Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
authorJakub Kicinski <kuba@kernel.org>
Thu, 9 Feb 2023 20:05:25 +0000 (12:05 -0800)
committerJakub Kicinski <kuba@kernel.org>
Thu, 9 Feb 2023 20:25:40 +0000 (12:25 -0800)
net/devlink/leftover.c / net/core/devlink.c:
  565b4824c39f ("devlink: change port event netdev notifier from per-net to global")
  f05bd8ebeb69 ("devlink: move code to a dedicated directory")
  687125b5799c ("devlink: split out core code")
https://lore.kernel.org/all/20230208094657.379f2b1a@canb.auug.org.au/

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
33 files changed:
1  2 
MAINTAINERS
drivers/net/dsa/mt7530.c
drivers/net/ethernet/cadence/macb_main.c
drivers/net/ethernet/intel/ice/ice_common.c
drivers/net/ethernet/intel/ice/ice_main.c
drivers/net/ethernet/intel/ice/ice_tc_lib.c
drivers/net/ethernet/intel/igc/igc_main.c
drivers/net/ethernet/mediatek/mtk_eth_soc.c
drivers/net/ethernet/mellanox/mlx5/core/diag/fw_tracer.c
drivers/net/ethernet/mellanox/mlx5/core/ecpf.c
drivers/net/ethernet/mellanox/mlx5/core/en_fs.c
drivers/net/ethernet/mellanox/mlx5/core/en_main.c
drivers/net/ethernet/mellanox/mlx5/core/ipoib/ethtool.c
drivers/net/ethernet/mellanox/mlx5/core/main.c
drivers/net/ethernet/mellanox/mlx5/core/pagealloc.c
drivers/net/ethernet/microchip/sparx5/sparx5_ptp.c
drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c
drivers/net/phy/phylink.c
include/linux/mlx5/driver.h
lib/Kconfig.debug
net/core/neighbour.c
net/core/sock.c
net/devlink/core.c
net/devlink/leftover.c
net/ipv4/af_inet.c
net/ipv4/inet_connection_sock.c
net/mptcp/pm_netlink.c
net/mptcp/protocol.c
net/mptcp/sockopt.c
net/mptcp/subflow.c
net/xfrm/xfrm_state.c
tools/testing/selftests/net/forwarding/lib.sh
tools/testing/selftests/net/mptcp/mptcp_join.sh

diff --cc MAINTAINERS
Simple merge
Simple merge
index b70e36025d922bdb885a37a3dd9c9a9d299c5c1e,cdc87ecae5d399ffdb95961e06330e5a074bd6b1..9a3878f9e58246a19d555ed27c33c5563e08da95
@@@ -89,13 -85,9 +89,13 @@@ void mlx5_ec_cleanup(struct mlx5_core_d
        if (!mlx5_core_is_ecpf(dev))
                return;
  
 +      /* Management PF don't have a peer PF */
 +      if (mlx5_core_is_management_pf(dev))
 +              return;
 +
        mlx5_host_pf_cleanup(dev);
  
-       err = mlx5_wait_for_pages(dev, &dev->priv.host_pf_pages);
+       err = mlx5_wait_for_pages(dev, &dev->priv.page_counters[MLX5_HOST_PF]);
        if (err)
                mlx5_core_warn(dev, "Timeout reclaiming external host PF pages err(%d)\n", err);
  }
Simple merge
Simple merge
Simple merge
Simple merge
diff --cc net/core/sock.c
Simple merge
index aeffd1b8206d6478998f00e1e34cb7130ca503bb,0000000000000000000000000000000000000000..a4f47dafb8644597f6a002889522321887a3508a
mode 100644,000000..100644
--- /dev/null
@@@ -1,321 -1,0 +1,320 @@@
-       ret = register_netdevice_notifier_net(net, &devlink->netdevice_nb);
 +// SPDX-License-Identifier: GPL-2.0-or-later
 +/*
 + * Copyright (c) 2016 Mellanox Technologies. All rights reserved.
 + * Copyright (c) 2016 Jiri Pirko <jiri@mellanox.com>
 + */
 +
 +#include <net/genetlink.h>
 +
 +#include "devl_internal.h"
 +
 +DEFINE_XARRAY_FLAGS(devlinks, XA_FLAGS_ALLOC);
 +
 +void *devlink_priv(struct devlink *devlink)
 +{
 +      return &devlink->priv;
 +}
 +EXPORT_SYMBOL_GPL(devlink_priv);
 +
 +struct devlink *priv_to_devlink(void *priv)
 +{
 +      return container_of(priv, struct devlink, priv);
 +}
 +EXPORT_SYMBOL_GPL(priv_to_devlink);
 +
 +struct device *devlink_to_dev(const struct devlink *devlink)
 +{
 +      return devlink->dev;
 +}
 +EXPORT_SYMBOL_GPL(devlink_to_dev);
 +
 +struct net *devlink_net(const struct devlink *devlink)
 +{
 +      return read_pnet(&devlink->_net);
 +}
 +EXPORT_SYMBOL_GPL(devlink_net);
 +
 +void devl_assert_locked(struct devlink *devlink)
 +{
 +      lockdep_assert_held(&devlink->lock);
 +}
 +EXPORT_SYMBOL_GPL(devl_assert_locked);
 +
 +#ifdef CONFIG_LOCKDEP
 +/* For use in conjunction with LOCKDEP only e.g. rcu_dereference_protected() */
 +bool devl_lock_is_held(struct devlink *devlink)
 +{
 +      return lockdep_is_held(&devlink->lock);
 +}
 +EXPORT_SYMBOL_GPL(devl_lock_is_held);
 +#endif
 +
 +void devl_lock(struct devlink *devlink)
 +{
 +      mutex_lock(&devlink->lock);
 +}
 +EXPORT_SYMBOL_GPL(devl_lock);
 +
 +int devl_trylock(struct devlink *devlink)
 +{
 +      return mutex_trylock(&devlink->lock);
 +}
 +EXPORT_SYMBOL_GPL(devl_trylock);
 +
 +void devl_unlock(struct devlink *devlink)
 +{
 +      mutex_unlock(&devlink->lock);
 +}
 +EXPORT_SYMBOL_GPL(devl_unlock);
 +
 +/**
 + * devlink_try_get() - try to obtain a reference on a devlink instance
 + * @devlink: instance to reference
 + *
 + * Obtain a reference on a devlink instance. A reference on a devlink instance
 + * only implies that it's safe to take the instance lock. It does not imply
 + * that the instance is registered, use devl_is_registered() after taking
 + * the instance lock to check registration status.
 + */
 +struct devlink *__must_check devlink_try_get(struct devlink *devlink)
 +{
 +      if (refcount_inc_not_zero(&devlink->refcount))
 +              return devlink;
 +      return NULL;
 +}
 +
 +static void devlink_release(struct work_struct *work)
 +{
 +      struct devlink *devlink;
 +
 +      devlink = container_of(to_rcu_work(work), struct devlink, rwork);
 +
 +      mutex_destroy(&devlink->lock);
 +      lockdep_unregister_key(&devlink->lock_key);
 +      kfree(devlink);
 +}
 +
 +void devlink_put(struct devlink *devlink)
 +{
 +      if (refcount_dec_and_test(&devlink->refcount))
 +              queue_rcu_work(system_wq, &devlink->rwork);
 +}
 +
 +struct devlink *devlinks_xa_find_get(struct net *net, unsigned long *indexp)
 +{
 +      struct devlink *devlink = NULL;
 +
 +      rcu_read_lock();
 +retry:
 +      devlink = xa_find(&devlinks, indexp, ULONG_MAX, DEVLINK_REGISTERED);
 +      if (!devlink)
 +              goto unlock;
 +
 +      if (!devlink_try_get(devlink))
 +              goto next;
 +      if (!net_eq(devlink_net(devlink), net)) {
 +              devlink_put(devlink);
 +              goto next;
 +      }
 +unlock:
 +      rcu_read_unlock();
 +      return devlink;
 +
 +next:
 +      (*indexp)++;
 +      goto retry;
 +}
 +
 +/**
 + * devl_register - Register devlink instance
 + * @devlink: devlink
 + */
 +int devl_register(struct devlink *devlink)
 +{
 +      ASSERT_DEVLINK_NOT_REGISTERED(devlink);
 +      devl_assert_locked(devlink);
 +
 +      xa_set_mark(&devlinks, devlink->index, DEVLINK_REGISTERED);
 +      devlink_notify_register(devlink);
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_register);
 +
 +void devlink_register(struct devlink *devlink)
 +{
 +      devl_lock(devlink);
 +      devl_register(devlink);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_register);
 +
 +/**
 + * devl_unregister - Unregister devlink instance
 + * @devlink: devlink
 + */
 +void devl_unregister(struct devlink *devlink)
 +{
 +      ASSERT_DEVLINK_REGISTERED(devlink);
 +      devl_assert_locked(devlink);
 +
 +      devlink_notify_unregister(devlink);
 +      xa_clear_mark(&devlinks, devlink->index, DEVLINK_REGISTERED);
 +}
 +EXPORT_SYMBOL_GPL(devl_unregister);
 +
 +void devlink_unregister(struct devlink *devlink)
 +{
 +      devl_lock(devlink);
 +      devl_unregister(devlink);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_unregister);
 +
 +/**
 + *    devlink_alloc_ns - Allocate new devlink instance resources
 + *    in specific namespace
 + *
 + *    @ops: ops
 + *    @priv_size: size of user private data
 + *    @net: net namespace
 + *    @dev: parent device
 + *
 + *    Allocate new devlink instance resources, including devlink index
 + *    and name.
 + */
 +struct devlink *devlink_alloc_ns(const struct devlink_ops *ops,
 +                               size_t priv_size, struct net *net,
 +                               struct device *dev)
 +{
 +      struct devlink *devlink;
 +      static u32 last_id;
 +      int ret;
 +
 +      WARN_ON(!ops || !dev);
 +      if (!devlink_reload_actions_valid(ops))
 +              return NULL;
 +
 +      devlink = kzalloc(sizeof(*devlink) + priv_size, GFP_KERNEL);
 +      if (!devlink)
 +              return NULL;
 +
 +      ret = xa_alloc_cyclic(&devlinks, &devlink->index, devlink, xa_limit_31b,
 +                            &last_id, GFP_KERNEL);
 +      if (ret < 0)
 +              goto err_xa_alloc;
 +
 +      devlink->netdevice_nb.notifier_call = devlink_port_netdevice_event;
-       WARN_ON_ONCE(unregister_netdevice_notifier_net(devlink_net(devlink),
-                                                      &devlink->netdevice_nb));
++      ret = register_netdevice_notifier(&devlink->netdevice_nb);
 +      if (ret)
 +              goto err_register_netdevice_notifier;
 +
 +      devlink->dev = dev;
 +      devlink->ops = ops;
 +      xa_init_flags(&devlink->ports, XA_FLAGS_ALLOC);
 +      xa_init_flags(&devlink->snapshot_ids, XA_FLAGS_ALLOC);
 +      write_pnet(&devlink->_net, net);
 +      INIT_LIST_HEAD(&devlink->rate_list);
 +      INIT_LIST_HEAD(&devlink->linecard_list);
 +      INIT_LIST_HEAD(&devlink->sb_list);
 +      INIT_LIST_HEAD_RCU(&devlink->dpipe_table_list);
 +      INIT_LIST_HEAD(&devlink->resource_list);
 +      INIT_LIST_HEAD(&devlink->param_list);
 +      INIT_LIST_HEAD(&devlink->region_list);
 +      INIT_LIST_HEAD(&devlink->reporter_list);
 +      INIT_LIST_HEAD(&devlink->trap_list);
 +      INIT_LIST_HEAD(&devlink->trap_group_list);
 +      INIT_LIST_HEAD(&devlink->trap_policer_list);
 +      INIT_RCU_WORK(&devlink->rwork, devlink_release);
 +      lockdep_register_key(&devlink->lock_key);
 +      mutex_init(&devlink->lock);
 +      lockdep_set_class(&devlink->lock, &devlink->lock_key);
 +      refcount_set(&devlink->refcount, 1);
 +
 +      return devlink;
 +
 +err_register_netdevice_notifier:
 +      xa_erase(&devlinks, devlink->index);
 +err_xa_alloc:
 +      kfree(devlink);
 +      return NULL;
 +}
 +EXPORT_SYMBOL_GPL(devlink_alloc_ns);
 +
 +/**
 + *    devlink_free - Free devlink instance resources
 + *
 + *    @devlink: devlink
 + */
 +void devlink_free(struct devlink *devlink)
 +{
 +      ASSERT_DEVLINK_NOT_REGISTERED(devlink);
 +
 +      WARN_ON(!list_empty(&devlink->trap_policer_list));
 +      WARN_ON(!list_empty(&devlink->trap_group_list));
 +      WARN_ON(!list_empty(&devlink->trap_list));
 +      WARN_ON(!list_empty(&devlink->reporter_list));
 +      WARN_ON(!list_empty(&devlink->region_list));
 +      WARN_ON(!list_empty(&devlink->param_list));
 +      WARN_ON(!list_empty(&devlink->resource_list));
 +      WARN_ON(!list_empty(&devlink->dpipe_table_list));
 +      WARN_ON(!list_empty(&devlink->sb_list));
 +      WARN_ON(!list_empty(&devlink->rate_list));
 +      WARN_ON(!list_empty(&devlink->linecard_list));
 +      WARN_ON(!xa_empty(&devlink->ports));
 +
 +      xa_destroy(&devlink->snapshot_ids);
 +      xa_destroy(&devlink->ports);
 +
++      WARN_ON_ONCE(unregister_netdevice_notifier(&devlink->netdevice_nb));
 +
 +      xa_erase(&devlinks, devlink->index);
 +
 +      devlink_put(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_free);
 +
 +static void __net_exit devlink_pernet_pre_exit(struct net *net)
 +{
 +      struct devlink *devlink;
 +      u32 actions_performed;
 +      unsigned long index;
 +      int err;
 +
 +      /* In case network namespace is getting destroyed, reload
 +       * all devlink instances from this namespace into init_net.
 +       */
 +      devlinks_xa_for_each_registered_get(net, index, devlink) {
 +              devl_lock(devlink);
 +              err = 0;
 +              if (devl_is_registered(devlink))
 +                      err = devlink_reload(devlink, &init_net,
 +                                           DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
 +                                           DEVLINK_RELOAD_LIMIT_UNSPEC,
 +                                           &actions_performed, NULL);
 +              devl_unlock(devlink);
 +              devlink_put(devlink);
 +              if (err && err != -EOPNOTSUPP)
 +                      pr_warn("Failed to reload devlink instance into init_net\n");
 +      }
 +}
 +
 +static struct pernet_operations devlink_pernet_ops __net_initdata = {
 +      .pre_exit = devlink_pernet_pre_exit,
 +};
 +
 +static int __init devlink_init(void)
 +{
 +      int err;
 +
 +      err = genl_register_family(&devlink_nl_family);
 +      if (err)
 +              goto out;
 +      err = register_pernet_subsys(&devlink_pernet_ops);
 +
 +out:
 +      WARN_ON(err);
 +      return err;
 +}
 +
 +subsys_initcall(devlink_init);
index 9d637360334045dc4fa42855c6bca2c432639b8a,0000000000000000000000000000000000000000..f05ab093d231e1ecfde189be40975c05ad9d7138
mode 100644,000000..100644
--- /dev/null
@@@ -1,10840 -1,0 +1,10844 @@@
 +// SPDX-License-Identifier: GPL-2.0-or-later
 +/*
 + * net/core/devlink.c - Network physical/parent device Netlink interface
 + *
 + * Heavily inspired by net/wireless/
 + * Copyright (c) 2016 Mellanox Technologies. All rights reserved.
 + * Copyright (c) 2016 Jiri Pirko <jiri@mellanox.com>
 + */
 +
 +#include <linux/etherdevice.h>
 +#include <linux/kernel.h>
 +#include <linux/module.h>
 +#include <linux/types.h>
 +#include <linux/slab.h>
 +#include <linux/gfp.h>
 +#include <linux/device.h>
 +#include <linux/list.h>
 +#include <linux/netdevice.h>
 +#include <linux/spinlock.h>
 +#include <linux/refcount.h>
 +#include <linux/workqueue.h>
 +#include <linux/u64_stats_sync.h>
 +#include <linux/timekeeping.h>
 +#include <rdma/ib_verbs.h>
 +#include <net/netlink.h>
 +#include <net/genetlink.h>
 +#include <net/rtnetlink.h>
 +#include <net/net_namespace.h>
 +#include <net/sock.h>
 +#include <net/devlink.h>
 +#define CREATE_TRACE_POINTS
 +#include <trace/events/devlink.h>
 +
 +#include "devl_internal.h"
 +
 +struct devlink_linecard {
 +      struct list_head list;
 +      struct devlink *devlink;
 +      unsigned int index;
 +      const struct devlink_linecard_ops *ops;
 +      void *priv;
 +      enum devlink_linecard_state state;
 +      struct mutex state_lock; /* Protects state */
 +      const char *type;
 +      struct devlink_linecard_type *types;
 +      unsigned int types_count;
 +      struct devlink *nested_devlink;
 +};
 +
 +/**
 + * struct devlink_resource - devlink resource
 + * @name: name of the resource
 + * @id: id, per devlink instance
 + * @size: size of the resource
 + * @size_new: updated size of the resource, reload is needed
 + * @size_valid: valid in case the total size of the resource is valid
 + *              including its children
 + * @parent: parent resource
 + * @size_params: size parameters
 + * @list: parent list
 + * @resource_list: list of child resources
 + * @occ_get: occupancy getter callback
 + * @occ_get_priv: occupancy getter callback priv
 + */
 +struct devlink_resource {
 +      const char *name;
 +      u64 id;
 +      u64 size;
 +      u64 size_new;
 +      bool size_valid;
 +      struct devlink_resource *parent;
 +      struct devlink_resource_size_params size_params;
 +      struct list_head list;
 +      struct list_head resource_list;
 +      devlink_resource_occ_get_t *occ_get;
 +      void *occ_get_priv;
 +};
 +
 +static struct devlink_dpipe_field devlink_dpipe_fields_ethernet[] = {
 +      {
 +              .name = "destination mac",
 +              .id = DEVLINK_DPIPE_FIELD_ETHERNET_DST_MAC,
 +              .bitwidth = 48,
 +      },
 +};
 +
 +struct devlink_dpipe_header devlink_dpipe_header_ethernet = {
 +      .name = "ethernet",
 +      .id = DEVLINK_DPIPE_HEADER_ETHERNET,
 +      .fields = devlink_dpipe_fields_ethernet,
 +      .fields_count = ARRAY_SIZE(devlink_dpipe_fields_ethernet),
 +      .global = true,
 +};
 +EXPORT_SYMBOL_GPL(devlink_dpipe_header_ethernet);
 +
 +static struct devlink_dpipe_field devlink_dpipe_fields_ipv4[] = {
 +      {
 +              .name = "destination ip",
 +              .id = DEVLINK_DPIPE_FIELD_IPV4_DST_IP,
 +              .bitwidth = 32,
 +      },
 +};
 +
 +struct devlink_dpipe_header devlink_dpipe_header_ipv4 = {
 +      .name = "ipv4",
 +      .id = DEVLINK_DPIPE_HEADER_IPV4,
 +      .fields = devlink_dpipe_fields_ipv4,
 +      .fields_count = ARRAY_SIZE(devlink_dpipe_fields_ipv4),
 +      .global = true,
 +};
 +EXPORT_SYMBOL_GPL(devlink_dpipe_header_ipv4);
 +
 +static struct devlink_dpipe_field devlink_dpipe_fields_ipv6[] = {
 +      {
 +              .name = "destination ip",
 +              .id = DEVLINK_DPIPE_FIELD_IPV6_DST_IP,
 +              .bitwidth = 128,
 +      },
 +};
 +
 +struct devlink_dpipe_header devlink_dpipe_header_ipv6 = {
 +      .name = "ipv6",
 +      .id = DEVLINK_DPIPE_HEADER_IPV6,
 +      .fields = devlink_dpipe_fields_ipv6,
 +      .fields_count = ARRAY_SIZE(devlink_dpipe_fields_ipv6),
 +      .global = true,
 +};
 +EXPORT_SYMBOL_GPL(devlink_dpipe_header_ipv6);
 +
 +EXPORT_TRACEPOINT_SYMBOL_GPL(devlink_hwmsg);
 +EXPORT_TRACEPOINT_SYMBOL_GPL(devlink_hwerr);
 +EXPORT_TRACEPOINT_SYMBOL_GPL(devlink_trap_report);
 +
 +#define DEVLINK_PORT_FN_CAPS_VALID_MASK \
 +      (_BITUL(__DEVLINK_PORT_FN_ATTR_CAPS_MAX) - 1)
 +
 +static const struct nla_policy devlink_function_nl_policy[DEVLINK_PORT_FUNCTION_ATTR_MAX + 1] = {
 +      [DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR] = { .type = NLA_BINARY },
 +      [DEVLINK_PORT_FN_ATTR_STATE] =
 +              NLA_POLICY_RANGE(NLA_U8, DEVLINK_PORT_FN_STATE_INACTIVE,
 +                               DEVLINK_PORT_FN_STATE_ACTIVE),
 +      [DEVLINK_PORT_FN_ATTR_CAPS] =
 +              NLA_POLICY_BITFIELD32(DEVLINK_PORT_FN_CAPS_VALID_MASK),
 +};
 +
 +#define ASSERT_DEVLINK_PORT_REGISTERED(devlink_port)                          \
 +      WARN_ON_ONCE(!(devlink_port)->registered)
 +#define ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port)                      \
 +      WARN_ON_ONCE((devlink_port)->registered)
 +#define ASSERT_DEVLINK_PORT_INITIALIZED(devlink_port)                         \
 +      WARN_ON_ONCE(!(devlink_port)->initialized)
 +
 +static struct devlink_port *devlink_port_get_by_index(struct devlink *devlink,
 +                                                    unsigned int port_index)
 +{
 +      return xa_load(&devlink->ports, port_index);
 +}
 +
 +static struct devlink_port *devlink_port_get_from_attrs(struct devlink *devlink,
 +                                                      struct nlattr **attrs)
 +{
 +      if (attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              u32 port_index = nla_get_u32(attrs[DEVLINK_ATTR_PORT_INDEX]);
 +              struct devlink_port *devlink_port;
 +
 +              devlink_port = devlink_port_get_by_index(devlink, port_index);
 +              if (!devlink_port)
 +                      return ERR_PTR(-ENODEV);
 +              return devlink_port;
 +      }
 +      return ERR_PTR(-EINVAL);
 +}
 +
 +struct devlink_port *devlink_port_get_from_info(struct devlink *devlink,
 +                                              struct genl_info *info)
 +{
 +      return devlink_port_get_from_attrs(devlink, info->attrs);
 +}
 +
 +static inline bool
 +devlink_rate_is_leaf(struct devlink_rate *devlink_rate)
 +{
 +      return devlink_rate->type == DEVLINK_RATE_TYPE_LEAF;
 +}
 +
 +static inline bool
 +devlink_rate_is_node(struct devlink_rate *devlink_rate)
 +{
 +      return devlink_rate->type == DEVLINK_RATE_TYPE_NODE;
 +}
 +
 +static struct devlink_rate *
 +devlink_rate_leaf_get_from_info(struct devlink *devlink, struct genl_info *info)
 +{
 +      struct devlink_rate *devlink_rate;
 +      struct devlink_port *devlink_port;
 +
 +      devlink_port = devlink_port_get_from_attrs(devlink, info->attrs);
 +      if (IS_ERR(devlink_port))
 +              return ERR_CAST(devlink_port);
 +      devlink_rate = devlink_port->devlink_rate;
 +      return devlink_rate ?: ERR_PTR(-ENODEV);
 +}
 +
 +static struct devlink_rate *
 +devlink_rate_node_get_by_name(struct devlink *devlink, const char *node_name)
 +{
 +      static struct devlink_rate *devlink_rate;
 +
 +      list_for_each_entry(devlink_rate, &devlink->rate_list, list) {
 +              if (devlink_rate_is_node(devlink_rate) &&
 +                  !strcmp(node_name, devlink_rate->name))
 +                      return devlink_rate;
 +      }
 +      return ERR_PTR(-ENODEV);
 +}
 +
 +static struct devlink_rate *
 +devlink_rate_node_get_from_attrs(struct devlink *devlink, struct nlattr **attrs)
 +{
 +      const char *rate_node_name;
 +      size_t len;
 +
 +      if (!attrs[DEVLINK_ATTR_RATE_NODE_NAME])
 +              return ERR_PTR(-EINVAL);
 +      rate_node_name = nla_data(attrs[DEVLINK_ATTR_RATE_NODE_NAME]);
 +      len = strlen(rate_node_name);
 +      /* Name cannot be empty or decimal number */
 +      if (!len || strspn(rate_node_name, "0123456789") == len)
 +              return ERR_PTR(-EINVAL);
 +
 +      return devlink_rate_node_get_by_name(devlink, rate_node_name);
 +}
 +
 +struct devlink_rate *
 +devlink_rate_node_get_from_info(struct devlink *devlink, struct genl_info *info)
 +{
 +      return devlink_rate_node_get_from_attrs(devlink, info->attrs);
 +}
 +
 +struct devlink_rate *
 +devlink_rate_get_from_info(struct devlink *devlink, struct genl_info *info)
 +{
 +      struct nlattr **attrs = info->attrs;
 +
 +      if (attrs[DEVLINK_ATTR_PORT_INDEX])
 +              return devlink_rate_leaf_get_from_info(devlink, info);
 +      else if (attrs[DEVLINK_ATTR_RATE_NODE_NAME])
 +              return devlink_rate_node_get_from_info(devlink, info);
 +      else
 +              return ERR_PTR(-EINVAL);
 +}
 +
 +static struct devlink_linecard *
 +devlink_linecard_get_by_index(struct devlink *devlink,
 +                            unsigned int linecard_index)
 +{
 +      struct devlink_linecard *devlink_linecard;
 +
 +      list_for_each_entry(devlink_linecard, &devlink->linecard_list, list) {
 +              if (devlink_linecard->index == linecard_index)
 +                      return devlink_linecard;
 +      }
 +      return NULL;
 +}
 +
 +static bool devlink_linecard_index_exists(struct devlink *devlink,
 +                                        unsigned int linecard_index)
 +{
 +      return devlink_linecard_get_by_index(devlink, linecard_index);
 +}
 +
 +static struct devlink_linecard *
 +devlink_linecard_get_from_attrs(struct devlink *devlink, struct nlattr **attrs)
 +{
 +      if (attrs[DEVLINK_ATTR_LINECARD_INDEX]) {
 +              u32 linecard_index = nla_get_u32(attrs[DEVLINK_ATTR_LINECARD_INDEX]);
 +              struct devlink_linecard *linecard;
 +
 +              linecard = devlink_linecard_get_by_index(devlink, linecard_index);
 +              if (!linecard)
 +                      return ERR_PTR(-ENODEV);
 +              return linecard;
 +      }
 +      return ERR_PTR(-EINVAL);
 +}
 +
 +struct devlink_linecard *
 +devlink_linecard_get_from_info(struct devlink *devlink, struct genl_info *info)
 +{
 +      return devlink_linecard_get_from_attrs(devlink, info->attrs);
 +}
 +
 +struct devlink_sb {
 +      struct list_head list;
 +      unsigned int index;
 +      u32 size;
 +      u16 ingress_pools_count;
 +      u16 egress_pools_count;
 +      u16 ingress_tc_count;
 +      u16 egress_tc_count;
 +};
 +
 +static u16 devlink_sb_pool_count(struct devlink_sb *devlink_sb)
 +{
 +      return devlink_sb->ingress_pools_count + devlink_sb->egress_pools_count;
 +}
 +
 +static struct devlink_sb *devlink_sb_get_by_index(struct devlink *devlink,
 +                                                unsigned int sb_index)
 +{
 +      struct devlink_sb *devlink_sb;
 +
 +      list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 +              if (devlink_sb->index == sb_index)
 +                      return devlink_sb;
 +      }
 +      return NULL;
 +}
 +
 +static bool devlink_sb_index_exists(struct devlink *devlink,
 +                                  unsigned int sb_index)
 +{
 +      return devlink_sb_get_by_index(devlink, sb_index);
 +}
 +
 +static struct devlink_sb *devlink_sb_get_from_attrs(struct devlink *devlink,
 +                                                  struct nlattr **attrs)
 +{
 +      if (attrs[DEVLINK_ATTR_SB_INDEX]) {
 +              u32 sb_index = nla_get_u32(attrs[DEVLINK_ATTR_SB_INDEX]);
 +              struct devlink_sb *devlink_sb;
 +
 +              devlink_sb = devlink_sb_get_by_index(devlink, sb_index);
 +              if (!devlink_sb)
 +                      return ERR_PTR(-ENODEV);
 +              return devlink_sb;
 +      }
 +      return ERR_PTR(-EINVAL);
 +}
 +
 +static struct devlink_sb *devlink_sb_get_from_info(struct devlink *devlink,
 +                                                 struct genl_info *info)
 +{
 +      return devlink_sb_get_from_attrs(devlink, info->attrs);
 +}
 +
 +static int devlink_sb_pool_index_get_from_attrs(struct devlink_sb *devlink_sb,
 +                                              struct nlattr **attrs,
 +                                              u16 *p_pool_index)
 +{
 +      u16 val;
 +
 +      if (!attrs[DEVLINK_ATTR_SB_POOL_INDEX])
 +              return -EINVAL;
 +
 +      val = nla_get_u16(attrs[DEVLINK_ATTR_SB_POOL_INDEX]);
 +      if (val >= devlink_sb_pool_count(devlink_sb))
 +              return -EINVAL;
 +      *p_pool_index = val;
 +      return 0;
 +}
 +
 +static int devlink_sb_pool_index_get_from_info(struct devlink_sb *devlink_sb,
 +                                             struct genl_info *info,
 +                                             u16 *p_pool_index)
 +{
 +      return devlink_sb_pool_index_get_from_attrs(devlink_sb, info->attrs,
 +                                                  p_pool_index);
 +}
 +
 +static int
 +devlink_sb_pool_type_get_from_attrs(struct nlattr **attrs,
 +                                  enum devlink_sb_pool_type *p_pool_type)
 +{
 +      u8 val;
 +
 +      if (!attrs[DEVLINK_ATTR_SB_POOL_TYPE])
 +              return -EINVAL;
 +
 +      val = nla_get_u8(attrs[DEVLINK_ATTR_SB_POOL_TYPE]);
 +      if (val != DEVLINK_SB_POOL_TYPE_INGRESS &&
 +          val != DEVLINK_SB_POOL_TYPE_EGRESS)
 +              return -EINVAL;
 +      *p_pool_type = val;
 +      return 0;
 +}
 +
 +static int
 +devlink_sb_pool_type_get_from_info(struct genl_info *info,
 +                                 enum devlink_sb_pool_type *p_pool_type)
 +{
 +      return devlink_sb_pool_type_get_from_attrs(info->attrs, p_pool_type);
 +}
 +
 +static int
 +devlink_sb_th_type_get_from_attrs(struct nlattr **attrs,
 +                                enum devlink_sb_threshold_type *p_th_type)
 +{
 +      u8 val;
 +
 +      if (!attrs[DEVLINK_ATTR_SB_POOL_THRESHOLD_TYPE])
 +              return -EINVAL;
 +
 +      val = nla_get_u8(attrs[DEVLINK_ATTR_SB_POOL_THRESHOLD_TYPE]);
 +      if (val != DEVLINK_SB_THRESHOLD_TYPE_STATIC &&
 +          val != DEVLINK_SB_THRESHOLD_TYPE_DYNAMIC)
 +              return -EINVAL;
 +      *p_th_type = val;
 +      return 0;
 +}
 +
 +static int
 +devlink_sb_th_type_get_from_info(struct genl_info *info,
 +                               enum devlink_sb_threshold_type *p_th_type)
 +{
 +      return devlink_sb_th_type_get_from_attrs(info->attrs, p_th_type);
 +}
 +
 +static int
 +devlink_sb_tc_index_get_from_attrs(struct devlink_sb *devlink_sb,
 +                                 struct nlattr **attrs,
 +                                 enum devlink_sb_pool_type pool_type,
 +                                 u16 *p_tc_index)
 +{
 +      u16 val;
 +
 +      if (!attrs[DEVLINK_ATTR_SB_TC_INDEX])
 +              return -EINVAL;
 +
 +      val = nla_get_u16(attrs[DEVLINK_ATTR_SB_TC_INDEX]);
 +      if (pool_type == DEVLINK_SB_POOL_TYPE_INGRESS &&
 +          val >= devlink_sb->ingress_tc_count)
 +              return -EINVAL;
 +      if (pool_type == DEVLINK_SB_POOL_TYPE_EGRESS &&
 +          val >= devlink_sb->egress_tc_count)
 +              return -EINVAL;
 +      *p_tc_index = val;
 +      return 0;
 +}
 +
 +static void devlink_port_fn_cap_fill(struct nla_bitfield32 *caps,
 +                                   u32 cap, bool is_enable)
 +{
 +      caps->selector |= cap;
 +      if (is_enable)
 +              caps->value |= cap;
 +}
 +
 +static int devlink_port_fn_roce_fill(const struct devlink_ops *ops,
 +                                   struct devlink_port *devlink_port,
 +                                   struct nla_bitfield32 *caps,
 +                                   struct netlink_ext_ack *extack)
 +{
 +      bool is_enable;
 +      int err;
 +
 +      if (!ops->port_fn_roce_get)
 +              return 0;
 +
 +      err = ops->port_fn_roce_get(devlink_port, &is_enable, extack);
 +      if (err) {
 +              if (err == -EOPNOTSUPP)
 +                      return 0;
 +              return err;
 +      }
 +
 +      devlink_port_fn_cap_fill(caps, DEVLINK_PORT_FN_CAP_ROCE, is_enable);
 +      return 0;
 +}
 +
 +static int devlink_port_fn_migratable_fill(const struct devlink_ops *ops,
 +                                         struct devlink_port *devlink_port,
 +                                         struct nla_bitfield32 *caps,
 +                                         struct netlink_ext_ack *extack)
 +{
 +      bool is_enable;
 +      int err;
 +
 +      if (!ops->port_fn_migratable_get ||
 +          devlink_port->attrs.flavour != DEVLINK_PORT_FLAVOUR_PCI_VF)
 +              return 0;
 +
 +      err = ops->port_fn_migratable_get(devlink_port, &is_enable, extack);
 +      if (err) {
 +              if (err == -EOPNOTSUPP)
 +                      return 0;
 +              return err;
 +      }
 +
 +      devlink_port_fn_cap_fill(caps, DEVLINK_PORT_FN_CAP_MIGRATABLE, is_enable);
 +      return 0;
 +}
 +
 +static int devlink_port_fn_caps_fill(const struct devlink_ops *ops,
 +                                   struct devlink_port *devlink_port,
 +                                   struct sk_buff *msg,
 +                                   struct netlink_ext_ack *extack,
 +                                   bool *msg_updated)
 +{
 +      struct nla_bitfield32 caps = {};
 +      int err;
 +
 +      err = devlink_port_fn_roce_fill(ops, devlink_port, &caps, extack);
 +      if (err)
 +              return err;
 +
 +      err = devlink_port_fn_migratable_fill(ops, devlink_port, &caps, extack);
 +      if (err)
 +              return err;
 +
 +      if (!caps.selector)
 +              return 0;
 +      err = nla_put_bitfield32(msg, DEVLINK_PORT_FN_ATTR_CAPS, caps.value,
 +                               caps.selector);
 +      if (err)
 +              return err;
 +
 +      *msg_updated = true;
 +      return 0;
 +}
 +
 +static int
 +devlink_sb_tc_index_get_from_info(struct devlink_sb *devlink_sb,
 +                                struct genl_info *info,
 +                                enum devlink_sb_pool_type pool_type,
 +                                u16 *p_tc_index)
 +{
 +      return devlink_sb_tc_index_get_from_attrs(devlink_sb, info->attrs,
 +                                                pool_type, p_tc_index);
 +}
 +
 +struct devlink_region {
 +      struct devlink *devlink;
 +      struct devlink_port *port;
 +      struct list_head list;
 +      union {
 +              const struct devlink_region_ops *ops;
 +              const struct devlink_port_region_ops *port_ops;
 +      };
 +      struct mutex snapshot_lock; /* protects snapshot_list,
 +                                   * max_snapshots and cur_snapshots
 +                                   * consistency.
 +                                   */
 +      struct list_head snapshot_list;
 +      u32 max_snapshots;
 +      u32 cur_snapshots;
 +      u64 size;
 +};
 +
 +struct devlink_snapshot {
 +      struct list_head list;
 +      struct devlink_region *region;
 +      u8 *data;
 +      u32 id;
 +};
 +
 +static struct devlink_region *
 +devlink_region_get_by_name(struct devlink *devlink, const char *region_name)
 +{
 +      struct devlink_region *region;
 +
 +      list_for_each_entry(region, &devlink->region_list, list)
 +              if (!strcmp(region->ops->name, region_name))
 +                      return region;
 +
 +      return NULL;
 +}
 +
 +static struct devlink_region *
 +devlink_port_region_get_by_name(struct devlink_port *port,
 +                              const char *region_name)
 +{
 +      struct devlink_region *region;
 +
 +      list_for_each_entry(region, &port->region_list, list)
 +              if (!strcmp(region->ops->name, region_name))
 +                      return region;
 +
 +      return NULL;
 +}
 +
 +static struct devlink_snapshot *
 +devlink_region_snapshot_get_by_id(struct devlink_region *region, u32 id)
 +{
 +      struct devlink_snapshot *snapshot;
 +
 +      list_for_each_entry(snapshot, &region->snapshot_list, list)
 +              if (snapshot->id == id)
 +                      return snapshot;
 +
 +      return NULL;
 +}
 +
 +static int devlink_nl_put_nested_handle(struct sk_buff *msg, struct devlink *devlink)
 +{
 +      struct nlattr *nested_attr;
 +
 +      nested_attr = nla_nest_start(msg, DEVLINK_ATTR_NESTED_DEVLINK);
 +      if (!nested_attr)
 +              return -EMSGSIZE;
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, nested_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, nested_attr);
 +      return -EMSGSIZE;
 +}
 +
 +int devlink_nl_port_handle_fill(struct sk_buff *msg, struct devlink_port *devlink_port)
 +{
 +      if (devlink_nl_put_handle(msg, devlink_port->devlink))
 +              return -EMSGSIZE;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, devlink_port->index))
 +              return -EMSGSIZE;
 +      return 0;
 +}
 +
 +size_t devlink_nl_port_handle_size(struct devlink_port *devlink_port)
 +{
 +      struct devlink *devlink = devlink_port->devlink;
 +
 +      return nla_total_size(strlen(devlink->dev->bus->name) + 1) /* DEVLINK_ATTR_BUS_NAME */
 +           + nla_total_size(strlen(dev_name(devlink->dev)) + 1) /* DEVLINK_ATTR_DEV_NAME */
 +           + nla_total_size(4); /* DEVLINK_ATTR_PORT_INDEX */
 +}
 +
 +static int devlink_nl_port_attrs_put(struct sk_buff *msg,
 +                                   struct devlink_port *devlink_port)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +
 +      if (!devlink_port->attrs_set)
 +              return 0;
 +      if (attrs->lanes) {
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_LANES, attrs->lanes))
 +                      return -EMSGSIZE;
 +      }
 +      if (nla_put_u8(msg, DEVLINK_ATTR_PORT_SPLITTABLE, attrs->splittable))
 +              return -EMSGSIZE;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_PORT_FLAVOUR, attrs->flavour))
 +              return -EMSGSIZE;
 +      switch (devlink_port->attrs.flavour) {
 +      case DEVLINK_PORT_FLAVOUR_PCI_PF:
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_CONTROLLER_NUMBER,
 +                              attrs->pci_pf.controller) ||
 +                  nla_put_u16(msg, DEVLINK_ATTR_PORT_PCI_PF_NUMBER, attrs->pci_pf.pf))
 +                      return -EMSGSIZE;
 +              if (nla_put_u8(msg, DEVLINK_ATTR_PORT_EXTERNAL, attrs->pci_pf.external))
 +                      return -EMSGSIZE;
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_PCI_VF:
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_CONTROLLER_NUMBER,
 +                              attrs->pci_vf.controller) ||
 +                  nla_put_u16(msg, DEVLINK_ATTR_PORT_PCI_PF_NUMBER, attrs->pci_vf.pf) ||
 +                  nla_put_u16(msg, DEVLINK_ATTR_PORT_PCI_VF_NUMBER, attrs->pci_vf.vf))
 +                      return -EMSGSIZE;
 +              if (nla_put_u8(msg, DEVLINK_ATTR_PORT_EXTERNAL, attrs->pci_vf.external))
 +                      return -EMSGSIZE;
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_PCI_SF:
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_CONTROLLER_NUMBER,
 +                              attrs->pci_sf.controller) ||
 +                  nla_put_u16(msg, DEVLINK_ATTR_PORT_PCI_PF_NUMBER,
 +                              attrs->pci_sf.pf) ||
 +                  nla_put_u32(msg, DEVLINK_ATTR_PORT_PCI_SF_NUMBER,
 +                              attrs->pci_sf.sf))
 +                      return -EMSGSIZE;
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_PHYSICAL:
 +      case DEVLINK_PORT_FLAVOUR_CPU:
 +      case DEVLINK_PORT_FLAVOUR_DSA:
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_NUMBER,
 +                              attrs->phys.port_number))
 +                      return -EMSGSIZE;
 +              if (!attrs->split)
 +                      return 0;
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_SPLIT_GROUP,
 +                              attrs->phys.port_number))
 +                      return -EMSGSIZE;
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_SPLIT_SUBPORT_NUMBER,
 +                              attrs->phys.split_subport_number))
 +                      return -EMSGSIZE;
 +              break;
 +      default:
 +              break;
 +      }
 +      return 0;
 +}
 +
 +static int devlink_port_fn_hw_addr_fill(const struct devlink_ops *ops,
 +                                      struct devlink_port *port,
 +                                      struct sk_buff *msg,
 +                                      struct netlink_ext_ack *extack,
 +                                      bool *msg_updated)
 +{
 +      u8 hw_addr[MAX_ADDR_LEN];
 +      int hw_addr_len;
 +      int err;
 +
 +      if (!ops->port_function_hw_addr_get)
 +              return 0;
 +
 +      err = ops->port_function_hw_addr_get(port, hw_addr, &hw_addr_len,
 +                                           extack);
 +      if (err) {
 +              if (err == -EOPNOTSUPP)
 +                      return 0;
 +              return err;
 +      }
 +      err = nla_put(msg, DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR, hw_addr_len, hw_addr);
 +      if (err)
 +              return err;
 +      *msg_updated = true;
 +      return 0;
 +}
 +
 +static int devlink_nl_rate_fill(struct sk_buff *msg,
 +                              struct devlink_rate *devlink_rate,
 +                              enum devlink_command cmd, u32 portid, u32 seq,
 +                              int flags, struct netlink_ext_ack *extack)
 +{
 +      struct devlink *devlink = devlink_rate->devlink;
 +      void *hdr;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u16(msg, DEVLINK_ATTR_RATE_TYPE, devlink_rate->type))
 +              goto nla_put_failure;
 +
 +      if (devlink_rate_is_leaf(devlink_rate)) {
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX,
 +                              devlink_rate->devlink_port->index))
 +                      goto nla_put_failure;
 +      } else if (devlink_rate_is_node(devlink_rate)) {
 +              if (nla_put_string(msg, DEVLINK_ATTR_RATE_NODE_NAME,
 +                                 devlink_rate->name))
 +                      goto nla_put_failure;
 +      }
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_RATE_TX_SHARE,
 +                            devlink_rate->tx_share, DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_RATE_TX_MAX,
 +                            devlink_rate->tx_max, DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u32(msg, DEVLINK_ATTR_RATE_TX_PRIORITY,
 +                      devlink_rate->tx_priority))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u32(msg, DEVLINK_ATTR_RATE_TX_WEIGHT,
 +                      devlink_rate->tx_weight))
 +              goto nla_put_failure;
 +
 +      if (devlink_rate->parent)
 +              if (nla_put_string(msg, DEVLINK_ATTR_RATE_PARENT_NODE_NAME,
 +                                 devlink_rate->parent->name))
 +                      goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static bool
 +devlink_port_fn_state_valid(enum devlink_port_fn_state state)
 +{
 +      return state == DEVLINK_PORT_FN_STATE_INACTIVE ||
 +             state == DEVLINK_PORT_FN_STATE_ACTIVE;
 +}
 +
 +static bool
 +devlink_port_fn_opstate_valid(enum devlink_port_fn_opstate opstate)
 +{
 +      return opstate == DEVLINK_PORT_FN_OPSTATE_DETACHED ||
 +             opstate == DEVLINK_PORT_FN_OPSTATE_ATTACHED;
 +}
 +
 +static int devlink_port_fn_state_fill(const struct devlink_ops *ops,
 +                                    struct devlink_port *port,
 +                                    struct sk_buff *msg,
 +                                    struct netlink_ext_ack *extack,
 +                                    bool *msg_updated)
 +{
 +      enum devlink_port_fn_opstate opstate;
 +      enum devlink_port_fn_state state;
 +      int err;
 +
 +      if (!ops->port_fn_state_get)
 +              return 0;
 +
 +      err = ops->port_fn_state_get(port, &state, &opstate, extack);
 +      if (err) {
 +              if (err == -EOPNOTSUPP)
 +                      return 0;
 +              return err;
 +      }
 +      if (!devlink_port_fn_state_valid(state)) {
 +              WARN_ON_ONCE(1);
 +              NL_SET_ERR_MSG_MOD(extack, "Invalid state read from driver");
 +              return -EINVAL;
 +      }
 +      if (!devlink_port_fn_opstate_valid(opstate)) {
 +              WARN_ON_ONCE(1);
 +              NL_SET_ERR_MSG_MOD(extack,
 +                                 "Invalid operational state read from driver");
 +              return -EINVAL;
 +      }
 +      if (nla_put_u8(msg, DEVLINK_PORT_FN_ATTR_STATE, state) ||
 +          nla_put_u8(msg, DEVLINK_PORT_FN_ATTR_OPSTATE, opstate))
 +              return -EMSGSIZE;
 +      *msg_updated = true;
 +      return 0;
 +}
 +
 +static int
 +devlink_port_fn_mig_set(struct devlink_port *devlink_port, bool enable,
 +                      struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_ops *ops = devlink_port->devlink->ops;
 +
 +      return ops->port_fn_migratable_set(devlink_port, enable, extack);
 +}
 +
 +static int
 +devlink_port_fn_roce_set(struct devlink_port *devlink_port, bool enable,
 +                       struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_ops *ops = devlink_port->devlink->ops;
 +
 +      return ops->port_fn_roce_set(devlink_port, enable, extack);
 +}
 +
 +static int devlink_port_fn_caps_set(struct devlink_port *devlink_port,
 +                                  const struct nlattr *attr,
 +                                  struct netlink_ext_ack *extack)
 +{
 +      struct nla_bitfield32 caps;
 +      u32 caps_value;
 +      int err;
 +
 +      caps = nla_get_bitfield32(attr);
 +      caps_value = caps.value & caps.selector;
 +      if (caps.selector & DEVLINK_PORT_FN_CAP_ROCE) {
 +              err = devlink_port_fn_roce_set(devlink_port,
 +                                             caps_value & DEVLINK_PORT_FN_CAP_ROCE,
 +                                             extack);
 +              if (err)
 +                      return err;
 +      }
 +      if (caps.selector & DEVLINK_PORT_FN_CAP_MIGRATABLE) {
 +              err = devlink_port_fn_mig_set(devlink_port, caps_value &
 +                                            DEVLINK_PORT_FN_CAP_MIGRATABLE,
 +                                            extack);
 +              if (err)
 +                      return err;
 +      }
 +      return 0;
 +}
 +
 +static int
 +devlink_nl_port_function_attrs_put(struct sk_buff *msg, struct devlink_port *port,
 +                                 struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_ops *ops;
 +      struct nlattr *function_attr;
 +      bool msg_updated = false;
 +      int err;
 +
 +      function_attr = nla_nest_start_noflag(msg, DEVLINK_ATTR_PORT_FUNCTION);
 +      if (!function_attr)
 +              return -EMSGSIZE;
 +
 +      ops = port->devlink->ops;
 +      err = devlink_port_fn_hw_addr_fill(ops, port, msg, extack,
 +                                         &msg_updated);
 +      if (err)
 +              goto out;
 +      err = devlink_port_fn_caps_fill(ops, port, msg, extack,
 +                                      &msg_updated);
 +      if (err)
 +              goto out;
 +      err = devlink_port_fn_state_fill(ops, port, msg, extack, &msg_updated);
 +out:
 +      if (err || !msg_updated)
 +              nla_nest_cancel(msg, function_attr);
 +      else
 +              nla_nest_end(msg, function_attr);
 +      return err;
 +}
 +
 +static int devlink_nl_port_fill(struct sk_buff *msg,
 +                              struct devlink_port *devlink_port,
 +                              enum devlink_command cmd, u32 portid, u32 seq,
 +                              int flags, struct netlink_ext_ack *extack)
 +{
 +      struct devlink *devlink = devlink_port->devlink;
 +      void *hdr;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, devlink_port->index))
 +              goto nla_put_failure;
 +
 +      spin_lock_bh(&devlink_port->type_lock);
 +      if (nla_put_u16(msg, DEVLINK_ATTR_PORT_TYPE, devlink_port->type))
 +              goto nla_put_failure_type_locked;
 +      if (devlink_port->desired_type != DEVLINK_PORT_TYPE_NOTSET &&
 +          nla_put_u16(msg, DEVLINK_ATTR_PORT_DESIRED_TYPE,
 +                      devlink_port->desired_type))
 +              goto nla_put_failure_type_locked;
 +      if (devlink_port->type == DEVLINK_PORT_TYPE_ETH) {
 +              if (devlink_port->type_eth.netdev &&
 +                  (nla_put_u32(msg, DEVLINK_ATTR_PORT_NETDEV_IFINDEX,
 +                               devlink_port->type_eth.ifindex) ||
 +                   nla_put_string(msg, DEVLINK_ATTR_PORT_NETDEV_NAME,
 +                                  devlink_port->type_eth.ifname)))
 +                      goto nla_put_failure_type_locked;
 +      }
 +      if (devlink_port->type == DEVLINK_PORT_TYPE_IB) {
 +              struct ib_device *ibdev = devlink_port->type_ib.ibdev;
 +
 +              if (ibdev &&
 +                  nla_put_string(msg, DEVLINK_ATTR_PORT_IBDEV_NAME,
 +                                 ibdev->name))
 +                      goto nla_put_failure_type_locked;
 +      }
 +      spin_unlock_bh(&devlink_port->type_lock);
 +      if (devlink_nl_port_attrs_put(msg, devlink_port))
 +              goto nla_put_failure;
 +      if (devlink_nl_port_function_attrs_put(msg, devlink_port, extack))
 +              goto nla_put_failure;
 +      if (devlink_port->linecard &&
 +          nla_put_u32(msg, DEVLINK_ATTR_LINECARD_INDEX,
 +                      devlink_port->linecard->index))
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure_type_locked:
 +      spin_unlock_bh(&devlink_port->type_lock);
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static void devlink_port_notify(struct devlink_port *devlink_port,
 +                              enum devlink_command cmd)
 +{
 +      struct devlink *devlink = devlink_port->devlink;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_PORT_NEW && cmd != DEVLINK_CMD_PORT_DEL);
 +
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_port_fill(msg, devlink_port, cmd, 0, 0, 0, NULL);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink), msg,
 +                              0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static void devlink_rate_notify(struct devlink_rate *devlink_rate,
 +                              enum devlink_command cmd)
 +{
 +      struct devlink *devlink = devlink_rate->devlink;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_RATE_NEW && cmd != DEVLINK_CMD_RATE_DEL);
 +
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_rate_fill(msg, devlink_rate, cmd, 0, 0, 0, NULL);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink), msg,
 +                              0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int
 +devlink_nl_cmd_rate_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                               struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_rate *devlink_rate;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(devlink_rate, &devlink->rate_list, list) {
 +              enum devlink_command cmd = DEVLINK_CMD_RATE_NEW;
 +              u32 id = NETLINK_CB(cb->skb).portid;
 +
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_rate_fill(msg, devlink_rate, cmd, id,
 +                                         cb->nlh->nlmsg_seq,
 +                                         NLM_F_MULTI, NULL);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_rate_get = {
 +      .dump_one               = devlink_nl_cmd_rate_get_dump_one,
 +};
 +
 +static int devlink_nl_cmd_rate_get_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink_rate *devlink_rate = info->user_ptr[1];
 +      struct sk_buff *msg;
 +      int err;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_rate_fill(msg, devlink_rate, DEVLINK_CMD_RATE_NEW,
 +                                 info->snd_portid, info->snd_seq, 0,
 +                                 info->extack);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static bool
 +devlink_rate_is_parent_node(struct devlink_rate *devlink_rate,
 +                          struct devlink_rate *parent)
 +{
 +      while (parent) {
 +              if (parent == devlink_rate)
 +                      return true;
 +              parent = parent->parent;
 +      }
 +      return false;
 +}
 +
 +static int devlink_nl_cmd_port_get_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct sk_buff *msg;
 +      int err;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_port_fill(msg, devlink_port, DEVLINK_CMD_PORT_NEW,
 +                                 info->snd_portid, info->snd_seq, 0,
 +                                 info->extack);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int
 +devlink_nl_cmd_port_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                               struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_port *devlink_port;
 +      unsigned long port_index;
 +      int idx = 0;
 +      int err = 0;
 +
 +      xa_for_each(&devlink->ports, port_index, devlink_port) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_port_fill(msg, devlink_port,
 +                                         DEVLINK_CMD_NEW,
 +                                         NETLINK_CB(cb->skb).portid,
 +                                         cb->nlh->nlmsg_seq,
 +                                         NLM_F_MULTI, cb->extack);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_port_get = {
 +      .dump_one               = devlink_nl_cmd_port_get_dump_one,
 +};
 +
 +static int devlink_port_type_set(struct devlink_port *devlink_port,
 +                               enum devlink_port_type port_type)
 +
 +{
 +      int err;
 +
 +      if (!devlink_port->devlink->ops->port_type_set)
 +              return -EOPNOTSUPP;
 +
 +      if (port_type == devlink_port->type)
 +              return 0;
 +
 +      err = devlink_port->devlink->ops->port_type_set(devlink_port,
 +                                                      port_type);
 +      if (err)
 +              return err;
 +
 +      devlink_port->desired_type = port_type;
 +      devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_NEW);
 +      return 0;
 +}
 +
 +static int devlink_port_function_hw_addr_set(struct devlink_port *port,
 +                                           const struct nlattr *attr,
 +                                           struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_ops *ops = port->devlink->ops;
 +      const u8 *hw_addr;
 +      int hw_addr_len;
 +
 +      hw_addr = nla_data(attr);
 +      hw_addr_len = nla_len(attr);
 +      if (hw_addr_len > MAX_ADDR_LEN) {
 +              NL_SET_ERR_MSG_MOD(extack, "Port function hardware address too long");
 +              return -EINVAL;
 +      }
 +      if (port->type == DEVLINK_PORT_TYPE_ETH) {
 +              if (hw_addr_len != ETH_ALEN) {
 +                      NL_SET_ERR_MSG_MOD(extack, "Address must be 6 bytes for Ethernet device");
 +                      return -EINVAL;
 +              }
 +              if (!is_unicast_ether_addr(hw_addr)) {
 +                      NL_SET_ERR_MSG_MOD(extack, "Non-unicast hardware address unsupported");
 +                      return -EINVAL;
 +              }
 +      }
 +
 +      return ops->port_function_hw_addr_set(port, hw_addr, hw_addr_len,
 +                                            extack);
 +}
 +
 +static int devlink_port_fn_state_set(struct devlink_port *port,
 +                                   const struct nlattr *attr,
 +                                   struct netlink_ext_ack *extack)
 +{
 +      enum devlink_port_fn_state state;
 +      const struct devlink_ops *ops;
 +
 +      state = nla_get_u8(attr);
 +      ops = port->devlink->ops;
 +      return ops->port_fn_state_set(port, state, extack);
 +}
 +
 +static int devlink_port_function_validate(struct devlink_port *devlink_port,
 +                                        struct nlattr **tb,
 +                                        struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_ops *ops = devlink_port->devlink->ops;
 +      struct nlattr *attr;
 +
 +      if (tb[DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR] &&
 +          !ops->port_function_hw_addr_set) {
 +              NL_SET_ERR_MSG_ATTR(extack, tb[DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR],
 +                                  "Port doesn't support function attributes");
 +              return -EOPNOTSUPP;
 +      }
 +      if (tb[DEVLINK_PORT_FN_ATTR_STATE] && !ops->port_fn_state_set) {
 +              NL_SET_ERR_MSG_ATTR(extack, tb[DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR],
 +                                  "Function does not support state setting");
 +              return -EOPNOTSUPP;
 +      }
 +      attr = tb[DEVLINK_PORT_FN_ATTR_CAPS];
 +      if (attr) {
 +              struct nla_bitfield32 caps;
 +
 +              caps = nla_get_bitfield32(attr);
 +              if (caps.selector & DEVLINK_PORT_FN_CAP_ROCE &&
 +                  !ops->port_fn_roce_set) {
 +                      NL_SET_ERR_MSG_ATTR(extack, attr,
 +                                          "Port doesn't support RoCE function attribute");
 +                      return -EOPNOTSUPP;
 +              }
 +              if (caps.selector & DEVLINK_PORT_FN_CAP_MIGRATABLE) {
 +                      if (!ops->port_fn_migratable_set) {
 +                              NL_SET_ERR_MSG_ATTR(extack, attr,
 +                                                  "Port doesn't support migratable function attribute");
 +                              return -EOPNOTSUPP;
 +                      }
 +                      if (devlink_port->attrs.flavour != DEVLINK_PORT_FLAVOUR_PCI_VF) {
 +                              NL_SET_ERR_MSG_ATTR(extack, attr,
 +                                                  "migratable function attribute supported for VFs only");
 +                              return -EOPNOTSUPP;
 +                      }
 +              }
 +      }
 +      return 0;
 +}
 +
 +static int devlink_port_function_set(struct devlink_port *port,
 +                                   const struct nlattr *attr,
 +                                   struct netlink_ext_ack *extack)
 +{
 +      struct nlattr *tb[DEVLINK_PORT_FUNCTION_ATTR_MAX + 1];
 +      int err;
 +
 +      err = nla_parse_nested(tb, DEVLINK_PORT_FUNCTION_ATTR_MAX, attr,
 +                             devlink_function_nl_policy, extack);
 +      if (err < 0) {
 +              NL_SET_ERR_MSG_MOD(extack, "Fail to parse port function attributes");
 +              return err;
 +      }
 +
 +      err = devlink_port_function_validate(port, tb, extack);
 +      if (err)
 +              return err;
 +
 +      attr = tb[DEVLINK_PORT_FUNCTION_ATTR_HW_ADDR];
 +      if (attr) {
 +              err = devlink_port_function_hw_addr_set(port, attr, extack);
 +              if (err)
 +                      return err;
 +      }
 +
 +      attr = tb[DEVLINK_PORT_FN_ATTR_CAPS];
 +      if (attr) {
 +              err = devlink_port_fn_caps_set(port, attr, extack);
 +              if (err)
 +                      return err;
 +      }
 +
 +      /* Keep this as the last function attribute set, so that when
 +       * multiple port function attributes are set along with state,
 +       * Those can be applied first before activating the state.
 +       */
 +      attr = tb[DEVLINK_PORT_FN_ATTR_STATE];
 +      if (attr)
 +              err = devlink_port_fn_state_set(port, attr, extack);
 +
 +      if (!err)
 +              devlink_port_notify(port, DEVLINK_CMD_PORT_NEW);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_port_set_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      int err;
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_TYPE]) {
 +              enum devlink_port_type port_type;
 +
 +              port_type = nla_get_u16(info->attrs[DEVLINK_ATTR_PORT_TYPE]);
 +              err = devlink_port_type_set(devlink_port, port_type);
 +              if (err)
 +                      return err;
 +      }
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_FUNCTION]) {
 +              struct nlattr *attr = info->attrs[DEVLINK_ATTR_PORT_FUNCTION];
 +              struct netlink_ext_ack *extack = info->extack;
 +
 +              err = devlink_port_function_set(devlink_port, attr, extack);
 +              if (err)
 +                      return err;
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_port_split_doit(struct sk_buff *skb,
 +                                        struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = info->user_ptr[0];
 +      u32 count;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_PORT_SPLIT_COUNT))
 +              return -EINVAL;
 +      if (!devlink->ops->port_split)
 +              return -EOPNOTSUPP;
 +
 +      count = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_SPLIT_COUNT]);
 +
 +      if (!devlink_port->attrs.splittable) {
 +              /* Split ports cannot be split. */
 +              if (devlink_port->attrs.split)
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Port cannot be split further");
 +              else
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Port cannot be split");
 +              return -EINVAL;
 +      }
 +
 +      if (count < 2 || !is_power_of_2(count) || count > devlink_port->attrs.lanes) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "Invalid split count");
 +              return -EINVAL;
 +      }
 +
 +      return devlink->ops->port_split(devlink, devlink_port, count,
 +                                      info->extack);
 +}
 +
 +static int devlink_nl_cmd_port_unsplit_doit(struct sk_buff *skb,
 +                                          struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = info->user_ptr[0];
 +
 +      if (!devlink->ops->port_unsplit)
 +              return -EOPNOTSUPP;
 +      return devlink->ops->port_unsplit(devlink, devlink_port, info->extack);
 +}
 +
 +static int devlink_port_new_notify(struct devlink *devlink,
 +                                 unsigned int port_index,
 +                                 struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      lockdep_assert_held(&devlink->lock);
 +      devlink_port = devlink_port_get_by_index(devlink, port_index);
 +      if (!devlink_port) {
 +              err = -ENODEV;
 +              goto out;
 +      }
 +
 +      err = devlink_nl_port_fill(msg, devlink_port, DEVLINK_CMD_NEW,
 +                                 info->snd_portid, info->snd_seq, 0, NULL);
 +      if (err)
 +              goto out;
 +
 +      return genlmsg_reply(msg, info);
 +
 +out:
 +      nlmsg_free(msg);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_port_new_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink_port_new_attrs new_attrs = {};
 +      struct devlink *devlink = info->user_ptr[0];
 +      unsigned int new_port_index;
 +      int err;
 +
 +      if (!devlink->ops->port_new || !devlink->ops->port_del)
 +              return -EOPNOTSUPP;
 +
 +      if (!info->attrs[DEVLINK_ATTR_PORT_FLAVOUR] ||
 +          !info->attrs[DEVLINK_ATTR_PORT_PCI_PF_NUMBER]) {
 +              NL_SET_ERR_MSG_MOD(extack, "Port flavour or PCI PF are not specified");
 +              return -EINVAL;
 +      }
 +      new_attrs.flavour = nla_get_u16(info->attrs[DEVLINK_ATTR_PORT_FLAVOUR]);
 +      new_attrs.pfnum =
 +              nla_get_u16(info->attrs[DEVLINK_ATTR_PORT_PCI_PF_NUMBER]);
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              /* Port index of the new port being created by driver. */
 +              new_attrs.port_index =
 +                      nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +              new_attrs.port_index_valid = true;
 +      }
 +      if (info->attrs[DEVLINK_ATTR_PORT_CONTROLLER_NUMBER]) {
 +              new_attrs.controller =
 +                      nla_get_u16(info->attrs[DEVLINK_ATTR_PORT_CONTROLLER_NUMBER]);
 +              new_attrs.controller_valid = true;
 +      }
 +      if (new_attrs.flavour == DEVLINK_PORT_FLAVOUR_PCI_SF &&
 +          info->attrs[DEVLINK_ATTR_PORT_PCI_SF_NUMBER]) {
 +              new_attrs.sfnum = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_PCI_SF_NUMBER]);
 +              new_attrs.sfnum_valid = true;
 +      }
 +
 +      err = devlink->ops->port_new(devlink, &new_attrs, extack,
 +                                   &new_port_index);
 +      if (err)
 +              return err;
 +
 +      err = devlink_port_new_notify(devlink, new_port_index, info);
 +      if (err && err != -ENODEV) {
 +              /* Fail to send the response; destroy newly created port. */
 +              devlink->ops->port_del(devlink, new_port_index, extack);
 +      }
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_port_del_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      unsigned int port_index;
 +
 +      if (!devlink->ops->port_del)
 +              return -EOPNOTSUPP;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_PORT_INDEX)) {
 +              NL_SET_ERR_MSG_MOD(extack, "Port index is not specified");
 +              return -EINVAL;
 +      }
 +      port_index = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +
 +      return devlink->ops->port_del(devlink, port_index, extack);
 +}
 +
 +static int
 +devlink_nl_rate_parent_node_set(struct devlink_rate *devlink_rate,
 +                              struct genl_info *info,
 +                              struct nlattr *nla_parent)
 +{
 +      struct devlink *devlink = devlink_rate->devlink;
 +      const char *parent_name = nla_data(nla_parent);
 +      const struct devlink_ops *ops = devlink->ops;
 +      size_t len = strlen(parent_name);
 +      struct devlink_rate *parent;
 +      int err = -EOPNOTSUPP;
 +
 +      parent = devlink_rate->parent;
 +
 +      if (parent && !len) {
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_parent_set(devlink_rate, NULL,
 +                                                      devlink_rate->priv, NULL,
 +                                                      info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_parent_set(devlink_rate, NULL,
 +                                                      devlink_rate->priv, NULL,
 +                                                      info->extack);
 +              if (err)
 +                      return err;
 +
 +              refcount_dec(&parent->refcnt);
 +              devlink_rate->parent = NULL;
 +      } else if (len) {
 +              parent = devlink_rate_node_get_by_name(devlink, parent_name);
 +              if (IS_ERR(parent))
 +                      return -ENODEV;
 +
 +              if (parent == devlink_rate) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Parent to self is not allowed");
 +                      return -EINVAL;
 +              }
 +
 +              if (devlink_rate_is_node(devlink_rate) &&
 +                  devlink_rate_is_parent_node(devlink_rate, parent->parent)) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Node is already a parent of parent node.");
 +                      return -EEXIST;
 +              }
 +
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_parent_set(devlink_rate, parent,
 +                                                      devlink_rate->priv, parent->priv,
 +                                                      info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_parent_set(devlink_rate, parent,
 +                                                      devlink_rate->priv, parent->priv,
 +                                                      info->extack);
 +              if (err)
 +                      return err;
 +
 +              if (devlink_rate->parent)
 +                      /* we're reassigning to other parent in this case */
 +                      refcount_dec(&devlink_rate->parent->refcnt);
 +
 +              refcount_inc(&parent->refcnt);
 +              devlink_rate->parent = parent;
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_rate_set(struct devlink_rate *devlink_rate,
 +                             const struct devlink_ops *ops,
 +                             struct genl_info *info)
 +{
 +      struct nlattr *nla_parent, **attrs = info->attrs;
 +      int err = -EOPNOTSUPP;
 +      u32 priority;
 +      u32 weight;
 +      u64 rate;
 +
 +      if (attrs[DEVLINK_ATTR_RATE_TX_SHARE]) {
 +              rate = nla_get_u64(attrs[DEVLINK_ATTR_RATE_TX_SHARE]);
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_tx_share_set(devlink_rate, devlink_rate->priv,
 +                                                        rate, info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_tx_share_set(devlink_rate, devlink_rate->priv,
 +                                                        rate, info->extack);
 +              if (err)
 +                      return err;
 +              devlink_rate->tx_share = rate;
 +      }
 +
 +      if (attrs[DEVLINK_ATTR_RATE_TX_MAX]) {
 +              rate = nla_get_u64(attrs[DEVLINK_ATTR_RATE_TX_MAX]);
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_tx_max_set(devlink_rate, devlink_rate->priv,
 +                                                      rate, info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_tx_max_set(devlink_rate, devlink_rate->priv,
 +                                                      rate, info->extack);
 +              if (err)
 +                      return err;
 +              devlink_rate->tx_max = rate;
 +      }
 +
 +      if (attrs[DEVLINK_ATTR_RATE_TX_PRIORITY]) {
 +              priority = nla_get_u32(attrs[DEVLINK_ATTR_RATE_TX_PRIORITY]);
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_tx_priority_set(devlink_rate, devlink_rate->priv,
 +                                                           priority, info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_tx_priority_set(devlink_rate, devlink_rate->priv,
 +                                                           priority, info->extack);
 +
 +              if (err)
 +                      return err;
 +              devlink_rate->tx_priority = priority;
 +      }
 +
 +      if (attrs[DEVLINK_ATTR_RATE_TX_WEIGHT]) {
 +              weight = nla_get_u32(attrs[DEVLINK_ATTR_RATE_TX_WEIGHT]);
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      err = ops->rate_leaf_tx_weight_set(devlink_rate, devlink_rate->priv,
 +                                                         weight, info->extack);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      err = ops->rate_node_tx_weight_set(devlink_rate, devlink_rate->priv,
 +                                                         weight, info->extack);
 +
 +              if (err)
 +                      return err;
 +              devlink_rate->tx_weight = weight;
 +      }
 +
 +      nla_parent = attrs[DEVLINK_ATTR_RATE_PARENT_NODE_NAME];
 +      if (nla_parent) {
 +              err = devlink_nl_rate_parent_node_set(devlink_rate, info,
 +                                                    nla_parent);
 +              if (err)
 +                      return err;
 +      }
 +
 +      return 0;
 +}
 +
 +static bool devlink_rate_set_ops_supported(const struct devlink_ops *ops,
 +                                         struct genl_info *info,
 +                                         enum devlink_rate_type type)
 +{
 +      struct nlattr **attrs = info->attrs;
 +
 +      if (type == DEVLINK_RATE_TYPE_LEAF) {
 +              if (attrs[DEVLINK_ATTR_RATE_TX_SHARE] && !ops->rate_leaf_tx_share_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "TX share set isn't supported for the leafs");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_MAX] && !ops->rate_leaf_tx_max_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "TX max set isn't supported for the leafs");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_PARENT_NODE_NAME] &&
 +                  !ops->rate_leaf_parent_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Parent set isn't supported for the leafs");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_PRIORITY] && !ops->rate_leaf_tx_priority_set) {
 +                      NL_SET_ERR_MSG_ATTR(info->extack,
 +                                          attrs[DEVLINK_ATTR_RATE_TX_PRIORITY],
 +                                          "TX priority set isn't supported for the leafs");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_WEIGHT] && !ops->rate_leaf_tx_weight_set) {
 +                      NL_SET_ERR_MSG_ATTR(info->extack,
 +                                          attrs[DEVLINK_ATTR_RATE_TX_WEIGHT],
 +                                          "TX weight set isn't supported for the leafs");
 +                      return false;
 +              }
 +      } else if (type == DEVLINK_RATE_TYPE_NODE) {
 +              if (attrs[DEVLINK_ATTR_RATE_TX_SHARE] && !ops->rate_node_tx_share_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "TX share set isn't supported for the nodes");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_MAX] && !ops->rate_node_tx_max_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "TX max set isn't supported for the nodes");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_PARENT_NODE_NAME] &&
 +                  !ops->rate_node_parent_set) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Parent set isn't supported for the nodes");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_PRIORITY] && !ops->rate_node_tx_priority_set) {
 +                      NL_SET_ERR_MSG_ATTR(info->extack,
 +                                          attrs[DEVLINK_ATTR_RATE_TX_PRIORITY],
 +                                          "TX priority set isn't supported for the nodes");
 +                      return false;
 +              }
 +              if (attrs[DEVLINK_ATTR_RATE_TX_WEIGHT] && !ops->rate_node_tx_weight_set) {
 +                      NL_SET_ERR_MSG_ATTR(info->extack,
 +                                          attrs[DEVLINK_ATTR_RATE_TX_WEIGHT],
 +                                          "TX weight set isn't supported for the nodes");
 +                      return false;
 +              }
 +      } else {
 +              WARN(1, "Unknown type of rate object");
 +              return false;
 +      }
 +
 +      return true;
 +}
 +
 +static int devlink_nl_cmd_rate_set_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink_rate *devlink_rate = info->user_ptr[1];
 +      struct devlink *devlink = devlink_rate->devlink;
 +      const struct devlink_ops *ops = devlink->ops;
 +      int err;
 +
 +      if (!ops || !devlink_rate_set_ops_supported(ops, info, devlink_rate->type))
 +              return -EOPNOTSUPP;
 +
 +      err = devlink_nl_rate_set(devlink_rate, ops, info);
 +
 +      if (!err)
 +              devlink_rate_notify(devlink_rate, DEVLINK_CMD_RATE_NEW);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_rate_new_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_rate *rate_node;
 +      const struct devlink_ops *ops;
 +      int err;
 +
 +      ops = devlink->ops;
 +      if (!ops || !ops->rate_node_new || !ops->rate_node_del) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "Rate nodes aren't supported");
 +              return -EOPNOTSUPP;
 +      }
 +
 +      if (!devlink_rate_set_ops_supported(ops, info, DEVLINK_RATE_TYPE_NODE))
 +              return -EOPNOTSUPP;
 +
 +      rate_node = devlink_rate_node_get_from_attrs(devlink, info->attrs);
 +      if (!IS_ERR(rate_node))
 +              return -EEXIST;
 +      else if (rate_node == ERR_PTR(-EINVAL))
 +              return -EINVAL;
 +
 +      rate_node = kzalloc(sizeof(*rate_node), GFP_KERNEL);
 +      if (!rate_node)
 +              return -ENOMEM;
 +
 +      rate_node->devlink = devlink;
 +      rate_node->type = DEVLINK_RATE_TYPE_NODE;
 +      rate_node->name = nla_strdup(info->attrs[DEVLINK_ATTR_RATE_NODE_NAME], GFP_KERNEL);
 +      if (!rate_node->name) {
 +              err = -ENOMEM;
 +              goto err_strdup;
 +      }
 +
 +      err = ops->rate_node_new(rate_node, &rate_node->priv, info->extack);
 +      if (err)
 +              goto err_node_new;
 +
 +      err = devlink_nl_rate_set(rate_node, ops, info);
 +      if (err)
 +              goto err_rate_set;
 +
 +      refcount_set(&rate_node->refcnt, 1);
 +      list_add(&rate_node->list, &devlink->rate_list);
 +      devlink_rate_notify(rate_node, DEVLINK_CMD_RATE_NEW);
 +      return 0;
 +
 +err_rate_set:
 +      ops->rate_node_del(rate_node, rate_node->priv, info->extack);
 +err_node_new:
 +      kfree(rate_node->name);
 +err_strdup:
 +      kfree(rate_node);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_rate_del_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink_rate *rate_node = info->user_ptr[1];
 +      struct devlink *devlink = rate_node->devlink;
 +      const struct devlink_ops *ops = devlink->ops;
 +      int err;
 +
 +      if (refcount_read(&rate_node->refcnt) > 1) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "Node has children. Cannot delete node.");
 +              return -EBUSY;
 +      }
 +
 +      devlink_rate_notify(rate_node, DEVLINK_CMD_RATE_DEL);
 +      err = ops->rate_node_del(rate_node, rate_node->priv, info->extack);
 +      if (rate_node->parent)
 +              refcount_dec(&rate_node->parent->refcnt);
 +      list_del(&rate_node->list);
 +      kfree(rate_node->name);
 +      kfree(rate_node);
 +      return err;
 +}
 +
 +struct devlink_linecard_type {
 +      const char *type;
 +      const void *priv;
 +};
 +
 +static int devlink_nl_linecard_fill(struct sk_buff *msg,
 +                                  struct devlink *devlink,
 +                                  struct devlink_linecard *linecard,
 +                                  enum devlink_command cmd, u32 portid,
 +                                  u32 seq, int flags,
 +                                  struct netlink_ext_ack *extack)
 +{
 +      struct devlink_linecard_type *linecard_type;
 +      struct nlattr *attr;
 +      void *hdr;
 +      int i;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_LINECARD_INDEX, linecard->index))
 +              goto nla_put_failure;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_LINECARD_STATE, linecard->state))
 +              goto nla_put_failure;
 +      if (linecard->type &&
 +          nla_put_string(msg, DEVLINK_ATTR_LINECARD_TYPE, linecard->type))
 +              goto nla_put_failure;
 +
 +      if (linecard->types_count) {
 +              attr = nla_nest_start(msg,
 +                                    DEVLINK_ATTR_LINECARD_SUPPORTED_TYPES);
 +              if (!attr)
 +                      goto nla_put_failure;
 +              for (i = 0; i < linecard->types_count; i++) {
 +                      linecard_type = &linecard->types[i];
 +                      if (nla_put_string(msg, DEVLINK_ATTR_LINECARD_TYPE,
 +                                         linecard_type->type)) {
 +                              nla_nest_cancel(msg, attr);
 +                              goto nla_put_failure;
 +                      }
 +              }
 +              nla_nest_end(msg, attr);
 +      }
 +
 +      if (linecard->nested_devlink &&
 +          devlink_nl_put_nested_handle(msg, linecard->nested_devlink))
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static void devlink_linecard_notify(struct devlink_linecard *linecard,
 +                                  enum devlink_command cmd)
 +{
 +      struct devlink *devlink = linecard->devlink;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_LINECARD_NEW &&
 +              cmd != DEVLINK_CMD_LINECARD_DEL);
 +
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_linecard_fill(msg, devlink, linecard, cmd, 0, 0, 0,
 +                                     NULL);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink),
 +                              msg, 0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int devlink_nl_cmd_linecard_get_doit(struct sk_buff *skb,
 +                                          struct genl_info *info)
 +{
 +      struct devlink_linecard *linecard = info->user_ptr[1];
 +      struct devlink *devlink = linecard->devlink;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      mutex_lock(&linecard->state_lock);
 +      err = devlink_nl_linecard_fill(msg, devlink, linecard,
 +                                     DEVLINK_CMD_LINECARD_NEW,
 +                                     info->snd_portid, info->snd_seq, 0,
 +                                     info->extack);
 +      mutex_unlock(&linecard->state_lock);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int devlink_nl_cmd_linecard_get_dump_one(struct sk_buff *msg,
 +                                              struct devlink *devlink,
 +                                              struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_linecard *linecard;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(linecard, &devlink->linecard_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              mutex_lock(&linecard->state_lock);
 +              err = devlink_nl_linecard_fill(msg, devlink, linecard,
 +                                             DEVLINK_CMD_LINECARD_NEW,
 +                                             NETLINK_CB(cb->skb).portid,
 +                                             cb->nlh->nlmsg_seq,
 +                                             NLM_F_MULTI,
 +                                             cb->extack);
 +              mutex_unlock(&linecard->state_lock);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_linecard_get = {
 +      .dump_one               = devlink_nl_cmd_linecard_get_dump_one,
 +};
 +
 +static struct devlink_linecard_type *
 +devlink_linecard_type_lookup(struct devlink_linecard *linecard,
 +                           const char *type)
 +{
 +      struct devlink_linecard_type *linecard_type;
 +      int i;
 +
 +      for (i = 0; i < linecard->types_count; i++) {
 +              linecard_type = &linecard->types[i];
 +              if (!strcmp(type, linecard_type->type))
 +                      return linecard_type;
 +      }
 +      return NULL;
 +}
 +
 +static int devlink_linecard_type_set(struct devlink_linecard *linecard,
 +                                   const char *type,
 +                                   struct netlink_ext_ack *extack)
 +{
 +      const struct devlink_linecard_ops *ops = linecard->ops;
 +      struct devlink_linecard_type *linecard_type;
 +      int err;
 +
 +      mutex_lock(&linecard->state_lock);
 +      if (linecard->state == DEVLINK_LINECARD_STATE_PROVISIONING) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card is currently being provisioned");
 +              err = -EBUSY;
 +              goto out;
 +      }
 +      if (linecard->state == DEVLINK_LINECARD_STATE_UNPROVISIONING) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card is currently being unprovisioned");
 +              err = -EBUSY;
 +              goto out;
 +      }
 +
 +      linecard_type = devlink_linecard_type_lookup(linecard, type);
 +      if (!linecard_type) {
 +              NL_SET_ERR_MSG_MOD(extack, "Unsupported line card type provided");
 +              err = -EINVAL;
 +              goto out;
 +      }
 +
 +      if (linecard->state != DEVLINK_LINECARD_STATE_UNPROVISIONED &&
 +          linecard->state != DEVLINK_LINECARD_STATE_PROVISIONING_FAILED) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card already provisioned");
 +              err = -EBUSY;
 +              /* Check if the line card is provisioned in the same
 +               * way the user asks. In case it is, make the operation
 +               * to return success.
 +               */
 +              if (ops->same_provision &&
 +                  ops->same_provision(linecard, linecard->priv,
 +                                      linecard_type->type,
 +                                      linecard_type->priv))
 +                      err = 0;
 +              goto out;
 +      }
 +
 +      linecard->state = DEVLINK_LINECARD_STATE_PROVISIONING;
 +      linecard->type = linecard_type->type;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +      err = ops->provision(linecard, linecard->priv, linecard_type->type,
 +                           linecard_type->priv, extack);
 +      if (err) {
 +              /* Provisioning failed. Assume the linecard is unprovisioned
 +               * for future operations.
 +               */
 +              mutex_lock(&linecard->state_lock);
 +              linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONED;
 +              linecard->type = NULL;
 +              devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +              mutex_unlock(&linecard->state_lock);
 +      }
 +      return err;
 +
 +out:
 +      mutex_unlock(&linecard->state_lock);
 +      return err;
 +}
 +
 +static int devlink_linecard_type_unset(struct devlink_linecard *linecard,
 +                                     struct netlink_ext_ack *extack)
 +{
 +      int err;
 +
 +      mutex_lock(&linecard->state_lock);
 +      if (linecard->state == DEVLINK_LINECARD_STATE_PROVISIONING) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card is currently being provisioned");
 +              err = -EBUSY;
 +              goto out;
 +      }
 +      if (linecard->state == DEVLINK_LINECARD_STATE_UNPROVISIONING) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card is currently being unprovisioned");
 +              err = -EBUSY;
 +              goto out;
 +      }
 +      if (linecard->state == DEVLINK_LINECARD_STATE_PROVISIONING_FAILED) {
 +              linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONED;
 +              linecard->type = NULL;
 +              devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +              err = 0;
 +              goto out;
 +      }
 +
 +      if (linecard->state == DEVLINK_LINECARD_STATE_UNPROVISIONED) {
 +              NL_SET_ERR_MSG_MOD(extack, "Line card is not provisioned");
 +              err = 0;
 +              goto out;
 +      }
 +      linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONING;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +      err = linecard->ops->unprovision(linecard, linecard->priv,
 +                                       extack);
 +      if (err) {
 +              /* Unprovisioning failed. Assume the linecard is unprovisioned
 +               * for future operations.
 +               */
 +              mutex_lock(&linecard->state_lock);
 +              linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONED;
 +              linecard->type = NULL;
 +              devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +              mutex_unlock(&linecard->state_lock);
 +      }
 +      return err;
 +
 +out:
 +      mutex_unlock(&linecard->state_lock);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_linecard_set_doit(struct sk_buff *skb,
 +                                          struct genl_info *info)
 +{
 +      struct devlink_linecard *linecard = info->user_ptr[1];
 +      struct netlink_ext_ack *extack = info->extack;
 +      int err;
 +
 +      if (info->attrs[DEVLINK_ATTR_LINECARD_TYPE]) {
 +              const char *type;
 +
 +              type = nla_data(info->attrs[DEVLINK_ATTR_LINECARD_TYPE]);
 +              if (strcmp(type, "")) {
 +                      err = devlink_linecard_type_set(linecard, type, extack);
 +                      if (err)
 +                              return err;
 +              } else {
 +                      err = devlink_linecard_type_unset(linecard, extack);
 +                      if (err)
 +                              return err;
 +              }
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_sb_fill(struct sk_buff *msg, struct devlink *devlink,
 +                            struct devlink_sb *devlink_sb,
 +                            enum devlink_command cmd, u32 portid,
 +                            u32 seq, int flags)
 +{
 +      void *hdr;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_INDEX, devlink_sb->index))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_SIZE, devlink_sb->size))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_INGRESS_POOL_COUNT,
 +                      devlink_sb->ingress_pools_count))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_EGRESS_POOL_COUNT,
 +                      devlink_sb->egress_pools_count))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_INGRESS_TC_COUNT,
 +                      devlink_sb->ingress_tc_count))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_EGRESS_TC_COUNT,
 +                      devlink_sb->egress_tc_count))
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_sb_get_doit(struct sk_buff *skb,
 +                                    struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_sb *devlink_sb;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_sb_fill(msg, devlink, devlink_sb,
 +                               DEVLINK_CMD_SB_NEW,
 +                               info->snd_portid, info->snd_seq, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int
 +devlink_nl_cmd_sb_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                             struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_sb *devlink_sb;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_sb_fill(msg, devlink, devlink_sb,
 +                                       DEVLINK_CMD_SB_NEW,
 +                                       NETLINK_CB(cb->skb).portid,
 +                                       cb->nlh->nlmsg_seq,
 +                                       NLM_F_MULTI);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_sb_get = {
 +      .dump_one               = devlink_nl_cmd_sb_get_dump_one,
 +};
 +
 +static int devlink_nl_sb_pool_fill(struct sk_buff *msg, struct devlink *devlink,
 +                                 struct devlink_sb *devlink_sb,
 +                                 u16 pool_index, enum devlink_command cmd,
 +                                 u32 portid, u32 seq, int flags)
 +{
 +      struct devlink_sb_pool_info pool_info;
 +      void *hdr;
 +      int err;
 +
 +      err = devlink->ops->sb_pool_get(devlink, devlink_sb->index,
 +                                      pool_index, &pool_info);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_INDEX, devlink_sb->index))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_POOL_INDEX, pool_index))
 +              goto nla_put_failure;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_SB_POOL_TYPE, pool_info.pool_type))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_POOL_SIZE, pool_info.size))
 +              goto nla_put_failure;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_SB_POOL_THRESHOLD_TYPE,
 +                     pool_info.threshold_type))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_POOL_CELL_SIZE,
 +                      pool_info.cell_size))
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_sb_pool_get_doit(struct sk_buff *skb,
 +                                         struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_sb *devlink_sb;
 +      struct sk_buff *msg;
 +      u16 pool_index;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_index_get_from_info(devlink_sb, info,
 +                                                &pool_index);
 +      if (err)
 +              return err;
 +
 +      if (!devlink->ops->sb_pool_get)
 +              return -EOPNOTSUPP;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_sb_pool_fill(msg, devlink, devlink_sb, pool_index,
 +                                    DEVLINK_CMD_SB_POOL_NEW,
 +                                    info->snd_portid, info->snd_seq, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int __sb_pool_get_dumpit(struct sk_buff *msg, int start, int *p_idx,
 +                              struct devlink *devlink,
 +                              struct devlink_sb *devlink_sb,
 +                              u32 portid, u32 seq)
 +{
 +      u16 pool_count = devlink_sb_pool_count(devlink_sb);
 +      u16 pool_index;
 +      int err;
 +
 +      for (pool_index = 0; pool_index < pool_count; pool_index++) {
 +              if (*p_idx < start) {
 +                      (*p_idx)++;
 +                      continue;
 +              }
 +              err = devlink_nl_sb_pool_fill(msg, devlink,
 +                                            devlink_sb,
 +                                            pool_index,
 +                                            DEVLINK_CMD_SB_POOL_NEW,
 +                                            portid, seq, NLM_F_MULTI);
 +              if (err)
 +                      return err;
 +              (*p_idx)++;
 +      }
 +      return 0;
 +}
 +
 +static int
 +devlink_nl_cmd_sb_pool_get_dump_one(struct sk_buff *msg,
 +                                  struct devlink *devlink,
 +                                  struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_sb *devlink_sb;
 +      int err = 0;
 +      int idx = 0;
 +
 +      if (!devlink->ops->sb_pool_get)
 +              return 0;
 +
 +      list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 +              err = __sb_pool_get_dumpit(msg, state->idx, &idx,
 +                                         devlink, devlink_sb,
 +                                         NETLINK_CB(cb->skb).portid,
 +                                         cb->nlh->nlmsg_seq);
 +              if (err == -EOPNOTSUPP) {
 +                      err = 0;
 +              } else if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_sb_pool_get = {
 +      .dump_one               = devlink_nl_cmd_sb_pool_get_dump_one,
 +};
 +
 +static int devlink_sb_pool_set(struct devlink *devlink, unsigned int sb_index,
 +                             u16 pool_index, u32 size,
 +                             enum devlink_sb_threshold_type threshold_type,
 +                             struct netlink_ext_ack *extack)
 +
 +{
 +      const struct devlink_ops *ops = devlink->ops;
 +
 +      if (ops->sb_pool_set)
 +              return ops->sb_pool_set(devlink, sb_index, pool_index,
 +                                      size, threshold_type, extack);
 +      return -EOPNOTSUPP;
 +}
 +
 +static int devlink_nl_cmd_sb_pool_set_doit(struct sk_buff *skb,
 +                                         struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      enum devlink_sb_threshold_type threshold_type;
 +      struct devlink_sb *devlink_sb;
 +      u16 pool_index;
 +      u32 size;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_index_get_from_info(devlink_sb, info,
 +                                                &pool_index);
 +      if (err)
 +              return err;
 +
 +      err = devlink_sb_th_type_get_from_info(info, &threshold_type);
 +      if (err)
 +              return err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_SB_POOL_SIZE))
 +              return -EINVAL;
 +
 +      size = nla_get_u32(info->attrs[DEVLINK_ATTR_SB_POOL_SIZE]);
 +      return devlink_sb_pool_set(devlink, devlink_sb->index,
 +                                 pool_index, size, threshold_type,
 +                                 info->extack);
 +}
 +
 +static int devlink_nl_sb_port_pool_fill(struct sk_buff *msg,
 +                                      struct devlink *devlink,
 +                                      struct devlink_port *devlink_port,
 +                                      struct devlink_sb *devlink_sb,
 +                                      u16 pool_index,
 +                                      enum devlink_command cmd,
 +                                      u32 portid, u32 seq, int flags)
 +{
 +      const struct devlink_ops *ops = devlink->ops;
 +      u32 threshold;
 +      void *hdr;
 +      int err;
 +
 +      err = ops->sb_port_pool_get(devlink_port, devlink_sb->index,
 +                                  pool_index, &threshold);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, devlink_port->index))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_INDEX, devlink_sb->index))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_POOL_INDEX, pool_index))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_THRESHOLD, threshold))
 +              goto nla_put_failure;
 +
 +      if (ops->sb_occ_port_pool_get) {
 +              u32 cur;
 +              u32 max;
 +
 +              err = ops->sb_occ_port_pool_get(devlink_port, devlink_sb->index,
 +                                              pool_index, &cur, &max);
 +              if (err && err != -EOPNOTSUPP)
 +                      goto sb_occ_get_failure;
 +              if (!err) {
 +                      if (nla_put_u32(msg, DEVLINK_ATTR_SB_OCC_CUR, cur))
 +                              goto nla_put_failure;
 +                      if (nla_put_u32(msg, DEVLINK_ATTR_SB_OCC_MAX, max))
 +                              goto nla_put_failure;
 +              }
 +      }
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +sb_occ_get_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_sb_port_pool_get_doit(struct sk_buff *skb,
 +                                              struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = devlink_port->devlink;
 +      struct devlink_sb *devlink_sb;
 +      struct sk_buff *msg;
 +      u16 pool_index;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_index_get_from_info(devlink_sb, info,
 +                                                &pool_index);
 +      if (err)
 +              return err;
 +
 +      if (!devlink->ops->sb_port_pool_get)
 +              return -EOPNOTSUPP;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_sb_port_pool_fill(msg, devlink, devlink_port,
 +                                         devlink_sb, pool_index,
 +                                         DEVLINK_CMD_SB_PORT_POOL_NEW,
 +                                         info->snd_portid, info->snd_seq, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int __sb_port_pool_get_dumpit(struct sk_buff *msg, int start, int *p_idx,
 +                                   struct devlink *devlink,
 +                                   struct devlink_sb *devlink_sb,
 +                                   u32 portid, u32 seq)
 +{
 +      struct devlink_port *devlink_port;
 +      u16 pool_count = devlink_sb_pool_count(devlink_sb);
 +      unsigned long port_index;
 +      u16 pool_index;
 +      int err;
 +
 +      xa_for_each(&devlink->ports, port_index, devlink_port) {
 +              for (pool_index = 0; pool_index < pool_count; pool_index++) {
 +                      if (*p_idx < start) {
 +                              (*p_idx)++;
 +                              continue;
 +                      }
 +                      err = devlink_nl_sb_port_pool_fill(msg, devlink,
 +                                                         devlink_port,
 +                                                         devlink_sb,
 +                                                         pool_index,
 +                                                         DEVLINK_CMD_SB_PORT_POOL_NEW,
 +                                                         portid, seq,
 +                                                         NLM_F_MULTI);
 +                      if (err)
 +                              return err;
 +                      (*p_idx)++;
 +              }
 +      }
 +      return 0;
 +}
 +
 +static int
 +devlink_nl_cmd_sb_port_pool_get_dump_one(struct sk_buff *msg,
 +                                       struct devlink *devlink,
 +                                       struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_sb *devlink_sb;
 +      int idx = 0;
 +      int err = 0;
 +
 +      if (!devlink->ops->sb_port_pool_get)
 +              return 0;
 +
 +      list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 +              err = __sb_port_pool_get_dumpit(msg, state->idx, &idx,
 +                                              devlink, devlink_sb,
 +                                              NETLINK_CB(cb->skb).portid,
 +                                              cb->nlh->nlmsg_seq);
 +              if (err == -EOPNOTSUPP) {
 +                      err = 0;
 +              } else if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_sb_port_pool_get = {
 +      .dump_one               = devlink_nl_cmd_sb_port_pool_get_dump_one,
 +};
 +
 +static int devlink_sb_port_pool_set(struct devlink_port *devlink_port,
 +                                  unsigned int sb_index, u16 pool_index,
 +                                  u32 threshold,
 +                                  struct netlink_ext_ack *extack)
 +
 +{
 +      const struct devlink_ops *ops = devlink_port->devlink->ops;
 +
 +      if (ops->sb_port_pool_set)
 +              return ops->sb_port_pool_set(devlink_port, sb_index,
 +                                           pool_index, threshold, extack);
 +      return -EOPNOTSUPP;
 +}
 +
 +static int devlink_nl_cmd_sb_port_pool_set_doit(struct sk_buff *skb,
 +                                              struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_sb *devlink_sb;
 +      u16 pool_index;
 +      u32 threshold;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_index_get_from_info(devlink_sb, info,
 +                                                &pool_index);
 +      if (err)
 +              return err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_SB_THRESHOLD))
 +              return -EINVAL;
 +
 +      threshold = nla_get_u32(info->attrs[DEVLINK_ATTR_SB_THRESHOLD]);
 +      return devlink_sb_port_pool_set(devlink_port, devlink_sb->index,
 +                                      pool_index, threshold, info->extack);
 +}
 +
 +static int
 +devlink_nl_sb_tc_pool_bind_fill(struct sk_buff *msg, struct devlink *devlink,
 +                              struct devlink_port *devlink_port,
 +                              struct devlink_sb *devlink_sb, u16 tc_index,
 +                              enum devlink_sb_pool_type pool_type,
 +                              enum devlink_command cmd,
 +                              u32 portid, u32 seq, int flags)
 +{
 +      const struct devlink_ops *ops = devlink->ops;
 +      u16 pool_index;
 +      u32 threshold;
 +      void *hdr;
 +      int err;
 +
 +      err = ops->sb_tc_pool_bind_get(devlink_port, devlink_sb->index,
 +                                     tc_index, pool_type,
 +                                     &pool_index, &threshold);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, devlink_port->index))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_INDEX, devlink_sb->index))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_TC_INDEX, tc_index))
 +              goto nla_put_failure;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_SB_POOL_TYPE, pool_type))
 +              goto nla_put_failure;
 +      if (nla_put_u16(msg, DEVLINK_ATTR_SB_POOL_INDEX, pool_index))
 +              goto nla_put_failure;
 +      if (nla_put_u32(msg, DEVLINK_ATTR_SB_THRESHOLD, threshold))
 +              goto nla_put_failure;
 +
 +      if (ops->sb_occ_tc_port_bind_get) {
 +              u32 cur;
 +              u32 max;
 +
 +              err = ops->sb_occ_tc_port_bind_get(devlink_port,
 +                                                 devlink_sb->index,
 +                                                 tc_index, pool_type,
 +                                                 &cur, &max);
 +              if (err && err != -EOPNOTSUPP)
 +                      return err;
 +              if (!err) {
 +                      if (nla_put_u32(msg, DEVLINK_ATTR_SB_OCC_CUR, cur))
 +                              goto nla_put_failure;
 +                      if (nla_put_u32(msg, DEVLINK_ATTR_SB_OCC_MAX, max))
 +                              goto nla_put_failure;
 +              }
 +      }
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_sb_tc_pool_bind_get_doit(struct sk_buff *skb,
 +                                                 struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = devlink_port->devlink;
 +      struct devlink_sb *devlink_sb;
 +      struct sk_buff *msg;
 +      enum devlink_sb_pool_type pool_type;
 +      u16 tc_index;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_type_get_from_info(info, &pool_type);
 +      if (err)
 +              return err;
 +
 +      err = devlink_sb_tc_index_get_from_info(devlink_sb, info,
 +                                              pool_type, &tc_index);
 +      if (err)
 +              return err;
 +
 +      if (!devlink->ops->sb_tc_pool_bind_get)
 +              return -EOPNOTSUPP;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_sb_tc_pool_bind_fill(msg, devlink, devlink_port,
 +                                            devlink_sb, tc_index, pool_type,
 +                                            DEVLINK_CMD_SB_TC_POOL_BIND_NEW,
 +                                            info->snd_portid,
 +                                            info->snd_seq, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int __sb_tc_pool_bind_get_dumpit(struct sk_buff *msg,
 +                                      int start, int *p_idx,
 +                                      struct devlink *devlink,
 +                                      struct devlink_sb *devlink_sb,
 +                                      u32 portid, u32 seq)
 +{
 +      struct devlink_port *devlink_port;
 +      unsigned long port_index;
 +      u16 tc_index;
 +      int err;
 +
 +      xa_for_each(&devlink->ports, port_index, devlink_port) {
 +              for (tc_index = 0;
 +                   tc_index < devlink_sb->ingress_tc_count; tc_index++) {
 +                      if (*p_idx < start) {
 +                              (*p_idx)++;
 +                              continue;
 +                      }
 +                      err = devlink_nl_sb_tc_pool_bind_fill(msg, devlink,
 +                                                            devlink_port,
 +                                                            devlink_sb,
 +                                                            tc_index,
 +                                                            DEVLINK_SB_POOL_TYPE_INGRESS,
 +                                                            DEVLINK_CMD_SB_TC_POOL_BIND_NEW,
 +                                                            portid, seq,
 +                                                            NLM_F_MULTI);
 +                      if (err)
 +                              return err;
 +                      (*p_idx)++;
 +              }
 +              for (tc_index = 0;
 +                   tc_index < devlink_sb->egress_tc_count; tc_index++) {
 +                      if (*p_idx < start) {
 +                              (*p_idx)++;
 +                              continue;
 +                      }
 +                      err = devlink_nl_sb_tc_pool_bind_fill(msg, devlink,
 +                                                            devlink_port,
 +                                                            devlink_sb,
 +                                                            tc_index,
 +                                                            DEVLINK_SB_POOL_TYPE_EGRESS,
 +                                                            DEVLINK_CMD_SB_TC_POOL_BIND_NEW,
 +                                                            portid, seq,
 +                                                            NLM_F_MULTI);
 +                      if (err)
 +                              return err;
 +                      (*p_idx)++;
 +              }
 +      }
 +      return 0;
 +}
 +
 +static int
 +devlink_nl_cmd_sb_tc_pool_bind_get_dump_one(struct sk_buff *msg,
 +                                          struct devlink *devlink,
 +                                          struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_sb *devlink_sb;
 +      int idx = 0;
 +      int err = 0;
 +
 +      if (!devlink->ops->sb_tc_pool_bind_get)
 +              return 0;
 +
 +      list_for_each_entry(devlink_sb, &devlink->sb_list, list) {
 +              err = __sb_tc_pool_bind_get_dumpit(msg, state->idx, &idx,
 +                                                 devlink, devlink_sb,
 +                                                 NETLINK_CB(cb->skb).portid,
 +                                                 cb->nlh->nlmsg_seq);
 +              if (err == -EOPNOTSUPP) {
 +                      err = 0;
 +              } else if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_sb_tc_pool_bind_get = {
 +      .dump_one               = devlink_nl_cmd_sb_tc_pool_bind_get_dump_one,
 +};
 +
 +static int devlink_sb_tc_pool_bind_set(struct devlink_port *devlink_port,
 +                                     unsigned int sb_index, u16 tc_index,
 +                                     enum devlink_sb_pool_type pool_type,
 +                                     u16 pool_index, u32 threshold,
 +                                     struct netlink_ext_ack *extack)
 +
 +{
 +      const struct devlink_ops *ops = devlink_port->devlink->ops;
 +
 +      if (ops->sb_tc_pool_bind_set)
 +              return ops->sb_tc_pool_bind_set(devlink_port, sb_index,
 +                                              tc_index, pool_type,
 +                                              pool_index, threshold, extack);
 +      return -EOPNOTSUPP;
 +}
 +
 +static int devlink_nl_cmd_sb_tc_pool_bind_set_doit(struct sk_buff *skb,
 +                                                 struct genl_info *info)
 +{
 +      struct devlink_port *devlink_port = info->user_ptr[1];
 +      struct devlink *devlink = info->user_ptr[0];
 +      enum devlink_sb_pool_type pool_type;
 +      struct devlink_sb *devlink_sb;
 +      u16 tc_index;
 +      u16 pool_index;
 +      u32 threshold;
 +      int err;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      err = devlink_sb_pool_type_get_from_info(info, &pool_type);
 +      if (err)
 +              return err;
 +
 +      err = devlink_sb_tc_index_get_from_info(devlink_sb, info,
 +                                              pool_type, &tc_index);
 +      if (err)
 +              return err;
 +
 +      err = devlink_sb_pool_index_get_from_info(devlink_sb, info,
 +                                                &pool_index);
 +      if (err)
 +              return err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_SB_THRESHOLD))
 +              return -EINVAL;
 +
 +      threshold = nla_get_u32(info->attrs[DEVLINK_ATTR_SB_THRESHOLD]);
 +      return devlink_sb_tc_pool_bind_set(devlink_port, devlink_sb->index,
 +                                         tc_index, pool_type,
 +                                         pool_index, threshold, info->extack);
 +}
 +
 +static int devlink_nl_cmd_sb_occ_snapshot_doit(struct sk_buff *skb,
 +                                             struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      const struct devlink_ops *ops = devlink->ops;
 +      struct devlink_sb *devlink_sb;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      if (ops->sb_occ_snapshot)
 +              return ops->sb_occ_snapshot(devlink, devlink_sb->index);
 +      return -EOPNOTSUPP;
 +}
 +
 +static int devlink_nl_cmd_sb_occ_max_clear_doit(struct sk_buff *skb,
 +                                              struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      const struct devlink_ops *ops = devlink->ops;
 +      struct devlink_sb *devlink_sb;
 +
 +      devlink_sb = devlink_sb_get_from_info(devlink, info);
 +      if (IS_ERR(devlink_sb))
 +              return PTR_ERR(devlink_sb);
 +
 +      if (ops->sb_occ_max_clear)
 +              return ops->sb_occ_max_clear(devlink, devlink_sb->index);
 +      return -EOPNOTSUPP;
 +}
 +
 +int devlink_rate_nodes_check(struct devlink *devlink, u16 mode,
 +                           struct netlink_ext_ack *extack)
 +{
 +      struct devlink_rate *devlink_rate;
 +
 +      list_for_each_entry(devlink_rate, &devlink->rate_list, list)
 +              if (devlink_rate_is_node(devlink_rate)) {
 +                      NL_SET_ERR_MSG_MOD(extack, "Rate node(s) exists.");
 +                      return -EBUSY;
 +              }
 +      return 0;
 +}
 +
 +int devlink_dpipe_match_put(struct sk_buff *skb,
 +                          struct devlink_dpipe_match *match)
 +{
 +      struct devlink_dpipe_header *header = match->header;
 +      struct devlink_dpipe_field *field = &header->fields[match->field_id];
 +      struct nlattr *match_attr;
 +
 +      match_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_MATCH);
 +      if (!match_attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_u32(skb, DEVLINK_ATTR_DPIPE_MATCH_TYPE, match->type) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_HEADER_INDEX, match->header_index) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_HEADER_ID, header->id) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_FIELD_ID, field->id) ||
 +          nla_put_u8(skb, DEVLINK_ATTR_DPIPE_HEADER_GLOBAL, header->global))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(skb, match_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, match_attr);
 +      return -EMSGSIZE;
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_match_put);
 +
 +static int devlink_dpipe_matches_put(struct devlink_dpipe_table *table,
 +                                   struct sk_buff *skb)
 +{
 +      struct nlattr *matches_attr;
 +
 +      matches_attr = nla_nest_start_noflag(skb,
 +                                           DEVLINK_ATTR_DPIPE_TABLE_MATCHES);
 +      if (!matches_attr)
 +              return -EMSGSIZE;
 +
 +      if (table->table_ops->matches_dump(table->priv, skb))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(skb, matches_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, matches_attr);
 +      return -EMSGSIZE;
 +}
 +
 +int devlink_dpipe_action_put(struct sk_buff *skb,
 +                           struct devlink_dpipe_action *action)
 +{
 +      struct devlink_dpipe_header *header = action->header;
 +      struct devlink_dpipe_field *field = &header->fields[action->field_id];
 +      struct nlattr *action_attr;
 +
 +      action_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_ACTION);
 +      if (!action_attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_u32(skb, DEVLINK_ATTR_DPIPE_ACTION_TYPE, action->type) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_HEADER_INDEX, action->header_index) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_HEADER_ID, header->id) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_FIELD_ID, field->id) ||
 +          nla_put_u8(skb, DEVLINK_ATTR_DPIPE_HEADER_GLOBAL, header->global))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(skb, action_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, action_attr);
 +      return -EMSGSIZE;
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_action_put);
 +
 +static int devlink_dpipe_actions_put(struct devlink_dpipe_table *table,
 +                                   struct sk_buff *skb)
 +{
 +      struct nlattr *actions_attr;
 +
 +      actions_attr = nla_nest_start_noflag(skb,
 +                                           DEVLINK_ATTR_DPIPE_TABLE_ACTIONS);
 +      if (!actions_attr)
 +              return -EMSGSIZE;
 +
 +      if (table->table_ops->actions_dump(table->priv, skb))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(skb, actions_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, actions_attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_dpipe_table_put(struct sk_buff *skb,
 +                                 struct devlink_dpipe_table *table)
 +{
 +      struct nlattr *table_attr;
 +      u64 table_size;
 +
 +      table_size = table->table_ops->size_get(table->priv);
 +      table_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_TABLE);
 +      if (!table_attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_string(skb, DEVLINK_ATTR_DPIPE_TABLE_NAME, table->name) ||
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_DPIPE_TABLE_SIZE, table_size,
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +      if (nla_put_u8(skb, DEVLINK_ATTR_DPIPE_TABLE_COUNTERS_ENABLED,
 +                     table->counters_enabled))
 +              goto nla_put_failure;
 +
 +      if (table->resource_valid) {
 +              if (nla_put_u64_64bit(skb, DEVLINK_ATTR_DPIPE_TABLE_RESOURCE_ID,
 +                                    table->resource_id, DEVLINK_ATTR_PAD) ||
 +                  nla_put_u64_64bit(skb, DEVLINK_ATTR_DPIPE_TABLE_RESOURCE_UNITS,
 +                                    table->resource_units, DEVLINK_ATTR_PAD))
 +                      goto nla_put_failure;
 +      }
 +      if (devlink_dpipe_matches_put(table, skb))
 +              goto nla_put_failure;
 +
 +      if (devlink_dpipe_actions_put(table, skb))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(skb, table_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, table_attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_dpipe_send_and_alloc_skb(struct sk_buff **pskb,
 +                                          struct genl_info *info)
 +{
 +      int err;
 +
 +      if (*pskb) {
 +              err = genlmsg_reply(*pskb, info);
 +              if (err)
 +                      return err;
 +      }
 +      *pskb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!*pskb)
 +              return -ENOMEM;
 +      return 0;
 +}
 +
 +static int devlink_dpipe_tables_fill(struct genl_info *info,
 +                                   enum devlink_command cmd, int flags,
 +                                   struct list_head *dpipe_tables,
 +                                   const char *table_name)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_dpipe_table *table;
 +      struct nlattr *tables_attr;
 +      struct sk_buff *skb = NULL;
 +      struct nlmsghdr *nlh;
 +      bool incomplete;
 +      void *hdr;
 +      int i;
 +      int err;
 +
 +      table = list_first_entry(dpipe_tables,
 +                               struct devlink_dpipe_table, list);
 +start_again:
 +      err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                        &devlink_nl_family, NLM_F_MULTI, cmd);
 +      if (!hdr) {
 +              nlmsg_free(skb);
 +              return -EMSGSIZE;
 +      }
 +
 +      if (devlink_nl_put_handle(skb, devlink))
 +              goto nla_put_failure;
 +      tables_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_TABLES);
 +      if (!tables_attr)
 +              goto nla_put_failure;
 +
 +      i = 0;
 +      incomplete = false;
 +      list_for_each_entry_from(table, dpipe_tables, list) {
 +              if (!table_name) {
 +                      err = devlink_dpipe_table_put(skb, table);
 +                      if (err) {
 +                              if (!i)
 +                                      goto err_table_put;
 +                              incomplete = true;
 +                              break;
 +                      }
 +              } else {
 +                      if (!strcmp(table->name, table_name)) {
 +                              err = devlink_dpipe_table_put(skb, table);
 +                              if (err)
 +                                      break;
 +                      }
 +              }
 +              i++;
 +      }
 +
 +      nla_nest_end(skb, tables_attr);
 +      genlmsg_end(skb, hdr);
 +      if (incomplete)
 +              goto start_again;
 +
 +send_done:
 +      nlh = nlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                      NLMSG_DONE, 0, flags | NLM_F_MULTI);
 +      if (!nlh) {
 +              err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +              if (err)
 +                      return err;
 +              goto send_done;
 +      }
 +
 +      return genlmsg_reply(skb, info);
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +err_table_put:
 +      nlmsg_free(skb);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_dpipe_table_get(struct sk_buff *skb,
 +                                        struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      const char *table_name =  NULL;
 +
 +      if (info->attrs[DEVLINK_ATTR_DPIPE_TABLE_NAME])
 +              table_name = nla_data(info->attrs[DEVLINK_ATTR_DPIPE_TABLE_NAME]);
 +
 +      return devlink_dpipe_tables_fill(info, DEVLINK_CMD_DPIPE_TABLE_GET, 0,
 +                                       &devlink->dpipe_table_list,
 +                                       table_name);
 +}
 +
 +static int devlink_dpipe_value_put(struct sk_buff *skb,
 +                                 struct devlink_dpipe_value *value)
 +{
 +      if (nla_put(skb, DEVLINK_ATTR_DPIPE_VALUE,
 +                  value->value_size, value->value))
 +              return -EMSGSIZE;
 +      if (value->mask)
 +              if (nla_put(skb, DEVLINK_ATTR_DPIPE_VALUE_MASK,
 +                          value->value_size, value->mask))
 +                      return -EMSGSIZE;
 +      if (value->mapping_valid)
 +              if (nla_put_u32(skb, DEVLINK_ATTR_DPIPE_VALUE_MAPPING,
 +                              value->mapping_value))
 +                      return -EMSGSIZE;
 +      return 0;
 +}
 +
 +static int devlink_dpipe_action_value_put(struct sk_buff *skb,
 +                                        struct devlink_dpipe_value *value)
 +{
 +      if (!value->action)
 +              return -EINVAL;
 +      if (devlink_dpipe_action_put(skb, value->action))
 +              return -EMSGSIZE;
 +      if (devlink_dpipe_value_put(skb, value))
 +              return -EMSGSIZE;
 +      return 0;
 +}
 +
 +static int devlink_dpipe_action_values_put(struct sk_buff *skb,
 +                                         struct devlink_dpipe_value *values,
 +                                         unsigned int values_count)
 +{
 +      struct nlattr *action_attr;
 +      int i;
 +      int err;
 +
 +      for (i = 0; i < values_count; i++) {
 +              action_attr = nla_nest_start_noflag(skb,
 +                                                  DEVLINK_ATTR_DPIPE_ACTION_VALUE);
 +              if (!action_attr)
 +                      return -EMSGSIZE;
 +              err = devlink_dpipe_action_value_put(skb, &values[i]);
 +              if (err)
 +                      goto err_action_value_put;
 +              nla_nest_end(skb, action_attr);
 +      }
 +      return 0;
 +
 +err_action_value_put:
 +      nla_nest_cancel(skb, action_attr);
 +      return err;
 +}
 +
 +static int devlink_dpipe_match_value_put(struct sk_buff *skb,
 +                                       struct devlink_dpipe_value *value)
 +{
 +      if (!value->match)
 +              return -EINVAL;
 +      if (devlink_dpipe_match_put(skb, value->match))
 +              return -EMSGSIZE;
 +      if (devlink_dpipe_value_put(skb, value))
 +              return -EMSGSIZE;
 +      return 0;
 +}
 +
 +static int devlink_dpipe_match_values_put(struct sk_buff *skb,
 +                                        struct devlink_dpipe_value *values,
 +                                        unsigned int values_count)
 +{
 +      struct nlattr *match_attr;
 +      int i;
 +      int err;
 +
 +      for (i = 0; i < values_count; i++) {
 +              match_attr = nla_nest_start_noflag(skb,
 +                                                 DEVLINK_ATTR_DPIPE_MATCH_VALUE);
 +              if (!match_attr)
 +                      return -EMSGSIZE;
 +              err = devlink_dpipe_match_value_put(skb, &values[i]);
 +              if (err)
 +                      goto err_match_value_put;
 +              nla_nest_end(skb, match_attr);
 +      }
 +      return 0;
 +
 +err_match_value_put:
 +      nla_nest_cancel(skb, match_attr);
 +      return err;
 +}
 +
 +static int devlink_dpipe_entry_put(struct sk_buff *skb,
 +                                 struct devlink_dpipe_entry *entry)
 +{
 +      struct nlattr *entry_attr, *matches_attr, *actions_attr;
 +      int err;
 +
 +      entry_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_ENTRY);
 +      if (!entry_attr)
 +              return  -EMSGSIZE;
 +
 +      if (nla_put_u64_64bit(skb, DEVLINK_ATTR_DPIPE_ENTRY_INDEX, entry->index,
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +      if (entry->counter_valid)
 +              if (nla_put_u64_64bit(skb, DEVLINK_ATTR_DPIPE_ENTRY_COUNTER,
 +                                    entry->counter, DEVLINK_ATTR_PAD))
 +                      goto nla_put_failure;
 +
 +      matches_attr = nla_nest_start_noflag(skb,
 +                                           DEVLINK_ATTR_DPIPE_ENTRY_MATCH_VALUES);
 +      if (!matches_attr)
 +              goto nla_put_failure;
 +
 +      err = devlink_dpipe_match_values_put(skb, entry->match_values,
 +                                           entry->match_values_count);
 +      if (err) {
 +              nla_nest_cancel(skb, matches_attr);
 +              goto err_match_values_put;
 +      }
 +      nla_nest_end(skb, matches_attr);
 +
 +      actions_attr = nla_nest_start_noflag(skb,
 +                                           DEVLINK_ATTR_DPIPE_ENTRY_ACTION_VALUES);
 +      if (!actions_attr)
 +              goto nla_put_failure;
 +
 +      err = devlink_dpipe_action_values_put(skb, entry->action_values,
 +                                            entry->action_values_count);
 +      if (err) {
 +              nla_nest_cancel(skb, actions_attr);
 +              goto err_action_values_put;
 +      }
 +      nla_nest_end(skb, actions_attr);
 +
 +      nla_nest_end(skb, entry_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +err_match_values_put:
 +err_action_values_put:
 +      nla_nest_cancel(skb, entry_attr);
 +      return err;
 +}
 +
 +static struct devlink_dpipe_table *
 +devlink_dpipe_table_find(struct list_head *dpipe_tables,
 +                       const char *table_name, struct devlink *devlink)
 +{
 +      struct devlink_dpipe_table *table;
 +      list_for_each_entry_rcu(table, dpipe_tables, list,
 +                              lockdep_is_held(&devlink->lock)) {
 +              if (!strcmp(table->name, table_name))
 +                      return table;
 +      }
 +      return NULL;
 +}
 +
 +int devlink_dpipe_entry_ctx_prepare(struct devlink_dpipe_dump_ctx *dump_ctx)
 +{
 +      struct devlink *devlink;
 +      int err;
 +
 +      err = devlink_dpipe_send_and_alloc_skb(&dump_ctx->skb,
 +                                             dump_ctx->info);
 +      if (err)
 +              return err;
 +
 +      dump_ctx->hdr = genlmsg_put(dump_ctx->skb,
 +                                  dump_ctx->info->snd_portid,
 +                                  dump_ctx->info->snd_seq,
 +                                  &devlink_nl_family, NLM_F_MULTI,
 +                                  dump_ctx->cmd);
 +      if (!dump_ctx->hdr)
 +              goto nla_put_failure;
 +
 +      devlink = dump_ctx->info->user_ptr[0];
 +      if (devlink_nl_put_handle(dump_ctx->skb, devlink))
 +              goto nla_put_failure;
 +      dump_ctx->nest = nla_nest_start_noflag(dump_ctx->skb,
 +                                             DEVLINK_ATTR_DPIPE_ENTRIES);
 +      if (!dump_ctx->nest)
 +              goto nla_put_failure;
 +      return 0;
 +
 +nla_put_failure:
 +      nlmsg_free(dump_ctx->skb);
 +      return -EMSGSIZE;
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_entry_ctx_prepare);
 +
 +int devlink_dpipe_entry_ctx_append(struct devlink_dpipe_dump_ctx *dump_ctx,
 +                                 struct devlink_dpipe_entry *entry)
 +{
 +      return devlink_dpipe_entry_put(dump_ctx->skb, entry);
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_entry_ctx_append);
 +
 +int devlink_dpipe_entry_ctx_close(struct devlink_dpipe_dump_ctx *dump_ctx)
 +{
 +      nla_nest_end(dump_ctx->skb, dump_ctx->nest);
 +      genlmsg_end(dump_ctx->skb, dump_ctx->hdr);
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_entry_ctx_close);
 +
 +void devlink_dpipe_entry_clear(struct devlink_dpipe_entry *entry)
 +
 +{
 +      unsigned int value_count, value_index;
 +      struct devlink_dpipe_value *value;
 +
 +      value = entry->action_values;
 +      value_count = entry->action_values_count;
 +      for (value_index = 0; value_index < value_count; value_index++) {
 +              kfree(value[value_index].value);
 +              kfree(value[value_index].mask);
 +      }
 +
 +      value = entry->match_values;
 +      value_count = entry->match_values_count;
 +      for (value_index = 0; value_index < value_count; value_index++) {
 +              kfree(value[value_index].value);
 +              kfree(value[value_index].mask);
 +      }
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_entry_clear);
 +
 +static int devlink_dpipe_entries_fill(struct genl_info *info,
 +                                    enum devlink_command cmd, int flags,
 +                                    struct devlink_dpipe_table *table)
 +{
 +      struct devlink_dpipe_dump_ctx dump_ctx;
 +      struct nlmsghdr *nlh;
 +      int err;
 +
 +      dump_ctx.skb = NULL;
 +      dump_ctx.cmd = cmd;
 +      dump_ctx.info = info;
 +
 +      err = table->table_ops->entries_dump(table->priv,
 +                                           table->counters_enabled,
 +                                           &dump_ctx);
 +      if (err)
 +              return err;
 +
 +send_done:
 +      nlh = nlmsg_put(dump_ctx.skb, info->snd_portid, info->snd_seq,
 +                      NLMSG_DONE, 0, flags | NLM_F_MULTI);
 +      if (!nlh) {
 +              err = devlink_dpipe_send_and_alloc_skb(&dump_ctx.skb, info);
 +              if (err)
 +                      return err;
 +              goto send_done;
 +      }
 +      return genlmsg_reply(dump_ctx.skb, info);
 +}
 +
 +static int devlink_nl_cmd_dpipe_entries_get(struct sk_buff *skb,
 +                                          struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_dpipe_table *table;
 +      const char *table_name;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_DPIPE_TABLE_NAME))
 +              return -EINVAL;
 +
 +      table_name = nla_data(info->attrs[DEVLINK_ATTR_DPIPE_TABLE_NAME]);
 +      table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 +                                       table_name, devlink);
 +      if (!table)
 +              return -EINVAL;
 +
 +      if (!table->table_ops->entries_dump)
 +              return -EINVAL;
 +
 +      return devlink_dpipe_entries_fill(info, DEVLINK_CMD_DPIPE_ENTRIES_GET,
 +                                        0, table);
 +}
 +
 +static int devlink_dpipe_fields_put(struct sk_buff *skb,
 +                                  const struct devlink_dpipe_header *header)
 +{
 +      struct devlink_dpipe_field *field;
 +      struct nlattr *field_attr;
 +      int i;
 +
 +      for (i = 0; i < header->fields_count; i++) {
 +              field = &header->fields[i];
 +              field_attr = nla_nest_start_noflag(skb,
 +                                                 DEVLINK_ATTR_DPIPE_FIELD);
 +              if (!field_attr)
 +                      return -EMSGSIZE;
 +              if (nla_put_string(skb, DEVLINK_ATTR_DPIPE_FIELD_NAME, field->name) ||
 +                  nla_put_u32(skb, DEVLINK_ATTR_DPIPE_FIELD_ID, field->id) ||
 +                  nla_put_u32(skb, DEVLINK_ATTR_DPIPE_FIELD_BITWIDTH, field->bitwidth) ||
 +                  nla_put_u32(skb, DEVLINK_ATTR_DPIPE_FIELD_MAPPING_TYPE, field->mapping_type))
 +                      goto nla_put_failure;
 +              nla_nest_end(skb, field_attr);
 +      }
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(skb, field_attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_dpipe_header_put(struct sk_buff *skb,
 +                                  struct devlink_dpipe_header *header)
 +{
 +      struct nlattr *fields_attr, *header_attr;
 +      int err;
 +
 +      header_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_HEADER);
 +      if (!header_attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_string(skb, DEVLINK_ATTR_DPIPE_HEADER_NAME, header->name) ||
 +          nla_put_u32(skb, DEVLINK_ATTR_DPIPE_HEADER_ID, header->id) ||
 +          nla_put_u8(skb, DEVLINK_ATTR_DPIPE_HEADER_GLOBAL, header->global))
 +              goto nla_put_failure;
 +
 +      fields_attr = nla_nest_start_noflag(skb,
 +                                          DEVLINK_ATTR_DPIPE_HEADER_FIELDS);
 +      if (!fields_attr)
 +              goto nla_put_failure;
 +
 +      err = devlink_dpipe_fields_put(skb, header);
 +      if (err) {
 +              nla_nest_cancel(skb, fields_attr);
 +              goto nla_put_failure;
 +      }
 +      nla_nest_end(skb, fields_attr);
 +      nla_nest_end(skb, header_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +      nla_nest_cancel(skb, header_attr);
 +      return err;
 +}
 +
 +static int devlink_dpipe_headers_fill(struct genl_info *info,
 +                                    enum devlink_command cmd, int flags,
 +                                    struct devlink_dpipe_headers *
 +                                    dpipe_headers)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct nlattr *headers_attr;
 +      struct sk_buff *skb = NULL;
 +      struct nlmsghdr *nlh;
 +      void *hdr;
 +      int i, j;
 +      int err;
 +
 +      i = 0;
 +start_again:
 +      err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                        &devlink_nl_family, NLM_F_MULTI, cmd);
 +      if (!hdr) {
 +              nlmsg_free(skb);
 +              return -EMSGSIZE;
 +      }
 +
 +      if (devlink_nl_put_handle(skb, devlink))
 +              goto nla_put_failure;
 +      headers_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_DPIPE_HEADERS);
 +      if (!headers_attr)
 +              goto nla_put_failure;
 +
 +      j = 0;
 +      for (; i < dpipe_headers->headers_count; i++) {
 +              err = devlink_dpipe_header_put(skb, dpipe_headers->headers[i]);
 +              if (err) {
 +                      if (!j)
 +                              goto err_table_put;
 +                      break;
 +              }
 +              j++;
 +      }
 +      nla_nest_end(skb, headers_attr);
 +      genlmsg_end(skb, hdr);
 +      if (i != dpipe_headers->headers_count)
 +              goto start_again;
 +
 +send_done:
 +      nlh = nlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                      NLMSG_DONE, 0, flags | NLM_F_MULTI);
 +      if (!nlh) {
 +              err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +              if (err)
 +                      return err;
 +              goto send_done;
 +      }
 +      return genlmsg_reply(skb, info);
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +err_table_put:
 +      nlmsg_free(skb);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_dpipe_headers_get(struct sk_buff *skb,
 +                                          struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +
 +      if (!devlink->dpipe_headers)
 +              return -EOPNOTSUPP;
 +      return devlink_dpipe_headers_fill(info, DEVLINK_CMD_DPIPE_HEADERS_GET,
 +                                        0, devlink->dpipe_headers);
 +}
 +
 +static int devlink_dpipe_table_counters_set(struct devlink *devlink,
 +                                          const char *table_name,
 +                                          bool enable)
 +{
 +      struct devlink_dpipe_table *table;
 +
 +      table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 +                                       table_name, devlink);
 +      if (!table)
 +              return -EINVAL;
 +
 +      if (table->counter_control_extern)
 +              return -EOPNOTSUPP;
 +
 +      if (!(table->counters_enabled ^ enable))
 +              return 0;
 +
 +      table->counters_enabled = enable;
 +      if (table->table_ops->counters_set_update)
 +              table->table_ops->counters_set_update(table->priv, enable);
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_dpipe_table_counters_set(struct sk_buff *skb,
 +                                                 struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      const char *table_name;
 +      bool counters_enable;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_DPIPE_TABLE_NAME) ||
 +          GENL_REQ_ATTR_CHECK(info,
 +                              DEVLINK_ATTR_DPIPE_TABLE_COUNTERS_ENABLED))
 +              return -EINVAL;
 +
 +      table_name = nla_data(info->attrs[DEVLINK_ATTR_DPIPE_TABLE_NAME]);
 +      counters_enable = !!nla_get_u8(info->attrs[DEVLINK_ATTR_DPIPE_TABLE_COUNTERS_ENABLED]);
 +
 +      return devlink_dpipe_table_counters_set(devlink, table_name,
 +                                              counters_enable);
 +}
 +
 +static struct devlink_resource *
 +devlink_resource_find(struct devlink *devlink,
 +                    struct devlink_resource *resource, u64 resource_id)
 +{
 +      struct list_head *resource_list;
 +
 +      if (resource)
 +              resource_list = &resource->resource_list;
 +      else
 +              resource_list = &devlink->resource_list;
 +
 +      list_for_each_entry(resource, resource_list, list) {
 +              struct devlink_resource *child_resource;
 +
 +              if (resource->id == resource_id)
 +                      return resource;
 +
 +              child_resource = devlink_resource_find(devlink, resource,
 +                                                     resource_id);
 +              if (child_resource)
 +                      return child_resource;
 +      }
 +      return NULL;
 +}
 +
 +static void
 +devlink_resource_validate_children(struct devlink_resource *resource)
 +{
 +      struct devlink_resource *child_resource;
 +      bool size_valid = true;
 +      u64 parts_size = 0;
 +
 +      if (list_empty(&resource->resource_list))
 +              goto out;
 +
 +      list_for_each_entry(child_resource, &resource->resource_list, list)
 +              parts_size += child_resource->size_new;
 +
 +      if (parts_size > resource->size_new)
 +              size_valid = false;
 +out:
 +      resource->size_valid = size_valid;
 +}
 +
 +static int
 +devlink_resource_validate_size(struct devlink_resource *resource, u64 size,
 +                             struct netlink_ext_ack *extack)
 +{
 +      u64 reminder;
 +      int err = 0;
 +
 +      if (size > resource->size_params.size_max) {
 +              NL_SET_ERR_MSG_MOD(extack, "Size larger than maximum");
 +              err = -EINVAL;
 +      }
 +
 +      if (size < resource->size_params.size_min) {
 +              NL_SET_ERR_MSG_MOD(extack, "Size smaller than minimum");
 +              err = -EINVAL;
 +      }
 +
 +      div64_u64_rem(size, resource->size_params.size_granularity, &reminder);
 +      if (reminder) {
 +              NL_SET_ERR_MSG_MOD(extack, "Wrong granularity");
 +              err = -EINVAL;
 +      }
 +
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_resource_set(struct sk_buff *skb,
 +                                     struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_resource *resource;
 +      u64 resource_id;
 +      u64 size;
 +      int err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_RESOURCE_ID) ||
 +          GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_RESOURCE_SIZE))
 +              return -EINVAL;
 +      resource_id = nla_get_u64(info->attrs[DEVLINK_ATTR_RESOURCE_ID]);
 +
 +      resource = devlink_resource_find(devlink, NULL, resource_id);
 +      if (!resource)
 +              return -EINVAL;
 +
 +      size = nla_get_u64(info->attrs[DEVLINK_ATTR_RESOURCE_SIZE]);
 +      err = devlink_resource_validate_size(resource, size, info->extack);
 +      if (err)
 +              return err;
 +
 +      resource->size_new = size;
 +      devlink_resource_validate_children(resource);
 +      if (resource->parent)
 +              devlink_resource_validate_children(resource->parent);
 +      return 0;
 +}
 +
 +static int
 +devlink_resource_size_params_put(struct devlink_resource *resource,
 +                               struct sk_buff *skb)
 +{
 +      struct devlink_resource_size_params *size_params;
 +
 +      size_params = &resource->size_params;
 +      if (nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE_GRAN,
 +                            size_params->size_granularity, DEVLINK_ATTR_PAD) ||
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE_MAX,
 +                            size_params->size_max, DEVLINK_ATTR_PAD) ||
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE_MIN,
 +                            size_params->size_min, DEVLINK_ATTR_PAD) ||
 +          nla_put_u8(skb, DEVLINK_ATTR_RESOURCE_UNIT, size_params->unit))
 +              return -EMSGSIZE;
 +      return 0;
 +}
 +
 +static int devlink_resource_occ_put(struct devlink_resource *resource,
 +                                  struct sk_buff *skb)
 +{
 +      if (!resource->occ_get)
 +              return 0;
 +      return nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_OCC,
 +                               resource->occ_get(resource->occ_get_priv),
 +                               DEVLINK_ATTR_PAD);
 +}
 +
 +static int devlink_resource_put(struct devlink *devlink, struct sk_buff *skb,
 +                              struct devlink_resource *resource)
 +{
 +      struct devlink_resource *child_resource;
 +      struct nlattr *child_resource_attr;
 +      struct nlattr *resource_attr;
 +
 +      resource_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_RESOURCE);
 +      if (!resource_attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_string(skb, DEVLINK_ATTR_RESOURCE_NAME, resource->name) ||
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE, resource->size,
 +                            DEVLINK_ATTR_PAD) ||
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_ID, resource->id,
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +      if (resource->size != resource->size_new &&
 +          nla_put_u64_64bit(skb, DEVLINK_ATTR_RESOURCE_SIZE_NEW,
 +                            resource->size_new, DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +      if (devlink_resource_occ_put(resource, skb))
 +              goto nla_put_failure;
 +      if (devlink_resource_size_params_put(resource, skb))
 +              goto nla_put_failure;
 +      if (list_empty(&resource->resource_list))
 +              goto out;
 +
 +      if (nla_put_u8(skb, DEVLINK_ATTR_RESOURCE_SIZE_VALID,
 +                     resource->size_valid))
 +              goto nla_put_failure;
 +
 +      child_resource_attr = nla_nest_start_noflag(skb,
 +                                                  DEVLINK_ATTR_RESOURCE_LIST);
 +      if (!child_resource_attr)
 +              goto nla_put_failure;
 +
 +      list_for_each_entry(child_resource, &resource->resource_list, list) {
 +              if (devlink_resource_put(devlink, skb, child_resource))
 +                      goto resource_put_failure;
 +      }
 +
 +      nla_nest_end(skb, child_resource_attr);
 +out:
 +      nla_nest_end(skb, resource_attr);
 +      return 0;
 +
 +resource_put_failure:
 +      nla_nest_cancel(skb, child_resource_attr);
 +nla_put_failure:
 +      nla_nest_cancel(skb, resource_attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_resource_fill(struct genl_info *info,
 +                               enum devlink_command cmd, int flags)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_resource *resource;
 +      struct nlattr *resources_attr;
 +      struct sk_buff *skb = NULL;
 +      struct nlmsghdr *nlh;
 +      bool incomplete;
 +      void *hdr;
 +      int i;
 +      int err;
 +
 +      resource = list_first_entry(&devlink->resource_list,
 +                                  struct devlink_resource, list);
 +start_again:
 +      err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +      if (err)
 +              return err;
 +
 +      hdr = genlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                        &devlink_nl_family, NLM_F_MULTI, cmd);
 +      if (!hdr) {
 +              nlmsg_free(skb);
 +              return -EMSGSIZE;
 +      }
 +
 +      if (devlink_nl_put_handle(skb, devlink))
 +              goto nla_put_failure;
 +
 +      resources_attr = nla_nest_start_noflag(skb,
 +                                             DEVLINK_ATTR_RESOURCE_LIST);
 +      if (!resources_attr)
 +              goto nla_put_failure;
 +
 +      incomplete = false;
 +      i = 0;
 +      list_for_each_entry_from(resource, &devlink->resource_list, list) {
 +              err = devlink_resource_put(devlink, skb, resource);
 +              if (err) {
 +                      if (!i)
 +                              goto err_resource_put;
 +                      incomplete = true;
 +                      break;
 +              }
 +              i++;
 +      }
 +      nla_nest_end(skb, resources_attr);
 +      genlmsg_end(skb, hdr);
 +      if (incomplete)
 +              goto start_again;
 +send_done:
 +      nlh = nlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                      NLMSG_DONE, 0, flags | NLM_F_MULTI);
 +      if (!nlh) {
 +              err = devlink_dpipe_send_and_alloc_skb(&skb, info);
 +              if (err)
 +                      return err;
 +              goto send_done;
 +      }
 +      return genlmsg_reply(skb, info);
 +
 +nla_put_failure:
 +      err = -EMSGSIZE;
 +err_resource_put:
 +      nlmsg_free(skb);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_resource_dump(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +
 +      if (list_empty(&devlink->resource_list))
 +              return -EOPNOTSUPP;
 +
 +      return devlink_resource_fill(info, DEVLINK_CMD_RESOURCE_DUMP, 0);
 +}
 +
 +int devlink_resources_validate(struct devlink *devlink,
 +                             struct devlink_resource *resource,
 +                             struct genl_info *info)
 +{
 +      struct list_head *resource_list;
 +      int err = 0;
 +
 +      if (resource)
 +              resource_list = &resource->resource_list;
 +      else
 +              resource_list = &devlink->resource_list;
 +
 +      list_for_each_entry(resource, resource_list, list) {
 +              if (!resource->size_valid)
 +                      return -EINVAL;
 +              err = devlink_resources_validate(devlink, resource, info);
 +              if (err)
 +                      return err;
 +      }
 +      return err;
 +}
 +
 +static const struct devlink_param devlink_param_generic[] = {
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_INT_ERR_RESET,
 +              .name = DEVLINK_PARAM_GENERIC_INT_ERR_RESET_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_INT_ERR_RESET_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_MAX_MACS,
 +              .name = DEVLINK_PARAM_GENERIC_MAX_MACS_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_MAX_MACS_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_SRIOV,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_SRIOV_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_SRIOV_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_REGION_SNAPSHOT,
 +              .name = DEVLINK_PARAM_GENERIC_REGION_SNAPSHOT_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_REGION_SNAPSHOT_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_IGNORE_ARI,
 +              .name = DEVLINK_PARAM_GENERIC_IGNORE_ARI_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_IGNORE_ARI_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MAX,
 +              .name = DEVLINK_PARAM_GENERIC_MSIX_VEC_PER_PF_MAX_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_MSIX_VEC_PER_PF_MAX_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_MSIX_VEC_PER_PF_MIN,
 +              .name = DEVLINK_PARAM_GENERIC_MSIX_VEC_PER_PF_MIN_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_MSIX_VEC_PER_PF_MIN_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_FW_LOAD_POLICY,
 +              .name = DEVLINK_PARAM_GENERIC_FW_LOAD_POLICY_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_FW_LOAD_POLICY_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_RESET_DEV_ON_DRV_PROBE,
 +              .name = DEVLINK_PARAM_GENERIC_RESET_DEV_ON_DRV_PROBE_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_RESET_DEV_ON_DRV_PROBE_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_REMOTE_DEV_RESET,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_ETH,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_ETH_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_ETH_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_RDMA,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_RDMA_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_RDMA_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_VNET,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_VNET_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_VNET_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_IWARP,
 +              .name = DEVLINK_PARAM_GENERIC_ENABLE_IWARP_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_ENABLE_IWARP_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_IO_EQ_SIZE,
 +              .name = DEVLINK_PARAM_GENERIC_IO_EQ_SIZE_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_IO_EQ_SIZE_TYPE,
 +      },
 +      {
 +              .id = DEVLINK_PARAM_GENERIC_ID_EVENT_EQ_SIZE,
 +              .name = DEVLINK_PARAM_GENERIC_EVENT_EQ_SIZE_NAME,
 +              .type = DEVLINK_PARAM_GENERIC_EVENT_EQ_SIZE_TYPE,
 +      },
 +};
 +
 +static int devlink_param_generic_verify(const struct devlink_param *param)
 +{
 +      /* verify it match generic parameter by id and name */
 +      if (param->id > DEVLINK_PARAM_GENERIC_ID_MAX)
 +              return -EINVAL;
 +      if (strcmp(param->name, devlink_param_generic[param->id].name))
 +              return -ENOENT;
 +
 +      WARN_ON(param->type != devlink_param_generic[param->id].type);
 +
 +      return 0;
 +}
 +
 +static int devlink_param_driver_verify(const struct devlink_param *param)
 +{
 +      int i;
 +
 +      if (param->id <= DEVLINK_PARAM_GENERIC_ID_MAX)
 +              return -EINVAL;
 +      /* verify no such name in generic params */
 +      for (i = 0; i <= DEVLINK_PARAM_GENERIC_ID_MAX; i++)
 +              if (!strcmp(param->name, devlink_param_generic[i].name))
 +                      return -EEXIST;
 +
 +      return 0;
 +}
 +
 +static struct devlink_param_item *
 +devlink_param_find_by_name(struct list_head *param_list,
 +                         const char *param_name)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      list_for_each_entry(param_item, param_list, list)
 +              if (!strcmp(param_item->param->name, param_name))
 +                      return param_item;
 +      return NULL;
 +}
 +
 +static struct devlink_param_item *
 +devlink_param_find_by_id(struct list_head *param_list, u32 param_id)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      list_for_each_entry(param_item, param_list, list)
 +              if (param_item->param->id == param_id)
 +                      return param_item;
 +      return NULL;
 +}
 +
 +static bool
 +devlink_param_cmode_is_supported(const struct devlink_param *param,
 +                               enum devlink_param_cmode cmode)
 +{
 +      return test_bit(cmode, &param->supported_cmodes);
 +}
 +
 +static int devlink_param_get(struct devlink *devlink,
 +                           const struct devlink_param *param,
 +                           struct devlink_param_gset_ctx *ctx)
 +{
 +      if (!param->get || devlink->reload_failed)
 +              return -EOPNOTSUPP;
 +      return param->get(devlink, param->id, ctx);
 +}
 +
 +static int devlink_param_set(struct devlink *devlink,
 +                           const struct devlink_param *param,
 +                           struct devlink_param_gset_ctx *ctx)
 +{
 +      if (!param->set || devlink->reload_failed)
 +              return -EOPNOTSUPP;
 +      return param->set(devlink, param->id, ctx);
 +}
 +
 +static int
 +devlink_param_type_to_nla_type(enum devlink_param_type param_type)
 +{
 +      switch (param_type) {
 +      case DEVLINK_PARAM_TYPE_U8:
 +              return NLA_U8;
 +      case DEVLINK_PARAM_TYPE_U16:
 +              return NLA_U16;
 +      case DEVLINK_PARAM_TYPE_U32:
 +              return NLA_U32;
 +      case DEVLINK_PARAM_TYPE_STRING:
 +              return NLA_STRING;
 +      case DEVLINK_PARAM_TYPE_BOOL:
 +              return NLA_FLAG;
 +      default:
 +              return -EINVAL;
 +      }
 +}
 +
 +static int
 +devlink_nl_param_value_fill_one(struct sk_buff *msg,
 +                              enum devlink_param_type type,
 +                              enum devlink_param_cmode cmode,
 +                              union devlink_param_value val)
 +{
 +      struct nlattr *param_value_attr;
 +
 +      param_value_attr = nla_nest_start_noflag(msg,
 +                                               DEVLINK_ATTR_PARAM_VALUE);
 +      if (!param_value_attr)
 +              goto nla_put_failure;
 +
 +      if (nla_put_u8(msg, DEVLINK_ATTR_PARAM_VALUE_CMODE, cmode))
 +              goto value_nest_cancel;
 +
 +      switch (type) {
 +      case DEVLINK_PARAM_TYPE_U8:
 +              if (nla_put_u8(msg, DEVLINK_ATTR_PARAM_VALUE_DATA, val.vu8))
 +                      goto value_nest_cancel;
 +              break;
 +      case DEVLINK_PARAM_TYPE_U16:
 +              if (nla_put_u16(msg, DEVLINK_ATTR_PARAM_VALUE_DATA, val.vu16))
 +                      goto value_nest_cancel;
 +              break;
 +      case DEVLINK_PARAM_TYPE_U32:
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PARAM_VALUE_DATA, val.vu32))
 +                      goto value_nest_cancel;
 +              break;
 +      case DEVLINK_PARAM_TYPE_STRING:
 +              if (nla_put_string(msg, DEVLINK_ATTR_PARAM_VALUE_DATA,
 +                                 val.vstr))
 +                      goto value_nest_cancel;
 +              break;
 +      case DEVLINK_PARAM_TYPE_BOOL:
 +              if (val.vbool &&
 +                  nla_put_flag(msg, DEVLINK_ATTR_PARAM_VALUE_DATA))
 +                      goto value_nest_cancel;
 +              break;
 +      }
 +
 +      nla_nest_end(msg, param_value_attr);
 +      return 0;
 +
 +value_nest_cancel:
 +      nla_nest_cancel(msg, param_value_attr);
 +nla_put_failure:
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_param_fill(struct sk_buff *msg, struct devlink *devlink,
 +                               unsigned int port_index,
 +                               struct devlink_param_item *param_item,
 +                               enum devlink_command cmd,
 +                               u32 portid, u32 seq, int flags)
 +{
 +      union devlink_param_value param_value[DEVLINK_PARAM_CMODE_MAX + 1];
 +      bool param_value_set[DEVLINK_PARAM_CMODE_MAX + 1] = {};
 +      const struct devlink_param *param = param_item->param;
 +      struct devlink_param_gset_ctx ctx;
 +      struct nlattr *param_values_list;
 +      struct nlattr *param_attr;
 +      int nla_type;
 +      void *hdr;
 +      int err;
 +      int i;
 +
 +      /* Get value from driver part to driverinit configuration mode */
 +      for (i = 0; i <= DEVLINK_PARAM_CMODE_MAX; i++) {
 +              if (!devlink_param_cmode_is_supported(param, i))
 +                      continue;
 +              if (i == DEVLINK_PARAM_CMODE_DRIVERINIT) {
 +                      if (!param_item->driverinit_value_valid)
 +                              return -EOPNOTSUPP;
 +                      param_value[i] = param_item->driverinit_value;
 +              } else {
 +                      ctx.cmode = i;
 +                      err = devlink_param_get(devlink, param, &ctx);
 +                      if (err)
 +                              return err;
 +                      param_value[i] = ctx.val;
 +              }
 +              param_value_set[i] = true;
 +      }
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto genlmsg_cancel;
 +
 +      if (cmd == DEVLINK_CMD_PORT_PARAM_GET ||
 +          cmd == DEVLINK_CMD_PORT_PARAM_NEW ||
 +          cmd == DEVLINK_CMD_PORT_PARAM_DEL)
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, port_index))
 +                      goto genlmsg_cancel;
 +
 +      param_attr = nla_nest_start_noflag(msg, DEVLINK_ATTR_PARAM);
 +      if (!param_attr)
 +              goto genlmsg_cancel;
 +      if (nla_put_string(msg, DEVLINK_ATTR_PARAM_NAME, param->name))
 +              goto param_nest_cancel;
 +      if (param->generic && nla_put_flag(msg, DEVLINK_ATTR_PARAM_GENERIC))
 +              goto param_nest_cancel;
 +
 +      nla_type = devlink_param_type_to_nla_type(param->type);
 +      if (nla_type < 0)
 +              goto param_nest_cancel;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_PARAM_TYPE, nla_type))
 +              goto param_nest_cancel;
 +
 +      param_values_list = nla_nest_start_noflag(msg,
 +                                                DEVLINK_ATTR_PARAM_VALUES_LIST);
 +      if (!param_values_list)
 +              goto param_nest_cancel;
 +
 +      for (i = 0; i <= DEVLINK_PARAM_CMODE_MAX; i++) {
 +              if (!param_value_set[i])
 +                      continue;
 +              err = devlink_nl_param_value_fill_one(msg, param->type,
 +                                                    i, param_value[i]);
 +              if (err)
 +                      goto values_list_nest_cancel;
 +      }
 +
 +      nla_nest_end(msg, param_values_list);
 +      nla_nest_end(msg, param_attr);
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +values_list_nest_cancel:
 +      nla_nest_end(msg, param_values_list);
 +param_nest_cancel:
 +      nla_nest_cancel(msg, param_attr);
 +genlmsg_cancel:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static void devlink_param_notify(struct devlink *devlink,
 +                               unsigned int port_index,
 +                               struct devlink_param_item *param_item,
 +                               enum devlink_command cmd)
 +{
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_PARAM_NEW && cmd != DEVLINK_CMD_PARAM_DEL &&
 +              cmd != DEVLINK_CMD_PORT_PARAM_NEW &&
 +              cmd != DEVLINK_CMD_PORT_PARAM_DEL);
 +
 +      /* devlink_notify_register() / devlink_notify_unregister()
 +       * will replay the notifications if the params are added/removed
 +       * outside of the lifetime of the instance.
 +       */
 +      if (!devl_is_registered(devlink))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +      err = devlink_nl_param_fill(msg, devlink, port_index, param_item, cmd,
 +                                  0, 0, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink),
 +                              msg, 0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int
 +devlink_nl_cmd_param_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                                struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_param_item *param_item;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(param_item, &devlink->param_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_param_fill(msg, devlink, 0, param_item,
 +                                          DEVLINK_CMD_PARAM_GET,
 +                                          NETLINK_CB(cb->skb).portid,
 +                                          cb->nlh->nlmsg_seq,
 +                                          NLM_F_MULTI);
 +              if (err == -EOPNOTSUPP) {
 +                      err = 0;
 +              } else if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_param_get = {
 +      .dump_one               = devlink_nl_cmd_param_get_dump_one,
 +};
 +
 +static int
 +devlink_param_type_get_from_info(struct genl_info *info,
 +                               enum devlink_param_type *param_type)
 +{
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_PARAM_TYPE))
 +              return -EINVAL;
 +
 +      switch (nla_get_u8(info->attrs[DEVLINK_ATTR_PARAM_TYPE])) {
 +      case NLA_U8:
 +              *param_type = DEVLINK_PARAM_TYPE_U8;
 +              break;
 +      case NLA_U16:
 +              *param_type = DEVLINK_PARAM_TYPE_U16;
 +              break;
 +      case NLA_U32:
 +              *param_type = DEVLINK_PARAM_TYPE_U32;
 +              break;
 +      case NLA_STRING:
 +              *param_type = DEVLINK_PARAM_TYPE_STRING;
 +              break;
 +      case NLA_FLAG:
 +              *param_type = DEVLINK_PARAM_TYPE_BOOL;
 +              break;
 +      default:
 +              return -EINVAL;
 +      }
 +
 +      return 0;
 +}
 +
 +static int
 +devlink_param_value_get_from_info(const struct devlink_param *param,
 +                                struct genl_info *info,
 +                                union devlink_param_value *value)
 +{
 +      struct nlattr *param_data;
 +      int len;
 +
 +      param_data = info->attrs[DEVLINK_ATTR_PARAM_VALUE_DATA];
 +
 +      if (param->type != DEVLINK_PARAM_TYPE_BOOL && !param_data)
 +              return -EINVAL;
 +
 +      switch (param->type) {
 +      case DEVLINK_PARAM_TYPE_U8:
 +              if (nla_len(param_data) != sizeof(u8))
 +                      return -EINVAL;
 +              value->vu8 = nla_get_u8(param_data);
 +              break;
 +      case DEVLINK_PARAM_TYPE_U16:
 +              if (nla_len(param_data) != sizeof(u16))
 +                      return -EINVAL;
 +              value->vu16 = nla_get_u16(param_data);
 +              break;
 +      case DEVLINK_PARAM_TYPE_U32:
 +              if (nla_len(param_data) != sizeof(u32))
 +                      return -EINVAL;
 +              value->vu32 = nla_get_u32(param_data);
 +              break;
 +      case DEVLINK_PARAM_TYPE_STRING:
 +              len = strnlen(nla_data(param_data), nla_len(param_data));
 +              if (len == nla_len(param_data) ||
 +                  len >= __DEVLINK_PARAM_MAX_STRING_VALUE)
 +                      return -EINVAL;
 +              strcpy(value->vstr, nla_data(param_data));
 +              break;
 +      case DEVLINK_PARAM_TYPE_BOOL:
 +              if (param_data && nla_len(param_data))
 +                      return -EINVAL;
 +              value->vbool = nla_get_flag(param_data);
 +              break;
 +      }
 +      return 0;
 +}
 +
 +static struct devlink_param_item *
 +devlink_param_get_from_info(struct list_head *param_list,
 +                          struct genl_info *info)
 +{
 +      char *param_name;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_PARAM_NAME))
 +              return NULL;
 +
 +      param_name = nla_data(info->attrs[DEVLINK_ATTR_PARAM_NAME]);
 +      return devlink_param_find_by_name(param_list, param_name);
 +}
 +
 +static int devlink_nl_cmd_param_get_doit(struct sk_buff *skb,
 +                                       struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_param_item *param_item;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      param_item = devlink_param_get_from_info(&devlink->param_list, info);
 +      if (!param_item)
 +              return -EINVAL;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_param_fill(msg, devlink, 0, param_item,
 +                                  DEVLINK_CMD_PARAM_GET,
 +                                  info->snd_portid, info->snd_seq, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int __devlink_nl_cmd_param_set_doit(struct devlink *devlink,
 +                                         unsigned int port_index,
 +                                         struct list_head *param_list,
 +                                         struct genl_info *info,
 +                                         enum devlink_command cmd)
 +{
 +      enum devlink_param_type param_type;
 +      struct devlink_param_gset_ctx ctx;
 +      enum devlink_param_cmode cmode;
 +      struct devlink_param_item *param_item;
 +      const struct devlink_param *param;
 +      union devlink_param_value value;
 +      int err = 0;
 +
 +      param_item = devlink_param_get_from_info(param_list, info);
 +      if (!param_item)
 +              return -EINVAL;
 +      param = param_item->param;
 +      err = devlink_param_type_get_from_info(info, &param_type);
 +      if (err)
 +              return err;
 +      if (param_type != param->type)
 +              return -EINVAL;
 +      err = devlink_param_value_get_from_info(param, info, &value);
 +      if (err)
 +              return err;
 +      if (param->validate) {
 +              err = param->validate(devlink, param->id, value, info->extack);
 +              if (err)
 +                      return err;
 +      }
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_PARAM_VALUE_CMODE))
 +              return -EINVAL;
 +      cmode = nla_get_u8(info->attrs[DEVLINK_ATTR_PARAM_VALUE_CMODE]);
 +      if (!devlink_param_cmode_is_supported(param, cmode))
 +              return -EOPNOTSUPP;
 +
 +      if (cmode == DEVLINK_PARAM_CMODE_DRIVERINIT) {
 +              if (param->type == DEVLINK_PARAM_TYPE_STRING)
 +                      strcpy(param_item->driverinit_value.vstr, value.vstr);
 +              else
 +                      param_item->driverinit_value = value;
 +              param_item->driverinit_value_valid = true;
 +      } else {
 +              if (!param->set)
 +                      return -EOPNOTSUPP;
 +              ctx.val = value;
 +              ctx.cmode = cmode;
 +              err = devlink_param_set(devlink, param, &ctx);
 +              if (err)
 +                      return err;
 +      }
 +
 +      devlink_param_notify(devlink, port_index, param_item, cmd);
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_param_set_doit(struct sk_buff *skb,
 +                                       struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +
 +      return __devlink_nl_cmd_param_set_doit(devlink, 0, &devlink->param_list,
 +                                             info, DEVLINK_CMD_PARAM_NEW);
 +}
 +
 +static int devlink_nl_cmd_port_param_get_dumpit(struct sk_buff *msg,
 +                                              struct netlink_callback *cb)
 +{
 +      NL_SET_ERR_MSG_MOD(cb->extack, "Port params are not supported");
 +      return msg->len;
 +}
 +
 +static int devlink_nl_cmd_port_param_get_doit(struct sk_buff *skb,
 +                                            struct genl_info *info)
 +{
 +      NL_SET_ERR_MSG_MOD(info->extack, "Port params are not supported");
 +      return -EINVAL;
 +}
 +
 +static int devlink_nl_cmd_port_param_set_doit(struct sk_buff *skb,
 +                                            struct genl_info *info)
 +{
 +      NL_SET_ERR_MSG_MOD(info->extack, "Port params are not supported");
 +      return -EINVAL;
 +}
 +
 +static int devlink_nl_region_snapshot_id_put(struct sk_buff *msg,
 +                                           struct devlink *devlink,
 +                                           struct devlink_snapshot *snapshot)
 +{
 +      struct nlattr *snap_attr;
 +      int err;
 +
 +      snap_attr = nla_nest_start_noflag(msg, DEVLINK_ATTR_REGION_SNAPSHOT);
 +      if (!snap_attr)
 +              return -EINVAL;
 +
 +      err = nla_put_u32(msg, DEVLINK_ATTR_REGION_SNAPSHOT_ID, snapshot->id);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, snap_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, snap_attr);
 +      return err;
 +}
 +
 +static int devlink_nl_region_snapshots_id_put(struct sk_buff *msg,
 +                                            struct devlink *devlink,
 +                                            struct devlink_region *region)
 +{
 +      struct devlink_snapshot *snapshot;
 +      struct nlattr *snapshots_attr;
 +      int err;
 +
 +      snapshots_attr = nla_nest_start_noflag(msg,
 +                                             DEVLINK_ATTR_REGION_SNAPSHOTS);
 +      if (!snapshots_attr)
 +              return -EINVAL;
 +
 +      list_for_each_entry(snapshot, &region->snapshot_list, list) {
 +              err = devlink_nl_region_snapshot_id_put(msg, devlink, snapshot);
 +              if (err)
 +                      goto nla_put_failure;
 +      }
 +
 +      nla_nest_end(msg, snapshots_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, snapshots_attr);
 +      return err;
 +}
 +
 +static int devlink_nl_region_fill(struct sk_buff *msg, struct devlink *devlink,
 +                                enum devlink_command cmd, u32 portid,
 +                                u32 seq, int flags,
 +                                struct devlink_region *region)
 +{
 +      void *hdr;
 +      int err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      err = devlink_nl_put_handle(msg, devlink);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      if (region->port) {
 +              err = nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX,
 +                                region->port->index);
 +              if (err)
 +                      goto nla_put_failure;
 +      }
 +
 +      err = nla_put_string(msg, DEVLINK_ATTR_REGION_NAME, region->ops->name);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      err = nla_put_u64_64bit(msg, DEVLINK_ATTR_REGION_SIZE,
 +                              region->size,
 +                              DEVLINK_ATTR_PAD);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      err = nla_put_u32(msg, DEVLINK_ATTR_REGION_MAX_SNAPSHOTS,
 +                        region->max_snapshots);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      err = devlink_nl_region_snapshots_id_put(msg, devlink, region);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return err;
 +}
 +
 +static struct sk_buff *
 +devlink_nl_region_notify_build(struct devlink_region *region,
 +                             struct devlink_snapshot *snapshot,
 +                             enum devlink_command cmd, u32 portid, u32 seq)
 +{
 +      struct devlink *devlink = region->devlink;
 +      struct sk_buff *msg;
 +      void *hdr;
 +      int err;
 +
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return ERR_PTR(-ENOMEM);
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, 0, cmd);
 +      if (!hdr) {
 +              err = -EMSGSIZE;
 +              goto out_free_msg;
 +      }
 +
 +      err = devlink_nl_put_handle(msg, devlink);
 +      if (err)
 +              goto out_cancel_msg;
 +
 +      if (region->port) {
 +              err = nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX,
 +                                region->port->index);
 +              if (err)
 +                      goto out_cancel_msg;
 +      }
 +
 +      err = nla_put_string(msg, DEVLINK_ATTR_REGION_NAME,
 +                           region->ops->name);
 +      if (err)
 +              goto out_cancel_msg;
 +
 +      if (snapshot) {
 +              err = nla_put_u32(msg, DEVLINK_ATTR_REGION_SNAPSHOT_ID,
 +                                snapshot->id);
 +              if (err)
 +                      goto out_cancel_msg;
 +      } else {
 +              err = nla_put_u64_64bit(msg, DEVLINK_ATTR_REGION_SIZE,
 +                                      region->size, DEVLINK_ATTR_PAD);
 +              if (err)
 +                      goto out_cancel_msg;
 +      }
 +      genlmsg_end(msg, hdr);
 +
 +      return msg;
 +
 +out_cancel_msg:
 +      genlmsg_cancel(msg, hdr);
 +out_free_msg:
 +      nlmsg_free(msg);
 +      return ERR_PTR(err);
 +}
 +
 +static void devlink_nl_region_notify(struct devlink_region *region,
 +                                   struct devlink_snapshot *snapshot,
 +                                   enum devlink_command cmd)
 +{
 +      struct devlink *devlink = region->devlink;
 +      struct sk_buff *msg;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_REGION_NEW && cmd != DEVLINK_CMD_REGION_DEL);
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = devlink_nl_region_notify_build(region, snapshot, cmd, 0, 0);
 +      if (IS_ERR(msg))
 +              return;
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink), msg,
 +                              0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +/**
 + * __devlink_snapshot_id_increment - Increment number of snapshots using an id
 + *    @devlink: devlink instance
 + *    @id: the snapshot id
 + *
 + *    Track when a new snapshot begins using an id. Load the count for the
 + *    given id from the snapshot xarray, increment it, and store it back.
 + *
 + *    Called when a new snapshot is created with the given id.
 + *
 + *    The id *must* have been previously allocated by
 + *    devlink_region_snapshot_id_get().
 + *
 + *    Returns 0 on success, or an error on failure.
 + */
 +static int __devlink_snapshot_id_increment(struct devlink *devlink, u32 id)
 +{
 +      unsigned long count;
 +      void *p;
 +      int err;
 +
 +      xa_lock(&devlink->snapshot_ids);
 +      p = xa_load(&devlink->snapshot_ids, id);
 +      if (WARN_ON(!p)) {
 +              err = -EINVAL;
 +              goto unlock;
 +      }
 +
 +      if (WARN_ON(!xa_is_value(p))) {
 +              err = -EINVAL;
 +              goto unlock;
 +      }
 +
 +      count = xa_to_value(p);
 +      count++;
 +
 +      err = xa_err(__xa_store(&devlink->snapshot_ids, id, xa_mk_value(count),
 +                              GFP_ATOMIC));
 +unlock:
 +      xa_unlock(&devlink->snapshot_ids);
 +      return err;
 +}
 +
 +/**
 + * __devlink_snapshot_id_decrement - Decrease number of snapshots using an id
 + *    @devlink: devlink instance
 + *    @id: the snapshot id
 + *
 + *    Track when a snapshot is deleted and stops using an id. Load the count
 + *    for the given id from the snapshot xarray, decrement it, and store it
 + *    back.
 + *
 + *    If the count reaches zero, erase this id from the xarray, freeing it
 + *    up for future re-use by devlink_region_snapshot_id_get().
 + *
 + *    Called when a snapshot using the given id is deleted, and when the
 + *    initial allocator of the id is finished using it.
 + */
 +static void __devlink_snapshot_id_decrement(struct devlink *devlink, u32 id)
 +{
 +      unsigned long count;
 +      void *p;
 +
 +      xa_lock(&devlink->snapshot_ids);
 +      p = xa_load(&devlink->snapshot_ids, id);
 +      if (WARN_ON(!p))
 +              goto unlock;
 +
 +      if (WARN_ON(!xa_is_value(p)))
 +              goto unlock;
 +
 +      count = xa_to_value(p);
 +
 +      if (count > 1) {
 +              count--;
 +              __xa_store(&devlink->snapshot_ids, id, xa_mk_value(count),
 +                         GFP_ATOMIC);
 +      } else {
 +              /* If this was the last user, we can erase this id */
 +              __xa_erase(&devlink->snapshot_ids, id);
 +      }
 +unlock:
 +      xa_unlock(&devlink->snapshot_ids);
 +}
 +
 +/**
 + *    __devlink_snapshot_id_insert - Insert a specific snapshot ID
 + *    @devlink: devlink instance
 + *    @id: the snapshot id
 + *
 + *    Mark the given snapshot id as used by inserting a zero value into the
 + *    snapshot xarray.
 + *
 + *    This must be called while holding the devlink instance lock. Unlike
 + *    devlink_snapshot_id_get, the initial reference count is zero, not one.
 + *    It is expected that the id will immediately be used before
 + *    releasing the devlink instance lock.
 + *
 + *    Returns zero on success, or an error code if the snapshot id could not
 + *    be inserted.
 + */
 +static int __devlink_snapshot_id_insert(struct devlink *devlink, u32 id)
 +{
 +      int err;
 +
 +      xa_lock(&devlink->snapshot_ids);
 +      if (xa_load(&devlink->snapshot_ids, id)) {
 +              xa_unlock(&devlink->snapshot_ids);
 +              return -EEXIST;
 +      }
 +      err = xa_err(__xa_store(&devlink->snapshot_ids, id, xa_mk_value(0),
 +                              GFP_ATOMIC));
 +      xa_unlock(&devlink->snapshot_ids);
 +      return err;
 +}
 +
 +/**
 + *    __devlink_region_snapshot_id_get - get snapshot ID
 + *    @devlink: devlink instance
 + *    @id: storage to return snapshot id
 + *
 + *    Allocates a new snapshot id. Returns zero on success, or a negative
 + *    error on failure. Must be called while holding the devlink instance
 + *    lock.
 + *
 + *    Snapshot IDs are tracked using an xarray which stores the number of
 + *    users of the snapshot id.
 + *
 + *    Note that the caller of this function counts as a 'user', in order to
 + *    avoid race conditions. The caller must release its hold on the
 + *    snapshot by using devlink_region_snapshot_id_put.
 + */
 +static int __devlink_region_snapshot_id_get(struct devlink *devlink, u32 *id)
 +{
 +      return xa_alloc(&devlink->snapshot_ids, id, xa_mk_value(1),
 +                      xa_limit_32b, GFP_KERNEL);
 +}
 +
 +/**
 + *    __devlink_region_snapshot_create - create a new snapshot
 + *    This will add a new snapshot of a region. The snapshot
 + *    will be stored on the region struct and can be accessed
 + *    from devlink. This is useful for future analyses of snapshots.
 + *    Multiple snapshots can be created on a region.
 + *    The @snapshot_id should be obtained using the getter function.
 + *
 + *    Must be called only while holding the region snapshot lock.
 + *
 + *    @region: devlink region of the snapshot
 + *    @data: snapshot data
 + *    @snapshot_id: snapshot id to be created
 + */
 +static int
 +__devlink_region_snapshot_create(struct devlink_region *region,
 +                               u8 *data, u32 snapshot_id)
 +{
 +      struct devlink *devlink = region->devlink;
 +      struct devlink_snapshot *snapshot;
 +      int err;
 +
 +      lockdep_assert_held(&region->snapshot_lock);
 +
 +      /* check if region can hold one more snapshot */
 +      if (region->cur_snapshots == region->max_snapshots)
 +              return -ENOSPC;
 +
 +      if (devlink_region_snapshot_get_by_id(region, snapshot_id))
 +              return -EEXIST;
 +
 +      snapshot = kzalloc(sizeof(*snapshot), GFP_KERNEL);
 +      if (!snapshot)
 +              return -ENOMEM;
 +
 +      err = __devlink_snapshot_id_increment(devlink, snapshot_id);
 +      if (err)
 +              goto err_snapshot_id_increment;
 +
 +      snapshot->id = snapshot_id;
 +      snapshot->region = region;
 +      snapshot->data = data;
 +
 +      list_add_tail(&snapshot->list, &region->snapshot_list);
 +
 +      region->cur_snapshots++;
 +
 +      devlink_nl_region_notify(region, snapshot, DEVLINK_CMD_REGION_NEW);
 +      return 0;
 +
 +err_snapshot_id_increment:
 +      kfree(snapshot);
 +      return err;
 +}
 +
 +static void devlink_region_snapshot_del(struct devlink_region *region,
 +                                      struct devlink_snapshot *snapshot)
 +{
 +      struct devlink *devlink = region->devlink;
 +
 +      lockdep_assert_held(&region->snapshot_lock);
 +
 +      devlink_nl_region_notify(region, snapshot, DEVLINK_CMD_REGION_DEL);
 +      region->cur_snapshots--;
 +      list_del(&snapshot->list);
 +      region->ops->destructor(snapshot->data);
 +      __devlink_snapshot_id_decrement(devlink, snapshot->id);
 +      kfree(snapshot);
 +}
 +
 +static int devlink_nl_cmd_region_get_doit(struct sk_buff *skb,
 +                                        struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_port *port = NULL;
 +      struct devlink_region *region;
 +      const char *region_name;
 +      struct sk_buff *msg;
 +      unsigned int index;
 +      int err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_REGION_NAME))
 +              return -EINVAL;
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              index = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +
 +              port = devlink_port_get_by_index(devlink, index);
 +              if (!port)
 +                      return -ENODEV;
 +      }
 +
 +      region_name = nla_data(info->attrs[DEVLINK_ATTR_REGION_NAME]);
 +      if (port)
 +              region = devlink_port_region_get_by_name(port, region_name);
 +      else
 +              region = devlink_region_get_by_name(devlink, region_name);
 +
 +      if (!region)
 +              return -EINVAL;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_region_fill(msg, devlink, DEVLINK_CMD_REGION_GET,
 +                                   info->snd_portid, info->snd_seq, 0,
 +                                   region);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int devlink_nl_cmd_region_get_port_dumpit(struct sk_buff *msg,
 +                                               struct netlink_callback *cb,
 +                                               struct devlink_port *port,
 +                                               int *idx,
 +                                               int start)
 +{
 +      struct devlink_region *region;
 +      int err = 0;
 +
 +      list_for_each_entry(region, &port->region_list, list) {
 +              if (*idx < start) {
 +                      (*idx)++;
 +                      continue;
 +              }
 +              err = devlink_nl_region_fill(msg, port->devlink,
 +                                           DEVLINK_CMD_REGION_GET,
 +                                           NETLINK_CB(cb->skb).portid,
 +                                           cb->nlh->nlmsg_seq,
 +                                           NLM_F_MULTI, region);
 +              if (err)
 +                      goto out;
 +              (*idx)++;
 +      }
 +
 +out:
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_region_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                                 struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_region *region;
 +      struct devlink_port *port;
 +      unsigned long port_index;
 +      int idx = 0;
 +      int err;
 +
 +      list_for_each_entry(region, &devlink->region_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_region_fill(msg, devlink,
 +                                           DEVLINK_CMD_REGION_GET,
 +                                           NETLINK_CB(cb->skb).portid,
 +                                           cb->nlh->nlmsg_seq,
 +                                           NLM_F_MULTI, region);
 +              if (err) {
 +                      state->idx = idx;
 +                      return err;
 +              }
 +              idx++;
 +      }
 +
 +      xa_for_each(&devlink->ports, port_index, port) {
 +              err = devlink_nl_cmd_region_get_port_dumpit(msg, cb, port, &idx,
 +                                                          state->idx);
 +              if (err) {
 +                      state->idx = idx;
 +                      return err;
 +              }
 +      }
 +
 +      return 0;
 +}
 +
 +const struct devlink_cmd devl_cmd_region_get = {
 +      .dump_one               = devlink_nl_cmd_region_get_dump_one,
 +};
 +
 +static int devlink_nl_cmd_region_del(struct sk_buff *skb,
 +                                   struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_snapshot *snapshot;
 +      struct devlink_port *port = NULL;
 +      struct devlink_region *region;
 +      const char *region_name;
 +      unsigned int index;
 +      u32 snapshot_id;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_REGION_NAME) ||
 +          GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_REGION_SNAPSHOT_ID))
 +              return -EINVAL;
 +
 +      region_name = nla_data(info->attrs[DEVLINK_ATTR_REGION_NAME]);
 +      snapshot_id = nla_get_u32(info->attrs[DEVLINK_ATTR_REGION_SNAPSHOT_ID]);
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              index = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +
 +              port = devlink_port_get_by_index(devlink, index);
 +              if (!port)
 +                      return -ENODEV;
 +      }
 +
 +      if (port)
 +              region = devlink_port_region_get_by_name(port, region_name);
 +      else
 +              region = devlink_region_get_by_name(devlink, region_name);
 +
 +      if (!region)
 +              return -EINVAL;
 +
 +      mutex_lock(&region->snapshot_lock);
 +      snapshot = devlink_region_snapshot_get_by_id(region, snapshot_id);
 +      if (!snapshot) {
 +              mutex_unlock(&region->snapshot_lock);
 +              return -EINVAL;
 +      }
 +
 +      devlink_region_snapshot_del(region, snapshot);
 +      mutex_unlock(&region->snapshot_lock);
 +      return 0;
 +}
 +
 +static int
 +devlink_nl_cmd_region_new(struct sk_buff *skb, struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_snapshot *snapshot;
 +      struct devlink_port *port = NULL;
 +      struct nlattr *snapshot_id_attr;
 +      struct devlink_region *region;
 +      const char *region_name;
 +      unsigned int index;
 +      u32 snapshot_id;
 +      u8 *data;
 +      int err;
 +
 +      if (GENL_REQ_ATTR_CHECK(info, DEVLINK_ATTR_REGION_NAME)) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "No region name provided");
 +              return -EINVAL;
 +      }
 +
 +      region_name = nla_data(info->attrs[DEVLINK_ATTR_REGION_NAME]);
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              index = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +
 +              port = devlink_port_get_by_index(devlink, index);
 +              if (!port)
 +                      return -ENODEV;
 +      }
 +
 +      if (port)
 +              region = devlink_port_region_get_by_name(port, region_name);
 +      else
 +              region = devlink_region_get_by_name(devlink, region_name);
 +
 +      if (!region) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "The requested region does not exist");
 +              return -EINVAL;
 +      }
 +
 +      if (!region->ops->snapshot) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "The requested region does not support taking an immediate snapshot");
 +              return -EOPNOTSUPP;
 +      }
 +
 +      mutex_lock(&region->snapshot_lock);
 +
 +      if (region->cur_snapshots == region->max_snapshots) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "The region has reached the maximum number of stored snapshots");
 +              err = -ENOSPC;
 +              goto unlock;
 +      }
 +
 +      snapshot_id_attr = info->attrs[DEVLINK_ATTR_REGION_SNAPSHOT_ID];
 +      if (snapshot_id_attr) {
 +              snapshot_id = nla_get_u32(snapshot_id_attr);
 +
 +              if (devlink_region_snapshot_get_by_id(region, snapshot_id)) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "The requested snapshot id is already in use");
 +                      err = -EEXIST;
 +                      goto unlock;
 +              }
 +
 +              err = __devlink_snapshot_id_insert(devlink, snapshot_id);
 +              if (err)
 +                      goto unlock;
 +      } else {
 +              err = __devlink_region_snapshot_id_get(devlink, &snapshot_id);
 +              if (err) {
 +                      NL_SET_ERR_MSG_MOD(info->extack, "Failed to allocate a new snapshot id");
 +                      goto unlock;
 +              }
 +      }
 +
 +      if (port)
 +              err = region->port_ops->snapshot(port, region->port_ops,
 +                                               info->extack, &data);
 +      else
 +              err = region->ops->snapshot(devlink, region->ops,
 +                                          info->extack, &data);
 +      if (err)
 +              goto err_snapshot_capture;
 +
 +      err = __devlink_region_snapshot_create(region, data, snapshot_id);
 +      if (err)
 +              goto err_snapshot_create;
 +
 +      if (!snapshot_id_attr) {
 +              struct sk_buff *msg;
 +
 +              snapshot = devlink_region_snapshot_get_by_id(region,
 +                                                           snapshot_id);
 +              if (WARN_ON(!snapshot)) {
 +                      err = -EINVAL;
 +                      goto unlock;
 +              }
 +
 +              msg = devlink_nl_region_notify_build(region, snapshot,
 +                                                   DEVLINK_CMD_REGION_NEW,
 +                                                   info->snd_portid,
 +                                                   info->snd_seq);
 +              err = PTR_ERR_OR_ZERO(msg);
 +              if (err)
 +                      goto err_notify;
 +
 +              err = genlmsg_reply(msg, info);
 +              if (err)
 +                      goto err_notify;
 +      }
 +
 +      mutex_unlock(&region->snapshot_lock);
 +      return 0;
 +
 +err_snapshot_create:
 +      region->ops->destructor(data);
 +err_snapshot_capture:
 +      __devlink_snapshot_id_decrement(devlink, snapshot_id);
 +      mutex_unlock(&region->snapshot_lock);
 +      return err;
 +
 +err_notify:
 +      devlink_region_snapshot_del(region, snapshot);
 +unlock:
 +      mutex_unlock(&region->snapshot_lock);
 +      return err;
 +}
 +
 +static int devlink_nl_cmd_region_read_chunk_fill(struct sk_buff *msg,
 +                                               u8 *chunk, u32 chunk_size,
 +                                               u64 addr)
 +{
 +      struct nlattr *chunk_attr;
 +      int err;
 +
 +      chunk_attr = nla_nest_start_noflag(msg, DEVLINK_ATTR_REGION_CHUNK);
 +      if (!chunk_attr)
 +              return -EINVAL;
 +
 +      err = nla_put(msg, DEVLINK_ATTR_REGION_CHUNK_DATA, chunk_size, chunk);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      err = nla_put_u64_64bit(msg, DEVLINK_ATTR_REGION_CHUNK_ADDR, addr,
 +                              DEVLINK_ATTR_PAD);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, chunk_attr);
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, chunk_attr);
 +      return err;
 +}
 +
 +#define DEVLINK_REGION_READ_CHUNK_SIZE 256
 +
 +typedef int devlink_chunk_fill_t(void *cb_priv, u8 *chunk, u32 chunk_size,
 +                               u64 curr_offset,
 +                               struct netlink_ext_ack *extack);
 +
 +static int
 +devlink_nl_region_read_fill(struct sk_buff *skb, devlink_chunk_fill_t *cb,
 +                          void *cb_priv, u64 start_offset, u64 end_offset,
 +                          u64 *new_offset, struct netlink_ext_ack *extack)
 +{
 +      u64 curr_offset = start_offset;
 +      int err = 0;
 +      u8 *data;
 +
 +      /* Allocate and re-use a single buffer */
 +      data = kmalloc(DEVLINK_REGION_READ_CHUNK_SIZE, GFP_KERNEL);
 +      if (!data)
 +              return -ENOMEM;
 +
 +      *new_offset = start_offset;
 +
 +      while (curr_offset < end_offset) {
 +              u32 data_size;
 +
 +              data_size = min_t(u32, end_offset - curr_offset,
 +                                DEVLINK_REGION_READ_CHUNK_SIZE);
 +
 +              err = cb(cb_priv, data, data_size, curr_offset, extack);
 +              if (err)
 +                      break;
 +
 +              err = devlink_nl_cmd_region_read_chunk_fill(skb, data, data_size, curr_offset);
 +              if (err)
 +                      break;
 +
 +              curr_offset += data_size;
 +      }
 +      *new_offset = curr_offset;
 +
 +      kfree(data);
 +
 +      return err;
 +}
 +
 +static int
 +devlink_region_snapshot_fill(void *cb_priv, u8 *chunk, u32 chunk_size,
 +                           u64 curr_offset,
 +                           struct netlink_ext_ack __always_unused *extack)
 +{
 +      struct devlink_snapshot *snapshot = cb_priv;
 +
 +      memcpy(chunk, &snapshot->data[curr_offset], chunk_size);
 +
 +      return 0;
 +}
 +
 +static int
 +devlink_region_port_direct_fill(void *cb_priv, u8 *chunk, u32 chunk_size,
 +                              u64 curr_offset, struct netlink_ext_ack *extack)
 +{
 +      struct devlink_region *region = cb_priv;
 +
 +      return region->port_ops->read(region->port, region->port_ops, extack,
 +                                    curr_offset, chunk_size, chunk);
 +}
 +
 +static int
 +devlink_region_direct_fill(void *cb_priv, u8 *chunk, u32 chunk_size,
 +                         u64 curr_offset, struct netlink_ext_ack *extack)
 +{
 +      struct devlink_region *region = cb_priv;
 +
 +      return region->ops->read(region->devlink, region->ops, extack,
 +                               curr_offset, chunk_size, chunk);
 +}
 +
 +static int devlink_nl_cmd_region_read_dumpit(struct sk_buff *skb,
 +                                           struct netlink_callback *cb)
 +{
 +      const struct genl_dumpit_info *info = genl_dumpit_info(cb);
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct nlattr *chunks_attr, *region_attr, *snapshot_attr;
 +      u64 ret_offset, start_offset, end_offset = U64_MAX;
 +      struct nlattr **attrs = info->attrs;
 +      struct devlink_port *port = NULL;
 +      devlink_chunk_fill_t *region_cb;
 +      struct devlink_region *region;
 +      const char *region_name;
 +      struct devlink *devlink;
 +      unsigned int index;
 +      void *region_cb_priv;
 +      void *hdr;
 +      int err;
 +
 +      start_offset = state->start_offset;
 +
 +      devlink = devlink_get_from_attrs_lock(sock_net(cb->skb->sk), attrs);
 +      if (IS_ERR(devlink))
 +              return PTR_ERR(devlink);
 +
 +      if (!attrs[DEVLINK_ATTR_REGION_NAME]) {
 +              NL_SET_ERR_MSG(cb->extack, "No region name provided");
 +              err = -EINVAL;
 +              goto out_unlock;
 +      }
 +
 +      if (info->attrs[DEVLINK_ATTR_PORT_INDEX]) {
 +              index = nla_get_u32(info->attrs[DEVLINK_ATTR_PORT_INDEX]);
 +
 +              port = devlink_port_get_by_index(devlink, index);
 +              if (!port) {
 +                      err = -ENODEV;
 +                      goto out_unlock;
 +              }
 +      }
 +
 +      region_attr = attrs[DEVLINK_ATTR_REGION_NAME];
 +      region_name = nla_data(region_attr);
 +
 +      if (port)
 +              region = devlink_port_region_get_by_name(port, region_name);
 +      else
 +              region = devlink_region_get_by_name(devlink, region_name);
 +
 +      if (!region) {
 +              NL_SET_ERR_MSG_ATTR(cb->extack, region_attr, "Requested region does not exist");
 +              err = -EINVAL;
 +              goto out_unlock;
 +      }
 +
 +      snapshot_attr = attrs[DEVLINK_ATTR_REGION_SNAPSHOT_ID];
 +      if (!snapshot_attr) {
 +              if (!nla_get_flag(attrs[DEVLINK_ATTR_REGION_DIRECT])) {
 +                      NL_SET_ERR_MSG(cb->extack, "No snapshot id provided");
 +                      err = -EINVAL;
 +                      goto out_unlock;
 +              }
 +
 +              if (!region->ops->read) {
 +                      NL_SET_ERR_MSG(cb->extack, "Requested region does not support direct read");
 +                      err = -EOPNOTSUPP;
 +                      goto out_unlock;
 +              }
 +
 +              if (port)
 +                      region_cb = &devlink_region_port_direct_fill;
 +              else
 +                      region_cb = &devlink_region_direct_fill;
 +              region_cb_priv = region;
 +      } else {
 +              struct devlink_snapshot *snapshot;
 +              u32 snapshot_id;
 +
 +              if (nla_get_flag(attrs[DEVLINK_ATTR_REGION_DIRECT])) {
 +                      NL_SET_ERR_MSG_ATTR(cb->extack, snapshot_attr, "Direct region read does not use snapshot");
 +                      err = -EINVAL;
 +                      goto out_unlock;
 +              }
 +
 +              snapshot_id = nla_get_u32(snapshot_attr);
 +              snapshot = devlink_region_snapshot_get_by_id(region, snapshot_id);
 +              if (!snapshot) {
 +                      NL_SET_ERR_MSG_ATTR(cb->extack, snapshot_attr, "Requested snapshot does not exist");
 +                      err = -EINVAL;
 +                      goto out_unlock;
 +              }
 +              region_cb = &devlink_region_snapshot_fill;
 +              region_cb_priv = snapshot;
 +      }
 +
 +      if (attrs[DEVLINK_ATTR_REGION_CHUNK_ADDR] &&
 +          attrs[DEVLINK_ATTR_REGION_CHUNK_LEN]) {
 +              if (!start_offset)
 +                      start_offset =
 +                              nla_get_u64(attrs[DEVLINK_ATTR_REGION_CHUNK_ADDR]);
 +
 +              end_offset = nla_get_u64(attrs[DEVLINK_ATTR_REGION_CHUNK_ADDR]);
 +              end_offset += nla_get_u64(attrs[DEVLINK_ATTR_REGION_CHUNK_LEN]);
 +      }
 +
 +      if (end_offset > region->size)
 +              end_offset = region->size;
 +
 +      /* return 0 if there is no further data to read */
 +      if (start_offset == end_offset) {
 +              err = 0;
 +              goto out_unlock;
 +      }
 +
 +      hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
 +                        &devlink_nl_family, NLM_F_ACK | NLM_F_MULTI,
 +                        DEVLINK_CMD_REGION_READ);
 +      if (!hdr) {
 +              err = -EMSGSIZE;
 +              goto out_unlock;
 +      }
 +
 +      err = devlink_nl_put_handle(skb, devlink);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      if (region->port) {
 +              err = nla_put_u32(skb, DEVLINK_ATTR_PORT_INDEX,
 +                                region->port->index);
 +              if (err)
 +                      goto nla_put_failure;
 +      }
 +
 +      err = nla_put_string(skb, DEVLINK_ATTR_REGION_NAME, region_name);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      chunks_attr = nla_nest_start_noflag(skb, DEVLINK_ATTR_REGION_CHUNKS);
 +      if (!chunks_attr) {
 +              err = -EMSGSIZE;
 +              goto nla_put_failure;
 +      }
 +
 +      err = devlink_nl_region_read_fill(skb, region_cb, region_cb_priv,
 +                                        start_offset, end_offset, &ret_offset,
 +                                        cb->extack);
 +
 +      if (err && err != -EMSGSIZE)
 +              goto nla_put_failure;
 +
 +      /* Check if there was any progress done to prevent infinite loop */
 +      if (ret_offset == start_offset) {
 +              err = -EINVAL;
 +              goto nla_put_failure;
 +      }
 +
 +      state->start_offset = ret_offset;
 +
 +      nla_nest_end(skb, chunks_attr);
 +      genlmsg_end(skb, hdr);
 +      devl_unlock(devlink);
 +      devlink_put(devlink);
 +      return skb->len;
 +
 +nla_put_failure:
 +      genlmsg_cancel(skb, hdr);
 +out_unlock:
 +      devl_unlock(devlink);
 +      devlink_put(devlink);
 +      return err;
 +}
 +
 +struct devlink_fmsg_item {
 +      struct list_head list;
 +      int attrtype;
 +      u8 nla_type;
 +      u16 len;
 +      int value[];
 +};
 +
 +struct devlink_fmsg {
 +      struct list_head item_list;
 +      bool putting_binary; /* This flag forces enclosing of binary data
 +                            * in an array brackets. It forces using
 +                            * of designated API:
 +                            * devlink_fmsg_binary_pair_nest_start()
 +                            * devlink_fmsg_binary_pair_nest_end()
 +                            */
 +};
 +
 +static struct devlink_fmsg *devlink_fmsg_alloc(void)
 +{
 +      struct devlink_fmsg *fmsg;
 +
 +      fmsg = kzalloc(sizeof(*fmsg), GFP_KERNEL);
 +      if (!fmsg)
 +              return NULL;
 +
 +      INIT_LIST_HEAD(&fmsg->item_list);
 +
 +      return fmsg;
 +}
 +
 +static void devlink_fmsg_free(struct devlink_fmsg *fmsg)
 +{
 +      struct devlink_fmsg_item *item, *tmp;
 +
 +      list_for_each_entry_safe(item, tmp, &fmsg->item_list, list) {
 +              list_del(&item->list);
 +              kfree(item);
 +      }
 +      kfree(fmsg);
 +}
 +
 +static int devlink_fmsg_nest_common(struct devlink_fmsg *fmsg,
 +                                  int attrtype)
 +{
 +      struct devlink_fmsg_item *item;
 +
 +      item = kzalloc(sizeof(*item), GFP_KERNEL);
 +      if (!item)
 +              return -ENOMEM;
 +
 +      item->attrtype = attrtype;
 +      list_add_tail(&item->list, &fmsg->item_list);
 +
 +      return 0;
 +}
 +
 +int devlink_fmsg_obj_nest_start(struct devlink_fmsg *fmsg)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_nest_common(fmsg, DEVLINK_ATTR_FMSG_OBJ_NEST_START);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_obj_nest_start);
 +
 +static int devlink_fmsg_nest_end(struct devlink_fmsg *fmsg)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_nest_common(fmsg, DEVLINK_ATTR_FMSG_NEST_END);
 +}
 +
 +int devlink_fmsg_obj_nest_end(struct devlink_fmsg *fmsg)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_nest_end(fmsg);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_obj_nest_end);
 +
 +#define DEVLINK_FMSG_MAX_SIZE (GENLMSG_DEFAULT_SIZE - GENL_HDRLEN - NLA_HDRLEN)
 +
 +static int devlink_fmsg_put_name(struct devlink_fmsg *fmsg, const char *name)
 +{
 +      struct devlink_fmsg_item *item;
 +
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      if (strlen(name) + 1 > DEVLINK_FMSG_MAX_SIZE)
 +              return -EMSGSIZE;
 +
 +      item = kzalloc(sizeof(*item) + strlen(name) + 1, GFP_KERNEL);
 +      if (!item)
 +              return -ENOMEM;
 +
 +      item->nla_type = NLA_NUL_STRING;
 +      item->len = strlen(name) + 1;
 +      item->attrtype = DEVLINK_ATTR_FMSG_OBJ_NAME;
 +      memcpy(&item->value, name, item->len);
 +      list_add_tail(&item->list, &fmsg->item_list);
 +
 +      return 0;
 +}
 +
 +int devlink_fmsg_pair_nest_start(struct devlink_fmsg *fmsg, const char *name)
 +{
 +      int err;
 +
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      err = devlink_fmsg_nest_common(fmsg, DEVLINK_ATTR_FMSG_PAIR_NEST_START);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_put_name(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_pair_nest_start);
 +
 +int devlink_fmsg_pair_nest_end(struct devlink_fmsg *fmsg)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_nest_end(fmsg);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_pair_nest_end);
 +
 +int devlink_fmsg_arr_pair_nest_start(struct devlink_fmsg *fmsg,
 +                                   const char *name)
 +{
 +      int err;
 +
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_nest_common(fmsg, DEVLINK_ATTR_FMSG_ARR_NEST_START);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_arr_pair_nest_start);
 +
 +int devlink_fmsg_arr_pair_nest_end(struct devlink_fmsg *fmsg)
 +{
 +      int err;
 +
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      err = devlink_fmsg_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_arr_pair_nest_end);
 +
 +int devlink_fmsg_binary_pair_nest_start(struct devlink_fmsg *fmsg,
 +                                      const char *name)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_arr_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      fmsg->putting_binary = true;
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_binary_pair_nest_start);
 +
 +int devlink_fmsg_binary_pair_nest_end(struct devlink_fmsg *fmsg)
 +{
 +      if (!fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      fmsg->putting_binary = false;
 +      return devlink_fmsg_arr_pair_nest_end(fmsg);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_binary_pair_nest_end);
 +
 +static int devlink_fmsg_put_value(struct devlink_fmsg *fmsg,
 +                                const void *value, u16 value_len,
 +                                u8 value_nla_type)
 +{
 +      struct devlink_fmsg_item *item;
 +
 +      if (value_len > DEVLINK_FMSG_MAX_SIZE)
 +              return -EMSGSIZE;
 +
 +      item = kzalloc(sizeof(*item) + value_len, GFP_KERNEL);
 +      if (!item)
 +              return -ENOMEM;
 +
 +      item->nla_type = value_nla_type;
 +      item->len = value_len;
 +      item->attrtype = DEVLINK_ATTR_FMSG_OBJ_VALUE_DATA;
 +      memcpy(&item->value, value, item->len);
 +      list_add_tail(&item->list, &fmsg->item_list);
 +
 +      return 0;
 +}
 +
 +static int devlink_fmsg_bool_put(struct devlink_fmsg *fmsg, bool value)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, &value, sizeof(value), NLA_FLAG);
 +}
 +
 +static int devlink_fmsg_u8_put(struct devlink_fmsg *fmsg, u8 value)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, &value, sizeof(value), NLA_U8);
 +}
 +
 +int devlink_fmsg_u32_put(struct devlink_fmsg *fmsg, u32 value)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, &value, sizeof(value), NLA_U32);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_u32_put);
 +
 +static int devlink_fmsg_u64_put(struct devlink_fmsg *fmsg, u64 value)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, &value, sizeof(value), NLA_U64);
 +}
 +
 +int devlink_fmsg_string_put(struct devlink_fmsg *fmsg, const char *value)
 +{
 +      if (fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, value, strlen(value) + 1,
 +                                    NLA_NUL_STRING);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_string_put);
 +
 +int devlink_fmsg_binary_put(struct devlink_fmsg *fmsg, const void *value,
 +                          u16 value_len)
 +{
 +      if (!fmsg->putting_binary)
 +              return -EINVAL;
 +
 +      return devlink_fmsg_put_value(fmsg, value, value_len, NLA_BINARY);
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_binary_put);
 +
 +int devlink_fmsg_bool_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                             bool value)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_bool_put(fmsg, value);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_pair_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_bool_pair_put);
 +
 +int devlink_fmsg_u8_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                           u8 value)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_u8_put(fmsg, value);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_pair_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_u8_pair_put);
 +
 +int devlink_fmsg_u32_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                            u32 value)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_u32_put(fmsg, value);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_pair_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_u32_pair_put);
 +
 +int devlink_fmsg_u64_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                            u64 value)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_u64_put(fmsg, value);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_pair_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_u64_pair_put);
 +
 +int devlink_fmsg_string_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                               const char *value)
 +{
 +      int err;
 +
 +      err = devlink_fmsg_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_string_put(fmsg, value);
 +      if (err)
 +              return err;
 +
 +      err = devlink_fmsg_pair_nest_end(fmsg);
 +      if (err)
 +              return err;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_string_pair_put);
 +
 +int devlink_fmsg_binary_pair_put(struct devlink_fmsg *fmsg, const char *name,
 +                               const void *value, u32 value_len)
 +{
 +      u32 data_size;
 +      int end_err;
 +      u32 offset;
 +      int err;
 +
 +      err = devlink_fmsg_binary_pair_nest_start(fmsg, name);
 +      if (err)
 +              return err;
 +
 +      for (offset = 0; offset < value_len; offset += data_size) {
 +              data_size = value_len - offset;
 +              if (data_size > DEVLINK_FMSG_MAX_SIZE)
 +                      data_size = DEVLINK_FMSG_MAX_SIZE;
 +              err = devlink_fmsg_binary_put(fmsg, value + offset, data_size);
 +              if (err)
 +                      break;
 +              /* Exit from loop with a break (instead of
 +               * return) to make sure putting_binary is turned off in
 +               * devlink_fmsg_binary_pair_nest_end
 +               */
 +      }
 +
 +      end_err = devlink_fmsg_binary_pair_nest_end(fmsg);
 +      if (end_err)
 +              err = end_err;
 +
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_fmsg_binary_pair_put);
 +
 +static int
 +devlink_fmsg_item_fill_type(struct devlink_fmsg_item *msg, struct sk_buff *skb)
 +{
 +      switch (msg->nla_type) {
 +      case NLA_FLAG:
 +      case NLA_U8:
 +      case NLA_U32:
 +      case NLA_U64:
 +      case NLA_NUL_STRING:
 +      case NLA_BINARY:
 +              return nla_put_u8(skb, DEVLINK_ATTR_FMSG_OBJ_VALUE_TYPE,
 +                                msg->nla_type);
 +      default:
 +              return -EINVAL;
 +      }
 +}
 +
 +static int
 +devlink_fmsg_item_fill_data(struct devlink_fmsg_item *msg, struct sk_buff *skb)
 +{
 +      int attrtype = DEVLINK_ATTR_FMSG_OBJ_VALUE_DATA;
 +      u8 tmp;
 +
 +      switch (msg->nla_type) {
 +      case NLA_FLAG:
 +              /* Always provide flag data, regardless of its value */
 +              tmp = *(bool *) msg->value;
 +
 +              return nla_put_u8(skb, attrtype, tmp);
 +      case NLA_U8:
 +              return nla_put_u8(skb, attrtype, *(u8 *) msg->value);
 +      case NLA_U32:
 +              return nla_put_u32(skb, attrtype, *(u32 *) msg->value);
 +      case NLA_U64:
 +              return nla_put_u64_64bit(skb, attrtype, *(u64 *) msg->value,
 +                                       DEVLINK_ATTR_PAD);
 +      case NLA_NUL_STRING:
 +              return nla_put_string(skb, attrtype, (char *) &msg->value);
 +      case NLA_BINARY:
 +              return nla_put(skb, attrtype, msg->len, (void *) &msg->value);
 +      default:
 +              return -EINVAL;
 +      }
 +}
 +
 +static int
 +devlink_fmsg_prepare_skb(struct devlink_fmsg *fmsg, struct sk_buff *skb,
 +                       int *start)
 +{
 +      struct devlink_fmsg_item *item;
 +      struct nlattr *fmsg_nlattr;
 +      int err = 0;
 +      int i = 0;
 +
 +      fmsg_nlattr = nla_nest_start_noflag(skb, DEVLINK_ATTR_FMSG);
 +      if (!fmsg_nlattr)
 +              return -EMSGSIZE;
 +
 +      list_for_each_entry(item, &fmsg->item_list, list) {
 +              if (i < *start) {
 +                      i++;
 +                      continue;
 +              }
 +
 +              switch (item->attrtype) {
 +              case DEVLINK_ATTR_FMSG_OBJ_NEST_START:
 +              case DEVLINK_ATTR_FMSG_PAIR_NEST_START:
 +              case DEVLINK_ATTR_FMSG_ARR_NEST_START:
 +              case DEVLINK_ATTR_FMSG_NEST_END:
 +                      err = nla_put_flag(skb, item->attrtype);
 +                      break;
 +              case DEVLINK_ATTR_FMSG_OBJ_VALUE_DATA:
 +                      err = devlink_fmsg_item_fill_type(item, skb);
 +                      if (err)
 +                              break;
 +                      err = devlink_fmsg_item_fill_data(item, skb);
 +                      break;
 +              case DEVLINK_ATTR_FMSG_OBJ_NAME:
 +                      err = nla_put_string(skb, item->attrtype,
 +                                           (char *) &item->value);
 +                      break;
 +              default:
 +                      err = -EINVAL;
 +                      break;
 +              }
 +              if (!err)
 +                      *start = ++i;
 +              else
 +                      break;
 +      }
 +
 +      nla_nest_end(skb, fmsg_nlattr);
 +      return err;
 +}
 +
 +static int devlink_fmsg_snd(struct devlink_fmsg *fmsg,
 +                          struct genl_info *info,
 +                          enum devlink_command cmd, int flags)
 +{
 +      struct nlmsghdr *nlh;
 +      struct sk_buff *skb;
 +      bool last = false;
 +      int index = 0;
 +      void *hdr;
 +      int err;
 +
 +      while (!last) {
 +              int tmp_index = index;
 +
 +              skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +              if (!skb)
 +                      return -ENOMEM;
 +
 +              hdr = genlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                                &devlink_nl_family, flags | NLM_F_MULTI, cmd);
 +              if (!hdr) {
 +                      err = -EMSGSIZE;
 +                      goto nla_put_failure;
 +              }
 +
 +              err = devlink_fmsg_prepare_skb(fmsg, skb, &index);
 +              if (!err)
 +                      last = true;
 +              else if (err != -EMSGSIZE || tmp_index == index)
 +                      goto nla_put_failure;
 +
 +              genlmsg_end(skb, hdr);
 +              err = genlmsg_reply(skb, info);
 +              if (err)
 +                      return err;
 +      }
 +
 +      skb = genlmsg_new(GENLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!skb)
 +              return -ENOMEM;
 +      nlh = nlmsg_put(skb, info->snd_portid, info->snd_seq,
 +                      NLMSG_DONE, 0, flags | NLM_F_MULTI);
 +      if (!nlh) {
 +              err = -EMSGSIZE;
 +              goto nla_put_failure;
 +      }
 +
 +      return genlmsg_reply(skb, info);
 +
 +nla_put_failure:
 +      nlmsg_free(skb);
 +      return err;
 +}
 +
 +static int devlink_fmsg_dumpit(struct devlink_fmsg *fmsg, struct sk_buff *skb,
 +                             struct netlink_callback *cb,
 +                             enum devlink_command cmd)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      int index = state->idx;
 +      int tmp_index = index;
 +      void *hdr;
 +      int err;
 +
 +      hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
 +                        &devlink_nl_family, NLM_F_ACK | NLM_F_MULTI, cmd);
 +      if (!hdr) {
 +              err = -EMSGSIZE;
 +              goto nla_put_failure;
 +      }
 +
 +      err = devlink_fmsg_prepare_skb(fmsg, skb, &index);
 +      if ((err && err != -EMSGSIZE) || tmp_index == index)
 +              goto nla_put_failure;
 +
 +      state->idx = index;
 +      genlmsg_end(skb, hdr);
 +      return skb->len;
 +
 +nla_put_failure:
 +      genlmsg_cancel(skb, hdr);
 +      return err;
 +}
 +
 +struct devlink_health_reporter {
 +      struct list_head list;
 +      void *priv;
 +      const struct devlink_health_reporter_ops *ops;
 +      struct devlink *devlink;
 +      struct devlink_port *devlink_port;
 +      struct devlink_fmsg *dump_fmsg;
 +      struct mutex dump_lock; /* lock parallel read/write from dump buffers */
 +      u64 graceful_period;
 +      bool auto_recover;
 +      bool auto_dump;
 +      u8 health_state;
 +      u64 dump_ts;
 +      u64 dump_real_ts;
 +      u64 error_count;
 +      u64 recovery_count;
 +      u64 last_recovery_ts;
 +};
 +
 +void *
 +devlink_health_reporter_priv(struct devlink_health_reporter *reporter)
 +{
 +      return reporter->priv;
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_reporter_priv);
 +
 +static struct devlink_health_reporter *
 +__devlink_health_reporter_find_by_name(struct list_head *reporter_list,
 +                                     const char *reporter_name)
 +{
 +      struct devlink_health_reporter *reporter;
 +
 +      list_for_each_entry(reporter, reporter_list, list)
 +              if (!strcmp(reporter->ops->name, reporter_name))
 +                      return reporter;
 +      return NULL;
 +}
 +
 +static struct devlink_health_reporter *
 +devlink_health_reporter_find_by_name(struct devlink *devlink,
 +                                   const char *reporter_name)
 +{
 +      return __devlink_health_reporter_find_by_name(&devlink->reporter_list,
 +                                                    reporter_name);
 +}
 +
 +static struct devlink_health_reporter *
 +devlink_port_health_reporter_find_by_name(struct devlink_port *devlink_port,
 +                                        const char *reporter_name)
 +{
 +      return __devlink_health_reporter_find_by_name(&devlink_port->reporter_list,
 +                                                    reporter_name);
 +}
 +
 +static struct devlink_health_reporter *
 +__devlink_health_reporter_create(struct devlink *devlink,
 +                               const struct devlink_health_reporter_ops *ops,
 +                               u64 graceful_period, void *priv)
 +{
 +      struct devlink_health_reporter *reporter;
 +
 +      if (WARN_ON(graceful_period && !ops->recover))
 +              return ERR_PTR(-EINVAL);
 +
 +      reporter = kzalloc(sizeof(*reporter), GFP_KERNEL);
 +      if (!reporter)
 +              return ERR_PTR(-ENOMEM);
 +
 +      reporter->priv = priv;
 +      reporter->ops = ops;
 +      reporter->devlink = devlink;
 +      reporter->graceful_period = graceful_period;
 +      reporter->auto_recover = !!ops->recover;
 +      reporter->auto_dump = !!ops->dump;
 +      mutex_init(&reporter->dump_lock);
 +      return reporter;
 +}
 +
 +/**
 + *    devl_port_health_reporter_create - create devlink health reporter for
 + *                                       specified port instance
 + *
 + *    @port: devlink_port which should contain the new reporter
 + *    @ops: ops
 + *    @graceful_period: to avoid recovery loops, in msecs
 + *    @priv: priv
 + */
 +struct devlink_health_reporter *
 +devl_port_health_reporter_create(struct devlink_port *port,
 +                               const struct devlink_health_reporter_ops *ops,
 +                               u64 graceful_period, void *priv)
 +{
 +      struct devlink_health_reporter *reporter;
 +
 +      devl_assert_locked(port->devlink);
 +
 +      if (__devlink_health_reporter_find_by_name(&port->reporter_list,
 +                                                 ops->name))
 +              return ERR_PTR(-EEXIST);
 +
 +      reporter = __devlink_health_reporter_create(port->devlink, ops,
 +                                                  graceful_period, priv);
 +      if (IS_ERR(reporter))
 +              return reporter;
 +
 +      reporter->devlink_port = port;
 +      list_add_tail(&reporter->list, &port->reporter_list);
 +      return reporter;
 +}
 +EXPORT_SYMBOL_GPL(devl_port_health_reporter_create);
 +
 +struct devlink_health_reporter *
 +devlink_port_health_reporter_create(struct devlink_port *port,
 +                                  const struct devlink_health_reporter_ops *ops,
 +                                  u64 graceful_period, void *priv)
 +{
 +      struct devlink_health_reporter *reporter;
 +      struct devlink *devlink = port->devlink;
 +
 +      devl_lock(devlink);
 +      reporter = devl_port_health_reporter_create(port, ops,
 +                                                  graceful_period, priv);
 +      devl_unlock(devlink);
 +      return reporter;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_health_reporter_create);
 +
 +/**
 + *    devl_health_reporter_create - create devlink health reporter
 + *
 + *    @devlink: devlink
 + *    @ops: ops
 + *    @graceful_period: to avoid recovery loops, in msecs
 + *    @priv: priv
 + */
 +struct devlink_health_reporter *
 +devl_health_reporter_create(struct devlink *devlink,
 +                          const struct devlink_health_reporter_ops *ops,
 +                          u64 graceful_period, void *priv)
 +{
 +      struct devlink_health_reporter *reporter;
 +
 +      devl_assert_locked(devlink);
 +
 +      if (devlink_health_reporter_find_by_name(devlink, ops->name))
 +              return ERR_PTR(-EEXIST);
 +
 +      reporter = __devlink_health_reporter_create(devlink, ops,
 +                                                  graceful_period, priv);
 +      if (IS_ERR(reporter))
 +              return reporter;
 +
 +      list_add_tail(&reporter->list, &devlink->reporter_list);
 +      return reporter;
 +}
 +EXPORT_SYMBOL_GPL(devl_health_reporter_create);
 +
 +struct devlink_health_reporter *
 +devlink_health_reporter_create(struct devlink *devlink,
 +                             const struct devlink_health_reporter_ops *ops,
 +                             u64 graceful_period, void *priv)
 +{
 +      struct devlink_health_reporter *reporter;
 +
 +      devl_lock(devlink);
 +      reporter = devl_health_reporter_create(devlink, ops,
 +                                             graceful_period, priv);
 +      devl_unlock(devlink);
 +      return reporter;
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_reporter_create);
 +
 +static void
 +devlink_health_reporter_free(struct devlink_health_reporter *reporter)
 +{
 +      mutex_destroy(&reporter->dump_lock);
 +      if (reporter->dump_fmsg)
 +              devlink_fmsg_free(reporter->dump_fmsg);
 +      kfree(reporter);
 +}
 +
 +/**
 + *    devl_health_reporter_destroy - destroy devlink health reporter
 + *
 + *    @reporter: devlink health reporter to destroy
 + */
 +void
 +devl_health_reporter_destroy(struct devlink_health_reporter *reporter)
 +{
 +      devl_assert_locked(reporter->devlink);
 +
 +      list_del(&reporter->list);
 +      devlink_health_reporter_free(reporter);
 +}
 +EXPORT_SYMBOL_GPL(devl_health_reporter_destroy);
 +
 +void
 +devlink_health_reporter_destroy(struct devlink_health_reporter *reporter)
 +{
 +      struct devlink *devlink = reporter->devlink;
 +
 +      devl_lock(devlink);
 +      devl_health_reporter_destroy(reporter);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_reporter_destroy);
 +
 +static int
 +devlink_nl_health_reporter_fill(struct sk_buff *msg,
 +                              struct devlink_health_reporter *reporter,
 +                              enum devlink_command cmd, u32 portid,
 +                              u32 seq, int flags)
 +{
 +      struct devlink *devlink = reporter->devlink;
 +      struct nlattr *reporter_attr;
 +      void *hdr;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto genlmsg_cancel;
 +
 +      if (reporter->devlink_port) {
 +              if (nla_put_u32(msg, DEVLINK_ATTR_PORT_INDEX, reporter->devlink_port->index))
 +                      goto genlmsg_cancel;
 +      }
 +      reporter_attr = nla_nest_start_noflag(msg,
 +                                            DEVLINK_ATTR_HEALTH_REPORTER);
 +      if (!reporter_attr)
 +              goto genlmsg_cancel;
 +      if (nla_put_string(msg, DEVLINK_ATTR_HEALTH_REPORTER_NAME,
 +                         reporter->ops->name))
 +              goto reporter_nest_cancel;
 +      if (nla_put_u8(msg, DEVLINK_ATTR_HEALTH_REPORTER_STATE,
 +                     reporter->health_state))
 +              goto reporter_nest_cancel;
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_HEALTH_REPORTER_ERR_COUNT,
 +                            reporter->error_count, DEVLINK_ATTR_PAD))
 +              goto reporter_nest_cancel;
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_HEALTH_REPORTER_RECOVER_COUNT,
 +                            reporter->recovery_count, DEVLINK_ATTR_PAD))
 +              goto reporter_nest_cancel;
 +      if (reporter->ops->recover &&
 +          nla_put_u64_64bit(msg, DEVLINK_ATTR_HEALTH_REPORTER_GRACEFUL_PERIOD,
 +                            reporter->graceful_period,
 +                            DEVLINK_ATTR_PAD))
 +              goto reporter_nest_cancel;
 +      if (reporter->ops->recover &&
 +          nla_put_u8(msg, DEVLINK_ATTR_HEALTH_REPORTER_AUTO_RECOVER,
 +                     reporter->auto_recover))
 +              goto reporter_nest_cancel;
 +      if (reporter->dump_fmsg &&
 +          nla_put_u64_64bit(msg, DEVLINK_ATTR_HEALTH_REPORTER_DUMP_TS,
 +                            jiffies_to_msecs(reporter->dump_ts),
 +                            DEVLINK_ATTR_PAD))
 +              goto reporter_nest_cancel;
 +      if (reporter->dump_fmsg &&
 +          nla_put_u64_64bit(msg, DEVLINK_ATTR_HEALTH_REPORTER_DUMP_TS_NS,
 +                            reporter->dump_real_ts, DEVLINK_ATTR_PAD))
 +              goto reporter_nest_cancel;
 +      if (reporter->ops->dump &&
 +          nla_put_u8(msg, DEVLINK_ATTR_HEALTH_REPORTER_AUTO_DUMP,
 +                     reporter->auto_dump))
 +              goto reporter_nest_cancel;
 +
 +      nla_nest_end(msg, reporter_attr);
 +      genlmsg_end(msg, hdr);
 +      return 0;
 +
 +reporter_nest_cancel:
 +      nla_nest_end(msg, reporter_attr);
 +genlmsg_cancel:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static void devlink_recover_notify(struct devlink_health_reporter *reporter,
 +                                 enum devlink_command cmd)
 +{
 +      struct devlink *devlink = reporter->devlink;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON(cmd != DEVLINK_CMD_HEALTH_REPORTER_RECOVER);
 +      WARN_ON(!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED));
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_health_reporter_fill(msg, reporter, cmd, 0, 0, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink), msg,
 +                              0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +void
 +devlink_health_reporter_recovery_done(struct devlink_health_reporter *reporter)
 +{
 +      reporter->recovery_count++;
 +      reporter->last_recovery_ts = jiffies;
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_reporter_recovery_done);
 +
 +static int
 +devlink_health_reporter_recover(struct devlink_health_reporter *reporter,
 +                              void *priv_ctx, struct netlink_ext_ack *extack)
 +{
 +      int err;
 +
 +      if (reporter->health_state == DEVLINK_HEALTH_REPORTER_STATE_HEALTHY)
 +              return 0;
 +
 +      if (!reporter->ops->recover)
 +              return -EOPNOTSUPP;
 +
 +      err = reporter->ops->recover(reporter, priv_ctx, extack);
 +      if (err)
 +              return err;
 +
 +      devlink_health_reporter_recovery_done(reporter);
 +      reporter->health_state = DEVLINK_HEALTH_REPORTER_STATE_HEALTHY;
 +      devlink_recover_notify(reporter, DEVLINK_CMD_HEALTH_REPORTER_RECOVER);
 +
 +      return 0;
 +}
 +
 +static void
 +devlink_health_dump_clear(struct devlink_health_reporter *reporter)
 +{
 +      if (!reporter->dump_fmsg)
 +              return;
 +      devlink_fmsg_free(reporter->dump_fmsg);
 +      reporter->dump_fmsg = NULL;
 +}
 +
 +static int devlink_health_do_dump(struct devlink_health_reporter *reporter,
 +                                void *priv_ctx,
 +                                struct netlink_ext_ack *extack)
 +{
 +      int err;
 +
 +      if (!reporter->ops->dump)
 +              return 0;
 +
 +      if (reporter->dump_fmsg)
 +              return 0;
 +
 +      reporter->dump_fmsg = devlink_fmsg_alloc();
 +      if (!reporter->dump_fmsg) {
 +              err = -ENOMEM;
 +              return err;
 +      }
 +
 +      err = devlink_fmsg_obj_nest_start(reporter->dump_fmsg);
 +      if (err)
 +              goto dump_err;
 +
 +      err = reporter->ops->dump(reporter, reporter->dump_fmsg,
 +                                priv_ctx, extack);
 +      if (err)
 +              goto dump_err;
 +
 +      err = devlink_fmsg_obj_nest_end(reporter->dump_fmsg);
 +      if (err)
 +              goto dump_err;
 +
 +      reporter->dump_ts = jiffies;
 +      reporter->dump_real_ts = ktime_get_real_ns();
 +
 +      return 0;
 +
 +dump_err:
 +      devlink_health_dump_clear(reporter);
 +      return err;
 +}
 +
 +int devlink_health_report(struct devlink_health_reporter *reporter,
 +                        const char *msg, void *priv_ctx)
 +{
 +      enum devlink_health_reporter_state prev_health_state;
 +      struct devlink *devlink = reporter->devlink;
 +      unsigned long recover_ts_threshold;
 +      int ret;
 +
 +      /* write a log message of the current error */
 +      WARN_ON(!msg);
 +      trace_devlink_health_report(devlink, reporter->ops->name, msg);
 +      reporter->error_count++;
 +      prev_health_state = reporter->health_state;
 +      reporter->health_state = DEVLINK_HEALTH_REPORTER_STATE_ERROR;
 +      devlink_recover_notify(reporter, DEVLINK_CMD_HEALTH_REPORTER_RECOVER);
 +
 +      /* abort if the previous error wasn't recovered */
 +      recover_ts_threshold = reporter->last_recovery_ts +
 +                             msecs_to_jiffies(reporter->graceful_period);
 +      if (reporter->auto_recover &&
 +          (prev_health_state != DEVLINK_HEALTH_REPORTER_STATE_HEALTHY ||
 +           (reporter->last_recovery_ts && reporter->recovery_count &&
 +            time_is_after_jiffies(recover_ts_threshold)))) {
 +              trace_devlink_health_recover_aborted(devlink,
 +                                                   reporter->ops->name,
 +                                                   reporter->health_state,
 +                                                   jiffies -
 +                                                   reporter->last_recovery_ts);
 +              return -ECANCELED;
 +      }
 +
 +      if (reporter->auto_dump) {
 +              mutex_lock(&reporter->dump_lock);
 +              /* store current dump of current error, for later analysis */
 +              devlink_health_do_dump(reporter, priv_ctx, NULL);
 +              mutex_unlock(&reporter->dump_lock);
 +      }
 +
 +      if (!reporter->auto_recover)
 +              return 0;
 +
 +      devl_lock(devlink);
 +      ret = devlink_health_reporter_recover(reporter, priv_ctx, NULL);
 +      devl_unlock(devlink);
 +
 +      return ret;
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_report);
 +
 +static struct devlink_health_reporter *
 +devlink_health_reporter_get_from_attrs(struct devlink *devlink,
 +                                     struct nlattr **attrs)
 +{
 +      struct devlink_port *devlink_port;
 +      char *reporter_name;
 +
 +      if (!attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME])
 +              return NULL;
 +
 +      reporter_name = nla_data(attrs[DEVLINK_ATTR_HEALTH_REPORTER_NAME]);
 +      devlink_port = devlink_port_get_from_attrs(devlink, attrs);
 +      if (IS_ERR(devlink_port))
 +              return devlink_health_reporter_find_by_name(devlink,
 +                                                          reporter_name);
 +      else
 +              return devlink_port_health_reporter_find_by_name(devlink_port,
 +                                                               reporter_name);
 +}
 +
 +static struct devlink_health_reporter *
 +devlink_health_reporter_get_from_info(struct devlink *devlink,
 +                                    struct genl_info *info)
 +{
 +      return devlink_health_reporter_get_from_attrs(devlink, info->attrs);
 +}
 +
 +static struct devlink_health_reporter *
 +devlink_health_reporter_get_from_cb(struct netlink_callback *cb)
 +{
 +      const struct genl_dumpit_info *info = genl_dumpit_info(cb);
 +      struct devlink_health_reporter *reporter;
 +      struct nlattr **attrs = info->attrs;
 +      struct devlink *devlink;
 +
 +      devlink = devlink_get_from_attrs_lock(sock_net(cb->skb->sk), attrs);
 +      if (IS_ERR(devlink))
 +              return NULL;
 +      devl_unlock(devlink);
 +
 +      reporter = devlink_health_reporter_get_from_attrs(devlink, attrs);
 +      devlink_put(devlink);
 +      return reporter;
 +}
 +
 +void
 +devlink_health_reporter_state_update(struct devlink_health_reporter *reporter,
 +                                   enum devlink_health_reporter_state state)
 +{
 +      if (WARN_ON(state != DEVLINK_HEALTH_REPORTER_STATE_HEALTHY &&
 +                  state != DEVLINK_HEALTH_REPORTER_STATE_ERROR))
 +              return;
 +
 +      if (reporter->health_state == state)
 +              return;
 +
 +      reporter->health_state = state;
 +      trace_devlink_health_reporter_state_update(reporter->devlink,
 +                                                 reporter->ops->name, state);
 +      devlink_recover_notify(reporter, DEVLINK_CMD_HEALTH_REPORTER_RECOVER);
 +}
 +EXPORT_SYMBOL_GPL(devlink_health_reporter_state_update);
 +
 +static int devlink_nl_cmd_health_reporter_get_doit(struct sk_buff *skb,
 +                                                 struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_health_reporter_fill(msg, reporter,
 +                                            DEVLINK_CMD_HEALTH_REPORTER_GET,
 +                                            info->snd_portid, info->snd_seq,
 +                                            0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return err;
 +      }
 +
 +      return genlmsg_reply(msg, info);
 +}
 +
 +static int
 +devlink_nl_cmd_health_reporter_get_dump_one(struct sk_buff *msg,
 +                                          struct devlink *devlink,
 +                                          struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_health_reporter *reporter;
 +      struct devlink_port *port;
 +      unsigned long port_index;
 +      int idx = 0;
 +      int err;
 +
 +      list_for_each_entry(reporter, &devlink->reporter_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_health_reporter_fill(msg, reporter,
 +                                                    DEVLINK_CMD_HEALTH_REPORTER_GET,
 +                                                    NETLINK_CB(cb->skb).portid,
 +                                                    cb->nlh->nlmsg_seq,
 +                                                    NLM_F_MULTI);
 +              if (err) {
 +                      state->idx = idx;
 +                      return err;
 +              }
 +              idx++;
 +      }
 +      xa_for_each(&devlink->ports, port_index, port) {
 +              list_for_each_entry(reporter, &port->reporter_list, list) {
 +                      if (idx < state->idx) {
 +                              idx++;
 +                              continue;
 +                      }
 +                      err = devlink_nl_health_reporter_fill(msg, reporter,
 +                                                            DEVLINK_CMD_HEALTH_REPORTER_GET,
 +                                                            NETLINK_CB(cb->skb).portid,
 +                                                            cb->nlh->nlmsg_seq,
 +                                                            NLM_F_MULTI);
 +                      if (err) {
 +                              state->idx = idx;
 +                              return err;
 +                      }
 +                      idx++;
 +              }
 +      }
 +
 +      return 0;
 +}
 +
 +const struct devlink_cmd devl_cmd_health_reporter_get = {
 +      .dump_one               = devlink_nl_cmd_health_reporter_get_dump_one,
 +};
 +
 +static int
 +devlink_nl_cmd_health_reporter_set_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      if (!reporter->ops->recover &&
 +          (info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_GRACEFUL_PERIOD] ||
 +           info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_RECOVER]))
 +              return -EOPNOTSUPP;
 +
 +      if (!reporter->ops->dump &&
 +          info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_DUMP])
 +              return -EOPNOTSUPP;
 +
 +      if (info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_GRACEFUL_PERIOD])
 +              reporter->graceful_period =
 +                      nla_get_u64(info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_GRACEFUL_PERIOD]);
 +
 +      if (info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_RECOVER])
 +              reporter->auto_recover =
 +                      nla_get_u8(info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_RECOVER]);
 +
 +      if (info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_DUMP])
 +              reporter->auto_dump =
 +              nla_get_u8(info->attrs[DEVLINK_ATTR_HEALTH_REPORTER_AUTO_DUMP]);
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_health_reporter_recover_doit(struct sk_buff *skb,
 +                                                     struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      return devlink_health_reporter_recover(reporter, NULL, info->extack);
 +}
 +
 +static int devlink_nl_cmd_health_reporter_diagnose_doit(struct sk_buff *skb,
 +                                                      struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +      struct devlink_fmsg *fmsg;
 +      int err;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      if (!reporter->ops->diagnose)
 +              return -EOPNOTSUPP;
 +
 +      fmsg = devlink_fmsg_alloc();
 +      if (!fmsg)
 +              return -ENOMEM;
 +
 +      err = devlink_fmsg_obj_nest_start(fmsg);
 +      if (err)
 +              goto out;
 +
 +      err = reporter->ops->diagnose(reporter, fmsg, info->extack);
 +      if (err)
 +              goto out;
 +
 +      err = devlink_fmsg_obj_nest_end(fmsg);
 +      if (err)
 +              goto out;
 +
 +      err = devlink_fmsg_snd(fmsg, info,
 +                             DEVLINK_CMD_HEALTH_REPORTER_DIAGNOSE, 0);
 +
 +out:
 +      devlink_fmsg_free(fmsg);
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_health_reporter_dump_get_dumpit(struct sk_buff *skb,
 +                                             struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_health_reporter *reporter;
 +      int err;
 +
 +      reporter = devlink_health_reporter_get_from_cb(cb);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      if (!reporter->ops->dump)
 +              return -EOPNOTSUPP;
 +
 +      mutex_lock(&reporter->dump_lock);
 +      if (!state->idx) {
 +              err = devlink_health_do_dump(reporter, NULL, cb->extack);
 +              if (err)
 +                      goto unlock;
 +              state->dump_ts = reporter->dump_ts;
 +      }
 +      if (!reporter->dump_fmsg || state->dump_ts != reporter->dump_ts) {
 +              NL_SET_ERR_MSG_MOD(cb->extack, "Dump trampled, please retry");
 +              err = -EAGAIN;
 +              goto unlock;
 +      }
 +
 +      err = devlink_fmsg_dumpit(reporter->dump_fmsg, skb, cb,
 +                                DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET);
 +unlock:
 +      mutex_unlock(&reporter->dump_lock);
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_health_reporter_dump_clear_doit(struct sk_buff *skb,
 +                                             struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      if (!reporter->ops->dump)
 +              return -EOPNOTSUPP;
 +
 +      mutex_lock(&reporter->dump_lock);
 +      devlink_health_dump_clear(reporter);
 +      mutex_unlock(&reporter->dump_lock);
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_health_reporter_test_doit(struct sk_buff *skb,
 +                                                  struct genl_info *info)
 +{
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_health_reporter *reporter;
 +
 +      reporter = devlink_health_reporter_get_from_info(devlink, info);
 +      if (!reporter)
 +              return -EINVAL;
 +
 +      if (!reporter->ops->test)
 +              return -EOPNOTSUPP;
 +
 +      return reporter->ops->test(reporter, info->extack);
 +}
 +
 +struct devlink_stats {
 +      u64_stats_t rx_bytes;
 +      u64_stats_t rx_packets;
 +      struct u64_stats_sync syncp;
 +};
 +
 +/**
 + * struct devlink_trap_policer_item - Packet trap policer attributes.
 + * @policer: Immutable packet trap policer attributes.
 + * @rate: Rate in packets / sec.
 + * @burst: Burst size in packets.
 + * @list: trap_policer_list member.
 + *
 + * Describes packet trap policer attributes. Created by devlink during trap
 + * policer registration.
 + */
 +struct devlink_trap_policer_item {
 +      const struct devlink_trap_policer *policer;
 +      u64 rate;
 +      u64 burst;
 +      struct list_head list;
 +};
 +
 +/**
 + * struct devlink_trap_group_item - Packet trap group attributes.
 + * @group: Immutable packet trap group attributes.
 + * @policer_item: Associated policer item. Can be NULL.
 + * @list: trap_group_list member.
 + * @stats: Trap group statistics.
 + *
 + * Describes packet trap group attributes. Created by devlink during trap
 + * group registration.
 + */
 +struct devlink_trap_group_item {
 +      const struct devlink_trap_group *group;
 +      struct devlink_trap_policer_item *policer_item;
 +      struct list_head list;
 +      struct devlink_stats __percpu *stats;
 +};
 +
 +/**
 + * struct devlink_trap_item - Packet trap attributes.
 + * @trap: Immutable packet trap attributes.
 + * @group_item: Associated group item.
 + * @list: trap_list member.
 + * @action: Trap action.
 + * @stats: Trap statistics.
 + * @priv: Driver private information.
 + *
 + * Describes both mutable and immutable packet trap attributes. Created by
 + * devlink during trap registration and used for all trap related operations.
 + */
 +struct devlink_trap_item {
 +      const struct devlink_trap *trap;
 +      struct devlink_trap_group_item *group_item;
 +      struct list_head list;
 +      enum devlink_trap_action action;
 +      struct devlink_stats __percpu *stats;
 +      void *priv;
 +};
 +
 +static struct devlink_trap_policer_item *
 +devlink_trap_policer_item_lookup(struct devlink *devlink, u32 id)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +
 +      list_for_each_entry(policer_item, &devlink->trap_policer_list, list) {
 +              if (policer_item->policer->id == id)
 +                      return policer_item;
 +      }
 +
 +      return NULL;
 +}
 +
 +static struct devlink_trap_item *
 +devlink_trap_item_lookup(struct devlink *devlink, const char *name)
 +{
 +      struct devlink_trap_item *trap_item;
 +
 +      list_for_each_entry(trap_item, &devlink->trap_list, list) {
 +              if (!strcmp(trap_item->trap->name, name))
 +                      return trap_item;
 +      }
 +
 +      return NULL;
 +}
 +
 +static struct devlink_trap_item *
 +devlink_trap_item_get_from_info(struct devlink *devlink,
 +                              struct genl_info *info)
 +{
 +      struct nlattr *attr;
 +
 +      if (!info->attrs[DEVLINK_ATTR_TRAP_NAME])
 +              return NULL;
 +      attr = info->attrs[DEVLINK_ATTR_TRAP_NAME];
 +
 +      return devlink_trap_item_lookup(devlink, nla_data(attr));
 +}
 +
 +static int
 +devlink_trap_action_get_from_info(struct genl_info *info,
 +                                enum devlink_trap_action *p_trap_action)
 +{
 +      u8 val;
 +
 +      val = nla_get_u8(info->attrs[DEVLINK_ATTR_TRAP_ACTION]);
 +      switch (val) {
 +      case DEVLINK_TRAP_ACTION_DROP:
 +      case DEVLINK_TRAP_ACTION_TRAP:
 +      case DEVLINK_TRAP_ACTION_MIRROR:
 +              *p_trap_action = val;
 +              break;
 +      default:
 +              return -EINVAL;
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_metadata_put(struct sk_buff *msg,
 +                                   const struct devlink_trap *trap)
 +{
 +      struct nlattr *attr;
 +
 +      attr = nla_nest_start(msg, DEVLINK_ATTR_TRAP_METADATA);
 +      if (!attr)
 +              return -EMSGSIZE;
 +
 +      if ((trap->metadata_cap & DEVLINK_TRAP_METADATA_TYPE_F_IN_PORT) &&
 +          nla_put_flag(msg, DEVLINK_ATTR_TRAP_METADATA_TYPE_IN_PORT))
 +              goto nla_put_failure;
 +      if ((trap->metadata_cap & DEVLINK_TRAP_METADATA_TYPE_F_FA_COOKIE) &&
 +          nla_put_flag(msg, DEVLINK_ATTR_TRAP_METADATA_TYPE_FA_COOKIE))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, attr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, attr);
 +      return -EMSGSIZE;
 +}
 +
 +static void devlink_trap_stats_read(struct devlink_stats __percpu *trap_stats,
 +                                  struct devlink_stats *stats)
 +{
 +      int i;
 +
 +      memset(stats, 0, sizeof(*stats));
 +      for_each_possible_cpu(i) {
 +              struct devlink_stats *cpu_stats;
 +              u64 rx_packets, rx_bytes;
 +              unsigned int start;
 +
 +              cpu_stats = per_cpu_ptr(trap_stats, i);
 +              do {
 +                      start = u64_stats_fetch_begin(&cpu_stats->syncp);
 +                      rx_packets = u64_stats_read(&cpu_stats->rx_packets);
 +                      rx_bytes = u64_stats_read(&cpu_stats->rx_bytes);
 +              } while (u64_stats_fetch_retry(&cpu_stats->syncp, start));
 +
 +              u64_stats_add(&stats->rx_packets, rx_packets);
 +              u64_stats_add(&stats->rx_bytes, rx_bytes);
 +      }
 +}
 +
 +static int
 +devlink_trap_group_stats_put(struct sk_buff *msg,
 +                           struct devlink_stats __percpu *trap_stats)
 +{
 +      struct devlink_stats stats;
 +      struct nlattr *attr;
 +
 +      devlink_trap_stats_read(trap_stats, &stats);
 +
 +      attr = nla_nest_start(msg, DEVLINK_ATTR_STATS);
 +      if (!attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_PACKETS,
 +                            u64_stats_read(&stats.rx_packets),
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_BYTES,
 +                            u64_stats_read(&stats.rx_bytes),
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, attr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_trap_stats_put(struct sk_buff *msg, struct devlink *devlink,
 +                                const struct devlink_trap_item *trap_item)
 +{
 +      struct devlink_stats stats;
 +      struct nlattr *attr;
 +      u64 drops = 0;
 +      int err;
 +
 +      if (devlink->ops->trap_drop_counter_get) {
 +              err = devlink->ops->trap_drop_counter_get(devlink,
 +                                                        trap_item->trap,
 +                                                        &drops);
 +              if (err)
 +                      return err;
 +      }
 +
 +      devlink_trap_stats_read(trap_item->stats, &stats);
 +
 +      attr = nla_nest_start(msg, DEVLINK_ATTR_STATS);
 +      if (!attr)
 +              return -EMSGSIZE;
 +
 +      if (devlink->ops->trap_drop_counter_get &&
 +          nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_DROPPED, drops,
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_PACKETS,
 +                            u64_stats_read(&stats.rx_packets),
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_BYTES,
 +                            u64_stats_read(&stats.rx_bytes),
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, attr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_trap_fill(struct sk_buff *msg, struct devlink *devlink,
 +                              const struct devlink_trap_item *trap_item,
 +                              enum devlink_command cmd, u32 portid, u32 seq,
 +                              int flags)
 +{
 +      struct devlink_trap_group_item *group_item = trap_item->group_item;
 +      void *hdr;
 +      int err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +
 +      if (nla_put_string(msg, DEVLINK_ATTR_TRAP_GROUP_NAME,
 +                         group_item->group->name))
 +              goto nla_put_failure;
 +
 +      if (nla_put_string(msg, DEVLINK_ATTR_TRAP_NAME, trap_item->trap->name))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u8(msg, DEVLINK_ATTR_TRAP_TYPE, trap_item->trap->type))
 +              goto nla_put_failure;
 +
 +      if (trap_item->trap->generic &&
 +          nla_put_flag(msg, DEVLINK_ATTR_TRAP_GENERIC))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u8(msg, DEVLINK_ATTR_TRAP_ACTION, trap_item->action))
 +              goto nla_put_failure;
 +
 +      err = devlink_trap_metadata_put(msg, trap_item->trap);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      err = devlink_trap_stats_put(msg, devlink, trap_item);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_trap_get_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_trap_item *trap_item;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      if (list_empty(&devlink->trap_list))
 +              return -EOPNOTSUPP;
 +
 +      trap_item = devlink_trap_item_get_from_info(devlink, info);
 +      if (!trap_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap");
 +              return -ENOENT;
 +      }
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_trap_fill(msg, devlink, trap_item,
 +                                 DEVLINK_CMD_TRAP_NEW, info->snd_portid,
 +                                 info->snd_seq, 0);
 +      if (err)
 +              goto err_trap_fill;
 +
 +      return genlmsg_reply(msg, info);
 +
 +err_trap_fill:
 +      nlmsg_free(msg);
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_trap_get_dump_one(struct sk_buff *msg, struct devlink *devlink,
 +                               struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_trap_item *trap_item;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(trap_item, &devlink->trap_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_trap_fill(msg, devlink, trap_item,
 +                                         DEVLINK_CMD_TRAP_NEW,
 +                                         NETLINK_CB(cb->skb).portid,
 +                                         cb->nlh->nlmsg_seq,
 +                                         NLM_F_MULTI);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_trap_get = {
 +      .dump_one               = devlink_nl_cmd_trap_get_dump_one,
 +};
 +
 +static int __devlink_trap_action_set(struct devlink *devlink,
 +                                   struct devlink_trap_item *trap_item,
 +                                   enum devlink_trap_action trap_action,
 +                                   struct netlink_ext_ack *extack)
 +{
 +      int err;
 +
 +      if (trap_item->action != trap_action &&
 +          trap_item->trap->type != DEVLINK_TRAP_TYPE_DROP) {
 +              NL_SET_ERR_MSG_MOD(extack, "Cannot change action of non-drop traps. Skipping");
 +              return 0;
 +      }
 +
 +      err = devlink->ops->trap_action_set(devlink, trap_item->trap,
 +                                          trap_action, extack);
 +      if (err)
 +              return err;
 +
 +      trap_item->action = trap_action;
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_action_set(struct devlink *devlink,
 +                                 struct devlink_trap_item *trap_item,
 +                                 struct genl_info *info)
 +{
 +      enum devlink_trap_action trap_action;
 +      int err;
 +
 +      if (!info->attrs[DEVLINK_ATTR_TRAP_ACTION])
 +              return 0;
 +
 +      err = devlink_trap_action_get_from_info(info, &trap_action);
 +      if (err) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "Invalid trap action");
 +              return -EINVAL;
 +      }
 +
 +      return __devlink_trap_action_set(devlink, trap_item, trap_action,
 +                                       info->extack);
 +}
 +
 +static int devlink_nl_cmd_trap_set_doit(struct sk_buff *skb,
 +                                      struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_trap_item *trap_item;
 +
 +      if (list_empty(&devlink->trap_list))
 +              return -EOPNOTSUPP;
 +
 +      trap_item = devlink_trap_item_get_from_info(devlink, info);
 +      if (!trap_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap");
 +              return -ENOENT;
 +      }
 +
 +      return devlink_trap_action_set(devlink, trap_item, info);
 +}
 +
 +static struct devlink_trap_group_item *
 +devlink_trap_group_item_lookup(struct devlink *devlink, const char *name)
 +{
 +      struct devlink_trap_group_item *group_item;
 +
 +      list_for_each_entry(group_item, &devlink->trap_group_list, list) {
 +              if (!strcmp(group_item->group->name, name))
 +                      return group_item;
 +      }
 +
 +      return NULL;
 +}
 +
 +static struct devlink_trap_group_item *
 +devlink_trap_group_item_lookup_by_id(struct devlink *devlink, u16 id)
 +{
 +      struct devlink_trap_group_item *group_item;
 +
 +      list_for_each_entry(group_item, &devlink->trap_group_list, list) {
 +              if (group_item->group->id == id)
 +                      return group_item;
 +      }
 +
 +      return NULL;
 +}
 +
 +static struct devlink_trap_group_item *
 +devlink_trap_group_item_get_from_info(struct devlink *devlink,
 +                                    struct genl_info *info)
 +{
 +      char *name;
 +
 +      if (!info->attrs[DEVLINK_ATTR_TRAP_GROUP_NAME])
 +              return NULL;
 +      name = nla_data(info->attrs[DEVLINK_ATTR_TRAP_GROUP_NAME]);
 +
 +      return devlink_trap_group_item_lookup(devlink, name);
 +}
 +
 +static int
 +devlink_nl_trap_group_fill(struct sk_buff *msg, struct devlink *devlink,
 +                         const struct devlink_trap_group_item *group_item,
 +                         enum devlink_command cmd, u32 portid, u32 seq,
 +                         int flags)
 +{
 +      void *hdr;
 +      int err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +
 +      if (nla_put_string(msg, DEVLINK_ATTR_TRAP_GROUP_NAME,
 +                         group_item->group->name))
 +              goto nla_put_failure;
 +
 +      if (group_item->group->generic &&
 +          nla_put_flag(msg, DEVLINK_ATTR_TRAP_GENERIC))
 +              goto nla_put_failure;
 +
 +      if (group_item->policer_item &&
 +          nla_put_u32(msg, DEVLINK_ATTR_TRAP_POLICER_ID,
 +                      group_item->policer_item->policer->id))
 +              goto nla_put_failure;
 +
 +      err = devlink_trap_group_stats_put(msg, group_item->stats);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_trap_group_get_doit(struct sk_buff *skb,
 +                                            struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_trap_group_item *group_item;
 +      struct sk_buff *msg;
 +      int err;
 +
 +      if (list_empty(&devlink->trap_group_list))
 +              return -EOPNOTSUPP;
 +
 +      group_item = devlink_trap_group_item_get_from_info(devlink, info);
 +      if (!group_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap group");
 +              return -ENOENT;
 +      }
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_trap_group_fill(msg, devlink, group_item,
 +                                       DEVLINK_CMD_TRAP_GROUP_NEW,
 +                                       info->snd_portid, info->snd_seq, 0);
 +      if (err)
 +              goto err_trap_group_fill;
 +
 +      return genlmsg_reply(msg, info);
 +
 +err_trap_group_fill:
 +      nlmsg_free(msg);
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_trap_group_get_dump_one(struct sk_buff *msg,
 +                                     struct devlink *devlink,
 +                                     struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_trap_group_item *group_item;
 +      int idx = 0;
 +      int err = 0;
 +
 +
 +      list_for_each_entry(group_item, &devlink->trap_group_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_trap_group_fill(msg, devlink, group_item,
 +                                               DEVLINK_CMD_TRAP_GROUP_NEW,
 +                                               NETLINK_CB(cb->skb).portid,
 +                                               cb->nlh->nlmsg_seq,
 +                                               NLM_F_MULTI);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_trap_group_get = {
 +      .dump_one               = devlink_nl_cmd_trap_group_get_dump_one,
 +};
 +
 +static int
 +__devlink_trap_group_action_set(struct devlink *devlink,
 +                              struct devlink_trap_group_item *group_item,
 +                              enum devlink_trap_action trap_action,
 +                              struct netlink_ext_ack *extack)
 +{
 +      const char *group_name = group_item->group->name;
 +      struct devlink_trap_item *trap_item;
 +      int err;
 +
 +      if (devlink->ops->trap_group_action_set) {
 +              err = devlink->ops->trap_group_action_set(devlink, group_item->group,
 +                                                        trap_action, extack);
 +              if (err)
 +                      return err;
 +
 +              list_for_each_entry(trap_item, &devlink->trap_list, list) {
 +                      if (strcmp(trap_item->group_item->group->name, group_name))
 +                              continue;
 +                      if (trap_item->action != trap_action &&
 +                          trap_item->trap->type != DEVLINK_TRAP_TYPE_DROP)
 +                              continue;
 +                      trap_item->action = trap_action;
 +              }
 +
 +              return 0;
 +      }
 +
 +      list_for_each_entry(trap_item, &devlink->trap_list, list) {
 +              if (strcmp(trap_item->group_item->group->name, group_name))
 +                      continue;
 +              err = __devlink_trap_action_set(devlink, trap_item,
 +                                              trap_action, extack);
 +              if (err)
 +                      return err;
 +      }
 +
 +      return 0;
 +}
 +
 +static int
 +devlink_trap_group_action_set(struct devlink *devlink,
 +                            struct devlink_trap_group_item *group_item,
 +                            struct genl_info *info, bool *p_modified)
 +{
 +      enum devlink_trap_action trap_action;
 +      int err;
 +
 +      if (!info->attrs[DEVLINK_ATTR_TRAP_ACTION])
 +              return 0;
 +
 +      err = devlink_trap_action_get_from_info(info, &trap_action);
 +      if (err) {
 +              NL_SET_ERR_MSG_MOD(info->extack, "Invalid trap action");
 +              return -EINVAL;
 +      }
 +
 +      err = __devlink_trap_group_action_set(devlink, group_item, trap_action,
 +                                            info->extack);
 +      if (err)
 +              return err;
 +
 +      *p_modified = true;
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_group_set(struct devlink *devlink,
 +                                struct devlink_trap_group_item *group_item,
 +                                struct genl_info *info)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      struct netlink_ext_ack *extack = info->extack;
 +      const struct devlink_trap_policer *policer;
 +      struct nlattr **attrs = info->attrs;
 +      u32 policer_id;
 +      int err;
 +
 +      if (!attrs[DEVLINK_ATTR_TRAP_POLICER_ID])
 +              return 0;
 +
 +      if (!devlink->ops->trap_group_set)
 +              return -EOPNOTSUPP;
 +
 +      policer_id = nla_get_u32(attrs[DEVLINK_ATTR_TRAP_POLICER_ID]);
 +      policer_item = devlink_trap_policer_item_lookup(devlink, policer_id);
 +      if (policer_id && !policer_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap policer");
 +              return -ENOENT;
 +      }
 +      policer = policer_item ? policer_item->policer : NULL;
 +
 +      err = devlink->ops->trap_group_set(devlink, group_item->group, policer,
 +                                         extack);
 +      if (err)
 +              return err;
 +
 +      group_item->policer_item = policer_item;
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_trap_group_set_doit(struct sk_buff *skb,
 +                                            struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct devlink_trap_group_item *group_item;
 +      bool modified = false;
 +      int err;
 +
 +      if (list_empty(&devlink->trap_group_list))
 +              return -EOPNOTSUPP;
 +
 +      group_item = devlink_trap_group_item_get_from_info(devlink, info);
 +      if (!group_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap group");
 +              return -ENOENT;
 +      }
 +
 +      err = devlink_trap_group_action_set(devlink, group_item, info,
 +                                          &modified);
 +      if (err)
 +              return err;
 +
 +      err = devlink_trap_group_set(devlink, group_item, info);
 +      if (err)
 +              goto err_trap_group_set;
 +
 +      return 0;
 +
 +err_trap_group_set:
 +      if (modified)
 +              NL_SET_ERR_MSG_MOD(extack, "Trap group set failed, but some changes were committed already");
 +      return err;
 +}
 +
 +static struct devlink_trap_policer_item *
 +devlink_trap_policer_item_get_from_info(struct devlink *devlink,
 +                                      struct genl_info *info)
 +{
 +      u32 id;
 +
 +      if (!info->attrs[DEVLINK_ATTR_TRAP_POLICER_ID])
 +              return NULL;
 +      id = nla_get_u32(info->attrs[DEVLINK_ATTR_TRAP_POLICER_ID]);
 +
 +      return devlink_trap_policer_item_lookup(devlink, id);
 +}
 +
 +static int
 +devlink_trap_policer_stats_put(struct sk_buff *msg, struct devlink *devlink,
 +                             const struct devlink_trap_policer *policer)
 +{
 +      struct nlattr *attr;
 +      u64 drops;
 +      int err;
 +
 +      if (!devlink->ops->trap_policer_counter_get)
 +              return 0;
 +
 +      err = devlink->ops->trap_policer_counter_get(devlink, policer, &drops);
 +      if (err)
 +              return err;
 +
 +      attr = nla_nest_start(msg, DEVLINK_ATTR_STATS);
 +      if (!attr)
 +              return -EMSGSIZE;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_STATS_RX_DROPPED, drops,
 +                            DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      nla_nest_end(msg, attr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      nla_nest_cancel(msg, attr);
 +      return -EMSGSIZE;
 +}
 +
 +static int
 +devlink_nl_trap_policer_fill(struct sk_buff *msg, struct devlink *devlink,
 +                           const struct devlink_trap_policer_item *policer_item,
 +                           enum devlink_command cmd, u32 portid, u32 seq,
 +                           int flags)
 +{
 +      void *hdr;
 +      int err;
 +
 +      hdr = genlmsg_put(msg, portid, seq, &devlink_nl_family, flags, cmd);
 +      if (!hdr)
 +              return -EMSGSIZE;
 +
 +      if (devlink_nl_put_handle(msg, devlink))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u32(msg, DEVLINK_ATTR_TRAP_POLICER_ID,
 +                      policer_item->policer->id))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_TRAP_POLICER_RATE,
 +                            policer_item->rate, DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      if (nla_put_u64_64bit(msg, DEVLINK_ATTR_TRAP_POLICER_BURST,
 +                            policer_item->burst, DEVLINK_ATTR_PAD))
 +              goto nla_put_failure;
 +
 +      err = devlink_trap_policer_stats_put(msg, devlink,
 +                                           policer_item->policer);
 +      if (err)
 +              goto nla_put_failure;
 +
 +      genlmsg_end(msg, hdr);
 +
 +      return 0;
 +
 +nla_put_failure:
 +      genlmsg_cancel(msg, hdr);
 +      return -EMSGSIZE;
 +}
 +
 +static int devlink_nl_cmd_trap_policer_get_doit(struct sk_buff *skb,
 +                                              struct genl_info *info)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +      struct sk_buff *msg;
 +      int err;
 +
 +      if (list_empty(&devlink->trap_policer_list))
 +              return -EOPNOTSUPP;
 +
 +      policer_item = devlink_trap_policer_item_get_from_info(devlink, info);
 +      if (!policer_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap policer");
 +              return -ENOENT;
 +      }
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return -ENOMEM;
 +
 +      err = devlink_nl_trap_policer_fill(msg, devlink, policer_item,
 +                                         DEVLINK_CMD_TRAP_POLICER_NEW,
 +                                         info->snd_portid, info->snd_seq, 0);
 +      if (err)
 +              goto err_trap_policer_fill;
 +
 +      return genlmsg_reply(msg, info);
 +
 +err_trap_policer_fill:
 +      nlmsg_free(msg);
 +      return err;
 +}
 +
 +static int
 +devlink_nl_cmd_trap_policer_get_dump_one(struct sk_buff *msg,
 +                                       struct devlink *devlink,
 +                                       struct netlink_callback *cb)
 +{
 +      struct devlink_nl_dump_state *state = devlink_dump_state(cb);
 +      struct devlink_trap_policer_item *policer_item;
 +      int idx = 0;
 +      int err = 0;
 +
 +      list_for_each_entry(policer_item, &devlink->trap_policer_list, list) {
 +              if (idx < state->idx) {
 +                      idx++;
 +                      continue;
 +              }
 +              err = devlink_nl_trap_policer_fill(msg, devlink, policer_item,
 +                                                 DEVLINK_CMD_TRAP_POLICER_NEW,
 +                                                 NETLINK_CB(cb->skb).portid,
 +                                                 cb->nlh->nlmsg_seq,
 +                                                 NLM_F_MULTI);
 +              if (err) {
 +                      state->idx = idx;
 +                      break;
 +              }
 +              idx++;
 +      }
 +
 +      return err;
 +}
 +
 +const struct devlink_cmd devl_cmd_trap_policer_get = {
 +      .dump_one               = devlink_nl_cmd_trap_policer_get_dump_one,
 +};
 +
 +static int
 +devlink_trap_policer_set(struct devlink *devlink,
 +                       struct devlink_trap_policer_item *policer_item,
 +                       struct genl_info *info)
 +{
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct nlattr **attrs = info->attrs;
 +      u64 rate, burst;
 +      int err;
 +
 +      rate = policer_item->rate;
 +      burst = policer_item->burst;
 +
 +      if (attrs[DEVLINK_ATTR_TRAP_POLICER_RATE])
 +              rate = nla_get_u64(attrs[DEVLINK_ATTR_TRAP_POLICER_RATE]);
 +
 +      if (attrs[DEVLINK_ATTR_TRAP_POLICER_BURST])
 +              burst = nla_get_u64(attrs[DEVLINK_ATTR_TRAP_POLICER_BURST]);
 +
 +      if (rate < policer_item->policer->min_rate) {
 +              NL_SET_ERR_MSG_MOD(extack, "Policer rate lower than limit");
 +              return -EINVAL;
 +      }
 +
 +      if (rate > policer_item->policer->max_rate) {
 +              NL_SET_ERR_MSG_MOD(extack, "Policer rate higher than limit");
 +              return -EINVAL;
 +      }
 +
 +      if (burst < policer_item->policer->min_burst) {
 +              NL_SET_ERR_MSG_MOD(extack, "Policer burst size lower than limit");
 +              return -EINVAL;
 +      }
 +
 +      if (burst > policer_item->policer->max_burst) {
 +              NL_SET_ERR_MSG_MOD(extack, "Policer burst size higher than limit");
 +              return -EINVAL;
 +      }
 +
 +      err = devlink->ops->trap_policer_set(devlink, policer_item->policer,
 +                                           rate, burst, info->extack);
 +      if (err)
 +              return err;
 +
 +      policer_item->rate = rate;
 +      policer_item->burst = burst;
 +
 +      return 0;
 +}
 +
 +static int devlink_nl_cmd_trap_policer_set_doit(struct sk_buff *skb,
 +                                              struct genl_info *info)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      struct netlink_ext_ack *extack = info->extack;
 +      struct devlink *devlink = info->user_ptr[0];
 +
 +      if (list_empty(&devlink->trap_policer_list))
 +              return -EOPNOTSUPP;
 +
 +      if (!devlink->ops->trap_policer_set)
 +              return -EOPNOTSUPP;
 +
 +      policer_item = devlink_trap_policer_item_get_from_info(devlink, info);
 +      if (!policer_item) {
 +              NL_SET_ERR_MSG_MOD(extack, "Device did not register this trap policer");
 +              return -ENOENT;
 +      }
 +
 +      return devlink_trap_policer_set(devlink, policer_item, info);
 +}
 +
 +const struct genl_small_ops devlink_nl_ops[56] = {
 +      {
 +              .cmd = DEVLINK_CMD_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RATE_GET,
 +              .doit = devlink_nl_cmd_rate_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_RATE,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RATE_SET,
 +              .doit = devlink_nl_cmd_rate_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_RATE,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RATE_NEW,
 +              .doit = devlink_nl_cmd_rate_new_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RATE_DEL,
 +              .doit = devlink_nl_cmd_rate_del_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_RATE_NODE,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_SPLIT,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_split_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_UNSPLIT,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_unsplit_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_NEW,
 +              .doit = devlink_nl_cmd_port_new_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_DEL,
 +              .doit = devlink_nl_cmd_port_del_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_LINECARD_GET,
 +              .doit = devlink_nl_cmd_linecard_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_LINECARD,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_LINECARD_SET,
 +              .doit = devlink_nl_cmd_linecard_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_LINECARD,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_POOL_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_pool_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_POOL_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_pool_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_PORT_POOL_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_port_pool_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_PORT_POOL_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_port_pool_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_TC_POOL_BIND_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_tc_pool_bind_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_TC_POOL_BIND_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_tc_pool_bind_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_OCC_SNAPSHOT,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_occ_snapshot_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SB_OCC_MAX_CLEAR,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_sb_occ_max_clear_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_ESWITCH_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_eswitch_get_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_ESWITCH_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_eswitch_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_DPIPE_TABLE_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_dpipe_table_get,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_DPIPE_ENTRIES_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_dpipe_entries_get,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_DPIPE_HEADERS_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_dpipe_headers_get,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_DPIPE_TABLE_COUNTERS_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_dpipe_table_counters_set,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RESOURCE_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_resource_set,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RESOURCE_DUMP,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_resource_dump,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_RELOAD,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_reload,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PARAM_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_param_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PARAM_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_param_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_PARAM_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_param_get_doit,
 +              .dumpit = devlink_nl_cmd_port_param_get_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_PORT_PARAM_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_port_param_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_REGION_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_region_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_REGION_NEW,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_region_new,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_REGION_DEL,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_region_del,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_REGION_READ,
 +              .validate = GENL_DONT_VALIDATE_STRICT |
 +                          GENL_DONT_VALIDATE_DUMP_STRICT,
 +              .dumpit = devlink_nl_cmd_region_read_dumpit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_INFO_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_info_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_SET,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_RECOVER,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_recover_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_DIAGNOSE,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_diagnose_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_DUMP_GET,
 +              .validate = GENL_DONT_VALIDATE_STRICT |
 +                          GENL_DONT_VALIDATE_DUMP_STRICT,
 +              .dumpit = devlink_nl_cmd_health_reporter_dump_get_dumpit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_DUMP_CLEAR,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_dump_clear_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_HEALTH_REPORTER_TEST,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_health_reporter_test_doit,
 +              .flags = GENL_ADMIN_PERM,
 +              .internal_flags = DEVLINK_NL_FLAG_NEED_DEVLINK_OR_PORT,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_FLASH_UPDATE,
 +              .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
 +              .doit = devlink_nl_cmd_flash_update,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_GET,
 +              .doit = devlink_nl_cmd_trap_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_SET,
 +              .doit = devlink_nl_cmd_trap_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_GROUP_GET,
 +              .doit = devlink_nl_cmd_trap_group_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_GROUP_SET,
 +              .doit = devlink_nl_cmd_trap_group_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_POLICER_GET,
 +              .doit = devlink_nl_cmd_trap_policer_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_TRAP_POLICER_SET,
 +              .doit = devlink_nl_cmd_trap_policer_set_doit,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SELFTESTS_GET,
 +              .doit = devlink_nl_cmd_selftests_get_doit,
 +              .dumpit = devlink_nl_instance_iter_dumpit,
 +              /* can be retrieved by unprivileged users */
 +      },
 +      {
 +              .cmd = DEVLINK_CMD_SELFTESTS_RUN,
 +              .doit = devlink_nl_cmd_selftests_run,
 +              .flags = GENL_ADMIN_PERM,
 +      },
 +      /* -- No new ops here! Use split ops going forward! -- */
 +};
 +
 +static void
 +devlink_trap_policer_notify(struct devlink *devlink,
 +                          const struct devlink_trap_policer_item *policer_item,
 +                          enum devlink_command cmd);
 +static void
 +devlink_trap_group_notify(struct devlink *devlink,
 +                        const struct devlink_trap_group_item *group_item,
 +                        enum devlink_command cmd);
 +static void devlink_trap_notify(struct devlink *devlink,
 +                              const struct devlink_trap_item *trap_item,
 +                              enum devlink_command cmd);
 +
 +void devlink_notify_register(struct devlink *devlink)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      struct devlink_trap_group_item *group_item;
 +      struct devlink_param_item *param_item;
 +      struct devlink_trap_item *trap_item;
 +      struct devlink_port *devlink_port;
 +      struct devlink_linecard *linecard;
 +      struct devlink_rate *rate_node;
 +      struct devlink_region *region;
 +      unsigned long port_index;
 +
 +      devlink_notify(devlink, DEVLINK_CMD_NEW);
 +      list_for_each_entry(linecard, &devlink->linecard_list, list)
 +              devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +
 +      xa_for_each(&devlink->ports, port_index, devlink_port)
 +              devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_NEW);
 +
 +      list_for_each_entry(policer_item, &devlink->trap_policer_list, list)
 +              devlink_trap_policer_notify(devlink, policer_item,
 +                                          DEVLINK_CMD_TRAP_POLICER_NEW);
 +
 +      list_for_each_entry(group_item, &devlink->trap_group_list, list)
 +              devlink_trap_group_notify(devlink, group_item,
 +                                        DEVLINK_CMD_TRAP_GROUP_NEW);
 +
 +      list_for_each_entry(trap_item, &devlink->trap_list, list)
 +              devlink_trap_notify(devlink, trap_item, DEVLINK_CMD_TRAP_NEW);
 +
 +      list_for_each_entry(rate_node, &devlink->rate_list, list)
 +              devlink_rate_notify(rate_node, DEVLINK_CMD_RATE_NEW);
 +
 +      list_for_each_entry(region, &devlink->region_list, list)
 +              devlink_nl_region_notify(region, NULL, DEVLINK_CMD_REGION_NEW);
 +
 +      list_for_each_entry(param_item, &devlink->param_list, list)
 +              devlink_param_notify(devlink, 0, param_item,
 +                                   DEVLINK_CMD_PARAM_NEW);
 +}
 +
 +void devlink_notify_unregister(struct devlink *devlink)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      struct devlink_trap_group_item *group_item;
 +      struct devlink_param_item *param_item;
 +      struct devlink_trap_item *trap_item;
 +      struct devlink_port *devlink_port;
 +      struct devlink_rate *rate_node;
 +      struct devlink_region *region;
 +      unsigned long port_index;
 +
 +      list_for_each_entry_reverse(param_item, &devlink->param_list, list)
 +              devlink_param_notify(devlink, 0, param_item,
 +                                   DEVLINK_CMD_PARAM_DEL);
 +
 +      list_for_each_entry_reverse(region, &devlink->region_list, list)
 +              devlink_nl_region_notify(region, NULL, DEVLINK_CMD_REGION_DEL);
 +
 +      list_for_each_entry_reverse(rate_node, &devlink->rate_list, list)
 +              devlink_rate_notify(rate_node, DEVLINK_CMD_RATE_DEL);
 +
 +      list_for_each_entry_reverse(trap_item, &devlink->trap_list, list)
 +              devlink_trap_notify(devlink, trap_item, DEVLINK_CMD_TRAP_DEL);
 +
 +      list_for_each_entry_reverse(group_item, &devlink->trap_group_list, list)
 +              devlink_trap_group_notify(devlink, group_item,
 +                                        DEVLINK_CMD_TRAP_GROUP_DEL);
 +      list_for_each_entry_reverse(policer_item, &devlink->trap_policer_list,
 +                                  list)
 +              devlink_trap_policer_notify(devlink, policer_item,
 +                                          DEVLINK_CMD_TRAP_POLICER_DEL);
 +
 +      xa_for_each(&devlink->ports, port_index, devlink_port)
 +              devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_DEL);
 +      devlink_notify(devlink, DEVLINK_CMD_DEL);
 +}
 +
 +static void devlink_port_type_warn(struct work_struct *work)
 +{
 +      WARN(true, "Type was not set for devlink port.");
 +}
 +
 +static bool devlink_port_type_should_warn(struct devlink_port *devlink_port)
 +{
 +      /* Ignore CPU and DSA flavours. */
 +      return devlink_port->attrs.flavour != DEVLINK_PORT_FLAVOUR_CPU &&
 +             devlink_port->attrs.flavour != DEVLINK_PORT_FLAVOUR_DSA &&
 +             devlink_port->attrs.flavour != DEVLINK_PORT_FLAVOUR_UNUSED;
 +}
 +
 +#define DEVLINK_PORT_TYPE_WARN_TIMEOUT (HZ * 3600)
 +
 +static void devlink_port_type_warn_schedule(struct devlink_port *devlink_port)
 +{
 +      if (!devlink_port_type_should_warn(devlink_port))
 +              return;
 +      /* Schedule a work to WARN in case driver does not set port
 +       * type within timeout.
 +       */
 +      schedule_delayed_work(&devlink_port->type_warn_dw,
 +                            DEVLINK_PORT_TYPE_WARN_TIMEOUT);
 +}
 +
 +static void devlink_port_type_warn_cancel(struct devlink_port *devlink_port)
 +{
 +      if (!devlink_port_type_should_warn(devlink_port))
 +              return;
 +      cancel_delayed_work_sync(&devlink_port->type_warn_dw);
 +}
 +
 +/**
 + * devlink_port_init() - Init devlink port
 + *
 + * @devlink: devlink
 + * @devlink_port: devlink port
 + *
 + * Initialize essencial stuff that is needed for functions
 + * that may be called before devlink port registration.
 + * Call to this function is optional and not needed
 + * in case the driver does not use such functions.
 + */
 +void devlink_port_init(struct devlink *devlink,
 +                     struct devlink_port *devlink_port)
 +{
 +      if (devlink_port->initialized)
 +              return;
 +      devlink_port->devlink = devlink;
 +      INIT_LIST_HEAD(&devlink_port->region_list);
 +      devlink_port->initialized = true;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_init);
 +
 +/**
 + * devlink_port_fini() - Deinitialize devlink port
 + *
 + * @devlink_port: devlink port
 + *
 + * Deinitialize essencial stuff that is in use for functions
 + * that may be called after devlink port unregistration.
 + * Call to this function is optional and not needed
 + * in case the driver does not use such functions.
 + */
 +void devlink_port_fini(struct devlink_port *devlink_port)
 +{
 +      WARN_ON(!list_empty(&devlink_port->region_list));
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_fini);
 +
 +/**
 + * devl_port_register() - Register devlink port
 + *
 + * @devlink: devlink
 + * @devlink_port: devlink port
 + * @port_index: driver-specific numerical identifier of the port
 + *
 + * Register devlink port with provided port index. User can use
 + * any indexing, even hw-related one. devlink_port structure
 + * is convenient to be embedded inside user driver private structure.
 + * Note that the caller should take care of zeroing the devlink_port
 + * structure.
 + */
 +int devl_port_register(struct devlink *devlink,
 +                     struct devlink_port *devlink_port,
 +                     unsigned int port_index)
 +{
 +      int err;
 +
 +      devl_assert_locked(devlink);
 +
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      devlink_port_init(devlink, devlink_port);
 +      devlink_port->registered = true;
 +      devlink_port->index = port_index;
 +      spin_lock_init(&devlink_port->type_lock);
 +      INIT_LIST_HEAD(&devlink_port->reporter_list);
 +      err = xa_insert(&devlink->ports, port_index, devlink_port, GFP_KERNEL);
 +      if (err)
 +              return err;
 +
 +      INIT_DELAYED_WORK(&devlink_port->type_warn_dw, &devlink_port_type_warn);
 +      devlink_port_type_warn_schedule(devlink_port);
 +      devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_NEW);
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_port_register);
 +
 +/**
 + *    devlink_port_register - Register devlink port
 + *
 + *    @devlink: devlink
 + *    @devlink_port: devlink port
 + *    @port_index: driver-specific numerical identifier of the port
 + *
 + *    Register devlink port with provided port index. User can use
 + *    any indexing, even hw-related one. devlink_port structure
 + *    is convenient to be embedded inside user driver private structure.
 + *    Note that the caller should take care of zeroing the devlink_port
 + *    structure.
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +int devlink_port_register(struct devlink *devlink,
 +                        struct devlink_port *devlink_port,
 +                        unsigned int port_index)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_port_register(devlink, devlink_port, port_index);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_register);
 +
 +/**
 + * devl_port_unregister() - Unregister devlink port
 + *
 + * @devlink_port: devlink port
 + */
 +void devl_port_unregister(struct devlink_port *devlink_port)
 +{
 +      lockdep_assert_held(&devlink_port->devlink->lock);
 +      WARN_ON(devlink_port->type != DEVLINK_PORT_TYPE_NOTSET);
 +
 +      devlink_port_type_warn_cancel(devlink_port);
 +      devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_DEL);
 +      xa_erase(&devlink_port->devlink->ports, devlink_port->index);
 +      WARN_ON(!list_empty(&devlink_port->reporter_list));
 +      devlink_port->registered = false;
 +}
 +EXPORT_SYMBOL_GPL(devl_port_unregister);
 +
 +/**
 + *    devlink_port_unregister - Unregister devlink port
 + *
 + *    @devlink_port: devlink port
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_port_unregister(struct devlink_port *devlink_port)
 +{
 +      struct devlink *devlink = devlink_port->devlink;
 +
 +      devl_lock(devlink);
 +      devl_port_unregister(devlink_port);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_unregister);
 +
 +static void devlink_port_type_netdev_checks(struct devlink_port *devlink_port,
 +                                          struct net_device *netdev)
 +{
 +      const struct net_device_ops *ops = netdev->netdev_ops;
 +
 +      /* If driver registers devlink port, it should set devlink port
 +       * attributes accordingly so the compat functions are called
 +       * and the original ops are not used.
 +       */
 +      if (ops->ndo_get_phys_port_name) {
 +              /* Some drivers use the same set of ndos for netdevs
 +               * that have devlink_port registered and also for
 +               * those who don't. Make sure that ndo_get_phys_port_name
 +               * returns -EOPNOTSUPP here in case it is defined.
 +               * Warn if not.
 +               */
 +              char name[IFNAMSIZ];
 +              int err;
 +
 +              err = ops->ndo_get_phys_port_name(netdev, name, sizeof(name));
 +              WARN_ON(err != -EOPNOTSUPP);
 +      }
 +      if (ops->ndo_get_port_parent_id) {
 +              /* Some drivers use the same set of ndos for netdevs
 +               * that have devlink_port registered and also for
 +               * those who don't. Make sure that ndo_get_port_parent_id
 +               * returns -EOPNOTSUPP here in case it is defined.
 +               * Warn if not.
 +               */
 +              struct netdev_phys_item_id ppid;
 +              int err;
 +
 +              err = ops->ndo_get_port_parent_id(netdev, &ppid);
 +              WARN_ON(err != -EOPNOTSUPP);
 +      }
 +}
 +
 +static void __devlink_port_type_set(struct devlink_port *devlink_port,
 +                                  enum devlink_port_type type,
 +                                  void *type_dev)
 +{
 +      struct net_device *netdev = type_dev;
 +
 +      ASSERT_DEVLINK_PORT_REGISTERED(devlink_port);
 +
 +      if (type == DEVLINK_PORT_TYPE_NOTSET) {
 +              devlink_port_type_warn_schedule(devlink_port);
 +      } else {
 +              devlink_port_type_warn_cancel(devlink_port);
 +              if (type == DEVLINK_PORT_TYPE_ETH && netdev)
 +                      devlink_port_type_netdev_checks(devlink_port, netdev);
 +      }
 +
 +      spin_lock_bh(&devlink_port->type_lock);
 +      devlink_port->type = type;
 +      switch (type) {
 +      case DEVLINK_PORT_TYPE_ETH:
 +              devlink_port->type_eth.netdev = netdev;
 +              if (netdev) {
 +                      ASSERT_RTNL();
 +                      devlink_port->type_eth.ifindex = netdev->ifindex;
 +                      BUILD_BUG_ON(sizeof(devlink_port->type_eth.ifname) !=
 +                                   sizeof(netdev->name));
 +                      strcpy(devlink_port->type_eth.ifname, netdev->name);
 +              }
 +              break;
 +      case DEVLINK_PORT_TYPE_IB:
 +              devlink_port->type_ib.ibdev = type_dev;
 +              break;
 +      default:
 +              break;
 +      }
 +      spin_unlock_bh(&devlink_port->type_lock);
 +      devlink_port_notify(devlink_port, DEVLINK_CMD_PORT_NEW);
 +}
 +
 +/**
 + *    devlink_port_type_eth_set - Set port type to Ethernet
 + *
 + *    @devlink_port: devlink port
 + *
 + *    If driver is calling this, most likely it is doing something wrong.
 + */
 +void devlink_port_type_eth_set(struct devlink_port *devlink_port)
 +{
 +      dev_warn(devlink_port->devlink->dev,
 +               "devlink port type for port %d set to Ethernet without a software interface reference, device type not supported by the kernel?\n",
 +               devlink_port->index);
 +      __devlink_port_type_set(devlink_port, DEVLINK_PORT_TYPE_ETH, NULL);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_type_eth_set);
 +
 +/**
 + *    devlink_port_type_ib_set - Set port type to InfiniBand
 + *
 + *    @devlink_port: devlink port
 + *    @ibdev: related IB device
 + */
 +void devlink_port_type_ib_set(struct devlink_port *devlink_port,
 +                            struct ib_device *ibdev)
 +{
 +      __devlink_port_type_set(devlink_port, DEVLINK_PORT_TYPE_IB, ibdev);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_type_ib_set);
 +
 +/**
 + *    devlink_port_type_clear - Clear port type
 + *
 + *    @devlink_port: devlink port
 + *
 + *    If driver is calling this for clearing Ethernet type, most likely
 + *    it is doing something wrong.
 + */
 +void devlink_port_type_clear(struct devlink_port *devlink_port)
 +{
 +      if (devlink_port->type == DEVLINK_PORT_TYPE_ETH)
 +              dev_warn(devlink_port->devlink->dev,
 +                       "devlink port type for port %d cleared without a software interface reference, device type not supported by the kernel?\n",
 +                       devlink_port->index);
 +      __devlink_port_type_set(devlink_port, DEVLINK_PORT_TYPE_NOTSET, NULL);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_type_clear);
 +
 +int devlink_port_netdevice_event(struct notifier_block *nb,
 +                               unsigned long event, void *ptr)
 +{
 +      struct net_device *netdev = netdev_notifier_info_to_dev(ptr);
 +      struct devlink_port *devlink_port = netdev->devlink_port;
 +      struct devlink *devlink;
 +
 +      devlink = container_of(nb, struct devlink, netdevice_nb);
 +
 +      if (!devlink_port || devlink_port->devlink != devlink)
 +              return NOTIFY_OK;
 +
 +      switch (event) {
 +      case NETDEV_POST_INIT:
 +              /* Set the type but not netdev pointer. It is going to be set
 +               * later on by NETDEV_REGISTER event. Happens once during
 +               * netdevice register
 +               */
 +              __devlink_port_type_set(devlink_port, DEVLINK_PORT_TYPE_ETH,
 +                                      NULL);
 +              break;
 +      case NETDEV_REGISTER:
 +      case NETDEV_CHANGENAME:
++              if (devlink_net(devlink) != dev_net(netdev))
++                      return NOTIFY_OK;
 +              /* Set the netdev on top of previously set type. Note this
 +               * event happens also during net namespace change so here
 +               * we take into account netdev pointer appearing in this
 +               * namespace.
 +               */
 +              __devlink_port_type_set(devlink_port, devlink_port->type,
 +                                      netdev);
 +              break;
 +      case NETDEV_UNREGISTER:
++              if (devlink_net(devlink) != dev_net(netdev))
++                      return NOTIFY_OK;
 +              /* Clear netdev pointer, but not the type. This event happens
 +               * also during net namespace change so we need to clear
 +               * pointer to netdev that is going to another net namespace.
 +               */
 +              __devlink_port_type_set(devlink_port, devlink_port->type,
 +                                      NULL);
 +              break;
 +      case NETDEV_PRE_UNINIT:
 +              /* Clear the type and the netdev pointer. Happens one during
 +               * netdevice unregister.
 +               */
 +              __devlink_port_type_set(devlink_port, DEVLINK_PORT_TYPE_NOTSET,
 +                                      NULL);
 +              break;
 +      }
 +
 +      return NOTIFY_OK;
 +}
 +
 +static int __devlink_port_attrs_set(struct devlink_port *devlink_port,
 +                                  enum devlink_port_flavour flavour)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +
 +      devlink_port->attrs_set = true;
 +      attrs->flavour = flavour;
 +      if (attrs->switch_id.id_len) {
 +              devlink_port->switch_port = true;
 +              if (WARN_ON(attrs->switch_id.id_len > MAX_PHYS_ITEM_ID_LEN))
 +                      attrs->switch_id.id_len = MAX_PHYS_ITEM_ID_LEN;
 +      } else {
 +              devlink_port->switch_port = false;
 +      }
 +      return 0;
 +}
 +
 +/**
 + *    devlink_port_attrs_set - Set port attributes
 + *
 + *    @devlink_port: devlink port
 + *    @attrs: devlink port attrs
 + */
 +void devlink_port_attrs_set(struct devlink_port *devlink_port,
 +                          struct devlink_port_attrs *attrs)
 +{
 +      int ret;
 +
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      devlink_port->attrs = *attrs;
 +      ret = __devlink_port_attrs_set(devlink_port, attrs->flavour);
 +      if (ret)
 +              return;
 +      WARN_ON(attrs->splittable && attrs->split);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_attrs_set);
 +
 +/**
 + *    devlink_port_attrs_pci_pf_set - Set PCI PF port attributes
 + *
 + *    @devlink_port: devlink port
 + *    @controller: associated controller number for the devlink port instance
 + *    @pf: associated PF for the devlink port instance
 + *    @external: indicates if the port is for an external controller
 + */
 +void devlink_port_attrs_pci_pf_set(struct devlink_port *devlink_port, u32 controller,
 +                                 u16 pf, bool external)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +      int ret;
 +
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      ret = __devlink_port_attrs_set(devlink_port,
 +                                     DEVLINK_PORT_FLAVOUR_PCI_PF);
 +      if (ret)
 +              return;
 +      attrs->pci_pf.controller = controller;
 +      attrs->pci_pf.pf = pf;
 +      attrs->pci_pf.external = external;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_attrs_pci_pf_set);
 +
 +/**
 + *    devlink_port_attrs_pci_vf_set - Set PCI VF port attributes
 + *
 + *    @devlink_port: devlink port
 + *    @controller: associated controller number for the devlink port instance
 + *    @pf: associated PF for the devlink port instance
 + *    @vf: associated VF of a PF for the devlink port instance
 + *    @external: indicates if the port is for an external controller
 + */
 +void devlink_port_attrs_pci_vf_set(struct devlink_port *devlink_port, u32 controller,
 +                                 u16 pf, u16 vf, bool external)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +      int ret;
 +
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      ret = __devlink_port_attrs_set(devlink_port,
 +                                     DEVLINK_PORT_FLAVOUR_PCI_VF);
 +      if (ret)
 +              return;
 +      attrs->pci_vf.controller = controller;
 +      attrs->pci_vf.pf = pf;
 +      attrs->pci_vf.vf = vf;
 +      attrs->pci_vf.external = external;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_attrs_pci_vf_set);
 +
 +/**
 + *    devlink_port_attrs_pci_sf_set - Set PCI SF port attributes
 + *
 + *    @devlink_port: devlink port
 + *    @controller: associated controller number for the devlink port instance
 + *    @pf: associated PF for the devlink port instance
 + *    @sf: associated SF of a PF for the devlink port instance
 + *    @external: indicates if the port is for an external controller
 + */
 +void devlink_port_attrs_pci_sf_set(struct devlink_port *devlink_port, u32 controller,
 +                                 u16 pf, u32 sf, bool external)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +      int ret;
 +
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      ret = __devlink_port_attrs_set(devlink_port,
 +                                     DEVLINK_PORT_FLAVOUR_PCI_SF);
 +      if (ret)
 +              return;
 +      attrs->pci_sf.controller = controller;
 +      attrs->pci_sf.pf = pf;
 +      attrs->pci_sf.sf = sf;
 +      attrs->pci_sf.external = external;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_attrs_pci_sf_set);
 +
 +/**
 + * devl_rate_node_create - create devlink rate node
 + * @devlink: devlink instance
 + * @priv: driver private data
 + * @node_name: name of the resulting node
 + * @parent: parent devlink_rate struct
 + *
 + * Create devlink rate object of type node
 + */
 +struct devlink_rate *
 +devl_rate_node_create(struct devlink *devlink, void *priv, char *node_name,
 +                    struct devlink_rate *parent)
 +{
 +      struct devlink_rate *rate_node;
 +
 +      rate_node = devlink_rate_node_get_by_name(devlink, node_name);
 +      if (!IS_ERR(rate_node))
 +              return ERR_PTR(-EEXIST);
 +
 +      rate_node = kzalloc(sizeof(*rate_node), GFP_KERNEL);
 +      if (!rate_node)
 +              return ERR_PTR(-ENOMEM);
 +
 +      if (parent) {
 +              rate_node->parent = parent;
 +              refcount_inc(&rate_node->parent->refcnt);
 +      }
 +
 +      rate_node->type = DEVLINK_RATE_TYPE_NODE;
 +      rate_node->devlink = devlink;
 +      rate_node->priv = priv;
 +
 +      rate_node->name = kstrdup(node_name, GFP_KERNEL);
 +      if (!rate_node->name) {
 +              kfree(rate_node);
 +              return ERR_PTR(-ENOMEM);
 +      }
 +
 +      refcount_set(&rate_node->refcnt, 1);
 +      list_add(&rate_node->list, &devlink->rate_list);
 +      devlink_rate_notify(rate_node, DEVLINK_CMD_RATE_NEW);
 +      return rate_node;
 +}
 +EXPORT_SYMBOL_GPL(devl_rate_node_create);
 +
 +/**
 + * devl_rate_leaf_create - create devlink rate leaf
 + * @devlink_port: devlink port object to create rate object on
 + * @priv: driver private data
 + * @parent: parent devlink_rate struct
 + *
 + * Create devlink rate object of type leaf on provided @devlink_port.
 + */
 +int devl_rate_leaf_create(struct devlink_port *devlink_port, void *priv,
 +                        struct devlink_rate *parent)
 +{
 +      struct devlink *devlink = devlink_port->devlink;
 +      struct devlink_rate *devlink_rate;
 +
 +      devl_assert_locked(devlink_port->devlink);
 +
 +      if (WARN_ON(devlink_port->devlink_rate))
 +              return -EBUSY;
 +
 +      devlink_rate = kzalloc(sizeof(*devlink_rate), GFP_KERNEL);
 +      if (!devlink_rate)
 +              return -ENOMEM;
 +
 +      if (parent) {
 +              devlink_rate->parent = parent;
 +              refcount_inc(&devlink_rate->parent->refcnt);
 +      }
 +
 +      devlink_rate->type = DEVLINK_RATE_TYPE_LEAF;
 +      devlink_rate->devlink = devlink;
 +      devlink_rate->devlink_port = devlink_port;
 +      devlink_rate->priv = priv;
 +      list_add_tail(&devlink_rate->list, &devlink->rate_list);
 +      devlink_port->devlink_rate = devlink_rate;
 +      devlink_rate_notify(devlink_rate, DEVLINK_CMD_RATE_NEW);
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_rate_leaf_create);
 +
 +/**
 + * devl_rate_leaf_destroy - destroy devlink rate leaf
 + *
 + * @devlink_port: devlink port linked to the rate object
 + *
 + * Destroy the devlink rate object of type leaf on provided @devlink_port.
 + */
 +void devl_rate_leaf_destroy(struct devlink_port *devlink_port)
 +{
 +      struct devlink_rate *devlink_rate = devlink_port->devlink_rate;
 +
 +      devl_assert_locked(devlink_port->devlink);
 +      if (!devlink_rate)
 +              return;
 +
 +      devlink_rate_notify(devlink_rate, DEVLINK_CMD_RATE_DEL);
 +      if (devlink_rate->parent)
 +              refcount_dec(&devlink_rate->parent->refcnt);
 +      list_del(&devlink_rate->list);
 +      devlink_port->devlink_rate = NULL;
 +      kfree(devlink_rate);
 +}
 +EXPORT_SYMBOL_GPL(devl_rate_leaf_destroy);
 +
 +/**
 + * devl_rate_nodes_destroy - destroy all devlink rate nodes on device
 + * @devlink: devlink instance
 + *
 + * Unset parent for all rate objects and destroy all rate nodes
 + * on specified device.
 + */
 +void devl_rate_nodes_destroy(struct devlink *devlink)
 +{
 +      static struct devlink_rate *devlink_rate, *tmp;
 +      const struct devlink_ops *ops = devlink->ops;
 +
 +      devl_assert_locked(devlink);
 +
 +      list_for_each_entry(devlink_rate, &devlink->rate_list, list) {
 +              if (!devlink_rate->parent)
 +                      continue;
 +
 +              refcount_dec(&devlink_rate->parent->refcnt);
 +              if (devlink_rate_is_leaf(devlink_rate))
 +                      ops->rate_leaf_parent_set(devlink_rate, NULL, devlink_rate->priv,
 +                                                NULL, NULL);
 +              else if (devlink_rate_is_node(devlink_rate))
 +                      ops->rate_node_parent_set(devlink_rate, NULL, devlink_rate->priv,
 +                                                NULL, NULL);
 +      }
 +      list_for_each_entry_safe(devlink_rate, tmp, &devlink->rate_list, list) {
 +              if (devlink_rate_is_node(devlink_rate)) {
 +                      ops->rate_node_del(devlink_rate, devlink_rate->priv, NULL);
 +                      list_del(&devlink_rate->list);
 +                      kfree(devlink_rate->name);
 +                      kfree(devlink_rate);
 +              }
 +      }
 +}
 +EXPORT_SYMBOL_GPL(devl_rate_nodes_destroy);
 +
 +/**
 + *    devlink_port_linecard_set - Link port with a linecard
 + *
 + *    @devlink_port: devlink port
 + *    @linecard: devlink linecard
 + */
 +void devlink_port_linecard_set(struct devlink_port *devlink_port,
 +                             struct devlink_linecard *linecard)
 +{
 +      ASSERT_DEVLINK_PORT_NOT_REGISTERED(devlink_port);
 +
 +      devlink_port->linecard = linecard;
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_linecard_set);
 +
 +static int __devlink_port_phys_port_name_get(struct devlink_port *devlink_port,
 +                                           char *name, size_t len)
 +{
 +      struct devlink_port_attrs *attrs = &devlink_port->attrs;
 +      int n = 0;
 +
 +      if (!devlink_port->attrs_set)
 +              return -EOPNOTSUPP;
 +
 +      switch (attrs->flavour) {
 +      case DEVLINK_PORT_FLAVOUR_PHYSICAL:
 +              if (devlink_port->linecard)
 +                      n = snprintf(name, len, "l%u",
 +                                   devlink_port->linecard->index);
 +              if (n < len)
 +                      n += snprintf(name + n, len - n, "p%u",
 +                                    attrs->phys.port_number);
 +              if (n < len && attrs->split)
 +                      n += snprintf(name + n, len - n, "s%u",
 +                                    attrs->phys.split_subport_number);
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_CPU:
 +      case DEVLINK_PORT_FLAVOUR_DSA:
 +      case DEVLINK_PORT_FLAVOUR_UNUSED:
 +              /* As CPU and DSA ports do not have a netdevice associated
 +               * case should not ever happen.
 +               */
 +              WARN_ON(1);
 +              return -EINVAL;
 +      case DEVLINK_PORT_FLAVOUR_PCI_PF:
 +              if (attrs->pci_pf.external) {
 +                      n = snprintf(name, len, "c%u", attrs->pci_pf.controller);
 +                      if (n >= len)
 +                              return -EINVAL;
 +                      len -= n;
 +                      name += n;
 +              }
 +              n = snprintf(name, len, "pf%u", attrs->pci_pf.pf);
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_PCI_VF:
 +              if (attrs->pci_vf.external) {
 +                      n = snprintf(name, len, "c%u", attrs->pci_vf.controller);
 +                      if (n >= len)
 +                              return -EINVAL;
 +                      len -= n;
 +                      name += n;
 +              }
 +              n = snprintf(name, len, "pf%uvf%u",
 +                           attrs->pci_vf.pf, attrs->pci_vf.vf);
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_PCI_SF:
 +              if (attrs->pci_sf.external) {
 +                      n = snprintf(name, len, "c%u", attrs->pci_sf.controller);
 +                      if (n >= len)
 +                              return -EINVAL;
 +                      len -= n;
 +                      name += n;
 +              }
 +              n = snprintf(name, len, "pf%usf%u", attrs->pci_sf.pf,
 +                           attrs->pci_sf.sf);
 +              break;
 +      case DEVLINK_PORT_FLAVOUR_VIRTUAL:
 +              return -EOPNOTSUPP;
 +      }
 +
 +      if (n >= len)
 +              return -EINVAL;
 +
 +      return 0;
 +}
 +
 +static int devlink_linecard_types_init(struct devlink_linecard *linecard)
 +{
 +      struct devlink_linecard_type *linecard_type;
 +      unsigned int count;
 +      int i;
 +
 +      count = linecard->ops->types_count(linecard, linecard->priv);
 +      linecard->types = kmalloc_array(count, sizeof(*linecard_type),
 +                                      GFP_KERNEL);
 +      if (!linecard->types)
 +              return -ENOMEM;
 +      linecard->types_count = count;
 +
 +      for (i = 0; i < count; i++) {
 +              linecard_type = &linecard->types[i];
 +              linecard->ops->types_get(linecard, linecard->priv, i,
 +                                       &linecard_type->type,
 +                                       &linecard_type->priv);
 +      }
 +      return 0;
 +}
 +
 +static void devlink_linecard_types_fini(struct devlink_linecard *linecard)
 +{
 +      kfree(linecard->types);
 +}
 +
 +/**
 + *    devl_linecard_create - Create devlink linecard
 + *
 + *    @devlink: devlink
 + *    @linecard_index: driver-specific numerical identifier of the linecard
 + *    @ops: linecards ops
 + *    @priv: user priv pointer
 + *
 + *    Create devlink linecard instance with provided linecard index.
 + *    Caller can use any indexing, even hw-related one.
 + *
 + *    Return: Line card structure or an ERR_PTR() encoded error code.
 + */
 +struct devlink_linecard *
 +devl_linecard_create(struct devlink *devlink, unsigned int linecard_index,
 +                   const struct devlink_linecard_ops *ops, void *priv)
 +{
 +      struct devlink_linecard *linecard;
 +      int err;
 +
 +      if (WARN_ON(!ops || !ops->provision || !ops->unprovision ||
 +                  !ops->types_count || !ops->types_get))
 +              return ERR_PTR(-EINVAL);
 +
 +      if (devlink_linecard_index_exists(devlink, linecard_index))
 +              return ERR_PTR(-EEXIST);
 +
 +      linecard = kzalloc(sizeof(*linecard), GFP_KERNEL);
 +      if (!linecard)
 +              return ERR_PTR(-ENOMEM);
 +
 +      linecard->devlink = devlink;
 +      linecard->index = linecard_index;
 +      linecard->ops = ops;
 +      linecard->priv = priv;
 +      linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONED;
 +      mutex_init(&linecard->state_lock);
 +
 +      err = devlink_linecard_types_init(linecard);
 +      if (err) {
 +              mutex_destroy(&linecard->state_lock);
 +              kfree(linecard);
 +              return ERR_PTR(err);
 +      }
 +
 +      list_add_tail(&linecard->list, &devlink->linecard_list);
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      return linecard;
 +}
 +EXPORT_SYMBOL_GPL(devl_linecard_create);
 +
 +/**
 + *    devl_linecard_destroy - Destroy devlink linecard
 + *
 + *    @linecard: devlink linecard
 + */
 +void devl_linecard_destroy(struct devlink_linecard *linecard)
 +{
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_DEL);
 +      list_del(&linecard->list);
 +      devlink_linecard_types_fini(linecard);
 +      mutex_destroy(&linecard->state_lock);
 +      kfree(linecard);
 +}
 +EXPORT_SYMBOL_GPL(devl_linecard_destroy);
 +
 +/**
 + *    devlink_linecard_provision_set - Set provisioning on linecard
 + *
 + *    @linecard: devlink linecard
 + *    @type: linecard type
 + *
 + *    This is either called directly from the provision() op call or
 + *    as a result of the provision() op call asynchronously.
 + */
 +void devlink_linecard_provision_set(struct devlink_linecard *linecard,
 +                                  const char *type)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      WARN_ON(linecard->type && strcmp(linecard->type, type));
 +      linecard->state = DEVLINK_LINECARD_STATE_PROVISIONED;
 +      linecard->type = type;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_provision_set);
 +
 +/**
 + *    devlink_linecard_provision_clear - Clear provisioning on linecard
 + *
 + *    @linecard: devlink linecard
 + *
 + *    This is either called directly from the unprovision() op call or
 + *    as a result of the unprovision() op call asynchronously.
 + */
 +void devlink_linecard_provision_clear(struct devlink_linecard *linecard)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      WARN_ON(linecard->nested_devlink);
 +      linecard->state = DEVLINK_LINECARD_STATE_UNPROVISIONED;
 +      linecard->type = NULL;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_provision_clear);
 +
 +/**
 + *    devlink_linecard_provision_fail - Fail provisioning on linecard
 + *
 + *    @linecard: devlink linecard
 + *
 + *    This is either called directly from the provision() op call or
 + *    as a result of the provision() op call asynchronously.
 + */
 +void devlink_linecard_provision_fail(struct devlink_linecard *linecard)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      WARN_ON(linecard->nested_devlink);
 +      linecard->state = DEVLINK_LINECARD_STATE_PROVISIONING_FAILED;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_provision_fail);
 +
 +/**
 + *    devlink_linecard_activate - Set linecard active
 + *
 + *    @linecard: devlink linecard
 + */
 +void devlink_linecard_activate(struct devlink_linecard *linecard)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      WARN_ON(linecard->state != DEVLINK_LINECARD_STATE_PROVISIONED);
 +      linecard->state = DEVLINK_LINECARD_STATE_ACTIVE;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_activate);
 +
 +/**
 + *    devlink_linecard_deactivate - Set linecard inactive
 + *
 + *    @linecard: devlink linecard
 + */
 +void devlink_linecard_deactivate(struct devlink_linecard *linecard)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      switch (linecard->state) {
 +      case DEVLINK_LINECARD_STATE_ACTIVE:
 +              linecard->state = DEVLINK_LINECARD_STATE_PROVISIONED;
 +              devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +              break;
 +      case DEVLINK_LINECARD_STATE_UNPROVISIONING:
 +              /* Line card is being deactivated as part
 +               * of unprovisioning flow.
 +               */
 +              break;
 +      default:
 +              WARN_ON(1);
 +              break;
 +      }
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_deactivate);
 +
 +/**
 + *    devlink_linecard_nested_dl_set - Attach/detach nested devlink
 + *                                     instance to linecard.
 + *
 + *    @linecard: devlink linecard
 + *    @nested_devlink: devlink instance to attach or NULL to detach
 + */
 +void devlink_linecard_nested_dl_set(struct devlink_linecard *linecard,
 +                                  struct devlink *nested_devlink)
 +{
 +      mutex_lock(&linecard->state_lock);
 +      linecard->nested_devlink = nested_devlink;
 +      devlink_linecard_notify(linecard, DEVLINK_CMD_LINECARD_NEW);
 +      mutex_unlock(&linecard->state_lock);
 +}
 +EXPORT_SYMBOL_GPL(devlink_linecard_nested_dl_set);
 +
 +int devl_sb_register(struct devlink *devlink, unsigned int sb_index,
 +                   u32 size, u16 ingress_pools_count,
 +                   u16 egress_pools_count, u16 ingress_tc_count,
 +                   u16 egress_tc_count)
 +{
 +      struct devlink_sb *devlink_sb;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      if (devlink_sb_index_exists(devlink, sb_index))
 +              return -EEXIST;
 +
 +      devlink_sb = kzalloc(sizeof(*devlink_sb), GFP_KERNEL);
 +      if (!devlink_sb)
 +              return -ENOMEM;
 +      devlink_sb->index = sb_index;
 +      devlink_sb->size = size;
 +      devlink_sb->ingress_pools_count = ingress_pools_count;
 +      devlink_sb->egress_pools_count = egress_pools_count;
 +      devlink_sb->ingress_tc_count = ingress_tc_count;
 +      devlink_sb->egress_tc_count = egress_tc_count;
 +      list_add_tail(&devlink_sb->list, &devlink->sb_list);
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_sb_register);
 +
 +int devlink_sb_register(struct devlink *devlink, unsigned int sb_index,
 +                      u32 size, u16 ingress_pools_count,
 +                      u16 egress_pools_count, u16 ingress_tc_count,
 +                      u16 egress_tc_count)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_sb_register(devlink, sb_index, size, ingress_pools_count,
 +                             egress_pools_count, ingress_tc_count,
 +                             egress_tc_count);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_sb_register);
 +
 +void devl_sb_unregister(struct devlink *devlink, unsigned int sb_index)
 +{
 +      struct devlink_sb *devlink_sb;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      devlink_sb = devlink_sb_get_by_index(devlink, sb_index);
 +      WARN_ON(!devlink_sb);
 +      list_del(&devlink_sb->list);
 +      kfree(devlink_sb);
 +}
 +EXPORT_SYMBOL_GPL(devl_sb_unregister);
 +
 +void devlink_sb_unregister(struct devlink *devlink, unsigned int sb_index)
 +{
 +      devl_lock(devlink);
 +      devl_sb_unregister(devlink, sb_index);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_sb_unregister);
 +
 +/**
 + * devl_dpipe_headers_register - register dpipe headers
 + *
 + * @devlink: devlink
 + * @dpipe_headers: dpipe header array
 + *
 + * Register the headers supported by hardware.
 + */
 +void devl_dpipe_headers_register(struct devlink *devlink,
 +                               struct devlink_dpipe_headers *dpipe_headers)
 +{
 +      lockdep_assert_held(&devlink->lock);
 +
 +      devlink->dpipe_headers = dpipe_headers;
 +}
 +EXPORT_SYMBOL_GPL(devl_dpipe_headers_register);
 +
 +/**
 + * devl_dpipe_headers_unregister - unregister dpipe headers
 + *
 + * @devlink: devlink
 + *
 + * Unregister the headers supported by hardware.
 + */
 +void devl_dpipe_headers_unregister(struct devlink *devlink)
 +{
 +      lockdep_assert_held(&devlink->lock);
 +
 +      devlink->dpipe_headers = NULL;
 +}
 +EXPORT_SYMBOL_GPL(devl_dpipe_headers_unregister);
 +
 +/**
 + *    devlink_dpipe_table_counter_enabled - check if counter allocation
 + *                                          required
 + *    @devlink: devlink
 + *    @table_name: tables name
 + *
 + *    Used by driver to check if counter allocation is required.
 + *    After counter allocation is turned on the table entries
 + *    are updated to include counter statistics.
 + *
 + *    After that point on the driver must respect the counter
 + *    state so that each entry added to the table is added
 + *    with a counter.
 + */
 +bool devlink_dpipe_table_counter_enabled(struct devlink *devlink,
 +                                       const char *table_name)
 +{
 +      struct devlink_dpipe_table *table;
 +      bool enabled;
 +
 +      rcu_read_lock();
 +      table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 +                                       table_name, devlink);
 +      enabled = false;
 +      if (table)
 +              enabled = table->counters_enabled;
 +      rcu_read_unlock();
 +      return enabled;
 +}
 +EXPORT_SYMBOL_GPL(devlink_dpipe_table_counter_enabled);
 +
 +/**
 + * devl_dpipe_table_register - register dpipe table
 + *
 + * @devlink: devlink
 + * @table_name: table name
 + * @table_ops: table ops
 + * @priv: priv
 + * @counter_control_extern: external control for counters
 + */
 +int devl_dpipe_table_register(struct devlink *devlink,
 +                            const char *table_name,
 +                            struct devlink_dpipe_table_ops *table_ops,
 +                            void *priv, bool counter_control_extern)
 +{
 +      struct devlink_dpipe_table *table;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      if (WARN_ON(!table_ops->size_get))
 +              return -EINVAL;
 +
 +      if (devlink_dpipe_table_find(&devlink->dpipe_table_list, table_name,
 +                                   devlink))
 +              return -EEXIST;
 +
 +      table = kzalloc(sizeof(*table), GFP_KERNEL);
 +      if (!table)
 +              return -ENOMEM;
 +
 +      table->name = table_name;
 +      table->table_ops = table_ops;
 +      table->priv = priv;
 +      table->counter_control_extern = counter_control_extern;
 +
 +      list_add_tail_rcu(&table->list, &devlink->dpipe_table_list);
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_dpipe_table_register);
 +
 +/**
 + * devl_dpipe_table_unregister - unregister dpipe table
 + *
 + * @devlink: devlink
 + * @table_name: table name
 + */
 +void devl_dpipe_table_unregister(struct devlink *devlink,
 +                               const char *table_name)
 +{
 +      struct devlink_dpipe_table *table;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 +                                       table_name, devlink);
 +      if (!table)
 +              return;
 +      list_del_rcu(&table->list);
 +      kfree_rcu(table, rcu);
 +}
 +EXPORT_SYMBOL_GPL(devl_dpipe_table_unregister);
 +
 +/**
 + * devl_resource_register - devlink resource register
 + *
 + * @devlink: devlink
 + * @resource_name: resource's name
 + * @resource_size: resource's size
 + * @resource_id: resource's id
 + * @parent_resource_id: resource's parent id
 + * @size_params: size parameters
 + *
 + * Generic resources should reuse the same names across drivers.
 + * Please see the generic resources list at:
 + * Documentation/networking/devlink/devlink-resource.rst
 + */
 +int devl_resource_register(struct devlink *devlink,
 +                         const char *resource_name,
 +                         u64 resource_size,
 +                         u64 resource_id,
 +                         u64 parent_resource_id,
 +                         const struct devlink_resource_size_params *size_params)
 +{
 +      struct devlink_resource *resource;
 +      struct list_head *resource_list;
 +      bool top_hierarchy;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      top_hierarchy = parent_resource_id == DEVLINK_RESOURCE_ID_PARENT_TOP;
 +
 +      resource = devlink_resource_find(devlink, NULL, resource_id);
 +      if (resource)
 +              return -EINVAL;
 +
 +      resource = kzalloc(sizeof(*resource), GFP_KERNEL);
 +      if (!resource)
 +              return -ENOMEM;
 +
 +      if (top_hierarchy) {
 +              resource_list = &devlink->resource_list;
 +      } else {
 +              struct devlink_resource *parent_resource;
 +
 +              parent_resource = devlink_resource_find(devlink, NULL,
 +                                                      parent_resource_id);
 +              if (parent_resource) {
 +                      resource_list = &parent_resource->resource_list;
 +                      resource->parent = parent_resource;
 +              } else {
 +                      kfree(resource);
 +                      return -EINVAL;
 +              }
 +      }
 +
 +      resource->name = resource_name;
 +      resource->size = resource_size;
 +      resource->size_new = resource_size;
 +      resource->id = resource_id;
 +      resource->size_valid = true;
 +      memcpy(&resource->size_params, size_params,
 +             sizeof(resource->size_params));
 +      INIT_LIST_HEAD(&resource->resource_list);
 +      list_add_tail(&resource->list, resource_list);
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_resource_register);
 +
 +/**
 + *    devlink_resource_register - devlink resource register
 + *
 + *    @devlink: devlink
 + *    @resource_name: resource's name
 + *    @resource_size: resource's size
 + *    @resource_id: resource's id
 + *    @parent_resource_id: resource's parent id
 + *    @size_params: size parameters
 + *
 + *    Generic resources should reuse the same names across drivers.
 + *    Please see the generic resources list at:
 + *    Documentation/networking/devlink/devlink-resource.rst
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +int devlink_resource_register(struct devlink *devlink,
 +                            const char *resource_name,
 +                            u64 resource_size,
 +                            u64 resource_id,
 +                            u64 parent_resource_id,
 +                            const struct devlink_resource_size_params *size_params)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_resource_register(devlink, resource_name, resource_size,
 +                                   resource_id, parent_resource_id, size_params);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_resource_register);
 +
 +static void devlink_resource_unregister(struct devlink *devlink,
 +                                      struct devlink_resource *resource)
 +{
 +      struct devlink_resource *tmp, *child_resource;
 +
 +      list_for_each_entry_safe(child_resource, tmp, &resource->resource_list,
 +                               list) {
 +              devlink_resource_unregister(devlink, child_resource);
 +              list_del(&child_resource->list);
 +              kfree(child_resource);
 +      }
 +}
 +
 +/**
 + * devl_resources_unregister - free all resources
 + *
 + * @devlink: devlink
 + */
 +void devl_resources_unregister(struct devlink *devlink)
 +{
 +      struct devlink_resource *tmp, *child_resource;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      list_for_each_entry_safe(child_resource, tmp, &devlink->resource_list,
 +                               list) {
 +              devlink_resource_unregister(devlink, child_resource);
 +              list_del(&child_resource->list);
 +              kfree(child_resource);
 +      }
 +}
 +EXPORT_SYMBOL_GPL(devl_resources_unregister);
 +
 +/**
 + *    devlink_resources_unregister - free all resources
 + *
 + *    @devlink: devlink
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_resources_unregister(struct devlink *devlink)
 +{
 +      devl_lock(devlink);
 +      devl_resources_unregister(devlink);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_resources_unregister);
 +
 +/**
 + * devl_resource_size_get - get and update size
 + *
 + * @devlink: devlink
 + * @resource_id: the requested resource id
 + * @p_resource_size: ptr to update
 + */
 +int devl_resource_size_get(struct devlink *devlink,
 +                         u64 resource_id,
 +                         u64 *p_resource_size)
 +{
 +      struct devlink_resource *resource;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      resource = devlink_resource_find(devlink, NULL, resource_id);
 +      if (!resource)
 +              return -EINVAL;
 +      *p_resource_size = resource->size_new;
 +      resource->size = resource->size_new;
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_resource_size_get);
 +
 +/**
 + * devl_dpipe_table_resource_set - set the resource id
 + *
 + * @devlink: devlink
 + * @table_name: table name
 + * @resource_id: resource id
 + * @resource_units: number of resource's units consumed per table's entry
 + */
 +int devl_dpipe_table_resource_set(struct devlink *devlink,
 +                                const char *table_name, u64 resource_id,
 +                                u64 resource_units)
 +{
 +      struct devlink_dpipe_table *table;
 +
 +      table = devlink_dpipe_table_find(&devlink->dpipe_table_list,
 +                                       table_name, devlink);
 +      if (!table)
 +              return -EINVAL;
 +
 +      table->resource_id = resource_id;
 +      table->resource_units = resource_units;
 +      table->resource_valid = true;
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_dpipe_table_resource_set);
 +
 +/**
 + * devl_resource_occ_get_register - register occupancy getter
 + *
 + * @devlink: devlink
 + * @resource_id: resource id
 + * @occ_get: occupancy getter callback
 + * @occ_get_priv: occupancy getter callback priv
 + */
 +void devl_resource_occ_get_register(struct devlink *devlink,
 +                                  u64 resource_id,
 +                                  devlink_resource_occ_get_t *occ_get,
 +                                  void *occ_get_priv)
 +{
 +      struct devlink_resource *resource;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      resource = devlink_resource_find(devlink, NULL, resource_id);
 +      if (WARN_ON(!resource))
 +              return;
 +      WARN_ON(resource->occ_get);
 +
 +      resource->occ_get = occ_get;
 +      resource->occ_get_priv = occ_get_priv;
 +}
 +EXPORT_SYMBOL_GPL(devl_resource_occ_get_register);
 +
 +/**
 + *    devlink_resource_occ_get_register - register occupancy getter
 + *
 + *    @devlink: devlink
 + *    @resource_id: resource id
 + *    @occ_get: occupancy getter callback
 + *    @occ_get_priv: occupancy getter callback priv
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_resource_occ_get_register(struct devlink *devlink,
 +                                     u64 resource_id,
 +                                     devlink_resource_occ_get_t *occ_get,
 +                                     void *occ_get_priv)
 +{
 +      devl_lock(devlink);
 +      devl_resource_occ_get_register(devlink, resource_id,
 +                                     occ_get, occ_get_priv);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_resource_occ_get_register);
 +
 +/**
 + * devl_resource_occ_get_unregister - unregister occupancy getter
 + *
 + * @devlink: devlink
 + * @resource_id: resource id
 + */
 +void devl_resource_occ_get_unregister(struct devlink *devlink,
 +                                    u64 resource_id)
 +{
 +      struct devlink_resource *resource;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      resource = devlink_resource_find(devlink, NULL, resource_id);
 +      if (WARN_ON(!resource))
 +              return;
 +      WARN_ON(!resource->occ_get);
 +
 +      resource->occ_get = NULL;
 +      resource->occ_get_priv = NULL;
 +}
 +EXPORT_SYMBOL_GPL(devl_resource_occ_get_unregister);
 +
 +/**
 + *    devlink_resource_occ_get_unregister - unregister occupancy getter
 + *
 + *    @devlink: devlink
 + *    @resource_id: resource id
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_resource_occ_get_unregister(struct devlink *devlink,
 +                                       u64 resource_id)
 +{
 +      devl_lock(devlink);
 +      devl_resource_occ_get_unregister(devlink, resource_id);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_resource_occ_get_unregister);
 +
 +static int devlink_param_verify(const struct devlink_param *param)
 +{
 +      if (!param || !param->name || !param->supported_cmodes)
 +              return -EINVAL;
 +      if (param->generic)
 +              return devlink_param_generic_verify(param);
 +      else
 +              return devlink_param_driver_verify(param);
 +}
 +
 +static int devlink_param_register(struct devlink *devlink,
 +                                const struct devlink_param *param)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      WARN_ON(devlink_param_verify(param));
 +      WARN_ON(devlink_param_find_by_name(&devlink->param_list, param->name));
 +
 +      if (param->supported_cmodes == BIT(DEVLINK_PARAM_CMODE_DRIVERINIT))
 +              WARN_ON(param->get || param->set);
 +      else
 +              WARN_ON(!param->get || !param->set);
 +
 +      param_item = kzalloc(sizeof(*param_item), GFP_KERNEL);
 +      if (!param_item)
 +              return -ENOMEM;
 +
 +      param_item->param = param;
 +
 +      list_add_tail(&param_item->list, &devlink->param_list);
 +      devlink_param_notify(devlink, 0, param_item, DEVLINK_CMD_PARAM_NEW);
 +      return 0;
 +}
 +
 +static void devlink_param_unregister(struct devlink *devlink,
 +                                   const struct devlink_param *param)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      param_item =
 +              devlink_param_find_by_name(&devlink->param_list, param->name);
 +      if (WARN_ON(!param_item))
 +              return;
 +      devlink_param_notify(devlink, 0, param_item, DEVLINK_CMD_PARAM_DEL);
 +      list_del(&param_item->list);
 +      kfree(param_item);
 +}
 +
 +/**
 + *    devl_params_register - register configuration parameters
 + *
 + *    @devlink: devlink
 + *    @params: configuration parameters array
 + *    @params_count: number of parameters provided
 + *
 + *    Register the configuration parameters supported by the driver.
 + */
 +int devl_params_register(struct devlink *devlink,
 +                       const struct devlink_param *params,
 +                       size_t params_count)
 +{
 +      const struct devlink_param *param = params;
 +      int i, err;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      for (i = 0; i < params_count; i++, param++) {
 +              err = devlink_param_register(devlink, param);
 +              if (err)
 +                      goto rollback;
 +      }
 +      return 0;
 +
 +rollback:
 +      if (!i)
 +              return err;
 +
 +      for (param--; i > 0; i--, param--)
 +              devlink_param_unregister(devlink, param);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devl_params_register);
 +
 +int devlink_params_register(struct devlink *devlink,
 +                          const struct devlink_param *params,
 +                          size_t params_count)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_params_register(devlink, params, params_count);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_params_register);
 +
 +/**
 + *    devl_params_unregister - unregister configuration parameters
 + *    @devlink: devlink
 + *    @params: configuration parameters to unregister
 + *    @params_count: number of parameters provided
 + */
 +void devl_params_unregister(struct devlink *devlink,
 +                          const struct devlink_param *params,
 +                          size_t params_count)
 +{
 +      const struct devlink_param *param = params;
 +      int i;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      for (i = 0; i < params_count; i++, param++)
 +              devlink_param_unregister(devlink, param);
 +}
 +EXPORT_SYMBOL_GPL(devl_params_unregister);
 +
 +void devlink_params_unregister(struct devlink *devlink,
 +                             const struct devlink_param *params,
 +                             size_t params_count)
 +{
 +      devl_lock(devlink);
 +      devl_params_unregister(devlink, params, params_count);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_params_unregister);
 +
 +/**
 + *    devl_param_driverinit_value_get - get configuration parameter
 + *                                      value for driver initializing
 + *
 + *    @devlink: devlink
 + *    @param_id: parameter ID
 + *    @init_val: value of parameter in driverinit configuration mode
 + *
 + *    This function should be used by the driver to get driverinit
 + *    configuration for initialization after reload command.
 + */
 +int devl_param_driverinit_value_get(struct devlink *devlink, u32 param_id,
 +                                  union devlink_param_value *init_val)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      lockdep_assert_held(&devlink->lock);
 +
 +      if (WARN_ON(!devlink_reload_supported(devlink->ops)))
 +              return -EOPNOTSUPP;
 +
 +      param_item = devlink_param_find_by_id(&devlink->param_list, param_id);
 +      if (!param_item)
 +              return -EINVAL;
 +
 +      if (!param_item->driverinit_value_valid)
 +              return -EOPNOTSUPP;
 +
 +      if (WARN_ON(!devlink_param_cmode_is_supported(param_item->param,
 +                                                    DEVLINK_PARAM_CMODE_DRIVERINIT)))
 +              return -EOPNOTSUPP;
 +
 +      if (param_item->param->type == DEVLINK_PARAM_TYPE_STRING)
 +              strcpy(init_val->vstr, param_item->driverinit_value.vstr);
 +      else
 +              *init_val = param_item->driverinit_value;
 +
 +      return 0;
 +}
 +EXPORT_SYMBOL_GPL(devl_param_driverinit_value_get);
 +
 +/**
 + *    devl_param_driverinit_value_set - set value of configuration
 + *                                      parameter for driverinit
 + *                                      configuration mode
 + *
 + *    @devlink: devlink
 + *    @param_id: parameter ID
 + *    @init_val: value of parameter to set for driverinit configuration mode
 + *
 + *    This function should be used by the driver to set driverinit
 + *    configuration mode default value.
 + */
 +void devl_param_driverinit_value_set(struct devlink *devlink, u32 param_id,
 +                                   union devlink_param_value init_val)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      param_item = devlink_param_find_by_id(&devlink->param_list, param_id);
 +      if (WARN_ON(!param_item))
 +              return;
 +
 +      if (WARN_ON(!devlink_param_cmode_is_supported(param_item->param,
 +                                                    DEVLINK_PARAM_CMODE_DRIVERINIT)))
 +              return;
 +
 +      if (param_item->param->type == DEVLINK_PARAM_TYPE_STRING)
 +              strcpy(param_item->driverinit_value.vstr, init_val.vstr);
 +      else
 +              param_item->driverinit_value = init_val;
 +      param_item->driverinit_value_valid = true;
 +
 +      devlink_param_notify(devlink, 0, param_item, DEVLINK_CMD_PARAM_NEW);
 +}
 +EXPORT_SYMBOL_GPL(devl_param_driverinit_value_set);
 +
 +/**
 + *    devl_param_value_changed - notify devlink on a parameter's value
 + *                               change. Should be called by the driver
 + *                               right after the change.
 + *
 + *    @devlink: devlink
 + *    @param_id: parameter ID
 + *
 + *    This function should be used by the driver to notify devlink on value
 + *    change, excluding driverinit configuration mode.
 + *    For driverinit configuration mode driver should use the function
 + */
 +void devl_param_value_changed(struct devlink *devlink, u32 param_id)
 +{
 +      struct devlink_param_item *param_item;
 +
 +      param_item = devlink_param_find_by_id(&devlink->param_list, param_id);
 +      WARN_ON(!param_item);
 +
 +      devlink_param_notify(devlink, 0, param_item, DEVLINK_CMD_PARAM_NEW);
 +}
 +EXPORT_SYMBOL_GPL(devl_param_value_changed);
 +
 +/**
 + * devl_region_create - create a new address region
 + *
 + * @devlink: devlink
 + * @ops: region operations and name
 + * @region_max_snapshots: Maximum supported number of snapshots for region
 + * @region_size: size of region
 + */
 +struct devlink_region *devl_region_create(struct devlink *devlink,
 +                                        const struct devlink_region_ops *ops,
 +                                        u32 region_max_snapshots,
 +                                        u64 region_size)
 +{
 +      struct devlink_region *region;
 +
 +      devl_assert_locked(devlink);
 +
 +      if (WARN_ON(!ops) || WARN_ON(!ops->destructor))
 +              return ERR_PTR(-EINVAL);
 +
 +      if (devlink_region_get_by_name(devlink, ops->name))
 +              return ERR_PTR(-EEXIST);
 +
 +      region = kzalloc(sizeof(*region), GFP_KERNEL);
 +      if (!region)
 +              return ERR_PTR(-ENOMEM);
 +
 +      region->devlink = devlink;
 +      region->max_snapshots = region_max_snapshots;
 +      region->ops = ops;
 +      region->size = region_size;
 +      INIT_LIST_HEAD(&region->snapshot_list);
 +      mutex_init(&region->snapshot_lock);
 +      list_add_tail(&region->list, &devlink->region_list);
 +      devlink_nl_region_notify(region, NULL, DEVLINK_CMD_REGION_NEW);
 +
 +      return region;
 +}
 +EXPORT_SYMBOL_GPL(devl_region_create);
 +
 +/**
 + *    devlink_region_create - create a new address region
 + *
 + *    @devlink: devlink
 + *    @ops: region operations and name
 + *    @region_max_snapshots: Maximum supported number of snapshots for region
 + *    @region_size: size of region
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +struct devlink_region *
 +devlink_region_create(struct devlink *devlink,
 +                    const struct devlink_region_ops *ops,
 +                    u32 region_max_snapshots, u64 region_size)
 +{
 +      struct devlink_region *region;
 +
 +      devl_lock(devlink);
 +      region = devl_region_create(devlink, ops, region_max_snapshots,
 +                                  region_size);
 +      devl_unlock(devlink);
 +      return region;
 +}
 +EXPORT_SYMBOL_GPL(devlink_region_create);
 +
 +/**
 + *    devlink_port_region_create - create a new address region for a port
 + *
 + *    @port: devlink port
 + *    @ops: region operations and name
 + *    @region_max_snapshots: Maximum supported number of snapshots for region
 + *    @region_size: size of region
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +struct devlink_region *
 +devlink_port_region_create(struct devlink_port *port,
 +                         const struct devlink_port_region_ops *ops,
 +                         u32 region_max_snapshots, u64 region_size)
 +{
 +      struct devlink *devlink = port->devlink;
 +      struct devlink_region *region;
 +      int err = 0;
 +
 +      ASSERT_DEVLINK_PORT_INITIALIZED(port);
 +
 +      if (WARN_ON(!ops) || WARN_ON(!ops->destructor))
 +              return ERR_PTR(-EINVAL);
 +
 +      devl_lock(devlink);
 +
 +      if (devlink_port_region_get_by_name(port, ops->name)) {
 +              err = -EEXIST;
 +              goto unlock;
 +      }
 +
 +      region = kzalloc(sizeof(*region), GFP_KERNEL);
 +      if (!region) {
 +              err = -ENOMEM;
 +              goto unlock;
 +      }
 +
 +      region->devlink = devlink;
 +      region->port = port;
 +      region->max_snapshots = region_max_snapshots;
 +      region->port_ops = ops;
 +      region->size = region_size;
 +      INIT_LIST_HEAD(&region->snapshot_list);
 +      mutex_init(&region->snapshot_lock);
 +      list_add_tail(&region->list, &port->region_list);
 +      devlink_nl_region_notify(region, NULL, DEVLINK_CMD_REGION_NEW);
 +
 +      devl_unlock(devlink);
 +      return region;
 +
 +unlock:
 +      devl_unlock(devlink);
 +      return ERR_PTR(err);
 +}
 +EXPORT_SYMBOL_GPL(devlink_port_region_create);
 +
 +/**
 + * devl_region_destroy - destroy address region
 + *
 + * @region: devlink region to destroy
 + */
 +void devl_region_destroy(struct devlink_region *region)
 +{
 +      struct devlink *devlink = region->devlink;
 +      struct devlink_snapshot *snapshot, *ts;
 +
 +      devl_assert_locked(devlink);
 +
 +      /* Free all snapshots of region */
 +      mutex_lock(&region->snapshot_lock);
 +      list_for_each_entry_safe(snapshot, ts, &region->snapshot_list, list)
 +              devlink_region_snapshot_del(region, snapshot);
 +      mutex_unlock(&region->snapshot_lock);
 +
 +      list_del(&region->list);
 +      mutex_destroy(&region->snapshot_lock);
 +
 +      devlink_nl_region_notify(region, NULL, DEVLINK_CMD_REGION_DEL);
 +      kfree(region);
 +}
 +EXPORT_SYMBOL_GPL(devl_region_destroy);
 +
 +/**
 + *    devlink_region_destroy - destroy address region
 + *
 + *    @region: devlink region to destroy
 + *
 + *    Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_region_destroy(struct devlink_region *region)
 +{
 +      struct devlink *devlink = region->devlink;
 +
 +      devl_lock(devlink);
 +      devl_region_destroy(region);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_region_destroy);
 +
 +/**
 + *    devlink_region_snapshot_id_get - get snapshot ID
 + *
 + *    This callback should be called when adding a new snapshot,
 + *    Driver should use the same id for multiple snapshots taken
 + *    on multiple regions at the same time/by the same trigger.
 + *
 + *    The caller of this function must use devlink_region_snapshot_id_put
 + *    when finished creating regions using this id.
 + *
 + *    Returns zero on success, or a negative error code on failure.
 + *
 + *    @devlink: devlink
 + *    @id: storage to return id
 + */
 +int devlink_region_snapshot_id_get(struct devlink *devlink, u32 *id)
 +{
 +      return __devlink_region_snapshot_id_get(devlink, id);
 +}
 +EXPORT_SYMBOL_GPL(devlink_region_snapshot_id_get);
 +
 +/**
 + *    devlink_region_snapshot_id_put - put snapshot ID reference
 + *
 + *    This should be called by a driver after finishing creating snapshots
 + *    with an id. Doing so ensures that the ID can later be released in the
 + *    event that all snapshots using it have been destroyed.
 + *
 + *    @devlink: devlink
 + *    @id: id to release reference on
 + */
 +void devlink_region_snapshot_id_put(struct devlink *devlink, u32 id)
 +{
 +      __devlink_snapshot_id_decrement(devlink, id);
 +}
 +EXPORT_SYMBOL_GPL(devlink_region_snapshot_id_put);
 +
 +/**
 + *    devlink_region_snapshot_create - create a new snapshot
 + *    This will add a new snapshot of a region. The snapshot
 + *    will be stored on the region struct and can be accessed
 + *    from devlink. This is useful for future analyses of snapshots.
 + *    Multiple snapshots can be created on a region.
 + *    The @snapshot_id should be obtained using the getter function.
 + *
 + *    @region: devlink region of the snapshot
 + *    @data: snapshot data
 + *    @snapshot_id: snapshot id to be created
 + */
 +int devlink_region_snapshot_create(struct devlink_region *region,
 +                                 u8 *data, u32 snapshot_id)
 +{
 +      int err;
 +
 +      mutex_lock(&region->snapshot_lock);
 +      err = __devlink_region_snapshot_create(region, data, snapshot_id);
 +      mutex_unlock(&region->snapshot_lock);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_region_snapshot_create);
 +
 +#define DEVLINK_TRAP(_id, _type)                                            \
 +      {                                                                     \
 +              .type = DEVLINK_TRAP_TYPE_##_type,                            \
 +              .id = DEVLINK_TRAP_GENERIC_ID_##_id,                          \
 +              .name = DEVLINK_TRAP_GENERIC_NAME_##_id,                      \
 +      }
 +
 +static const struct devlink_trap devlink_trap_generic[] = {
 +      DEVLINK_TRAP(SMAC_MC, DROP),
 +      DEVLINK_TRAP(VLAN_TAG_MISMATCH, DROP),
 +      DEVLINK_TRAP(INGRESS_VLAN_FILTER, DROP),
 +      DEVLINK_TRAP(INGRESS_STP_FILTER, DROP),
 +      DEVLINK_TRAP(EMPTY_TX_LIST, DROP),
 +      DEVLINK_TRAP(PORT_LOOPBACK_FILTER, DROP),
 +      DEVLINK_TRAP(BLACKHOLE_ROUTE, DROP),
 +      DEVLINK_TRAP(TTL_ERROR, EXCEPTION),
 +      DEVLINK_TRAP(TAIL_DROP, DROP),
 +      DEVLINK_TRAP(NON_IP_PACKET, DROP),
 +      DEVLINK_TRAP(UC_DIP_MC_DMAC, DROP),
 +      DEVLINK_TRAP(DIP_LB, DROP),
 +      DEVLINK_TRAP(SIP_MC, DROP),
 +      DEVLINK_TRAP(SIP_LB, DROP),
 +      DEVLINK_TRAP(CORRUPTED_IP_HDR, DROP),
 +      DEVLINK_TRAP(IPV4_SIP_BC, DROP),
 +      DEVLINK_TRAP(IPV6_MC_DIP_RESERVED_SCOPE, DROP),
 +      DEVLINK_TRAP(IPV6_MC_DIP_INTERFACE_LOCAL_SCOPE, DROP),
 +      DEVLINK_TRAP(MTU_ERROR, EXCEPTION),
 +      DEVLINK_TRAP(UNRESOLVED_NEIGH, EXCEPTION),
 +      DEVLINK_TRAP(RPF, EXCEPTION),
 +      DEVLINK_TRAP(REJECT_ROUTE, EXCEPTION),
 +      DEVLINK_TRAP(IPV4_LPM_UNICAST_MISS, EXCEPTION),
 +      DEVLINK_TRAP(IPV6_LPM_UNICAST_MISS, EXCEPTION),
 +      DEVLINK_TRAP(NON_ROUTABLE, DROP),
 +      DEVLINK_TRAP(DECAP_ERROR, EXCEPTION),
 +      DEVLINK_TRAP(OVERLAY_SMAC_MC, DROP),
 +      DEVLINK_TRAP(INGRESS_FLOW_ACTION_DROP, DROP),
 +      DEVLINK_TRAP(EGRESS_FLOW_ACTION_DROP, DROP),
 +      DEVLINK_TRAP(STP, CONTROL),
 +      DEVLINK_TRAP(LACP, CONTROL),
 +      DEVLINK_TRAP(LLDP, CONTROL),
 +      DEVLINK_TRAP(IGMP_QUERY, CONTROL),
 +      DEVLINK_TRAP(IGMP_V1_REPORT, CONTROL),
 +      DEVLINK_TRAP(IGMP_V2_REPORT, CONTROL),
 +      DEVLINK_TRAP(IGMP_V3_REPORT, CONTROL),
 +      DEVLINK_TRAP(IGMP_V2_LEAVE, CONTROL),
 +      DEVLINK_TRAP(MLD_QUERY, CONTROL),
 +      DEVLINK_TRAP(MLD_V1_REPORT, CONTROL),
 +      DEVLINK_TRAP(MLD_V2_REPORT, CONTROL),
 +      DEVLINK_TRAP(MLD_V1_DONE, CONTROL),
 +      DEVLINK_TRAP(IPV4_DHCP, CONTROL),
 +      DEVLINK_TRAP(IPV6_DHCP, CONTROL),
 +      DEVLINK_TRAP(ARP_REQUEST, CONTROL),
 +      DEVLINK_TRAP(ARP_RESPONSE, CONTROL),
 +      DEVLINK_TRAP(ARP_OVERLAY, CONTROL),
 +      DEVLINK_TRAP(IPV6_NEIGH_SOLICIT, CONTROL),
 +      DEVLINK_TRAP(IPV6_NEIGH_ADVERT, CONTROL),
 +      DEVLINK_TRAP(IPV4_BFD, CONTROL),
 +      DEVLINK_TRAP(IPV6_BFD, CONTROL),
 +      DEVLINK_TRAP(IPV4_OSPF, CONTROL),
 +      DEVLINK_TRAP(IPV6_OSPF, CONTROL),
 +      DEVLINK_TRAP(IPV4_BGP, CONTROL),
 +      DEVLINK_TRAP(IPV6_BGP, CONTROL),
 +      DEVLINK_TRAP(IPV4_VRRP, CONTROL),
 +      DEVLINK_TRAP(IPV6_VRRP, CONTROL),
 +      DEVLINK_TRAP(IPV4_PIM, CONTROL),
 +      DEVLINK_TRAP(IPV6_PIM, CONTROL),
 +      DEVLINK_TRAP(UC_LB, CONTROL),
 +      DEVLINK_TRAP(LOCAL_ROUTE, CONTROL),
 +      DEVLINK_TRAP(EXTERNAL_ROUTE, CONTROL),
 +      DEVLINK_TRAP(IPV6_UC_DIP_LINK_LOCAL_SCOPE, CONTROL),
 +      DEVLINK_TRAP(IPV6_DIP_ALL_NODES, CONTROL),
 +      DEVLINK_TRAP(IPV6_DIP_ALL_ROUTERS, CONTROL),
 +      DEVLINK_TRAP(IPV6_ROUTER_SOLICIT, CONTROL),
 +      DEVLINK_TRAP(IPV6_ROUTER_ADVERT, CONTROL),
 +      DEVLINK_TRAP(IPV6_REDIRECT, CONTROL),
 +      DEVLINK_TRAP(IPV4_ROUTER_ALERT, CONTROL),
 +      DEVLINK_TRAP(IPV6_ROUTER_ALERT, CONTROL),
 +      DEVLINK_TRAP(PTP_EVENT, CONTROL),
 +      DEVLINK_TRAP(PTP_GENERAL, CONTROL),
 +      DEVLINK_TRAP(FLOW_ACTION_SAMPLE, CONTROL),
 +      DEVLINK_TRAP(FLOW_ACTION_TRAP, CONTROL),
 +      DEVLINK_TRAP(EARLY_DROP, DROP),
 +      DEVLINK_TRAP(VXLAN_PARSING, DROP),
 +      DEVLINK_TRAP(LLC_SNAP_PARSING, DROP),
 +      DEVLINK_TRAP(VLAN_PARSING, DROP),
 +      DEVLINK_TRAP(PPPOE_PPP_PARSING, DROP),
 +      DEVLINK_TRAP(MPLS_PARSING, DROP),
 +      DEVLINK_TRAP(ARP_PARSING, DROP),
 +      DEVLINK_TRAP(IP_1_PARSING, DROP),
 +      DEVLINK_TRAP(IP_N_PARSING, DROP),
 +      DEVLINK_TRAP(GRE_PARSING, DROP),
 +      DEVLINK_TRAP(UDP_PARSING, DROP),
 +      DEVLINK_TRAP(TCP_PARSING, DROP),
 +      DEVLINK_TRAP(IPSEC_PARSING, DROP),
 +      DEVLINK_TRAP(SCTP_PARSING, DROP),
 +      DEVLINK_TRAP(DCCP_PARSING, DROP),
 +      DEVLINK_TRAP(GTP_PARSING, DROP),
 +      DEVLINK_TRAP(ESP_PARSING, DROP),
 +      DEVLINK_TRAP(BLACKHOLE_NEXTHOP, DROP),
 +      DEVLINK_TRAP(DMAC_FILTER, DROP),
 +      DEVLINK_TRAP(EAPOL, CONTROL),
 +      DEVLINK_TRAP(LOCKED_PORT, DROP),
 +};
 +
 +#define DEVLINK_TRAP_GROUP(_id)                                                     \
 +      {                                                                     \
 +              .id = DEVLINK_TRAP_GROUP_GENERIC_ID_##_id,                    \
 +              .name = DEVLINK_TRAP_GROUP_GENERIC_NAME_##_id,                \
 +      }
 +
 +static const struct devlink_trap_group devlink_trap_group_generic[] = {
 +      DEVLINK_TRAP_GROUP(L2_DROPS),
 +      DEVLINK_TRAP_GROUP(L3_DROPS),
 +      DEVLINK_TRAP_GROUP(L3_EXCEPTIONS),
 +      DEVLINK_TRAP_GROUP(BUFFER_DROPS),
 +      DEVLINK_TRAP_GROUP(TUNNEL_DROPS),
 +      DEVLINK_TRAP_GROUP(ACL_DROPS),
 +      DEVLINK_TRAP_GROUP(STP),
 +      DEVLINK_TRAP_GROUP(LACP),
 +      DEVLINK_TRAP_GROUP(LLDP),
 +      DEVLINK_TRAP_GROUP(MC_SNOOPING),
 +      DEVLINK_TRAP_GROUP(DHCP),
 +      DEVLINK_TRAP_GROUP(NEIGH_DISCOVERY),
 +      DEVLINK_TRAP_GROUP(BFD),
 +      DEVLINK_TRAP_GROUP(OSPF),
 +      DEVLINK_TRAP_GROUP(BGP),
 +      DEVLINK_TRAP_GROUP(VRRP),
 +      DEVLINK_TRAP_GROUP(PIM),
 +      DEVLINK_TRAP_GROUP(UC_LB),
 +      DEVLINK_TRAP_GROUP(LOCAL_DELIVERY),
 +      DEVLINK_TRAP_GROUP(EXTERNAL_DELIVERY),
 +      DEVLINK_TRAP_GROUP(IPV6),
 +      DEVLINK_TRAP_GROUP(PTP_EVENT),
 +      DEVLINK_TRAP_GROUP(PTP_GENERAL),
 +      DEVLINK_TRAP_GROUP(ACL_SAMPLE),
 +      DEVLINK_TRAP_GROUP(ACL_TRAP),
 +      DEVLINK_TRAP_GROUP(PARSER_ERROR_DROPS),
 +      DEVLINK_TRAP_GROUP(EAPOL),
 +};
 +
 +static int devlink_trap_generic_verify(const struct devlink_trap *trap)
 +{
 +      if (trap->id > DEVLINK_TRAP_GENERIC_ID_MAX)
 +              return -EINVAL;
 +
 +      if (strcmp(trap->name, devlink_trap_generic[trap->id].name))
 +              return -EINVAL;
 +
 +      if (trap->type != devlink_trap_generic[trap->id].type)
 +              return -EINVAL;
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_driver_verify(const struct devlink_trap *trap)
 +{
 +      int i;
 +
 +      if (trap->id <= DEVLINK_TRAP_GENERIC_ID_MAX)
 +              return -EINVAL;
 +
 +      for (i = 0; i < ARRAY_SIZE(devlink_trap_generic); i++) {
 +              if (!strcmp(trap->name, devlink_trap_generic[i].name))
 +                      return -EEXIST;
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_verify(const struct devlink_trap *trap)
 +{
 +      if (!trap || !trap->name)
 +              return -EINVAL;
 +
 +      if (trap->generic)
 +              return devlink_trap_generic_verify(trap);
 +      else
 +              return devlink_trap_driver_verify(trap);
 +}
 +
 +static int
 +devlink_trap_group_generic_verify(const struct devlink_trap_group *group)
 +{
 +      if (group->id > DEVLINK_TRAP_GROUP_GENERIC_ID_MAX)
 +              return -EINVAL;
 +
 +      if (strcmp(group->name, devlink_trap_group_generic[group->id].name))
 +              return -EINVAL;
 +
 +      return 0;
 +}
 +
 +static int
 +devlink_trap_group_driver_verify(const struct devlink_trap_group *group)
 +{
 +      int i;
 +
 +      if (group->id <= DEVLINK_TRAP_GROUP_GENERIC_ID_MAX)
 +              return -EINVAL;
 +
 +      for (i = 0; i < ARRAY_SIZE(devlink_trap_group_generic); i++) {
 +              if (!strcmp(group->name, devlink_trap_group_generic[i].name))
 +                      return -EEXIST;
 +      }
 +
 +      return 0;
 +}
 +
 +static int devlink_trap_group_verify(const struct devlink_trap_group *group)
 +{
 +      if (group->generic)
 +              return devlink_trap_group_generic_verify(group);
 +      else
 +              return devlink_trap_group_driver_verify(group);
 +}
 +
 +static void
 +devlink_trap_group_notify(struct devlink *devlink,
 +                        const struct devlink_trap_group_item *group_item,
 +                        enum devlink_command cmd)
 +{
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON_ONCE(cmd != DEVLINK_CMD_TRAP_GROUP_NEW &&
 +                   cmd != DEVLINK_CMD_TRAP_GROUP_DEL);
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_trap_group_fill(msg, devlink, group_item, cmd, 0, 0,
 +                                       0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink),
 +                              msg, 0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int
 +devlink_trap_item_group_link(struct devlink *devlink,
 +                           struct devlink_trap_item *trap_item)
 +{
 +      u16 group_id = trap_item->trap->init_group_id;
 +      struct devlink_trap_group_item *group_item;
 +
 +      group_item = devlink_trap_group_item_lookup_by_id(devlink, group_id);
 +      if (WARN_ON_ONCE(!group_item))
 +              return -EINVAL;
 +
 +      trap_item->group_item = group_item;
 +
 +      return 0;
 +}
 +
 +static void devlink_trap_notify(struct devlink *devlink,
 +                              const struct devlink_trap_item *trap_item,
 +                              enum devlink_command cmd)
 +{
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON_ONCE(cmd != DEVLINK_CMD_TRAP_NEW &&
 +                   cmd != DEVLINK_CMD_TRAP_DEL);
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_trap_fill(msg, devlink, trap_item, cmd, 0, 0, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink),
 +                              msg, 0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int
 +devlink_trap_register(struct devlink *devlink,
 +                    const struct devlink_trap *trap, void *priv)
 +{
 +      struct devlink_trap_item *trap_item;
 +      int err;
 +
 +      if (devlink_trap_item_lookup(devlink, trap->name))
 +              return -EEXIST;
 +
 +      trap_item = kzalloc(sizeof(*trap_item), GFP_KERNEL);
 +      if (!trap_item)
 +              return -ENOMEM;
 +
 +      trap_item->stats = netdev_alloc_pcpu_stats(struct devlink_stats);
 +      if (!trap_item->stats) {
 +              err = -ENOMEM;
 +              goto err_stats_alloc;
 +      }
 +
 +      trap_item->trap = trap;
 +      trap_item->action = trap->init_action;
 +      trap_item->priv = priv;
 +
 +      err = devlink_trap_item_group_link(devlink, trap_item);
 +      if (err)
 +              goto err_group_link;
 +
 +      err = devlink->ops->trap_init(devlink, trap, trap_item);
 +      if (err)
 +              goto err_trap_init;
 +
 +      list_add_tail(&trap_item->list, &devlink->trap_list);
 +      devlink_trap_notify(devlink, trap_item, DEVLINK_CMD_TRAP_NEW);
 +
 +      return 0;
 +
 +err_trap_init:
 +err_group_link:
 +      free_percpu(trap_item->stats);
 +err_stats_alloc:
 +      kfree(trap_item);
 +      return err;
 +}
 +
 +static void devlink_trap_unregister(struct devlink *devlink,
 +                                  const struct devlink_trap *trap)
 +{
 +      struct devlink_trap_item *trap_item;
 +
 +      trap_item = devlink_trap_item_lookup(devlink, trap->name);
 +      if (WARN_ON_ONCE(!trap_item))
 +              return;
 +
 +      devlink_trap_notify(devlink, trap_item, DEVLINK_CMD_TRAP_DEL);
 +      list_del(&trap_item->list);
 +      if (devlink->ops->trap_fini)
 +              devlink->ops->trap_fini(devlink, trap, trap_item);
 +      free_percpu(trap_item->stats);
 +      kfree(trap_item);
 +}
 +
 +static void devlink_trap_disable(struct devlink *devlink,
 +                               const struct devlink_trap *trap)
 +{
 +      struct devlink_trap_item *trap_item;
 +
 +      trap_item = devlink_trap_item_lookup(devlink, trap->name);
 +      if (WARN_ON_ONCE(!trap_item))
 +              return;
 +
 +      devlink->ops->trap_action_set(devlink, trap, DEVLINK_TRAP_ACTION_DROP,
 +                                    NULL);
 +      trap_item->action = DEVLINK_TRAP_ACTION_DROP;
 +}
 +
 +/**
 + * devl_traps_register - Register packet traps with devlink.
 + * @devlink: devlink.
 + * @traps: Packet traps.
 + * @traps_count: Count of provided packet traps.
 + * @priv: Driver private information.
 + *
 + * Return: Non-zero value on failure.
 + */
 +int devl_traps_register(struct devlink *devlink,
 +                      const struct devlink_trap *traps,
 +                      size_t traps_count, void *priv)
 +{
 +      int i, err;
 +
 +      if (!devlink->ops->trap_init || !devlink->ops->trap_action_set)
 +              return -EINVAL;
 +
 +      devl_assert_locked(devlink);
 +      for (i = 0; i < traps_count; i++) {
 +              const struct devlink_trap *trap = &traps[i];
 +
 +              err = devlink_trap_verify(trap);
 +              if (err)
 +                      goto err_trap_verify;
 +
 +              err = devlink_trap_register(devlink, trap, priv);
 +              if (err)
 +                      goto err_trap_register;
 +      }
 +
 +      return 0;
 +
 +err_trap_register:
 +err_trap_verify:
 +      for (i--; i >= 0; i--)
 +              devlink_trap_unregister(devlink, &traps[i]);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devl_traps_register);
 +
 +/**
 + * devlink_traps_register - Register packet traps with devlink.
 + * @devlink: devlink.
 + * @traps: Packet traps.
 + * @traps_count: Count of provided packet traps.
 + * @priv: Driver private information.
 + *
 + * Context: Takes and release devlink->lock <mutex>.
 + *
 + * Return: Non-zero value on failure.
 + */
 +int devlink_traps_register(struct devlink *devlink,
 +                         const struct devlink_trap *traps,
 +                         size_t traps_count, void *priv)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_traps_register(devlink, traps, traps_count, priv);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_traps_register);
 +
 +/**
 + * devl_traps_unregister - Unregister packet traps from devlink.
 + * @devlink: devlink.
 + * @traps: Packet traps.
 + * @traps_count: Count of provided packet traps.
 + */
 +void devl_traps_unregister(struct devlink *devlink,
 +                         const struct devlink_trap *traps,
 +                         size_t traps_count)
 +{
 +      int i;
 +
 +      devl_assert_locked(devlink);
 +      /* Make sure we do not have any packets in-flight while unregistering
 +       * traps by disabling all of them and waiting for a grace period.
 +       */
 +      for (i = traps_count - 1; i >= 0; i--)
 +              devlink_trap_disable(devlink, &traps[i]);
 +      synchronize_rcu();
 +      for (i = traps_count - 1; i >= 0; i--)
 +              devlink_trap_unregister(devlink, &traps[i]);
 +}
 +EXPORT_SYMBOL_GPL(devl_traps_unregister);
 +
 +/**
 + * devlink_traps_unregister - Unregister packet traps from devlink.
 + * @devlink: devlink.
 + * @traps: Packet traps.
 + * @traps_count: Count of provided packet traps.
 + *
 + * Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_traps_unregister(struct devlink *devlink,
 +                            const struct devlink_trap *traps,
 +                            size_t traps_count)
 +{
 +      devl_lock(devlink);
 +      devl_traps_unregister(devlink, traps, traps_count);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_traps_unregister);
 +
 +static void
 +devlink_trap_stats_update(struct devlink_stats __percpu *trap_stats,
 +                        size_t skb_len)
 +{
 +      struct devlink_stats *stats;
 +
 +      stats = this_cpu_ptr(trap_stats);
 +      u64_stats_update_begin(&stats->syncp);
 +      u64_stats_add(&stats->rx_bytes, skb_len);
 +      u64_stats_inc(&stats->rx_packets);
 +      u64_stats_update_end(&stats->syncp);
 +}
 +
 +static void
 +devlink_trap_report_metadata_set(struct devlink_trap_metadata *metadata,
 +                               const struct devlink_trap_item *trap_item,
 +                               struct devlink_port *in_devlink_port,
 +                               const struct flow_action_cookie *fa_cookie)
 +{
 +      metadata->trap_name = trap_item->trap->name;
 +      metadata->trap_group_name = trap_item->group_item->group->name;
 +      metadata->fa_cookie = fa_cookie;
 +      metadata->trap_type = trap_item->trap->type;
 +
 +      spin_lock(&in_devlink_port->type_lock);
 +      if (in_devlink_port->type == DEVLINK_PORT_TYPE_ETH)
 +              metadata->input_dev = in_devlink_port->type_eth.netdev;
 +      spin_unlock(&in_devlink_port->type_lock);
 +}
 +
 +/**
 + * devlink_trap_report - Report trapped packet to drop monitor.
 + * @devlink: devlink.
 + * @skb: Trapped packet.
 + * @trap_ctx: Trap context.
 + * @in_devlink_port: Input devlink port.
 + * @fa_cookie: Flow action cookie. Could be NULL.
 + */
 +void devlink_trap_report(struct devlink *devlink, struct sk_buff *skb,
 +                       void *trap_ctx, struct devlink_port *in_devlink_port,
 +                       const struct flow_action_cookie *fa_cookie)
 +
 +{
 +      struct devlink_trap_item *trap_item = trap_ctx;
 +
 +      devlink_trap_stats_update(trap_item->stats, skb->len);
 +      devlink_trap_stats_update(trap_item->group_item->stats, skb->len);
 +
 +      if (trace_devlink_trap_report_enabled()) {
 +              struct devlink_trap_metadata metadata = {};
 +
 +              devlink_trap_report_metadata_set(&metadata, trap_item,
 +                                               in_devlink_port, fa_cookie);
 +              trace_devlink_trap_report(devlink, skb, &metadata);
 +      }
 +}
 +EXPORT_SYMBOL_GPL(devlink_trap_report);
 +
 +/**
 + * devlink_trap_ctx_priv - Trap context to driver private information.
 + * @trap_ctx: Trap context.
 + *
 + * Return: Driver private information passed during registration.
 + */
 +void *devlink_trap_ctx_priv(void *trap_ctx)
 +{
 +      struct devlink_trap_item *trap_item = trap_ctx;
 +
 +      return trap_item->priv;
 +}
 +EXPORT_SYMBOL_GPL(devlink_trap_ctx_priv);
 +
 +static int
 +devlink_trap_group_item_policer_link(struct devlink *devlink,
 +                                   struct devlink_trap_group_item *group_item)
 +{
 +      u32 policer_id = group_item->group->init_policer_id;
 +      struct devlink_trap_policer_item *policer_item;
 +
 +      if (policer_id == 0)
 +              return 0;
 +
 +      policer_item = devlink_trap_policer_item_lookup(devlink, policer_id);
 +      if (WARN_ON_ONCE(!policer_item))
 +              return -EINVAL;
 +
 +      group_item->policer_item = policer_item;
 +
 +      return 0;
 +}
 +
 +static int
 +devlink_trap_group_register(struct devlink *devlink,
 +                          const struct devlink_trap_group *group)
 +{
 +      struct devlink_trap_group_item *group_item;
 +      int err;
 +
 +      if (devlink_trap_group_item_lookup(devlink, group->name))
 +              return -EEXIST;
 +
 +      group_item = kzalloc(sizeof(*group_item), GFP_KERNEL);
 +      if (!group_item)
 +              return -ENOMEM;
 +
 +      group_item->stats = netdev_alloc_pcpu_stats(struct devlink_stats);
 +      if (!group_item->stats) {
 +              err = -ENOMEM;
 +              goto err_stats_alloc;
 +      }
 +
 +      group_item->group = group;
 +
 +      err = devlink_trap_group_item_policer_link(devlink, group_item);
 +      if (err)
 +              goto err_policer_link;
 +
 +      if (devlink->ops->trap_group_init) {
 +              err = devlink->ops->trap_group_init(devlink, group);
 +              if (err)
 +                      goto err_group_init;
 +      }
 +
 +      list_add_tail(&group_item->list, &devlink->trap_group_list);
 +      devlink_trap_group_notify(devlink, group_item,
 +                                DEVLINK_CMD_TRAP_GROUP_NEW);
 +
 +      return 0;
 +
 +err_group_init:
 +err_policer_link:
 +      free_percpu(group_item->stats);
 +err_stats_alloc:
 +      kfree(group_item);
 +      return err;
 +}
 +
 +static void
 +devlink_trap_group_unregister(struct devlink *devlink,
 +                            const struct devlink_trap_group *group)
 +{
 +      struct devlink_trap_group_item *group_item;
 +
 +      group_item = devlink_trap_group_item_lookup(devlink, group->name);
 +      if (WARN_ON_ONCE(!group_item))
 +              return;
 +
 +      devlink_trap_group_notify(devlink, group_item,
 +                                DEVLINK_CMD_TRAP_GROUP_DEL);
 +      list_del(&group_item->list);
 +      free_percpu(group_item->stats);
 +      kfree(group_item);
 +}
 +
 +/**
 + * devl_trap_groups_register - Register packet trap groups with devlink.
 + * @devlink: devlink.
 + * @groups: Packet trap groups.
 + * @groups_count: Count of provided packet trap groups.
 + *
 + * Return: Non-zero value on failure.
 + */
 +int devl_trap_groups_register(struct devlink *devlink,
 +                            const struct devlink_trap_group *groups,
 +                            size_t groups_count)
 +{
 +      int i, err;
 +
 +      devl_assert_locked(devlink);
 +      for (i = 0; i < groups_count; i++) {
 +              const struct devlink_trap_group *group = &groups[i];
 +
 +              err = devlink_trap_group_verify(group);
 +              if (err)
 +                      goto err_trap_group_verify;
 +
 +              err = devlink_trap_group_register(devlink, group);
 +              if (err)
 +                      goto err_trap_group_register;
 +      }
 +
 +      return 0;
 +
 +err_trap_group_register:
 +err_trap_group_verify:
 +      for (i--; i >= 0; i--)
 +              devlink_trap_group_unregister(devlink, &groups[i]);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devl_trap_groups_register);
 +
 +/**
 + * devlink_trap_groups_register - Register packet trap groups with devlink.
 + * @devlink: devlink.
 + * @groups: Packet trap groups.
 + * @groups_count: Count of provided packet trap groups.
 + *
 + * Context: Takes and release devlink->lock <mutex>.
 + *
 + * Return: Non-zero value on failure.
 + */
 +int devlink_trap_groups_register(struct devlink *devlink,
 +                               const struct devlink_trap_group *groups,
 +                               size_t groups_count)
 +{
 +      int err;
 +
 +      devl_lock(devlink);
 +      err = devl_trap_groups_register(devlink, groups, groups_count);
 +      devl_unlock(devlink);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devlink_trap_groups_register);
 +
 +/**
 + * devl_trap_groups_unregister - Unregister packet trap groups from devlink.
 + * @devlink: devlink.
 + * @groups: Packet trap groups.
 + * @groups_count: Count of provided packet trap groups.
 + */
 +void devl_trap_groups_unregister(struct devlink *devlink,
 +                               const struct devlink_trap_group *groups,
 +                               size_t groups_count)
 +{
 +      int i;
 +
 +      devl_assert_locked(devlink);
 +      for (i = groups_count - 1; i >= 0; i--)
 +              devlink_trap_group_unregister(devlink, &groups[i]);
 +}
 +EXPORT_SYMBOL_GPL(devl_trap_groups_unregister);
 +
 +/**
 + * devlink_trap_groups_unregister - Unregister packet trap groups from devlink.
 + * @devlink: devlink.
 + * @groups: Packet trap groups.
 + * @groups_count: Count of provided packet trap groups.
 + *
 + * Context: Takes and release devlink->lock <mutex>.
 + */
 +void devlink_trap_groups_unregister(struct devlink *devlink,
 +                                  const struct devlink_trap_group *groups,
 +                                  size_t groups_count)
 +{
 +      devl_lock(devlink);
 +      devl_trap_groups_unregister(devlink, groups, groups_count);
 +      devl_unlock(devlink);
 +}
 +EXPORT_SYMBOL_GPL(devlink_trap_groups_unregister);
 +
 +static void
 +devlink_trap_policer_notify(struct devlink *devlink,
 +                          const struct devlink_trap_policer_item *policer_item,
 +                          enum devlink_command cmd)
 +{
 +      struct sk_buff *msg;
 +      int err;
 +
 +      WARN_ON_ONCE(cmd != DEVLINK_CMD_TRAP_POLICER_NEW &&
 +                   cmd != DEVLINK_CMD_TRAP_POLICER_DEL);
 +      if (!xa_get_mark(&devlinks, devlink->index, DEVLINK_REGISTERED))
 +              return;
 +
 +      msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
 +      if (!msg)
 +              return;
 +
 +      err = devlink_nl_trap_policer_fill(msg, devlink, policer_item, cmd, 0,
 +                                         0, 0);
 +      if (err) {
 +              nlmsg_free(msg);
 +              return;
 +      }
 +
 +      genlmsg_multicast_netns(&devlink_nl_family, devlink_net(devlink),
 +                              msg, 0, DEVLINK_MCGRP_CONFIG, GFP_KERNEL);
 +}
 +
 +static int
 +devlink_trap_policer_register(struct devlink *devlink,
 +                            const struct devlink_trap_policer *policer)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +      int err;
 +
 +      if (devlink_trap_policer_item_lookup(devlink, policer->id))
 +              return -EEXIST;
 +
 +      policer_item = kzalloc(sizeof(*policer_item), GFP_KERNEL);
 +      if (!policer_item)
 +              return -ENOMEM;
 +
 +      policer_item->policer = policer;
 +      policer_item->rate = policer->init_rate;
 +      policer_item->burst = policer->init_burst;
 +
 +      if (devlink->ops->trap_policer_init) {
 +              err = devlink->ops->trap_policer_init(devlink, policer);
 +              if (err)
 +                      goto err_policer_init;
 +      }
 +
 +      list_add_tail(&policer_item->list, &devlink->trap_policer_list);
 +      devlink_trap_policer_notify(devlink, policer_item,
 +                                  DEVLINK_CMD_TRAP_POLICER_NEW);
 +
 +      return 0;
 +
 +err_policer_init:
 +      kfree(policer_item);
 +      return err;
 +}
 +
 +static void
 +devlink_trap_policer_unregister(struct devlink *devlink,
 +                              const struct devlink_trap_policer *policer)
 +{
 +      struct devlink_trap_policer_item *policer_item;
 +
 +      policer_item = devlink_trap_policer_item_lookup(devlink, policer->id);
 +      if (WARN_ON_ONCE(!policer_item))
 +              return;
 +
 +      devlink_trap_policer_notify(devlink, policer_item,
 +                                  DEVLINK_CMD_TRAP_POLICER_DEL);
 +      list_del(&policer_item->list);
 +      if (devlink->ops->trap_policer_fini)
 +              devlink->ops->trap_policer_fini(devlink, policer);
 +      kfree(policer_item);
 +}
 +
 +/**
 + * devl_trap_policers_register - Register packet trap policers with devlink.
 + * @devlink: devlink.
 + * @policers: Packet trap policers.
 + * @policers_count: Count of provided packet trap policers.
 + *
 + * Return: Non-zero value on failure.
 + */
 +int
 +devl_trap_policers_register(struct devlink *devlink,
 +                          const struct devlink_trap_policer *policers,
 +                          size_t policers_count)
 +{
 +      int i, err;
 +
 +      devl_assert_locked(devlink);
 +      for (i = 0; i < policers_count; i++) {
 +              const struct devlink_trap_policer *policer = &policers[i];
 +
 +              if (WARN_ON(policer->id == 0 ||
 +                          policer->max_rate < policer->min_rate ||
 +                          policer->max_burst < policer->min_burst)) {
 +                      err = -EINVAL;
 +                      goto err_trap_policer_verify;
 +              }
 +
 +              err = devlink_trap_policer_register(devlink, policer);
 +              if (err)
 +                      goto err_trap_policer_register;
 +      }
 +      return 0;
 +
 +err_trap_policer_register:
 +err_trap_policer_verify:
 +      for (i--; i >= 0; i--)
 +              devlink_trap_policer_unregister(devlink, &policers[i]);
 +      return err;
 +}
 +EXPORT_SYMBOL_GPL(devl_trap_policers_register);
 +
 +/**
 + * devl_trap_policers_unregister - Unregister packet trap policers from devlink.
 + * @devlink: devlink.
 + * @policers: Packet trap policers.
 + * @policers_count: Count of provided packet trap policers.
 + */
 +void
 +devl_trap_policers_unregister(struct devlink *devlink,
 +                            const struct devlink_trap_policer *policers,
 +                            size_t policers_count)
 +{
 +      int i;
 +
 +      devl_assert_locked(devlink);
 +      for (i = policers_count - 1; i >= 0; i--)
 +              devlink_trap_policer_unregister(devlink, &policers[i]);
 +}
 +EXPORT_SYMBOL_GPL(devl_trap_policers_unregister);
 +
 +int devlink_compat_phys_port_name_get(struct net_device *dev,
 +                                    char *name, size_t len)
 +{
 +      struct devlink_port *devlink_port;
 +
 +      /* RTNL mutex is held here which ensures that devlink_port
 +       * instance cannot disappear in the middle. No need to take
 +       * any devlink lock as only permanent values are accessed.
 +       */
 +      ASSERT_RTNL();
 +
 +      devlink_port = dev->devlink_port;
 +      if (!devlink_port)
 +              return -EOPNOTSUPP;
 +
 +      return __devlink_port_phys_port_name_get(devlink_port, name, len);
 +}
 +
 +int devlink_compat_switch_id_get(struct net_device *dev,
 +                               struct netdev_phys_item_id *ppid)
 +{
 +      struct devlink_port *devlink_port;
 +
 +      /* Caller must hold RTNL mutex or reference to dev, which ensures that
 +       * devlink_port instance cannot disappear in the middle. No need to take
 +       * any devlink lock as only permanent values are accessed.
 +       */
 +      devlink_port = dev->devlink_port;
 +      if (!devlink_port || !devlink_port->switch_port)
 +              return -EOPNOTSUPP;
 +
 +      memcpy(ppid, &devlink_port->attrs.switch_id, sizeof(*ppid));
 +
 +      return 0;
 +}
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge