1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 #ifndef _NET_CORE_DEV_H 3 #define _NET_CORE_DEV_H 4 5 #include <linux/cleanup.h> 6 #include <linux/types.h> 7 #include <linux/rwsem.h> 8 #include <linux/netdevice.h> 9 #include <net/netdev_lock.h> 10 11 struct net; 12 struct netlink_ext_ack; 13 struct netdev_queue_config; 14 struct cpumask; 15 16 /* Random bits of netdevice that don't need to be exposed */ 17 #define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */ 18 struct sd_flow_limit { 19 struct rcu_head rcu; 20 unsigned int count; 21 u8 log_buckets; 22 unsigned int history_head; 23 u16 history[FLOW_LIMIT_HISTORY]; 24 u8 buckets[]; 25 }; 26 27 extern int netdev_flow_limit_table_len; 28 29 struct napi_struct * 30 netdev_napi_by_id_lock(struct net *net, unsigned int napi_id); 31 struct net_device *dev_get_by_napi_id(unsigned int napi_id); 32 33 struct net_device *__netdev_put_lock(struct net_device *dev, struct net *net); 34 struct net_device *netdev_put_lock(struct net_device *dev, struct net *net, 35 netdevice_tracker *tracker); 36 struct net_device * 37 netdev_xa_find_lock(struct net *net, struct net_device *dev, 38 unsigned long *index); 39 40 DEFINE_FREE(netdev_unlock, struct net_device *, if (_T) netdev_unlock(_T)); 41 42 #define for_each_netdev_lock_scoped(net, var_name, ifindex) \ 43 for (struct net_device *var_name __free(netdev_unlock) = NULL; \ 44 (var_name = netdev_xa_find_lock(net, var_name, &ifindex)); \ 45 ifindex++) 46 47 struct net_device * 48 netdev_get_by_index_lock_ops_compat(struct net *net, int ifindex); 49 struct net_device * 50 netdev_xa_find_lock_ops_compat(struct net *net, struct net_device *dev, 51 unsigned long *index); 52 53 DEFINE_FREE(netdev_unlock_ops_compat, struct net_device *, 54 if (_T) netdev_unlock_ops_compat(_T)); 55 56 #define for_each_netdev_lock_ops_compat_scoped(net, var_name, ifindex) \ 57 for (struct net_device *var_name __free(netdev_unlock_ops_compat) = NULL; \ 58 (var_name = netdev_xa_find_lock_ops_compat(net, var_name, \ 59 &ifindex)); \ 60 ifindex++) 61 62 #ifdef CONFIG_PROC_FS 63 int __init dev_proc_init(void); 64 #else 65 #define dev_proc_init() 0 66 #endif 67 68 void linkwatch_init_dev(struct net_device *dev); 69 void linkwatch_run_queue(void); 70 71 void dev_addr_flush(struct net_device *dev); 72 int dev_addr_init(struct net_device *dev); 73 void dev_addr_check(struct net_device *dev); 74 75 #if IS_ENABLED(CONFIG_NET_SHAPER) 76 void net_shaper_flush_netdev(struct net_device *dev); 77 void net_shaper_set_real_num_tx_queues(struct net_device *dev, 78 unsigned int txq); 79 #else 80 static inline void net_shaper_flush_netdev(struct net_device *dev) {} 81 static inline void net_shaper_set_real_num_tx_queues(struct net_device *dev, 82 unsigned int txq) {} 83 #endif 84 85 /* sysctls not referred to from outside net/core/ */ 86 extern int netdev_unregister_timeout_secs; 87 extern int weight_p; 88 extern int dev_weight_rx_bias; 89 extern int dev_weight_tx_bias; 90 91 extern struct rw_semaphore dev_addr_sem; 92 93 /* rtnl helpers */ 94 extern struct list_head net_todo_list; 95 void netdev_run_todo(void); 96 97 int netdev_queue_config_validate(struct net_device *dev, int rxq_idx, 98 struct netdev_queue_config *qcfg, 99 struct netlink_ext_ack *extack); 100 101 bool netif_rxq_has_mp(struct net_device *dev, unsigned int rxq_idx); 102 bool netif_rxq_is_leased(struct net_device *dev, unsigned int rxq_idx); 103 104 /* netdev management, shared between various uAPI entry points */ 105 struct netdev_name_node { 106 struct hlist_node hlist; 107 struct list_head list; 108 struct net_device *dev; 109 const char *name; 110 struct rcu_head rcu; 111 }; 112 113 int netdev_get_name(struct net *net, char *name, int ifindex); 114 int netif_change_name(struct net_device *dev, const char *newname); 115 int dev_change_name(struct net_device *dev, const char *newname); 116 117 #define netdev_for_each_altname(dev, namenode) \ 118 list_for_each_entry((namenode), &(dev)->name_node->list, list) 119 #define netdev_for_each_altname_safe(dev, namenode, next) \ 120 list_for_each_entry_safe((namenode), (next), &(dev)->name_node->list, \ 121 list) 122 123 int netdev_name_node_alt_create(struct net_device *dev, const char *name); 124 int netdev_name_node_alt_destroy(struct net_device *dev, const char *name); 125 126 int dev_validate_mtu(struct net_device *dev, int mtu, 127 struct netlink_ext_ack *extack); 128 int netif_set_mtu_ext(struct net_device *dev, int new_mtu, 129 struct netlink_ext_ack *extack); 130 131 int dev_get_phys_port_id(struct net_device *dev, 132 struct netdev_phys_item_id *ppid); 133 int dev_get_phys_port_name(struct net_device *dev, 134 char *name, size_t len); 135 136 int netif_change_proto_down(struct net_device *dev, bool proto_down); 137 int dev_change_proto_down(struct net_device *dev, bool proto_down); 138 void netdev_change_proto_down_reason_locked(struct net_device *dev, 139 unsigned long mask, u32 value); 140 141 typedef int (*bpf_op_t)(struct net_device *dev, struct netdev_bpf *bpf); 142 int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, 143 int fd, int expected_fd, u32 flags); 144 145 int netif_change_tx_queue_len(struct net_device *dev, unsigned long new_len); 146 int dev_change_tx_queue_len(struct net_device *dev, unsigned long new_len); 147 void netif_set_group(struct net_device *dev, int new_group); 148 void dev_set_group(struct net_device *dev, int new_group); 149 int netif_change_carrier(struct net_device *dev, bool new_carrier); 150 int dev_change_carrier(struct net_device *dev, bool new_carrier); 151 152 void __dev_set_rx_mode(struct net_device *dev); 153 154 void __dev_notify_flags(struct net_device *dev, unsigned int old_flags, 155 unsigned int gchanges, u32 portid, 156 const struct nlmsghdr *nlh); 157 158 void unregister_netdevice_many_notify(struct list_head *head, 159 u32 portid, const struct nlmsghdr *nlh); 160 161 static inline void netif_set_up(struct net_device *dev, bool value) 162 { 163 if (value) 164 dev->flags |= IFF_UP; 165 else 166 dev->flags &= ~IFF_UP; 167 168 if (!netdev_need_ops_lock(dev)) 169 netdev_lock(dev); 170 dev->up = value; 171 if (!netdev_need_ops_lock(dev)) 172 netdev_unlock(dev); 173 } 174 175 static inline void netif_set_gso_max_size(struct net_device *dev, 176 unsigned int size) 177 { 178 /* dev->gso_max_size is read locklessly from sk_setup_caps() */ 179 WRITE_ONCE(dev->gso_max_size, size); 180 if (size <= GSO_LEGACY_MAX_SIZE) 181 WRITE_ONCE(dev->gso_ipv4_max_size, size); 182 } 183 184 static inline void netif_set_gso_max_segs(struct net_device *dev, 185 unsigned int segs) 186 { 187 /* dev->gso_max_segs is read locklessly from sk_setup_caps() */ 188 WRITE_ONCE(dev->gso_max_segs, segs); 189 } 190 191 static inline void netif_set_gro_max_size(struct net_device *dev, 192 unsigned int size) 193 { 194 /* This pairs with the READ_ONCE() in skb_gro_receive() */ 195 WRITE_ONCE(dev->gro_max_size, size); 196 if (size <= GRO_LEGACY_MAX_SIZE) 197 WRITE_ONCE(dev->gro_ipv4_max_size, size); 198 } 199 200 static inline void netif_set_gso_ipv4_max_size(struct net_device *dev, 201 unsigned int size) 202 { 203 /* dev->gso_ipv4_max_size is read locklessly from sk_setup_caps() */ 204 WRITE_ONCE(dev->gso_ipv4_max_size, size); 205 } 206 207 static inline void netif_set_gro_ipv4_max_size(struct net_device *dev, 208 unsigned int size) 209 { 210 /* This pairs with the READ_ONCE() in skb_gro_receive() */ 211 WRITE_ONCE(dev->gro_ipv4_max_size, size); 212 } 213 214 /** 215 * napi_get_defer_hard_irqs - get the NAPI's defer_hard_irqs 216 * @n: napi struct to get the defer_hard_irqs field from 217 * 218 * Return: the per-NAPI value of the defar_hard_irqs field. 219 */ 220 static inline u32 napi_get_defer_hard_irqs(const struct napi_struct *n) 221 { 222 return READ_ONCE(n->defer_hard_irqs); 223 } 224 225 /** 226 * napi_set_defer_hard_irqs - set the defer_hard_irqs for a napi 227 * @n: napi_struct to set the defer_hard_irqs field 228 * @defer: the value the field should be set to 229 */ 230 static inline void napi_set_defer_hard_irqs(struct napi_struct *n, u32 defer) 231 { 232 WRITE_ONCE(n->defer_hard_irqs, defer); 233 } 234 235 /** 236 * netdev_set_defer_hard_irqs - set defer_hard_irqs for all NAPIs of a netdev 237 * @netdev: the net_device for which all NAPIs will have defer_hard_irqs set 238 * @defer: the defer_hard_irqs value to set 239 */ 240 static inline void netdev_set_defer_hard_irqs(struct net_device *netdev, 241 u32 defer) 242 { 243 unsigned int count = max(netdev->num_rx_queues, 244 netdev->num_tx_queues); 245 struct napi_struct *napi; 246 int i; 247 248 WRITE_ONCE(netdev->napi_defer_hard_irqs, defer); 249 list_for_each_entry(napi, &netdev->napi_list, dev_list) 250 napi_set_defer_hard_irqs(napi, defer); 251 252 for (i = 0; i < count; i++) 253 netdev->napi_config[i].defer_hard_irqs = defer; 254 } 255 256 /** 257 * napi_get_gro_flush_timeout - get the gro_flush_timeout 258 * @n: napi struct to get the gro_flush_timeout from 259 * 260 * Return: the per-NAPI value of the gro_flush_timeout field. 261 */ 262 static inline unsigned long 263 napi_get_gro_flush_timeout(const struct napi_struct *n) 264 { 265 return READ_ONCE(n->gro_flush_timeout); 266 } 267 268 /** 269 * napi_set_gro_flush_timeout - set the gro_flush_timeout for a napi 270 * @n: napi struct to set the gro_flush_timeout 271 * @timeout: timeout value to set 272 * 273 * napi_set_gro_flush_timeout sets the per-NAPI gro_flush_timeout 274 */ 275 static inline void napi_set_gro_flush_timeout(struct napi_struct *n, 276 unsigned long timeout) 277 { 278 WRITE_ONCE(n->gro_flush_timeout, timeout); 279 } 280 281 /** 282 * netdev_set_gro_flush_timeout - set gro_flush_timeout of a netdev's NAPIs 283 * @netdev: the net_device for which all NAPIs will have gro_flush_timeout set 284 * @timeout: the timeout value to set 285 */ 286 static inline void netdev_set_gro_flush_timeout(struct net_device *netdev, 287 unsigned long timeout) 288 { 289 unsigned int count = max(netdev->num_rx_queues, 290 netdev->num_tx_queues); 291 struct napi_struct *napi; 292 int i; 293 294 WRITE_ONCE(netdev->gro_flush_timeout, timeout); 295 list_for_each_entry(napi, &netdev->napi_list, dev_list) 296 napi_set_gro_flush_timeout(napi, timeout); 297 298 for (i = 0; i < count; i++) 299 netdev->napi_config[i].gro_flush_timeout = timeout; 300 } 301 302 /** 303 * napi_get_irq_suspend_timeout - get the irq_suspend_timeout 304 * @n: napi struct to get the irq_suspend_timeout from 305 * 306 * Return: the per-NAPI value of the irq_suspend_timeout field. 307 */ 308 static inline unsigned long 309 napi_get_irq_suspend_timeout(const struct napi_struct *n) 310 { 311 return READ_ONCE(n->irq_suspend_timeout); 312 } 313 314 /** 315 * napi_set_irq_suspend_timeout - set the irq_suspend_timeout for a napi 316 * @n: napi struct to set the irq_suspend_timeout 317 * @timeout: timeout value to set 318 * 319 * napi_set_irq_suspend_timeout sets the per-NAPI irq_suspend_timeout 320 */ 321 static inline void napi_set_irq_suspend_timeout(struct napi_struct *n, 322 unsigned long timeout) 323 { 324 WRITE_ONCE(n->irq_suspend_timeout, timeout); 325 } 326 327 static inline enum netdev_napi_threaded napi_get_threaded(struct napi_struct *n) 328 { 329 if (test_bit(NAPI_STATE_THREADED_BUSY_POLL, &n->state)) 330 return NETDEV_NAPI_THREADED_BUSY_POLL; 331 332 if (test_bit(NAPI_STATE_THREADED, &n->state)) 333 return NETDEV_NAPI_THREADED_ENABLED; 334 335 return NETDEV_NAPI_THREADED_DISABLED; 336 } 337 338 static inline enum netdev_napi_threaded 339 napi_get_threaded_config(struct net_device *dev, struct napi_struct *n) 340 { 341 if (n->config) 342 return n->config->threaded; 343 return dev->threaded; 344 } 345 346 int napi_set_threaded(struct napi_struct *n, 347 enum netdev_napi_threaded threaded); 348 349 int netif_set_threaded(struct net_device *dev, 350 enum netdev_napi_threaded threaded); 351 352 int rps_cpumask_housekeeping(struct cpumask *mask); 353 354 #if defined(CONFIG_DEBUG_NET) && defined(CONFIG_BPF_SYSCALL) 355 void xdp_do_check_flushed(struct napi_struct *napi); 356 #else 357 static inline void xdp_do_check_flushed(struct napi_struct *napi) { } 358 #endif 359 360 /* Best effort check that NAPI is not idle (can't be scheduled to run) */ 361 static inline void napi_assert_will_not_race(const struct napi_struct *napi) 362 { 363 /* uninitialized instance, can't race */ 364 if (!napi->poll_list.next) 365 return; 366 367 /* SCHED bit is set on disabled instances */ 368 WARN_ON(!test_bit(NAPI_STATE_SCHED, &napi->state)); 369 WARN_ON(READ_ONCE(napi->list_owner) != -1); 370 } 371 372 void kick_defer_list_purge(unsigned int cpu); 373 374 int dev_set_hwtstamp_phylib(struct net_device *dev, 375 struct kernel_hwtstamp_config *cfg, 376 struct netlink_ext_ack *extack); 377 int dev_get_hwtstamp_phylib(struct net_device *dev, 378 struct kernel_hwtstamp_config *cfg); 379 int net_hwtstamp_validate(const struct kernel_hwtstamp_config *cfg); 380 381 #endif 382