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 
31 #include "common.h"
32 
33 #include <linux/kthread.h>
34 #include <linux/sched/task.h>
35 #include <linux/ethtool.h>
36 #include <linux/rtnetlink.h>
37 #include <linux/if_vlan.h>
38 #include <linux/vmalloc.h>
39 
40 #include <xen/events.h>
41 #include <asm/xen/hypercall.h>
42 #include <xen/balloon.h>
43 
44 #define XENVIF_QUEUE_LENGTH 32
45 #define XENVIF_NAPI_WEIGHT  64
46 
47 /* Number of bytes allowed on the internal guest Rx queue. */
48 #define XENVIF_RX_QUEUE_BYTES (XEN_NETIF_RX_RING_SIZE/2 * PAGE_SIZE)
49 
50 /* This function is used to set SKBTX_DEV_ZEROCOPY as well as
51  * increasing the inflight counter. We need to increase the inflight
52  * counter because core driver calls into xenvif_zerocopy_callback
53  * which calls xenvif_skb_zerocopy_complete.
54  */
55 void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue,
56 				 struct sk_buff *skb)
57 {
58 	skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY;
59 	atomic_inc(&queue->inflight_packets);
60 }
61 
62 void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue)
63 {
64 	atomic_dec(&queue->inflight_packets);
65 
66 	/* Wake the dealloc thread _after_ decrementing inflight_packets so
67 	 * that if kthread_stop() has already been called, the dealloc thread
68 	 * does not wait forever with nothing to wake it.
69 	 */
70 	wake_up(&queue->dealloc_wq);
71 }
72 
73 int xenvif_schedulable(struct xenvif *vif)
74 {
75 	return netif_running(vif->dev) &&
76 		test_bit(VIF_STATUS_CONNECTED, &vif->status) &&
77 		!vif->disabled;
78 }
79 
80 static irqreturn_t xenvif_tx_interrupt(int irq, void *dev_id)
81 {
82 	struct xenvif_queue *queue = dev_id;
83 
84 	if (RING_HAS_UNCONSUMED_REQUESTS(&queue->tx))
85 		napi_schedule(&queue->napi);
86 
87 	return IRQ_HANDLED;
88 }
89 
90 static int xenvif_poll(struct napi_struct *napi, int budget)
91 {
92 	struct xenvif_queue *queue =
93 		container_of(napi, struct xenvif_queue, napi);
94 	int work_done;
95 
96 	/* This vif is rogue, we pretend we've there is nothing to do
97 	 * for this vif to deschedule it from NAPI. But this interface
98 	 * will be turned off in thread context later.
99 	 */
100 	if (unlikely(queue->vif->disabled)) {
101 		napi_complete(napi);
102 		return 0;
103 	}
104 
105 	work_done = xenvif_tx_action(queue, budget);
106 
107 	if (work_done < budget) {
108 		napi_complete_done(napi, work_done);
109 		/* If the queue is rate-limited, it shall be
110 		 * rescheduled in the timer callback.
111 		 */
112 		if (likely(!queue->rate_limited))
113 			xenvif_napi_schedule_or_enable_events(queue);
114 	}
115 
116 	return work_done;
117 }
118 
119 static irqreturn_t xenvif_rx_interrupt(int irq, void *dev_id)
120 {
121 	struct xenvif_queue *queue = dev_id;
122 
123 	xenvif_kick_thread(queue);
124 
125 	return IRQ_HANDLED;
126 }
127 
128 irqreturn_t xenvif_interrupt(int irq, void *dev_id)
129 {
130 	xenvif_tx_interrupt(irq, dev_id);
131 	xenvif_rx_interrupt(irq, dev_id);
132 
133 	return IRQ_HANDLED;
134 }
135 
136 int xenvif_queue_stopped(struct xenvif_queue *queue)
137 {
138 	struct net_device *dev = queue->vif->dev;
139 	unsigned int id = queue->id;
140 	return netif_tx_queue_stopped(netdev_get_tx_queue(dev, id));
141 }
142 
143 void xenvif_wake_queue(struct xenvif_queue *queue)
144 {
145 	struct net_device *dev = queue->vif->dev;
146 	unsigned int id = queue->id;
147 	netif_tx_wake_queue(netdev_get_tx_queue(dev, id));
148 }
149 
150 static u16 xenvif_select_queue(struct net_device *dev, struct sk_buff *skb,
151 			       struct net_device *sb_dev)
152 {
153 	struct xenvif *vif = netdev_priv(dev);
154 	unsigned int size = vif->hash.size;
155 	unsigned int num_queues;
156 
157 	/* If queues are not set up internally - always return 0
158 	 * as the packet going to be dropped anyway */
159 	num_queues = READ_ONCE(vif->num_queues);
160 	if (num_queues < 1)
161 		return 0;
162 
163 	if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE)
164 		return netdev_pick_tx(dev, skb, NULL) %
165 		       dev->real_num_tx_queues;
166 
167 	xenvif_set_skb_hash(vif, skb);
168 
169 	if (size == 0)
170 		return skb_get_hash_raw(skb) % dev->real_num_tx_queues;
171 
172 	return vif->hash.mapping[vif->hash.mapping_sel]
173 				[skb_get_hash_raw(skb) % size];
174 }
175 
176 static netdev_tx_t
177 xenvif_start_xmit(struct sk_buff *skb, struct net_device *dev)
178 {
179 	struct xenvif *vif = netdev_priv(dev);
180 	struct xenvif_queue *queue = NULL;
181 	unsigned int num_queues;
182 	u16 index;
183 	struct xenvif_rx_cb *cb;
184 
185 	BUG_ON(skb->dev != dev);
186 
187 	/* Drop the packet if queues are not set up.
188 	 * This handler should be called inside an RCU read section
189 	 * so we don't need to enter it here explicitly.
190 	 */
191 	num_queues = READ_ONCE(vif->num_queues);
192 	if (num_queues < 1)
193 		goto drop;
194 
195 	/* Obtain the queue to be used to transmit this packet */
196 	index = skb_get_queue_mapping(skb);
197 	if (index >= num_queues) {
198 		pr_warn_ratelimited("Invalid queue %hu for packet on interface %s\n",
199 				    index, vif->dev->name);
200 		index %= num_queues;
201 	}
202 	queue = &vif->queues[index];
203 
204 	/* Drop the packet if queue is not ready */
205 	if (queue->task == NULL ||
206 	    queue->dealloc_task == NULL ||
207 	    !xenvif_schedulable(vif))
208 		goto drop;
209 
210 	if (vif->multicast_control && skb->pkt_type == PACKET_MULTICAST) {
211 		struct ethhdr *eth = (struct ethhdr *)skb->data;
212 
213 		if (!xenvif_mcast_match(vif, eth->h_dest))
214 			goto drop;
215 	}
216 
217 	cb = XENVIF_RX_CB(skb);
218 	cb->expires = jiffies + vif->drain_timeout;
219 
220 	/* If there is no hash algorithm configured then make sure there
221 	 * is no hash information in the socket buffer otherwise it
222 	 * would be incorrectly forwarded to the frontend.
223 	 */
224 	if (vif->hash.alg == XEN_NETIF_CTRL_HASH_ALGORITHM_NONE)
225 		skb_clear_hash(skb);
226 
227 	xenvif_rx_queue_tail(queue, skb);
228 	xenvif_kick_thread(queue);
229 
230 	return NETDEV_TX_OK;
231 
232  drop:
233 	vif->dev->stats.tx_dropped++;
234 	dev_kfree_skb(skb);
235 	return NETDEV_TX_OK;
236 }
237 
238 static struct net_device_stats *xenvif_get_stats(struct net_device *dev)
239 {
240 	struct xenvif *vif = netdev_priv(dev);
241 	struct xenvif_queue *queue = NULL;
242 	unsigned int num_queues;
243 	u64 rx_bytes = 0;
244 	u64 rx_packets = 0;
245 	u64 tx_bytes = 0;
246 	u64 tx_packets = 0;
247 	unsigned int index;
248 
249 	rcu_read_lock();
250 	num_queues = READ_ONCE(vif->num_queues);
251 
252 	/* Aggregate tx and rx stats from each queue */
253 	for (index = 0; index < num_queues; ++index) {
254 		queue = &vif->queues[index];
255 		rx_bytes += queue->stats.rx_bytes;
256 		rx_packets += queue->stats.rx_packets;
257 		tx_bytes += queue->stats.tx_bytes;
258 		tx_packets += queue->stats.tx_packets;
259 	}
260 
261 	rcu_read_unlock();
262 
263 	vif->dev->stats.rx_bytes = rx_bytes;
264 	vif->dev->stats.rx_packets = rx_packets;
265 	vif->dev->stats.tx_bytes = tx_bytes;
266 	vif->dev->stats.tx_packets = tx_packets;
267 
268 	return &vif->dev->stats;
269 }
270 
271 static void xenvif_up(struct xenvif *vif)
272 {
273 	struct xenvif_queue *queue = NULL;
274 	unsigned int num_queues = vif->num_queues;
275 	unsigned int queue_index;
276 
277 	for (queue_index = 0; queue_index < num_queues; ++queue_index) {
278 		queue = &vif->queues[queue_index];
279 		napi_enable(&queue->napi);
280 		enable_irq(queue->tx_irq);
281 		if (queue->tx_irq != queue->rx_irq)
282 			enable_irq(queue->rx_irq);
283 		xenvif_napi_schedule_or_enable_events(queue);
284 	}
285 }
286 
287 static void xenvif_down(struct xenvif *vif)
288 {
289 	struct xenvif_queue *queue = NULL;
290 	unsigned int num_queues = vif->num_queues;
291 	unsigned int queue_index;
292 
293 	for (queue_index = 0; queue_index < num_queues; ++queue_index) {
294 		queue = &vif->queues[queue_index];
295 		disable_irq(queue->tx_irq);
296 		if (queue->tx_irq != queue->rx_irq)
297 			disable_irq(queue->rx_irq);
298 		napi_disable(&queue->napi);
299 		del_timer_sync(&queue->credit_timeout);
300 	}
301 }
302 
303 static int xenvif_open(struct net_device *dev)
304 {
305 	struct xenvif *vif = netdev_priv(dev);
306 	if (test_bit(VIF_STATUS_CONNECTED, &vif->status))
307 		xenvif_up(vif);
308 	netif_tx_start_all_queues(dev);
309 	return 0;
310 }
311 
312 static int xenvif_close(struct net_device *dev)
313 {
314 	struct xenvif *vif = netdev_priv(dev);
315 	if (test_bit(VIF_STATUS_CONNECTED, &vif->status))
316 		xenvif_down(vif);
317 	netif_tx_stop_all_queues(dev);
318 	return 0;
319 }
320 
321 static int xenvif_change_mtu(struct net_device *dev, int mtu)
322 {
323 	struct xenvif *vif = netdev_priv(dev);
324 	int max = vif->can_sg ? ETH_MAX_MTU - VLAN_ETH_HLEN : ETH_DATA_LEN;
325 
326 	if (mtu > max)
327 		return -EINVAL;
328 	dev->mtu = mtu;
329 	return 0;
330 }
331 
332 static netdev_features_t xenvif_fix_features(struct net_device *dev,
333 	netdev_features_t features)
334 {
335 	struct xenvif *vif = netdev_priv(dev);
336 
337 	if (!vif->can_sg)
338 		features &= ~NETIF_F_SG;
339 	if (~(vif->gso_mask) & GSO_BIT(TCPV4))
340 		features &= ~NETIF_F_TSO;
341 	if (~(vif->gso_mask) & GSO_BIT(TCPV6))
342 		features &= ~NETIF_F_TSO6;
343 	if (!vif->ip_csum)
344 		features &= ~NETIF_F_IP_CSUM;
345 	if (!vif->ipv6_csum)
346 		features &= ~NETIF_F_IPV6_CSUM;
347 
348 	return features;
349 }
350 
351 static const struct xenvif_stat {
352 	char name[ETH_GSTRING_LEN];
353 	u16 offset;
354 } xenvif_stats[] = {
355 	{
356 		"rx_gso_checksum_fixup",
357 		offsetof(struct xenvif_stats, rx_gso_checksum_fixup)
358 	},
359 	/* If (sent != success + fail), there are probably packets never
360 	 * freed up properly!
361 	 */
362 	{
363 		"tx_zerocopy_sent",
364 		offsetof(struct xenvif_stats, tx_zerocopy_sent),
365 	},
366 	{
367 		"tx_zerocopy_success",
368 		offsetof(struct xenvif_stats, tx_zerocopy_success),
369 	},
370 	{
371 		"tx_zerocopy_fail",
372 		offsetof(struct xenvif_stats, tx_zerocopy_fail)
373 	},
374 	/* Number of packets exceeding MAX_SKB_FRAG slots. You should use
375 	 * a guest with the same MAX_SKB_FRAG
376 	 */
377 	{
378 		"tx_frag_overflow",
379 		offsetof(struct xenvif_stats, tx_frag_overflow)
380 	},
381 };
382 
383 static int xenvif_get_sset_count(struct net_device *dev, int string_set)
384 {
385 	switch (string_set) {
386 	case ETH_SS_STATS:
387 		return ARRAY_SIZE(xenvif_stats);
388 	default:
389 		return -EINVAL;
390 	}
391 }
392 
393 static void xenvif_get_ethtool_stats(struct net_device *dev,
394 				     struct ethtool_stats *stats, u64 * data)
395 {
396 	struct xenvif *vif = netdev_priv(dev);
397 	unsigned int num_queues;
398 	int i;
399 	unsigned int queue_index;
400 
401 	rcu_read_lock();
402 	num_queues = READ_ONCE(vif->num_queues);
403 
404 	for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) {
405 		unsigned long accum = 0;
406 		for (queue_index = 0; queue_index < num_queues; ++queue_index) {
407 			void *vif_stats = &vif->queues[queue_index].stats;
408 			accum += *(unsigned long *)(vif_stats + xenvif_stats[i].offset);
409 		}
410 		data[i] = accum;
411 	}
412 
413 	rcu_read_unlock();
414 }
415 
416 static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data)
417 {
418 	int i;
419 
420 	switch (stringset) {
421 	case ETH_SS_STATS:
422 		for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++)
423 			memcpy(data + i * ETH_GSTRING_LEN,
424 			       xenvif_stats[i].name, ETH_GSTRING_LEN);
425 		break;
426 	}
427 }
428 
429 static const struct ethtool_ops xenvif_ethtool_ops = {
430 	.get_link	= ethtool_op_get_link,
431 
432 	.get_sset_count = xenvif_get_sset_count,
433 	.get_ethtool_stats = xenvif_get_ethtool_stats,
434 	.get_strings = xenvif_get_strings,
435 };
436 
437 static const struct net_device_ops xenvif_netdev_ops = {
438 	.ndo_select_queue = xenvif_select_queue,
439 	.ndo_start_xmit	= xenvif_start_xmit,
440 	.ndo_get_stats	= xenvif_get_stats,
441 	.ndo_open	= xenvif_open,
442 	.ndo_stop	= xenvif_close,
443 	.ndo_change_mtu	= xenvif_change_mtu,
444 	.ndo_fix_features = xenvif_fix_features,
445 	.ndo_set_mac_address = eth_mac_addr,
446 	.ndo_validate_addr   = eth_validate_addr,
447 };
448 
449 struct xenvif *xenvif_alloc(struct device *parent, domid_t domid,
450 			    unsigned int handle)
451 {
452 	int err;
453 	struct net_device *dev;
454 	struct xenvif *vif;
455 	char name[IFNAMSIZ] = {};
456 
457 	snprintf(name, IFNAMSIZ - 1, "vif%u.%u", domid, handle);
458 	/* Allocate a netdev with the max. supported number of queues.
459 	 * When the guest selects the desired number, it will be updated
460 	 * via netif_set_real_num_*_queues().
461 	 */
462 	dev = alloc_netdev_mq(sizeof(struct xenvif), name, NET_NAME_UNKNOWN,
463 			      ether_setup, xenvif_max_queues);
464 	if (dev == NULL) {
465 		pr_warn("Could not allocate netdev for %s\n", name);
466 		return ERR_PTR(-ENOMEM);
467 	}
468 
469 	SET_NETDEV_DEV(dev, parent);
470 
471 	vif = netdev_priv(dev);
472 
473 	vif->domid  = domid;
474 	vif->handle = handle;
475 	vif->can_sg = 1;
476 	vif->ip_csum = 1;
477 	vif->dev = dev;
478 	vif->disabled = false;
479 	vif->drain_timeout = msecs_to_jiffies(rx_drain_timeout_msecs);
480 	vif->stall_timeout = msecs_to_jiffies(rx_stall_timeout_msecs);
481 
482 	/* Start out with no queues. */
483 	vif->queues = NULL;
484 	vif->num_queues = 0;
485 
486 	spin_lock_init(&vif->lock);
487 	INIT_LIST_HEAD(&vif->fe_mcast_addr);
488 
489 	dev->netdev_ops	= &xenvif_netdev_ops;
490 	dev->hw_features = NETIF_F_SG |
491 		NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
492 		NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_FRAGLIST;
493 	dev->features = dev->hw_features | NETIF_F_RXCSUM;
494 	dev->ethtool_ops = &xenvif_ethtool_ops;
495 
496 	dev->tx_queue_len = XENVIF_QUEUE_LENGTH;
497 
498 	dev->min_mtu = ETH_MIN_MTU;
499 	dev->max_mtu = ETH_MAX_MTU - VLAN_ETH_HLEN;
500 
501 	/*
502 	 * Initialise a dummy MAC address. We choose the numerically
503 	 * largest non-broadcast address to prevent the address getting
504 	 * stolen by an Ethernet bridge for STP purposes.
505 	 * (FE:FF:FF:FF:FF:FF)
506 	 */
507 	eth_broadcast_addr(dev->dev_addr);
508 	dev->dev_addr[0] &= ~0x01;
509 
510 	netif_carrier_off(dev);
511 
512 	err = register_netdev(dev);
513 	if (err) {
514 		netdev_warn(dev, "Could not register device: err=%d\n", err);
515 		free_netdev(dev);
516 		return ERR_PTR(err);
517 	}
518 
519 	netdev_dbg(dev, "Successfully created xenvif\n");
520 
521 	__module_get(THIS_MODULE);
522 
523 	return vif;
524 }
525 
526 int xenvif_init_queue(struct xenvif_queue *queue)
527 {
528 	int err, i;
529 
530 	queue->credit_bytes = queue->remaining_credit = ~0UL;
531 	queue->credit_usec  = 0UL;
532 	timer_setup(&queue->credit_timeout, xenvif_tx_credit_callback, 0);
533 	queue->credit_window_start = get_jiffies_64();
534 
535 	queue->rx_queue_max = XENVIF_RX_QUEUE_BYTES;
536 
537 	skb_queue_head_init(&queue->rx_queue);
538 	skb_queue_head_init(&queue->tx_queue);
539 
540 	queue->pending_cons = 0;
541 	queue->pending_prod = MAX_PENDING_REQS;
542 	for (i = 0; i < MAX_PENDING_REQS; ++i)
543 		queue->pending_ring[i] = i;
544 
545 	spin_lock_init(&queue->callback_lock);
546 	spin_lock_init(&queue->response_lock);
547 
548 	/* If ballooning is disabled, this will consume real memory, so you
549 	 * better enable it. The long term solution would be to use just a
550 	 * bunch of valid page descriptors, without dependency on ballooning
551 	 */
552 	err = gnttab_alloc_pages(MAX_PENDING_REQS,
553 				 queue->mmap_pages);
554 	if (err) {
555 		netdev_err(queue->vif->dev, "Could not reserve mmap_pages\n");
556 		return -ENOMEM;
557 	}
558 
559 	for (i = 0; i < MAX_PENDING_REQS; i++) {
560 		queue->pending_tx_info[i].callback_struct = (struct ubuf_info)
561 			{ .callback = xenvif_zerocopy_callback,
562 			  { { .ctx = NULL,
563 			      .desc = i } } };
564 		queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE;
565 	}
566 
567 	return 0;
568 }
569 
570 void xenvif_carrier_on(struct xenvif *vif)
571 {
572 	rtnl_lock();
573 	if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN)
574 		dev_set_mtu(vif->dev, ETH_DATA_LEN);
575 	netdev_update_features(vif->dev);
576 	set_bit(VIF_STATUS_CONNECTED, &vif->status);
577 	if (netif_running(vif->dev))
578 		xenvif_up(vif);
579 	rtnl_unlock();
580 }
581 
582 int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref,
583 			unsigned int evtchn)
584 {
585 	struct net_device *dev = vif->dev;
586 	void *addr;
587 	struct xen_netif_ctrl_sring *shared;
588 	int err;
589 
590 	err = xenbus_map_ring_valloc(xenvif_to_xenbus_device(vif),
591 				     &ring_ref, 1, &addr);
592 	if (err)
593 		goto err;
594 
595 	shared = (struct xen_netif_ctrl_sring *)addr;
596 	BACK_RING_INIT(&vif->ctrl, shared, XEN_PAGE_SIZE);
597 
598 	err = bind_interdomain_evtchn_to_irq(vif->domid, evtchn);
599 	if (err < 0)
600 		goto err_unmap;
601 
602 	vif->ctrl_irq = err;
603 
604 	xenvif_init_hash(vif);
605 
606 	err = request_threaded_irq(vif->ctrl_irq, NULL, xenvif_ctrl_irq_fn,
607 				   IRQF_ONESHOT, "xen-netback-ctrl", vif);
608 	if (err) {
609 		pr_warn("Could not setup irq handler for %s\n", dev->name);
610 		goto err_deinit;
611 	}
612 
613 	return 0;
614 
615 err_deinit:
616 	xenvif_deinit_hash(vif);
617 	unbind_from_irqhandler(vif->ctrl_irq, vif);
618 	vif->ctrl_irq = 0;
619 
620 err_unmap:
621 	xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
622 				vif->ctrl.sring);
623 	vif->ctrl.sring = NULL;
624 
625 err:
626 	return err;
627 }
628 
629 int xenvif_connect_data(struct xenvif_queue *queue,
630 			unsigned long tx_ring_ref,
631 			unsigned long rx_ring_ref,
632 			unsigned int tx_evtchn,
633 			unsigned int rx_evtchn)
634 {
635 	struct task_struct *task;
636 	int err = -ENOMEM;
637 
638 	BUG_ON(queue->tx_irq);
639 	BUG_ON(queue->task);
640 	BUG_ON(queue->dealloc_task);
641 
642 	err = xenvif_map_frontend_data_rings(queue, tx_ring_ref,
643 					     rx_ring_ref);
644 	if (err < 0)
645 		goto err;
646 
647 	init_waitqueue_head(&queue->wq);
648 	init_waitqueue_head(&queue->dealloc_wq);
649 	atomic_set(&queue->inflight_packets, 0);
650 
651 	netif_napi_add(queue->vif->dev, &queue->napi, xenvif_poll,
652 			XENVIF_NAPI_WEIGHT);
653 
654 	if (tx_evtchn == rx_evtchn) {
655 		/* feature-split-event-channels == 0 */
656 		err = bind_interdomain_evtchn_to_irqhandler(
657 			queue->vif->domid, tx_evtchn, xenvif_interrupt, 0,
658 			queue->name, queue);
659 		if (err < 0)
660 			goto err_unmap;
661 		queue->tx_irq = queue->rx_irq = err;
662 		disable_irq(queue->tx_irq);
663 	} else {
664 		/* feature-split-event-channels == 1 */
665 		snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name),
666 			 "%s-tx", queue->name);
667 		err = bind_interdomain_evtchn_to_irqhandler(
668 			queue->vif->domid, tx_evtchn, xenvif_tx_interrupt, 0,
669 			queue->tx_irq_name, queue);
670 		if (err < 0)
671 			goto err_unmap;
672 		queue->tx_irq = err;
673 		disable_irq(queue->tx_irq);
674 
675 		snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name),
676 			 "%s-rx", queue->name);
677 		err = bind_interdomain_evtchn_to_irqhandler(
678 			queue->vif->domid, rx_evtchn, xenvif_rx_interrupt, 0,
679 			queue->rx_irq_name, queue);
680 		if (err < 0)
681 			goto err_tx_unbind;
682 		queue->rx_irq = err;
683 		disable_irq(queue->rx_irq);
684 	}
685 
686 	queue->stalled = true;
687 
688 	task = kthread_create(xenvif_kthread_guest_rx,
689 			      (void *)queue, "%s-guest-rx", queue->name);
690 	if (IS_ERR(task)) {
691 		pr_warn("Could not allocate kthread for %s\n", queue->name);
692 		err = PTR_ERR(task);
693 		goto err_rx_unbind;
694 	}
695 	queue->task = task;
696 	get_task_struct(task);
697 
698 	task = kthread_create(xenvif_dealloc_kthread,
699 			      (void *)queue, "%s-dealloc", queue->name);
700 	if (IS_ERR(task)) {
701 		pr_warn("Could not allocate kthread for %s\n", queue->name);
702 		err = PTR_ERR(task);
703 		goto err_rx_unbind;
704 	}
705 	queue->dealloc_task = task;
706 
707 	wake_up_process(queue->task);
708 	wake_up_process(queue->dealloc_task);
709 
710 	return 0;
711 
712 err_rx_unbind:
713 	unbind_from_irqhandler(queue->rx_irq, queue);
714 	queue->rx_irq = 0;
715 err_tx_unbind:
716 	unbind_from_irqhandler(queue->tx_irq, queue);
717 	queue->tx_irq = 0;
718 err_unmap:
719 	xenvif_unmap_frontend_data_rings(queue);
720 	netif_napi_del(&queue->napi);
721 err:
722 	module_put(THIS_MODULE);
723 	return err;
724 }
725 
726 void xenvif_carrier_off(struct xenvif *vif)
727 {
728 	struct net_device *dev = vif->dev;
729 
730 	rtnl_lock();
731 	if (test_and_clear_bit(VIF_STATUS_CONNECTED, &vif->status)) {
732 		netif_carrier_off(dev); /* discard queued packets */
733 		if (netif_running(dev))
734 			xenvif_down(vif);
735 	}
736 	rtnl_unlock();
737 }
738 
739 void xenvif_disconnect_data(struct xenvif *vif)
740 {
741 	struct xenvif_queue *queue = NULL;
742 	unsigned int num_queues = vif->num_queues;
743 	unsigned int queue_index;
744 
745 	xenvif_carrier_off(vif);
746 
747 	for (queue_index = 0; queue_index < num_queues; ++queue_index) {
748 		queue = &vif->queues[queue_index];
749 
750 		netif_napi_del(&queue->napi);
751 
752 		if (queue->task) {
753 			kthread_stop(queue->task);
754 			put_task_struct(queue->task);
755 			queue->task = NULL;
756 		}
757 
758 		if (queue->dealloc_task) {
759 			kthread_stop(queue->dealloc_task);
760 			queue->dealloc_task = NULL;
761 		}
762 
763 		if (queue->tx_irq) {
764 			if (queue->tx_irq == queue->rx_irq)
765 				unbind_from_irqhandler(queue->tx_irq, queue);
766 			else {
767 				unbind_from_irqhandler(queue->tx_irq, queue);
768 				unbind_from_irqhandler(queue->rx_irq, queue);
769 			}
770 			queue->tx_irq = 0;
771 		}
772 
773 		xenvif_unmap_frontend_data_rings(queue);
774 	}
775 
776 	xenvif_mcast_addr_list_free(vif);
777 }
778 
779 void xenvif_disconnect_ctrl(struct xenvif *vif)
780 {
781 	if (vif->ctrl_irq) {
782 		xenvif_deinit_hash(vif);
783 		unbind_from_irqhandler(vif->ctrl_irq, vif);
784 		vif->ctrl_irq = 0;
785 	}
786 
787 	if (vif->ctrl.sring) {
788 		xenbus_unmap_ring_vfree(xenvif_to_xenbus_device(vif),
789 					vif->ctrl.sring);
790 		vif->ctrl.sring = NULL;
791 	}
792 }
793 
794 /* Reverse the relevant parts of xenvif_init_queue().
795  * Used for queue teardown from xenvif_free(), and on the
796  * error handling paths in xenbus.c:connect().
797  */
798 void xenvif_deinit_queue(struct xenvif_queue *queue)
799 {
800 	gnttab_free_pages(MAX_PENDING_REQS, queue->mmap_pages);
801 }
802 
803 void xenvif_free(struct xenvif *vif)
804 {
805 	struct xenvif_queue *queues = vif->queues;
806 	unsigned int num_queues = vif->num_queues;
807 	unsigned int queue_index;
808 
809 	unregister_netdev(vif->dev);
810 	free_netdev(vif->dev);
811 
812 	for (queue_index = 0; queue_index < num_queues; ++queue_index)
813 		xenvif_deinit_queue(&queues[queue_index]);
814 	vfree(queues);
815 
816 	module_put(THIS_MODULE);
817 }
818