Merge branch 'net-marvell-prestera-add-nexthop-routes-offloading'

Yevhen Orlov says:

====================
net: marvell: prestera: add nexthop routes offloading

Add support for nexthop routes for Marvell Prestera driver.
Subscribe on NEIGH_UPDATE events.

Add features:
 - Support connected route adding
   e.g.: "ip address add 1.1.1.1/24 dev sw1p1"
   e.g.: "ip route add 6.6.6/24 dev sw1p1"
 - Support nexthop route adding
   e.g.: "ip route add 5.5.5/24 via 1.1.1.2"
 - Support ECMP route adding
   e.g.: "ip route add 5.5.5/24 nexthop via 1.1.1.2 nexthop via 1.1.1.3"
 - Support "offload" and "trap" flags per each nexthop
 - Support "offload" flag for neighbours

Limitations:
 - Only "local" and "main" tables supported
 - Only generic interfaces supported for router (no bridges or vlans)

Flags meaning:
  ip route add 5.5.5/24 nexthop via 2.2.2.2 nexthop via 2.2.2.3
  ip route show
  ...
  5.5.5.0/24 rt_offload
        nexthop via 2.2.2.2 dev sw1p31 weight 1 trap
        nexthop via 2.2.2.3 dev sw1p31 weight 1 trap
  ...
  # When you just add route - lpm entry became occupied
  # in HW ("rt_offload" flag), but related to nexthops neighbours
  # still not resolved ("trap" flag).
  #
  # After some time...
  ip route show
  ...
  5.5.5.0/24 rt_offload
        nexthop via 2.2.2.2 dev sw1p31 weight 1 offload
        nexthop via 2.2.2.3 dev sw1p31 weight 1 offload
  ...
  # You will see, that appropriate neighbours was resolved and nexthop
  # entries occupied in HW too ("offload" flag)
Co-developed-by: Taras Chornyi <tchornyi@marvell.com>
Signed-off-by: Taras Chornyi <tchornyi@marvell.com>
Co-developed-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: Oleksandr Mazur <oleksandr.mazur@plvision.eu>
Signed-off-by: Yevhen Orlov <yevhen.orlov@plvision.eu>

Changes for v2:
* Add more reviewers in CC
* Check if route nexthop or direct with fib_nh_gw_family instead of fib_nh_scope
  This is needed after,
  747c1430 ("ip: fix dflt addr selection for connected nexthop"),
  because direct route is now with the same scope as nexthop (RT_SCOPE_LINK)

Changes for v3:
* Resolve "unused functions" warnings, after
  patch ("net: marvell: prestera: Add heplers to interact ... "), and before
  patch ("net: marvell: prestera: Add neighbour cache accounting")

Changes for v4:
* Rebase to the latest master to resolve patch applying issues

Changes for v5:
* Repack structures to prevent holes
* Remove unused variables
* Fix misspeling issues

Changes for v6:
* Rebase on top of master
* Fix smatch warnings

Changes for v7:
* Rebase on top of master
* Refactor: use "fib_lookup" instead of "fib_new_table"+"fib_table_lookup",
  according to Paolo Abeni suggestion
* Refactor: use "rhashtable_free_and_destroy" instead of rhashtable
  walk, according to Paolo Abeni suggestion
====================

Link: https://lore.kernel.org/r/20221001093417.22388-1-yevhen.orlov@plvision.euSigned-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/ethernet/marvell/prestera/prestera.h

@@ -306,17 +306,27 @@ struct prestera_switch {
 	struct prestera_counter *counter;
 	u8 lag_member_max;
 	u8 lag_max;
+	u32 size_tbl_router_nexthop;
 };
 
 struct prestera_router {
 	struct prestera_switch *sw;
 	struct list_head vr_list;
 	struct list_head rif_entry_list;
+	struct rhashtable nh_neigh_ht;
+	struct rhashtable nexthop_group_ht;
 	struct rhashtable fib_ht;
+	struct rhashtable kern_neigh_cache_ht;
 	struct rhashtable kern_fib_cache_ht;
 	struct notifier_block inetaddr_nb;
 	struct notifier_block inetaddr_valid_nb;
 	struct notifier_block fib_nb;
+	struct notifier_block netevent_nb;
+	u8 *nhgrp_hw_state_cache; /* Bitmap cached hw state of nhs */
+	unsigned long nhgrp_hw_cache_kick; /* jiffies */
+	struct {
+		struct delayed_work dw;
+	} neighs_update;
 };
 
 struct prestera_rxtx_params {
@@ -362,6 +372,8 @@ int prestera_port_cfg_mac_write(struct prestera_port *port,
 struct prestera_port *prestera_port_dev_lower_find(struct net_device *dev);
 
 void prestera_queue_work(struct work_struct *work);
+void prestera_queue_delayed_work(struct delayed_work *work, unsigned long delay);
+void prestera_queue_drain(void);
 
 int prestera_port_learning_set(struct prestera_port *port, bool learn_enable);
 int prestera_port_uc_flood_set(struct prestera_port *port, bool flood);

drivers/net/ethernet/marvell/prestera/prestera_hw.c

@@ -10,11 +10,14 @@
 #include "prestera_hw.h"
 #include "prestera_acl.h"
 #include "prestera_counter.h"
+#include "prestera_router_hw.h"
 
 #define PRESTERA_SWITCH_INIT_TIMEOUT_MS (30 * 1000)
 
 #define PRESTERA_MIN_MTU 64
 
+#define PRESTERA_MSG_CHUNK_SIZE 1024
+
 enum prestera_cmd_type_t {
 	PRESTERA_CMD_TYPE_SWITCH_INIT = 0x1,
 	PRESTERA_CMD_TYPE_SWITCH_ATTR_SET = 0x2,
@@ -57,6 +60,10 @@ enum prestera_cmd_type_t {
 	PRESTERA_CMD_TYPE_ROUTER_RIF_DELETE = 0x601,
 	PRESTERA_CMD_TYPE_ROUTER_LPM_ADD = 0x610,
 	PRESTERA_CMD_TYPE_ROUTER_LPM_DELETE = 0x611,
+	PRESTERA_CMD_TYPE_ROUTER_NH_GRP_SET = 0x622,
+	PRESTERA_CMD_TYPE_ROUTER_NH_GRP_BLK_GET = 0x645,
+	PRESTERA_CMD_TYPE_ROUTER_NH_GRP_ADD = 0x623,
+	PRESTERA_CMD_TYPE_ROUTER_NH_GRP_DELETE = 0x624,
 	PRESTERA_CMD_TYPE_ROUTER_VR_CREATE = 0x630,
 	PRESTERA_CMD_TYPE_ROUTER_VR_DELETE = 0x631,
 
@@ -542,6 +549,14 @@ struct prestera_msg_ip_addr {
 	u8 __pad[3];
 };
 
+struct prestera_msg_nh {
+	struct prestera_msg_iface oif;
+	__le32 hw_id;
+	u8 mac[ETH_ALEN];
+	u8 is_active;
+	u8 pad;
+};
+
 struct prestera_msg_rif_req {
 	struct prestera_msg_cmd cmd;
 	struct prestera_msg_iface iif;
@@ -567,6 +582,34 @@ struct prestera_msg_lpm_req {
 	u8 __pad[2];
 };
 
+struct prestera_msg_nh_req {
+	struct prestera_msg_cmd cmd;
+	struct prestera_msg_nh nh[PRESTERA_NHGR_SIZE_MAX];
+	__le32 size;
+	__le32 grp_id;
+};
+
+struct prestera_msg_nh_chunk_req {
+	struct prestera_msg_cmd cmd;
+	__le32 offset;
+};
+
+struct prestera_msg_nh_chunk_resp {
+	struct prestera_msg_ret ret;
+	u8 hw_state[PRESTERA_MSG_CHUNK_SIZE];
+};
+
+struct prestera_msg_nh_grp_req {
+	struct prestera_msg_cmd cmd;
+	__le32 grp_id;
+	__le32 size;
+};
+
+struct prestera_msg_nh_grp_resp {
+	struct prestera_msg_ret ret;
+	__le32 grp_id;
+};
+
 struct prestera_msg_vr_req {
 	struct prestera_msg_cmd cmd;
 	__le16 vr_id;
@@ -729,11 +772,15 @@ static void prestera_hw_build_tests(void)
 	BUILD_BUG_ON(sizeof(struct prestera_msg_flood_domain_ports_reset_req) != 8);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_mdb_create_req) != 16);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_mdb_destroy_req) != 16);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh_req) != 124);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh_chunk_req) != 8);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh_grp_req) != 12);
 
 	/*  structure that are part of req/resp fw messages */
 	BUILD_BUG_ON(sizeof(struct prestera_msg_iface) != 16);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_ip_addr) != 20);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_flood_domain_port) != 12);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh) != 28);
 
 	/* check responses */
 	BUILD_BUG_ON(sizeof(struct prestera_msg_common_resp) != 8);
@@ -750,6 +797,8 @@ static void prestera_hw_build_tests(void)
 	BUILD_BUG_ON(sizeof(struct prestera_msg_vr_resp) != 12);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_policer_resp) != 12);
 	BUILD_BUG_ON(sizeof(struct prestera_msg_flood_domain_create_resp) != 12);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh_chunk_resp) != 1032);
+	BUILD_BUG_ON(sizeof(struct prestera_msg_nh_grp_resp) != 12);
 
 	/* check events */
 	BUILD_BUG_ON(sizeof(struct prestera_msg_event_port) != 20);
@@ -1027,6 +1076,8 @@ int prestera_hw_switch_init(struct prestera_switch *sw)
 	sw->id = resp.switch_id;
 	sw->lag_member_max = resp.lag_member_max;
 	sw->lag_max = resp.lag_max;
+	sw->size_tbl_router_nexthop =
+		__le32_to_cpu(resp.size_tbl_router_nexthop);
 
 	return 0;
 }
@@ -2037,6 +2088,85 @@ int prestera_hw_lpm_del(struct prestera_switch *sw, u16 vr_id,
 			    sizeof(req));
 }
 
+int prestera_hw_nh_entries_set(struct prestera_switch *sw, int count,
+			       struct prestera_neigh_info *nhs, u32 grp_id)
+{
+	struct prestera_msg_nh_req req = { .size = __cpu_to_le32((u32)count),
+			.grp_id = __cpu_to_le32(grp_id) };
+	int i, err;
+
+	for (i = 0; i < count; i++) {
+		req.nh[i].is_active = nhs[i].connected;
+		memcpy(&req.nh[i].mac, nhs[i].ha, ETH_ALEN);
+		err = prestera_iface_to_msg(&nhs[i].iface, &req.nh[i].oif);
+		if (err)
+			return err;
+	}
+
+	return prestera_cmd(sw, PRESTERA_CMD_TYPE_ROUTER_NH_GRP_SET, &req.cmd,
+			    sizeof(req));
+}
+
+int prestera_hw_nhgrp_blk_get(struct prestera_switch *sw,
+			      u8 *hw_state, u32 buf_size /* Buffer in bytes */)
+{
+	static struct prestera_msg_nh_chunk_resp resp;
+	struct prestera_msg_nh_chunk_req req;
+	u32 buf_offset;
+	int err;
+
+	memset(&hw_state[0], 0, buf_size);
+	buf_offset = 0;
+	while (1) {
+		if (buf_offset >= buf_size)
+			break;
+
+		memset(&req, 0, sizeof(req));
+		req.offset = __cpu_to_le32(buf_offset * 8); /* 8 bits in u8 */
+		err = prestera_cmd_ret(sw,
+				       PRESTERA_CMD_TYPE_ROUTER_NH_GRP_BLK_GET,
+				       &req.cmd, sizeof(req), &resp.ret,
+				       sizeof(resp));
+		if (err)
+			return err;
+
+		memcpy(&hw_state[buf_offset], &resp.hw_state[0],
+		       buf_offset + PRESTERA_MSG_CHUNK_SIZE > buf_size ?
+			buf_size - buf_offset : PRESTERA_MSG_CHUNK_SIZE);
+		buf_offset += PRESTERA_MSG_CHUNK_SIZE;
+	}
+
+	return 0;
+}
+
+int prestera_hw_nh_group_create(struct prestera_switch *sw, u16 nh_count,
+				u32 *grp_id)
+{
+	struct prestera_msg_nh_grp_req req = { .size = __cpu_to_le32((u32)nh_count) };
+	struct prestera_msg_nh_grp_resp resp;
+	int err;
+
+	err = prestera_cmd_ret(sw, PRESTERA_CMD_TYPE_ROUTER_NH_GRP_ADD,
+			       &req.cmd, sizeof(req), &resp.ret, sizeof(resp));
+	if (err)
+		return err;
+
+	*grp_id = __le32_to_cpu(resp.grp_id);
+	return err;
+}
+
+int prestera_hw_nh_group_delete(struct prestera_switch *sw, u16 nh_count,
+				u32 grp_id)
+{
+	struct prestera_msg_nh_grp_req req = {
+	    .grp_id = __cpu_to_le32(grp_id),
+	    .size = __cpu_to_le32(nh_count)
+	};
+
+	return prestera_cmd(sw, PRESTERA_CMD_TYPE_ROUTER_NH_GRP_DELETE,
+			    &req.cmd, sizeof(req));
+}
+
 int prestera_hw_rxtx_init(struct prestera_switch *sw,
 			  struct prestera_rxtx_params *params)
 {

drivers/net/ethernet/marvell/prestera/prestera_hw.h

@@ -146,6 +146,7 @@ struct prestera_counter_stats;
 struct prestera_iface;
 struct prestera_flood_domain;
 struct prestera_mdb_entry;
+struct prestera_neigh_info;
 
 /* Switch API */
 int prestera_hw_switch_init(struct prestera_switch *sw);
@@ -266,6 +267,16 @@ int prestera_hw_lpm_add(struct prestera_switch *sw, u16 vr_id,
 int prestera_hw_lpm_del(struct prestera_switch *sw, u16 vr_id,
 			__be32 dst, u32 dst_len);
 
+/* NH API */
+int prestera_hw_nh_entries_set(struct prestera_switch *sw, int count,
+			       struct prestera_neigh_info *nhs, u32 grp_id);
+int prestera_hw_nhgrp_blk_get(struct prestera_switch *sw,
+			      u8 *hw_state, u32 buf_size /* Buffer in bytes */);
+int prestera_hw_nh_group_create(struct prestera_switch *sw, u16 nh_count,
+				u32 *grp_id);
+int prestera_hw_nh_group_delete(struct prestera_switch *sw, u16 nh_count,
+				u32 grp_id);
+
 /* Event handlers */
 int prestera_hw_event_handler_register(struct prestera_switch *sw,
 				       enum prestera_event_type type,

drivers/net/ethernet/marvell/prestera/prestera_main.c

@@ -36,6 +36,17 @@ void prestera_queue_work(struct work_struct *work)
 	queue_work(prestera_owq, work);
 }
 
+void prestera_queue_delayed_work(struct delayed_work *work, unsigned long delay)
+{
+	queue_delayed_work(prestera_wq, work, delay);
+}
+
+void prestera_queue_drain(void)
+{
+	drain_workqueue(prestera_wq);
+	drain_workqueue(prestera_owq);
+}
+
 int prestera_port_learning_set(struct prestera_port *port, bool learn)
 {
 	return prestera_hw_port_learning_set(port, learn);

drivers/net/ethernet/marvell/prestera/prestera_router_hw.h

@@ -31,6 +31,63 @@ struct prestera_ip_addr {
 		PRESTERA_IPV4 = 0,
 		PRESTERA_IPV6
 	} v;
+#define PRESTERA_IP_ADDR_PLEN(V) ((V) == PRESTERA_IPV4 ? 32 : \
+				  /* (V) == PRESTERA_IPV6 ? */ 128 /* : 0 */)
+};
+
+struct prestera_nh_neigh_key {
+	struct prestera_ip_addr addr;
+	/* Seems like rif is obsolete, because there is iface in info ?
+	 * Key can contain functional fields, or fields, which is used to
+	 * filter duplicate objects on logical level (before you pass it to
+	 * HW)... also key can be used to cover hardware restrictions.
+	 * In our case rif - is logical interface (even can be VLAN), which
+	 * is used in combination with IP address (which is also not related to
+	 * hardware nexthop) to provide logical compression of created nexthops.
+	 * You even can imagine, that rif+IPaddr is just cookie.
+	 */
+	/* struct prestera_rif *rif; */
+	/* Use just as cookie, to divide ARP domains (in order with addr) */
+	void *rif;
+};
+
+/* Used for hw call */
+struct prestera_neigh_info {
+	struct prestera_iface iface;
+	unsigned char ha[ETH_ALEN];
+	u8 connected; /* bool. indicate, if mac/oif valid */
+	u8 __pad[1];
+};
+
+/* Used to notify nh about neigh change */
+struct prestera_nh_neigh {
+	struct prestera_nh_neigh_key key;
+	struct prestera_neigh_info info;
+	struct rhash_head ht_node; /* node of prestera_vr */
+	struct list_head nexthop_group_list;
+};
+
+#define PRESTERA_NHGR_SIZE_MAX 4
+
+struct prestera_nexthop_group {
+	struct prestera_nexthop_group_key {
+		struct prestera_nh_neigh_key neigh[PRESTERA_NHGR_SIZE_MAX];
+	} key;
+	/* Store intermediate object here.
+	 * This prevent overhead kzalloc call.
+	 */
+	/* nh_neigh is used only to notify nexthop_group */
+	struct prestera_nh_neigh_head {
+		struct prestera_nexthop_group *this;
+		struct list_head head;
+		/* ptr to neigh is not necessary.
+		 * It used to prevent lookup of nh_neigh by key (n) on destroy
+		 */
+		struct prestera_nh_neigh *neigh;
+	} nh_neigh_head[PRESTERA_NHGR_SIZE_MAX];
+	struct rhash_head ht_node; /* node of prestera_vr */
+	refcount_t refcount;
+	u32 grp_id; /* hw */
 };
 
 struct prestera_fib_key {
@@ -44,12 +101,16 @@ struct prestera_fib_info {
 	struct list_head vr_node;
 	enum prestera_fib_type {
 		PRESTERA_FIB_TYPE_INVALID = 0,
+		/* must be pointer to nh_grp id */
+		PRESTERA_FIB_TYPE_UC_NH,
 		/* It can be connected route
 		 * and will be overlapped with neighbours
 		 */
 		PRESTERA_FIB_TYPE_TRAP,
 		PRESTERA_FIB_TYPE_DROP
 	} type;
+	/* Valid only if type = UC_NH*/
+	struct prestera_nexthop_group *nh_grp;
 };
 
 struct prestera_fib_node {
@@ -67,6 +128,18 @@ struct prestera_rif_entry *
 prestera_rif_entry_create(struct prestera_switch *sw,
 			  struct prestera_rif_entry_key *k,
 			  u32 tb_id, const unsigned char *addr);
+struct prestera_nh_neigh *
+prestera_nh_neigh_find(struct prestera_switch *sw,
+		       struct prestera_nh_neigh_key *key);
+struct prestera_nh_neigh *
+prestera_nh_neigh_get(struct prestera_switch *sw,
+		      struct prestera_nh_neigh_key *key);
+void prestera_nh_neigh_put(struct prestera_switch *sw,
+			   struct prestera_nh_neigh *neigh);
+int prestera_nh_neigh_set(struct prestera_switch *sw,
+			  struct prestera_nh_neigh *neigh);
+bool prestera_nh_neigh_util_hw_state(struct prestera_switch *sw,
+				     struct prestera_nh_neigh *nh_neigh);
 struct prestera_fib_node *prestera_fib_node_find(struct prestera_switch *sw,
 						 struct prestera_fib_key *key);
 void prestera_fib_node_destroy(struct prestera_switch *sw,
@@ -74,7 +147,8 @@ void prestera_fib_node_destroy(struct prestera_switch *sw,
 struct prestera_fib_node *
 prestera_fib_node_create(struct prestera_switch *sw,
 			 struct prestera_fib_key *key,
-			 enum prestera_fib_type fib_type);
+			 enum prestera_fib_type fib_type,
+			 struct prestera_nexthop_group_key *nh_grp_key);
 int prestera_router_hw_init(struct prestera_switch *sw);
 void prestera_router_hw_fini(struct prestera_switch *sw);