]>
Commit | Line | Data |
---|---|---|
f942dc25 IC |
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 | ||
b3f980bd | 33 | #include <linux/kthread.h> |
f942dc25 IC |
34 | #include <linux/ethtool.h> |
35 | #include <linux/rtnetlink.h> | |
36 | #include <linux/if_vlan.h> | |
e7b599d7 | 37 | #include <linux/vmalloc.h> |
f942dc25 IC |
38 | |
39 | #include <xen/events.h> | |
40 | #include <asm/xen/hypercall.h> | |
f53c3fe8 | 41 | #include <xen/balloon.h> |
f942dc25 IC |
42 | |
43 | #define XENVIF_QUEUE_LENGTH 32 | |
b3f980bd | 44 | #define XENVIF_NAPI_WEIGHT 64 |
f942dc25 | 45 | |
f48da8b1 DV |
46 | /* Number of bytes allowed on the internal guest Rx queue. */ |
47 | #define XENVIF_RX_QUEUE_BYTES (XEN_NETIF_RX_RING_SIZE/2 * PAGE_SIZE) | |
48 | ||
a64bd934 WL |
49 | /* This function is used to set SKBTX_DEV_ZEROCOPY as well as |
50 | * increasing the inflight counter. We need to increase the inflight | |
51 | * counter because core driver calls into xenvif_zerocopy_callback | |
52 | * which calls xenvif_skb_zerocopy_complete. | |
53 | */ | |
54 | void xenvif_skb_zerocopy_prepare(struct xenvif_queue *queue, | |
55 | struct sk_buff *skb) | |
56 | { | |
57 | skb_shinfo(skb)->tx_flags |= SKBTX_DEV_ZEROCOPY; | |
58 | atomic_inc(&queue->inflight_packets); | |
59 | } | |
60 | ||
61 | void xenvif_skb_zerocopy_complete(struct xenvif_queue *queue) | |
62 | { | |
63 | atomic_dec(&queue->inflight_packets); | |
57b22906 RL |
64 | |
65 | /* Wake the dealloc thread _after_ decrementing inflight_packets so | |
66 | * that if kthread_stop() has already been called, the dealloc thread | |
67 | * does not wait forever with nothing to wake it. | |
68 | */ | |
69 | wake_up(&queue->dealloc_wq); | |
a64bd934 WL |
70 | } |
71 | ||
f942dc25 IC |
72 | int xenvif_schedulable(struct xenvif *vif) |
73 | { | |
3d1af1df | 74 | return netif_running(vif->dev) && |
f48da8b1 DV |
75 | test_bit(VIF_STATUS_CONNECTED, &vif->status) && |
76 | !vif->disabled; | |
f942dc25 IC |
77 | } |
78 | ||
e1f00a69 | 79 | static irqreturn_t xenvif_tx_interrupt(int irq, void *dev_id) |
f942dc25 | 80 | { |
e9ce7cb6 | 81 | struct xenvif_queue *queue = dev_id; |
f942dc25 | 82 | |
e9ce7cb6 WL |
83 | if (RING_HAS_UNCONSUMED_REQUESTS(&queue->tx)) |
84 | napi_schedule(&queue->napi); | |
f942dc25 | 85 | |
e1f00a69 WL |
86 | return IRQ_HANDLED; |
87 | } | |
88 | ||
38741d50 | 89 | static int xenvif_poll(struct napi_struct *napi, int budget) |
b3f980bd | 90 | { |
e9ce7cb6 WL |
91 | struct xenvif_queue *queue = |
92 | container_of(napi, struct xenvif_queue, napi); | |
b3f980bd WL |
93 | int work_done; |
94 | ||
e9d8b2c2 WL |
95 | /* This vif is rogue, we pretend we've there is nothing to do |
96 | * for this vif to deschedule it from NAPI. But this interface | |
97 | * will be turned off in thread context later. | |
98 | */ | |
2561cc15 | 99 | if (unlikely(queue->vif->disabled)) { |
e9d8b2c2 WL |
100 | napi_complete(napi); |
101 | return 0; | |
102 | } | |
103 | ||
e9ce7cb6 | 104 | work_done = xenvif_tx_action(queue, budget); |
b3f980bd WL |
105 | |
106 | if (work_done < budget) { | |
0d08fceb | 107 | napi_complete(napi); |
e9ce7cb6 | 108 | xenvif_napi_schedule_or_enable_events(queue); |
b3f980bd WL |
109 | } |
110 | ||
111 | return work_done; | |
112 | } | |
113 | ||
e1f00a69 WL |
114 | static irqreturn_t xenvif_rx_interrupt(int irq, void *dev_id) |
115 | { | |
e9ce7cb6 | 116 | struct xenvif_queue *queue = dev_id; |
e1f00a69 | 117 | |
e9ce7cb6 | 118 | xenvif_kick_thread(queue); |
f942dc25 IC |
119 | |
120 | return IRQ_HANDLED; | |
121 | } | |
122 | ||
f51de243 | 123 | irqreturn_t xenvif_interrupt(int irq, void *dev_id) |
e1f00a69 WL |
124 | { |
125 | xenvif_tx_interrupt(irq, dev_id); | |
126 | xenvif_rx_interrupt(irq, dev_id); | |
127 | ||
128 | return IRQ_HANDLED; | |
129 | } | |
130 | ||
e9ce7cb6 WL |
131 | int xenvif_queue_stopped(struct xenvif_queue *queue) |
132 | { | |
133 | struct net_device *dev = queue->vif->dev; | |
134 | unsigned int id = queue->id; | |
135 | return netif_tx_queue_stopped(netdev_get_tx_queue(dev, id)); | |
136 | } | |
137 | ||
138 | void xenvif_wake_queue(struct xenvif_queue *queue) | |
139 | { | |
140 | struct net_device *dev = queue->vif->dev; | |
141 | unsigned int id = queue->id; | |
142 | netif_tx_wake_queue(netdev_get_tx_queue(dev, id)); | |
143 | } | |
144 | ||
f942dc25 IC |
145 | static int xenvif_start_xmit(struct sk_buff *skb, struct net_device *dev) |
146 | { | |
147 | struct xenvif *vif = netdev_priv(dev); | |
e9ce7cb6 | 148 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 149 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 | 150 | u16 index; |
f48da8b1 | 151 | struct xenvif_rx_cb *cb; |
f942dc25 IC |
152 | |
153 | BUG_ON(skb->dev != dev); | |
154 | ||
e9ce7cb6 WL |
155 | /* Drop the packet if queues are not set up */ |
156 | if (num_queues < 1) | |
157 | goto drop; | |
158 | ||
159 | /* Obtain the queue to be used to transmit this packet */ | |
160 | index = skb_get_queue_mapping(skb); | |
161 | if (index >= num_queues) { | |
162 | pr_warn_ratelimited("Invalid queue %hu for packet on interface %s\n.", | |
163 | index, vif->dev->name); | |
164 | index %= num_queues; | |
165 | } | |
166 | queue = &vif->queues[index]; | |
167 | ||
168 | /* Drop the packet if queue is not ready */ | |
169 | if (queue->task == NULL || | |
170 | queue->dealloc_task == NULL || | |
f53c3fe8 | 171 | !xenvif_schedulable(vif)) |
f942dc25 IC |
172 | goto drop; |
173 | ||
f48da8b1 | 174 | cb = XENVIF_RX_CB(skb); |
26c0e102 | 175 | cb->expires = jiffies + vif->drain_timeout; |
f942dc25 | 176 | |
f48da8b1 | 177 | xenvif_rx_queue_tail(queue, skb); |
e9ce7cb6 | 178 | xenvif_kick_thread(queue); |
f942dc25 IC |
179 | |
180 | return NETDEV_TX_OK; | |
181 | ||
182 | drop: | |
183 | vif->dev->stats.tx_dropped++; | |
184 | dev_kfree_skb(skb); | |
185 | return NETDEV_TX_OK; | |
186 | } | |
187 | ||
f942dc25 IC |
188 | static struct net_device_stats *xenvif_get_stats(struct net_device *dev) |
189 | { | |
190 | struct xenvif *vif = netdev_priv(dev); | |
e9ce7cb6 | 191 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 192 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 WL |
193 | unsigned long rx_bytes = 0; |
194 | unsigned long rx_packets = 0; | |
195 | unsigned long tx_bytes = 0; | |
196 | unsigned long tx_packets = 0; | |
197 | unsigned int index; | |
198 | ||
199 | if (vif->queues == NULL) | |
200 | goto out; | |
201 | ||
202 | /* Aggregate tx and rx stats from each queue */ | |
203 | for (index = 0; index < num_queues; ++index) { | |
204 | queue = &vif->queues[index]; | |
205 | rx_bytes += queue->stats.rx_bytes; | |
206 | rx_packets += queue->stats.rx_packets; | |
207 | tx_bytes += queue->stats.tx_bytes; | |
208 | tx_packets += queue->stats.tx_packets; | |
209 | } | |
210 | ||
211 | out: | |
212 | vif->dev->stats.rx_bytes = rx_bytes; | |
213 | vif->dev->stats.rx_packets = rx_packets; | |
214 | vif->dev->stats.tx_bytes = tx_bytes; | |
215 | vif->dev->stats.tx_packets = tx_packets; | |
216 | ||
f942dc25 IC |
217 | return &vif->dev->stats; |
218 | } | |
219 | ||
220 | static void xenvif_up(struct xenvif *vif) | |
221 | { | |
e9ce7cb6 | 222 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 223 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 WL |
224 | unsigned int queue_index; |
225 | ||
226 | for (queue_index = 0; queue_index < num_queues; ++queue_index) { | |
227 | queue = &vif->queues[queue_index]; | |
228 | napi_enable(&queue->napi); | |
229 | enable_irq(queue->tx_irq); | |
230 | if (queue->tx_irq != queue->rx_irq) | |
231 | enable_irq(queue->rx_irq); | |
232 | xenvif_napi_schedule_or_enable_events(queue); | |
233 | } | |
f942dc25 IC |
234 | } |
235 | ||
236 | static void xenvif_down(struct xenvif *vif) | |
237 | { | |
e9ce7cb6 | 238 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 239 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 WL |
240 | unsigned int queue_index; |
241 | ||
242 | for (queue_index = 0; queue_index < num_queues; ++queue_index) { | |
243 | queue = &vif->queues[queue_index]; | |
e9ce7cb6 WL |
244 | disable_irq(queue->tx_irq); |
245 | if (queue->tx_irq != queue->rx_irq) | |
246 | disable_irq(queue->rx_irq); | |
8fe78989 | 247 | napi_disable(&queue->napi); |
e9ce7cb6 WL |
248 | del_timer_sync(&queue->credit_timeout); |
249 | } | |
f942dc25 IC |
250 | } |
251 | ||
252 | static int xenvif_open(struct net_device *dev) | |
253 | { | |
254 | struct xenvif *vif = netdev_priv(dev); | |
3d1af1df | 255 | if (test_bit(VIF_STATUS_CONNECTED, &vif->status)) |
f942dc25 | 256 | xenvif_up(vif); |
e9ce7cb6 | 257 | netif_tx_start_all_queues(dev); |
f942dc25 IC |
258 | return 0; |
259 | } | |
260 | ||
261 | static int xenvif_close(struct net_device *dev) | |
262 | { | |
263 | struct xenvif *vif = netdev_priv(dev); | |
3d1af1df | 264 | if (test_bit(VIF_STATUS_CONNECTED, &vif->status)) |
f942dc25 | 265 | xenvif_down(vif); |
e9ce7cb6 | 266 | netif_tx_stop_all_queues(dev); |
f942dc25 IC |
267 | return 0; |
268 | } | |
269 | ||
270 | static int xenvif_change_mtu(struct net_device *dev, int mtu) | |
271 | { | |
272 | struct xenvif *vif = netdev_priv(dev); | |
273 | int max = vif->can_sg ? 65535 - VLAN_ETH_HLEN : ETH_DATA_LEN; | |
274 | ||
275 | if (mtu > max) | |
276 | return -EINVAL; | |
277 | dev->mtu = mtu; | |
278 | return 0; | |
279 | } | |
280 | ||
c8f44aff MM |
281 | static netdev_features_t xenvif_fix_features(struct net_device *dev, |
282 | netdev_features_t features) | |
f942dc25 IC |
283 | { |
284 | struct xenvif *vif = netdev_priv(dev); | |
f942dc25 | 285 | |
47103041 MM |
286 | if (!vif->can_sg) |
287 | features &= ~NETIF_F_SG; | |
82cada22 | 288 | if (~(vif->gso_mask | vif->gso_prefix_mask) & GSO_BIT(TCPV4)) |
47103041 | 289 | features &= ~NETIF_F_TSO; |
82cada22 PD |
290 | if (~(vif->gso_mask | vif->gso_prefix_mask) & GSO_BIT(TCPV6)) |
291 | features &= ~NETIF_F_TSO6; | |
146c8a77 | 292 | if (!vif->ip_csum) |
47103041 | 293 | features &= ~NETIF_F_IP_CSUM; |
146c8a77 PD |
294 | if (!vif->ipv6_csum) |
295 | features &= ~NETIF_F_IPV6_CSUM; | |
f942dc25 | 296 | |
47103041 | 297 | return features; |
f942dc25 IC |
298 | } |
299 | ||
300 | static const struct xenvif_stat { | |
301 | char name[ETH_GSTRING_LEN]; | |
302 | u16 offset; | |
303 | } xenvif_stats[] = { | |
304 | { | |
305 | "rx_gso_checksum_fixup", | |
e9ce7cb6 | 306 | offsetof(struct xenvif_stats, rx_gso_checksum_fixup) |
f942dc25 | 307 | }, |
1bb332af ZK |
308 | /* If (sent != success + fail), there are probably packets never |
309 | * freed up properly! | |
310 | */ | |
311 | { | |
312 | "tx_zerocopy_sent", | |
e9ce7cb6 | 313 | offsetof(struct xenvif_stats, tx_zerocopy_sent), |
1bb332af ZK |
314 | }, |
315 | { | |
316 | "tx_zerocopy_success", | |
e9ce7cb6 | 317 | offsetof(struct xenvif_stats, tx_zerocopy_success), |
1bb332af ZK |
318 | }, |
319 | { | |
320 | "tx_zerocopy_fail", | |
e9ce7cb6 | 321 | offsetof(struct xenvif_stats, tx_zerocopy_fail) |
1bb332af | 322 | }, |
e3377f36 ZK |
323 | /* Number of packets exceeding MAX_SKB_FRAG slots. You should use |
324 | * a guest with the same MAX_SKB_FRAG | |
325 | */ | |
326 | { | |
327 | "tx_frag_overflow", | |
e9ce7cb6 | 328 | offsetof(struct xenvif_stats, tx_frag_overflow) |
e3377f36 | 329 | }, |
f942dc25 IC |
330 | }; |
331 | ||
332 | static int xenvif_get_sset_count(struct net_device *dev, int string_set) | |
333 | { | |
334 | switch (string_set) { | |
335 | case ETH_SS_STATS: | |
336 | return ARRAY_SIZE(xenvif_stats); | |
337 | default: | |
338 | return -EINVAL; | |
339 | } | |
340 | } | |
341 | ||
342 | static void xenvif_get_ethtool_stats(struct net_device *dev, | |
343 | struct ethtool_stats *stats, u64 * data) | |
344 | { | |
e9ce7cb6 | 345 | struct xenvif *vif = netdev_priv(dev); |
f7b50c4e | 346 | unsigned int num_queues = vif->num_queues; |
f942dc25 | 347 | int i; |
e9ce7cb6 | 348 | unsigned int queue_index; |
e9ce7cb6 WL |
349 | |
350 | for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) { | |
351 | unsigned long accum = 0; | |
352 | for (queue_index = 0; queue_index < num_queues; ++queue_index) { | |
d63951d7 | 353 | void *vif_stats = &vif->queues[queue_index].stats; |
e9ce7cb6 WL |
354 | accum += *(unsigned long *)(vif_stats + xenvif_stats[i].offset); |
355 | } | |
356 | data[i] = accum; | |
357 | } | |
f942dc25 IC |
358 | } |
359 | ||
360 | static void xenvif_get_strings(struct net_device *dev, u32 stringset, u8 * data) | |
361 | { | |
362 | int i; | |
363 | ||
364 | switch (stringset) { | |
365 | case ETH_SS_STATS: | |
366 | for (i = 0; i < ARRAY_SIZE(xenvif_stats); i++) | |
367 | memcpy(data + i * ETH_GSTRING_LEN, | |
368 | xenvif_stats[i].name, ETH_GSTRING_LEN); | |
369 | break; | |
370 | } | |
371 | } | |
372 | ||
813abbba | 373 | static const struct ethtool_ops xenvif_ethtool_ops = { |
f942dc25 IC |
374 | .get_link = ethtool_op_get_link, |
375 | ||
376 | .get_sset_count = xenvif_get_sset_count, | |
377 | .get_ethtool_stats = xenvif_get_ethtool_stats, | |
378 | .get_strings = xenvif_get_strings, | |
379 | }; | |
380 | ||
813abbba | 381 | static const struct net_device_ops xenvif_netdev_ops = { |
f942dc25 IC |
382 | .ndo_start_xmit = xenvif_start_xmit, |
383 | .ndo_get_stats = xenvif_get_stats, | |
384 | .ndo_open = xenvif_open, | |
385 | .ndo_stop = xenvif_close, | |
386 | .ndo_change_mtu = xenvif_change_mtu, | |
47103041 | 387 | .ndo_fix_features = xenvif_fix_features, |
4a633a60 MW |
388 | .ndo_set_mac_address = eth_mac_addr, |
389 | .ndo_validate_addr = eth_validate_addr, | |
f942dc25 IC |
390 | }; |
391 | ||
392 | struct xenvif *xenvif_alloc(struct device *parent, domid_t domid, | |
393 | unsigned int handle) | |
394 | { | |
395 | int err; | |
396 | struct net_device *dev; | |
397 | struct xenvif *vif; | |
398 | char name[IFNAMSIZ] = {}; | |
399 | ||
400 | snprintf(name, IFNAMSIZ - 1, "vif%u.%u", domid, handle); | |
8d3d53b3 AB |
401 | /* Allocate a netdev with the max. supported number of queues. |
402 | * When the guest selects the desired number, it will be updated | |
f7b50c4e | 403 | * via netif_set_real_num_*_queues(). |
8d3d53b3 | 404 | */ |
c835a677 TG |
405 | dev = alloc_netdev_mq(sizeof(struct xenvif), name, NET_NAME_UNKNOWN, |
406 | ether_setup, xenvif_max_queues); | |
f942dc25 | 407 | if (dev == NULL) { |
b3f980bd | 408 | pr_warn("Could not allocate netdev for %s\n", name); |
f942dc25 IC |
409 | return ERR_PTR(-ENOMEM); |
410 | } | |
411 | ||
412 | SET_NETDEV_DEV(dev, parent); | |
413 | ||
414 | vif = netdev_priv(dev); | |
ac3d5ac2 | 415 | |
f942dc25 IC |
416 | vif->domid = domid; |
417 | vif->handle = handle; | |
f942dc25 | 418 | vif->can_sg = 1; |
146c8a77 | 419 | vif->ip_csum = 1; |
f942dc25 | 420 | vif->dev = dev; |
e9d8b2c2 | 421 | vif->disabled = false; |
26c0e102 DV |
422 | vif->drain_timeout = msecs_to_jiffies(rx_drain_timeout_msecs); |
423 | vif->stall_timeout = msecs_to_jiffies(rx_stall_timeout_msecs); | |
e9d8b2c2 | 424 | |
f7b50c4e | 425 | /* Start out with no queues. */ |
e9ce7cb6 | 426 | vif->queues = NULL; |
f7b50c4e | 427 | vif->num_queues = 0; |
09350788 | 428 | |
ecf08d2d DV |
429 | spin_lock_init(&vif->lock); |
430 | ||
f942dc25 | 431 | dev->netdev_ops = &xenvif_netdev_ops; |
146c8a77 PD |
432 | dev->hw_features = NETIF_F_SG | |
433 | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | | |
82cada22 | 434 | NETIF_F_TSO | NETIF_F_TSO6; |
7365bcfa | 435 | dev->features = dev->hw_features | NETIF_F_RXCSUM; |
7ad24ea4 | 436 | dev->ethtool_ops = &xenvif_ethtool_ops; |
f942dc25 IC |
437 | |
438 | dev->tx_queue_len = XENVIF_QUEUE_LENGTH; | |
439 | ||
440 | /* | |
441 | * Initialise a dummy MAC address. We choose the numerically | |
442 | * largest non-broadcast address to prevent the address getting | |
443 | * stolen by an Ethernet bridge for STP purposes. | |
444 | * (FE:FF:FF:FF:FF:FF) | |
445 | */ | |
3b6ed26d | 446 | eth_broadcast_addr(dev->dev_addr); |
f942dc25 IC |
447 | dev->dev_addr[0] &= ~0x01; |
448 | ||
449 | netif_carrier_off(dev); | |
450 | ||
451 | err = register_netdev(dev); | |
452 | if (err) { | |
453 | netdev_warn(dev, "Could not register device: err=%d\n", err); | |
454 | free_netdev(dev); | |
455 | return ERR_PTR(err); | |
456 | } | |
457 | ||
458 | netdev_dbg(dev, "Successfully created xenvif\n"); | |
279f438e PD |
459 | |
460 | __module_get(THIS_MODULE); | |
461 | ||
f942dc25 IC |
462 | return vif; |
463 | } | |
464 | ||
e9ce7cb6 WL |
465 | int xenvif_init_queue(struct xenvif_queue *queue) |
466 | { | |
467 | int err, i; | |
468 | ||
469 | queue->credit_bytes = queue->remaining_credit = ~0UL; | |
470 | queue->credit_usec = 0UL; | |
471 | init_timer(&queue->credit_timeout); | |
edafc132 | 472 | queue->credit_timeout.function = xenvif_tx_credit_callback; |
e9ce7cb6 WL |
473 | queue->credit_window_start = get_jiffies_64(); |
474 | ||
f48da8b1 DV |
475 | queue->rx_queue_max = XENVIF_RX_QUEUE_BYTES; |
476 | ||
e9ce7cb6 WL |
477 | skb_queue_head_init(&queue->rx_queue); |
478 | skb_queue_head_init(&queue->tx_queue); | |
479 | ||
480 | queue->pending_cons = 0; | |
481 | queue->pending_prod = MAX_PENDING_REQS; | |
482 | for (i = 0; i < MAX_PENDING_REQS; ++i) | |
483 | queue->pending_ring[i] = i; | |
484 | ||
485 | spin_lock_init(&queue->callback_lock); | |
486 | spin_lock_init(&queue->response_lock); | |
487 | ||
488 | /* If ballooning is disabled, this will consume real memory, so you | |
489 | * better enable it. The long term solution would be to use just a | |
490 | * bunch of valid page descriptors, without dependency on ballooning | |
491 | */ | |
ff4b156f DV |
492 | err = gnttab_alloc_pages(MAX_PENDING_REQS, |
493 | queue->mmap_pages); | |
e9ce7cb6 WL |
494 | if (err) { |
495 | netdev_err(queue->vif->dev, "Could not reserve mmap_pages\n"); | |
496 | return -ENOMEM; | |
497 | } | |
498 | ||
499 | for (i = 0; i < MAX_PENDING_REQS; i++) { | |
500 | queue->pending_tx_info[i].callback_struct = (struct ubuf_info) | |
501 | { .callback = xenvif_zerocopy_callback, | |
502 | .ctx = NULL, | |
503 | .desc = i }; | |
504 | queue->grant_tx_handle[i] = NETBACK_INVALID_HANDLE; | |
505 | } | |
506 | ||
e9ce7cb6 WL |
507 | return 0; |
508 | } | |
509 | ||
510 | void xenvif_carrier_on(struct xenvif *vif) | |
511 | { | |
512 | rtnl_lock(); | |
513 | if (!vif->can_sg && vif->dev->mtu > ETH_DATA_LEN) | |
514 | dev_set_mtu(vif->dev, ETH_DATA_LEN); | |
515 | netdev_update_features(vif->dev); | |
3d1af1df | 516 | set_bit(VIF_STATUS_CONNECTED, &vif->status); |
e9ce7cb6 WL |
517 | if (netif_running(vif->dev)) |
518 | xenvif_up(vif); | |
519 | rtnl_unlock(); | |
520 | } | |
521 | ||
522 | int xenvif_connect(struct xenvif_queue *queue, unsigned long tx_ring_ref, | |
e1f00a69 WL |
523 | unsigned long rx_ring_ref, unsigned int tx_evtchn, |
524 | unsigned int rx_evtchn) | |
f942dc25 | 525 | { |
67fa3660 | 526 | struct task_struct *task; |
f942dc25 IC |
527 | int err = -ENOMEM; |
528 | ||
e9ce7cb6 WL |
529 | BUG_ON(queue->tx_irq); |
530 | BUG_ON(queue->task); | |
531 | BUG_ON(queue->dealloc_task); | |
f942dc25 | 532 | |
e9ce7cb6 | 533 | err = xenvif_map_frontend_rings(queue, tx_ring_ref, rx_ring_ref); |
f942dc25 IC |
534 | if (err < 0) |
535 | goto err; | |
536 | ||
e9ce7cb6 WL |
537 | init_waitqueue_head(&queue->wq); |
538 | init_waitqueue_head(&queue->dealloc_wq); | |
a64bd934 | 539 | atomic_set(&queue->inflight_packets, 0); |
ca2f09f2 | 540 | |
e24f8191 WL |
541 | netif_napi_add(queue->vif->dev, &queue->napi, xenvif_poll, |
542 | XENVIF_NAPI_WEIGHT); | |
543 | ||
e1f00a69 WL |
544 | if (tx_evtchn == rx_evtchn) { |
545 | /* feature-split-event-channels == 0 */ | |
546 | err = bind_interdomain_evtchn_to_irqhandler( | |
e9ce7cb6 WL |
547 | queue->vif->domid, tx_evtchn, xenvif_interrupt, 0, |
548 | queue->name, queue); | |
e1f00a69 WL |
549 | if (err < 0) |
550 | goto err_unmap; | |
e9ce7cb6 WL |
551 | queue->tx_irq = queue->rx_irq = err; |
552 | disable_irq(queue->tx_irq); | |
e1f00a69 WL |
553 | } else { |
554 | /* feature-split-event-channels == 1 */ | |
e9ce7cb6 WL |
555 | snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name), |
556 | "%s-tx", queue->name); | |
e1f00a69 | 557 | err = bind_interdomain_evtchn_to_irqhandler( |
e9ce7cb6 WL |
558 | queue->vif->domid, tx_evtchn, xenvif_tx_interrupt, 0, |
559 | queue->tx_irq_name, queue); | |
e1f00a69 WL |
560 | if (err < 0) |
561 | goto err_unmap; | |
e9ce7cb6 WL |
562 | queue->tx_irq = err; |
563 | disable_irq(queue->tx_irq); | |
e1f00a69 | 564 | |
e9ce7cb6 WL |
565 | snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name), |
566 | "%s-rx", queue->name); | |
e1f00a69 | 567 | err = bind_interdomain_evtchn_to_irqhandler( |
e9ce7cb6 WL |
568 | queue->vif->domid, rx_evtchn, xenvif_rx_interrupt, 0, |
569 | queue->rx_irq_name, queue); | |
e1f00a69 WL |
570 | if (err < 0) |
571 | goto err_tx_unbind; | |
e9ce7cb6 WL |
572 | queue->rx_irq = err; |
573 | disable_irq(queue->rx_irq); | |
e1f00a69 | 574 | } |
f942dc25 | 575 | |
ecf08d2d DV |
576 | queue->stalled = true; |
577 | ||
121fa4b7 | 578 | task = kthread_create(xenvif_kthread_guest_rx, |
e9ce7cb6 | 579 | (void *)queue, "%s-guest-rx", queue->name); |
67fa3660 | 580 | if (IS_ERR(task)) { |
e9ce7cb6 | 581 | pr_warn("Could not allocate kthread for %s\n", queue->name); |
67fa3660 | 582 | err = PTR_ERR(task); |
b3f980bd WL |
583 | goto err_rx_unbind; |
584 | } | |
e9ce7cb6 | 585 | queue->task = task; |
42b5212f | 586 | get_task_struct(task); |
67fa3660 | 587 | |
f53c3fe8 | 588 | task = kthread_create(xenvif_dealloc_kthread, |
e9ce7cb6 | 589 | (void *)queue, "%s-dealloc", queue->name); |
f53c3fe8 | 590 | if (IS_ERR(task)) { |
e9ce7cb6 | 591 | pr_warn("Could not allocate kthread for %s\n", queue->name); |
f53c3fe8 ZK |
592 | err = PTR_ERR(task); |
593 | goto err_rx_unbind; | |
594 | } | |
e9ce7cb6 | 595 | queue->dealloc_task = task; |
f53c3fe8 | 596 | |
e9ce7cb6 WL |
597 | wake_up_process(queue->task); |
598 | wake_up_process(queue->dealloc_task); | |
b3f980bd | 599 | |
f942dc25 | 600 | return 0; |
b3f980bd WL |
601 | |
602 | err_rx_unbind: | |
e9ce7cb6 WL |
603 | unbind_from_irqhandler(queue->rx_irq, queue); |
604 | queue->rx_irq = 0; | |
e1f00a69 | 605 | err_tx_unbind: |
e9ce7cb6 WL |
606 | unbind_from_irqhandler(queue->tx_irq, queue); |
607 | queue->tx_irq = 0; | |
f942dc25 | 608 | err_unmap: |
e9ce7cb6 | 609 | xenvif_unmap_frontend_rings(queue); |
f942dc25 | 610 | err: |
b103f358 | 611 | module_put(THIS_MODULE); |
f942dc25 IC |
612 | return err; |
613 | } | |
614 | ||
48856286 | 615 | void xenvif_carrier_off(struct xenvif *vif) |
f942dc25 IC |
616 | { |
617 | struct net_device *dev = vif->dev; | |
48856286 IC |
618 | |
619 | rtnl_lock(); | |
3d1af1df ZK |
620 | if (test_and_clear_bit(VIF_STATUS_CONNECTED, &vif->status)) { |
621 | netif_carrier_off(dev); /* discard queued packets */ | |
622 | if (netif_running(dev)) | |
623 | xenvif_down(vif); | |
624 | } | |
48856286 | 625 | rtnl_unlock(); |
48856286 IC |
626 | } |
627 | ||
628 | void xenvif_disconnect(struct xenvif *vif) | |
629 | { | |
e9ce7cb6 | 630 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 631 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 WL |
632 | unsigned int queue_index; |
633 | ||
3d1af1df | 634 | xenvif_carrier_off(vif); |
f942dc25 | 635 | |
e9ce7cb6 WL |
636 | for (queue_index = 0; queue_index < num_queues; ++queue_index) { |
637 | queue = &vif->queues[queue_index]; | |
db739ef3 | 638 | |
ea2c5e13 WL |
639 | netif_napi_del(&queue->napi); |
640 | ||
e9ce7cb6 | 641 | if (queue->task) { |
e9ce7cb6 | 642 | kthread_stop(queue->task); |
42b5212f | 643 | put_task_struct(queue->task); |
e9ce7cb6 WL |
644 | queue->task = NULL; |
645 | } | |
f53c3fe8 | 646 | |
e9ce7cb6 WL |
647 | if (queue->dealloc_task) { |
648 | kthread_stop(queue->dealloc_task); | |
649 | queue->dealloc_task = NULL; | |
650 | } | |
651 | ||
652 | if (queue->tx_irq) { | |
653 | if (queue->tx_irq == queue->rx_irq) | |
654 | unbind_from_irqhandler(queue->tx_irq, queue); | |
655 | else { | |
656 | unbind_from_irqhandler(queue->tx_irq, queue); | |
657 | unbind_from_irqhandler(queue->rx_irq, queue); | |
658 | } | |
659 | queue->tx_irq = 0; | |
e1f00a69 | 660 | } |
f942dc25 | 661 | |
e9ce7cb6 WL |
662 | xenvif_unmap_frontend_rings(queue); |
663 | } | |
279f438e PD |
664 | } |
665 | ||
8d3d53b3 AB |
666 | /* Reverse the relevant parts of xenvif_init_queue(). |
667 | * Used for queue teardown from xenvif_free(), and on the | |
668 | * error handling paths in xenbus.c:connect(). | |
669 | */ | |
670 | void xenvif_deinit_queue(struct xenvif_queue *queue) | |
671 | { | |
ff4b156f | 672 | gnttab_free_pages(MAX_PENDING_REQS, queue->mmap_pages); |
8d3d53b3 AB |
673 | } |
674 | ||
279f438e PD |
675 | void xenvif_free(struct xenvif *vif) |
676 | { | |
e9ce7cb6 | 677 | struct xenvif_queue *queue = NULL; |
f7b50c4e | 678 | unsigned int num_queues = vif->num_queues; |
e9ce7cb6 | 679 | unsigned int queue_index; |
f53c3fe8 | 680 | |
e9ce7cb6 | 681 | unregister_netdev(vif->dev); |
f53c3fe8 | 682 | |
e9ce7cb6 WL |
683 | for (queue_index = 0; queue_index < num_queues; ++queue_index) { |
684 | queue = &vif->queues[queue_index]; | |
8d3d53b3 | 685 | xenvif_deinit_queue(queue); |
e9ce7cb6 WL |
686 | } |
687 | ||
e9ce7cb6 WL |
688 | vfree(vif->queues); |
689 | vif->queues = NULL; | |
f7b50c4e | 690 | vif->num_queues = 0; |
f942dc25 | 691 | |
f942dc25 | 692 | free_netdev(vif->dev); |
b103f358 | 693 | |
279f438e | 694 | module_put(THIS_MODULE); |
f942dc25 | 695 | } |