linux-rdma.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Mark Zhang <markzhang@nvidia.com>
To: <jgg@nvidia.com>, <dledford@redhat.com>, <saeedm@nvidia.com>
Cc: <linux-rdma@vger.kernel.org>, <netdev@vger.kernel.org>,
	<aharonl@nvidia.com>, <netao@nvidia.com>, <leonro@nvidia.com>,
	Mark Zhang <markzhang@nvidia.com>
Subject: [PATCH rdma-next 08/10] RDMA/nldev: Add support to get optional counters statistics
Date: Wed, 18 Aug 2021 14:24:26 +0300	[thread overview]
Message-ID: <20210818112428.209111-9-markzhang@nvidia.com> (raw)
In-Reply-To: <20210818112428.209111-1-markzhang@nvidia.com>

From: Neta Ostrovsky <netao@nvidia.com>

This patch adds the ability to return per-port optional counter
statistisc through RDMA netlink.

Examples:

$ rdma statistic show link rocep8s0f0/1
link rocep8s0f0/1 rx_write_requests 0 rx_read_requests 0 rx_atomic_requests 0 out_of_buffer 0
out_of_sequence 0 duplicate_request 0 rnr_nak_retry_err 0 packet_seq_err 0 implied_nak_seq_err 0
local_ack_timeout_err 0 resp_local_length_error 0 resp_cqe_error 0 req_cqe_error 0
req_remote_invalid_request 0 req_remote_access_errors 0 resp_remote_access_errors 0
resp_cqe_flush_error 0 req_cqe_flush_error 0 roce_adp_retrans 0 roce_adp_retrans_to 0
roce_slow_restart 0 roce_slow_restart_cnps 0 roce_slow_restart_trans 0 rp_cnp_ignored 0
rp_cnp_handled 0 np_ecn_marked_roce_packets 0 np_cnp_sent 0 rx_icrc_encapsulated 0
    Optional-set: cc_rx_ce_pkts 0

Signed-off-by: Aharon Landau <aharonl@nvidia.com>
Signed-off-by: Neta Ostrovsky <netao@nvidia.com>
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
---
 drivers/infiniband/core/counters.c |  18 +++++
 drivers/infiniband/core/device.c   |   1 +
 drivers/infiniband/core/nldev.c    | 117 +++++++++++++++++++++++------
 include/rdma/ib_verbs.h            |   3 +
 include/rdma/rdma_counter.h        |   3 +
 5 files changed, 119 insertions(+), 23 deletions(-)

diff --git a/drivers/infiniband/core/counters.c b/drivers/infiniband/core/counters.c
index fa04178aa0eb..5f7a12b8f1bb 100644
--- a/drivers/infiniband/core/counters.c
+++ b/drivers/infiniband/core/counters.c
@@ -459,6 +459,24 @@ u64 rdma_counter_get_hwstat_value(struct ib_device *dev, u32 port, u32 index)
 	return sum;
 }
 
+/*
+ * rdma_opcounter_query_stats - Query the per-port optional counter values
+ */
+int rdma_opcounter_query_stats(struct rdma_op_stats *opstats,
+			       struct ib_device *dev, u32 port)
+{
+	int ret = 0;
+
+	if (!dev->ops.get_op_stats)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&opstats->lock);
+	ret = dev->ops.get_op_stats(dev, port, opstats);
+	mutex_unlock(&opstats->lock);
+
+	return ret;
+}
+
 static struct ib_qp *rdma_counter_get_qp(struct ib_device *dev, u32 qp_num)
 {
 	struct rdma_restrack_entry *res = NULL;
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index b9138f20f9a8..efd4b75b7752 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -2657,6 +2657,7 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops)
 	SET_DEVICE_OP(dev_ops, get_link_layer);
 	SET_DEVICE_OP(dev_ops, get_netdev);
 	SET_DEVICE_OP(dev_ops, get_numa_node);
+	SET_DEVICE_OP(dev_ops, get_op_stats);
 	SET_DEVICE_OP(dev_ops, get_port_immutable);
 	SET_DEVICE_OP(dev_ops, get_vector_affinity);
 	SET_DEVICE_OP(dev_ops, get_vf_config);
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 17d55d89f11c..b665651dfb1d 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -945,6 +945,30 @@ int rdma_nl_stat_hwcounter_entry(struct sk_buff *msg, const char *name,
 }
 EXPORT_SYMBOL(rdma_nl_stat_hwcounter_entry);
 
+static int rdma_nl_stat_opcounter_entry(struct sk_buff *msg, const char *name,
+					u64 value)
+{
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_OPCOUNTER_ENTRY);
+	if (!entry_attr)
+		return -EMSGSIZE;
+
+	if (nla_put_string(msg, RDMA_NLDEV_ATTR_STAT_OPCOUNTER_ENTRY_NAME,
+			   name))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_STAT_OPCOUNTER_ENTRY_VALUE,
+			      value, RDMA_NLDEV_ATTR_PAD))
+		goto err;
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+	return -EMSGSIZE;
+}
+
 static int fill_stat_mr_entry(struct sk_buff *msg, bool has_cap_net_admin,
 			      struct rdma_restrack_entry *res, uint32_t port)
 {
@@ -2124,15 +2148,52 @@ static int nldev_stat_del_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return ret;
 }
 
+static int stat_get_optional_counter(struct sk_buff *msg,
+				     struct ib_device *device, u32 port)
+{
+	struct rdma_op_stats *opstats;
+	struct nlattr *opstats_table;
+	int i, ret = 0;
+
+	opstats = device->port_data[port].port_counter.opstats;
+	if (!opstats)
+		return 0;
+
+	ret = rdma_opcounter_query_stats(opstats, device, port);
+	if (ret)
+		return ret;
+
+	opstats_table = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_OPCOUNTERS);
+	if (!opstats_table)
+		return -EMSGSIZE;
+
+	for (i = 0; i < opstats->num_opcounters; i++) {
+		if (!(opstats->opcounters[i].enabled))
+			continue;
+		ret = rdma_nl_stat_opcounter_entry(msg,
+						   opstats->opcounters[i].name,
+						   opstats->opcounters[i].value);
+		if (ret)
+			goto err;
+	}
+	nla_nest_end(msg, opstats_table);
+
+	return 0;
+
+err:
+	nla_nest_cancel(msg, opstats_table);
+	return ret;
+}
+
 static int stat_get_doit_default_counter(struct sk_buff *skb,
 					 struct nlmsghdr *nlh,
 					 struct netlink_ext_ack *extack,
 					 struct nlattr *tb[])
 {
-	struct rdma_hw_stats *stats;
-	struct nlattr *table_attr;
+	struct rdma_hw_stats *hwstats;
+	struct nlattr *hwstats_table;
 	struct ib_device *device;
-	int ret, num_cnts, i;
+	int ret, num_hwstats, i;
 	struct sk_buff *msg;
 	u32 index, port;
 	u64 v;
@@ -2145,14 +2206,19 @@ static int stat_get_doit_default_counter(struct sk_buff *skb,
 	if (!device)
 		return -EINVAL;
 
+	port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
+	if (!rdma_is_port_valid(device, port)) {
+		ret = -EINVAL;
+		goto err;
+	}
+
 	if (!device->ops.alloc_hw_port_stats || !device->ops.get_hw_stats) {
 		ret = -EINVAL;
 		goto err;
 	}
 
-	port = nla_get_u32(tb[RDMA_NLDEV_ATTR_PORT_INDEX]);
-	stats = ib_get_hw_stats_port(device, port);
-	if (!stats) {
+	hwstats = ib_get_hw_stats_port(device, port);
+	if (!hwstats) {
 		ret = -EINVAL;
 		goto err;
 	}
@@ -2174,38 +2240,43 @@ static int stat_get_doit_default_counter(struct sk_buff *skb,
 		goto err_msg;
 	}
 
-	mutex_lock(&stats->lock);
+	mutex_lock(&hwstats->lock);
 
-	num_cnts = device->ops.get_hw_stats(device, stats, port, 0);
-	if (num_cnts < 0) {
+	num_hwstats = device->ops.get_hw_stats(device, hwstats, port, 0);
+	if (num_hwstats < 0) {
 		ret = -EINVAL;
-		goto err_stats;
+		goto err_hwstats;
 	}
 
-	table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS);
-	if (!table_attr) {
+	hwstats_table = nla_nest_start(msg, RDMA_NLDEV_ATTR_STAT_HWCOUNTERS);
+	if (!hwstats_table) {
 		ret = -EMSGSIZE;
-		goto err_stats;
+		goto err_hwstats;
 	}
-	for (i = 0; i < num_cnts; i++) {
-		v = stats->value[i] +
+	for (i = 0; i < num_hwstats; i++) {
+		v = hwstats->value[i] +
 			rdma_counter_get_hwstat_value(device, port, i);
-		if (rdma_nl_stat_hwcounter_entry(msg, stats->names[i], v)) {
+		if (rdma_nl_stat_hwcounter_entry(msg, hwstats->names[i], v)) {
 			ret = -EMSGSIZE;
-			goto err_table;
+			goto err_hwstats_table;
 		}
 	}
-	nla_nest_end(msg, table_attr);
+	nla_nest_end(msg, hwstats_table);
+
+	mutex_unlock(&hwstats->lock);
+
+	ret = stat_get_optional_counter(msg, device, port);
+	if (ret)
+		goto err_msg;
 
-	mutex_unlock(&stats->lock);
 	nlmsg_end(msg, nlh);
 	ib_device_put(device);
 	return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid);
 
-err_table:
-	nla_nest_cancel(msg, table_attr);
-err_stats:
-	mutex_unlock(&stats->lock);
+err_hwstats_table:
+	nla_nest_cancel(msg, hwstats_table);
+err_hwstats:
+	mutex_unlock(&hwstats->lock);
 err_msg:
 	nlmsg_free(msg);
 err:
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index fa9e668b9b14..d85f2e842a1d 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2602,6 +2602,9 @@ struct ib_device_ops {
 			   int optional_stat);
 	int (*remove_op_stat)(struct ib_device *device, u32 port,
 			      int optional_stat);
+	int (*get_op_stats)(struct ib_device *device, u32 port,
+			    struct rdma_op_stats *stats);
+
 	/**
 	 * Allows rdma drivers to add their own restrack attributes.
 	 */
diff --git a/include/rdma/rdma_counter.h b/include/rdma/rdma_counter.h
index 48086a7248ac..31686a234c77 100644
--- a/include/rdma/rdma_counter.h
+++ b/include/rdma/rdma_counter.h
@@ -67,5 +67,8 @@ int rdma_opcounter_add(struct ib_device *dev, u32 port,
 		       const char *name);
 int rdma_opcounter_remove(struct ib_device *dev, u32 port,
 			  const char *name);
+int rdma_opcounter_query_stats(struct rdma_op_stats *opstats,
+			       struct ib_device *dev, u32 port);
+
 
 #endif /* _RDMA_COUNTER_H_ */
-- 
2.26.2


  parent reply	other threads:[~2021-08-18 11:25 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-18 11:24 [PATCH rdma-next 00/10] Optional counter statistics support Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 01/10] net/mlx5: Add support in bth_opcode as a match criteria Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 02/10] net/mlx5: Add priorities for counters in RDMA namespaces Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 03/10] RDMA/counters: Support to allocate per-port optional counter statistics Mark Zhang
2021-08-23 19:30   ` Jason Gunthorpe
2021-08-24  6:22     ` Mark Zhang
2021-08-24 13:14       ` Jason Gunthorpe
2021-08-18 11:24 ` [PATCH rdma-next 04/10] RDMA/mlx5: Add alloc_op_port_stats() support Mark Zhang
2021-08-23 19:19   ` Jason Gunthorpe
2021-08-18 11:24 ` [PATCH rdma-next 05/10] RDMA/mlx5: Add steering support in optional flow counters Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 06/10] RDMA/nldev: Add support to add and remove optional counters Mark Zhang
2021-08-23 19:42   ` Jason Gunthorpe
2021-08-24  2:09     ` Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 07/10] RDMA/mlx5: Add add_op_stat() and remove_op_stat() support Mark Zhang
2021-08-18 11:24 ` Mark Zhang [this message]
2021-08-18 11:24 ` [PATCH rdma-next 09/10] RDMA/mlx5: Add get_op_stats() support Mark Zhang
2021-08-18 11:24 ` [PATCH rdma-next 10/10] RDMA/nldev: Add support to get current enabled optional counters Mark Zhang
2021-08-23 19:44   ` Jason Gunthorpe
2021-08-24  2:13     ` Mark Zhang
2021-08-24 13:13       ` Jason Gunthorpe
2021-08-23 19:33 ` [PATCH rdma-next 00/10] Optional counter statistics support Jason Gunthorpe
2021-08-24  1:44   ` Mark Zhang
2021-08-24 13:11     ` Jason Gunthorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210818112428.209111-9-markzhang@nvidia.com \
    --to=markzhang@nvidia.com \
    --cc=aharonl@nvidia.com \
    --cc=dledford@redhat.com \
    --cc=jgg@nvidia.com \
    --cc=leonro@nvidia.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=netao@nvidia.com \
    --cc=netdev@vger.kernel.org \
    --cc=saeedm@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).