[PATCH rdma-next 7/9] RDMA/nldev: Add command to get FRMR pools

[Edward Srouji <edwards@nvidia.com>]

From: Michael Guralnik <michaelgur@nvidia.com>

Add support for a new command in netlink to dump to user the state of
the FRMR pools on the devices.
Expose each pool with its key and the usage statistics for it.

Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Reviewed-by: Patrisious Haddad <phaddad@nvidia.com>
Signed-off-by: Edward Srouji <edwards@nvidia.com>
---
 drivers/infiniband/core/nldev.c  | 254 +++++++++++++++++++++++++++++++++++++++
 include/uapi/rdma/rdma_netlink.h |  17 +++
 2 files changed, 271 insertions(+)

diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 2220a2dfab240eaef2eb64d8e45cb221dfa25614..6cdf6073fdf9c51ee291a63bb86ac690b094aa9f 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -37,11 +37,13 @@
 #include <net/netlink.h>
 #include <rdma/rdma_cm.h>
 #include <rdma/rdma_netlink.h>
+#include <rdma/frmr_pools.h>
 
 #include "core_priv.h"
 #include "cma_priv.h"
 #include "restrack.h"
 #include "uverbs.h"
+#include "frmr_pools.h"
 
 /*
  * This determines whether a non-privileged user is allowed to specify a
@@ -172,6 +174,16 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
 	[RDMA_NLDEV_ATTR_NAME_ASSIGN_TYPE]	= { .type = NLA_U8 },
 	[RDMA_NLDEV_ATTR_EVENT_TYPE]		= { .type = NLA_U8 },
 	[RDMA_NLDEV_ATTR_STAT_OPCOUNTER_ENABLED] = { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_FRMR_POOLS]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_ENTRY]	= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_KEY]		= { .type = NLA_NESTED },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ATS]	= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ACCESS_FLAGS] = { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_KEY_VENDOR_KEY] = { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_KEY_NUM_DMA_BLOCKS] = { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_QUEUE_HANDLES] = { .type = NLA_U32 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_MAX_IN_USE]	= { .type = NLA_U64 },
+	[RDMA_NLDEV_ATTR_FRMR_POOL_IN_USE]	= { .type = NLA_U64 },
 };
 
 static int put_driver_name_print_type(struct sk_buff *msg, const char *name,
@@ -2637,6 +2649,244 @@ static int nldev_deldev(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return ib_del_sub_device_and_put(device);
 }
 
+static int fill_frmr_pool_key(struct sk_buff *msg, struct ib_frmr_key *key)
+{
+	struct nlattr *key_attr;
+
+	key_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_FRMR_POOL_KEY);
+	if (!key_attr)
+		return -EMSGSIZE;
+
+	if (nla_put_u8(msg, RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ATS, key->ats))
+		goto err;
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ACCESS_FLAGS,
+			key->access_flags))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_FRMR_POOL_KEY_VENDOR_KEY,
+			      key->vendor_key, RDMA_NLDEV_ATTR_PAD))
+		goto err;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_FRMR_POOL_KEY_NUM_DMA_BLOCKS,
+			      key->num_dma_blocks, RDMA_NLDEV_ATTR_PAD))
+		goto err;
+
+	nla_nest_end(msg, key_attr);
+	return 0;
+
+err:
+	return -EMSGSIZE;
+}
+
+static int fill_frmr_pool_entry(struct sk_buff *msg, struct ib_frmr_pool *pool)
+{
+	if (fill_frmr_pool_key(msg, &pool->key))
+		return -EMSGSIZE;
+
+	spin_lock(&pool->lock);
+	if (nla_put_u32(msg, RDMA_NLDEV_ATTR_FRMR_POOL_QUEUE_HANDLES,
+			pool->queue.ci + pool->inactive_queue.ci))
+		goto err_unlock;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_FRMR_POOL_MAX_IN_USE,
+			      pool->max_in_use, RDMA_NLDEV_ATTR_PAD))
+		goto err_unlock;
+	if (nla_put_u64_64bit(msg, RDMA_NLDEV_ATTR_FRMR_POOL_IN_USE,
+			      pool->in_use, RDMA_NLDEV_ATTR_PAD))
+		goto err_unlock;
+	spin_unlock(&pool->lock);
+
+	return 0;
+
+err_unlock:
+	spin_unlock(&pool->lock);
+	return -EMSGSIZE;
+}
+
+static int fill_frmr_pools_info(struct sk_buff *msg, struct ib_device *device)
+{
+	struct ib_frmr_pools *pools = device->frmr_pools;
+	struct ib_frmr_pool *pool;
+	struct nlattr *table_attr;
+	struct rb_node *node;
+
+	if (!pools)
+		return 0;
+
+	read_lock(&pools->rb_lock);
+	if (RB_EMPTY_ROOT(&pools->rb_root)) {
+		read_unlock(&pools->rb_lock);
+		return 0;
+	}
+	read_unlock(&pools->rb_lock);
+
+	table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_FRMR_POOLS);
+	if (!table_attr)
+		return -EMSGSIZE;
+
+	read_lock(&pools->rb_lock);
+	for (node = rb_first(&pools->rb_root); node; node = rb_next(node)) {
+		pool = rb_entry(node, struct ib_frmr_pool, node);
+		if (fill_frmr_pool_entry(msg, pool))
+			goto err;
+	}
+	read_unlock(&pools->rb_lock);
+
+	nla_nest_end(msg, table_attr);
+	return 0;
+
+err:
+	read_unlock(&pools->rb_lock);
+	nla_nest_cancel(msg, table_attr);
+	return -EMSGSIZE;
+}
+
+static int nldev_frmr_pools_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
+				     struct netlink_ext_ack *extack)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct ib_device *device;
+	struct sk_buff *msg;
+	u32 index;
+	int ret;
+
+	ret = __nlmsg_parse(nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1, nldev_policy,
+			    NL_VALIDATE_LIBERAL, extack);
+	if (ret || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	index = nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]);
+	device = ib_device_get_by_index(sock_net(skb->sk), index);
+	if (!device)
+		return -EINVAL;
+
+	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+	if (!msg) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	nlh = nlmsg_put(msg, NETLINK_CB(skb).portid, nlh->nlmsg_seq,
+			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV,
+					 RDMA_NLDEV_CMD_FRMR_POOLS_GET),
+			0, 0);
+	if (!nlh || fill_nldev_handle(msg, device)) {
+		ret = -EMSGSIZE;
+		goto err_free;
+	}
+
+	ret = fill_frmr_pools_info(msg, device);
+	if (ret)
+		goto err_free;
+
+	nlmsg_end(msg, nlh);
+	ib_device_put(device);
+
+	return rdma_nl_unicast(sock_net(skb->sk), msg, NETLINK_CB(skb).portid);
+
+err_free:
+	nlmsg_free(msg);
+err:
+	ib_device_put(device);
+	return ret;
+}
+
+static int nldev_frmr_pools_get_dumpit(struct sk_buff *skb,
+				       struct netlink_callback *cb)
+{
+	struct nlattr *tb[RDMA_NLDEV_ATTR_MAX];
+	struct ib_frmr_pools *pools;
+	int err, ret = 0, idx = 0;
+	struct ib_frmr_pool *pool;
+	struct nlattr *table_attr;
+	struct nlattr *entry_attr;
+	struct ib_device *device;
+	int start = cb->args[0];
+	struct rb_node *node;
+	struct nlmsghdr *nlh;
+	bool filled = false;
+
+	err = __nlmsg_parse(cb->nlh, 0, tb, RDMA_NLDEV_ATTR_MAX - 1,
+			    nldev_policy, NL_VALIDATE_LIBERAL, NULL);
+	if (err || !tb[RDMA_NLDEV_ATTR_DEV_INDEX])
+		return -EINVAL;
+
+	device = ib_device_get_by_index(
+		sock_net(skb->sk), nla_get_u32(tb[RDMA_NLDEV_ATTR_DEV_INDEX]));
+	if (!device)
+		return -EINVAL;
+
+	pools = device->frmr_pools;
+	if (!pools) {
+		ib_device_put(device);
+		return 0;
+	}
+
+	nlh = nlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
+			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV,
+					 RDMA_NLDEV_CMD_FRMR_POOLS_GET),
+			0, NLM_F_MULTI);
+
+	if (!nlh || fill_nldev_handle(skb, device)) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	table_attr = nla_nest_start_noflag(skb, RDMA_NLDEV_ATTR_FRMR_POOLS);
+	if (!table_attr) {
+		ret = -EMSGSIZE;
+		goto err;
+	}
+
+	read_lock(&pools->rb_lock);
+	for (node = rb_first(&pools->rb_root); node; node = rb_next(node)) {
+		pool = rb_entry(node, struct ib_frmr_pool, node);
+		if (pool->key.kernel_vendor_key)
+			continue;
+
+		if (idx < start) {
+			idx++;
+			continue;
+		}
+
+		filled = true;
+
+		entry_attr = nla_nest_start_noflag(
+			skb, RDMA_NLDEV_ATTR_FRMR_POOL_ENTRY);
+		if (!entry_attr) {
+			ret = -EMSGSIZE;
+			goto end_msg;
+		}
+
+		if (fill_frmr_pool_entry(skb, pool)) {
+			nla_nest_cancel(skb, entry_attr);
+			ret = -EMSGSIZE;
+			goto end_msg;
+		}
+
+		nla_nest_end(skb, entry_attr);
+		idx++;
+	}
+end_msg:
+	read_unlock(&pools->rb_lock);
+
+	nla_nest_end(skb, table_attr);
+	nlmsg_end(skb, nlh);
+	cb->args[0] = idx;
+
+	/*
+	 * No more entries to fill, cancel the message and
+	 * return 0 to mark end of dumpit.
+	 */
+	if (!filled)
+		goto err;
+
+	ib_device_put(device);
+	return skb->len;
+
+err:
+	nlmsg_cancel(skb, nlh);
+	ib_device_put(device);
+	return ret;
+}
+
 static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	[RDMA_NLDEV_CMD_GET] = {
 		.doit = nldev_get_doit,
@@ -2743,6 +2993,10 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 		.doit = nldev_deldev,
 		.flags = RDMA_NL_ADMIN_PERM,
 	},
+	[RDMA_NLDEV_CMD_FRMR_POOLS_GET] = {
+		.doit = nldev_frmr_pools_get_doit,
+		.dump = nldev_frmr_pools_get_dumpit,
+	},
 };
 
 static int fill_mon_netdev_rename(struct sk_buff *msg,
diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
index f41f0228fcd0e0b74e74b4d87611546b00f799a1..8f17ffe0190cb86131109209c45caec155ab36da 100644
--- a/include/uapi/rdma/rdma_netlink.h
+++ b/include/uapi/rdma/rdma_netlink.h
@@ -308,6 +308,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_MONITOR,
 
+	RDMA_NLDEV_CMD_FRMR_POOLS_GET, /* can dump */
+
 	RDMA_NLDEV_NUM_OPS
 };
 
@@ -582,6 +584,21 @@ enum rdma_nldev_attr {
 	RDMA_NLDEV_SYS_ATTR_MONITOR_MODE,	/* u8 */
 
 	RDMA_NLDEV_ATTR_STAT_OPCOUNTER_ENABLED,	/* u8 */
+
+	/*
+	 * FRMR Pools attributes
+	 */
+	RDMA_NLDEV_ATTR_FRMR_POOLS,		/* nested table */
+	RDMA_NLDEV_ATTR_FRMR_POOL_ENTRY,	/* nested table */
+	RDMA_NLDEV_ATTR_FRMR_POOL_KEY,		/* nested table */
+	RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ATS,	/* u8 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_KEY_ACCESS_FLAGS,	/* u32 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_KEY_VENDOR_KEY,	/* u64 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_KEY_NUM_DMA_BLOCKS,	/* u64 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_QUEUE_HANDLES,	/* u32 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_MAX_IN_USE,	/* u64 */
+	RDMA_NLDEV_ATTR_FRMR_POOL_IN_USE,	/* u64 */
+
 	/*
 	 * Always the end
 	 */

-- 
2.47.1