Merge branch 'net-smc-stats'

Karsten Graul says:

====================
net/smc: Add SMC statistic support

Please apply the following patch series for smc to netdev's net-next tree.

This v2 is a resend of the code contained in v1 but with an updated
cover letter to describe why we have chosen to use the generic netlink
mechanism to access the smc protocol's statistic data.

The patchset adds statistic support to the SMC protocol. Per-cpu
variables are used to collect the statistic information for better
performance and for reducing concurrency pitfalls. The code that is
collecting statistic data is implemented in macros to increase code
reuse and readability.
The generic netlink mechanism in SMC is extended to provide the
collected statistics to userspace.
Network namespace awareness is also part of the statistics
implementation.

SMC is a protocol interacting with PCI devices (like RoCE Cards) and
runs on top of the TCP protocol. As SMC is a network protocol and not
an ethernet device driver, we decided to use the generic netlink
interface. This should be comparable to what other protocols in the
net subsystem like tipc, ncsi, ieee802154 or tcp, et al, do.
There is already an established internal generic netlink interface
mechanism in SMC which is used to collect SMC Protocol internal
information. This patchset extends that existing mechanism.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/net_namespace.h

@@ -32,6 +32,7 @@
 #include <net/netns/mpls.h>
 #include <net/netns/can.h>
 #include <net/netns/xdp.h>
+#include <net/netns/smc.h>
 #include <net/netns/bpf.h>
 #include <linux/ns_common.h>
 #include <linux/idr.h>
@@ -170,6 +171,9 @@ struct net {
 	struct sock		*crypto_nlsk;
 #endif
 	struct sock		*diag_nlsk;
+#if IS_ENABLED(CONFIG_SMC)
+	struct netns_smc	smc;
+#endif
 } __randomize_layout;
 
 #include <linux/seq_file_net.h>

include/net/netns/smc.h

+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __NETNS_SMC_H__
+#define __NETNS_SMC_H__
+#include <linux/mutex.h>
+#include <linux/percpu.h>
+
+struct smc_stats_rsn;
+struct smc_stats;
+struct netns_smc {
+	/* per cpu counters for SMC */
+	struct smc_stats __percpu	*smc_stats;
+	/* protect fback_rsn */
+	struct mutex			mutex_fback_rsn;
+	struct smc_stats_rsn		*fback_rsn;
+};
+#endif

include/uapi/linux/smc.h

@@ -47,6 +47,8 @@ enum {
 	SMC_NETLINK_GET_LGR_SMCD,
 	SMC_NETLINK_GET_DEV_SMCD,
 	SMC_NETLINK_GET_DEV_SMCR,
+	SMC_NETLINK_GET_STATS,
+	SMC_NETLINK_GET_FBACK_STATS,
 };
 
 /* SMC_GENL_FAMILY top level attributes */
@@ -58,6 +60,8 @@ enum {
 	SMC_GEN_LGR_SMCD,		/* nest */
 	SMC_GEN_DEV_SMCD,		/* nest */
 	SMC_GEN_DEV_SMCR,		/* nest */
+	SMC_GEN_STATS,			/* nest */
+	SMC_GEN_FBACK_STATS,		/* nest */
 	__SMC_GEN_MAX,
 	SMC_GEN_MAX = __SMC_GEN_MAX - 1
 };
@@ -159,4 +163,83 @@ enum {
 	SMC_NLA_DEV_MAX = __SMC_NLA_DEV_MAX - 1
 };
 
+/* SMC_NLA_STATS_T_TX(RX)_RMB_SIZE nested attributes */
+/* SMC_NLA_STATS_TX(RX)PLOAD_SIZE nested attributes */
+enum {
+	SMC_NLA_STATS_PLOAD_PAD,
+	SMC_NLA_STATS_PLOAD_8K,		/* u64 */
+	SMC_NLA_STATS_PLOAD_16K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_32K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_64K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_128K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_256K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_512K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_1024K,	/* u64 */
+	SMC_NLA_STATS_PLOAD_G_1024K,	/* u64 */
+	__SMC_NLA_STATS_PLOAD_MAX,
+	SMC_NLA_STATS_PLOAD_MAX = __SMC_NLA_STATS_PLOAD_MAX - 1
+};
+
+/* SMC_NLA_STATS_T_TX(RX)_RMB_STATS nested attributes */
+enum {
+	SMC_NLA_STATS_RMB_PAD,
+	SMC_NLA_STATS_RMB_SIZE_SM_PEER_CNT,	/* u64 */
+	SMC_NLA_STATS_RMB_SIZE_SM_CNT,		/* u64 */
+	SMC_NLA_STATS_RMB_FULL_PEER_CNT,	/* u64 */
+	SMC_NLA_STATS_RMB_FULL_CNT,		/* u64 */
+	SMC_NLA_STATS_RMB_REUSE_CNT,		/* u64 */
+	SMC_NLA_STATS_RMB_ALLOC_CNT,		/* u64 */
+	SMC_NLA_STATS_RMB_DGRADE_CNT,		/* u64 */
+	__SMC_NLA_STATS_RMB_MAX,
+	SMC_NLA_STATS_RMB_MAX = __SMC_NLA_STATS_RMB_MAX - 1
+};
+
+/* SMC_NLA_STATS_SMCD_TECH and _SMCR_TECH nested attributes */
+enum {
+	SMC_NLA_STATS_T_PAD,
+	SMC_NLA_STATS_T_TX_RMB_SIZE,	/* nest */
+	SMC_NLA_STATS_T_RX_RMB_SIZE,	/* nest */
+	SMC_NLA_STATS_T_TXPLOAD_SIZE,	/* nest */
+	SMC_NLA_STATS_T_RXPLOAD_SIZE,	/* nest */
+	SMC_NLA_STATS_T_TX_RMB_STATS,	/* nest */
+	SMC_NLA_STATS_T_RX_RMB_STATS,	/* nest */
+	SMC_NLA_STATS_T_CLNT_V1_SUCC,	/* u64 */
+	SMC_NLA_STATS_T_CLNT_V2_SUCC,	/* u64 */
+	SMC_NLA_STATS_T_SRV_V1_SUCC,	/* u64 */
+	SMC_NLA_STATS_T_SRV_V2_SUCC,	/* u64 */
+	SMC_NLA_STATS_T_SENDPAGE_CNT,	/* u64 */
+	SMC_NLA_STATS_T_SPLICE_CNT,	/* u64 */
+	SMC_NLA_STATS_T_CORK_CNT,	/* u64 */
+	SMC_NLA_STATS_T_NDLY_CNT,	/* u64 */
+	SMC_NLA_STATS_T_URG_DATA_CNT,	/* u64 */
+	SMC_NLA_STATS_T_RX_BYTES,	/* u64 */
+	SMC_NLA_STATS_T_TX_BYTES,	/* u64 */
+	SMC_NLA_STATS_T_RX_CNT,		/* u64 */
+	SMC_NLA_STATS_T_TX_CNT,		/* u64 */
+	__SMC_NLA_STATS_T_MAX,
+	SMC_NLA_STATS_T_MAX = __SMC_NLA_STATS_T_MAX - 1
+};
+
+/* SMC_GEN_STATS attributes */
+enum {
+	SMC_NLA_STATS_PAD,
+	SMC_NLA_STATS_SMCD_TECH,	/* nest */
+	SMC_NLA_STATS_SMCR_TECH,	/* nest */
+	SMC_NLA_STATS_CLNT_HS_ERR_CNT,	/* u64 */
+	SMC_NLA_STATS_SRV_HS_ERR_CNT,	/* u64 */
+	__SMC_NLA_STATS_MAX,
+	SMC_NLA_STATS_MAX = __SMC_NLA_STATS_MAX - 1
+};
+
+/* SMC_GEN_FBACK_STATS attributes */
+enum {
+	SMC_NLA_FBACK_STATS_PAD,
+	SMC_NLA_FBACK_STATS_TYPE,	/* u8 */
+	SMC_NLA_FBACK_STATS_SRV_CNT,	/* u64 */
+	SMC_NLA_FBACK_STATS_CLNT_CNT,	/* u64 */
+	SMC_NLA_FBACK_STATS_RSN_CODE,	/* u32 */
+	SMC_NLA_FBACK_STATS_RSN_CNT,	/* u16 */
+	__SMC_NLA_FBACK_STATS_MAX,
+	SMC_NLA_FBACK_STATS_MAX = __SMC_NLA_FBACK_STATS_MAX - 1
+};
 #endif /* _UAPI_LINUX_SMC_H */

net/smc/Makefile

@@ -2,4 +2,4 @@
 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_ism.o smc_netlink.o
+smc-y += smc_cdc.o smc_tx.o smc_rx.o smc_close.o smc_ism.o smc_netlink.o smc_stats.o

net/smc/af_smc.c

@@ -49,6 +49,7 @@
 #include "smc_tx.h"
 #include "smc_rx.h"
 #include "smc_close.h"
+#include "smc_stats.h"
 
 static DEFINE_MUTEX(smc_server_lgr_pending);	/* serialize link group
 						 * creation on server
@@ -508,9 +509,44 @@ static void smc_link_save_peer_info(struct smc_link *link,
 	link->peer_mtu = clc->r0.qp_mtu;
 }
 
-static void smc_switch_to_fallback(struct smc_sock *smc)
+static void smc_stat_inc_fback_rsn_cnt(struct smc_sock *smc,
+				       struct smc_stats_fback *fback_arr)
+{
+	int cnt;
+
+	for (cnt = 0; cnt < SMC_MAX_FBACK_RSN_CNT; cnt++) {
+		if (fback_arr[cnt].fback_code == smc->fallback_rsn) {
+			fback_arr[cnt].count++;
+			break;
+		}
+		if (!fback_arr[cnt].fback_code) {
+			fback_arr[cnt].fback_code = smc->fallback_rsn;
+			fback_arr[cnt].count++;
+			break;
+		}
+	}
+}
+
+static void smc_stat_fallback(struct smc_sock *smc)
+{
+	struct net *net = sock_net(&smc->sk);
+
+	mutex_lock(&net->smc.mutex_fback_rsn);
+	if (smc->listen_smc) {
+		smc_stat_inc_fback_rsn_cnt(smc, net->smc.fback_rsn->srv);
+		net->smc.fback_rsn->srv_fback_cnt++;
+	} else {
+		smc_stat_inc_fback_rsn_cnt(smc, net->smc.fback_rsn->clnt);
+		net->smc.fback_rsn->clnt_fback_cnt++;
+	}
+	mutex_unlock(&net->smc.mutex_fback_rsn);
+}
+
+static void smc_switch_to_fallback(struct smc_sock *smc, int reason_code)
 {
 	smc->use_fallback = true;
+	smc->fallback_rsn = reason_code;
+	smc_stat_fallback(smc);
 	if (smc->sk.sk_socket && smc->sk.sk_socket->file) {
 		smc->clcsock->file = smc->sk.sk_socket->file;
 		smc->clcsock->file->private_data = smc->clcsock;
@@ -522,8 +558,7 @@ static void smc_switch_to_fallback(struct smc_sock *smc)
 /* fall back during connect */
 static int smc_connect_fallback(struct smc_sock *smc, int reason_code)
 {
-	smc_switch_to_fallback(smc);
-	smc->fallback_rsn = reason_code;
+	smc_switch_to_fallback(smc, reason_code);
 	smc_copy_sock_settings_to_clc(smc);
 	smc->connect_nonblock = 0;
 	if (smc->sk.sk_state == SMC_INIT)
@@ -535,9 +570,11 @@ static int smc_connect_fallback(struct smc_sock *smc, int reason_code)
 static int smc_connect_decline_fallback(struct smc_sock *smc, int reason_code,
 					u8 version)
 {
+	struct net *net = sock_net(&smc->sk);
 	int rc;
 
 	if (reason_code < 0) { /* error, fallback is not possible */
+		this_cpu_inc(net->smc.smc_stats->clnt_hshake_err_cnt);
 		if (smc->sk.sk_state == SMC_INIT)
 			sock_put(&smc->sk); /* passive closing */
 		return reason_code;
@@ -545,6 +582,7 @@ static int smc_connect_decline_fallback(struct smc_sock *smc, int reason_code,
 	if (reason_code != SMC_CLC_DECL_PEERDECL) {
 		rc = smc_clc_send_decline(smc, reason_code, version);
 		if (rc < 0) {
+			this_cpu_inc(net->smc.smc_stats->clnt_hshake_err_cnt);
 			if (smc->sk.sk_state == SMC_INIT)
 				sock_put(&smc->sk); /* passive closing */
 			return rc;
@@ -992,6 +1030,7 @@ static int __smc_connect(struct smc_sock *smc)
 	if (rc)
 		goto vlan_cleanup;
 
+	SMC_STAT_CLNT_SUCC_INC(sock_net(smc->clcsock->sk), aclc);
 	smc_connect_ism_vlan_cleanup(smc, ini);
 	kfree(buf);
 	kfree(ini);
@@ -1307,7 +1346,9 @@ static void smc_listen_out_connected(struct smc_sock *new_smc)
 static void smc_listen_out_err(struct smc_sock *new_smc)
 {
 	struct sock *newsmcsk = &new_smc->sk;
+	struct net *net = sock_net(newsmcsk);
 
+	this_cpu_inc(net->smc.smc_stats->srv_hshake_err_cnt);
 	if (newsmcsk->sk_state == SMC_INIT)
 		sock_put(&new_smc->sk); /* passive closing */
 	newsmcsk->sk_state = SMC_CLOSED;
@@ -1325,8 +1366,7 @@ static void smc_listen_decline(struct smc_sock *new_smc, int reason_code,
 		smc_listen_out_err(new_smc);
 		return;
 	}
-	smc_switch_to_fallback(new_smc);
-	new_smc->fallback_rsn = reason_code;
+	smc_switch_to_fallback(new_smc, reason_code);
 	if (reason_code && reason_code != SMC_CLC_DECL_PEERDECL) {
 		if (smc_clc_send_decline(new_smc, reason_code, version) < 0) {
 			smc_listen_out_err(new_smc);
@@ -1699,8 +1739,7 @@ static void smc_listen_work(struct work_struct *work)
 
 	/* check if peer is smc capable */
 	if (!tcp_sk(newclcsock->sk)->syn_smc) {
-		smc_switch_to_fallback(new_smc);
-		new_smc->fallback_rsn = SMC_CLC_DECL_PEERNOSMC;
+		smc_switch_to_fallback(new_smc, SMC_CLC_DECL_PEERNOSMC);
 		smc_listen_out_connected(new_smc);
 		return;
 	}
@@ -1778,6 +1817,7 @@ static void smc_listen_work(struct work_struct *work)
 	}
 	smc_conn_save_peer_info(new_smc, cclc);
 	smc_listen_out_connected(new_smc);
+	SMC_STAT_SERV_SUCC_INC(sock_net(newclcsock->sk), ini);
 	goto out_free;
 
 out_unlock:
@@ -1984,18 +2024,19 @@ static int smc_sendmsg(struct socket *sock, struct msghdr *msg, size_t len)
 
 	if (msg->msg_flags & MSG_FASTOPEN) {
 		if (sk->sk_state == SMC_INIT && !smc->connect_nonblock) {
-			smc_switch_to_fallback(smc);
-			smc->fallback_rsn = SMC_CLC_DECL_OPTUNSUPP;
+			smc_switch_to_fallback(smc, SMC_CLC_DECL_OPTUNSUPP);
 		} else {
 			rc = -EINVAL;
 			goto out;
 		}
 	}
 
-	if (smc->use_fallback)
+	if (smc->use_fallback) {
 		rc = smc->clcsock->ops->sendmsg(smc->clcsock, msg, len);
-	else
+	} else {
 		rc = smc_tx_sendmsg(smc, msg, len);
+		SMC_STAT_TX_PAYLOAD(smc, len, rc);
+	}
 out:
 	release_sock(sk);
 	return rc;
@@ -2030,6 +2071,7 @@ static int smc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
 	} else {
 		msg->msg_namelen = 0;
 		rc = smc_rx_recvmsg(smc, msg, NULL, len, flags);
+		SMC_STAT_RX_PAYLOAD(smc, rc, rc);
 	}
 
 out:
@@ -2194,8 +2236,7 @@ static int smc_setsockopt(struct socket *sock, int level, int optname,
 	case TCP_FASTOPEN_NO_COOKIE:
 		/* option not supported by SMC */
 		if (sk->sk_state == SMC_INIT && !smc->connect_nonblock) {
-			smc_switch_to_fallback(smc);
-			smc->fallback_rsn = SMC_CLC_DECL_OPTUNSUPP;
+			smc_switch_to_fallback(smc, SMC_CLC_DECL_OPTUNSUPP);
 		} else {
 			rc = -EINVAL;
 		}
@@ -2204,19 +2245,23 @@ static int smc_setsockopt(struct socket *sock, int level, int optname,
 		if (sk->sk_state != SMC_INIT &&
 		    sk->sk_state != SMC_LISTEN &&
 		    sk->sk_state != SMC_CLOSED) {
-			if (val)
+			if (val) {
+				SMC_STAT_INC(smc, ndly_cnt);
 				mod_delayed_work(smc->conn.lgr->tx_wq,
 						 &smc->conn.tx_work, 0);
 			}
+		}
 		break;
 	case TCP_CORK:
 		if (sk->sk_state != SMC_INIT &&
 		    sk->sk_state != SMC_LISTEN &&
 		    sk->sk_state != SMC_CLOSED) {
-			if (!val)
+			if (!val) {
+				SMC_STAT_INC(smc, cork_cnt);
 				mod_delayed_work(smc->conn.lgr->tx_wq,
 						 &smc->conn.tx_work, 0);
 			}
+		}
 		break;
 	case TCP_DEFER_ACCEPT:
 		smc->sockopt_defer_accept = val;
@@ -2338,11 +2383,13 @@ static ssize_t smc_sendpage(struct socket *sock, struct page *page,
 		goto out;
 	}
 	release_sock(sk);
-	if (smc->use_fallback)
+	if (smc->use_fallback) {
 		rc = kernel_sendpage(smc->clcsock, page, offset,
 				     size, flags);
-	else
+	} else {
+		SMC_STAT_INC(smc, sendpage_cnt);
 		rc = sock_no_sendpage(sock, page, offset, size, flags);
+	}
 
 out:
 	return rc;
@@ -2391,6 +2438,7 @@ static ssize_t smc_splice_read(struct socket *sock, loff_t *ppos,
 			flags = MSG_DONTWAIT;
 		else
 			flags = 0;
+		SMC_STAT_INC(smc, splice_cnt);
 		rc = smc_rx_recvmsg(smc, NULL, pipe, len, flags);
 	}
 out:
@@ -2479,6 +2527,16 @@ static void __net_exit smc_net_exit(struct net *net)
 	smc_pnet_net_exit(net);
 }
 
+static __net_init int smc_net_stat_init(struct net *net)
+{
+	return smc_stats_init(net);
+}
+
+static void __net_exit smc_net_stat_exit(struct net *net)
+{
+	smc_stats_exit(net);
+}
+
 static struct pernet_operations smc_net_ops = {
 	.init = smc_net_init,
 	.exit = smc_net_exit,
@@ -2486,6 +2544,11 @@ static struct pernet_operations smc_net_ops = {
 	.size = sizeof(struct smc_net),
 };
 
+static struct pernet_operations smc_net_stat_ops = {
+	.init = smc_net_stat_init,
+	.exit = smc_net_stat_exit,
+};
+
 static int __init smc_init(void)
 {
 	int rc;
@@ -2494,6 +2557,10 @@ static int __init smc_init(void)
 	if (rc)
 		return rc;
 
+	rc = register_pernet_subsys(&smc_net_stat_ops);
+	if (rc)
+		return rc;
+
 	smc_ism_init();
 	smc_clc_init();
 
@@ -2595,6 +2662,7 @@ static void __exit smc_exit(void)
 	proto_unregister(&smc_proto);
 	smc_pnet_exit();
 	smc_nl_exit();
+	unregister_pernet_subsys(&smc_net_stat_ops);
 	unregister_pernet_subsys(&smc_net_ops);
 	rcu_barrier();
 }

net/smc/smc_core.c

@@ -33,6 +33,7 @@
 #include "smc_close.h"
 #include "smc_ism.h"
 #include "smc_netlink.h"
+#include "smc_stats.h"
 
 #define SMC_LGR_NUM_INCR		256
 #define SMC_LGR_FREE_DELAY_SERV		(600 * HZ)
@@ -2029,6 +2030,7 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_smcd, bool is_rmb)
 	struct smc_link_group *lgr = conn->lgr;
 	struct list_head *buf_list;
 	int bufsize, bufsize_short;
+	bool is_dgraded = false;
 	struct mutex *lock;	/* lock buffer list */
 	int sk_buf_size;
 
@@ -2056,6 +2058,8 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_smcd, bool is_rmb)
 		/* check for reusable slot in the link group */
 		buf_desc = smc_buf_get_slot(bufsize_short, lock, buf_list);
 		if (buf_desc) {
+			SMC_STAT_RMB_SIZE(smc, is_smcd, is_rmb, bufsize);
+			SMC_STAT_BUF_REUSE(smc, is_smcd, is_rmb);
 			memset(buf_desc->cpu_addr, 0, bufsize);
 			break; /* found reusable slot */
 		}
@@ -2067,9 +2071,16 @@ static int __smc_buf_create(struct smc_sock *smc, bool is_smcd, bool is_rmb)
 
 		if (PTR_ERR(buf_desc) == -ENOMEM)
 			break;
-		if (IS_ERR(buf_desc))
+		if (IS_ERR(buf_desc)) {
+			if (!is_dgraded) {
+				is_dgraded = true;
+				SMC_STAT_RMB_DOWNGRADED(smc, is_smcd, is_rmb);
+			}
 			continue;
+		}
 
+		SMC_STAT_RMB_ALLOC(smc, is_smcd, is_rmb);
+		SMC_STAT_RMB_SIZE(smc, is_smcd, is_rmb, bufsize);
 		buf_desc->used = 1;
 		mutex_lock(lock);
 		list_add(&buf_desc->list, buf_list);

net/smc/smc_netlink.c

@@ -19,6 +19,7 @@
 #include "smc_core.h"
 #include "smc_ism.h"
 #include "smc_ib.h"
+#include "smc_stats.h"
 #include "smc_netlink.h"
 
 #define SMC_CMD_MAX_ATTR 1
@@ -55,6 +56,16 @@ static const struct genl_ops smc_gen_nl_ops[] = {
 		/* can be retrieved by unprivileged users */
 		.dumpit = smcr_nl_get_device,
 	},
+	{
+		.cmd = SMC_NETLINK_GET_STATS,
+		/* can be retrieved by unprivileged users */
+		.dumpit = smc_nl_get_stats,
+	},
+	{
+		.cmd = SMC_NETLINK_GET_FBACK_STATS,
+		/* can be retrieved by unprivileged users */
+		.dumpit = smc_nl_get_fback_stats,
+	},
 };
 
 static const struct nla_policy smc_gen_nl_policy[2] = {

net/smc/smc_netlink.h

@@ -18,7 +18,7 @@
 extern struct genl_family smc_gen_nl_family;
 
 struct smc_nl_dmp_ctx {
-	int pos[2];
+	int pos[3];
 };
 
 static inline struct smc_nl_dmp_ctx *smc_nl_dmp_ctx(struct netlink_callback *c)

net/smc/smc_rx.c

@@ -21,6 +21,7 @@
 #include "smc_cdc.h"
 #include "smc_tx.h" /* smc_tx_consumer_update() */
 #include "smc_rx.h"
+#include "smc_stats.h"
 
 /* callback implementation to wakeup consumers blocked with smc_rx_wait().
  * indirectly called by smc_cdc_msg_recv_action().
@@ -227,6 +228,7 @@ static int smc_rx_recv_urg(struct smc_sock *smc, struct msghdr *msg, int len,
 	    conn->urg_state == SMC_URG_READ)
 		return -EINVAL;
 
+	SMC_STAT_INC(smc, urg_data_cnt);
 	if (conn->urg_state == SMC_URG_VALID) {
 		if (!(flags & MSG_PEEK))
 			smc->conn.urg_state = SMC_URG_READ;
@@ -303,6 +305,12 @@ int smc_rx_recvmsg(struct smc_sock *smc, struct msghdr *msg,
 	timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
 	target = sock_rcvlowat(sk, flags & MSG_WAITALL, len);
 
+	readable = atomic_read(&conn->bytes_to_rcv);
+	if (readable >= conn->rmb_desc->len)
+		SMC_STAT_RMB_RX_FULL(smc, !conn->lnk);
+
+	if (len < readable)
+		SMC_STAT_RMB_RX_SIZE_SMALL(smc, !conn->lnk);
 	/* we currently use 1 RMBE per RMB, so RMBE == RMB base addr */
 	rcvbuf_base = conn->rx_off + conn->rmb_desc->cpu_addr;
 

net/smc/smc_stats.h

+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Shared Memory Communications over RDMA (SMC-R) and RoCE
+ *
+ * Macros for SMC statistics
+ *
+ * Copyright IBM Corp. 2021
+ *
+ * Author(s):  Guvenc Gulce
+ */
+
+#ifndef NET_SMC_SMC_STATS_H_
+#define NET_SMC_SMC_STATS_H_
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <linux/percpu.h>
+#include <linux/ctype.h>
+#include <linux/smc.h>
+
+#include "smc_clc.h"
+
+#define SMC_MAX_FBACK_RSN_CNT 30
+
+enum {
+	SMC_BUF_8K,
+	SMC_BUF_16K,
+	SMC_BUF_32K,
+	SMC_BUF_64K,
+	SMC_BUF_128K,
+	SMC_BUF_256K,
+	SMC_BUF_512K,
+	SMC_BUF_1024K,
+	SMC_BUF_G_1024K,
+	SMC_BUF_MAX,
+};
+
+struct smc_stats_fback {
+	int	fback_code;
+	u16	count;
+};
+
+struct smc_stats_rsn {
+	struct	smc_stats_fback srv[SMC_MAX_FBACK_RSN_CNT];
+	struct	smc_stats_fback clnt[SMC_MAX_FBACK_RSN_CNT];
+	u64			srv_fback_cnt;
+	u64			clnt_fback_cnt;
+};
+
+struct smc_stats_rmbcnt {
+	u64	buf_size_small_peer_cnt;
+	u64	buf_size_small_cnt;
+	u64	buf_full_peer_cnt;
+	u64	buf_full_cnt;
+	u64	reuse_cnt;
+	u64	alloc_cnt;
+	u64	dgrade_cnt;
+};
+
+struct smc_stats_memsize {
+	u64	buf[SMC_BUF_MAX];
+};
+
+struct smc_stats_tech {
+	struct smc_stats_memsize tx_rmbsize;
+	struct smc_stats_memsize rx_rmbsize;
+	struct smc_stats_memsize tx_pd;
+	struct smc_stats_memsize rx_pd;
+	struct smc_stats_rmbcnt rmb_tx;
+	struct smc_stats_rmbcnt rmb_rx;
+	u64			clnt_v1_succ_cnt;
+	u64			clnt_v2_succ_cnt;
+	u64			srv_v1_succ_cnt;
+	u64			srv_v2_succ_cnt;
+	u64			sendpage_cnt;
+	u64			urg_data_cnt;
+	u64			splice_cnt;
+	u64			cork_cnt;
+	u64			ndly_cnt;
+	u64			rx_bytes;
+	u64			tx_bytes;
+	u64			rx_cnt;
+	u64			tx_cnt;
+};
+
+struct smc_stats {
+	struct smc_stats_tech	smc[2];
+	u64			clnt_hshake_err_cnt;
+	u64			srv_hshake_err_cnt;
+};
+
+#define SMC_STAT_PAYLOAD_SUB(_smc_stats, _tech, key, _len, _rc) \
+do { \
+	typeof(_smc_stats) stats = (_smc_stats); \
+	typeof(_tech) t = (_tech); \
+	typeof(_len) l = (_len); \
+	int _pos = fls64((l) >> 13); \
+	typeof(_rc) r = (_rc); \
+	int m = SMC_BUF_MAX - 1; \
+	this_cpu_inc((*stats).smc[t].key ## _cnt); \
+	if (r <= 0) \
+		break; \
+	_pos = (_pos < m) ? ((l == 1 << (_pos + 12)) ? _pos - 1 : _pos) : m; \
+	this_cpu_inc((*stats).smc[t].key ## _pd.buf[_pos]); \
+	this_cpu_add((*stats).smc[t].key ## _bytes, r); \
+} \
+while (0)
+
+#define SMC_STAT_TX_PAYLOAD(_smc, length, rcode) \
+do { \
+	typeof(_smc) __smc = _smc; \
+	struct net *_net = sock_net(&__smc->sk); \
+	struct smc_stats __percpu *_smc_stats = _net->smc.smc_stats; \
+	typeof(length) _len = (length); \
+	typeof(rcode) _rc = (rcode); \
+	bool is_smcd = !__smc->conn.lnk; \
+	if (is_smcd) \
+		SMC_STAT_PAYLOAD_SUB(_smc_stats, SMC_TYPE_D, tx, _len, _rc); \
+	else \
+		SMC_STAT_PAYLOAD_SUB(_smc_stats, SMC_TYPE_R, tx, _len, _rc); \
+} \
+while (0)
+
+#define SMC_STAT_RX_PAYLOAD(_smc, length, rcode) \
+do { \
+	typeof(_smc) __smc = _smc; \
+	struct net *_net = sock_net(&__smc->sk); \
+	struct smc_stats __percpu *_smc_stats = _net->smc.smc_stats; \
+	typeof(length) _len = (length); \
+	typeof(rcode) _rc = (rcode); \
+	bool is_smcd = !__smc->conn.lnk; \
+	if (is_smcd) \
+		SMC_STAT_PAYLOAD_SUB(_smc_stats, SMC_TYPE_D, rx, _len, _rc); \
+	else \
+		SMC_STAT_PAYLOAD_SUB(_smc_stats, SMC_TYPE_R, rx, _len, _rc); \
+} \
+while (0)
+
+#define SMC_STAT_RMB_SIZE_SUB(_smc_stats, _tech, k, _len) \
+do { \
+	typeof(_len) _l = (_len); \
+	typeof(_tech) t = (_tech); \
+	int _pos = fls((_l) >> 13); \
+	int m = SMC_BUF_MAX - 1; \
+	_pos = (_pos < m) ? ((_l == 1 << (_pos + 12)) ? _pos - 1 : _pos) : m; \
+	this_cpu_inc((*(_smc_stats)).smc[t].k ## _rmbsize.buf[_pos]); \
+} \
+while (0)
+
+#define SMC_STAT_RMB_SUB(_smc_stats, type, t, key) \
+	this_cpu_inc((*(_smc_stats)).smc[t].rmb ## _ ## key.type ## _cnt)
+
+#define SMC_STAT_RMB_SIZE(_smc, _is_smcd, _is_rx, _len) \
+do { \
+	struct net *_net = sock_net(&(_smc)->sk); \
+	struct smc_stats __percpu *_smc_stats = _net->smc.smc_stats; \
+	typeof(_is_smcd) is_d = (_is_smcd); \
+	typeof(_is_rx) is_r = (_is_rx); \
+	typeof(_len) l = (_len); \
+	if ((is_d) && (is_r)) \
+		SMC_STAT_RMB_SIZE_SUB(_smc_stats, SMC_TYPE_D, rx, l); \
+	if ((is_d) && !(is_r)) \
+		SMC_STAT_RMB_SIZE_SUB(_smc_stats, SMC_TYPE_D, tx, l); \
+	if (!(is_d) && (is_r)) \
+		SMC_STAT_RMB_SIZE_SUB(_smc_stats, SMC_TYPE_R, rx, l); \
+	if (!(is_d) && !(is_r)) \
+		SMC_STAT_RMB_SIZE_SUB(_smc_stats, SMC_TYPE_R, tx, l); \
+} \
+while (0)
+
+#define SMC_STAT_RMB(_smc, type, _is_smcd, _is_rx) \
+do { \
+	struct net *net = sock_net(&(_smc)->sk); \
+	struct smc_stats __percpu *_smc_stats = net->smc.smc_stats; \
+	typeof(_is_smcd) is_d = (_is_smcd); \
+	typeof(_is_rx) is_r = (_is_rx); \
+	if ((is_d) && (is_r)) \
+		SMC_STAT_RMB_SUB(_smc_stats, type, SMC_TYPE_D, rx); \
+	if ((is_d) && !(is_r)) \
+		SMC_STAT_RMB_SUB(_smc_stats, type, SMC_TYPE_D, tx); \
+	if (!(is_d) && (is_r)) \
+		SMC_STAT_RMB_SUB(_smc_stats, type, SMC_TYPE_R, rx); \
+	if (!(is_d) && !(is_r)) \
+		SMC_STAT_RMB_SUB(_smc_stats, type, SMC_TYPE_R, tx); \
+} \
+while (0)
+
+#define SMC_STAT_BUF_REUSE(smc, is_smcd, is_rx) \
+	SMC_STAT_RMB(smc, reuse, is_smcd, is_rx)
+
+#define SMC_STAT_RMB_ALLOC(smc, is_smcd, is_rx) \
+	SMC_STAT_RMB(smc, alloc, is_smcd, is_rx)
+
+#define SMC_STAT_RMB_DOWNGRADED(smc, is_smcd, is_rx) \
+	SMC_STAT_RMB(smc, dgrade, is_smcd, is_rx)
+
+#define SMC_STAT_RMB_TX_PEER_FULL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_full_peer, is_smcd, false)
+
+#define SMC_STAT_RMB_TX_FULL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_full, is_smcd, false)
+
+#define SMC_STAT_RMB_TX_PEER_SIZE_SMALL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_size_small_peer, is_smcd, false)
+
+#define SMC_STAT_RMB_TX_SIZE_SMALL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_size_small, is_smcd, false)
+
+#define SMC_STAT_RMB_RX_SIZE_SMALL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_size_small, is_smcd, true)
+
+#define SMC_STAT_RMB_RX_FULL(smc, is_smcd) \
+	SMC_STAT_RMB(smc, buf_full, is_smcd, true)
+
+#define SMC_STAT_INC(_smc, type) \
+do { \
+	typeof(_smc) __smc = _smc; \
+	bool is_smcd = !(__smc)->conn.lnk; \
+	struct net *net = sock_net(&(__smc)->sk); \
+	struct smc_stats __percpu *smc_stats = net->smc.smc_stats; \
+	if ((is_smcd)) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_D].type); \
+	else \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_R].type); \
+} \
+while (0)
+
+#define SMC_STAT_CLNT_SUCC_INC(net, _aclc) \
+do { \
+	typeof(_aclc) acl = (_aclc); \
+	bool is_v2 = (acl->hdr.version == SMC_V2); \
+	bool is_smcd = (acl->hdr.typev1 == SMC_TYPE_D); \
+	struct smc_stats __percpu *smc_stats = (net)->smc.smc_stats; \
+	if (is_v2 && is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_D].clnt_v2_succ_cnt); \
+	else if (is_v2 && !is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_R].clnt_v2_succ_cnt); \
+	else if (!is_v2 && is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_D].clnt_v1_succ_cnt); \
+	else if (!is_v2 && !is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_R].clnt_v1_succ_cnt); \
+} \
+while (0)
+
+#define SMC_STAT_SERV_SUCC_INC(net, _ini) \
+do { \
+	typeof(_ini) i = (_ini); \
+	bool is_v2 = (i->smcd_version & SMC_V2); \
+	bool is_smcd = (i->is_smcd); \
+	typeof(net->smc.smc_stats) smc_stats = (net)->smc.smc_stats; \
+	if (is_v2 && is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_D].srv_v2_succ_cnt); \
+	else if (is_v2 && !is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_R].srv_v2_succ_cnt); \
+	else if (!is_v2 && is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_D].srv_v1_succ_cnt); \
+	else if (!is_v2 && !is_smcd) \
+		this_cpu_inc(smc_stats->smc[SMC_TYPE_R].srv_v1_succ_cnt); \
+} \
+while (0)
+
+int smc_nl_get_stats(struct sk_buff *skb, struct netlink_callback *cb);
+int smc_nl_get_fback_stats(struct sk_buff *skb, struct netlink_callback *cb);
+int smc_stats_init(struct net *net);
+void smc_stats_exit(struct net *net);
+
+#endif /* NET_SMC_SMC_STATS_H_ */

net/smc/smc_tx.c

@@ -27,6 +27,7 @@
 #include "smc_close.h"
 #include "smc_ism.h"
 #include "smc_tx.h"
+#include "smc_stats.h"
 
 #define SMC_TX_WORK_DELAY	0
 #define SMC_TX_CORK_DELAY	(HZ >> 2)	/* 250 ms */
@@ -45,6 +46,8 @@ static void smc_tx_write_space(struct sock *sk)
 
 	/* similar to sk_stream_write_space */
 	if (atomic_read(&smc->conn.sndbuf_space) && sock) {
+		if (test_bit(SOCK_NOSPACE, &sock->flags))
+			SMC_STAT_RMB_TX_FULL(smc, !smc->conn.lnk);
 		clear_bit(SOCK_NOSPACE, &sock->flags);
 		rcu_read_lock();
 		wq = rcu_dereference(sk->sk_wq);
@@ -151,6 +154,15 @@ int smc_tx_sendmsg(struct smc_sock *smc, struct msghdr *msg, size_t len)
 		goto out_err;
 	}
 
+	if (len > conn->sndbuf_desc->len)
+		SMC_STAT_RMB_TX_SIZE_SMALL(smc, !conn->lnk);
+
+	if (len > conn->peer_rmbe_size)
+		SMC_STAT_RMB_TX_PEER_SIZE_SMALL(smc, !conn->lnk);
+
+	if (msg->msg_flags & MSG_OOB)
+		SMC_STAT_INC(smc, urg_data_cnt);
+
 	while (msg_data_left(msg)) {
 		if (sk->sk_state == SMC_INIT)
 			return -ENOTCONN;
@@ -419,8 +431,12 @@ static int smc_tx_rdma_writes(struct smc_connection *conn,
 	/* destination: RMBE */
 	/* cf. snd_wnd */
 	rmbespace = atomic_read(&conn->peer_rmbe_space);
-	if (rmbespace <= 0)
+	if (rmbespace <= 0) {
+		struct smc_sock *smc = container_of(conn, struct smc_sock,
+						    conn);
+		SMC_STAT_RMB_TX_PEER_FULL(smc, !conn->lnk);
 		return 0;
+	}
 	smc_curs_copy(&prod, &conn->local_tx_ctrl.prod, conn);
 	smc_curs_copy(&cons, &conn->local_rx_ctrl.cons, conn);