*** SUBJECT HERE ***

index 203519e..61ea800 100644
@@ -466,6 +466,9 @@ static void _cma_attach_to_dev(struct rdma_id_private *id_priv,
 	id_priv->id.route.addr.dev_addr.transport =
 		rdma_node_get_transport(cma_dev->device->node_type);
 	list_add_tail(&id_priv->list, &cma_dev->id_list);
+	id_priv->id.res.type = RDMA_RESTRACK_CM_ID;
+	id_priv->id.res.kern_name = id_priv->id.caller;
+	rdma_restrack_add(&id_priv->id.res);
 }
 
 static void cma_attach_to_dev(struct rdma_id_private *id_priv,

@@ -738,10 +741,10 @@ static void cma_deref_id(struct rdma_id_private *id_priv)
 		complete(&id_priv->comp);
 }
 
-struct rdma_cm_id *rdma_create_id(struct net *net,
-				  rdma_cm_event_handler event_handler,
-				  void *context, enum rdma_port_space ps,
-				  enum ib_qp_type qp_type)
+struct rdma_cm_id *__rdma_create_id(struct net *net,
+				    rdma_cm_event_handler event_handler,
+				    void *context, enum rdma_port_space ps,
+				    enum ib_qp_type qp_type, const char *caller)
 {
 	struct rdma_id_private *id_priv;
 

@@ -749,7 +752,10 @@ struct rdma_cm_id *rdma_create_id(struct net *net,
 	if (!id_priv)
 		return ERR_PTR(-ENOMEM);
 
-	id_priv->owner = task_pid_nr(current);
+	if (caller)
+		id_priv->id.caller = caller;
+	else
+		id_priv->owner = task_pid_nr(current);
 	id_priv->state = RDMA_CM_IDLE;
 	id_priv->id.context = context;
 	id_priv->id.event_handler = event_handler;

@@ -769,7 +775,7 @@ struct rdma_cm_id *rdma_create_id(struct net *net,
 
 	return &id_priv->id;
 }
-EXPORT_SYMBOL(rdma_create_id);
+EXPORT_SYMBOL(__rdma_create_id);
 
 static int cma_init_ud_qp(struct rdma_id_private *id_priv, struct ib_qp *qp)
 {

@@ -1629,6 +1635,7 @@ void rdma_destroy_id(struct rdma_cm_id *id)
 	mutex_unlock(&id_priv->handler_mutex);
 
 	if (id_priv->cma_dev) {
+		rdma_restrack_del(&id_priv->id.res);
 		if (rdma_cap_ib_cm(id_priv->id.device, 1)) {
 			if (id_priv->cm_id.ib)
 				ib_destroy_cm_id(id_priv->cm_id.ib);

@@ -1787,9 +1794,10 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
 		ib_event->param.req_rcvd.primary_path->service_id;
 	int ret;
 
-	id = rdma_create_id(listen_id->route.addr.dev_addr.net,
+	id = __rdma_create_id(listen_id->route.addr.dev_addr.net,
 			    listen_id->event_handler, listen_id->context,
-			    listen_id->ps, ib_event->param.req_rcvd.qp_type);
+			    listen_id->ps, ib_event->param.req_rcvd.qp_type,
+			    listen_id->caller);
 	if (IS_ERR(id))
 		return NULL;
 

@@ -1844,8 +1852,8 @@ static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id,
 	struct net *net = listen_id->route.addr.dev_addr.net;
 	int ret;
 
-	id = rdma_create_id(net, listen_id->event_handler, listen_id->context,
-			    listen_id->ps, IB_QPT_UD);
+	id = __rdma_create_id(net, listen_id->event_handler, listen_id->context,
+			      listen_id->ps, IB_QPT_UD, listen_id->caller);
 	if (IS_ERR(id))
 		return NULL;
 

@@ -2111,10 +2119,11 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		goto out;
 
 	/* Create a new RDMA id for the new IW CM ID */
-	new_cm_id = rdma_create_id(listen_id->id.route.addr.dev_addr.net,
-				   listen_id->id.event_handler,
-				   listen_id->id.context,
-				   RDMA_PS_TCP, IB_QPT_RC);
+	new_cm_id = __rdma_create_id(listen_id->id.route.addr.dev_addr.net,
+				     listen_id->id.event_handler,
+				     listen_id->id.context,
+				     RDMA_PS_TCP, IB_QPT_RC,
+				     listen_id->id.caller);
 	if (IS_ERR(new_cm_id)) {
 		ret = -ENOMEM;
 		goto out;

@@ -2239,8 +2248,8 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv,
 	if (cma_family(id_priv) == AF_IB && !rdma_cap_ib_cm(cma_dev->device, 1))
 		return;
 
-	id = rdma_create_id(net, cma_listen_handler, id_priv, id_priv->id.ps,
-			    id_priv->id.qp_type);
+	id = __rdma_create_id(net, cma_listen_handler, id_priv, id_priv->id.ps,
+			      id_priv->id.qp_type, id_priv->id.caller);
 	if (IS_ERR(id))
 		return;
 

@@ -3348,8 +3357,10 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
 
 	return 0;
 err2:
-	if (id_priv->cma_dev)
+	if (id_priv->cma_dev) {
+		rdma_restrack_del(&id_priv->id.res);
 		cma_release_dev(id_priv);
+	}
 err1:
 	cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE);
 	return ret;

@@ -3732,14 +3743,18 @@ static int cma_send_sidr_rep(struct rdma_id_private *id_priv,
 	return ib_send_cm_sidr_rep(id_priv->cm_id.ib, &rep);
 }
 
-int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
+int __rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param,
+		  const char *caller)
 {
 	struct rdma_id_private *id_priv;
 	int ret;
 
 	id_priv = container_of(id, struct rdma_id_private, id);
 
-	id_priv->owner = task_pid_nr(current);
+	if (caller)
+		id_priv->id.caller = caller;
+	else
+		id_priv->owner = task_pid_nr(current);
 
 	if (!cma_comp(id_priv, RDMA_CM_CONNECT))
 		return -EINVAL;

@@ -3779,7 +3794,7 @@ int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
 	rdma_reject(id, NULL, 0);
 	return ret;
 }
-EXPORT_SYMBOL(rdma_accept);
+EXPORT_SYMBOL(__rdma_accept);
 
 int rdma_notify(struct rdma_cm_id *id, enum ib_event_type event)
 {

285 lines not on this page · raw message

index 83bce7e..6385914 100644
@@ -3,12 +3,15 @@
  * Copyright (c) 2017-2018 Mellanox Technologies. All rights reserved.
  */
 
+#include <rdma/rdma_cm.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/restrack.h>
 #include <linux/mutex.h>
 #include <linux/sched/task.h>
 #include <linux/pid_namespace.h>
 
+#include "cma_priv.h"
+
 void rdma_restrack_init(struct rdma_restrack_root *res)
 {
 	init_rwsem(&res->rwsem);

@@ -44,7 +47,7 @@ static void set_kern_name(struct rdma_restrack_entry *res)
 	struct ib_qp *qp;
 
 	if (type != RDMA_RESTRACK_QP)
-		/* PD and CQ types already have this name embedded in */
+		/* Other types already have this name embedded in */
 		return;
 
 	qp = container_of(res, struct ib_qp, res);

@@ -61,6 +64,7 @@ static void set_kern_name(struct rdma_restrack_entry *res)
 static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 {
 	enum rdma_restrack_type type = res->type;
+	struct rdma_cm_id *cm_id;
 	struct ib_device *dev;
 	struct ib_xrcd *xrcd;
 	struct ib_pd *pd;

@@ -84,6 +88,10 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 		xrcd = container_of(res, struct ib_xrcd, res);
 		dev = xrcd->device;
 		break;
+	case RDMA_RESTRACK_CM_ID:
+		cm_id = container_of(res, struct rdma_cm_id, res);
+		dev = cm_id->device;
+		break;
 	default:
 		WARN_ONCE(true, "Wrong resource tracking type %u\n", type);
 		return NULL;

@@ -95,6 +103,7 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 static bool res_is_user(struct rdma_restrack_entry *res)
 {
 	enum rdma_restrack_type type = res->type;
+	struct rdma_cm_id *cm_id;
 	struct ib_xrcd *xrcd;
 	struct ib_pd *pd;
 	struct ib_cq *cq;

@@ -119,6 +128,10 @@ static bool res_is_user(struct rdma_restrack_entry *res)
 		xrcd = container_of(res, struct ib_xrcd, res);
 		is_user = xrcd->inode;
 		break;
+	case RDMA_RESTRACK_CM_ID:
+		cm_id = container_of(res, struct rdma_cm_id, res);
+		is_user = !cm_id->caller;
+		break;
 	default:
 		WARN_ONCE(true, "Wrong resource tracking type %u\n", type);
 	}

index d67219d..f7f0282 100644
@@ -476,8 +476,8 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
 		return -ENOMEM;
 
 	ctx->uid = cmd.uid;
-	ctx->cm_id = rdma_create_id(current->nsproxy->net_ns,
-				    ucma_event_handler, ctx, cmd.ps, qp_type);
+	ctx->cm_id = __rdma_create_id(current->nsproxy->net_ns,
+			      ucma_event_handler, ctx, cmd.ps, qp_type, NULL);
 	if (IS_ERR(ctx->cm_id)) {
 		ret = PTR_ERR(ctx->cm_id);
 		goto err1;

@@ -1084,12 +1084,12 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
 	if (cmd.conn_param.valid) {
 		ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param);
 		mutex_lock(&file->mut);
-		ret = rdma_accept(ctx->cm_id, &conn_param);
+		ret = __rdma_accept(ctx->cm_id, &conn_param, NULL);
 		if (!ret)
 			ctx->uid = cmd.uid;
 		mutex_unlock(&file->mut);
 	} else
-		ret = rdma_accept(ctx->cm_id, NULL);
+		ret = __rdma_accept(ctx->cm_id, NULL, NULL);
 
 	ucma_put_ctx(ctx);
 	return ret;

index 6538a5c..3e90501 100644
@@ -155,8 +155,19 @@ struct rdma_cm_id {
 	enum rdma_port_space	 ps;
 	enum ib_qp_type		 qp_type;
 	u8			 port_num;
+	const char *caller;
+
+	/*
+	 * Internal to RDMA/core, don't use in the drivers
+	 */
+	struct rdma_restrack_entry     res;
 };
 
+struct rdma_cm_id *__rdma_create_id(struct net *net,
+				  rdma_cm_event_handler event_handler,
+				  void *context, enum rdma_port_space ps,
+				  enum ib_qp_type qp_type, const char *caller);
+
 /**
  * rdma_create_id - Create an RDMA identifier.
  *

@@ -169,10 +180,9 @@ struct rdma_cm_id {
  *
  * The id holds a reference on the network namespace until it is destroyed.
  */
-struct rdma_cm_id *rdma_create_id(struct net *net,
-				  rdma_cm_event_handler event_handler,
-				  void *context, enum rdma_port_space ps,
-				  enum ib_qp_type qp_type);
+#define rdma_create_id(net, event_handler, context, ps, qp_type) \
+	__rdma_create_id((net), (event_handler), (context), (ps), (qp_type), \
+			 KBUILD_MODNAME)
 
 /**
   * rdma_destroy_id - Destroys an RDMA identifier.

@@ -284,6 +294,9 @@ int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
  */
 int rdma_listen(struct rdma_cm_id *id, int backlog);
 
+int __rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param,
+		  const char *caller);
+
 /**
  * rdma_accept - Called to accept a connection request or response.
  * @id: Connection identifier associated with the request.

@@ -299,7 +312,8 @@ int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
  * state of the qp associated with the id is modified to error, such that any
  * previously posted receive buffers would be flushed.
  */
-int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param);
+#define rdma_accept(id, conn_param) \
+	__rdma_accept((id), (conn_param),  KBUILD_MODNAME)
 
 /**
  * rdma_notify - Notifies the RDMA CM of an asynchronous event that has

index c2d8116..a794e0e 100644
@@ -33,6 +33,10 @@ enum rdma_restrack_type {
 	 */
 	RDMA_RESTRACK_XRCD,
 	/**
+	 * @RDMA_RESTRACK_CM_ID: Connection Manager ID (CM_ID)
+	 */
+	RDMA_RESTRACK_CM_ID,
+	/**
 	 * @RDMA_RESTRACK_MAX: Last entry, used for array dclarations
 	 */
 	RDMA_RESTRACK_MAX

index 17e59be..13f0bed 100644
@@ -240,6 +240,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_RES_QP_GET, /* can dump */
 
+	RDMA_NLDEV_CMD_RES_CM_ID_GET, /* can dump */
+
 	RDMA_NLDEV_NUM_OPS
 };
 

@@ -352,6 +354,34 @@ enum rdma_nldev_attr {
 	 */
 	RDMA_NLDEV_ATTR_RES_KERN_NAME,		/* string */
 
+	RDMA_NLDEV_ATTR_RES_CM_ID,		/* nested table */
+	RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY,	/* nested table */
+	/*
+	 * rdma_cm_id port space.
+	 */
+	RDMA_NLDEV_ATTR_RES_PS,			/* u32 */
+	/*
+	 * Source and destination IP address and port attributes.
+	 */
+	RDMA_NLDEV_ATTR_RES_IPV4_SADDR,		/* u8[4] */
+	RDMA_NLDEV_ATTR_RES_IPV4_DADDR,		/* u8[4] */
+	RDMA_NLDEV_ATTR_RES_IPV6_SADDR,		/* u8[16] */
+	RDMA_NLDEV_ATTR_RES_IPV6_DADDR,		/* u8[16] */
+	RDMA_NLDEV_ATTR_RES_IP_SPORT,		/* u16 */
+	RDMA_NLDEV_ATTR_RES_IP_DPORT,		/* u16 */
+	/*
+	 * ARPHRD_INFINIBAND, ARPHRD_ETHER, ...
+	 */
+	RDMA_NLDEV_ATTR_RES_DEV_TYPE,		/* u8 */
+	/*
+	 * enum enum rdma_transport_type (IB, IWARP, ...)
+	 */
+	RDMA_NLDEV_ATTR_RES_TRANSPORT_TYPE,	/* u8 */
+	/*
+	 * enum rdma_network_type (IB, IPv4, IPv6,...)
+	 */
+	RDMA_NLDEV_ATTR_RES_NETWORK_TYPE,	/* u8 */
+
 	RDMA_NLDEV_ATTR_MAX
 };
 #endif /* _UAPI_RDMA_NETLINK_H */

index e66963c..203519e 100644
@@ -62,6 +62,7 @@
 #include <rdma/iw_cm.h>
 
 #include "core_priv.h"
+#include "cma_priv.h"
 
 MODULE_AUTHOR("Sean Hefty");
 MODULE_DESCRIPTION("Generic RDMA CM Agent");

@@ -327,46 +328,6 @@ struct ib_device *cma_get_ib_dev(struct cma_device *cma_dev)
  * We do this by disabling removal notification while a callback is in process,
  * and reporting it after the callback completes.
  */
-struct rdma_id_private {
-	struct rdma_cm_id	id;
-
-	struct rdma_bind_list	*bind_list;
-	struct hlist_node	node;
-	struct list_head	list; /* listen_any_list or cma_device.list */
-	struct list_head	listen_list; /* per device listens */
-	struct cma_device	*cma_dev;
-	struct list_head	mc_list;
-
-	int			internal_id;
-	enum rdma_cm_state	state;
-	spinlock_t		lock;
-	struct mutex		qp_mutex;
-
-	struct completion	comp;
-	atomic_t		refcount;
-	struct mutex		handler_mutex;
-
-	int			backlog;
-	int			timeout_ms;
-	struct ib_sa_query	*query;
-	int			query_id;
-	union {
-		struct ib_cm_id	*ib;
-		struct iw_cm_id	*iw;
-	} cm_id;
-
-	u32			seq_num;
-	u32			qkey;
-	u32			qp_num;
-	pid_t			owner;
-	u32			options;
-	u8			srq;
-	u8			tos;
-	bool			tos_set;
-	u8			reuseaddr;
-	u8			afonly;
-	enum ib_gid_type	gid_type;
-};
 
 struct cma_multicast {
 	struct rdma_id_private *id_priv;

75 lines not on this page · raw message

index 13f5c46..34fb0d3 100644
@@ -89,6 +89,11 @@
 	[RDMA_NLDEV_ATTR_RES_DEV_TYPE]		= { .type = NLA_U8 },
 	[RDMA_NLDEV_ATTR_RES_TRANSPORT_TYPE]	= { .type = NLA_U8 },
 	[RDMA_NLDEV_ATTR_RES_NETWORK_TYPE]	= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_RES_CQ]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_CQ_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_CQE]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_USECNT]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_POLL_CTX]		= { .type = NLA_U8 },
 };
 
 static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)

@@ -396,6 +401,51 @@ static int fill_res_cm_id_entry(struct sk_buff *msg,
 	return -EMSGSIZE;
 }
 
+static int fill_res_cq_entry(struct sk_buff *msg,
+			     struct ib_cq *cq)
+{
+	struct rdma_restrack_entry *res = &cq->res;
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_CQ_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_CQE, cq->cqe))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT,
+			      atomic_read(&cq->usecnt), 0))
+		goto err;
+
+	/* Poll context is only valid for kernel CQs */
+	if (rdma_is_kernel_res(res) &&
+	    nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_POLL_CTX, cq->poll_ctx))
+		goto err;
+
+	/*
+	 * Existence of task means that it is user CQ and netlink
+	 * user is invited to go and read /proc/PID/comm to get name
+	 * of the task file and res->task_com should be NULL.
+	 */
+	if (rdma_is_kernel_res(res)) {
+		if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME,
+				   res->kern_name))
+			goto err;
+	} else {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID,
+				task_pid_vnr(res->task)))
+			goto err;
+	}
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
 static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {

@@ -929,6 +979,125 @@ static int nldev_res_get_cm_id_dumpit(struct sk_buff *skb,
 	put_device(&device->dev);
 	return ret;
 }
+
+static int nldev_res_get_cq_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct rdma_restrack_entry *res;
+	int err, ret = 0, idx = 0;
+	struct nlattr *table_attr;
+	struct ib_device *device;
+	int start = cb->args[0];
+	struct ib_cq *cq = NULL;
+	struct nlmsghdr *nlh;
+	u32 index;
+
+	err = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+			  nldev_policy, NULL);
+	/*
+	 * Right now, we are expecting the device index to get CQ information,
+	 * but it is possible to extend this code to return all devices in
+	 * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX.
+	 * if it doesn't exist, we will iterate over all devices.
+	 *
+	 * But it is not needed for now.
+	 */
+	if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(index);
+	if (!device)
+		return -EINVAL;
+
+	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
+		RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_CQ_GET),
+		0, NLM_F_MULTI);
+
+	if (fill_nldev_handle(skb, device)) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	table_attr = nla_nest_start(skb, RDMA_NLDEV_ATTR_RES_CQ);
+	if (!table_attr) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	down_read(&device->res.rwsem);
+	hash_for_each_possible(device->res.hash, res, node, RDMA_RESTRACK_CQ) {
+		if (idx < start)
+			goto next;
+
+		if ((rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) != &init_pid_ns) ||
+		    (!rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) !=
+		     task_active_pid_ns(res->task)))
+			/*
+			 * 1. Kernel CQs should be visible in init namspace only
+			 * 2. Present only CQs visible in the current namespace
+			 */
+			goto next;
+
+		if (!rdma_restrack_get(res))
+			/*
+			 * Resource is under release now, but we are not
+			 * relesing lock now, so it will be released in
+			 * our next pass, once we will get ->next pointer.
+			 */
+			goto next;
+
+		cq = container_of(res, struct ib_cq, res);
+
+		up_read(&device->res.rwsem);
+		ret = fill_res_cq_entry(skb, cq);
+		down_read(&device->res.rwsem);
+		/*
+		 * Return resource back, but it won't be released till
+		 * the &device->res.rwsem will be released for write.
+		 */
+		rdma_restrack_put(res);
+
+		if (ret == -EMSGSIZE)
+			/*
+			 * There is a chance to optimize here.
+			 * It can be done by using list_prepare_entry
+			 * and list_for_each_entry_continue afterwards.
+			 */
+			break;
+		if (ret)
+			goto res_err;
+next:		idx++;
+	}
+	up_read(&device->res.rwsem);
+
+	nla_nest_end(skb, table_attr);
+	nlmsg_end(skb, nlh);
+	cb->args[0] = idx;
+
+	/*
+	 * No more CQs to fill, cancel the message and
+	 * return 0 to mark end of dumpit.
+	 */
+	if (!cq)
+		goto err;
+
+	put_device(&device->dev);
+	return skb->len;
+
+res_err:
+	nla_nest_cancel(skb, table_attr);
+	up_read(&device->res.rwsem);
+
+err:
+	nlmsg_cancel(skb, nlh);
+	put_device(&device->dev);
+	return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,

@@ -958,6 +1127,9 @@ static int nldev_res_get_cm_id_dumpit(struct sk_buff *skb,
 	[RDMA_NLDEV_CMD_RES_CM_ID_GET] = {
 		.dump = nldev_res_get_cm_id_dumpit,
 	},
+	[RDMA_NLDEV_CMD_RES_CQ_GET] = {
+		.dump = nldev_res_get_cq_dumpit,
+	},
 };
 
 void __init nldev_init(void)

index 13f0bed..fa677ef 100644
@@ -242,6 +242,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_RES_CM_ID_GET, /* can dump */
 
+	RDMA_NLDEV_CMD_RES_CQ_GET, /* can dump */
+
 	RDMA_NLDEV_NUM_OPS
 };
 

@@ -382,6 +384,12 @@ enum rdma_nldev_attr {
 	 */
 	RDMA_NLDEV_ATTR_RES_NETWORK_TYPE,	/* u8 */
 
+	RDMA_NLDEV_ATTR_RES_CQ,			/* nested table */
+	RDMA_NLDEV_ATTR_RES_CQ_ENTRY,		/* nested table */
+	RDMA_NLDEV_ATTR_RES_CQE,		/* u32 */
+	RDMA_NLDEV_ATTR_RES_USECNT,		/* u64 */
+	RDMA_NLDEV_ATTR_RES_POLL_CTX,		/* u8 */
+
 	RDMA_NLDEV_ATTR_MAX
 };
 #endif /* _UAPI_RDMA_NETLINK_H */

index 34fb0d3..8d96f3e 100644
@@ -94,6 +94,13 @@
 	[RDMA_NLDEV_ATTR_RES_CQE]		= { .type = NLA_U32 },
 	[RDMA_NLDEV_ATTR_RES_USECNT]		= { .type = NLA_U64 },
 	[RDMA_NLDEV_ATTR_RES_POLL_CTX]		= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_RES_MR]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_MR_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_RKEY]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_LKEY]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_IOVA]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_MRLEN]		= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_RES_PGSIZE]		= { .type = NLA_U32 },
 };
 
 static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)

@@ -206,6 +213,7 @@ static int fill_res_info(struct sk_buff *msg, struct ib_device *device)
 		[RDMA_RESTRACK_CQ] = "cq",
 		[RDMA_RESTRACK_QP] = "qp",
 		[RDMA_RESTRACK_CM_ID] = "cm_id",
+		[RDMA_RESTRACK_MR] = "mr",
 	};
 
 	struct rdma_restrack_root *res = &device->res;

@@ -446,6 +454,51 @@ static int fill_res_cq_entry(struct sk_buff *msg,
 	return -EMSGSIZE;
 }
 
+static int fill_res_mr_entry(struct sk_buff *msg,
+			     struct ib_mr *mr)
+{
+	struct rdma_restrack_entry *res = &mr->res;
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_MR_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_RKEY, mr->rkey))
+		goto err;
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LKEY, mr->lkey))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_IOVA, mr->iova, 0))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_MRLEN, mr->length, 0))
+		goto err;
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PGSIZE, mr->page_size))
+		goto err;
+
+	/*
+	 * Existence of task means that it is user MR and netlink
+	 * user is invited to go and read /proc/PID/comm to get name
+	 * of the task file and res->task_com should be NULL.
+	 */
+	if (rdma_is_kernel_res(res)) {
+		if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME,
+				   res->kern_name))
+			goto err;
+	} else {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID,
+				task_pid_vnr(res->task)))
+			goto err;
+	}
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
 static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {

@@ -1098,6 +1151,124 @@ static int nldev_res_get_cq_dumpit(struct sk_buff *skb,
 	return ret;
 }
 
+static int nldev_res_get_mr_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct rdma_restrack_entry *res;
+	int err, ret = 0, idx = 0;
+	struct nlattr *table_attr;
+	struct ib_device *device;
+	int start = cb->args[0];
+	struct ib_mr *mr = NULL;
+	struct nlmsghdr *nlh;
+	u32 index;
+
+	err = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+			  nldev_policy, NULL);
+	/*
+	 * Right now, we are expecting the device index to get MR information,
+	 * but it is possible to extend this code to return all devices in
+	 * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX.
+	 * if it doesn't exist, we will iterate over all devices.
+	 *
+	 * But it is not needed for now.
+	 */
+	if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(index);
+	if (!device)
+		return -EINVAL;
+
+	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
+		RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_MR_GET),
+		0, NLM_F_MULTI);
+
+	if (fill_nldev_handle(skb, device)) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	table_attr = nla_nest_start(skb, RDMA_NLDEV_ATTR_RES_MR);
+	if (!table_attr) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	down_read(&device->res.rwsem);
+	hash_for_each_possible(device->res.hash, res, node, RDMA_RESTRACK_MR) {
+		if (idx < start)
+			goto next;
+
+		if ((rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) != &init_pid_ns) ||
+		    (!rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) !=
+		     task_active_pid_ns(res->task)))
+			/*
+			 * 1. Kernel MRs should be visible in init namspace only
+			 * 2. Present only MRs visible in the current namespace
+			 */
+			goto next;
+
+		if (!rdma_restrack_get(res))
+			/*
+			 * Resource is under release now, but we are not
+			 * relesing lock now, so it will be released in
+			 * our next pass, once we will get ->next pointer.
+			 */
+			goto next;
+
+		mr = container_of(res, struct ib_mr, res);
+
+		up_read(&device->res.rwsem);
+		ret = fill_res_mr_entry(skb, mr);
+		down_read(&device->res.rwsem);
+		/*
+		 * Return resource back, but it won't be released till
+		 * the &device->res.rwsem will be released for write.
+		 */
+		rdma_restrack_put(res);
+
+		if (ret == -EMSGSIZE)
+			/*
+			 * There is a chance to optimize here.
+			 * It can be done by using list_prepare_entry
+			 * and list_for_each_entry_continue afterwards.
+			 */
+			break;
+		if (ret)
+			goto res_err;
+next:		idx++;
+	}
+	up_read(&device->res.rwsem);
+
+	nla_nest_end(skb, table_attr);
+	nlmsg_end(skb, nlh);
+	cb->args[0] = idx;
+
+	/*
+	 * No more MRs to fill, cancel the message and
+	 * return 0 to mark end of dumpit.
+	 */
+	if (!mr)
+		goto err;
+
+	put_device(&device->dev);
+	return skb->len;
+
+res_err:
+	nla_nest_cancel(skb, table_attr);
+	up_read(&device->res.rwsem);
+
+err:
+	nlmsg_cancel(skb, nlh);
+	put_device(&device->dev);
+	return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,

@@ -1130,6 +1301,9 @@ static int nldev_res_get_cq_dumpit(struct sk_buff *skb,
 	[RDMA_NLDEV_CMD_RES_CQ_GET] = {
 		.dump = nldev_res_get_cq_dumpit,
 	},
+	[RDMA_NLDEV_CMD_RES_MR_GET] = {
+		.dump = nldev_res_get_mr_dumpit,
+	},
 };
 
 void __init nldev_init(void)

82 lines not on this page · raw message

index 256934d..3f026c4 100644
@@ -694,6 +694,8 @@ ssize_t ib_uverbs_reg_mr(struct ib_uverbs_file *file,
 	mr->pd      = pd;
 	mr->uobject = uobj;
 	atomic_inc(&pd->usecnt);
+	mr->res.type = RDMA_RESTRACK_MR;
+	rdma_restrack_add(&mr->res);
 
 	uobj->object = mr;
 

@@ -819,6 +821,7 @@ ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file,
 	struct ib_uverbs_dereg_mr cmd;
 	struct ib_uobject	 *uobj;
 	int                       ret = -EINVAL;
+	struct ib_mr *mr;
 
 	if (copy_from_user(&cmd, buf, sizeof cmd))
 		return -EFAULT;

@@ -828,6 +831,9 @@ ssize_t ib_uverbs_dereg_mr(struct ib_uverbs_file *file,
 	if (IS_ERR(uobj))
 		return PTR_ERR(uobj);
 
+	mr = uobj->object;
+	rdma_restrack_del(&mr->res);
+
 	ret = uobj_remove_commit(uobj);
 
 	return ret ?: in_len;

index 16ebc63..c3265f7 100644
@@ -1623,6 +1623,7 @@ int ib_dereg_mr(struct ib_mr *mr)
 	struct ib_pd *pd = mr->pd;
 	int ret;
 
+	rdma_restrack_del(&mr->res);
 	ret = mr->device->dereg_mr(mr);
 	if (!ret)
 		atomic_dec(&pd->usecnt);

@@ -1659,6 +1660,8 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
 		mr->uobject = NULL;
 		atomic_inc(&pd->usecnt);
 		mr->need_inval = false;
+		mr->res.type = RDMA_RESTRACK_MR;
+		rdma_restrack_add(&mr->res);
 	}
 
 	return mr;

index 5263c86..33d8c5d 100644
@@ -1771,6 +1771,11 @@ struct ib_mr {
 		struct ib_uobject	*uobject;	/* user */
 		struct list_head	qp_entry;	/* FR */
 	};
+
+	/*
+	 * Implementation details of the RDMA core, don't use in drivers:
+	 */
+	struct rdma_restrack_entry res;
 };
 
 struct ib_mw {

index a794e0e..bfd1140 100644
@@ -37,6 +37,10 @@ enum rdma_restrack_type {
 	 */
 	RDMA_RESTRACK_CM_ID,
 	/**
+	 * @RDMA_RESTRACK_MR: Memory Region (MR)
+	 */
+	RDMA_RESTRACK_MR,
+	/**
 	 * @RDMA_RESTRACK_MAX: Last entry, used for array dclarations
 	 */
 	RDMA_RESTRACK_MAX

index fa677ef..6adaeaa 100644
@@ -244,6 +244,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_RES_CQ_GET, /* can dump */
 
+	RDMA_NLDEV_CMD_RES_MR_GET, /* can dump */
+
 	RDMA_NLDEV_NUM_OPS
 };
 

@@ -390,6 +392,14 @@ enum rdma_nldev_attr {
 	RDMA_NLDEV_ATTR_RES_USECNT,		/* u64 */
 	RDMA_NLDEV_ATTR_RES_POLL_CTX,		/* u8 */
 
+	RDMA_NLDEV_ATTR_RES_MR,			/* nested table */
+	RDMA_NLDEV_ATTR_RES_MR_ENTRY,		/* nested table */
+	RDMA_NLDEV_ATTR_RES_RKEY,		/* u32 */
+	RDMA_NLDEV_ATTR_RES_LKEY,		/* u32 */
+	RDMA_NLDEV_ATTR_RES_IOVA,		/* u64 */
+	RDMA_NLDEV_ATTR_RES_MRLEN,		/* u64 */
+	RDMA_NLDEV_ATTR_RES_PGSIZE,		/* u32 */
+
 	RDMA_NLDEV_ATTR_MAX
 };
 #endif /* _UAPI_RDMA_NETLINK_H */

index 7e0eb20..e90f2fd 100644
@@ -391,6 +391,9 @@ static int finish_mem_reg(struct c4iw_mr *mhp, u32 stag)
 	mhp->attr.stag = stag;
 	mmid = stag >> 8;
 	mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
+	mhp->ibmr.length = mhp->attr.len;
+	mhp->ibmr.iova = mhp->attr.va_fbo;
+	mhp->ibmr.page_size = 1U << (mhp->attr.page_size + 12);
 	pr_debug("mmid 0x%x mhp %p\n", mmid, mhp);
 	return insert_handle(mhp->rhp, &mhp->rhp->mmidr, mhp, mmid);
 }

index 8d96f3e..42cbec0 100644
@@ -101,6 +101,11 @@
 	[RDMA_NLDEV_ATTR_RES_IOVA]		= { .type = NLA_U64 },
 	[RDMA_NLDEV_ATTR_RES_MRLEN]		= { .type = NLA_U64 },
 	[RDMA_NLDEV_ATTR_RES_PGSIZE]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_PD]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_PD_ENTRY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY]	= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_PD_FLAGS]		= { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY] = { .type = NLA_U32 },
 };
 
 static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)

@@ -499,6 +504,53 @@ static int fill_res_mr_entry(struct sk_buff *msg,
 	return -EMSGSIZE;
 }
 
+static int fill_res_pd_entry(struct sk_buff *msg,
+			     struct ib_pd *pd)
+{
+	struct rdma_restrack_entry *res = &pd->res;
+	struct nlattr *entry_attr;
+
+	entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_PD_ENTRY);
+	if (!entry_attr)
+		goto out;
+
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY,
+			pd->local_dma_lkey))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_RES_USECNT,
+			      atomic_read(&pd->usecnt), 0))
+		goto err;
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PD_FLAGS, pd->flags))
+		goto err;
+	if ((pd->flags & IB_PD_UNSAFE_GLOBAL_RKEY) &&
+	    nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY,
+			pd->unsafe_global_rkey))
+		goto err;
+
+	/*
+	 * Existence of task means that it is user PD and netlink
+	 * user is invited to go and read /proc/PID/comm to get name
+	 * of the task file and res->task_com should be NULL.
+	 */
+	if (rdma_is_kernel_res(res)) {
+		if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME,
+				   res->kern_name))
+			goto err;
+	} else {
+		if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID,
+				task_pid_vnr(res->task)))
+			goto err;
+	}
+
+	nla_nest_end(msg, entry_attr);
+	return 0;
+
+err:
+	nla_nest_cancel(msg, entry_attr);
+out:
+	return -EMSGSIZE;
+}
+
 static int nldev_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {

@@ -1269,6 +1321,124 @@ static int nldev_res_get_mr_dumpit(struct sk_buff *skb,
 	return ret;
 }
 
+static int nldev_res_get_pd_dumpit(struct sk_buff *skb,
+				   struct netlink_callback *cb)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct rdma_restrack_entry *res;
+	int err, ret = 0, idx = 0;
+	struct nlattr *table_attr;
+	struct ib_device *device;
+	int start = cb->args[0];
+	struct ib_pd *pd = NULL;
+	struct nlmsghdr *nlh;
+	u32 index;
+
+	err = nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+			  nldev_policy, NULL);
+	/*
+	 * Right now, we are expecting the device index to get PD information,
+	 * but it is possible to extend this code to return all devices in
+	 * one shot by checking the existence of RDMA_NLDEV_ATTR_DEV_INDEX.
+	 * if it doesn't exist, we will iterate over all devices.
+	 *
+	 * But it is not needed for now.
+	 */
+	if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(index);
+	if (!device)
+		return -EINVAL;
+
+	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
+		RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_RES_PD_GET),
+		0, NLM_F_MULTI);
+
+	if (fill_nldev_handle(skb, device)) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	table_attr = nla_nest_start(skb, RDMA_NLDEV_ATTR_RES_PD);
+	if (!table_attr) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	down_read(&device->res.rwsem);
+	hash_for_each_possible(device->res.hash, res, node, RDMA_RESTRACK_PD) {
+		if (idx < start)
+			goto next;
+
+		if ((rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) != &init_pid_ns) ||
+		    (!rdma_is_kernel_res(res) &&
+		     task_active_pid_ns(current) !=
+		     task_active_pid_ns(res->task)))
+			/*
+			 * 1. Kernel PDs should be visible in init namspace only
+			 * 2. Present only PDs visible in the current namespace
+			 */
+			goto next;
+
+		if (!rdma_restrack_get(res))
+			/*
+			 * Resource is under release now, but we are not
+			 * relesing lock now, so it will be released in
+			 * our next pass, once we will get ->next pointer.
+			 */
+			goto next;
+
+		pd = container_of(res, struct ib_pd, res);
+
+		up_read(&device->res.rwsem);
+		ret = fill_res_pd_entry(skb, pd);
+		down_read(&device->res.rwsem);
+		/*
+		 * Return resource back, but it won't be released till
+		 * the &device->res.rwsem will be released for write.
+		 */
+		rdma_restrack_put(res);
+
+		if (ret == -EMSGSIZE)
+			/*
+			 * There is a chance to optimize here.
+			 * It can be done by using list_prepare_entry
+			 * and list_for_each_entry_continue afterwards.
+			 */
+			break;
+		if (ret)
+			goto res_err;
+next:		idx++;
+	}
+	up_read(&device->res.rwsem);
+
+	nla_nest_end(skb, table_attr);
+	nlmsg_end(skb, nlh);
+	cb->args[0] = idx;
+
+	/*
+	 * No more PDs to fill, cancel the message and
+	 * return 0 to mark end of dumpit.
+	 */
+	if (!pd)
+		goto err;
+
+	put_device(&device->dev);
+	return skb->len;
+
+res_err:
+	nla_nest_cancel(skb, table_attr);
+	up_read(&device->res.rwsem);
+
+err:
+	nlmsg_cancel(skb, nlh);
+	put_device(&device->dev);
+	return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,

@@ -1304,6 +1474,9 @@ static int nldev_res_get_mr_dumpit(struct sk_buff *skb,
 	[RDMA_NLDEV_CMD_RES_MR_GET] = {
 		.dump = nldev_res_get_mr_dumpit,
 	},
+	[RDMA_NLDEV_CMD_RES_PD_GET] = {
+		.dump = nldev_res_get_pd_dumpit,
+	},
 };
 
 void __init nldev_init(void)

index 6adaeaa..4aeb59c 100644
@@ -246,6 +246,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_RES_MR_GET, /* can dump */
 
+	RDMA_NLDEV_CMD_RES_PD_GET, /* can dump */
+
 	RDMA_NLDEV_NUM_OPS
 };
 

@@ -400,6 +402,12 @@ enum rdma_nldev_attr {
 	RDMA_NLDEV_ATTR_RES_MRLEN,		/* u64 */
 	RDMA_NLDEV_ATTR_RES_PGSIZE,		/* u32 */
 
+	RDMA_NLDEV_ATTR_RES_PD,			/* nested table */
+	RDMA_NLDEV_ATTR_RES_PD_ENTRY,		/* nested table */
+	RDMA_NLDEV_ATTR_RES_LOCAL_DMA_LKEY,	/* u32 */
+	RDMA_NLDEV_ATTR_RES_PD_FLAGS,		/* u32 */
+	RDMA_NLDEV_ATTR_RES_UNSAFE_GLOBAL_RKEY,	/* u32 */
+
 	RDMA_NLDEV_ATTR_MAX
 };
 #endif /* _UAPI_RDMA_NETLINK_H */