net: Fix offloading indirect devices dependency on qdisc order creation

Currently, when creating an ingress qdisc on an indirect device before
the driver registered for callbacks, the driver will not have a chance
to register its filter configuration callbacks.

To fix that, modify the code such that it keeps track of all the ingress
qdiscs that call flow_indr_dev_setup_offload(). When a driver calls
flow_indr_dev_register(),  go through the list of tracked ingress qdiscs
and call the driver callback entry point so as to give it a chance to
register its callback.

Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/net/flow_offload.h

@@ -453,6 +453,7 @@ struct flow_block_offload {
 	struct list_head *driver_block_list;
 	struct netlink_ext_ack *extack;
 	struct Qdisc *sch;
+	struct list_head *cb_list_head;
 };
 
 enum tc_setup_type;

net/core/flow_offload.c

@@ -321,6 +321,7 @@ EXPORT_SYMBOL(flow_block_cb_setup_simple);
 static DEFINE_MUTEX(flow_indr_block_lock);
 static LIST_HEAD(flow_block_indr_list);
 static LIST_HEAD(flow_block_indr_dev_list);
+static LIST_HEAD(flow_indir_dev_list);
 
 struct flow_indr_dev {
 	struct list_head		list;
@@ -345,6 +346,33 @@ static struct flow_indr_dev *flow_indr_dev_alloc(flow_indr_block_bind_cb_t *cb,
 	return indr_dev;
 }
 
+struct flow_indir_dev_info {
+	void *data;
+	struct net_device *dev;
+	struct Qdisc *sch;
+	enum tc_setup_type type;
+	void (*cleanup)(struct flow_block_cb *block_cb);
+	struct list_head list;
+	enum flow_block_command command;
+	enum flow_block_binder_type binder_type;
+	struct list_head *cb_list;
+};
+
+static void existing_qdiscs_register(flow_indr_block_bind_cb_t *cb, void *cb_priv)
+{
+	struct flow_block_offload bo;
+	struct flow_indir_dev_info *cur;
+
+	list_for_each_entry(cur, &flow_indir_dev_list, list) {
+		memset(&bo, 0, sizeof(bo));
+		bo.command = cur->command;
+		bo.binder_type = cur->binder_type;
+		INIT_LIST_HEAD(&bo.cb_list);
+		cb(cur->dev, cur->sch, cb_priv, cur->type, &bo, cur->data, cur->cleanup);
+		list_splice(&bo.cb_list, cur->cb_list);
+	}
+}
+
 int flow_indr_dev_register(flow_indr_block_bind_cb_t *cb, void *cb_priv)
 {
 	struct flow_indr_dev *indr_dev;
@@ -366,6 +394,7 @@ int flow_indr_dev_register(flow_indr_block_bind_cb_t *cb, void *cb_priv)
 	}
 
 	list_add(&indr_dev->list, &flow_block_indr_dev_list);
+	existing_qdiscs_register(cb, cb_priv);
 	mutex_unlock(&flow_indr_block_lock);
 
 	return 0;
@@ -462,7 +491,59 @@ out:
 }
 EXPORT_SYMBOL(flow_indr_block_cb_alloc);
 
-int flow_indr_dev_setup_offload(struct net_device *dev, struct Qdisc *sch,
+static struct flow_indir_dev_info *find_indir_dev(void *data)
+{
+	struct flow_indir_dev_info *cur;
+
+	list_for_each_entry(cur, &flow_indir_dev_list, list) {
+		if (cur->data == data)
+			return cur;
+	}
+	return NULL;
+}
+
+static int indir_dev_add(void *data, struct net_device *dev, struct Qdisc *sch,
+			 enum tc_setup_type type, void (*cleanup)(struct flow_block_cb *block_cb),
+			 struct flow_block_offload *bo)
+{
+	struct flow_indir_dev_info *info;
+
+	info = find_indir_dev(data);
+	if (info)
+		return -EEXIST;
+
+	info = kzalloc(sizeof(*info), GFP_KERNEL);
+	if (!info)
+		return -ENOMEM;
+
+	info->data = data;
+	info->dev = dev;
+	info->sch = sch;
+	info->type = type;
+	info->cleanup = cleanup;
+	info->command = bo->command;
+	info->binder_type = bo->binder_type;
+	info->cb_list = bo->cb_list_head;
+
+	list_add(&info->list, &flow_indir_dev_list);
+	return 0;
+}
+
+static int indir_dev_remove(void *data)
+{
+	struct flow_indir_dev_info *info;
+
+	info = find_indir_dev(data);
+	if (!info)
+		return -ENOENT;
+
+	list_del(&info->list);
+
+	kfree(info);
+	return 0;
+}
+
+int flow_indr_dev_setup_offload(struct net_device *dev,	struct Qdisc *sch,
 				enum tc_setup_type type, void *data,
 				struct flow_block_offload *bo,
 				void (*cleanup)(struct flow_block_cb *block_cb))
@@ -470,6 +551,12 @@ int flow_indr_dev_setup_offload(struct net_device *dev, struct Qdisc *sch,
 	struct flow_indr_dev *this;
 
 	mutex_lock(&flow_indr_block_lock);
+
+	if (bo->command == FLOW_BLOCK_BIND)
+		indir_dev_add(data, dev, sch, type, cleanup, bo);
+	else if (bo->command == FLOW_BLOCK_UNBIND)
+		indir_dev_remove(data);
+
 	list_for_each_entry(this, &flow_block_indr_dev_list, list)
 		this->cb(dev, sch, this->cb_priv, type, bo, data, cleanup);
 

net/netfilter/nf_flow_table_offload.c

@@ -1096,6 +1096,7 @@ static void nf_flow_table_block_offload_init(struct flow_block_offload *bo,
 	bo->command	= cmd;
 	bo->binder_type	= FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS;
 	bo->extack	= extack;
+	bo->cb_list_head = &flowtable->flow_block.cb_list;
 	INIT_LIST_HEAD(&bo->cb_list);
 }
 

net/netfilter/nf_tables_offload.c

@@ -353,6 +353,7 @@ static void nft_flow_block_offload_init(struct flow_block_offload *bo,
 	bo->command	= cmd;
 	bo->binder_type	= FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS;
 	bo->extack	= extack;
+	bo->cb_list_head = &basechain->flow_block.cb_list;
 	INIT_LIST_HEAD(&bo->cb_list);
 }
 

net/sched/cls_api.c

@@ -634,6 +634,7 @@ static void tcf_block_offload_init(struct flow_block_offload *bo,
 	bo->block_shared = shared;
 	bo->extack = extack;
 	bo->sch = sch;
+	bo->cb_list_head = &flow_block->cb_list;
 	INIT_LIST_HEAD(&bo->cb_list);
 }