1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Linux network device link state notification 4 * 5 * Author: 6 * Stefan Rompf <sux@loplof.de> 7 */ 8 9 #include <linux/module.h> 10 #include <linux/netdevice.h> 11 #include <linux/if.h> 12 #include <net/sock.h> 13 #include <net/pkt_sched.h> 14 #include <linux/rtnetlink.h> 15 #include <linux/jiffies.h> 16 #include <linux/spinlock.h> 17 #include <linux/workqueue.h> 18 #include <linux/bitops.h> 19 #include <linux/types.h> 20 21 #include "dev.h" 22 23 enum lw_bits { 24 LW_URGENT = 0, 25 }; 26 27 static unsigned long linkwatch_flags; 28 static unsigned long linkwatch_nextevent; 29 30 static void linkwatch_event(struct work_struct *dummy); 31 static DECLARE_DELAYED_WORK(linkwatch_work, linkwatch_event); 32 33 static LIST_HEAD(lweventlist); 34 static DEFINE_SPINLOCK(lweventlist_lock); 35 36 static unsigned int default_operstate(const struct net_device *dev) 37 { 38 if (netif_testing(dev)) 39 return IF_OPER_TESTING; 40 41 /* Some uppers (DSA) have additional sources for being down, so 42 * first check whether lower is indeed the source of its down state. 43 */ 44 if (!netif_carrier_ok(dev)) { 45 struct net_device *peer; 46 int iflink; 47 48 /* If called from netdev_run_todo()/linkwatch_sync_dev(), 49 * dev_net(dev) can be already freed, and RTNL is not held. 50 */ 51 if (dev->reg_state <= NETREG_REGISTERED) 52 iflink = dev_get_iflink(dev); 53 else 54 iflink = dev->ifindex; 55 56 if (iflink == dev->ifindex) 57 return IF_OPER_DOWN; 58 59 ASSERT_RTNL(); 60 peer = __dev_get_by_index(dev_net(dev), iflink); 61 if (!peer) 62 return IF_OPER_DOWN; 63 64 return netif_carrier_ok(peer) ? IF_OPER_DOWN : 65 IF_OPER_LOWERLAYERDOWN; 66 } 67 68 if (netif_dormant(dev)) 69 return IF_OPER_DORMANT; 70 71 return IF_OPER_UP; 72 } 73 74 static void rfc2863_policy(struct net_device *dev) 75 { 76 unsigned int operstate = default_operstate(dev); 77 78 if (operstate == READ_ONCE(dev->operstate)) 79 return; 80 81 switch(dev->link_mode) { 82 case IF_LINK_MODE_TESTING: 83 if (operstate == IF_OPER_UP) 84 operstate = IF_OPER_TESTING; 85 break; 86 87 case IF_LINK_MODE_DORMANT: 88 if (operstate == IF_OPER_UP) 89 operstate = IF_OPER_DORMANT; 90 break; 91 case IF_LINK_MODE_DEFAULT: 92 default: 93 break; 94 } 95 96 WRITE_ONCE(dev->operstate, operstate); 97 } 98 99 100 void linkwatch_init_dev(struct net_device *dev) 101 { 102 /* Handle pre-registration link state changes */ 103 if (!netif_carrier_ok(dev) || netif_dormant(dev) || 104 netif_testing(dev)) 105 rfc2863_policy(dev); 106 } 107 108 109 static bool linkwatch_urgent_event(struct net_device *dev) 110 { 111 if (!netif_running(dev)) 112 return false; 113 114 if (dev->ifindex != dev_get_iflink(dev)) 115 return true; 116 117 if (netif_is_lag_port(dev) || netif_is_lag_master(dev)) 118 return true; 119 120 return netif_carrier_ok(dev) && qdisc_tx_changing(dev); 121 } 122 123 124 static void linkwatch_add_event(struct net_device *dev) 125 { 126 unsigned long flags; 127 128 spin_lock_irqsave(&lweventlist_lock, flags); 129 if (list_empty(&dev->link_watch_list)) { 130 list_add_tail(&dev->link_watch_list, &lweventlist); 131 netdev_hold(dev, &dev->linkwatch_dev_tracker, GFP_ATOMIC); 132 } 133 spin_unlock_irqrestore(&lweventlist_lock, flags); 134 } 135 136 137 static void linkwatch_schedule_work(int urgent) 138 { 139 unsigned long delay = linkwatch_nextevent - jiffies; 140 141 if (test_bit(LW_URGENT, &linkwatch_flags)) 142 return; 143 144 /* Minimise down-time: drop delay for up event. */ 145 if (urgent) { 146 if (test_and_set_bit(LW_URGENT, &linkwatch_flags)) 147 return; 148 delay = 0; 149 } 150 151 /* If we wrap around we'll delay it by at most HZ. */ 152 if (delay > HZ) 153 delay = 0; 154 155 /* 156 * If urgent, schedule immediate execution; otherwise, don't 157 * override the existing timer. 158 */ 159 if (test_bit(LW_URGENT, &linkwatch_flags)) 160 mod_delayed_work(system_dfl_wq, &linkwatch_work, 0); 161 else 162 queue_delayed_work(system_dfl_wq, &linkwatch_work, delay); 163 } 164 165 166 static void linkwatch_do_dev(struct net_device *dev) 167 { 168 /* 169 * Make sure the above read is complete since it can be 170 * rewritten as soon as we clear the bit below. 171 */ 172 smp_mb__before_atomic(); 173 174 /* We are about to handle this device, 175 * so new events can be accepted 176 */ 177 clear_bit(__LINK_STATE_LINKWATCH_PENDING, &dev->state); 178 179 rfc2863_policy(dev); 180 if (dev->flags & IFF_UP) { 181 if (netif_carrier_ok(dev)) 182 dev_activate(dev); 183 else 184 dev_deactivate(dev); 185 186 netif_state_change(dev); 187 } 188 } 189 190 static void __linkwatch_run_queue(int urgent_only) 191 { 192 #define MAX_DO_DEV_PER_LOOP 100 193 194 int do_dev = MAX_DO_DEV_PER_LOOP; 195 /* Use a local list here since we add non-urgent 196 * events back to the global one when called with 197 * urgent_only=1. 198 */ 199 LIST_HEAD(wrk); 200 201 /* Give urgent case more budget */ 202 if (urgent_only) 203 do_dev += MAX_DO_DEV_PER_LOOP; 204 205 /* 206 * Limit the number of linkwatch events to one 207 * per second so that a runaway driver does not 208 * cause a storm of messages on the netlink 209 * socket. This limit does not apply to up events 210 * while the device qdisc is down. 211 */ 212 if (!urgent_only) 213 linkwatch_nextevent = jiffies + HZ; 214 /* Limit wrap-around effect on delay. */ 215 else if (time_after(linkwatch_nextevent, jiffies + HZ)) 216 linkwatch_nextevent = jiffies; 217 218 clear_bit(LW_URGENT, &linkwatch_flags); 219 220 spin_lock_irq(&lweventlist_lock); 221 list_splice_init(&lweventlist, &wrk); 222 223 while (!list_empty(&wrk) && do_dev > 0) { 224 struct net_device *dev; 225 226 dev = list_first_entry(&wrk, struct net_device, link_watch_list); 227 list_del_init(&dev->link_watch_list); 228 229 if (!netif_device_present(dev) || 230 (urgent_only && !linkwatch_urgent_event(dev))) { 231 list_add_tail(&dev->link_watch_list, &lweventlist); 232 continue; 233 } 234 /* We must free netdev tracker under 235 * the spinlock protection. 236 */ 237 netdev_tracker_free(dev, &dev->linkwatch_dev_tracker); 238 spin_unlock_irq(&lweventlist_lock); 239 netdev_lock_ops(dev); 240 linkwatch_do_dev(dev); 241 netdev_unlock_ops(dev); 242 /* Use __dev_put() because netdev_tracker_free() was already 243 * called above. Must be after netdev_unlock_ops() to prevent 244 * netdev_run_todo() from freeing the device while still in use. 245 */ 246 __dev_put(dev); 247 do_dev--; 248 spin_lock_irq(&lweventlist_lock); 249 } 250 251 /* Add the remaining work back to lweventlist */ 252 list_splice_init(&wrk, &lweventlist); 253 254 if (!list_empty(&lweventlist)) 255 linkwatch_schedule_work(0); 256 spin_unlock_irq(&lweventlist_lock); 257 } 258 259 static bool linkwatch_clean_dev(struct net_device *dev) 260 { 261 unsigned long flags; 262 bool clean = false; 263 264 spin_lock_irqsave(&lweventlist_lock, flags); 265 if (!list_empty(&dev->link_watch_list)) { 266 list_del_init(&dev->link_watch_list); 267 clean = true; 268 /* We must release netdev tracker under 269 * the spinlock protection. 270 */ 271 netdev_tracker_free(dev, &dev->linkwatch_dev_tracker); 272 } 273 spin_unlock_irqrestore(&lweventlist_lock, flags); 274 275 return clean; 276 } 277 278 void __linkwatch_sync_dev(struct net_device *dev) 279 { 280 netdev_ops_assert_locked(dev); 281 282 if (linkwatch_clean_dev(dev)) { 283 linkwatch_do_dev(dev); 284 /* Use __dev_put() because netdev_tracker_free() was already 285 * called inside linkwatch_clean_dev(). 286 */ 287 __dev_put(dev); 288 } 289 } 290 291 void linkwatch_sync_dev(struct net_device *dev) 292 { 293 if (linkwatch_clean_dev(dev)) { 294 netdev_lock_ops(dev); 295 linkwatch_do_dev(dev); 296 netdev_unlock_ops(dev); 297 /* Use __dev_put() because netdev_tracker_free() was already 298 * called inside linkwatch_clean_dev(). 299 */ 300 __dev_put(dev); 301 } 302 } 303 304 /* Must be called with the rtnl semaphore held */ 305 void linkwatch_run_queue(void) 306 { 307 __linkwatch_run_queue(0); 308 } 309 310 311 static void linkwatch_event(struct work_struct *dummy) 312 { 313 rtnl_lock(); 314 __linkwatch_run_queue(time_after(linkwatch_nextevent, jiffies)); 315 rtnl_unlock(); 316 } 317 318 319 void linkwatch_fire_event(struct net_device *dev) 320 { 321 bool urgent = linkwatch_urgent_event(dev); 322 323 if (!test_and_set_bit(__LINK_STATE_LINKWATCH_PENDING, &dev->state)) { 324 linkwatch_add_event(dev); 325 } else if (!urgent) 326 return; 327 328 linkwatch_schedule_work(urgent); 329 } 330 EXPORT_SYMBOL(linkwatch_fire_event); 331