);
 }
@@ -748,6 +879,8 @@ void smc_sndbuf_sync_sg_for_device(struct smc_connection *conn)
 {
 	struct smc_link_group *lgr = conn->lgr;
 
+	if (!conn->lgr || conn->lgr->is_smcd)
+		return;
 	smc_ib_sync_sg_for_device(lgr->lnk[SMC_SINGLE_LINK].smcibdev,
 				  conn->sndbuf_desc, DMA_TO_DEVICE);
 }
@@ -756,6 +889,8 @@ void smc_rmb_sync_sg_for_cpu(struct smc_connection *conn)
 {
 	struct smc_link_group *lgr = conn->lgr;
 
+	if (!conn->lgr || conn->lgr->is_smcd)
+		return;
 	smc_ib_sync_sg_for_cpu(lgr->lnk[SMC_SINGLE_LINK].smcibdev,
 			       conn->rmb_desc, DMA_FROM_DEVICE);
 }
@@ -764,6 +899,8 @@ void smc_rmb_sync_sg_for_device(struct smc_connection *conn)
 {
 	struct smc_link_group *lgr = conn->lgr;
 
+	if (!conn->lgr || conn->lgr->is_smcd)
+		return;
 	smc_ib_sync_sg_for_device(lgr->lnk[SMC_SINGLE_LINK].smcibdev,
 				  conn->rmb_desc, DMA_FROM_DEVICE);
 }
@@ -774,16 +911,16 @@ void smc_rmb_sync_sg_for_device(struct smc_connection *conn)
  * the Linux implementation uses just one RMB-element per RMB, i.e. uses an
  * extra RMB for every connection in a link group
  */
-int smc_buf_create(struct smc_sock *smc)
+int smc_buf_create(struct smc_sock *smc, bool is_smcd)
 {
 	int rc;
 
 	/* create send buffer */
-	rc = __smc_buf_create(smc, false);
+	rc = __smc_buf_create(smc, is_smcd, false);
 	if (rc)
 		return rc;
 	/* create rmb */
-	rc = __smc_buf_create(smc, true);
+	rc = __smc_buf_create(smc, is_smcd, true);
 	if (rc)
 		smc_buf_free(smc->conn.lgr, false, smc->conn.sndbuf_desc);
 	return rc;
@@ -865,7 +1002,8 @@ void smc_core_exit(void)
 	spin_unlock_bh(&smc_lgr_list.lock);
 	list_for_each_entry_safe(lgr, lg, &lgr_freeing_list, list) {
 		list_del_init(&lgr->list);
-		smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
+		if (!lgr->is_smcd)
+			smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
 		cancel_delayed_work_sync(&lgr->free_work);
 		smc_lgr_free(lgr); /* free link group */
 	}

net/smc/smc_core.h

@@ -124,15 +124,28 @@ struct smc_buf_desc {
 	void			*cpu_addr;	/* virtual address of buffer */
 	struct page		*pages;
 	int			len;		/* length of buffer */
-	struct sg_table		sgt[SMC_LINKS_PER_LGR_MAX];/* virtual buffer */
-	struct ib_mr		*mr_rx[SMC_LINKS_PER_LGR_MAX];
-						/* for rmb only: memory region
-						 * incl. rkey provided to peer
-						 */
-	u32			order;		/* allocation order */
 	u32			used;		/* currently used / unused */
 	u8			reused	: 1;	/* new created / reused */
 	u8			regerr	: 1;	/* err during registration */
+	union {
+		struct { /* SMC-R */
+			struct sg_table		sgt[SMC_LINKS_PER_LGR_MAX];
+						/* virtual buffer */
+			struct ib_mr		*mr_rx[SMC_LINKS_PER_LGR_MAX];
+						/* for rmb only: memory region
+						 * incl. rkey provided to peer
+						 */
+			u32			order;	/* allocation order */
+		};
+		struct { /* SMC-D */
+			unsigned short		sba_idx;
+						/* SBA index number */
+			u64			token;
+						/* DMB token number */
+			dma_addr_t		dma_addr;
+						/* DMA address */
+		};
+	};
 };
 
 struct smc_rtoken {				/* address/key of remote RMB */
@@ -148,12 +161,10 @@ struct smc_rtoken {				/* address/key of remote RMB */
  * struct smc_clc_msg_accept_confirm.rmbe_size being a 4 bit value (0..15)
  */
 
+struct smcd_dev;
+
 struct smc_link_group {
 	struct list_head	list;
-	enum smc_lgr_role	role;		/* client or server */
-	struct smc_link		lnk[SMC_LINKS_PER_LGR_MAX];	/* smc link */
-	char			peer_systemid[SMC_SYSTEMID_LEN];
-						/* unique system_id of peer */
 	struct rb_root		conns_all;	/* connection tree */
 	rwlock_t		conns_lock;	/* protects conns_all */
 	unsigned int		conns_num;	/* current # of connections */
@@ -163,17 +174,35 @@ struct smc_link_group {
 	rwlock_t		sndbufs_lock;	/* protects tx buffers */
 	struct list_head	rmbs[SMC_RMBE_SIZES];	/* rx buffers */
 	rwlock_t		rmbs_lock;	/* protects rx buffers */
-	struct smc_rtoken	rtokens[SMC_RMBS_PER_LGR_MAX]
-				       [SMC_LINKS_PER_LGR_MAX];
-						/* remote addr/key pairs */
-	unsigned long		rtokens_used_mask[BITS_TO_LONGS(
-							SMC_RMBS_PER_LGR_MAX)];
-						/* used rtoken elements */
 
 	u8			id[SMC_LGR_ID_SIZE];	/* unique lgr id */
 	struct delayed_work	free_work;	/* delayed freeing of an lgr */
 	u8			sync_err : 1;	/* lgr no longer fits to peer */
 	u8			terminating : 1;/* lgr is terminating */
+
+	bool			is_smcd;	/* SMC-R or SMC-D */
+	union {
+		struct { /* SMC-R */
+			enum smc_lgr_role	role;
+						/* client or server */
+			struct smc_link		lnk[SMC_LINKS_PER_LGR_MAX];
+						/* smc link */
+			char			peer_systemid[SMC_SYSTEMID_LEN];
+						/* unique system_id of peer */
+			struct smc_rtoken	rtokens[SMC_RMBS_PER_LGR_MAX]
+						[SMC_LINKS_PER_LGR_MAX];
+						/* remote addr/key pairs */
+			unsigned long		rtokens_used_mask[BITS_TO_LONGS
+							(SMC_RMBS_PER_LGR_MAX)];
+						/* used rtoken elements */
+		};
+		struct { /* SMC-D */
+			u64			peer_gid;
+						/* Peer GID (remote) */
+			struct smcd_dev		*smcd;
+						/* ISM device for VLAN reg. */
+		};
+	};
 };
 
 /* Find the connection associated with the given alert token in the link group.
@@ -217,7 +246,8 @@ void smc_lgr_free(struct smc_link_group *lgr);
 void smc_lgr_forget(struct smc_link_group *lgr);
 void smc_lgr_terminate(struct smc_link_group *lgr);
 void smc_port_terminate(struct smc_ib_device *smcibdev, u8 ibport);
-int smc_buf_create(struct smc_sock *smc);
+void smc_smcd_terminate(struct smcd_dev *dev, u64 peer_gid);
+int smc_buf_create(struct smc_sock *smc, bool is_smcd);
 int smc_uncompress_bufsize(u8 compressed);
 int smc_rmb_rtoken_handling(struct smc_connection *conn,
 			    struct smc_clc_msg_accept_confirm *clc);
@@ -227,9 +257,12 @@ 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);
 void smc_rmb_sync_sg_for_device(struct smc_connection *conn);
+
 void smc_conn_free(struct smc_connection *conn);
-int smc_conn_create(struct smc_sock *smc,
+int smc_conn_create(struct smc_sock *smc, bool is_smcd, int srv_first_contact,
 		    struct smc_ib_device *smcibdev, u8 ibport,
-		    struct smc_clc_msg_local *lcl, int srv_first_contact);
+		    struct smc_clc_msg_local *lcl, struct smcd_dev *smcd,
+		    u64 peer_gid);
+void smcd_conn_free(struct smc_connection *conn);
 void smc_core_exit(void);
 #endif

net/smc/smc_diag.c

@@ -136,7 +136,8 @@ static int __smc_diag_dump(struct sock *sk, struct sk_buff *skb,
 			goto errout;
 	}
 
-	if ((req->diag_ext & (1 << (SMC_DIAG_LGRINFO - 1))) && smc->conn.lgr &&
+	if (smc->conn.lgr && !smc->conn.lgr->is_smcd &&
+	    (req->diag_ext & (1 << (SMC_DIAG_LGRINFO - 1))) &&
 	    !list_empty(&smc->conn.lgr->list)) {
 		struct smc_diag_lgrinfo linfo = {
 			.role = smc->conn.lgr->role,

net/smc/smc_ism.c

@@ -0,0 +1,304 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Shared Memory Communications Direct over ISM devices (SMC-D)
+ *
+ * Functions for ISM device.
+ *
+ * Copyright IBM Corp. 2018
+ */
+
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <asm/page.h>
+
+#include "smc.h"
+#include "smc_core.h"
+#include "smc_ism.h"
+
+struct smcd_dev_list smcd_dev_list = {
+	.list = LIST_HEAD_INIT(smcd_dev_list.list),
+	.lock = __SPIN_LOCK_UNLOCKED(smcd_dev_list.lock)
+};
+
+/* Test if an ISM communication is possible. */
+int smc_ism_cantalk(u64 peer_gid, unsigned short vlan_id, struct smcd_dev *smcd)
+{
+	return smcd->ops->query_remote_gid(smcd, peer_gid, vlan_id ? 1 : 0,
+					   vlan_id);
+}
+
+int smc_ism_write(struct smcd_dev *smcd, const struct smc_ism_position *pos,
+		  void *data, size_t len)
+{
+	int rc;
+
+	rc = smcd->ops->move_data(smcd, pos->token, pos->index, pos->signal,
+				  pos->offset, data, len);
+
+	return rc < 0 ? rc : 0;
+}
+
+/* Set a connection using this DMBE. */
+void smc_ism_set_conn(struct smc_connection *conn)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&conn->lgr->smcd->lock, flags);
+	conn->lgr->smcd->conn[conn->rmb_desc->sba_idx] = conn;
+	spin_unlock_irqrestore(&conn->lgr->smcd->lock, flags);
+}
+
+/* Unset a connection using this DMBE. */
+void smc_ism_unset_conn(struct smc_connection *conn)
+{
+	unsigned long flags;
+
+	if (!conn->rmb_desc)
+		return;
+
+	spin_lock_irqsave(&conn->lgr->smcd->lock, flags);
+	conn->lgr->smcd->conn[conn->rmb_desc->sba_idx] = NULL;
+	spin_unlock_irqrestore(&conn->lgr->smcd->lock, flags);
+}
+
+/* Register a VLAN identifier with the ISM device. Use a reference count
+ * and add a VLAN identifier only when the first DMB using this VLAN is
+ * registered.
+ */
+int smc_ism_get_vlan(struct smcd_dev *smcd, unsigned short vlanid)
+{
+	struct smc_ism_vlanid *new_vlan, *vlan;
+	unsigned long flags;
+	int rc = 0;
+
+	if (!vlanid)			/* No valid vlan id */
+		return -EINVAL;
+
+	/* create new vlan entry, in case we need it */
+	new_vlan = kzalloc(sizeof(*new_vlan), GFP_KERNEL);
+	if (!new_vlan)
+		return -ENOMEM;
+	new_vlan->vlanid = vlanid;
+	refcount_set(&new_vlan->refcnt, 1);
+
+	/* if there is an existing entry, increase count and return */
+	spin_lock_irqsave(&smcd->lock, flags);
+	list_for_each_entry(vlan, &smcd->vlan, list) {
+		if (vlan->vlanid == vlanid) {
+			refcount_inc(&vlan->refcnt);
+			kfree(new_vlan);
+			goto out;
+		}
+	}
+
+	/* no existing entry found.
+	 * add new entry to device; might fail, e.g., if HW limit reached
+	 */
+	if (smcd->ops->add_vlan_id(smcd, vlanid)) {
+		kfree(new_vlan);
+		rc = -EIO;
+		goto out;
+	}
+	list_add_tail(&new_vlan->list, &smcd->vlan);
+out:
+	spin_unlock_irqrestore(&smcd->lock, flags);
+	return rc;
+}
+
+/* Unregister a VLAN identifier with the ISM device. Use a reference count
+ * and remove a VLAN identifier only when the last DMB using this VLAN is
+ * unregistered.
+ */
+int smc_ism_put_vlan(struct smcd_dev *smcd, unsigned short vlanid)
+{
+	struct smc_ism_vlanid *vlan;
+	unsigned long flags;
+	bool found = false;
+	int rc = 0;
+
+	if (!vlanid)			/* No valid vlan id */
+		return -EINVAL;
+
+	spin_lock_irqsave(&smcd->lock, flags);
+	list_for_each_entry(vlan, &smcd->vlan, list) {
+		if (vlan->vlanid == vlanid) {
+			if (!refcount_dec_and_test(&vlan->refcnt))
+				goto out;
+			found = true;
+			break;
+		}
+	}
+	if (!found) {
+		rc = -ENOENT;
+		goto out;		/* VLAN id not in table */
+	}
+
+	/* Found and the last reference just gone */
+	if (smcd->ops->del_vlan_id(smcd, vlanid))
+		rc = -EIO;
+	list_del(&vlan->list);
+	kfree(vlan);
+out:
+	spin_unlock_irqrestore(&smcd->lock, flags);
+	return rc;
+}
+
+int smc_ism_unregister_dmb(struct smcd_dev *smcd, struct smc_buf_desc *dmb_desc)
+{
+	struct smcd_dmb dmb;
+
+	memset(&dmb, 0, sizeof(dmb));
+	dmb.dmb_tok = dmb_desc->token;
+	dmb.sba_idx = dmb_desc->sba_idx;
+	dmb.cpu_addr = dmb_desc->cpu_addr;
+	dmb.dma_addr = dmb_desc->dma_addr;
+	dmb.dmb_len = dmb_desc->len;
+	return smcd->ops->unregister_dmb(smcd, &dmb);
+}
+
+int smc_ism_register_dmb(struct smc_link_group *lgr, int dmb_len,
+			 struct smc_buf_desc *dmb_desc)
+{
+	struct smcd_dmb dmb;
+	int rc;
+
+	memset(&dmb, 0, sizeof(dmb));
+	dmb.dmb_len = dmb_len;
+	dmb.sba_idx = dmb_desc->sba_idx;
+	dmb.vlan_id = lgr->vlan_id;
+	dmb.rgid = lgr->peer_gid;
+	rc = lgr->smcd->ops->register_dmb(lgr->smcd, &dmb);
+	if (!rc) {
+		dmb_desc->sba_idx = dmb.sba_idx;
+		dmb_desc->token = dmb.dmb_tok;
+		dmb_desc->cpu_addr = dmb.cpu_addr;
+		dmb_desc->dma_addr = dmb.dma_addr;
+		dmb_desc->len = dmb.dmb_len;
+	}
+	return rc;
+}
+
+struct smc_ism_event_work {
+	struct work_struct work;
+	struct smcd_dev *smcd;
+	struct smcd_event event;
+};
+
+/* worker for SMC-D events */
+static void smc_ism_event_work(struct work_struct *work)
+{
+	struct smc_ism_event_work *wrk =
+		container_of(work, struct smc_ism_event_work, work);
+
+	switch (wrk->event.type) {
+	case ISM_EVENT_GID:	/* GID event, token is peer GID */
+		smc_smcd_terminate(wrk->smcd, wrk->event.tok);
+		break;
+	case ISM_EVENT_DMB:
+		break;
+	}
+	kfree(wrk);
+}
+
+static void smcd_release(struct device *dev)
+{
+	struct smcd_dev *smcd = container_of(dev, struct smcd_dev, dev);
+
+	kfree(smcd->conn);
+	kfree(smcd);
+}
+
+struct smcd_dev *smcd_alloc_dev(struct device *parent, const char *name,
+				const struct smcd_ops *ops, int max_dmbs)
+{
+	struct smcd_dev *smcd;
+
+	smcd = kzalloc(sizeof(*smcd), GFP_KERNEL);
+	if (!smcd)
+		return NULL;
+	smcd->conn = kcalloc(max_dmbs, sizeof(struct smc_connection *),
+			     GFP_KERNEL);
+	if (!smcd->conn) {
+		kfree(smcd);
+		return NULL;
+	}
+
+	smcd->dev.parent = parent;
+	smcd->dev.release = smcd_release;
+	device_initialize(&smcd->dev);
+	dev_set_name(&smcd->dev, name);
+	smcd->ops = ops;
+
+	spin_lock_init(&smcd->lock);
+	INIT_LIST_HEAD(&smcd->vlan);
+	smcd->event_wq = alloc_ordered_workqueue("ism_evt_wq-%s)",
+						 WQ_MEM_RECLAIM, name);
+	return smcd;
+}
+EXPORT_SYMBOL_GPL(smcd_alloc_dev);
+
+int smcd_register_dev(struct smcd_dev *smcd)
+{
+	spin_lock(&smcd_dev_list.lock);
+	list_add_tail(&smcd->list, &smcd_dev_list.list);
+	spin_unlock(&smcd_dev_list.lock);
+
+	return device_add(&smcd->dev);
+}
+EXPORT_SYMBOL_GPL(smcd_register_dev);
+
+void smcd_unregister_dev(struct smcd_dev *smcd)
+{
+	spin_lock(&smcd_dev_list.lock);
+	list_del(&smcd->list);
+	spin_unlock(&smcd_dev_list.lock);
+	flush_workqueue(smcd->event_wq);
+	destroy_workqueue(smcd->event_wq);
+	smc_smcd_terminate(smcd, 0);
+
+	device_del(&smcd->dev);
+}
+EXPORT_SYMBOL_GPL(smcd_unregister_dev);
+
+void smcd_free_dev(struct smcd_dev *smcd)
+{
+	put_device(&smcd->dev);
+}
+EXPORT_SYMBOL_GPL(smcd_free_dev);
+
+/* SMCD Device event handler. Called from ISM device interrupt handler.
+ * Parameters are smcd device pointer,
+ * - event->type (0 --> DMB, 1 --> GID),
+ * - event->code (event code),
+ * - event->tok (either DMB token when event type 0, or GID when event type 1)
+ * - event->time (time of day)
+ * - event->info (debug info).
+ *
+ * Context:
+ * - Function called in IRQ context from ISM device driver event handler.
+ */
+void smcd_handle_event(struct smcd_dev *smcd, struct smcd_event *event)
+{
+	struct smc_ism_event_work *wrk;
+
+	/* copy event to event work queue, and let it be handled there */
+	wrk = kmalloc(sizeof(*wrk), GFP_ATOMIC);
+	if (!wrk)
+		return;
+	INIT_WORK(&wrk->work, smc_ism_event_work);
+	wrk->smcd = smcd;
+	wrk->event = *event;
+	queue_work(smcd->event_wq, &wrk->work);
+}
+EXPORT_SYMBOL_GPL(smcd_handle_event);
+
+/* SMCD Device interrupt handler. Called from ISM device interrupt handler.
+ * Parameters are smcd device pointer and DMB number. Find the connection and
+ * schedule the tasklet for this connection.
+ *
+ * Context:
+ * - Function called in IRQ context from ISM device driver IRQ handler.
+ */
+void smcd_handle_irq(struct smcd_dev *smcd, unsigned int dmbno)
+{
+}
+EXPORT_SYMBOL_GPL(smcd_handle_irq);

net/smc/smc_ism.h

@@ -0,0 +1,48 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Shared Memory Communications Direct over ISM devices (SMC-D)
+ *
+ * SMC-D ISM device structure definitions.
+ *
+ * Copyright IBM Corp. 2018
+ */
+
+#ifndef SMCD_ISM_H
+#define SMCD_ISM_H
+
+#include <linux/uio.h>
+
+#include "smc.h"
+
+struct smcd_dev_list {	/* List of SMCD devices */
+	struct list_head list;
+	spinlock_t lock;	/* Protects list of devices */
+};
+
+extern struct smcd_dev_list	smcd_dev_list; /* list of smcd devices */
+
+struct smc_ism_vlanid {			/* VLAN id set on ISM device */
+	struct list_head list;
+	unsigned short vlanid;		/* Vlan id */
+	refcount_t refcnt;		/* Reference count */
+};
+
+struct smc_ism_position {	/* ISM device position to write to */
+	u64 token;		/* Token of DMB */
+	u32 offset;		/* Offset into DMBE */
+	u8 index;		/* Index of DMBE */
+	u8 signal;		/* Generate interrupt on owner side */
+};
+
+struct smcd_dev;
+
+int smc_ism_cantalk(u64 peer_gid, unsigned short vlan_id, struct smcd_dev *dev);
+void smc_ism_set_conn(struct smc_connection *conn);
+void smc_ism_unset_conn(struct smc_connection *conn);
+int smc_ism_get_vlan(struct smcd_dev *dev, unsigned short vlan_id);
+int smc_ism_put_vlan(struct smcd_dev *dev, unsigned short vlan_id);
+int smc_ism_register_dmb(struct smc_link_group *lgr, int buf_size,
+			 struct smc_buf_desc *dmb_desc);
+int smc_ism_unregister_dmb(struct smcd_dev *dev, struct smc_buf_desc *dmb_desc);
+int smc_ism_write(struct smcd_dev *dev, const struct smc_ism_position *pos,
+		  void *data, size_t len);
+#endif