Implement RDMA nldev netlink interface to get detailed CM_ID information.
Because cm_id's are attached to rdma devices in various work queue
contexts, the pid and task information at device-attach time is sometimes
not useful. For example, an nvme/f host connection cm_id ends up being
bound to a device in a work queue context and the resulting pid at attach
time no longer exists after connection setup. So instead we mark all
cm_id's created via the rdma_ucm as "user", and all others as "kernel".
This required tweaking the restrack code a little. It also required
wrapping some rdma_cm functions to allow passing the module name string.
Signed-off-by: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
---
drivers/infiniband/core/cma.c | 59 ++++++++++++++++---------
drivers/infiniband/core/cma_priv.h | 6 +++
drivers/infiniband/core/nldev.c | 88 +++++++++++++++++++++++++++++++++++++-
drivers/infiniband/core/restrack.c | 10 ++++-
drivers/infiniband/core/ucma.c | 8 ++--
include/rdma/rdma_cm.h | 18 +++++---
include/rdma/restrack.h | 4 ++
include/uapi/rdma/rdma_netlink.h | 14 ++++++
8 files changed, 176 insertions(+), 31 deletions(-)
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 203519e..e76d0c1 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -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->res.type = RDMA_RESTRACK_CM_ID;
+ id_priv->res.kern_name = id_priv->caller;
+ rdma_restrack_add(&id_priv->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->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->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);
@@ -1778,6 +1785,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
struct ib_cm_event *ib_event,
struct net_device *net_dev)
{
+ struct rdma_id_private *listen_id_priv;
struct rdma_id_private *id_priv;
struct rdma_cm_id *id;
struct rdma_route *rt;
@@ -1787,9 +1795,11 @@ 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,
+ listen_id_priv = container_of(listen_id, struct rdma_id_private, id);
+ 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_priv->caller);
if (IS_ERR(id))
return NULL;
@@ -1838,14 +1848,16 @@ static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id,
struct ib_cm_event *ib_event,
struct net_device *net_dev)
{
+ struct rdma_id_private *listen_id_priv;
struct rdma_id_private *id_priv;
struct rdma_cm_id *id;
const sa_family_t ss_family = listen_id->route.addr.src_addr.ss_family;
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);
+ listen_id_priv = container_of(listen_id, struct rdma_id_private, id);
+ id = __rdma_create_id(net, listen_id->event_handler, listen_id->context,
+ listen_id->ps, IB_QPT_UD, listen_id_priv->caller);
if (IS_ERR(id))
return NULL;
@@ -2111,10 +2123,11 @@ 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->caller);
if (IS_ERR(new_cm_id)) {
ret = -ENOMEM;
goto out;
@@ -2239,8 +2252,8 @@ 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->caller);
if (IS_ERR(id))
return;
@@ -3348,8 +3361,10 @@ 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->res);
cma_release_dev(id_priv);
+ }
err1:
cma_comp_exch(id_priv, RDMA_CM_ADDR_BOUND, RDMA_CM_IDLE);
return ret;
@@ -3732,14 +3747,18 @@ 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->caller = caller;
+ else
+ id_priv->owner = task_pid_nr(current);
if (!cma_comp(id_priv, RDMA_CM_CONNECT))
return -EINVAL;
@@ -3779,7 +3798,7 @@ 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)
{
diff --git a/drivers/infiniband/core/cma_priv.h b/drivers/infiniband/core/cma_priv.h
index 11a41be..a3f208c 100644
--- a/drivers/infiniband/core/cma_priv.h
+++ b/drivers/infiniband/core/cma_priv.h
@@ -75,5 +75,11 @@ struct rdma_id_private {
u8 reuseaddr;
u8 afonly;
enum ib_gid_type gid_type;
+ const char *caller;
+
+ /*
+ * Internal to RDMA/core, don't use in the drivers
+ */
+ struct rdma_restrack_entry res;
};
#endif /* _CMA_PRIV_H */
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 4f1cfa6..46ce553 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -34,9 +34,11 @@
#include <linux/pid.h>
#include <linux/pid_namespace.h>
#include <net/netlink.h>
+#include <rdma/rdma_cm.h>
#include <rdma/rdma_netlink.h>
#include "core_priv.h"
+#include "cma_priv.h"
static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
[RDMA_NLDEV_ATTR_DEV_INDEX] = { .type = NLA_U32 },
@@ -71,6 +73,13 @@
[RDMA_NLDEV_ATTR_RES_PID] = { .type = NLA_U32 },
[RDMA_NLDEV_ATTR_RES_KERN_NAME] = { .type = NLA_NUL_STRING,
.len = TASK_COMM_LEN },
+ [RDMA_NLDEV_ATTR_RES_CM_ID] = { .type = NLA_NESTED },
+ [RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY] = { .type = NLA_NESTED },
+ [RDMA_NLDEV_ATTR_RES_PS] = { .type = NLA_U32 },
+ [RDMA_NLDEV_ATTR_RES_SRC_ADDR] = {
+ .len = sizeof(struct __kernel_sockaddr_storage) },
+ [RDMA_NLDEV_ATTR_RES_DST_ADDR] = {
+ .len = sizeof(struct __kernel_sockaddr_storage) },
};
static int fill_nldev_handle(struct sk_buff *msg, struct ib_device *device)
@@ -182,6 +191,7 @@ static int fill_res_info(struct sk_buff *msg, struct ib_device *device)
[RDMA_RESTRACK_PD] = "pd",
[RDMA_RESTRACK_CQ] = "cq",
[RDMA_RESTRACK_QP] = "qp",
+ [RDMA_RESTRACK_CM_ID] = "cm_id",
};
struct rdma_restrack_root *res = &device->res;
@@ -284,6 +294,66 @@ static int fill_res_qp_entry(struct sk_buff *msg, struct netlink_callback *cb,
return -EMSGSIZE;
}
+static int fill_res_cm_id_entry(struct sk_buff *msg,
+ struct netlink_callback *cb,
+ struct rdma_restrack_entry *res, uint32_t port)
+{
+ struct rdma_id_private *id_priv =
+ container_of(res, struct rdma_id_private, res);
+ struct rdma_cm_id *cm_id = &id_priv->id;
+ struct nlattr *entry_attr;
+
+ if (port && port != cm_id->port_num)
+ return 0;
+
+ entry_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_RES_CM_ID_ENTRY);
+ if (!entry_attr)
+ goto out;
+
+ if (cm_id->port_num &&
+ nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, cm_id->port_num))
+ goto err;
+
+ if (id_priv->qp_num &&
+ nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_LQPN, id_priv->qp_num))
+ goto err;
+
+ if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PS, cm_id->ps))
+ goto err;
+
+ if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_TYPE, cm_id->qp_type))
+ goto err;
+ if (nla_put_u8(msg, RDMA_NLDEV_ATTR_RES_STATE, id_priv->state))
+ goto err;
+
+ if (nla_put(msg, RDMA_NLDEV_ATTR_RES_SRC_ADDR,
+ sizeof(cm_id->route.addr.src_addr),
+ &cm_id->route.addr.src_addr))
+ goto err;
+ if (nla_put(msg, RDMA_NLDEV_ATTR_RES_DST_ADDR,
+ sizeof(cm_id->route.addr.dst_addr),
+ &cm_id->route.addr.dst_addr))
+ goto err;
+
+ if (id_priv->caller) {
+ if (nla_put_string(msg, RDMA_NLDEV_ATTR_RES_KERN_NAME,
+ id_priv->caller))
+ goto err;
+ } else {
+ /* CMA keeps the owning pid. */
+ if (nla_put_u32(msg, RDMA_NLDEV_ATTR_RES_PID, id_priv->owner))
+ 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)
{
@@ -583,7 +653,7 @@ static int res_get_common_dumpit(struct sk_buff *skb,
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 QP information,
+ * Right now, we are expecting the device index to get res 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.
@@ -707,6 +777,12 @@ static int res_get_common_dumpit(struct sk_buff *skb,
.nldev_cmd = RDMA_NLDEV_CMD_RES_QP_GET,
.nldev_attr = RDMA_NLDEV_ATTR_RES_QP,
},
+ [RDMA_RESTRACK_CM_ID] = {
+ .fill_res_func = fill_res_cm_id_entry,
+ .res_type = RDMA_RESTRACK_CM_ID,
+ .nldev_cmd = RDMA_NLDEV_CMD_RES_CM_ID_GET,
+ .nldev_attr = RDMA_NLDEV_ATTR_RES_CM_ID,
+ },
};
static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
@@ -715,6 +791,13 @@ static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
return res_get_common_dumpit(skb, cb, &fill_entries[RDMA_RESTRACK_QP]);
}
+static int nldev_res_get_cm_id_dumpit(struct sk_buff *skb,
+ struct netlink_callback *cb)
+{
+ return res_get_common_dumpit(skb, cb,
+ &fill_entries[RDMA_RESTRACK_CM_ID]);
+}
+
static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
[RDMA_NLDEV_CMD_GET] = {
.doit = nldev_get_doit,
@@ -741,6 +824,9 @@ static int nldev_res_get_qp_dumpit(struct sk_buff *skb,
* too.
*/
},
+ [RDMA_NLDEV_CMD_RES_CM_ID_GET] = {
+ .dump = nldev_res_get_cm_id_dumpit,
+ },
};
void __init nldev_init(void)
diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c
index 41a7800..a85d8f5 100644
--- a/drivers/infiniband/core/restrack.c
+++ b/drivers/infiniband/core/restrack.c
@@ -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);
@@ -67,6 +70,9 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
return container_of(res, struct ib_cq, res)->device;
case RDMA_RESTRACK_QP:
return container_of(res, struct ib_qp, res)->device;
+ case RDMA_RESTRACK_CM_ID:
+ return container_of(res, struct rdma_id_private,
+ res)->id.device;
default:
WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
return NULL;
@@ -82,6 +88,8 @@ static bool res_is_user(struct rdma_restrack_entry *res)
return container_of(res, struct ib_cq, res)->uobject;
case RDMA_RESTRACK_QP:
return container_of(res, struct ib_qp, res)->uobject;
+ case RDMA_RESTRACK_CM_ID:
+ return !container_of(res, struct rdma_id_private, res)->caller;
default:
WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
return false;
diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c
index f015f1b..4764626 100644
--- a/drivers/infiniband/core/ucma.c
+++ b/drivers/infiniband/core/ucma.c
@@ -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;
diff --git a/include/rdma/rdma_cm.h b/include/rdma/rdma_cm.h
index 6538a5c..62caae8 100644
--- a/include/rdma/rdma_cm.h
+++ b/include/rdma/rdma_cm.h
@@ -157,6 +157,11 @@ struct rdma_cm_id {
u8 port_num;
};
+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 +174,9 @@ 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 +288,9 @@ 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 +306,8 @@ 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
diff --git a/include/rdma/restrack.h b/include/rdma/restrack.h
index 2cdf8dc..bbb1a8f 100644
--- a/include/rdma/restrack.h
+++ b/include/rdma/restrack.h
@@ -29,6 +29,10 @@ enum rdma_restrack_type {
*/
RDMA_RESTRACK_QP,
/**
+ * @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
diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
index 4c77e2a..0399aed 100644
--- a/include/uapi/rdma/rdma_netlink.h
+++ b/include/uapi/rdma/rdma_netlink.h
@@ -238,6 +238,8 @@ 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
};
@@ -350,6 +352,18 @@ 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 socket addresses
+ */
+ RDMA_NLDEV_ATTR_RES_SRC_ADDR, /* __kernel_sockaddr_storage */
+ RDMA_NLDEV_ATTR_RES_DST_ADDR, /* __kernel_sockaddr_storage */
+
RDMA_NLDEV_ATTR_MAX
};
#endif /* _UAPI_RDMA_NETLINK_H */
--
1.8.3.1
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
