Implement the RDMA nldev netlink interface for dumping detailed
MR information.
Signed-off-by: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
---
drivers/infiniband/core/nldev.c | 174 +++++++++++++++++++++++++++++++++++
drivers/infiniband/core/restrack.c | 36 +++++---
drivers/infiniband/core/uverbs_cmd.c | 2 +
drivers/infiniband/core/verbs.c | 3 +
include/rdma/ib_verbs.h | 5 +
include/rdma/restrack.h | 4 +
include/uapi/rdma/rdma_netlink.h | 10 ++
7 files changed, 221 insertions(+), 13 deletions(-)
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 32eda39..ace272b 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -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;
@@ -445,6 +453,51 @@ 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)
{
@@ -1099,6 +1152,124 @@ 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,
@@ -1131,6 +1302,9 @@ 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)
diff --git a/drivers/infiniband/core/restrack.c b/drivers/infiniband/core/restrack.c
index a85d8f5..58b05a9 100644
--- a/drivers/infiniband/core/restrack.c
+++ b/drivers/infiniband/core/restrack.c
@@ -43,22 +43,28 @@ int rdma_restrack_count(struct rdma_restrack_root *res,
static void set_kern_name(struct rdma_restrack_entry *res)
{
- enum rdma_restrack_type type = res->type;
- struct ib_qp *qp;
+ struct ib_pd *pd;
- if (type != RDMA_RESTRACK_QP)
- /* Other types already have this name embedded in */
- return;
-
- qp = container_of(res, struct ib_qp, res);
- if (!qp->pd) {
- WARN_ONCE(true, "XRC QPs are not supported\n");
- /* Survive, despite the programmer's error */
- res->kern_name = " ";
- return;
+ switch (res->type) {
+ case RDMA_RESTRACK_QP:
+ pd = container_of(res, struct ib_qp, res)->pd;
+ if (!pd) {
+ WARN_ONCE(true, "XRC QPs are not supported\n");
+ /* Survive, despite the programmer's error */
+ res->kern_name = " ";
+ }
+ break;
+ case RDMA_RESTRACK_MR:
+ pd = container_of(res, struct ib_mr, res)->pd;
+ break;
+ default:
+ /* Other types set kern_name directly */
+ pd = NULL;
+ break;
}
- res->kern_name = qp->pd->res.kern_name;
+ if (pd)
+ res->kern_name = pd->res.kern_name;
}
static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
@@ -73,6 +79,8 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
case RDMA_RESTRACK_CM_ID:
return container_of(res, struct rdma_id_private,
res)->id.device;
+ case RDMA_RESTRACK_MR:
+ return container_of(res, struct ib_mr, res)->device;
default:
WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
return NULL;
@@ -90,6 +98,8 @@ static bool res_is_user(struct rdma_restrack_entry *res)
return container_of(res, struct ib_qp, res)->uobject;
case RDMA_RESTRACK_CM_ID:
return !container_of(res, struct rdma_id_private, res)->caller;
+ case RDMA_RESTRACK_MR:
+ return container_of(res, struct ib_mr, res)->pd->uobject;
default:
WARN_ONCE(true, "Wrong resource tracking type %u\n", res->type);
return false;
diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 25a0e0e..99f01c4 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -693,6 +693,8 @@ 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;
diff --git a/drivers/infiniband/core/verbs.c b/drivers/infiniband/core/verbs.c
index 93025d2..d80dbe1 100644
--- a/drivers/infiniband/core/verbs.c
+++ b/drivers/infiniband/core/verbs.c
@@ -1622,6 +1622,7 @@ 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);
@@ -1658,6 +1659,8 @@ 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;
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 73b2387..7df3274 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -1772,6 +1772,11 @@ 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 {
diff --git a/include/rdma/restrack.h b/include/rdma/restrack.h
index bbb1a8f..2257cf5 100644
--- a/include/rdma/restrack.h
+++ b/include/rdma/restrack.h
@@ -33,6 +33,10 @@ 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
diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
index 49fa044..d21f174 100644
--- a/include/uapi/rdma/rdma_netlink.h
+++ b/include/uapi/rdma/rdma_netlink.h
@@ -242,6 +242,8 @@ enum rdma_nldev_command {
RDMA_NLDEV_CMD_RES_CQ_GET, /* can dump */
+ RDMA_NLDEV_CMD_RES_MR_GET, /* can dump */
+
RDMA_NLDEV_NUM_OPS
};
@@ -388,6 +390,14 @@ 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 */
--
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
