net/smc: Add SMC statistics support

Add the ability to collect SMC statistics information. 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. Signed-off-by: Guvenc Gulce <[email protected]> Signed-off-by: Karsten Graul <[email protected]> Signed-off-by: David S. Miller <[email protected]>
angolini · Jun 16, 2021 · e0e4b8f · e0e4b8f
1 parent fb0a1da
commit e0e4b8f
Show file tree

Hide file tree

Showing 7 changed files with 395 additions and 21 deletions.
diff --git a/net/smc/Makefile b/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
diff --git a/net/smc/af_smc.c b/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 +509,42 @@ 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)
+{
+	mutex_lock(&smc_stat_fback_rsn);
+	if (smc->listen_smc) {
+		smc_stat_inc_fback_rsn_cnt(smc, fback_rsn.srv);
+		fback_rsn.srv_fback_cnt++;
+	} else {
+		smc_stat_inc_fback_rsn_cnt(smc, fback_rsn.clnt);
+		fback_rsn.clnt_fback_cnt++;
+	}
+	mutex_unlock(&smc_stat_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 +556,7 @@ 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)
@@ -538,13 +571,15 @@ static int smc_connect_decline_fallback(struct smc_sock *smc, int reason_code,
 	int rc;
 
 	if (reason_code < 0) { /* error, fallback is not possible */
+		this_cpu_inc(smc_stats->clnt_hshake_err_cnt);
 		if (smc->sk.sk_state == SMC_INIT)
 			sock_put(&smc->sk); /* passive closing */
 		return reason_code;
 	}
 	if (reason_code != SMC_CLC_DECL_PEERDECL) {
 		rc = smc_clc_send_decline(smc, reason_code, version);
 		if (rc < 0) {
+			this_cpu_inc(smc_stats->clnt_hshake_err_cnt);
 			if (smc->sk.sk_state == SMC_INIT)
 				sock_put(&smc->sk); /* passive closing */
 			return rc;
@@ -992,6 +1027,7 @@ static int __smc_connect(struct smc_sock *smc)
 	if (rc)
 		goto vlan_cleanup;
 
+	SMC_STAT_CLNT_SUCC_INC(aclc);
 	smc_connect_ism_vlan_cleanup(smc, ini);
 	kfree(buf);
 	kfree(ini);
@@ -1308,6 +1344,7 @@ static void smc_listen_out_err(struct smc_sock *new_smc)
 {
 	struct sock *newsmcsk = &new_smc->sk;
 
+	this_cpu_inc(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 +1362,7 @@ 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 +1735,7 @@ 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 +1813,7 @@ 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(ini);
 	goto out_free;
 
 out_unlock:
@@ -1984,18 +2020,19 @@ 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 +2067,7 @@ 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 +2232,7 @@ 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;
 		}
@@ -2204,18 +2241,22 @@ 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->conn.lnk, 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->conn.lnk, cork_cnt);
 				mod_delayed_work(smc->conn.lgr->tx_wq,
 						 &smc->conn.tx_work, 0);
+			}
 		}
 		break;
 	case TCP_DEFER_ACCEPT:
@@ -2338,11 +2379,13 @@ 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->conn.lnk, sendpage_cnt);
 		rc = sock_no_sendpage(sock, page, offset, size, flags);
+	}
 
 out:
 	return rc;
@@ -2391,6 +2434,7 @@ static ssize_t smc_splice_read(struct socket *sock, loff_t *ppos,
 			flags = MSG_DONTWAIT;
 		else
 			flags = 0;
+		SMC_STAT_INC(!smc->conn.lnk, splice_cnt);
 		rc = smc_rx_recvmsg(smc, NULL, pipe, len, flags);
 	}
 out:
@@ -2514,10 +2558,16 @@ static int __init smc_init(void)
 	if (!smc_close_wq)
 		goto out_alloc_hs_wq;
 
+	rc = smc_stats_init();
+	if (rc) {
+		pr_err("%s: smc_stats_init fails with %d\n", __func__, rc);
+		goto out_alloc_wqs;
+	}
+
 	rc = smc_core_init();
 	if (rc) {
 		pr_err("%s: smc_core_init fails with %d\n", __func__, rc);
-		goto out_alloc_wqs;
+		goto out_smc_stat;
 	}
 
 	rc = smc_llc_init();
@@ -2569,6 +2619,8 @@ static int __init smc_init(void)
 	proto_unregister(&smc_proto);
 out_core:
 	smc_core_exit();
+out_smc_stat:
+	smc_stats_exit();
 out_alloc_wqs:
 	destroy_workqueue(smc_close_wq);
 out_alloc_hs_wq:
@@ -2591,6 +2643,7 @@ static void __exit smc_exit(void)
 	smc_ib_unregister_client();
 	destroy_workqueue(smc_close_wq);
 	destroy_workqueue(smc_hs_wq);
+	smc_stats_exit();
 	proto_unregister(&smc_proto6);
 	proto_unregister(&smc_proto);
 	smc_pnet_exit();

diff --git a/net/smc/smc_core.c b/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(is_smcd, is_rmb, bufsize);
+			SMC_STAT_BUF_REUSE(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(is_smcd, is_rmb);
+			}
 			continue;
+		}
 
+		SMC_STAT_RMB_ALLOC(is_smcd, is_rmb);
+		SMC_STAT_RMB_SIZE(is_smcd, is_rmb, bufsize);
 		buf_desc->used = 1;
 		mutex_lock(lock);
 		list_add(&buf_desc->list, buf_list);

diff --git a/net/smc/smc_rx.c b/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(!conn->lnk, 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(!conn->lnk);
+
+	if (len < readable)
+		SMC_STAT_RMB_RX_SIZE_SMALL(!conn->lnk);
 	/* we currently use 1 RMBE per RMB, so RMBE == RMB base addr */
 	rcvbuf_base = conn->rx_off + conn->rmb_desc->cpu_addr;
 

diff --git a/net/smc/smc_stats.c b/net/smc/smc_stats.c
@@ -0,0 +1,35 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Shared Memory Communications over RDMA (SMC-R) and RoCE
+ *
+ * SMC statistics netlink routines
+ *
+ * Copyright IBM Corp. 2021
+ *
+ * Author(s):  Guvenc Gulce
+ */
+#include <linux/init.h>
+#include <linux/mutex.h>
+#include <linux/percpu.h>
+#include <linux/ctype.h>
+#include "smc_stats.h"
+
+/* serialize fallback reason statistic gathering */
+DEFINE_MUTEX(smc_stat_fback_rsn);
+struct smc_stats __percpu *smc_stats;	/* per cpu counters for SMC */
+struct smc_stats_reason fback_rsn;
+
+int __init smc_stats_init(void)
+{
+	memset(&fback_rsn, 0, sizeof(fback_rsn));
+	smc_stats = alloc_percpu(struct smc_stats);
+	if (!smc_stats)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void smc_stats_exit(void)
+{
+	free_percpu(smc_stats);
+}