interface.c 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793
  1. /*
  2. * Network-device interface management.
  3. *
  4. * Copyright (c) 2004-2005, Keir Fraser
  5. *
  6. * This program is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU General Public License version 2
  8. * as published by the Free Software Foundation; or, when distributed
  9. * separately from the Linux kernel or incorporated into other
  10. * software packages, subject to the following license:
  11. *
  12. * Permission is hereby granted, free of charge, to any person obtaining a copy
  13. * of this source file (the "Software"), to deal in the Software without
  14. * restriction, including without limitation the rights to use, copy, modify,
  15. * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  16. * and to permit persons to whom the Software is furnished to do so, subject to
  17. * the following conditions:
  18. *
  19. * The above copyright notice and this permission notice shall be included in
  20. * all copies or substantial portions of the Software.
  21. *
  22. * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  23. * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  24. * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  25. * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  26. * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  27. * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  28. * IN THE SOFTWARE.
  29. */
  30. #include "common.h"
  31. #include <linux/kthread.h>
  32. #include <linux/ethtool.h>
  33. #include <linux/rtnetlink.h>
  34. #include <linux/if_vlan.h>
  35. #include <linux/vmalloc.h>
  36. #include <xen/events.h>
  37. #include <asm/xen/hypercall.h>
  38. #include <xen/balloon.h>
  39. #define XENVIF_QUEUE_LENGTH 32
  40. #define XENVIF_NAPI_WEIGHT 64
  41. /* Number of bytes allowed on the internal guest Rx queue. */
  42. #define XENVIF_RX_QUEUE_BYTES (XEN_NETIF_RX_RING_SIZE/2 * PAGE_SIZE)
  43. /* This function is used to set SKBTX_DEV_ZEROCOPY as well as
  44. * increasing the inflight counter. We need to increase the inflight
  45. * counter because core driver calls into xenvif_zerocopy_callback
  46. * which calls xenvif_skb_zerocopy_complete.
  47. */
  48. void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue,
  49. struct sk_buff *skb)
  50. {
  51. skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
  52. atomic_inc(&queue->inflight_packets);
  53. }
  54. void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue)
  55. {
  56. atomic_dec(&queue->inflight_packets);
  57. /* Wake the dealloc thread _after_ decrementing inflight_packets so
  58. * that if kthread_stop() has already been called, the dealloc thread
  59. * does not wait forever with nothing to wake it.
  60. */
  61. wake_up(&queue->dealloc_wq);
  62. }
  63. int xenvif_schedulable(struct xenvif *vif)
  64. {
  65. return netif_running(vif->dev) &&
  66. test_bit(VIF_STATUS_CONNECTED, &vif->status) &&
  67. !vif->disabled;
  68. }
  69. static irqreturn_t xenvif_tx_interrupt(int irq, void *dev_id)
  70. {
  71. struct xenvif_queue *queue = dev_id;
  72. if (RING_HAS_UNCONSUMED_REQUESTS(&queue->tx))
  73. napi_schedule(&queue->napi);
  74. return IRQ_HANDLED;
  75. }
  76. static int xenvif_poll(struct napi_struct *napi, int budget)
  77. {
  78. struct xenvif_queue *queue =
  79. container_of(napi, struct xenvif_queue, napi);
  80. int work_done;
  81. /* This vif is rogue, we pretend we've there is nothing to do
  82. * for this vif to deschedule it from NAPI. But this interface
  83. * will be turned off in thread context later.
  84. */
  85. if (unlikely(queue->vif->disabled)) {
  86. napi_complete(napi);
  87. return 0;
  88. }
  89. work_done = xenvif_tx_action(queue, budget);
  90. if (work_done < budget) {
  91. napi_complete(napi);
  92. xenvif_napi_schedule_or_enable_events(queue);
  93. }
  94. return work_done;
  95. }
  96. static irqreturn_t xenvif_rx_interrupt(int irq, void *dev_id)
  97. {
  98. struct xenvif_queue *queue = dev_id;
  99. xenvif_kick_thread(queue);
  100. return IRQ_HANDLED;
  101. }
  102. irqreturn_t xenvif_interrupt(int irq, void *dev_id)
  103. {
  104. xenvif_tx_interrupt(irq, dev_id);
  105. xenvif_rx_interrupt(irq, dev_id);
  106. return IRQ_HANDLED;
  107. }
  108. int xenvif_queue_stopped(struct xenvif_queue *queue)
  109. {
  110. struct net_device *dev = queue->vif->dev;
  111. unsigned int id = queue->id;
  112. return netif_tx_queue_stopped(netdev_get_tx_queue(dev, id));
  113. }
  114. void xenvif_wake_queue(struct xenvif_queue *queue)
  115. {
  116. struct net_device *dev = queue->vif->dev;
  117. unsigned int id = queue->id;
  118. netif_tx_wake_queue(netdev_get_tx_queue(dev, id));
  119. }
  120. static u16 xenvif_select_queue(struct net_device *dev, struct sk_buff *skb,
  121. void *accel_priv,
  122. select_queue_fallback_t fallback)
  123. {
  124. struct xenvif *vif = netdev_priv(dev);
  125. unsigned int size = vif->hash.size;
  126. if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE) {
  127. u16 index = fallback(dev, skb) % dev->real_num_tx_queues;
  128. /* Make sure there is no hash information in the socket
  129. * buffer otherwise it would be incorrectly forwarded
  130. * to the frontend.
  131. */
  132. skb_clear_hash(skb);
  133. return index;
  134. }
  135. xenvif_set_skb_hash(vif, skb);
  136. if (size == 0)
  137. return skb_get_hash_raw(skb) % dev->real_num_tx_queues;
  138. return vif->hash.mapping[skb_get_hash_raw(skb) % size];
  139. }
  140. static int xenvif_start_xmit(struct sk_buff *skb, struct net_device *dev)
  141. {
  142. struct xenvif *vif = netdev_priv(dev);
  143. struct xenvif_queue *queue = NULL;
  144. unsigned int num_queues = vif->num_queues;
  145. u16 index;
  146. struct xenvif_rx_cb *cb;
  147. BUG_ON(skb->dev != dev);
  148. /* Drop the packet if queues are not set up */
  149. if (num_queues < 1)
  150. goto drop;
  151. /* Obtain the queue to be used to transmit this packet */
  152. index = skb_get_queue_mapping(skb);
  153. if (index >= num_queues) {
  154. pr_warn_ratelimited("Invalid queue %hu for packet on interface %s\n.",
  155. index, vif->dev->name);
  156. index %= num_queues;
  157. }
  158. queue = &vif->queues[index];
  159. /* Drop the packet if queue is not ready */
  160. if (queue->task == NULL ||
  161. queue->dealloc_task == NULL ||
  162. !xenvif_schedulable(vif))
  163. goto drop;
  164. if (vif->multicast_control && skb->pkt_type == PACKET_MULTICAST) {
  165. struct ethhdr *eth = (struct ethhdr *)skb->data;
  166. if (!xenvif_mcast_match(vif, eth->h_dest))
  167. goto drop;
  168. }
  169. cb = XENVIF_RX_CB(skb);
  170. cb->expires = jiffies + vif->drain_timeout;
  171. xenvif_rx_queue_tail(queue, skb);
  172. xenvif_kick_thread(queue);
  173. return NETDEV_TX_OK;
  174. drop:
  175. vif->dev->stats.tx_dropped++;
  176. dev_kfree_skb(skb);
  177. return NETDEV_TX_OK;
  178. }
  179. static struct net_device_stats *xenvif_get_stats(struct net_device *dev)
  180. {
  181. struct xenvif *vif = netdev_priv(dev);
  182. struct xenvif_queue *queue = NULL;
  183. unsigned int num_queues = vif->num_queues;
  184. unsigned long rx_bytes = 0;
  185. unsigned long rx_packets = 0;
  186. unsigned long tx_bytes = 0;
  187. unsigned long tx_packets = 0;
  188. unsigned int index;
  189. if (vif->queues == NULL)
  190. goto out;
  191. /* Aggregate tx and rx stats from each queue */
  192. for (index = 0; index < num_queues; ++index) {
  193. queue = &vif->queues[index];
  194. rx_bytes += queue->stats.rx_bytes;
  195. rx_packets += queue->stats.rx_packets;
  196. tx_bytes += queue->stats.tx_bytes;
  197. tx_packets += queue->stats.tx_packets;
  198. }
  199. out:
  200. vif->dev->stats.rx_bytes = rx_bytes;
  201. vif->dev->stats.rx_packets = rx_packets;
  202. vif->dev->stats.tx_bytes = tx_bytes;
  203. vif->dev->stats.tx_packets = tx_packets;
  204. return &vif->dev->stats;
  205. }
  206. static void xenvif_up(struct xenvif *vif)
  207. {
  208. struct xenvif_queue *queue = NULL;
  209. unsigned int num_queues = vif->num_queues;
  210. unsigned int queue_index;
  211. for (queue_index = 0; queue_index < num_queues; ++queue_index) {
  212. queue = &vif->queues[queue_index];
  213. napi_enable(&queue->napi);
  214. enable_irq(queue->tx_irq);
  215. if (queue->tx_irq != queue->rx_irq)
  216. enable_irq(queue->rx_irq);
  217. xenvif_napi_schedule_or_enable_events(queue);
  218. }
  219. }
  220. static void xenvif_down(struct xenvif *vif)
  221. {
  222. struct xenvif_queue *queue = NULL;
  223. unsigned int num_queues = vif->num_queues;
  224. unsigned int queue_index;
  225. for (queue_index = 0; queue_index < num_queues; ++queue_index) {
  226. queue = &vif->queues[queue_index];
  227. disable_irq(queue->tx_irq);
  228. if (queue->tx_irq != queue->rx_irq)
  229. disable_irq(queue->rx_irq);
  230. napi_disable(&queue->napi);
  231. del_timer_sync(&queue->credit_timeout);
  232. }
  233. }
  234. static int xenvif_open(struct net_device *dev)
  235. {
  236. struct xenvif *vif = netdev_priv(dev);
  237. if (test_bit(VIF_STATUS_CONNECTED, &vif->status))
  238. xenvif_up(vif);
  239. netif_tx_start_all_queues(dev);
  240. return 0;
  241. }
  242. static int xenvif_close(struct net_device *dev)
  243. {
  244. struct xenvif *vif = netdev_priv(dev);
  245. if (test_bit(VIF_STATUS_CONNECTED, &vif->status))
  246. xenvif_down(vif);
  247. netif_tx_stop_all_queues(dev);
  248. return 0;
  249. }
  250. static int xenvif_change_mtu(struct net_device *dev, int mtu)
  251. {
  252. struct xenvif *vif = netdev_priv(dev);
  253. int max = vif->can_sg ? 65535 - VLAN_ETH_HLEN : ETH_DATA_LEN;
  254. if (mtu > max)
  255. return -EINVAL;
  256. dev->mtu = mtu;
  257. return 0;
  258. }
  259. static netdev_features_t xenvif_fix_features(struct net_device *dev,
  260. netdev_features_t features)
  261. {
  262. struct xenvif *vif = netdev_priv(dev);
  263. if (!vif->can_sg)
  264. features &= ~NETIF_F_SG;
  265. if (~(vif->gso_mask) & GSO_BIT(TCPV4))
  266. features &= ~NETIF_F_TSO;
  267. if (~(vif->gso_mask) & GSO_BIT(TCPV6))
  268. features &= ~NETIF_F_TSO6;
  269. if (!vif->ip_csum)
  270. features &= ~NETIF_F_IP_CSUM;
  271. if (!vif->ipv6_csum)
  272. features &= ~NETIF_F_IPV6_CSUM;
  273. return features;
  274. }
  275. static const struct xenvif_stat {
  276. char name[ETH_GSTRING_LEN];
  277. u16 offset;
  278. } xenvif_stats[] = {
  279. {
  280. "rx_gso_checksum_fixup",
  281. offsetof(struct xenvif_stats, rx_gso_checksum_fixup)
  282. },
  283. /* If (sent != success + fail), there are probably packets never
  284. * freed up properly!
  285. */
  286. {
  287. "tx_zerocopy_sent",
  288. offsetof(struct xenvif_stats, tx_zerocopy_sent),
  289. },
  290. {
  291. "tx_zerocopy_success",
  292. offsetof(struct xenvif_stats, tx_zerocopy_success),
  293. },
  294. {
  295. "tx_zerocopy_fail",
  296. offsetof(struct xenvif_stats, tx_zerocopy_fail)
  297. },
  298. /* Number of packets exceeding MAX_SKB_FRAG slots. You should use
  299. * a guest with the same MAX_SKB_FRAG
  300. */
  301. {
  302. "tx_frag_overflow",
  303. offsetof(struct xenvif_stats, tx_frag_overflow)
  304. },
  305. };
  306. static int xenvif_get_sset_count(struct net_device *dev, int string_set)
  307. {
  308. switch (string_set) {
  309. case ETH_SS_STATS:
  310. return ARRAY_SIZE(xenvif_stats);
  311. default:
  312. return -EINVAL;
  313. }
  314. }
  315. static void xenvif_get_ethtool_stats(struct net_device *dev,
  316. struct ethtool_stats *stats, u64 * data)
  317. {
  318. struct xenvif *vif = netdev_priv(dev);
  319. unsigned int num_queues = vif->num_queues;
  320. int i;
  321. unsigned int queue_index;
  322. for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) {
  323. unsigned long accum = 0;
  324. for (queue_index = 0; queue_index < num_queues; ++queue_index) {
  325. void *vif_stats = &vif->queues[queue_index].stats;
  326. accum += *(unsigned long *)(vif_stats + xenvif_stats[i].offset);
  327. }
  328. data[i] = accum;
  329. }
  330. }
  331. static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data)
  332. {
  333. int i;
  334. switch (stringset) {
  335. case ETH_SS_STATS:
  336. for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++)
  337. memcpy(data + i * ETH_GSTRING_LEN,
  338. xenvif_stats[i].name, ETH_GSTRING_LEN);
  339. break;
  340. }
  341. }
  342. static const struct ethtool_ops xenvif_ethtool_ops = {
  343. .get_link = ethtool_op_get_link,
  344. .get_sset_count = xenvif_get_sset_count,
  345. .get_ethtool_stats = xenvif_get_ethtool_stats,
  346. .get_strings = xenvif_get_strings,
  347. };
  348. static const struct net_device_ops xenvif_netdev_ops = {
  349. .ndo_select_queue = xenvif_select_queue,
  350. .ndo_start_xmit = xenvif_start_xmit,
  351. .ndo_get_stats = xenvif_get_stats,
  352. .ndo_open = xenvif_open,
  353. .ndo_stop = xenvif_close,
  354. .ndo_change_mtu = xenvif_change_mtu,
  355. .ndo_fix_features = xenvif_fix_features,
  356. .ndo_set_mac_address = eth_mac_addr,
  357. .ndo_validate_addr = eth_validate_addr,
  358. };
  359. struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
  360. unsigned int handle)
  361. {
  362. int err;
  363. struct net_device *dev;
  364. struct xenvif *vif;
  365. char name[IFNAMSIZ] = {};
  366. snprintf(name, IFNAMSIZ - 1, "vif%u.%u", domid, handle);
  367. /* Allocate a netdev with the max. supported number of queues.
  368. * When the guest selects the desired number, it will be updated
  369. * via netif_set_real_num_*_queues().
  370. */
  371. dev = alloc_netdev_mq(sizeof(struct xenvif), name, NET_NAME_UNKNOWN,
  372. ether_setup, xenvif_max_queues);
  373. if (dev == NULL) {
  374. pr_warn("Could not allocate netdev for %s\n", name);
  375. return ERR_PTR(-ENOMEM);
  376. }
  377. SET_NETDEV_DEV(dev, parent);
  378. vif = netdev_priv(dev);
  379. vif->domid = domid;
  380. vif->handle = handle;
  381. vif->can_sg = 1;
  382. vif->ip_csum = 1;
  383. vif->dev = dev;
  384. vif->disabled = false;
  385. vif->drain_timeout = msecs_to_jiffies(rx_drain_timeout_msecs);
  386. vif->stall_timeout = msecs_to_jiffies(rx_stall_timeout_msecs);
  387. /* Start out with no queues. */
  388. vif->queues = NULL;
  389. vif->num_queues = 0;
  390. spin_lock_init(&vif->lock);
  391. INIT_LIST_HEAD(&vif->fe_mcast_addr);
  392. dev->netdev_ops = &xenvif_netdev_ops;
  393. dev->hw_features = NETIF_F_SG |
  394. NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
  395. NETIF_F_TSO | NETIF_F_TSO6;
  396. dev->features = dev->hw_features | NETIF_F_RXCSUM;
  397. dev->ethtool_ops = &xenvif_ethtool_ops;
  398. dev->tx_queue_len = XENVIF_QUEUE_LENGTH;
  399. /*
  400. * Initialise a dummy MAC address. We choose the numerically
  401. * largest non-broadcast address to prevent the address getting
  402. * stolen by an Ethernet bridge for STP purposes.
  403. * (FE:FF:FF:FF:FF:FF)
  404. */
  405. eth_broadcast_addr(dev->dev_addr);
  406. dev->dev_addr[0] &= ~0x01;
  407. netif_carrier_off(dev);
  408. err = register_netdev(dev);
  409. if (err) {
  410. netdev_warn(dev, "Could not register device: err=%d\n", err);
  411. free_netdev(dev);
  412. return ERR_PTR(err);
  413. }
  414. netdev_dbg(dev, "Successfully created xenvif\n");
  415. __module_get(THIS_MODULE);
  416. return vif;
  417. }
  418. int xenvif_init_queue(struct xenvif_queue *queue)
  419. {
  420. int err, i;
  421. queue->credit_bytes = queue->remaining_credit = ~0UL;
  422. queue->credit_usec = 0UL;
  423. init_timer(&queue->credit_timeout);
  424. queue->credit_timeout.function = xenvif_tx_credit_callback;
  425. queue->credit_window_start = get_jiffies_64();
  426. queue->rx_queue_max = XENVIF_RX_QUEUE_BYTES;
  427. skb_queue_head_init(&queue->rx_queue);
  428. skb_queue_head_init(&queue->tx_queue);
  429. queue->pending_cons = 0;
  430. queue->pending_prod = MAX_PENDING_REQS;
  431. for (i = 0; i < MAX_PENDING_REQS; ++i)
  432. queue->pending_ring[i] = i;
  433. spin_lock_init(&queue->callback_lock);
  434. spin_lock_init(&queue->response_lock);
  435. /* If ballooning is disabled, this will consume real memory, so you
  436. * better enable it. The long term solution would be to use just a
  437. * bunch of valid page descriptors, without dependency on ballooning
  438. */
  439. err = gnttab_alloc_pages(MAX_PENDING_REQS,
  440. queue->mmap_pages);
  441. if (err) {
  442. netdev_err(queue->vif->dev, "Could not reserve mmap_pages\n");
  443. return -ENOMEM;
  444. }
  445. for (i = 0; i < MAX_PENDING_REQS; i++) {
  446. queue->pending_tx_info[i].callback_struct = (struct ubuf_info)
  447. { .callback = xenvif_zerocopy_callback,
  448. .ctx = NULL,
  449. .desc = i };
  450. queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE;
  451. }
  452. return 0;
  453. }
  454. void xenvif_carrier_on(struct xenvif *vif)
  455. {
  456. rtnl_lock();
  457. if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN)
  458. dev_set_mtu(vif->dev, ETH_DATA_LEN);
  459. netdev_update_features(vif->dev);
  460. set_bit(VIF_STATUS_CONNECTED, &vif->status);
  461. if (netif_running(vif->dev))
  462. xenvif_up(vif);
  463. rtnl_unlock();
  464. }
  465. int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
  466. unsigned int evtchn)
  467. {
  468. struct net_device *dev = vif->dev;
  469. void *addr;
  470. struct xen_netif_ctrl_sring *shared;
  471. int err;
  472. err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(vif),
  473. &ring_ref, 1, &addr);
  474. if (err)
  475. goto err;
  476. shared = (struct xen_netif_ctrl_sring *)addr;
  477. BACK_RING_INIT(&vif->ctrl, shared, XEN_PAGE_SIZE);
  478. err = bind_interdomain_evtchn_to_irq(vif->domid, evtchn);
  479. if (err < 0)
  480. goto err_unmap;
  481. vif->ctrl_irq = err;
  482. xenvif_init_hash(vif);
  483. err = request_threaded_irq(vif->ctrl_irq, NULL, xenvif_ctrl_irq_fn,
  484. IRQF_ONESHOT, "xen-netback-ctrl", vif);
  485. if (err) {
  486. pr_warn("Could not setup irq handler for %s\n", dev->name);
  487. goto err_deinit;
  488. }
  489. return 0;
  490. err_deinit:
  491. xenvif_deinit_hash(vif);
  492. unbind_from_irqhandler(vif->ctrl_irq, vif);
  493. vif->ctrl_irq = 0;
  494. err_unmap:
  495. xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
  496. vif->ctrl.sring);
  497. vif->ctrl.sring = NULL;
  498. err:
  499. return err;
  500. }
  501. int xenvif_connect_data(struct xenvif_queue *queue,
  502. unsigned long tx_ring_ref,
  503. unsigned long rx_ring_ref,
  504. unsigned int tx_evtchn,
  505. unsigned int rx_evtchn)
  506. {
  507. struct task_struct *task;
  508. int err = -ENOMEM;
  509. BUG_ON(queue->tx_irq);
  510. BUG_ON(queue->task);
  511. BUG_ON(queue->dealloc_task);
  512. err = xenvif_map_frontend_data_rings(queue, tx_ring_ref,
  513. rx_ring_ref);
  514. if (err < 0)
  515. goto err;
  516. init_waitqueue_head(&queue->wq);
  517. init_waitqueue_head(&queue->dealloc_wq);
  518. atomic_set(&queue->inflight_packets, 0);
  519. netif_napi_add(queue->vif->dev, &queue->napi, xenvif_poll,
  520. XENVIF_NAPI_WEIGHT);
  521. if (tx_evtchn == rx_evtchn) {
  522. /* feature-split-event-channels == 0 */
  523. err = bind_interdomain_evtchn_to_irqhandler(
  524. queue->vif->domid, tx_evtchn, xenvif_interrupt, 0,
  525. queue->name, queue);
  526. if (err < 0)
  527. goto err_unmap;
  528. queue->tx_irq = queue->rx_irq = err;
  529. disable_irq(queue->tx_irq);
  530. } else {
  531. /* feature-split-event-channels == 1 */
  532. snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name),
  533. "%s-tx", queue->name);
  534. err = bind_interdomain_evtchn_to_irqhandler(
  535. queue->vif->domid, tx_evtchn, xenvif_tx_interrupt, 0,
  536. queue->tx_irq_name, queue);
  537. if (err < 0)
  538. goto err_unmap;
  539. queue->tx_irq = err;
  540. disable_irq(queue->tx_irq);
  541. snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name),
  542. "%s-rx", queue->name);
  543. err = bind_interdomain_evtchn_to_irqhandler(
  544. queue->vif->domid, rx_evtchn, xenvif_rx_interrupt, 0,
  545. queue->rx_irq_name, queue);
  546. if (err < 0)
  547. goto err_tx_unbind;
  548. queue->rx_irq = err;
  549. disable_irq(queue->rx_irq);
  550. }
  551. queue->stalled = true;
  552. task = kthread_create(xenvif_kthread_guest_rx,
  553. (void *)queue, "%s-guest-rx", queue->name);
  554. if (IS_ERR(task)) {
  555. pr_warn("Could not allocate kthread for %s\n", queue->name);
  556. err = PTR_ERR(task);
  557. goto err_rx_unbind;
  558. }
  559. queue->task = task;
  560. get_task_struct(task);
  561. task = kthread_create(xenvif_dealloc_kthread,
  562. (void *)queue, "%s-dealloc", queue->name);
  563. if (IS_ERR(task)) {
  564. pr_warn("Could not allocate kthread for %s\n", queue->name);
  565. err = PTR_ERR(task);
  566. goto err_rx_unbind;
  567. }
  568. queue->dealloc_task = task;
  569. wake_up_process(queue->task);
  570. wake_up_process(queue->dealloc_task);
  571. return 0;
  572. err_rx_unbind:
  573. unbind_from_irqhandler(queue->rx_irq, queue);
  574. queue->rx_irq = 0;
  575. err_tx_unbind:
  576. unbind_from_irqhandler(queue->tx_irq, queue);
  577. queue->tx_irq = 0;
  578. err_unmap:
  579. xenvif_unmap_frontend_data_rings(queue);
  580. netif_napi_del(&queue->napi);
  581. err:
  582. module_put(THIS_MODULE);
  583. return err;
  584. }
  585. void xenvif_carrier_off(struct xenvif *vif)
  586. {
  587. struct net_device *dev = vif->dev;
  588. rtnl_lock();
  589. if (test_and_clear_bit(VIF_STATUS_CONNECTED, &vif->status)) {
  590. netif_carrier_off(dev); /* discard queued packets */
  591. if (netif_running(dev))
  592. xenvif_down(vif);
  593. }
  594. rtnl_unlock();
  595. }
  596. void xenvif_disconnect_data(struct xenvif *vif)
  597. {
  598. struct xenvif_queue *queue = NULL;
  599. unsigned int num_queues = vif->num_queues;
  600. unsigned int queue_index;
  601. xenvif_carrier_off(vif);
  602. for (queue_index = 0; queue_index < num_queues; ++queue_index) {
  603. queue = &vif->queues[queue_index];
  604. netif_napi_del(&queue->napi);
  605. if (queue->task) {
  606. kthread_stop(queue->task);
  607. put_task_struct(queue->task);
  608. queue->task = NULL;
  609. }
  610. if (queue->dealloc_task) {
  611. kthread_stop(queue->dealloc_task);
  612. queue->dealloc_task = NULL;
  613. }
  614. if (queue->tx_irq) {
  615. if (queue->tx_irq == queue->rx_irq)
  616. unbind_from_irqhandler(queue->tx_irq, queue);
  617. else {
  618. unbind_from_irqhandler(queue->tx_irq, queue);
  619. unbind_from_irqhandler(queue->rx_irq, queue);
  620. }
  621. queue->tx_irq = 0;
  622. }
  623. xenvif_unmap_frontend_data_rings(queue);
  624. }
  625. xenvif_mcast_addr_list_free(vif);
  626. }
  627. void xenvif_disconnect_ctrl(struct xenvif *vif)
  628. {
  629. if (vif->ctrl_irq) {
  630. xenvif_deinit_hash(vif);
  631. unbind_from_irqhandler(vif->ctrl_irq, vif);
  632. vif->ctrl_irq = 0;
  633. }
  634. if (vif->ctrl.sring) {
  635. xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
  636. vif->ctrl.sring);
  637. vif->ctrl.sring = NULL;
  638. }
  639. }
  640. /* Reverse the relevant parts of xenvif_init_queue().
  641. * Used for queue teardown from xenvif_free(), and on the
  642. * error handling paths in xenbus.c:connect().
  643. */
  644. void xenvif_deinit_queue(struct xenvif_queue *queue)
  645. {
  646. gnttab_free_pages(MAX_PENDING_REQS, queue->mmap_pages);
  647. }
  648. void xenvif_free(struct xenvif *vif)
  649. {
  650. struct xenvif_queue *queues = vif->queues;
  651. unsigned int num_queues = vif->num_queues;
  652. unsigned int queue_index;
  653. unregister_netdev(vif->dev);
  654. free_netdev(vif->dev);
  655. for (queue_index = 0; queue_index < num_queues; ++queue_index)
  656. xenvif_deinit_queue(&queues[queue_index]);
  657. vfree(queues);
  658. module_put(THIS_MODULE);
  659. }