]>
Commit | Line | Data |
---|---|---|
fbe78f4f AL |
1 | /* |
2 | * Virtio Network Device | |
3 | * | |
4 | * Copyright IBM, Corp. 2007 | |
5 | * | |
6 | * Authors: | |
7 | * Anthony Liguori <aliguori@us.ibm.com> | |
8 | * | |
9 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
10 | * the COPYING file in the top-level directory. | |
11 | * | |
12 | */ | |
13 | ||
1de7afc9 | 14 | #include "qemu/iov.h" |
0d09e41a | 15 | #include "hw/virtio/virtio.h" |
1422e32d | 16 | #include "net/net.h" |
7200ac3c | 17 | #include "net/checksum.h" |
a8ed73f7 | 18 | #include "net/tap.h" |
1de7afc9 PB |
19 | #include "qemu/error-report.h" |
20 | #include "qemu/timer.h" | |
0d09e41a PB |
21 | #include "hw/virtio/virtio-net.h" |
22 | #include "net/vhost_net.h" | |
17ec5a86 | 23 | #include "hw/virtio/virtio-bus.h" |
b1be4280 AK |
24 | #include "qapi/qmp/qjson.h" |
25 | #include "monitor/monitor.h" | |
fbe78f4f | 26 | |
0ce0e8f4 | 27 | #define VIRTIO_NET_VM_VERSION 11 |
b6503ed9 | 28 | |
4ffb17f5 | 29 | #define MAC_TABLE_ENTRIES 64 |
f21c0ed9 | 30 | #define MAX_VLAN (1 << 12) /* Per 802.1Q definition */ |
9d6271b8 | 31 | |
14f9b664 JL |
32 | /* |
33 | * Calculate the number of bytes up to and including the given 'field' of | |
34 | * 'container'. | |
35 | */ | |
36 | #define endof(container, field) \ | |
37 | (offsetof(container, field) + sizeof(((container *)0)->field)) | |
38 | ||
39 | typedef struct VirtIOFeature { | |
40 | uint32_t flags; | |
41 | size_t end; | |
42 | } VirtIOFeature; | |
43 | ||
44 | static VirtIOFeature feature_sizes[] = { | |
45 | {.flags = 1 << VIRTIO_NET_F_MAC, | |
46 | .end = endof(struct virtio_net_config, mac)}, | |
47 | {.flags = 1 << VIRTIO_NET_F_STATUS, | |
48 | .end = endof(struct virtio_net_config, status)}, | |
49 | {.flags = 1 << VIRTIO_NET_F_MQ, | |
50 | .end = endof(struct virtio_net_config, max_virtqueue_pairs)}, | |
51 | {} | |
52 | }; | |
53 | ||
fed699f9 | 54 | static VirtIONetQueue *virtio_net_get_subqueue(NetClientState *nc) |
0c87e93e JW |
55 | { |
56 | VirtIONet *n = qemu_get_nic_opaque(nc); | |
57 | ||
fed699f9 | 58 | return &n->vqs[nc->queue_index]; |
0c87e93e | 59 | } |
fed699f9 JW |
60 | |
61 | static int vq2q(int queue_index) | |
62 | { | |
63 | return queue_index / 2; | |
64 | } | |
65 | ||
fbe78f4f AL |
66 | /* TODO |
67 | * - we could suppress RX interrupt if we were so inclined. | |
68 | */ | |
69 | ||
0f03eca6 | 70 | static void virtio_net_get_config(VirtIODevice *vdev, uint8_t *config) |
fbe78f4f | 71 | { |
17a0ca55 | 72 | VirtIONet *n = VIRTIO_NET(vdev); |
fbe78f4f AL |
73 | struct virtio_net_config netcfg; |
74 | ||
b46d97f2 | 75 | stw_p(&netcfg.status, n->status); |
fed699f9 | 76 | stw_p(&netcfg.max_virtqueue_pairs, n->max_queues); |
79674068 | 77 | memcpy(netcfg.mac, n->mac, ETH_ALEN); |
14f9b664 | 78 | memcpy(config, &netcfg, n->config_size); |
fbe78f4f AL |
79 | } |
80 | ||
0f03eca6 AL |
81 | static void virtio_net_set_config(VirtIODevice *vdev, const uint8_t *config) |
82 | { | |
17a0ca55 | 83 | VirtIONet *n = VIRTIO_NET(vdev); |
14f9b664 | 84 | struct virtio_net_config netcfg = {}; |
0f03eca6 | 85 | |
14f9b664 | 86 | memcpy(&netcfg, config, n->config_size); |
0f03eca6 | 87 | |
17a0ca55 | 88 | if (!(vdev->guest_features >> VIRTIO_NET_F_CTRL_MAC_ADDR & 1) && |
c1943a3f | 89 | memcmp(netcfg.mac, n->mac, ETH_ALEN)) { |
79674068 | 90 | memcpy(n->mac, netcfg.mac, ETH_ALEN); |
b356f76d | 91 | qemu_format_nic_info_str(qemu_get_queue(n->nic), n->mac); |
0f03eca6 AL |
92 | } |
93 | } | |
94 | ||
783e7706 MT |
95 | static bool virtio_net_started(VirtIONet *n, uint8_t status) |
96 | { | |
17a0ca55 | 97 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
783e7706 | 98 | return (status & VIRTIO_CONFIG_S_DRIVER_OK) && |
17a0ca55 | 99 | (n->status & VIRTIO_NET_S_LINK_UP) && vdev->vm_running; |
783e7706 MT |
100 | } |
101 | ||
f57fcf70 JW |
102 | static void virtio_net_announce_timer(void *opaque) |
103 | { | |
104 | VirtIONet *n = opaque; | |
105 | VirtIODevice *vdev = VIRTIO_DEVICE(n); | |
106 | ||
107 | n->announce_counter--; | |
108 | n->status |= VIRTIO_NET_S_ANNOUNCE; | |
109 | virtio_notify_config(vdev); | |
110 | } | |
111 | ||
783e7706 | 112 | static void virtio_net_vhost_status(VirtIONet *n, uint8_t status) |
afbaa7b4 | 113 | { |
17a0ca55 | 114 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
b356f76d | 115 | NetClientState *nc = qemu_get_queue(n->nic); |
fed699f9 | 116 | int queues = n->multiqueue ? n->max_queues : 1; |
b356f76d | 117 | |
ed8b4afe | 118 | if (!get_vhost_net(nc->peer)) { |
afbaa7b4 MT |
119 | return; |
120 | } | |
fed699f9 | 121 | |
d7108d90 JW |
122 | if (!!n->vhost_started == |
123 | (virtio_net_started(n, status) && !nc->peer->link_down)) { | |
afbaa7b4 MT |
124 | return; |
125 | } | |
126 | if (!n->vhost_started) { | |
5430a28f | 127 | int r; |
ed8b4afe | 128 | if (!vhost_net_query(get_vhost_net(nc->peer), vdev)) { |
5430a28f MT |
129 | return; |
130 | } | |
1830b80f | 131 | n->vhost_started = 1; |
17a0ca55 | 132 | r = vhost_net_start(vdev, n->nic->ncs, queues); |
afbaa7b4 | 133 | if (r < 0) { |
e7b43f7e SH |
134 | error_report("unable to start vhost net: %d: " |
135 | "falling back on userspace virtio", -r); | |
1830b80f | 136 | n->vhost_started = 0; |
afbaa7b4 MT |
137 | } |
138 | } else { | |
17a0ca55 | 139 | vhost_net_stop(vdev, n->nic->ncs, queues); |
afbaa7b4 MT |
140 | n->vhost_started = 0; |
141 | } | |
142 | } | |
143 | ||
783e7706 MT |
144 | static void virtio_net_set_status(struct VirtIODevice *vdev, uint8_t status) |
145 | { | |
17a0ca55 | 146 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 JW |
147 | VirtIONetQueue *q; |
148 | int i; | |
149 | uint8_t queue_status; | |
783e7706 MT |
150 | |
151 | virtio_net_vhost_status(n, status); | |
152 | ||
fed699f9 JW |
153 | for (i = 0; i < n->max_queues; i++) { |
154 | q = &n->vqs[i]; | |
783e7706 | 155 | |
fed699f9 JW |
156 | if ((!n->multiqueue && i != 0) || i >= n->curr_queues) { |
157 | queue_status = 0; | |
783e7706 | 158 | } else { |
fed699f9 | 159 | queue_status = status; |
783e7706 | 160 | } |
fed699f9 JW |
161 | |
162 | if (!q->tx_waiting) { | |
163 | continue; | |
164 | } | |
165 | ||
166 | if (virtio_net_started(n, queue_status) && !n->vhost_started) { | |
167 | if (q->tx_timer) { | |
bc72ad67 AB |
168 | timer_mod(q->tx_timer, |
169 | qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); | |
fed699f9 JW |
170 | } else { |
171 | qemu_bh_schedule(q->tx_bh); | |
172 | } | |
783e7706 | 173 | } else { |
fed699f9 | 174 | if (q->tx_timer) { |
bc72ad67 | 175 | timer_del(q->tx_timer); |
fed699f9 JW |
176 | } else { |
177 | qemu_bh_cancel(q->tx_bh); | |
178 | } | |
783e7706 MT |
179 | } |
180 | } | |
181 | } | |
182 | ||
4e68f7a0 | 183 | static void virtio_net_set_link_status(NetClientState *nc) |
554c97dd | 184 | { |
cc1f0f45 | 185 | VirtIONet *n = qemu_get_nic_opaque(nc); |
17a0ca55 | 186 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
554c97dd AL |
187 | uint16_t old_status = n->status; |
188 | ||
eb6b6c12 | 189 | if (nc->link_down) |
554c97dd AL |
190 | n->status &= ~VIRTIO_NET_S_LINK_UP; |
191 | else | |
192 | n->status |= VIRTIO_NET_S_LINK_UP; | |
193 | ||
194 | if (n->status != old_status) | |
17a0ca55 | 195 | virtio_notify_config(vdev); |
afbaa7b4 | 196 | |
17a0ca55 | 197 | virtio_net_set_status(vdev, vdev->status); |
554c97dd AL |
198 | } |
199 | ||
b1be4280 AK |
200 | static void rxfilter_notify(NetClientState *nc) |
201 | { | |
202 | QObject *event_data; | |
203 | VirtIONet *n = qemu_get_nic_opaque(nc); | |
204 | ||
205 | if (nc->rxfilter_notify_enabled) { | |
96e35046 | 206 | gchar *path = object_get_canonical_path(OBJECT(n->qdev)); |
b1be4280 AK |
207 | if (n->netclient_name) { |
208 | event_data = qobject_from_jsonf("{ 'name': %s, 'path': %s }", | |
96e35046 | 209 | n->netclient_name, path); |
b1be4280 | 210 | } else { |
96e35046 | 211 | event_data = qobject_from_jsonf("{ 'path': %s }", path); |
b1be4280 AK |
212 | } |
213 | monitor_protocol_event(QEVENT_NIC_RX_FILTER_CHANGED, event_data); | |
214 | qobject_decref(event_data); | |
96e35046 | 215 | g_free(path); |
b1be4280 AK |
216 | |
217 | /* disable event notification to avoid events flooding */ | |
218 | nc->rxfilter_notify_enabled = 0; | |
219 | } | |
220 | } | |
221 | ||
222 | static char *mac_strdup_printf(const uint8_t *mac) | |
223 | { | |
224 | return g_strdup_printf("%.2x:%.2x:%.2x:%.2x:%.2x:%.2x", mac[0], | |
225 | mac[1], mac[2], mac[3], mac[4], mac[5]); | |
226 | } | |
227 | ||
f7bc8ef8 AK |
228 | static intList *get_vlan_table(VirtIONet *n) |
229 | { | |
230 | intList *list, *entry; | |
231 | int i, j; | |
232 | ||
233 | list = NULL; | |
234 | for (i = 0; i < MAX_VLAN >> 5; i++) { | |
235 | for (j = 0; n->vlans[i] && j <= 0x1f; j++) { | |
236 | if (n->vlans[i] & (1U << j)) { | |
237 | entry = g_malloc0(sizeof(*entry)); | |
238 | entry->value = (i << 5) + j; | |
239 | entry->next = list; | |
240 | list = entry; | |
241 | } | |
242 | } | |
243 | } | |
244 | ||
245 | return list; | |
246 | } | |
247 | ||
b1be4280 AK |
248 | static RxFilterInfo *virtio_net_query_rxfilter(NetClientState *nc) |
249 | { | |
250 | VirtIONet *n = qemu_get_nic_opaque(nc); | |
f7bc8ef8 | 251 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
b1be4280 AK |
252 | RxFilterInfo *info; |
253 | strList *str_list, *entry; | |
f7bc8ef8 | 254 | int i; |
b1be4280 AK |
255 | |
256 | info = g_malloc0(sizeof(*info)); | |
257 | info->name = g_strdup(nc->name); | |
258 | info->promiscuous = n->promisc; | |
259 | ||
260 | if (n->nouni) { | |
261 | info->unicast = RX_STATE_NONE; | |
262 | } else if (n->alluni) { | |
263 | info->unicast = RX_STATE_ALL; | |
264 | } else { | |
265 | info->unicast = RX_STATE_NORMAL; | |
266 | } | |
267 | ||
268 | if (n->nomulti) { | |
269 | info->multicast = RX_STATE_NONE; | |
270 | } else if (n->allmulti) { | |
271 | info->multicast = RX_STATE_ALL; | |
272 | } else { | |
273 | info->multicast = RX_STATE_NORMAL; | |
274 | } | |
275 | ||
276 | info->broadcast_allowed = n->nobcast; | |
277 | info->multicast_overflow = n->mac_table.multi_overflow; | |
278 | info->unicast_overflow = n->mac_table.uni_overflow; | |
279 | ||
280 | info->main_mac = mac_strdup_printf(n->mac); | |
281 | ||
282 | str_list = NULL; | |
283 | for (i = 0; i < n->mac_table.first_multi; i++) { | |
284 | entry = g_malloc0(sizeof(*entry)); | |
285 | entry->value = mac_strdup_printf(n->mac_table.macs + i * ETH_ALEN); | |
286 | entry->next = str_list; | |
287 | str_list = entry; | |
288 | } | |
289 | info->unicast_table = str_list; | |
290 | ||
291 | str_list = NULL; | |
292 | for (i = n->mac_table.first_multi; i < n->mac_table.in_use; i++) { | |
293 | entry = g_malloc0(sizeof(*entry)); | |
294 | entry->value = mac_strdup_printf(n->mac_table.macs + i * ETH_ALEN); | |
295 | entry->next = str_list; | |
296 | str_list = entry; | |
297 | } | |
298 | info->multicast_table = str_list; | |
f7bc8ef8 | 299 | info->vlan_table = get_vlan_table(n); |
b1be4280 | 300 | |
f7bc8ef8 AK |
301 | if (!((1 << VIRTIO_NET_F_CTRL_VLAN) & vdev->guest_features)) { |
302 | info->vlan = RX_STATE_ALL; | |
303 | } else if (!info->vlan_table) { | |
304 | info->vlan = RX_STATE_NONE; | |
305 | } else { | |
306 | info->vlan = RX_STATE_NORMAL; | |
b1be4280 | 307 | } |
b1be4280 AK |
308 | |
309 | /* enable event notification after query */ | |
310 | nc->rxfilter_notify_enabled = 1; | |
311 | ||
312 | return info; | |
313 | } | |
314 | ||
002437cd AL |
315 | static void virtio_net_reset(VirtIODevice *vdev) |
316 | { | |
17a0ca55 | 317 | VirtIONet *n = VIRTIO_NET(vdev); |
002437cd AL |
318 | |
319 | /* Reset back to compatibility mode */ | |
320 | n->promisc = 1; | |
321 | n->allmulti = 0; | |
015cb166 AW |
322 | n->alluni = 0; |
323 | n->nomulti = 0; | |
324 | n->nouni = 0; | |
325 | n->nobcast = 0; | |
fed699f9 JW |
326 | /* multiqueue is disabled by default */ |
327 | n->curr_queues = 1; | |
f57fcf70 JW |
328 | timer_del(n->announce_timer); |
329 | n->announce_counter = 0; | |
330 | n->status &= ~VIRTIO_NET_S_ANNOUNCE; | |
b6503ed9 | 331 | |
f21c0ed9 | 332 | /* Flush any MAC and VLAN filter table state */ |
b6503ed9 | 333 | n->mac_table.in_use = 0; |
2d9aba39 | 334 | n->mac_table.first_multi = 0; |
8fd2a2f1 AW |
335 | n->mac_table.multi_overflow = 0; |
336 | n->mac_table.uni_overflow = 0; | |
b6503ed9 | 337 | memset(n->mac_table.macs, 0, MAC_TABLE_ENTRIES * ETH_ALEN); |
41dc8a67 | 338 | memcpy(&n->mac[0], &n->nic->conf->macaddr, sizeof(n->mac)); |
702d66a8 | 339 | qemu_format_nic_info_str(qemu_get_queue(n->nic), n->mac); |
f21c0ed9 | 340 | memset(n->vlans, 0, MAX_VLAN >> 3); |
002437cd AL |
341 | } |
342 | ||
6e371ab8 | 343 | static void peer_test_vnet_hdr(VirtIONet *n) |
3a330134 | 344 | { |
b356f76d JW |
345 | NetClientState *nc = qemu_get_queue(n->nic); |
346 | if (!nc->peer) { | |
6e371ab8 | 347 | return; |
b356f76d | 348 | } |
3a330134 | 349 | |
d6085e3a | 350 | n->has_vnet_hdr = qemu_has_vnet_hdr(nc->peer); |
6e371ab8 | 351 | } |
3a330134 | 352 | |
6e371ab8 MT |
353 | static int peer_has_vnet_hdr(VirtIONet *n) |
354 | { | |
3a330134 MM |
355 | return n->has_vnet_hdr; |
356 | } | |
357 | ||
0ce0e8f4 MM |
358 | static int peer_has_ufo(VirtIONet *n) |
359 | { | |
360 | if (!peer_has_vnet_hdr(n)) | |
361 | return 0; | |
362 | ||
d6085e3a | 363 | n->has_ufo = qemu_has_ufo(qemu_get_queue(n->nic)->peer); |
0ce0e8f4 MM |
364 | |
365 | return n->has_ufo; | |
366 | } | |
367 | ||
ff3a8066 MT |
368 | static void virtio_net_set_mrg_rx_bufs(VirtIONet *n, int mergeable_rx_bufs) |
369 | { | |
fed699f9 JW |
370 | int i; |
371 | NetClientState *nc; | |
372 | ||
ff3a8066 MT |
373 | n->mergeable_rx_bufs = mergeable_rx_bufs; |
374 | ||
375 | n->guest_hdr_len = n->mergeable_rx_bufs ? | |
376 | sizeof(struct virtio_net_hdr_mrg_rxbuf) : sizeof(struct virtio_net_hdr); | |
377 | ||
fed699f9 JW |
378 | for (i = 0; i < n->max_queues; i++) { |
379 | nc = qemu_get_subqueue(n->nic, i); | |
380 | ||
381 | if (peer_has_vnet_hdr(n) && | |
d6085e3a SH |
382 | qemu_has_vnet_hdr_len(nc->peer, n->guest_hdr_len)) { |
383 | qemu_set_vnet_hdr_len(nc->peer, n->guest_hdr_len); | |
fed699f9 JW |
384 | n->host_hdr_len = n->guest_hdr_len; |
385 | } | |
ff3a8066 MT |
386 | } |
387 | } | |
388 | ||
fed699f9 JW |
389 | static int peer_attach(VirtIONet *n, int index) |
390 | { | |
391 | NetClientState *nc = qemu_get_subqueue(n->nic, index); | |
392 | ||
393 | if (!nc->peer) { | |
394 | return 0; | |
395 | } | |
396 | ||
397 | if (nc->peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) { | |
398 | return 0; | |
399 | } | |
400 | ||
401 | return tap_enable(nc->peer); | |
402 | } | |
403 | ||
404 | static int peer_detach(VirtIONet *n, int index) | |
405 | { | |
406 | NetClientState *nc = qemu_get_subqueue(n->nic, index); | |
407 | ||
408 | if (!nc->peer) { | |
409 | return 0; | |
410 | } | |
411 | ||
412 | if (nc->peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) { | |
413 | return 0; | |
414 | } | |
415 | ||
416 | return tap_disable(nc->peer); | |
417 | } | |
418 | ||
419 | static void virtio_net_set_queues(VirtIONet *n) | |
420 | { | |
421 | int i; | |
ddfa83ea | 422 | int r; |
fed699f9 JW |
423 | |
424 | for (i = 0; i < n->max_queues; i++) { | |
425 | if (i < n->curr_queues) { | |
ddfa83ea JS |
426 | r = peer_attach(n, i); |
427 | assert(!r); | |
fed699f9 | 428 | } else { |
ddfa83ea JS |
429 | r = peer_detach(n, i); |
430 | assert(!r); | |
fed699f9 JW |
431 | } |
432 | } | |
433 | } | |
434 | ||
ec57db16 | 435 | static void virtio_net_set_multiqueue(VirtIONet *n, int multiqueue); |
fed699f9 | 436 | |
8172539d | 437 | static uint32_t virtio_net_get_features(VirtIODevice *vdev, uint32_t features) |
fbe78f4f | 438 | { |
17a0ca55 | 439 | VirtIONet *n = VIRTIO_NET(vdev); |
b356f76d | 440 | NetClientState *nc = qemu_get_queue(n->nic); |
fbe78f4f | 441 | |
c9f79a3f MT |
442 | features |= (1 << VIRTIO_NET_F_MAC); |
443 | ||
6e371ab8 | 444 | if (!peer_has_vnet_hdr(n)) { |
8172539d MT |
445 | features &= ~(0x1 << VIRTIO_NET_F_CSUM); |
446 | features &= ~(0x1 << VIRTIO_NET_F_HOST_TSO4); | |
447 | features &= ~(0x1 << VIRTIO_NET_F_HOST_TSO6); | |
448 | features &= ~(0x1 << VIRTIO_NET_F_HOST_ECN); | |
449 | ||
450 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_CSUM); | |
451 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_TSO4); | |
452 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_TSO6); | |
453 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_ECN); | |
454 | } | |
3a330134 | 455 | |
8172539d MT |
456 | if (!peer_has_vnet_hdr(n) || !peer_has_ufo(n)) { |
457 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_UFO); | |
458 | features &= ~(0x1 << VIRTIO_NET_F_HOST_UFO); | |
3a330134 MM |
459 | } |
460 | ||
ed8b4afe | 461 | if (!get_vhost_net(nc->peer)) { |
9bc6304c MT |
462 | return features; |
463 | } | |
ed8b4afe | 464 | return vhost_net_get_features(get_vhost_net(nc->peer), features); |
fbe78f4f AL |
465 | } |
466 | ||
8eca6b1b AL |
467 | static uint32_t virtio_net_bad_features(VirtIODevice *vdev) |
468 | { | |
469 | uint32_t features = 0; | |
470 | ||
471 | /* Linux kernel 2.6.25. It understood MAC (as everyone must), | |
472 | * but also these: */ | |
473 | features |= (1 << VIRTIO_NET_F_MAC); | |
184bd048 DK |
474 | features |= (1 << VIRTIO_NET_F_CSUM); |
475 | features |= (1 << VIRTIO_NET_F_HOST_TSO4); | |
476 | features |= (1 << VIRTIO_NET_F_HOST_TSO6); | |
477 | features |= (1 << VIRTIO_NET_F_HOST_ECN); | |
8eca6b1b | 478 | |
8172539d | 479 | return features; |
8eca6b1b AL |
480 | } |
481 | ||
644c9858 DF |
482 | static void virtio_net_apply_guest_offloads(VirtIONet *n) |
483 | { | |
ad37bb3b | 484 | qemu_set_offload(qemu_get_queue(n->nic)->peer, |
644c9858 DF |
485 | !!(n->curr_guest_offloads & (1ULL << VIRTIO_NET_F_GUEST_CSUM)), |
486 | !!(n->curr_guest_offloads & (1ULL << VIRTIO_NET_F_GUEST_TSO4)), | |
487 | !!(n->curr_guest_offloads & (1ULL << VIRTIO_NET_F_GUEST_TSO6)), | |
488 | !!(n->curr_guest_offloads & (1ULL << VIRTIO_NET_F_GUEST_ECN)), | |
489 | !!(n->curr_guest_offloads & (1ULL << VIRTIO_NET_F_GUEST_UFO))); | |
490 | } | |
491 | ||
492 | static uint64_t virtio_net_guest_offloads_by_features(uint32_t features) | |
493 | { | |
494 | static const uint64_t guest_offloads_mask = | |
495 | (1ULL << VIRTIO_NET_F_GUEST_CSUM) | | |
496 | (1ULL << VIRTIO_NET_F_GUEST_TSO4) | | |
497 | (1ULL << VIRTIO_NET_F_GUEST_TSO6) | | |
498 | (1ULL << VIRTIO_NET_F_GUEST_ECN) | | |
499 | (1ULL << VIRTIO_NET_F_GUEST_UFO); | |
500 | ||
501 | return guest_offloads_mask & features; | |
502 | } | |
503 | ||
504 | static inline uint64_t virtio_net_supported_guest_offloads(VirtIONet *n) | |
505 | { | |
506 | VirtIODevice *vdev = VIRTIO_DEVICE(n); | |
507 | return virtio_net_guest_offloads_by_features(vdev->guest_features); | |
508 | } | |
509 | ||
fbe78f4f AL |
510 | static void virtio_net_set_features(VirtIODevice *vdev, uint32_t features) |
511 | { | |
17a0ca55 | 512 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 JW |
513 | int i; |
514 | ||
ec57db16 | 515 | virtio_net_set_multiqueue(n, !!(features & (1 << VIRTIO_NET_F_MQ))); |
fbe78f4f | 516 | |
ff3a8066 | 517 | virtio_net_set_mrg_rx_bufs(n, !!(features & (1 << VIRTIO_NET_F_MRG_RXBUF))); |
f5436dd9 MM |
518 | |
519 | if (n->has_vnet_hdr) { | |
644c9858 DF |
520 | n->curr_guest_offloads = |
521 | virtio_net_guest_offloads_by_features(features); | |
522 | virtio_net_apply_guest_offloads(n); | |
f5436dd9 | 523 | } |
fed699f9 JW |
524 | |
525 | for (i = 0; i < n->max_queues; i++) { | |
526 | NetClientState *nc = qemu_get_subqueue(n->nic, i); | |
527 | ||
ed8b4afe | 528 | if (!get_vhost_net(nc->peer)) { |
fed699f9 JW |
529 | continue; |
530 | } | |
ed8b4afe | 531 | vhost_net_ack_features(get_vhost_net(nc->peer), features); |
dc14a397 | 532 | } |
0b1eaa88 SF |
533 | |
534 | if ((1 << VIRTIO_NET_F_CTRL_VLAN) & features) { | |
535 | memset(n->vlans, 0, MAX_VLAN >> 3); | |
536 | } else { | |
537 | memset(n->vlans, 0xff, MAX_VLAN >> 3); | |
538 | } | |
fbe78f4f AL |
539 | } |
540 | ||
002437cd | 541 | static int virtio_net_handle_rx_mode(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 542 | struct iovec *iov, unsigned int iov_cnt) |
002437cd AL |
543 | { |
544 | uint8_t on; | |
921ac5d0 | 545 | size_t s; |
b1be4280 | 546 | NetClientState *nc = qemu_get_queue(n->nic); |
002437cd | 547 | |
921ac5d0 MT |
548 | s = iov_to_buf(iov, iov_cnt, 0, &on, sizeof(on)); |
549 | if (s != sizeof(on)) { | |
550 | return VIRTIO_NET_ERR; | |
002437cd AL |
551 | } |
552 | ||
dd23454b | 553 | if (cmd == VIRTIO_NET_CTRL_RX_PROMISC) { |
002437cd | 554 | n->promisc = on; |
dd23454b | 555 | } else if (cmd == VIRTIO_NET_CTRL_RX_ALLMULTI) { |
002437cd | 556 | n->allmulti = on; |
dd23454b | 557 | } else if (cmd == VIRTIO_NET_CTRL_RX_ALLUNI) { |
015cb166 | 558 | n->alluni = on; |
dd23454b | 559 | } else if (cmd == VIRTIO_NET_CTRL_RX_NOMULTI) { |
015cb166 | 560 | n->nomulti = on; |
dd23454b | 561 | } else if (cmd == VIRTIO_NET_CTRL_RX_NOUNI) { |
015cb166 | 562 | n->nouni = on; |
dd23454b | 563 | } else if (cmd == VIRTIO_NET_CTRL_RX_NOBCAST) { |
015cb166 | 564 | n->nobcast = on; |
921ac5d0 | 565 | } else { |
002437cd | 566 | return VIRTIO_NET_ERR; |
921ac5d0 | 567 | } |
002437cd | 568 | |
b1be4280 AK |
569 | rxfilter_notify(nc); |
570 | ||
002437cd AL |
571 | return VIRTIO_NET_OK; |
572 | } | |
573 | ||
644c9858 DF |
574 | static int virtio_net_handle_offloads(VirtIONet *n, uint8_t cmd, |
575 | struct iovec *iov, unsigned int iov_cnt) | |
576 | { | |
577 | VirtIODevice *vdev = VIRTIO_DEVICE(n); | |
578 | uint64_t offloads; | |
579 | size_t s; | |
580 | ||
581 | if (!((1 << VIRTIO_NET_F_CTRL_GUEST_OFFLOADS) & vdev->guest_features)) { | |
582 | return VIRTIO_NET_ERR; | |
583 | } | |
584 | ||
585 | s = iov_to_buf(iov, iov_cnt, 0, &offloads, sizeof(offloads)); | |
586 | if (s != sizeof(offloads)) { | |
587 | return VIRTIO_NET_ERR; | |
588 | } | |
589 | ||
590 | if (cmd == VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET) { | |
591 | uint64_t supported_offloads; | |
592 | ||
593 | if (!n->has_vnet_hdr) { | |
594 | return VIRTIO_NET_ERR; | |
595 | } | |
596 | ||
597 | supported_offloads = virtio_net_supported_guest_offloads(n); | |
598 | if (offloads & ~supported_offloads) { | |
599 | return VIRTIO_NET_ERR; | |
600 | } | |
601 | ||
602 | n->curr_guest_offloads = offloads; | |
603 | virtio_net_apply_guest_offloads(n); | |
604 | ||
605 | return VIRTIO_NET_OK; | |
606 | } else { | |
607 | return VIRTIO_NET_ERR; | |
608 | } | |
609 | } | |
610 | ||
b6503ed9 | 611 | static int virtio_net_handle_mac(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 612 | struct iovec *iov, unsigned int iov_cnt) |
b6503ed9 AL |
613 | { |
614 | struct virtio_net_ctrl_mac mac_data; | |
921ac5d0 | 615 | size_t s; |
b1be4280 | 616 | NetClientState *nc = qemu_get_queue(n->nic); |
b6503ed9 | 617 | |
c1943a3f AK |
618 | if (cmd == VIRTIO_NET_CTRL_MAC_ADDR_SET) { |
619 | if (iov_size(iov, iov_cnt) != sizeof(n->mac)) { | |
620 | return VIRTIO_NET_ERR; | |
621 | } | |
622 | s = iov_to_buf(iov, iov_cnt, 0, &n->mac, sizeof(n->mac)); | |
623 | assert(s == sizeof(n->mac)); | |
b356f76d | 624 | qemu_format_nic_info_str(qemu_get_queue(n->nic), n->mac); |
b1be4280 AK |
625 | rxfilter_notify(nc); |
626 | ||
c1943a3f AK |
627 | return VIRTIO_NET_OK; |
628 | } | |
629 | ||
921ac5d0 | 630 | if (cmd != VIRTIO_NET_CTRL_MAC_TABLE_SET) { |
b6503ed9 | 631 | return VIRTIO_NET_ERR; |
921ac5d0 | 632 | } |
b6503ed9 | 633 | |
cae2e556 AK |
634 | int in_use = 0; |
635 | int first_multi = 0; | |
636 | uint8_t uni_overflow = 0; | |
637 | uint8_t multi_overflow = 0; | |
638 | uint8_t *macs = g_malloc0(MAC_TABLE_ENTRIES * ETH_ALEN); | |
b6503ed9 | 639 | |
921ac5d0 MT |
640 | s = iov_to_buf(iov, iov_cnt, 0, &mac_data.entries, |
641 | sizeof(mac_data.entries)); | |
642 | mac_data.entries = ldl_p(&mac_data.entries); | |
643 | if (s != sizeof(mac_data.entries)) { | |
b1be4280 | 644 | goto error; |
921ac5d0 MT |
645 | } |
646 | iov_discard_front(&iov, &iov_cnt, s); | |
b6503ed9 | 647 | |
921ac5d0 | 648 | if (mac_data.entries * ETH_ALEN > iov_size(iov, iov_cnt)) { |
b1be4280 | 649 | goto error; |
921ac5d0 | 650 | } |
b6503ed9 AL |
651 | |
652 | if (mac_data.entries <= MAC_TABLE_ENTRIES) { | |
cae2e556 | 653 | s = iov_to_buf(iov, iov_cnt, 0, macs, |
921ac5d0 MT |
654 | mac_data.entries * ETH_ALEN); |
655 | if (s != mac_data.entries * ETH_ALEN) { | |
b1be4280 | 656 | goto error; |
921ac5d0 | 657 | } |
cae2e556 | 658 | in_use += mac_data.entries; |
b6503ed9 | 659 | } else { |
cae2e556 | 660 | uni_overflow = 1; |
b6503ed9 AL |
661 | } |
662 | ||
921ac5d0 MT |
663 | iov_discard_front(&iov, &iov_cnt, mac_data.entries * ETH_ALEN); |
664 | ||
cae2e556 | 665 | first_multi = in_use; |
2d9aba39 | 666 | |
921ac5d0 MT |
667 | s = iov_to_buf(iov, iov_cnt, 0, &mac_data.entries, |
668 | sizeof(mac_data.entries)); | |
669 | mac_data.entries = ldl_p(&mac_data.entries); | |
670 | if (s != sizeof(mac_data.entries)) { | |
b1be4280 | 671 | goto error; |
921ac5d0 MT |
672 | } |
673 | ||
674 | iov_discard_front(&iov, &iov_cnt, s); | |
b6503ed9 | 675 | |
921ac5d0 | 676 | if (mac_data.entries * ETH_ALEN != iov_size(iov, iov_cnt)) { |
b1be4280 | 677 | goto error; |
921ac5d0 | 678 | } |
b6503ed9 | 679 | |
edc24385 | 680 | if (mac_data.entries <= MAC_TABLE_ENTRIES - in_use) { |
cae2e556 | 681 | s = iov_to_buf(iov, iov_cnt, 0, &macs[in_use * ETH_ALEN], |
921ac5d0 MT |
682 | mac_data.entries * ETH_ALEN); |
683 | if (s != mac_data.entries * ETH_ALEN) { | |
b1be4280 | 684 | goto error; |
8fd2a2f1 | 685 | } |
cae2e556 | 686 | in_use += mac_data.entries; |
921ac5d0 | 687 | } else { |
cae2e556 | 688 | multi_overflow = 1; |
b6503ed9 AL |
689 | } |
690 | ||
cae2e556 AK |
691 | n->mac_table.in_use = in_use; |
692 | n->mac_table.first_multi = first_multi; | |
693 | n->mac_table.uni_overflow = uni_overflow; | |
694 | n->mac_table.multi_overflow = multi_overflow; | |
695 | memcpy(n->mac_table.macs, macs, MAC_TABLE_ENTRIES * ETH_ALEN); | |
696 | g_free(macs); | |
b1be4280 AK |
697 | rxfilter_notify(nc); |
698 | ||
b6503ed9 | 699 | return VIRTIO_NET_OK; |
b1be4280 AK |
700 | |
701 | error: | |
cae2e556 | 702 | g_free(macs); |
b1be4280 | 703 | return VIRTIO_NET_ERR; |
b6503ed9 AL |
704 | } |
705 | ||
f21c0ed9 | 706 | static int virtio_net_handle_vlan_table(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 707 | struct iovec *iov, unsigned int iov_cnt) |
f21c0ed9 AL |
708 | { |
709 | uint16_t vid; | |
921ac5d0 | 710 | size_t s; |
b1be4280 | 711 | NetClientState *nc = qemu_get_queue(n->nic); |
f21c0ed9 | 712 | |
921ac5d0 MT |
713 | s = iov_to_buf(iov, iov_cnt, 0, &vid, sizeof(vid)); |
714 | vid = lduw_p(&vid); | |
715 | if (s != sizeof(vid)) { | |
f21c0ed9 AL |
716 | return VIRTIO_NET_ERR; |
717 | } | |
718 | ||
f21c0ed9 AL |
719 | if (vid >= MAX_VLAN) |
720 | return VIRTIO_NET_ERR; | |
721 | ||
722 | if (cmd == VIRTIO_NET_CTRL_VLAN_ADD) | |
723 | n->vlans[vid >> 5] |= (1U << (vid & 0x1f)); | |
724 | else if (cmd == VIRTIO_NET_CTRL_VLAN_DEL) | |
725 | n->vlans[vid >> 5] &= ~(1U << (vid & 0x1f)); | |
726 | else | |
727 | return VIRTIO_NET_ERR; | |
728 | ||
b1be4280 AK |
729 | rxfilter_notify(nc); |
730 | ||
f21c0ed9 AL |
731 | return VIRTIO_NET_OK; |
732 | } | |
733 | ||
f57fcf70 JW |
734 | static int virtio_net_handle_announce(VirtIONet *n, uint8_t cmd, |
735 | struct iovec *iov, unsigned int iov_cnt) | |
736 | { | |
737 | if (cmd == VIRTIO_NET_CTRL_ANNOUNCE_ACK && | |
738 | n->status & VIRTIO_NET_S_ANNOUNCE) { | |
739 | n->status &= ~VIRTIO_NET_S_ANNOUNCE; | |
740 | if (n->announce_counter) { | |
741 | timer_mod(n->announce_timer, | |
742 | qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
743 | self_announce_delay(n->announce_counter)); | |
744 | } | |
745 | return VIRTIO_NET_OK; | |
746 | } else { | |
747 | return VIRTIO_NET_ERR; | |
748 | } | |
749 | } | |
750 | ||
fed699f9 | 751 | static int virtio_net_handle_mq(VirtIONet *n, uint8_t cmd, |
f8f7c533 | 752 | struct iovec *iov, unsigned int iov_cnt) |
fed699f9 | 753 | { |
17a0ca55 | 754 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
f8f7c533 JW |
755 | struct virtio_net_ctrl_mq mq; |
756 | size_t s; | |
757 | uint16_t queues; | |
fed699f9 | 758 | |
f8f7c533 JW |
759 | s = iov_to_buf(iov, iov_cnt, 0, &mq, sizeof(mq)); |
760 | if (s != sizeof(mq)) { | |
fed699f9 JW |
761 | return VIRTIO_NET_ERR; |
762 | } | |
763 | ||
764 | if (cmd != VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET) { | |
765 | return VIRTIO_NET_ERR; | |
766 | } | |
767 | ||
f8f7c533 | 768 | queues = lduw_p(&mq.virtqueue_pairs); |
fed699f9 | 769 | |
f8f7c533 JW |
770 | if (queues < VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN || |
771 | queues > VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX || | |
772 | queues > n->max_queues || | |
fed699f9 JW |
773 | !n->multiqueue) { |
774 | return VIRTIO_NET_ERR; | |
775 | } | |
776 | ||
f8f7c533 | 777 | n->curr_queues = queues; |
fed699f9 JW |
778 | /* stop the backend before changing the number of queues to avoid handling a |
779 | * disabled queue */ | |
17a0ca55 | 780 | virtio_net_set_status(vdev, vdev->status); |
fed699f9 JW |
781 | virtio_net_set_queues(n); |
782 | ||
783 | return VIRTIO_NET_OK; | |
784 | } | |
3d11d36c AL |
785 | static void virtio_net_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq) |
786 | { | |
17a0ca55 | 787 | VirtIONet *n = VIRTIO_NET(vdev); |
3d11d36c AL |
788 | struct virtio_net_ctrl_hdr ctrl; |
789 | virtio_net_ctrl_ack status = VIRTIO_NET_ERR; | |
790 | VirtQueueElement elem; | |
921ac5d0 MT |
791 | size_t s; |
792 | struct iovec *iov; | |
793 | unsigned int iov_cnt; | |
3d11d36c AL |
794 | |
795 | while (virtqueue_pop(vq, &elem)) { | |
921ac5d0 MT |
796 | if (iov_size(elem.in_sg, elem.in_num) < sizeof(status) || |
797 | iov_size(elem.out_sg, elem.out_num) < sizeof(ctrl)) { | |
e7b43f7e | 798 | error_report("virtio-net ctrl missing headers"); |
3d11d36c AL |
799 | exit(1); |
800 | } | |
801 | ||
921ac5d0 MT |
802 | iov = elem.out_sg; |
803 | iov_cnt = elem.out_num; | |
804 | s = iov_to_buf(iov, iov_cnt, 0, &ctrl, sizeof(ctrl)); | |
805 | iov_discard_front(&iov, &iov_cnt, sizeof(ctrl)); | |
806 | if (s != sizeof(ctrl)) { | |
807 | status = VIRTIO_NET_ERR; | |
dd23454b | 808 | } else if (ctrl.class == VIRTIO_NET_CTRL_RX) { |
921ac5d0 MT |
809 | status = virtio_net_handle_rx_mode(n, ctrl.cmd, iov, iov_cnt); |
810 | } else if (ctrl.class == VIRTIO_NET_CTRL_MAC) { | |
811 | status = virtio_net_handle_mac(n, ctrl.cmd, iov, iov_cnt); | |
812 | } else if (ctrl.class == VIRTIO_NET_CTRL_VLAN) { | |
813 | status = virtio_net_handle_vlan_table(n, ctrl.cmd, iov, iov_cnt); | |
f57fcf70 JW |
814 | } else if (ctrl.class == VIRTIO_NET_CTRL_ANNOUNCE) { |
815 | status = virtio_net_handle_announce(n, ctrl.cmd, iov, iov_cnt); | |
fed699f9 | 816 | } else if (ctrl.class == VIRTIO_NET_CTRL_MQ) { |
f8f7c533 | 817 | status = virtio_net_handle_mq(n, ctrl.cmd, iov, iov_cnt); |
644c9858 DF |
818 | } else if (ctrl.class == VIRTIO_NET_CTRL_GUEST_OFFLOADS) { |
819 | status = virtio_net_handle_offloads(n, ctrl.cmd, iov, iov_cnt); | |
3d11d36c AL |
820 | } |
821 | ||
921ac5d0 MT |
822 | s = iov_from_buf(elem.in_sg, elem.in_num, 0, &status, sizeof(status)); |
823 | assert(s == sizeof(status)); | |
3d11d36c AL |
824 | |
825 | virtqueue_push(vq, &elem, sizeof(status)); | |
826 | virtio_notify(vdev, vq); | |
827 | } | |
828 | } | |
829 | ||
fbe78f4f AL |
830 | /* RX */ |
831 | ||
832 | static void virtio_net_handle_rx(VirtIODevice *vdev, VirtQueue *vq) | |
833 | { | |
17a0ca55 | 834 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 | 835 | int queue_index = vq2q(virtio_get_queue_index(vq)); |
8aeff62d | 836 | |
fed699f9 | 837 | qemu_flush_queued_packets(qemu_get_subqueue(n->nic, queue_index)); |
fbe78f4f AL |
838 | } |
839 | ||
4e68f7a0 | 840 | static int virtio_net_can_receive(NetClientState *nc) |
fbe78f4f | 841 | { |
cc1f0f45 | 842 | VirtIONet *n = qemu_get_nic_opaque(nc); |
17a0ca55 | 843 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
fed699f9 | 844 | VirtIONetQueue *q = virtio_net_get_subqueue(nc); |
0c87e93e | 845 | |
17a0ca55 | 846 | if (!vdev->vm_running) { |
95477323 MT |
847 | return 0; |
848 | } | |
cdd5cc12 | 849 | |
fed699f9 JW |
850 | if (nc->queue_index >= n->curr_queues) { |
851 | return 0; | |
852 | } | |
853 | ||
0c87e93e | 854 | if (!virtio_queue_ready(q->rx_vq) || |
17a0ca55 | 855 | !(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) { |
fbe78f4f | 856 | return 0; |
0c87e93e | 857 | } |
fbe78f4f | 858 | |
cdd5cc12 MM |
859 | return 1; |
860 | } | |
861 | ||
0c87e93e | 862 | static int virtio_net_has_buffers(VirtIONetQueue *q, int bufsize) |
cdd5cc12 | 863 | { |
0c87e93e JW |
864 | VirtIONet *n = q->n; |
865 | if (virtio_queue_empty(q->rx_vq) || | |
fbe78f4f | 866 | (n->mergeable_rx_bufs && |
0c87e93e JW |
867 | !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) { |
868 | virtio_queue_set_notification(q->rx_vq, 1); | |
06b12970 TL |
869 | |
870 | /* To avoid a race condition where the guest has made some buffers | |
871 | * available after the above check but before notification was | |
872 | * enabled, check for available buffers again. | |
873 | */ | |
0c87e93e | 874 | if (virtio_queue_empty(q->rx_vq) || |
06b12970 | 875 | (n->mergeable_rx_bufs && |
0c87e93e | 876 | !virtqueue_avail_bytes(q->rx_vq, bufsize, 0))) { |
06b12970 | 877 | return 0; |
0c87e93e | 878 | } |
fbe78f4f AL |
879 | } |
880 | ||
0c87e93e | 881 | virtio_queue_set_notification(q->rx_vq, 0); |
fbe78f4f AL |
882 | return 1; |
883 | } | |
884 | ||
1d41b0c1 AL |
885 | /* dhclient uses AF_PACKET but doesn't pass auxdata to the kernel so |
886 | * it never finds out that the packets don't have valid checksums. This | |
887 | * causes dhclient to get upset. Fedora's carried a patch for ages to | |
888 | * fix this with Xen but it hasn't appeared in an upstream release of | |
889 | * dhclient yet. | |
890 | * | |
891 | * To avoid breaking existing guests, we catch udp packets and add | |
892 | * checksums. This is terrible but it's better than hacking the guest | |
893 | * kernels. | |
894 | * | |
895 | * N.B. if we introduce a zero-copy API, this operation is no longer free so | |
896 | * we should provide a mechanism to disable it to avoid polluting the host | |
897 | * cache. | |
898 | */ | |
899 | static void work_around_broken_dhclient(struct virtio_net_hdr *hdr, | |
22cc84db | 900 | uint8_t *buf, size_t size) |
1d41b0c1 AL |
901 | { |
902 | if ((hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && /* missing csum */ | |
903 | (size > 27 && size < 1500) && /* normal sized MTU */ | |
904 | (buf[12] == 0x08 && buf[13] == 0x00) && /* ethertype == IPv4 */ | |
905 | (buf[23] == 17) && /* ip.protocol == UDP */ | |
906 | (buf[34] == 0 && buf[35] == 67)) { /* udp.srcport == bootps */ | |
22cc84db | 907 | net_checksum_calculate(buf, size); |
1d41b0c1 AL |
908 | hdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM; |
909 | } | |
910 | } | |
911 | ||
280598b7 MT |
912 | static void receive_header(VirtIONet *n, const struct iovec *iov, int iov_cnt, |
913 | const void *buf, size_t size) | |
fbe78f4f | 914 | { |
3a330134 | 915 | if (n->has_vnet_hdr) { |
22cc84db MT |
916 | /* FIXME this cast is evil */ |
917 | void *wbuf = (void *)buf; | |
280598b7 MT |
918 | work_around_broken_dhclient(wbuf, wbuf + n->host_hdr_len, |
919 | size - n->host_hdr_len); | |
920 | iov_from_buf(iov, iov_cnt, 0, buf, sizeof(struct virtio_net_hdr)); | |
22cc84db MT |
921 | } else { |
922 | struct virtio_net_hdr hdr = { | |
923 | .flags = 0, | |
924 | .gso_type = VIRTIO_NET_HDR_GSO_NONE | |
925 | }; | |
926 | iov_from_buf(iov, iov_cnt, 0, &hdr, sizeof hdr); | |
3a330134 | 927 | } |
fbe78f4f AL |
928 | } |
929 | ||
3831ab20 AL |
930 | static int receive_filter(VirtIONet *n, const uint8_t *buf, int size) |
931 | { | |
932 | static const uint8_t bcast[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}; | |
f21c0ed9 | 933 | static const uint8_t vlan[] = {0x81, 0x00}; |
3831ab20 | 934 | uint8_t *ptr = (uint8_t *)buf; |
b6503ed9 | 935 | int i; |
3831ab20 AL |
936 | |
937 | if (n->promisc) | |
938 | return 1; | |
939 | ||
e043ebc6 | 940 | ptr += n->host_hdr_len; |
3a330134 | 941 | |
f21c0ed9 AL |
942 | if (!memcmp(&ptr[12], vlan, sizeof(vlan))) { |
943 | int vid = be16_to_cpup((uint16_t *)(ptr + 14)) & 0xfff; | |
944 | if (!(n->vlans[vid >> 5] & (1U << (vid & 0x1f)))) | |
945 | return 0; | |
946 | } | |
947 | ||
bbe2f399 AW |
948 | if (ptr[0] & 1) { // multicast |
949 | if (!memcmp(ptr, bcast, sizeof(bcast))) { | |
015cb166 AW |
950 | return !n->nobcast; |
951 | } else if (n->nomulti) { | |
952 | return 0; | |
8fd2a2f1 | 953 | } else if (n->allmulti || n->mac_table.multi_overflow) { |
bbe2f399 AW |
954 | return 1; |
955 | } | |
2d9aba39 AW |
956 | |
957 | for (i = n->mac_table.first_multi; i < n->mac_table.in_use; i++) { | |
958 | if (!memcmp(ptr, &n->mac_table.macs[i * ETH_ALEN], ETH_ALEN)) { | |
959 | return 1; | |
960 | } | |
961 | } | |
bbe2f399 | 962 | } else { // unicast |
015cb166 AW |
963 | if (n->nouni) { |
964 | return 0; | |
965 | } else if (n->alluni || n->mac_table.uni_overflow) { | |
8fd2a2f1 AW |
966 | return 1; |
967 | } else if (!memcmp(ptr, n->mac, ETH_ALEN)) { | |
bbe2f399 AW |
968 | return 1; |
969 | } | |
3831ab20 | 970 | |
2d9aba39 AW |
971 | for (i = 0; i < n->mac_table.first_multi; i++) { |
972 | if (!memcmp(ptr, &n->mac_table.macs[i * ETH_ALEN], ETH_ALEN)) { | |
973 | return 1; | |
974 | } | |
975 | } | |
b6503ed9 AL |
976 | } |
977 | ||
3831ab20 AL |
978 | return 0; |
979 | } | |
980 | ||
4e68f7a0 | 981 | static ssize_t virtio_net_receive(NetClientState *nc, const uint8_t *buf, size_t size) |
fbe78f4f | 982 | { |
cc1f0f45 | 983 | VirtIONet *n = qemu_get_nic_opaque(nc); |
fed699f9 | 984 | VirtIONetQueue *q = virtio_net_get_subqueue(nc); |
17a0ca55 | 985 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
63c58728 MT |
986 | struct iovec mhdr_sg[VIRTQUEUE_MAX_SIZE]; |
987 | struct virtio_net_hdr_mrg_rxbuf mhdr; | |
988 | unsigned mhdr_cnt = 0; | |
22cc84db | 989 | size_t offset, i, guest_offset; |
fbe78f4f | 990 | |
fed699f9 | 991 | if (!virtio_net_can_receive(nc)) { |
cdd5cc12 | 992 | return -1; |
b356f76d | 993 | } |
cdd5cc12 | 994 | |
940cda94 | 995 | /* hdr_len refers to the header we supply to the guest */ |
0c87e93e | 996 | if (!virtio_net_has_buffers(q, size + n->guest_hdr_len - n->host_hdr_len)) { |
8aeff62d | 997 | return 0; |
0c87e93e | 998 | } |
fbe78f4f | 999 | |
3831ab20 | 1000 | if (!receive_filter(n, buf, size)) |
4f1c942b | 1001 | return size; |
3831ab20 | 1002 | |
fbe78f4f AL |
1003 | offset = i = 0; |
1004 | ||
1005 | while (offset < size) { | |
1006 | VirtQueueElement elem; | |
1007 | int len, total; | |
22cc84db | 1008 | const struct iovec *sg = elem.in_sg; |
fbe78f4f | 1009 | |
22c253d9 | 1010 | total = 0; |
fbe78f4f | 1011 | |
0c87e93e | 1012 | if (virtqueue_pop(q->rx_vq, &elem) == 0) { |
fbe78f4f | 1013 | if (i == 0) |
4f1c942b | 1014 | return -1; |
e7b43f7e | 1015 | error_report("virtio-net unexpected empty queue: " |
279a4253 | 1016 | "i %zd mergeable %d offset %zd, size %zd, " |
e7b43f7e | 1017 | "guest hdr len %zd, host hdr len %zd guest features 0x%x", |
279a4253 | 1018 | i, n->mergeable_rx_bufs, offset, size, |
17a0ca55 | 1019 | n->guest_hdr_len, n->host_hdr_len, vdev->guest_features); |
fbe78f4f AL |
1020 | exit(1); |
1021 | } | |
1022 | ||
1023 | if (elem.in_num < 1) { | |
e7b43f7e | 1024 | error_report("virtio-net receive queue contains no in buffers"); |
fbe78f4f AL |
1025 | exit(1); |
1026 | } | |
1027 | ||
fbe78f4f | 1028 | if (i == 0) { |
c8d28e7e | 1029 | assert(offset == 0); |
63c58728 MT |
1030 | if (n->mergeable_rx_bufs) { |
1031 | mhdr_cnt = iov_copy(mhdr_sg, ARRAY_SIZE(mhdr_sg), | |
1032 | sg, elem.in_num, | |
1033 | offsetof(typeof(mhdr), num_buffers), | |
1034 | sizeof(mhdr.num_buffers)); | |
1035 | } | |
fbe78f4f | 1036 | |
c8d28e7e MT |
1037 | receive_header(n, sg, elem.in_num, buf, size); |
1038 | offset = n->host_hdr_len; | |
e35e23f6 | 1039 | total += n->guest_hdr_len; |
22cc84db MT |
1040 | guest_offset = n->guest_hdr_len; |
1041 | } else { | |
1042 | guest_offset = 0; | |
fbe78f4f AL |
1043 | } |
1044 | ||
1045 | /* copy in packet. ugh */ | |
22cc84db | 1046 | len = iov_from_buf(sg, elem.in_num, guest_offset, |
dcf6f5e1 | 1047 | buf + offset, size - offset); |
fbe78f4f | 1048 | total += len; |
279a4253 MT |
1049 | offset += len; |
1050 | /* If buffers can't be merged, at this point we | |
1051 | * must have consumed the complete packet. | |
1052 | * Otherwise, drop it. */ | |
1053 | if (!n->mergeable_rx_bufs && offset < size) { | |
1054 | #if 0 | |
e7b43f7e SH |
1055 | error_report("virtio-net truncated non-mergeable packet: " |
1056 | "i %zd mergeable %d offset %zd, size %zd, " | |
1057 | "guest hdr len %zd, host hdr len %zd", | |
1058 | i, n->mergeable_rx_bufs, | |
e35e23f6 | 1059 | offset, size, n->guest_hdr_len, n->host_hdr_len); |
279a4253 MT |
1060 | #endif |
1061 | return size; | |
1062 | } | |
fbe78f4f AL |
1063 | |
1064 | /* signal other side */ | |
0c87e93e | 1065 | virtqueue_fill(q->rx_vq, &elem, total, i++); |
fbe78f4f AL |
1066 | } |
1067 | ||
63c58728 MT |
1068 | if (mhdr_cnt) { |
1069 | stw_p(&mhdr.num_buffers, i); | |
1070 | iov_from_buf(mhdr_sg, mhdr_cnt, | |
1071 | 0, | |
1072 | &mhdr.num_buffers, sizeof mhdr.num_buffers); | |
44b15bc5 | 1073 | } |
fbe78f4f | 1074 | |
0c87e93e | 1075 | virtqueue_flush(q->rx_vq, i); |
17a0ca55 | 1076 | virtio_notify(vdev, q->rx_vq); |
4f1c942b MM |
1077 | |
1078 | return size; | |
fbe78f4f AL |
1079 | } |
1080 | ||
0c87e93e | 1081 | static int32_t virtio_net_flush_tx(VirtIONetQueue *q); |
6243375f | 1082 | |
4e68f7a0 | 1083 | static void virtio_net_tx_complete(NetClientState *nc, ssize_t len) |
6243375f | 1084 | { |
cc1f0f45 | 1085 | VirtIONet *n = qemu_get_nic_opaque(nc); |
fed699f9 | 1086 | VirtIONetQueue *q = virtio_net_get_subqueue(nc); |
17a0ca55 | 1087 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
6243375f | 1088 | |
0c87e93e | 1089 | virtqueue_push(q->tx_vq, &q->async_tx.elem, 0); |
17a0ca55 | 1090 | virtio_notify(vdev, q->tx_vq); |
6243375f | 1091 | |
0c87e93e | 1092 | q->async_tx.elem.out_num = q->async_tx.len = 0; |
6243375f | 1093 | |
0c87e93e JW |
1094 | virtio_queue_set_notification(q->tx_vq, 1); |
1095 | virtio_net_flush_tx(q); | |
6243375f MM |
1096 | } |
1097 | ||
fbe78f4f | 1098 | /* TX */ |
0c87e93e | 1099 | static int32_t virtio_net_flush_tx(VirtIONetQueue *q) |
fbe78f4f | 1100 | { |
0c87e93e | 1101 | VirtIONet *n = q->n; |
17a0ca55 | 1102 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
fbe78f4f | 1103 | VirtQueueElement elem; |
e3f30488 | 1104 | int32_t num_packets = 0; |
fed699f9 | 1105 | int queue_index = vq2q(virtio_get_queue_index(q->tx_vq)); |
17a0ca55 | 1106 | if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) { |
e3f30488 AW |
1107 | return num_packets; |
1108 | } | |
fbe78f4f | 1109 | |
17a0ca55 | 1110 | assert(vdev->vm_running); |
783e7706 | 1111 | |
0c87e93e JW |
1112 | if (q->async_tx.elem.out_num) { |
1113 | virtio_queue_set_notification(q->tx_vq, 0); | |
e3f30488 | 1114 | return num_packets; |
6243375f MM |
1115 | } |
1116 | ||
0c87e93e | 1117 | while (virtqueue_pop(q->tx_vq, &elem)) { |
14761f9c | 1118 | ssize_t ret, len; |
fbe78f4f AL |
1119 | unsigned int out_num = elem.out_num; |
1120 | struct iovec *out_sg = &elem.out_sg[0]; | |
14761f9c | 1121 | struct iovec sg[VIRTQUEUE_MAX_SIZE]; |
fbe78f4f | 1122 | |
7b80d08e | 1123 | if (out_num < 1) { |
e7b43f7e | 1124 | error_report("virtio-net header not in first element"); |
fbe78f4f AL |
1125 | exit(1); |
1126 | } | |
1127 | ||
14761f9c MT |
1128 | /* |
1129 | * If host wants to see the guest header as is, we can | |
1130 | * pass it on unchanged. Otherwise, copy just the parts | |
1131 | * that host is interested in. | |
1132 | */ | |
1133 | assert(n->host_hdr_len <= n->guest_hdr_len); | |
1134 | if (n->host_hdr_len != n->guest_hdr_len) { | |
1135 | unsigned sg_num = iov_copy(sg, ARRAY_SIZE(sg), | |
1136 | out_sg, out_num, | |
1137 | 0, n->host_hdr_len); | |
1138 | sg_num += iov_copy(sg + sg_num, ARRAY_SIZE(sg) - sg_num, | |
1139 | out_sg, out_num, | |
1140 | n->guest_hdr_len, -1); | |
1141 | out_num = sg_num; | |
1142 | out_sg = sg; | |
fbe78f4f AL |
1143 | } |
1144 | ||
7b80d08e | 1145 | len = n->guest_hdr_len; |
14761f9c | 1146 | |
fed699f9 JW |
1147 | ret = qemu_sendv_packet_async(qemu_get_subqueue(n->nic, queue_index), |
1148 | out_sg, out_num, virtio_net_tx_complete); | |
6243375f | 1149 | if (ret == 0) { |
0c87e93e JW |
1150 | virtio_queue_set_notification(q->tx_vq, 0); |
1151 | q->async_tx.elem = elem; | |
1152 | q->async_tx.len = len; | |
e3f30488 | 1153 | return -EBUSY; |
6243375f MM |
1154 | } |
1155 | ||
1156 | len += ret; | |
fbe78f4f | 1157 | |
0c87e93e | 1158 | virtqueue_push(q->tx_vq, &elem, 0); |
17a0ca55 | 1159 | virtio_notify(vdev, q->tx_vq); |
e3f30488 AW |
1160 | |
1161 | if (++num_packets >= n->tx_burst) { | |
1162 | break; | |
1163 | } | |
fbe78f4f | 1164 | } |
e3f30488 | 1165 | return num_packets; |
fbe78f4f AL |
1166 | } |
1167 | ||
a697a334 | 1168 | static void virtio_net_handle_tx_timer(VirtIODevice *vdev, VirtQueue *vq) |
fbe78f4f | 1169 | { |
17a0ca55 | 1170 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 | 1171 | VirtIONetQueue *q = &n->vqs[vq2q(virtio_get_queue_index(vq))]; |
fbe78f4f | 1172 | |
783e7706 | 1173 | /* This happens when device was stopped but VCPU wasn't. */ |
17a0ca55 | 1174 | if (!vdev->vm_running) { |
0c87e93e | 1175 | q->tx_waiting = 1; |
783e7706 MT |
1176 | return; |
1177 | } | |
1178 | ||
0c87e93e | 1179 | if (q->tx_waiting) { |
fbe78f4f | 1180 | virtio_queue_set_notification(vq, 1); |
bc72ad67 | 1181 | timer_del(q->tx_timer); |
0c87e93e JW |
1182 | q->tx_waiting = 0; |
1183 | virtio_net_flush_tx(q); | |
fbe78f4f | 1184 | } else { |
bc72ad67 AB |
1185 | timer_mod(q->tx_timer, |
1186 | qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + n->tx_timeout); | |
0c87e93e | 1187 | q->tx_waiting = 1; |
fbe78f4f AL |
1188 | virtio_queue_set_notification(vq, 0); |
1189 | } | |
1190 | } | |
1191 | ||
a697a334 AW |
1192 | static void virtio_net_handle_tx_bh(VirtIODevice *vdev, VirtQueue *vq) |
1193 | { | |
17a0ca55 | 1194 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 | 1195 | VirtIONetQueue *q = &n->vqs[vq2q(virtio_get_queue_index(vq))]; |
a697a334 | 1196 | |
0c87e93e | 1197 | if (unlikely(q->tx_waiting)) { |
a697a334 AW |
1198 | return; |
1199 | } | |
0c87e93e | 1200 | q->tx_waiting = 1; |
783e7706 | 1201 | /* This happens when device was stopped but VCPU wasn't. */ |
17a0ca55 | 1202 | if (!vdev->vm_running) { |
783e7706 MT |
1203 | return; |
1204 | } | |
a697a334 | 1205 | virtio_queue_set_notification(vq, 0); |
0c87e93e | 1206 | qemu_bh_schedule(q->tx_bh); |
a697a334 AW |
1207 | } |
1208 | ||
fbe78f4f AL |
1209 | static void virtio_net_tx_timer(void *opaque) |
1210 | { | |
0c87e93e JW |
1211 | VirtIONetQueue *q = opaque; |
1212 | VirtIONet *n = q->n; | |
17a0ca55 FK |
1213 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
1214 | assert(vdev->vm_running); | |
fbe78f4f | 1215 | |
0c87e93e | 1216 | q->tx_waiting = 0; |
fbe78f4f AL |
1217 | |
1218 | /* Just in case the driver is not ready on more */ | |
17a0ca55 | 1219 | if (!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK)) { |
fbe78f4f | 1220 | return; |
17a0ca55 | 1221 | } |
fbe78f4f | 1222 | |
0c87e93e JW |
1223 | virtio_queue_set_notification(q->tx_vq, 1); |
1224 | virtio_net_flush_tx(q); | |
fbe78f4f AL |
1225 | } |
1226 | ||
a697a334 AW |
1227 | static void virtio_net_tx_bh(void *opaque) |
1228 | { | |
0c87e93e JW |
1229 | VirtIONetQueue *q = opaque; |
1230 | VirtIONet *n = q->n; | |
17a0ca55 | 1231 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
a697a334 AW |
1232 | int32_t ret; |
1233 | ||
17a0ca55 | 1234 | assert(vdev->vm_running); |
783e7706 | 1235 | |
0c87e93e | 1236 | q->tx_waiting = 0; |
a697a334 AW |
1237 | |
1238 | /* Just in case the driver is not ready on more */ | |
17a0ca55 | 1239 | if (unlikely(!(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK))) { |
a697a334 | 1240 | return; |
17a0ca55 | 1241 | } |
a697a334 | 1242 | |
0c87e93e | 1243 | ret = virtio_net_flush_tx(q); |
a697a334 AW |
1244 | if (ret == -EBUSY) { |
1245 | return; /* Notification re-enable handled by tx_complete */ | |
1246 | } | |
1247 | ||
1248 | /* If we flush a full burst of packets, assume there are | |
1249 | * more coming and immediately reschedule */ | |
1250 | if (ret >= n->tx_burst) { | |
0c87e93e JW |
1251 | qemu_bh_schedule(q->tx_bh); |
1252 | q->tx_waiting = 1; | |
a697a334 AW |
1253 | return; |
1254 | } | |
1255 | ||
1256 | /* If less than a full burst, re-enable notification and flush | |
1257 | * anything that may have come in while we weren't looking. If | |
1258 | * we find something, assume the guest is still active and reschedule */ | |
0c87e93e JW |
1259 | virtio_queue_set_notification(q->tx_vq, 1); |
1260 | if (virtio_net_flush_tx(q) > 0) { | |
1261 | virtio_queue_set_notification(q->tx_vq, 0); | |
1262 | qemu_bh_schedule(q->tx_bh); | |
1263 | q->tx_waiting = 1; | |
a697a334 AW |
1264 | } |
1265 | } | |
1266 | ||
ec57db16 | 1267 | static void virtio_net_set_multiqueue(VirtIONet *n, int multiqueue) |
fed699f9 | 1268 | { |
17a0ca55 | 1269 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
fed699f9 JW |
1270 | int i, max = multiqueue ? n->max_queues : 1; |
1271 | ||
1272 | n->multiqueue = multiqueue; | |
1273 | ||
1274 | for (i = 2; i <= n->max_queues * 2 + 1; i++) { | |
1275 | virtio_del_queue(vdev, i); | |
1276 | } | |
1277 | ||
1278 | for (i = 1; i < max; i++) { | |
1279 | n->vqs[i].rx_vq = virtio_add_queue(vdev, 256, virtio_net_handle_rx); | |
1280 | if (n->vqs[i].tx_timer) { | |
1281 | n->vqs[i].tx_vq = | |
1282 | virtio_add_queue(vdev, 256, virtio_net_handle_tx_timer); | |
bc72ad67 | 1283 | n->vqs[i].tx_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, |
fed699f9 JW |
1284 | virtio_net_tx_timer, |
1285 | &n->vqs[i]); | |
1286 | } else { | |
1287 | n->vqs[i].tx_vq = | |
1288 | virtio_add_queue(vdev, 256, virtio_net_handle_tx_bh); | |
1289 | n->vqs[i].tx_bh = qemu_bh_new(virtio_net_tx_bh, &n->vqs[i]); | |
1290 | } | |
1291 | ||
1292 | n->vqs[i].tx_waiting = 0; | |
1293 | n->vqs[i].n = n; | |
1294 | } | |
1295 | ||
ec57db16 JW |
1296 | /* Note: Minux Guests (version 3.2.1) use ctrl vq but don't ack |
1297 | * VIRTIO_NET_F_CTRL_VQ. Create ctrl vq unconditionally to avoid | |
1298 | * breaking them. | |
1299 | */ | |
1300 | n->ctrl_vq = virtio_add_queue(vdev, 64, virtio_net_handle_ctrl); | |
fed699f9 JW |
1301 | |
1302 | virtio_net_set_queues(n); | |
1303 | } | |
1304 | ||
fbe78f4f AL |
1305 | static void virtio_net_save(QEMUFile *f, void *opaque) |
1306 | { | |
5f800801 | 1307 | int i; |
fbe78f4f | 1308 | VirtIONet *n = opaque; |
17a0ca55 | 1309 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
fbe78f4f | 1310 | |
afbaa7b4 MT |
1311 | /* At this point, backend must be stopped, otherwise |
1312 | * it might keep writing to memory. */ | |
1313 | assert(!n->vhost_started); | |
17a0ca55 | 1314 | virtio_save(vdev, f); |
fbe78f4f | 1315 | |
79674068 | 1316 | qemu_put_buffer(f, n->mac, ETH_ALEN); |
5f800801 | 1317 | qemu_put_be32(f, n->vqs[0].tx_waiting); |
e46cb38f | 1318 | qemu_put_be32(f, n->mergeable_rx_bufs); |
9d6271b8 | 1319 | qemu_put_be16(f, n->status); |
f10c592e AW |
1320 | qemu_put_byte(f, n->promisc); |
1321 | qemu_put_byte(f, n->allmulti); | |
b6503ed9 AL |
1322 | qemu_put_be32(f, n->mac_table.in_use); |
1323 | qemu_put_buffer(f, n->mac_table.macs, n->mac_table.in_use * ETH_ALEN); | |
f21c0ed9 | 1324 | qemu_put_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3); |
3a330134 | 1325 | qemu_put_be32(f, n->has_vnet_hdr); |
8fd2a2f1 AW |
1326 | qemu_put_byte(f, n->mac_table.multi_overflow); |
1327 | qemu_put_byte(f, n->mac_table.uni_overflow); | |
015cb166 AW |
1328 | qemu_put_byte(f, n->alluni); |
1329 | qemu_put_byte(f, n->nomulti); | |
1330 | qemu_put_byte(f, n->nouni); | |
1331 | qemu_put_byte(f, n->nobcast); | |
0ce0e8f4 | 1332 | qemu_put_byte(f, n->has_ufo); |
5f800801 JW |
1333 | if (n->max_queues > 1) { |
1334 | qemu_put_be16(f, n->max_queues); | |
1335 | qemu_put_be16(f, n->curr_queues); | |
1336 | for (i = 1; i < n->curr_queues; i++) { | |
1337 | qemu_put_be32(f, n->vqs[i].tx_waiting); | |
1338 | } | |
1339 | } | |
644c9858 DF |
1340 | |
1341 | if ((1 << VIRTIO_NET_F_CTRL_GUEST_OFFLOADS) & vdev->guest_features) { | |
1342 | qemu_put_be64(f, n->curr_guest_offloads); | |
1343 | } | |
fbe78f4f AL |
1344 | } |
1345 | ||
1346 | static int virtio_net_load(QEMUFile *f, void *opaque, int version_id) | |
1347 | { | |
1348 | VirtIONet *n = opaque; | |
17a0ca55 | 1349 | VirtIODevice *vdev = VIRTIO_DEVICE(n); |
5f800801 | 1350 | int ret, i, link_down; |
fbe78f4f | 1351 | |
9d6271b8 | 1352 | if (version_id < 2 || version_id > VIRTIO_NET_VM_VERSION) |
fbe78f4f AL |
1353 | return -EINVAL; |
1354 | ||
17a0ca55 | 1355 | ret = virtio_load(vdev, f); |
2a633c46 OW |
1356 | if (ret) { |
1357 | return ret; | |
1358 | } | |
fbe78f4f | 1359 | |
79674068 | 1360 | qemu_get_buffer(f, n->mac, ETH_ALEN); |
5f800801 | 1361 | n->vqs[0].tx_waiting = qemu_get_be32(f); |
ff3a8066 MT |
1362 | |
1363 | virtio_net_set_mrg_rx_bufs(n, qemu_get_be32(f)); | |
fbe78f4f | 1364 | |
9d6271b8 AL |
1365 | if (version_id >= 3) |
1366 | n->status = qemu_get_be16(f); | |
1367 | ||
002437cd | 1368 | if (version_id >= 4) { |
f10c592e AW |
1369 | if (version_id < 8) { |
1370 | n->promisc = qemu_get_be32(f); | |
1371 | n->allmulti = qemu_get_be32(f); | |
1372 | } else { | |
1373 | n->promisc = qemu_get_byte(f); | |
1374 | n->allmulti = qemu_get_byte(f); | |
1375 | } | |
002437cd AL |
1376 | } |
1377 | ||
b6503ed9 AL |
1378 | if (version_id >= 5) { |
1379 | n->mac_table.in_use = qemu_get_be32(f); | |
1380 | /* MAC_TABLE_ENTRIES may be different from the saved image */ | |
1381 | if (n->mac_table.in_use <= MAC_TABLE_ENTRIES) { | |
1382 | qemu_get_buffer(f, n->mac_table.macs, | |
1383 | n->mac_table.in_use * ETH_ALEN); | |
98f93ddd MT |
1384 | } else { |
1385 | int64_t i; | |
1386 | ||
1387 | /* Overflow detected - can happen if source has a larger MAC table. | |
1388 | * We simply set overflow flag so there's no need to maintain the | |
1389 | * table of addresses, discard them all. | |
1390 | * Note: 64 bit math to avoid integer overflow. | |
1391 | */ | |
1392 | for (i = 0; i < (int64_t)n->mac_table.in_use * ETH_ALEN; ++i) { | |
1393 | qemu_get_byte(f); | |
1394 | } | |
8fd2a2f1 | 1395 | n->mac_table.multi_overflow = n->mac_table.uni_overflow = 1; |
b6503ed9 AL |
1396 | n->mac_table.in_use = 0; |
1397 | } | |
1398 | } | |
1399 | ||
f21c0ed9 AL |
1400 | if (version_id >= 6) |
1401 | qemu_get_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3); | |
1402 | ||
3a330134 MM |
1403 | if (version_id >= 7) { |
1404 | if (qemu_get_be32(f) && !peer_has_vnet_hdr(n)) { | |
1ecda02b | 1405 | error_report("virtio-net: saved image requires vnet_hdr=on"); |
3a330134 MM |
1406 | return -1; |
1407 | } | |
6c042c16 AW |
1408 | } |
1409 | ||
8fd2a2f1 AW |
1410 | if (version_id >= 9) { |
1411 | n->mac_table.multi_overflow = qemu_get_byte(f); | |
1412 | n->mac_table.uni_overflow = qemu_get_byte(f); | |
1413 | } | |
1414 | ||
015cb166 AW |
1415 | if (version_id >= 10) { |
1416 | n->alluni = qemu_get_byte(f); | |
1417 | n->nomulti = qemu_get_byte(f); | |
1418 | n->nouni = qemu_get_byte(f); | |
1419 | n->nobcast = qemu_get_byte(f); | |
1420 | } | |
1421 | ||
0ce0e8f4 MM |
1422 | if (version_id >= 11) { |
1423 | if (qemu_get_byte(f) && !peer_has_ufo(n)) { | |
1ecda02b | 1424 | error_report("virtio-net: saved image requires TUN_F_UFO support"); |
0ce0e8f4 MM |
1425 | return -1; |
1426 | } | |
1427 | } | |
1428 | ||
5f800801 JW |
1429 | if (n->max_queues > 1) { |
1430 | if (n->max_queues != qemu_get_be16(f)) { | |
1431 | error_report("virtio-net: different max_queues "); | |
1432 | return -1; | |
1433 | } | |
1434 | ||
1435 | n->curr_queues = qemu_get_be16(f); | |
eea750a5 MT |
1436 | if (n->curr_queues > n->max_queues) { |
1437 | error_report("virtio-net: curr_queues %x > max_queues %x", | |
1438 | n->curr_queues, n->max_queues); | |
1439 | return -1; | |
1440 | } | |
5f800801 JW |
1441 | for (i = 1; i < n->curr_queues; i++) { |
1442 | n->vqs[i].tx_waiting = qemu_get_be32(f); | |
1443 | } | |
1444 | } | |
1445 | ||
644c9858 DF |
1446 | if ((1 << VIRTIO_NET_F_CTRL_GUEST_OFFLOADS) & vdev->guest_features) { |
1447 | n->curr_guest_offloads = qemu_get_be64(f); | |
1448 | } else { | |
1449 | n->curr_guest_offloads = virtio_net_supported_guest_offloads(n); | |
1450 | } | |
1451 | ||
1452 | if (peer_has_vnet_hdr(n)) { | |
1453 | virtio_net_apply_guest_offloads(n); | |
1454 | } | |
1455 | ||
5f800801 JW |
1456 | virtio_net_set_queues(n); |
1457 | ||
2d9aba39 AW |
1458 | /* Find the first multicast entry in the saved MAC filter */ |
1459 | for (i = 0; i < n->mac_table.in_use; i++) { | |
1460 | if (n->mac_table.macs[i * ETH_ALEN] & 1) { | |
1461 | break; | |
1462 | } | |
1463 | } | |
1464 | n->mac_table.first_multi = i; | |
98991481 AK |
1465 | |
1466 | /* nc.link_down can't be migrated, so infer link_down according | |
1467 | * to link status bit in n->status */ | |
5f800801 JW |
1468 | link_down = (n->status & VIRTIO_NET_S_LINK_UP) == 0; |
1469 | for (i = 0; i < n->max_queues; i++) { | |
1470 | qemu_get_subqueue(n->nic, i)->link_down = link_down; | |
1471 | } | |
98991481 | 1472 | |
f57fcf70 JW |
1473 | if (vdev->guest_features & (0x1 << VIRTIO_NET_F_GUEST_ANNOUNCE) && |
1474 | vdev->guest_features & (0x1 << VIRTIO_NET_F_CTRL_VQ)) { | |
1475 | n->announce_counter = SELF_ANNOUNCE_ROUNDS; | |
1476 | timer_mod(n->announce_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL)); | |
1477 | } | |
1478 | ||
fbe78f4f AL |
1479 | return 0; |
1480 | } | |
1481 | ||
4e68f7a0 | 1482 | static void virtio_net_cleanup(NetClientState *nc) |
b946a153 | 1483 | { |
cc1f0f45 | 1484 | VirtIONet *n = qemu_get_nic_opaque(nc); |
b946a153 | 1485 | |
eb6b6c12 | 1486 | n->nic = NULL; |
b946a153 AL |
1487 | } |
1488 | ||
eb6b6c12 | 1489 | static NetClientInfo net_virtio_info = { |
2be64a68 | 1490 | .type = NET_CLIENT_OPTIONS_KIND_NIC, |
eb6b6c12 MM |
1491 | .size = sizeof(NICState), |
1492 | .can_receive = virtio_net_can_receive, | |
1493 | .receive = virtio_net_receive, | |
cb77e358 | 1494 | .cleanup = virtio_net_cleanup, |
eb6b6c12 | 1495 | .link_status_changed = virtio_net_set_link_status, |
b1be4280 | 1496 | .query_rx_filter = virtio_net_query_rxfilter, |
eb6b6c12 MM |
1497 | }; |
1498 | ||
f56a1247 MT |
1499 | static bool virtio_net_guest_notifier_pending(VirtIODevice *vdev, int idx) |
1500 | { | |
17a0ca55 | 1501 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 | 1502 | NetClientState *nc = qemu_get_subqueue(n->nic, vq2q(idx)); |
f56a1247 | 1503 | assert(n->vhost_started); |
ed8b4afe | 1504 | return vhost_net_virtqueue_pending(get_vhost_net(nc->peer), idx); |
f56a1247 MT |
1505 | } |
1506 | ||
1507 | static void virtio_net_guest_notifier_mask(VirtIODevice *vdev, int idx, | |
1508 | bool mask) | |
1509 | { | |
17a0ca55 | 1510 | VirtIONet *n = VIRTIO_NET(vdev); |
fed699f9 | 1511 | NetClientState *nc = qemu_get_subqueue(n->nic, vq2q(idx)); |
f56a1247 | 1512 | assert(n->vhost_started); |
ed8b4afe | 1513 | vhost_net_virtqueue_mask(get_vhost_net(nc->peer), |
f56a1247 MT |
1514 | vdev, idx, mask); |
1515 | } | |
1516 | ||
17ec5a86 | 1517 | void virtio_net_set_config_size(VirtIONet *n, uint32_t host_features) |
fbe78f4f | 1518 | { |
14f9b664 | 1519 | int i, config_size = 0; |
e9016ee2 | 1520 | host_features |= (1 << VIRTIO_NET_F_MAC); |
14f9b664 JL |
1521 | for (i = 0; feature_sizes[i].flags != 0; i++) { |
1522 | if (host_features & feature_sizes[i].flags) { | |
1523 | config_size = MAX(feature_sizes[i].end, config_size); | |
1524 | } | |
1525 | } | |
17ec5a86 FK |
1526 | n->config_size = config_size; |
1527 | } | |
1528 | ||
8a253ec2 FK |
1529 | void virtio_net_set_netclient_name(VirtIONet *n, const char *name, |
1530 | const char *type) | |
1531 | { | |
1532 | /* | |
1533 | * The name can be NULL, the netclient name will be type.x. | |
1534 | */ | |
1535 | assert(type != NULL); | |
1536 | ||
9e288406 | 1537 | g_free(n->netclient_name); |
9e288406 | 1538 | g_free(n->netclient_type); |
80e0090a | 1539 | n->netclient_name = g_strdup(name); |
8a253ec2 FK |
1540 | n->netclient_type = g_strdup(type); |
1541 | } | |
1542 | ||
e6f746b3 | 1543 | static void virtio_net_device_realize(DeviceState *dev, Error **errp) |
17ec5a86 | 1544 | { |
e6f746b3 | 1545 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); |
284a32f0 | 1546 | VirtIONet *n = VIRTIO_NET(dev); |
b1be4280 | 1547 | NetClientState *nc; |
284a32f0 | 1548 | int i; |
1773d9ee | 1549 | |
284a32f0 | 1550 | virtio_init(vdev, "virtio-net", VIRTIO_ID_NET, n->config_size); |
fbe78f4f | 1551 | |
1773d9ee | 1552 | n->max_queues = MAX(n->nic_conf.queues, 1); |
f6b26cf2 | 1553 | n->vqs = g_malloc0(sizeof(VirtIONetQueue) * n->max_queues); |
17a0ca55 | 1554 | n->vqs[0].rx_vq = virtio_add_queue(vdev, 256, virtio_net_handle_rx); |
fed699f9 JW |
1555 | n->curr_queues = 1; |
1556 | n->vqs[0].n = n; | |
1773d9ee | 1557 | n->tx_timeout = n->net_conf.txtimer; |
a697a334 | 1558 | |
1773d9ee FK |
1559 | if (n->net_conf.tx && strcmp(n->net_conf.tx, "timer") |
1560 | && strcmp(n->net_conf.tx, "bh")) { | |
e7b43f7e SH |
1561 | error_report("virtio-net: " |
1562 | "Unknown option tx=%s, valid options: \"timer\" \"bh\"", | |
1773d9ee | 1563 | n->net_conf.tx); |
e7b43f7e | 1564 | error_report("Defaulting to \"bh\""); |
a697a334 AW |
1565 | } |
1566 | ||
1773d9ee | 1567 | if (n->net_conf.tx && !strcmp(n->net_conf.tx, "timer")) { |
17a0ca55 | 1568 | n->vqs[0].tx_vq = virtio_add_queue(vdev, 256, |
fed699f9 | 1569 | virtio_net_handle_tx_timer); |
bc72ad67 | 1570 | n->vqs[0].tx_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, virtio_net_tx_timer, |
fed699f9 | 1571 | &n->vqs[0]); |
a697a334 | 1572 | } else { |
17a0ca55 | 1573 | n->vqs[0].tx_vq = virtio_add_queue(vdev, 256, |
fed699f9 JW |
1574 | virtio_net_handle_tx_bh); |
1575 | n->vqs[0].tx_bh = qemu_bh_new(virtio_net_tx_bh, &n->vqs[0]); | |
a697a334 | 1576 | } |
17a0ca55 | 1577 | n->ctrl_vq = virtio_add_queue(vdev, 64, virtio_net_handle_ctrl); |
1773d9ee FK |
1578 | qemu_macaddr_default_if_unset(&n->nic_conf.macaddr); |
1579 | memcpy(&n->mac[0], &n->nic_conf.macaddr, sizeof(n->mac)); | |
554c97dd | 1580 | n->status = VIRTIO_NET_S_LINK_UP; |
f57fcf70 JW |
1581 | n->announce_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL, |
1582 | virtio_net_announce_timer, n); | |
fbe78f4f | 1583 | |
8a253ec2 FK |
1584 | if (n->netclient_type) { |
1585 | /* | |
1586 | * Happen when virtio_net_set_netclient_name has been called. | |
1587 | */ | |
1588 | n->nic = qemu_new_nic(&net_virtio_info, &n->nic_conf, | |
1589 | n->netclient_type, n->netclient_name, n); | |
1590 | } else { | |
1591 | n->nic = qemu_new_nic(&net_virtio_info, &n->nic_conf, | |
284a32f0 | 1592 | object_get_typename(OBJECT(dev)), dev->id, n); |
8a253ec2 FK |
1593 | } |
1594 | ||
6e371ab8 MT |
1595 | peer_test_vnet_hdr(n); |
1596 | if (peer_has_vnet_hdr(n)) { | |
fed699f9 | 1597 | for (i = 0; i < n->max_queues; i++) { |
d6085e3a | 1598 | qemu_using_vnet_hdr(qemu_get_subqueue(n->nic, i)->peer, true); |
fed699f9 | 1599 | } |
6e371ab8 MT |
1600 | n->host_hdr_len = sizeof(struct virtio_net_hdr); |
1601 | } else { | |
1602 | n->host_hdr_len = 0; | |
1603 | } | |
eb6b6c12 | 1604 | |
1773d9ee | 1605 | qemu_format_nic_info_str(qemu_get_queue(n->nic), n->nic_conf.macaddr.a); |
96d5e201 | 1606 | |
fed699f9 | 1607 | n->vqs[0].tx_waiting = 0; |
1773d9ee | 1608 | n->tx_burst = n->net_conf.txburst; |
ff3a8066 | 1609 | virtio_net_set_mrg_rx_bufs(n, 0); |
002437cd | 1610 | n->promisc = 1; /* for compatibility */ |
fbe78f4f | 1611 | |
7267c094 | 1612 | n->mac_table.macs = g_malloc0(MAC_TABLE_ENTRIES * ETH_ALEN); |
b6503ed9 | 1613 | |
7267c094 | 1614 | n->vlans = g_malloc0(MAX_VLAN >> 3); |
f21c0ed9 | 1615 | |
b1be4280 AK |
1616 | nc = qemu_get_queue(n->nic); |
1617 | nc->rxfilter_notify_enabled = 1; | |
1618 | ||
284a32f0 AF |
1619 | n->qdev = dev; |
1620 | register_savevm(dev, "virtio-net", -1, VIRTIO_NET_VM_VERSION, | |
fbe78f4f | 1621 | virtio_net_save, virtio_net_load, n); |
cf21e106 | 1622 | |
284a32f0 | 1623 | add_boot_device_path(n->nic_conf.bootindex, dev, "/ethernet-phy@0"); |
17ec5a86 FK |
1624 | } |
1625 | ||
306ec6c3 | 1626 | static void virtio_net_device_unrealize(DeviceState *dev, Error **errp) |
17ec5a86 | 1627 | { |
306ec6c3 AF |
1628 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); |
1629 | VirtIONet *n = VIRTIO_NET(dev); | |
17ec5a86 FK |
1630 | int i; |
1631 | ||
1632 | /* This will stop vhost backend if appropriate. */ | |
1633 | virtio_net_set_status(vdev, 0); | |
1634 | ||
306ec6c3 | 1635 | unregister_savevm(dev, "virtio-net", n); |
17ec5a86 | 1636 | |
9e288406 MA |
1637 | g_free(n->netclient_name); |
1638 | n->netclient_name = NULL; | |
1639 | g_free(n->netclient_type); | |
1640 | n->netclient_type = NULL; | |
8a253ec2 | 1641 | |
17ec5a86 FK |
1642 | g_free(n->mac_table.macs); |
1643 | g_free(n->vlans); | |
1644 | ||
1645 | for (i = 0; i < n->max_queues; i++) { | |
1646 | VirtIONetQueue *q = &n->vqs[i]; | |
1647 | NetClientState *nc = qemu_get_subqueue(n->nic, i); | |
1648 | ||
1649 | qemu_purge_queued_packets(nc); | |
1650 | ||
1651 | if (q->tx_timer) { | |
bc72ad67 AB |
1652 | timer_del(q->tx_timer); |
1653 | timer_free(q->tx_timer); | |
fe2dafa0 | 1654 | } else if (q->tx_bh) { |
17ec5a86 FK |
1655 | qemu_bh_delete(q->tx_bh); |
1656 | } | |
1657 | } | |
1658 | ||
f57fcf70 JW |
1659 | timer_del(n->announce_timer); |
1660 | timer_free(n->announce_timer); | |
17ec5a86 FK |
1661 | g_free(n->vqs); |
1662 | qemu_del_nic(n->nic); | |
6a1a8cc7 | 1663 | virtio_cleanup(vdev); |
17ec5a86 FK |
1664 | } |
1665 | ||
1666 | static void virtio_net_instance_init(Object *obj) | |
1667 | { | |
1668 | VirtIONet *n = VIRTIO_NET(obj); | |
1669 | ||
1670 | /* | |
1671 | * The default config_size is sizeof(struct virtio_net_config). | |
1672 | * Can be overriden with virtio_net_set_config_size. | |
1673 | */ | |
1674 | n->config_size = sizeof(struct virtio_net_config); | |
1675 | } | |
1676 | ||
1677 | static Property virtio_net_properties[] = { | |
1678 | DEFINE_NIC_PROPERTIES(VirtIONet, nic_conf), | |
1679 | DEFINE_PROP_UINT32("x-txtimer", VirtIONet, net_conf.txtimer, | |
1680 | TX_TIMER_INTERVAL), | |
1681 | DEFINE_PROP_INT32("x-txburst", VirtIONet, net_conf.txburst, TX_BURST), | |
1682 | DEFINE_PROP_STRING("tx", VirtIONet, net_conf.tx), | |
1683 | DEFINE_PROP_END_OF_LIST(), | |
1684 | }; | |
1685 | ||
1686 | static void virtio_net_class_init(ObjectClass *klass, void *data) | |
1687 | { | |
1688 | DeviceClass *dc = DEVICE_CLASS(klass); | |
1689 | VirtioDeviceClass *vdc = VIRTIO_DEVICE_CLASS(klass); | |
e6f746b3 | 1690 | |
17ec5a86 | 1691 | dc->props = virtio_net_properties; |
125ee0ed | 1692 | set_bit(DEVICE_CATEGORY_NETWORK, dc->categories); |
e6f746b3 | 1693 | vdc->realize = virtio_net_device_realize; |
306ec6c3 | 1694 | vdc->unrealize = virtio_net_device_unrealize; |
17ec5a86 FK |
1695 | vdc->get_config = virtio_net_get_config; |
1696 | vdc->set_config = virtio_net_set_config; | |
1697 | vdc->get_features = virtio_net_get_features; | |
1698 | vdc->set_features = virtio_net_set_features; | |
1699 | vdc->bad_features = virtio_net_bad_features; | |
1700 | vdc->reset = virtio_net_reset; | |
1701 | vdc->set_status = virtio_net_set_status; | |
1702 | vdc->guest_notifier_mask = virtio_net_guest_notifier_mask; | |
1703 | vdc->guest_notifier_pending = virtio_net_guest_notifier_pending; | |
1704 | } | |
1705 | ||
1706 | static const TypeInfo virtio_net_info = { | |
1707 | .name = TYPE_VIRTIO_NET, | |
1708 | .parent = TYPE_VIRTIO_DEVICE, | |
1709 | .instance_size = sizeof(VirtIONet), | |
1710 | .instance_init = virtio_net_instance_init, | |
1711 | .class_init = virtio_net_class_init, | |
1712 | }; | |
1713 | ||
1714 | static void virtio_register_types(void) | |
1715 | { | |
1716 | type_register_static(&virtio_net_info); | |
1717 | } | |
1718 | ||
1719 | type_init(virtio_register_types) |