~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/core/netpoll.c

Version: ~ [ linux-6.11.5 ] ~ [ linux-6.10.14 ] ~ [ linux-6.9.12 ] ~ [ linux-6.8.12 ] ~ [ linux-6.7.12 ] ~ [ linux-6.6.58 ] ~ [ linux-6.5.13 ] ~ [ linux-6.4.16 ] ~ [ linux-6.3.13 ] ~ [ linux-6.2.16 ] ~ [ linux-6.1.114 ] ~ [ linux-6.0.19 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.169 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.228 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.284 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.322 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.336 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.337 ] ~ [ linux-4.4.302 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0-only
  2 /*
  3  * Common framework for low-level network console, dump, and debugger code
  4  *
  5  * Sep 8 2003  Matt Mackall <mpm@selenic.com>
  6  *
  7  * based on the netconsole code from:
  8  *
  9  * Copyright (C) 2001  Ingo Molnar <mingo@redhat.com>
 10  * Copyright (C) 2002  Red Hat, Inc.
 11  */
 12 
 13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 14 
 15 #include <linux/moduleparam.h>
 16 #include <linux/kernel.h>
 17 #include <linux/netdevice.h>
 18 #include <linux/etherdevice.h>
 19 #include <linux/string.h>
 20 #include <linux/if_arp.h>
 21 #include <linux/inetdevice.h>
 22 #include <linux/inet.h>
 23 #include <linux/interrupt.h>
 24 #include <linux/netpoll.h>
 25 #include <linux/sched.h>
 26 #include <linux/delay.h>
 27 #include <linux/rcupdate.h>
 28 #include <linux/workqueue.h>
 29 #include <linux/slab.h>
 30 #include <linux/export.h>
 31 #include <linux/if_vlan.h>
 32 #include <net/tcp.h>
 33 #include <net/udp.h>
 34 #include <net/addrconf.h>
 35 #include <net/ndisc.h>
 36 #include <net/ip6_checksum.h>
 37 #include <asm/unaligned.h>
 38 #include <trace/events/napi.h>
 39 #include <linux/kconfig.h>
 40 
 41 /*
 42  * We maintain a small pool of fully-sized skbs, to make sure the
 43  * message gets out even in extreme OOM situations.
 44  */
 45 
 46 #define MAX_UDP_CHUNK 1460
 47 #define MAX_SKBS 32
 48 
 49 static struct sk_buff_head skb_pool;
 50 
 51 DEFINE_STATIC_SRCU(netpoll_srcu);
 52 
 53 #define USEC_PER_POLL   50
 54 
 55 #define MAX_SKB_SIZE                                                    \
 56         (sizeof(struct ethhdr) +                                        \
 57          sizeof(struct iphdr) +                                         \
 58          sizeof(struct udphdr) +                                        \
 59          MAX_UDP_CHUNK)
 60 
 61 static void zap_completion_queue(void);
 62 
 63 static unsigned int carrier_timeout = 4;
 64 module_param(carrier_timeout, uint, 0644);
 65 
 66 #define np_info(np, fmt, ...)                           \
 67         pr_info("%s: " fmt, np->name, ##__VA_ARGS__)
 68 #define np_err(np, fmt, ...)                            \
 69         pr_err("%s: " fmt, np->name, ##__VA_ARGS__)
 70 #define np_notice(np, fmt, ...)                         \
 71         pr_notice("%s: " fmt, np->name, ##__VA_ARGS__)
 72 
 73 static netdev_tx_t netpoll_start_xmit(struct sk_buff *skb,
 74                                       struct net_device *dev,
 75                                       struct netdev_queue *txq)
 76 {
 77         netdev_tx_t status = NETDEV_TX_OK;
 78         netdev_features_t features;
 79 
 80         features = netif_skb_features(skb);
 81 
 82         if (skb_vlan_tag_present(skb) &&
 83             !vlan_hw_offload_capable(features, skb->vlan_proto)) {
 84                 skb = __vlan_hwaccel_push_inside(skb);
 85                 if (unlikely(!skb)) {
 86                         /* This is actually a packet drop, but we
 87                          * don't want the code that calls this
 88                          * function to try and operate on a NULL skb.
 89                          */
 90                         goto out;
 91                 }
 92         }
 93 
 94         status = netdev_start_xmit(skb, dev, txq, false);
 95 
 96 out:
 97         return status;
 98 }
 99 
100 static void queue_process(struct work_struct *work)
101 {
102         struct netpoll_info *npinfo =
103                 container_of(work, struct netpoll_info, tx_work.work);
104         struct sk_buff *skb;
105         unsigned long flags;
106 
107         while ((skb = skb_dequeue(&npinfo->txq))) {
108                 struct net_device *dev = skb->dev;
109                 struct netdev_queue *txq;
110                 unsigned int q_index;
111 
112                 if (!netif_device_present(dev) || !netif_running(dev)) {
113                         kfree_skb(skb);
114                         continue;
115                 }
116 
117                 local_irq_save(flags);
118                 /* check if skb->queue_mapping is still valid */
119                 q_index = skb_get_queue_mapping(skb);
120                 if (unlikely(q_index >= dev->real_num_tx_queues)) {
121                         q_index = q_index % dev->real_num_tx_queues;
122                         skb_set_queue_mapping(skb, q_index);
123                 }
124                 txq = netdev_get_tx_queue(dev, q_index);
125                 HARD_TX_LOCK(dev, txq, smp_processor_id());
126                 if (netif_xmit_frozen_or_stopped(txq) ||
127                     !dev_xmit_complete(netpoll_start_xmit(skb, dev, txq))) {
128                         skb_queue_head(&npinfo->txq, skb);
129                         HARD_TX_UNLOCK(dev, txq);
130                         local_irq_restore(flags);
131 
132                         schedule_delayed_work(&npinfo->tx_work, HZ/10);
133                         return;
134                 }
135                 HARD_TX_UNLOCK(dev, txq);
136                 local_irq_restore(flags);
137         }
138 }
139 
140 static int netif_local_xmit_active(struct net_device *dev)
141 {
142         int i;
143 
144         for (i = 0; i < dev->num_tx_queues; i++) {
145                 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
146 
147                 if (READ_ONCE(txq->xmit_lock_owner) == smp_processor_id())
148                         return 1;
149         }
150 
151         return 0;
152 }
153 
154 static void poll_one_napi(struct napi_struct *napi)
155 {
156         int work;
157 
158         /* If we set this bit but see that it has already been set,
159          * that indicates that napi has been disabled and we need
160          * to abort this operation
161          */
162         if (test_and_set_bit(NAPI_STATE_NPSVC, &napi->state))
163                 return;
164 
165         /* We explicilty pass the polling call a budget of 0 to
166          * indicate that we are clearing the Tx path only.
167          */
168         work = napi->poll(napi, 0);
169         WARN_ONCE(work, "%pS exceeded budget in poll\n", napi->poll);
170         trace_napi_poll(napi, work, 0);
171 
172         clear_bit(NAPI_STATE_NPSVC, &napi->state);
173 }
174 
175 static void poll_napi(struct net_device *dev)
176 {
177         struct napi_struct *napi;
178         int cpu = smp_processor_id();
179 
180         list_for_each_entry_rcu(napi, &dev->napi_list, dev_list) {
181                 if (cmpxchg(&napi->poll_owner, -1, cpu) == -1) {
182                         poll_one_napi(napi);
183                         smp_store_release(&napi->poll_owner, -1);
184                 }
185         }
186 }
187 
188 void netpoll_poll_dev(struct net_device *dev)
189 {
190         struct netpoll_info *ni = rcu_dereference_bh(dev->npinfo);
191         const struct net_device_ops *ops;
192 
193         /* Don't do any rx activity if the dev_lock mutex is held
194          * the dev_open/close paths use this to block netpoll activity
195          * while changing device state
196          */
197         if (!ni || down_trylock(&ni->dev_lock))
198                 return;
199 
200         /* Some drivers will take the same locks in poll and xmit,
201          * we can't poll if local CPU is already in xmit.
202          */
203         if (!netif_running(dev) || netif_local_xmit_active(dev)) {
204                 up(&ni->dev_lock);
205                 return;
206         }
207 
208         ops = dev->netdev_ops;
209         if (ops->ndo_poll_controller)
210                 ops->ndo_poll_controller(dev);
211 
212         poll_napi(dev);
213 
214         up(&ni->dev_lock);
215 
216         zap_completion_queue();
217 }
218 EXPORT_SYMBOL(netpoll_poll_dev);
219 
220 void netpoll_poll_disable(struct net_device *dev)
221 {
222         struct netpoll_info *ni;
223         int idx;
224         might_sleep();
225         idx = srcu_read_lock(&netpoll_srcu);
226         ni = srcu_dereference(dev->npinfo, &netpoll_srcu);
227         if (ni)
228                 down(&ni->dev_lock);
229         srcu_read_unlock(&netpoll_srcu, idx);
230 }
231 
232 void netpoll_poll_enable(struct net_device *dev)
233 {
234         struct netpoll_info *ni;
235         rcu_read_lock();
236         ni = rcu_dereference(dev->npinfo);
237         if (ni)
238                 up(&ni->dev_lock);
239         rcu_read_unlock();
240 }
241 
242 static void refill_skbs(void)
243 {
244         struct sk_buff *skb;
245         unsigned long flags;
246 
247         spin_lock_irqsave(&skb_pool.lock, flags);
248         while (skb_pool.qlen < MAX_SKBS) {
249                 skb = alloc_skb(MAX_SKB_SIZE, GFP_ATOMIC);
250                 if (!skb)
251                         break;
252 
253                 __skb_queue_tail(&skb_pool, skb);
254         }
255         spin_unlock_irqrestore(&skb_pool.lock, flags);
256 }
257 
258 static void zap_completion_queue(void)
259 {
260         unsigned long flags;
261         struct softnet_data *sd = &get_cpu_var(softnet_data);
262 
263         if (sd->completion_queue) {
264                 struct sk_buff *clist;
265 
266                 local_irq_save(flags);
267                 clist = sd->completion_queue;
268                 sd->completion_queue = NULL;
269                 local_irq_restore(flags);
270 
271                 while (clist != NULL) {
272                         struct sk_buff *skb = clist;
273                         clist = clist->next;
274                         if (!skb_irq_freeable(skb)) {
275                                 refcount_set(&skb->users, 1);
276                                 dev_kfree_skb_any(skb); /* put this one back */
277                         } else {
278                                 __kfree_skb(skb);
279                         }
280                 }
281         }
282 
283         put_cpu_var(softnet_data);
284 }
285 
286 static struct sk_buff *find_skb(struct netpoll *np, int len, int reserve)
287 {
288         int count = 0;
289         struct sk_buff *skb;
290 
291         zap_completion_queue();
292         refill_skbs();
293 repeat:
294 
295         skb = alloc_skb(len, GFP_ATOMIC);
296         if (!skb)
297                 skb = skb_dequeue(&skb_pool);
298 
299         if (!skb) {
300                 if (++count < 10) {
301                         netpoll_poll_dev(np->dev);
302                         goto repeat;
303                 }
304                 return NULL;
305         }
306 
307         refcount_set(&skb->users, 1);
308         skb_reserve(skb, reserve);
309         return skb;
310 }
311 
312 static int netpoll_owner_active(struct net_device *dev)
313 {
314         struct napi_struct *napi;
315 
316         list_for_each_entry_rcu(napi, &dev->napi_list, dev_list) {
317                 if (READ_ONCE(napi->poll_owner) == smp_processor_id())
318                         return 1;
319         }
320         return 0;
321 }
322 
323 /* call with IRQ disabled */
324 static netdev_tx_t __netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
325 {
326         netdev_tx_t status = NETDEV_TX_BUSY;
327         struct net_device *dev;
328         unsigned long tries;
329         /* It is up to the caller to keep npinfo alive. */
330         struct netpoll_info *npinfo;
331 
332         lockdep_assert_irqs_disabled();
333 
334         dev = np->dev;
335         npinfo = rcu_dereference_bh(dev->npinfo);
336 
337         if (!npinfo || !netif_running(dev) || !netif_device_present(dev)) {
338                 dev_kfree_skb_irq(skb);
339                 return NET_XMIT_DROP;
340         }
341 
342         /* don't get messages out of order, and no recursion */
343         if (skb_queue_len(&npinfo->txq) == 0 && !netpoll_owner_active(dev)) {
344                 struct netdev_queue *txq;
345 
346                 txq = netdev_core_pick_tx(dev, skb, NULL);
347 
348                 /* try until next clock tick */
349                 for (tries = jiffies_to_usecs(1)/USEC_PER_POLL;
350                      tries > 0; --tries) {
351                         if (HARD_TX_TRYLOCK(dev, txq)) {
352                                 if (!netif_xmit_stopped(txq))
353                                         status = netpoll_start_xmit(skb, dev, txq);
354 
355                                 HARD_TX_UNLOCK(dev, txq);
356 
357                                 if (dev_xmit_complete(status))
358                                         break;
359 
360                         }
361 
362                         /* tickle device maybe there is some cleanup */
363                         netpoll_poll_dev(np->dev);
364 
365                         udelay(USEC_PER_POLL);
366                 }
367 
368                 WARN_ONCE(!irqs_disabled(),
369                         "netpoll_send_skb_on_dev(): %s enabled interrupts in poll (%pS)\n",
370                         dev->name, dev->netdev_ops->ndo_start_xmit);
371 
372         }
373 
374         if (!dev_xmit_complete(status)) {
375                 skb_queue_tail(&npinfo->txq, skb);
376                 schedule_delayed_work(&npinfo->tx_work,0);
377         }
378         return NETDEV_TX_OK;
379 }
380 
381 netdev_tx_t netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
382 {
383         unsigned long flags;
384         netdev_tx_t ret;
385 
386         if (unlikely(!np)) {
387                 dev_kfree_skb_irq(skb);
388                 ret = NET_XMIT_DROP;
389         } else {
390                 local_irq_save(flags);
391                 ret = __netpoll_send_skb(np, skb);
392                 local_irq_restore(flags);
393         }
394         return ret;
395 }
396 EXPORT_SYMBOL(netpoll_send_skb);
397 
398 void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
399 {
400         int total_len, ip_len, udp_len;
401         struct sk_buff *skb;
402         struct udphdr *udph;
403         struct iphdr *iph;
404         struct ethhdr *eth;
405         static atomic_t ip_ident;
406         struct ipv6hdr *ip6h;
407 
408         if (!IS_ENABLED(CONFIG_PREEMPT_RT))
409                 WARN_ON_ONCE(!irqs_disabled());
410 
411         udp_len = len + sizeof(*udph);
412         if (np->ipv6)
413                 ip_len = udp_len + sizeof(*ip6h);
414         else
415                 ip_len = udp_len + sizeof(*iph);
416 
417         total_len = ip_len + LL_RESERVED_SPACE(np->dev);
418 
419         skb = find_skb(np, total_len + np->dev->needed_tailroom,
420                        total_len - len);
421         if (!skb)
422                 return;
423 
424         skb_copy_to_linear_data(skb, msg, len);
425         skb_put(skb, len);
426 
427         skb_push(skb, sizeof(*udph));
428         skb_reset_transport_header(skb);
429         udph = udp_hdr(skb);
430         udph->source = htons(np->local_port);
431         udph->dest = htons(np->remote_port);
432         udph->len = htons(udp_len);
433 
434         if (np->ipv6) {
435                 udph->check = 0;
436                 udph->check = csum_ipv6_magic(&np->local_ip.in6,
437                                               &np->remote_ip.in6,
438                                               udp_len, IPPROTO_UDP,
439                                               csum_partial(udph, udp_len, 0));
440                 if (udph->check == 0)
441                         udph->check = CSUM_MANGLED_0;
442 
443                 skb_push(skb, sizeof(*ip6h));
444                 skb_reset_network_header(skb);
445                 ip6h = ipv6_hdr(skb);
446 
447                 /* ip6h->version = 6; ip6h->priority = 0; */
448                 *(unsigned char *)ip6h = 0x60;
449                 ip6h->flow_lbl[0] = 0;
450                 ip6h->flow_lbl[1] = 0;
451                 ip6h->flow_lbl[2] = 0;
452 
453                 ip6h->payload_len = htons(sizeof(struct udphdr) + len);
454                 ip6h->nexthdr = IPPROTO_UDP;
455                 ip6h->hop_limit = 32;
456                 ip6h->saddr = np->local_ip.in6;
457                 ip6h->daddr = np->remote_ip.in6;
458 
459                 eth = skb_push(skb, ETH_HLEN);
460                 skb_reset_mac_header(skb);
461                 skb->protocol = eth->h_proto = htons(ETH_P_IPV6);
462         } else {
463                 udph->check = 0;
464                 udph->check = csum_tcpudp_magic(np->local_ip.ip,
465                                                 np->remote_ip.ip,
466                                                 udp_len, IPPROTO_UDP,
467                                                 csum_partial(udph, udp_len, 0));
468                 if (udph->check == 0)
469                         udph->check = CSUM_MANGLED_0;
470 
471                 skb_push(skb, sizeof(*iph));
472                 skb_reset_network_header(skb);
473                 iph = ip_hdr(skb);
474 
475                 /* iph->version = 4; iph->ihl = 5; */
476                 *(unsigned char *)iph = 0x45;
477                 iph->tos      = 0;
478                 put_unaligned(htons(ip_len), &(iph->tot_len));
479                 iph->id       = htons(atomic_inc_return(&ip_ident));
480                 iph->frag_off = 0;
481                 iph->ttl      = 64;
482                 iph->protocol = IPPROTO_UDP;
483                 iph->check    = 0;
484                 put_unaligned(np->local_ip.ip, &(iph->saddr));
485                 put_unaligned(np->remote_ip.ip, &(iph->daddr));
486                 iph->check    = ip_fast_csum((unsigned char *)iph, iph->ihl);
487 
488                 eth = skb_push(skb, ETH_HLEN);
489                 skb_reset_mac_header(skb);
490                 skb->protocol = eth->h_proto = htons(ETH_P_IP);
491         }
492 
493         ether_addr_copy(eth->h_source, np->dev->dev_addr);
494         ether_addr_copy(eth->h_dest, np->remote_mac);
495 
496         skb->dev = np->dev;
497 
498         netpoll_send_skb(np, skb);
499 }
500 EXPORT_SYMBOL(netpoll_send_udp);
501 
502 void netpoll_print_options(struct netpoll *np)
503 {
504         np_info(np, "local port %d\n", np->local_port);
505         if (np->ipv6)
506                 np_info(np, "local IPv6 address %pI6c\n", &np->local_ip.in6);
507         else
508                 np_info(np, "local IPv4 address %pI4\n", &np->local_ip.ip);
509         np_info(np, "interface '%s'\n", np->dev_name);
510         np_info(np, "remote port %d\n", np->remote_port);
511         if (np->ipv6)
512                 np_info(np, "remote IPv6 address %pI6c\n", &np->remote_ip.in6);
513         else
514                 np_info(np, "remote IPv4 address %pI4\n", &np->remote_ip.ip);
515         np_info(np, "remote ethernet address %pM\n", np->remote_mac);
516 }
517 EXPORT_SYMBOL(netpoll_print_options);
518 
519 static int netpoll_parse_ip_addr(const char *str, union inet_addr *addr)
520 {
521         const char *end;
522 
523         if (!strchr(str, ':') &&
524             in4_pton(str, -1, (void *)addr, -1, &end) > 0) {
525                 if (!*end)
526                         return 0;
527         }
528         if (in6_pton(str, -1, addr->in6.s6_addr, -1, &end) > 0) {
529 #if IS_ENABLED(CONFIG_IPV6)
530                 if (!*end)
531                         return 1;
532 #else
533                 return -1;
534 #endif
535         }
536         return -1;
537 }
538 
539 int netpoll_parse_options(struct netpoll *np, char *opt)
540 {
541         char *cur=opt, *delim;
542         int ipv6;
543         bool ipversion_set = false;
544 
545         if (*cur != '@') {
546                 if ((delim = strchr(cur, '@')) == NULL)
547                         goto parse_failed;
548                 *delim = 0;
549                 if (kstrtou16(cur, 10, &np->local_port))
550                         goto parse_failed;
551                 cur = delim;
552         }
553         cur++;
554 
555         if (*cur != '/') {
556                 ipversion_set = true;
557                 if ((delim = strchr(cur, '/')) == NULL)
558                         goto parse_failed;
559                 *delim = 0;
560                 ipv6 = netpoll_parse_ip_addr(cur, &np->local_ip);
561                 if (ipv6 < 0)
562                         goto parse_failed;
563                 else
564                         np->ipv6 = (bool)ipv6;
565                 cur = delim;
566         }
567         cur++;
568 
569         if (*cur != ',') {
570                 /* parse out dev name */
571                 if ((delim = strchr(cur, ',')) == NULL)
572                         goto parse_failed;
573                 *delim = 0;
574                 strscpy(np->dev_name, cur, sizeof(np->dev_name));
575                 cur = delim;
576         }
577         cur++;
578 
579         if (*cur != '@') {
580                 /* dst port */
581                 if ((delim = strchr(cur, '@')) == NULL)
582                         goto parse_failed;
583                 *delim = 0;
584                 if (*cur == ' ' || *cur == '\t')
585                         np_info(np, "warning: whitespace is not allowed\n");
586                 if (kstrtou16(cur, 10, &np->remote_port))
587                         goto parse_failed;
588                 cur = delim;
589         }
590         cur++;
591 
592         /* dst ip */
593         if ((delim = strchr(cur, '/')) == NULL)
594                 goto parse_failed;
595         *delim = 0;
596         ipv6 = netpoll_parse_ip_addr(cur, &np->remote_ip);
597         if (ipv6 < 0)
598                 goto parse_failed;
599         else if (ipversion_set && np->ipv6 != (bool)ipv6)
600                 goto parse_failed;
601         else
602                 np->ipv6 = (bool)ipv6;
603         cur = delim + 1;
604 
605         if (*cur != 0) {
606                 /* MAC address */
607                 if (!mac_pton(cur, np->remote_mac))
608                         goto parse_failed;
609         }
610 
611         netpoll_print_options(np);
612 
613         return 0;
614 
615  parse_failed:
616         np_info(np, "couldn't parse config at '%s'!\n", cur);
617         return -1;
618 }
619 EXPORT_SYMBOL(netpoll_parse_options);
620 
621 int __netpoll_setup(struct netpoll *np, struct net_device *ndev)
622 {
623         struct netpoll_info *npinfo;
624         const struct net_device_ops *ops;
625         int err;
626 
627         if (ndev->priv_flags & IFF_DISABLE_NETPOLL) {
628                 np_err(np, "%s doesn't support polling, aborting\n",
629                        ndev->name);
630                 err = -ENOTSUPP;
631                 goto out;
632         }
633 
634         if (!ndev->npinfo) {
635                 npinfo = kmalloc(sizeof(*npinfo), GFP_KERNEL);
636                 if (!npinfo) {
637                         err = -ENOMEM;
638                         goto out;
639                 }
640 
641                 sema_init(&npinfo->dev_lock, 1);
642                 skb_queue_head_init(&npinfo->txq);
643                 INIT_DELAYED_WORK(&npinfo->tx_work, queue_process);
644 
645                 refcount_set(&npinfo->refcnt, 1);
646 
647                 ops = ndev->netdev_ops;
648                 if (ops->ndo_netpoll_setup) {
649                         err = ops->ndo_netpoll_setup(ndev, npinfo);
650                         if (err)
651                                 goto free_npinfo;
652                 }
653         } else {
654                 npinfo = rtnl_dereference(ndev->npinfo);
655                 refcount_inc(&npinfo->refcnt);
656         }
657 
658         np->dev = ndev;
659         strscpy(np->dev_name, ndev->name, IFNAMSIZ);
660         npinfo->netpoll = np;
661 
662         /* last thing to do is link it to the net device structure */
663         rcu_assign_pointer(ndev->npinfo, npinfo);
664 
665         return 0;
666 
667 free_npinfo:
668         kfree(npinfo);
669 out:
670         return err;
671 }
672 EXPORT_SYMBOL_GPL(__netpoll_setup);
673 
674 int netpoll_setup(struct netpoll *np)
675 {
676         struct net_device *ndev = NULL;
677         bool ip_overwritten = false;
678         struct in_device *in_dev;
679         int err;
680 
681         rtnl_lock();
682         if (np->dev_name[0]) {
683                 struct net *net = current->nsproxy->net_ns;
684                 ndev = __dev_get_by_name(net, np->dev_name);
685         }
686         if (!ndev) {
687                 np_err(np, "%s doesn't exist, aborting\n", np->dev_name);
688                 err = -ENODEV;
689                 goto unlock;
690         }
691         netdev_hold(ndev, &np->dev_tracker, GFP_KERNEL);
692 
693         if (netdev_master_upper_dev_get(ndev)) {
694                 np_err(np, "%s is a slave device, aborting\n", np->dev_name);
695                 err = -EBUSY;
696                 goto put;
697         }
698 
699         if (!netif_running(ndev)) {
700                 unsigned long atmost;
701 
702                 np_info(np, "device %s not up yet, forcing it\n", np->dev_name);
703 
704                 err = dev_open(ndev, NULL);
705 
706                 if (err) {
707                         np_err(np, "failed to open %s\n", ndev->name);
708                         goto put;
709                 }
710 
711                 rtnl_unlock();
712                 atmost = jiffies + carrier_timeout * HZ;
713                 while (!netif_carrier_ok(ndev)) {
714                         if (time_after(jiffies, atmost)) {
715                                 np_notice(np, "timeout waiting for carrier\n");
716                                 break;
717                         }
718                         msleep(1);
719                 }
720 
721                 rtnl_lock();
722         }
723 
724         if (!np->local_ip.ip) {
725                 if (!np->ipv6) {
726                         const struct in_ifaddr *ifa;
727 
728                         in_dev = __in_dev_get_rtnl(ndev);
729                         if (!in_dev)
730                                 goto put_noaddr;
731 
732                         ifa = rtnl_dereference(in_dev->ifa_list);
733                         if (!ifa) {
734 put_noaddr:
735                                 np_err(np, "no IP address for %s, aborting\n",
736                                        np->dev_name);
737                                 err = -EDESTADDRREQ;
738                                 goto put;
739                         }
740 
741                         np->local_ip.ip = ifa->ifa_local;
742                         ip_overwritten = true;
743                         np_info(np, "local IP %pI4\n", &np->local_ip.ip);
744                 } else {
745 #if IS_ENABLED(CONFIG_IPV6)
746                         struct inet6_dev *idev;
747 
748                         err = -EDESTADDRREQ;
749                         idev = __in6_dev_get(ndev);
750                         if (idev) {
751                                 struct inet6_ifaddr *ifp;
752 
753                                 read_lock_bh(&idev->lock);
754                                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
755                                         if (!!(ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL) !=
756                                             !!(ipv6_addr_type(&np->remote_ip.in6) & IPV6_ADDR_LINKLOCAL))
757                                                 continue;
758                                         np->local_ip.in6 = ifp->addr;
759                                         ip_overwritten = true;
760                                         err = 0;
761                                         break;
762                                 }
763                                 read_unlock_bh(&idev->lock);
764                         }
765                         if (err) {
766                                 np_err(np, "no IPv6 address for %s, aborting\n",
767                                        np->dev_name);
768                                 goto put;
769                         } else
770                                 np_info(np, "local IPv6 %pI6c\n", &np->local_ip.in6);
771 #else
772                         np_err(np, "IPv6 is not supported %s, aborting\n",
773                                np->dev_name);
774                         err = -EINVAL;
775                         goto put;
776 #endif
777                 }
778         }
779 
780         /* fill up the skb queue */
781         refill_skbs();
782 
783         err = __netpoll_setup(np, ndev);
784         if (err)
785                 goto put;
786         rtnl_unlock();
787         return 0;
788 
789 put:
790         DEBUG_NET_WARN_ON_ONCE(np->dev);
791         if (ip_overwritten)
792                 memset(&np->local_ip, 0, sizeof(np->local_ip));
793         netdev_put(ndev, &np->dev_tracker);
794 unlock:
795         rtnl_unlock();
796         return err;
797 }
798 EXPORT_SYMBOL(netpoll_setup);
799 
800 static int __init netpoll_init(void)
801 {
802         skb_queue_head_init(&skb_pool);
803         return 0;
804 }
805 core_initcall(netpoll_init);
806 
807 static void rcu_cleanup_netpoll_info(struct rcu_head *rcu_head)
808 {
809         struct netpoll_info *npinfo =
810                         container_of(rcu_head, struct netpoll_info, rcu);
811 
812         skb_queue_purge(&npinfo->txq);
813 
814         /* we can't call cancel_delayed_work_sync here, as we are in softirq */
815         cancel_delayed_work(&npinfo->tx_work);
816 
817         /* clean after last, unfinished work */
818         __skb_queue_purge(&npinfo->txq);
819         /* now cancel it again */
820         cancel_delayed_work(&npinfo->tx_work);
821         kfree(npinfo);
822 }
823 
824 void __netpoll_cleanup(struct netpoll *np)
825 {
826         struct netpoll_info *npinfo;
827 
828         npinfo = rtnl_dereference(np->dev->npinfo);
829         if (!npinfo)
830                 return;
831 
832         synchronize_srcu(&netpoll_srcu);
833 
834         if (refcount_dec_and_test(&npinfo->refcnt)) {
835                 const struct net_device_ops *ops;
836 
837                 ops = np->dev->netdev_ops;
838                 if (ops->ndo_netpoll_cleanup)
839                         ops->ndo_netpoll_cleanup(np->dev);
840 
841                 RCU_INIT_POINTER(np->dev->npinfo, NULL);
842                 call_rcu(&npinfo->rcu, rcu_cleanup_netpoll_info);
843         } else
844                 RCU_INIT_POINTER(np->dev->npinfo, NULL);
845 }
846 EXPORT_SYMBOL_GPL(__netpoll_cleanup);
847 
848 void __netpoll_free(struct netpoll *np)
849 {
850         ASSERT_RTNL();
851 
852         /* Wait for transmitting packets to finish before freeing. */
853         synchronize_rcu();
854         __netpoll_cleanup(np);
855         kfree(np);
856 }
857 EXPORT_SYMBOL_GPL(__netpoll_free);
858 
859 void netpoll_cleanup(struct netpoll *np)
860 {
861         rtnl_lock();
862         if (!np->dev)
863                 goto out;
864         __netpoll_cleanup(np);
865         netdev_put(np->dev, &np->dev_tracker);
866         np->dev = NULL;
867 out:
868         rtnl_unlock();
869 }
870 EXPORT_SYMBOL(netpoll_cleanup);
871 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

sflogo.php