]>
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" |
fbe78f4f | 15 | #include "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" | |
fbe78f4f | 21 | #include "virtio-net.h" |
9bc6304c | 22 | #include "vhost_net.h" |
fbe78f4f | 23 | |
0ce0e8f4 | 24 | #define VIRTIO_NET_VM_VERSION 11 |
b6503ed9 | 25 | |
4ffb17f5 | 26 | #define MAC_TABLE_ENTRIES 64 |
f21c0ed9 | 27 | #define MAX_VLAN (1 << 12) /* Per 802.1Q definition */ |
9d6271b8 | 28 | |
fbe78f4f AL |
29 | typedef struct VirtIONet |
30 | { | |
31 | VirtIODevice vdev; | |
79674068 | 32 | uint8_t mac[ETH_ALEN]; |
554c97dd | 33 | uint16_t status; |
fbe78f4f AL |
34 | VirtQueue *rx_vq; |
35 | VirtQueue *tx_vq; | |
3d11d36c | 36 | VirtQueue *ctrl_vq; |
eb6b6c12 | 37 | NICState *nic; |
fbe78f4f | 38 | QEMUTimer *tx_timer; |
a697a334 | 39 | QEMUBH *tx_bh; |
f0c07c7c | 40 | uint32_t tx_timeout; |
e3f30488 | 41 | int32_t tx_burst; |
4b4b8d36 | 42 | int tx_waiting; |
3a330134 | 43 | uint32_t has_vnet_hdr; |
e35e23f6 MT |
44 | size_t host_hdr_len; |
45 | size_t guest_hdr_len; | |
0ce0e8f4 | 46 | uint8_t has_ufo; |
6243375f MM |
47 | struct { |
48 | VirtQueueElement elem; | |
49 | ssize_t len; | |
50 | } async_tx; | |
fbe78f4f | 51 | int mergeable_rx_bufs; |
f10c592e AW |
52 | uint8_t promisc; |
53 | uint8_t allmulti; | |
015cb166 AW |
54 | uint8_t alluni; |
55 | uint8_t nomulti; | |
56 | uint8_t nouni; | |
57 | uint8_t nobcast; | |
9bc6304c | 58 | uint8_t vhost_started; |
b6503ed9 AL |
59 | struct { |
60 | int in_use; | |
2d9aba39 | 61 | int first_multi; |
8fd2a2f1 AW |
62 | uint8_t multi_overflow; |
63 | uint8_t uni_overflow; | |
b6503ed9 AL |
64 | uint8_t *macs; |
65 | } mac_table; | |
f21c0ed9 | 66 | uint32_t *vlans; |
01657c86 | 67 | DeviceState *qdev; |
fbe78f4f AL |
68 | } VirtIONet; |
69 | ||
70 | /* TODO | |
71 | * - we could suppress RX interrupt if we were so inclined. | |
72 | */ | |
73 | ||
74 | static VirtIONet *to_virtio_net(VirtIODevice *vdev) | |
75 | { | |
76 | return (VirtIONet *)vdev; | |
77 | } | |
78 | ||
0f03eca6 | 79 | static void virtio_net_get_config(VirtIODevice *vdev, uint8_t *config) |
fbe78f4f AL |
80 | { |
81 | VirtIONet *n = to_virtio_net(vdev); | |
82 | struct virtio_net_config netcfg; | |
83 | ||
b46d97f2 | 84 | stw_p(&netcfg.status, n->status); |
79674068 | 85 | memcpy(netcfg.mac, n->mac, ETH_ALEN); |
fbe78f4f AL |
86 | memcpy(config, &netcfg, sizeof(netcfg)); |
87 | } | |
88 | ||
0f03eca6 AL |
89 | static void virtio_net_set_config(VirtIODevice *vdev, const uint8_t *config) |
90 | { | |
91 | VirtIONet *n = to_virtio_net(vdev); | |
92 | struct virtio_net_config netcfg; | |
93 | ||
94 | memcpy(&netcfg, config, sizeof(netcfg)); | |
95 | ||
c1943a3f AK |
96 | if (!(n->vdev.guest_features >> VIRTIO_NET_F_CTRL_MAC_ADDR & 1) && |
97 | memcmp(netcfg.mac, n->mac, ETH_ALEN)) { | |
79674068 | 98 | memcpy(n->mac, netcfg.mac, ETH_ALEN); |
eb6b6c12 | 99 | qemu_format_nic_info_str(&n->nic->nc, n->mac); |
0f03eca6 AL |
100 | } |
101 | } | |
102 | ||
783e7706 MT |
103 | static bool virtio_net_started(VirtIONet *n, uint8_t status) |
104 | { | |
105 | return (status & VIRTIO_CONFIG_S_DRIVER_OK) && | |
85cf2a8d | 106 | (n->status & VIRTIO_NET_S_LINK_UP) && n->vdev.vm_running; |
783e7706 MT |
107 | } |
108 | ||
109 | static void virtio_net_vhost_status(VirtIONet *n, uint8_t status) | |
afbaa7b4 | 110 | { |
afbaa7b4 MT |
111 | if (!n->nic->nc.peer) { |
112 | return; | |
113 | } | |
2be64a68 | 114 | if (n->nic->nc.peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) { |
afbaa7b4 MT |
115 | return; |
116 | } | |
117 | ||
118 | if (!tap_get_vhost_net(n->nic->nc.peer)) { | |
119 | return; | |
120 | } | |
32993698 MT |
121 | if (!!n->vhost_started == virtio_net_started(n, status) && |
122 | !n->nic->nc.peer->link_down) { | |
afbaa7b4 MT |
123 | return; |
124 | } | |
125 | if (!n->vhost_started) { | |
5430a28f MT |
126 | int r; |
127 | if (!vhost_net_query(tap_get_vhost_net(n->nic->nc.peer), &n->vdev)) { | |
128 | return; | |
129 | } | |
1830b80f | 130 | n->vhost_started = 1; |
5430a28f | 131 | r = vhost_net_start(tap_get_vhost_net(n->nic->nc.peer), &n->vdev); |
afbaa7b4 | 132 | if (r < 0) { |
e7b43f7e SH |
133 | error_report("unable to start vhost net: %d: " |
134 | "falling back on userspace virtio", -r); | |
1830b80f | 135 | n->vhost_started = 0; |
afbaa7b4 MT |
136 | } |
137 | } else { | |
138 | vhost_net_stop(tap_get_vhost_net(n->nic->nc.peer), &n->vdev); | |
139 | n->vhost_started = 0; | |
140 | } | |
141 | } | |
142 | ||
783e7706 MT |
143 | static void virtio_net_set_status(struct VirtIODevice *vdev, uint8_t status) |
144 | { | |
145 | VirtIONet *n = to_virtio_net(vdev); | |
146 | ||
147 | virtio_net_vhost_status(n, status); | |
148 | ||
149 | if (!n->tx_waiting) { | |
150 | return; | |
151 | } | |
152 | ||
153 | if (virtio_net_started(n, status) && !n->vhost_started) { | |
154 | if (n->tx_timer) { | |
155 | qemu_mod_timer(n->tx_timer, | |
74475455 | 156 | qemu_get_clock_ns(vm_clock) + n->tx_timeout); |
783e7706 MT |
157 | } else { |
158 | qemu_bh_schedule(n->tx_bh); | |
159 | } | |
160 | } else { | |
161 | if (n->tx_timer) { | |
162 | qemu_del_timer(n->tx_timer); | |
163 | } else { | |
164 | qemu_bh_cancel(n->tx_bh); | |
165 | } | |
166 | } | |
167 | } | |
168 | ||
4e68f7a0 | 169 | static void virtio_net_set_link_status(NetClientState *nc) |
554c97dd | 170 | { |
eb6b6c12 | 171 | VirtIONet *n = DO_UPCAST(NICState, nc, nc)->opaque; |
554c97dd AL |
172 | uint16_t old_status = n->status; |
173 | ||
eb6b6c12 | 174 | if (nc->link_down) |
554c97dd AL |
175 | n->status &= ~VIRTIO_NET_S_LINK_UP; |
176 | else | |
177 | n->status |= VIRTIO_NET_S_LINK_UP; | |
178 | ||
179 | if (n->status != old_status) | |
180 | virtio_notify_config(&n->vdev); | |
afbaa7b4 MT |
181 | |
182 | virtio_net_set_status(&n->vdev, n->vdev.status); | |
554c97dd AL |
183 | } |
184 | ||
002437cd AL |
185 | static void virtio_net_reset(VirtIODevice *vdev) |
186 | { | |
187 | VirtIONet *n = to_virtio_net(vdev); | |
188 | ||
189 | /* Reset back to compatibility mode */ | |
190 | n->promisc = 1; | |
191 | n->allmulti = 0; | |
015cb166 AW |
192 | n->alluni = 0; |
193 | n->nomulti = 0; | |
194 | n->nouni = 0; | |
195 | n->nobcast = 0; | |
b6503ed9 | 196 | |
f21c0ed9 | 197 | /* Flush any MAC and VLAN filter table state */ |
b6503ed9 | 198 | n->mac_table.in_use = 0; |
2d9aba39 | 199 | n->mac_table.first_multi = 0; |
8fd2a2f1 AW |
200 | n->mac_table.multi_overflow = 0; |
201 | n->mac_table.uni_overflow = 0; | |
b6503ed9 | 202 | memset(n->mac_table.macs, 0, MAC_TABLE_ENTRIES * ETH_ALEN); |
41dc8a67 | 203 | memcpy(&n->mac[0], &n->nic->conf->macaddr, sizeof(n->mac)); |
f21c0ed9 | 204 | memset(n->vlans, 0, MAX_VLAN >> 3); |
002437cd AL |
205 | } |
206 | ||
6e371ab8 | 207 | static void peer_test_vnet_hdr(VirtIONet *n) |
3a330134 | 208 | { |
eb6b6c12 | 209 | if (!n->nic->nc.peer) |
6e371ab8 | 210 | return; |
3a330134 | 211 | |
2be64a68 | 212 | if (n->nic->nc.peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) |
6e371ab8 | 213 | return; |
3a330134 | 214 | |
eb6b6c12 | 215 | n->has_vnet_hdr = tap_has_vnet_hdr(n->nic->nc.peer); |
6e371ab8 | 216 | } |
3a330134 | 217 | |
6e371ab8 MT |
218 | static int peer_has_vnet_hdr(VirtIONet *n) |
219 | { | |
3a330134 MM |
220 | return n->has_vnet_hdr; |
221 | } | |
222 | ||
0ce0e8f4 MM |
223 | static int peer_has_ufo(VirtIONet *n) |
224 | { | |
225 | if (!peer_has_vnet_hdr(n)) | |
226 | return 0; | |
227 | ||
eb6b6c12 | 228 | n->has_ufo = tap_has_ufo(n->nic->nc.peer); |
0ce0e8f4 MM |
229 | |
230 | return n->has_ufo; | |
231 | } | |
232 | ||
ff3a8066 MT |
233 | static void virtio_net_set_mrg_rx_bufs(VirtIONet *n, int mergeable_rx_bufs) |
234 | { | |
235 | n->mergeable_rx_bufs = mergeable_rx_bufs; | |
236 | ||
237 | n->guest_hdr_len = n->mergeable_rx_bufs ? | |
238 | sizeof(struct virtio_net_hdr_mrg_rxbuf) : sizeof(struct virtio_net_hdr); | |
239 | ||
240 | if (peer_has_vnet_hdr(n) && | |
241 | tap_has_vnet_hdr_len(n->nic->nc.peer, n->guest_hdr_len)) { | |
242 | tap_set_vnet_hdr_len(n->nic->nc.peer, n->guest_hdr_len); | |
243 | n->host_hdr_len = n->guest_hdr_len; | |
244 | } | |
245 | } | |
246 | ||
8172539d | 247 | static uint32_t virtio_net_get_features(VirtIODevice *vdev, uint32_t features) |
fbe78f4f | 248 | { |
3a330134 | 249 | VirtIONet *n = to_virtio_net(vdev); |
fbe78f4f | 250 | |
c9f79a3f MT |
251 | features |= (1 << VIRTIO_NET_F_MAC); |
252 | ||
6e371ab8 | 253 | if (!peer_has_vnet_hdr(n)) { |
8172539d MT |
254 | features &= ~(0x1 << VIRTIO_NET_F_CSUM); |
255 | features &= ~(0x1 << VIRTIO_NET_F_HOST_TSO4); | |
256 | features &= ~(0x1 << VIRTIO_NET_F_HOST_TSO6); | |
257 | features &= ~(0x1 << VIRTIO_NET_F_HOST_ECN); | |
258 | ||
259 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_CSUM); | |
260 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_TSO4); | |
261 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_TSO6); | |
262 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_ECN); | |
263 | } | |
3a330134 | 264 | |
8172539d MT |
265 | if (!peer_has_vnet_hdr(n) || !peer_has_ufo(n)) { |
266 | features &= ~(0x1 << VIRTIO_NET_F_GUEST_UFO); | |
267 | features &= ~(0x1 << VIRTIO_NET_F_HOST_UFO); | |
3a330134 MM |
268 | } |
269 | ||
9bc6304c | 270 | if (!n->nic->nc.peer || |
2be64a68 | 271 | n->nic->nc.peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) { |
9bc6304c MT |
272 | return features; |
273 | } | |
274 | if (!tap_get_vhost_net(n->nic->nc.peer)) { | |
275 | return features; | |
276 | } | |
277 | return vhost_net_get_features(tap_get_vhost_net(n->nic->nc.peer), features); | |
fbe78f4f AL |
278 | } |
279 | ||
8eca6b1b AL |
280 | static uint32_t virtio_net_bad_features(VirtIODevice *vdev) |
281 | { | |
282 | uint32_t features = 0; | |
283 | ||
284 | /* Linux kernel 2.6.25. It understood MAC (as everyone must), | |
285 | * but also these: */ | |
286 | features |= (1 << VIRTIO_NET_F_MAC); | |
184bd048 DK |
287 | features |= (1 << VIRTIO_NET_F_CSUM); |
288 | features |= (1 << VIRTIO_NET_F_HOST_TSO4); | |
289 | features |= (1 << VIRTIO_NET_F_HOST_TSO6); | |
290 | features |= (1 << VIRTIO_NET_F_HOST_ECN); | |
8eca6b1b | 291 | |
8172539d | 292 | return features; |
8eca6b1b AL |
293 | } |
294 | ||
fbe78f4f AL |
295 | static void virtio_net_set_features(VirtIODevice *vdev, uint32_t features) |
296 | { | |
297 | VirtIONet *n = to_virtio_net(vdev); | |
298 | ||
ff3a8066 | 299 | virtio_net_set_mrg_rx_bufs(n, !!(features & (1 << VIRTIO_NET_F_MRG_RXBUF))); |
f5436dd9 MM |
300 | |
301 | if (n->has_vnet_hdr) { | |
eb6b6c12 | 302 | tap_set_offload(n->nic->nc.peer, |
f5436dd9 MM |
303 | (features >> VIRTIO_NET_F_GUEST_CSUM) & 1, |
304 | (features >> VIRTIO_NET_F_GUEST_TSO4) & 1, | |
305 | (features >> VIRTIO_NET_F_GUEST_TSO6) & 1, | |
6c9f58ba SS |
306 | (features >> VIRTIO_NET_F_GUEST_ECN) & 1, |
307 | (features >> VIRTIO_NET_F_GUEST_UFO) & 1); | |
f5436dd9 | 308 | } |
dc14a397 | 309 | if (!n->nic->nc.peer || |
2be64a68 | 310 | n->nic->nc.peer->info->type != NET_CLIENT_OPTIONS_KIND_TAP) { |
dc14a397 DS |
311 | return; |
312 | } | |
313 | if (!tap_get_vhost_net(n->nic->nc.peer)) { | |
314 | return; | |
315 | } | |
57c3229b | 316 | vhost_net_ack_features(tap_get_vhost_net(n->nic->nc.peer), features); |
fbe78f4f AL |
317 | } |
318 | ||
002437cd | 319 | static int virtio_net_handle_rx_mode(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 320 | struct iovec *iov, unsigned int iov_cnt) |
002437cd AL |
321 | { |
322 | uint8_t on; | |
921ac5d0 | 323 | size_t s; |
002437cd | 324 | |
921ac5d0 MT |
325 | s = iov_to_buf(iov, iov_cnt, 0, &on, sizeof(on)); |
326 | if (s != sizeof(on)) { | |
327 | return VIRTIO_NET_ERR; | |
002437cd AL |
328 | } |
329 | ||
921ac5d0 | 330 | if (cmd == VIRTIO_NET_CTRL_RX_MODE_PROMISC) { |
002437cd | 331 | n->promisc = on; |
921ac5d0 | 332 | } else if (cmd == VIRTIO_NET_CTRL_RX_MODE_ALLMULTI) { |
002437cd | 333 | n->allmulti = on; |
921ac5d0 | 334 | } else if (cmd == VIRTIO_NET_CTRL_RX_MODE_ALLUNI) { |
015cb166 | 335 | n->alluni = on; |
921ac5d0 | 336 | } else if (cmd == VIRTIO_NET_CTRL_RX_MODE_NOMULTI) { |
015cb166 | 337 | n->nomulti = on; |
921ac5d0 | 338 | } else if (cmd == VIRTIO_NET_CTRL_RX_MODE_NOUNI) { |
015cb166 | 339 | n->nouni = on; |
921ac5d0 | 340 | } else if (cmd == VIRTIO_NET_CTRL_RX_MODE_NOBCAST) { |
015cb166 | 341 | n->nobcast = on; |
921ac5d0 | 342 | } else { |
002437cd | 343 | return VIRTIO_NET_ERR; |
921ac5d0 | 344 | } |
002437cd AL |
345 | |
346 | return VIRTIO_NET_OK; | |
347 | } | |
348 | ||
b6503ed9 | 349 | static int virtio_net_handle_mac(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 350 | struct iovec *iov, unsigned int iov_cnt) |
b6503ed9 AL |
351 | { |
352 | struct virtio_net_ctrl_mac mac_data; | |
921ac5d0 | 353 | size_t s; |
b6503ed9 | 354 | |
c1943a3f AK |
355 | if (cmd == VIRTIO_NET_CTRL_MAC_ADDR_SET) { |
356 | if (iov_size(iov, iov_cnt) != sizeof(n->mac)) { | |
357 | return VIRTIO_NET_ERR; | |
358 | } | |
359 | s = iov_to_buf(iov, iov_cnt, 0, &n->mac, sizeof(n->mac)); | |
360 | assert(s == sizeof(n->mac)); | |
361 | qemu_format_nic_info_str(&n->nic->nc, n->mac); | |
362 | return VIRTIO_NET_OK; | |
363 | } | |
364 | ||
921ac5d0 | 365 | if (cmd != VIRTIO_NET_CTRL_MAC_TABLE_SET) { |
b6503ed9 | 366 | return VIRTIO_NET_ERR; |
921ac5d0 | 367 | } |
b6503ed9 AL |
368 | |
369 | n->mac_table.in_use = 0; | |
2d9aba39 | 370 | n->mac_table.first_multi = 0; |
8fd2a2f1 AW |
371 | n->mac_table.uni_overflow = 0; |
372 | n->mac_table.multi_overflow = 0; | |
b6503ed9 AL |
373 | memset(n->mac_table.macs, 0, MAC_TABLE_ENTRIES * ETH_ALEN); |
374 | ||
921ac5d0 MT |
375 | s = iov_to_buf(iov, iov_cnt, 0, &mac_data.entries, |
376 | sizeof(mac_data.entries)); | |
377 | mac_data.entries = ldl_p(&mac_data.entries); | |
378 | if (s != sizeof(mac_data.entries)) { | |
379 | return VIRTIO_NET_ERR; | |
380 | } | |
381 | iov_discard_front(&iov, &iov_cnt, s); | |
b6503ed9 | 382 | |
921ac5d0 | 383 | if (mac_data.entries * ETH_ALEN > iov_size(iov, iov_cnt)) { |
b6503ed9 | 384 | return VIRTIO_NET_ERR; |
921ac5d0 | 385 | } |
b6503ed9 AL |
386 | |
387 | if (mac_data.entries <= MAC_TABLE_ENTRIES) { | |
921ac5d0 MT |
388 | s = iov_to_buf(iov, iov_cnt, 0, n->mac_table.macs, |
389 | mac_data.entries * ETH_ALEN); | |
390 | if (s != mac_data.entries * ETH_ALEN) { | |
391 | return VIRTIO_NET_ERR; | |
392 | } | |
b6503ed9 AL |
393 | n->mac_table.in_use += mac_data.entries; |
394 | } else { | |
8fd2a2f1 | 395 | n->mac_table.uni_overflow = 1; |
b6503ed9 AL |
396 | } |
397 | ||
921ac5d0 MT |
398 | iov_discard_front(&iov, &iov_cnt, mac_data.entries * ETH_ALEN); |
399 | ||
2d9aba39 AW |
400 | n->mac_table.first_multi = n->mac_table.in_use; |
401 | ||
921ac5d0 MT |
402 | s = iov_to_buf(iov, iov_cnt, 0, &mac_data.entries, |
403 | sizeof(mac_data.entries)); | |
404 | mac_data.entries = ldl_p(&mac_data.entries); | |
405 | if (s != sizeof(mac_data.entries)) { | |
406 | return VIRTIO_NET_ERR; | |
407 | } | |
408 | ||
409 | iov_discard_front(&iov, &iov_cnt, s); | |
b6503ed9 | 410 | |
921ac5d0 | 411 | if (mac_data.entries * ETH_ALEN != iov_size(iov, iov_cnt)) { |
b6503ed9 | 412 | return VIRTIO_NET_ERR; |
921ac5d0 | 413 | } |
b6503ed9 | 414 | |
921ac5d0 MT |
415 | if (n->mac_table.in_use + mac_data.entries <= MAC_TABLE_ENTRIES) { |
416 | s = iov_to_buf(iov, iov_cnt, 0, n->mac_table.macs, | |
417 | mac_data.entries * ETH_ALEN); | |
418 | if (s != mac_data.entries * ETH_ALEN) { | |
419 | return VIRTIO_NET_ERR; | |
8fd2a2f1 | 420 | } |
921ac5d0 MT |
421 | n->mac_table.in_use += mac_data.entries; |
422 | } else { | |
423 | n->mac_table.multi_overflow = 1; | |
b6503ed9 AL |
424 | } |
425 | ||
426 | return VIRTIO_NET_OK; | |
427 | } | |
428 | ||
f21c0ed9 | 429 | static int virtio_net_handle_vlan_table(VirtIONet *n, uint8_t cmd, |
921ac5d0 | 430 | struct iovec *iov, unsigned int iov_cnt) |
f21c0ed9 AL |
431 | { |
432 | uint16_t vid; | |
921ac5d0 | 433 | size_t s; |
f21c0ed9 | 434 | |
921ac5d0 MT |
435 | s = iov_to_buf(iov, iov_cnt, 0, &vid, sizeof(vid)); |
436 | vid = lduw_p(&vid); | |
437 | if (s != sizeof(vid)) { | |
f21c0ed9 AL |
438 | return VIRTIO_NET_ERR; |
439 | } | |
440 | ||
f21c0ed9 AL |
441 | if (vid >= MAX_VLAN) |
442 | return VIRTIO_NET_ERR; | |
443 | ||
444 | if (cmd == VIRTIO_NET_CTRL_VLAN_ADD) | |
445 | n->vlans[vid >> 5] |= (1U << (vid & 0x1f)); | |
446 | else if (cmd == VIRTIO_NET_CTRL_VLAN_DEL) | |
447 | n->vlans[vid >> 5] &= ~(1U << (vid & 0x1f)); | |
448 | else | |
449 | return VIRTIO_NET_ERR; | |
450 | ||
451 | return VIRTIO_NET_OK; | |
452 | } | |
453 | ||
3d11d36c AL |
454 | static void virtio_net_handle_ctrl(VirtIODevice *vdev, VirtQueue *vq) |
455 | { | |
002437cd | 456 | VirtIONet *n = to_virtio_net(vdev); |
3d11d36c AL |
457 | struct virtio_net_ctrl_hdr ctrl; |
458 | virtio_net_ctrl_ack status = VIRTIO_NET_ERR; | |
459 | VirtQueueElement elem; | |
921ac5d0 MT |
460 | size_t s; |
461 | struct iovec *iov; | |
462 | unsigned int iov_cnt; | |
3d11d36c AL |
463 | |
464 | while (virtqueue_pop(vq, &elem)) { | |
921ac5d0 MT |
465 | if (iov_size(elem.in_sg, elem.in_num) < sizeof(status) || |
466 | iov_size(elem.out_sg, elem.out_num) < sizeof(ctrl)) { | |
e7b43f7e | 467 | error_report("virtio-net ctrl missing headers"); |
3d11d36c AL |
468 | exit(1); |
469 | } | |
470 | ||
921ac5d0 MT |
471 | iov = elem.out_sg; |
472 | iov_cnt = elem.out_num; | |
473 | s = iov_to_buf(iov, iov_cnt, 0, &ctrl, sizeof(ctrl)); | |
474 | iov_discard_front(&iov, &iov_cnt, sizeof(ctrl)); | |
475 | if (s != sizeof(ctrl)) { | |
476 | status = VIRTIO_NET_ERR; | |
477 | } else if (ctrl.class == VIRTIO_NET_CTRL_RX_MODE) { | |
478 | status = virtio_net_handle_rx_mode(n, ctrl.cmd, iov, iov_cnt); | |
479 | } else if (ctrl.class == VIRTIO_NET_CTRL_MAC) { | |
480 | status = virtio_net_handle_mac(n, ctrl.cmd, iov, iov_cnt); | |
481 | } else if (ctrl.class == VIRTIO_NET_CTRL_VLAN) { | |
482 | status = virtio_net_handle_vlan_table(n, ctrl.cmd, iov, iov_cnt); | |
3d11d36c AL |
483 | } |
484 | ||
921ac5d0 MT |
485 | s = iov_from_buf(elem.in_sg, elem.in_num, 0, &status, sizeof(status)); |
486 | assert(s == sizeof(status)); | |
3d11d36c AL |
487 | |
488 | virtqueue_push(vq, &elem, sizeof(status)); | |
489 | virtio_notify(vdev, vq); | |
490 | } | |
491 | } | |
492 | ||
fbe78f4f AL |
493 | /* RX */ |
494 | ||
495 | static void virtio_net_handle_rx(VirtIODevice *vdev, VirtQueue *vq) | |
496 | { | |
8aeff62d MM |
497 | VirtIONet *n = to_virtio_net(vdev); |
498 | ||
eb6b6c12 | 499 | qemu_flush_queued_packets(&n->nic->nc); |
fbe78f4f AL |
500 | } |
501 | ||
4e68f7a0 | 502 | static int virtio_net_can_receive(NetClientState *nc) |
fbe78f4f | 503 | { |
eb6b6c12 | 504 | VirtIONet *n = DO_UPCAST(NICState, nc, nc)->opaque; |
85cf2a8d | 505 | if (!n->vdev.vm_running) { |
95477323 MT |
506 | return 0; |
507 | } | |
cdd5cc12 | 508 | |
fbe78f4f AL |
509 | if (!virtio_queue_ready(n->rx_vq) || |
510 | !(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK)) | |
511 | return 0; | |
512 | ||
cdd5cc12 MM |
513 | return 1; |
514 | } | |
515 | ||
516 | static int virtio_net_has_buffers(VirtIONet *n, int bufsize) | |
517 | { | |
fbe78f4f AL |
518 | if (virtio_queue_empty(n->rx_vq) || |
519 | (n->mergeable_rx_bufs && | |
520 | !virtqueue_avail_bytes(n->rx_vq, bufsize, 0))) { | |
521 | virtio_queue_set_notification(n->rx_vq, 1); | |
06b12970 TL |
522 | |
523 | /* To avoid a race condition where the guest has made some buffers | |
524 | * available after the above check but before notification was | |
525 | * enabled, check for available buffers again. | |
526 | */ | |
527 | if (virtio_queue_empty(n->rx_vq) || | |
528 | (n->mergeable_rx_bufs && | |
529 | !virtqueue_avail_bytes(n->rx_vq, bufsize, 0))) | |
530 | return 0; | |
fbe78f4f AL |
531 | } |
532 | ||
533 | virtio_queue_set_notification(n->rx_vq, 0); | |
534 | return 1; | |
535 | } | |
536 | ||
1d41b0c1 AL |
537 | /* dhclient uses AF_PACKET but doesn't pass auxdata to the kernel so |
538 | * it never finds out that the packets don't have valid checksums. This | |
539 | * causes dhclient to get upset. Fedora's carried a patch for ages to | |
540 | * fix this with Xen but it hasn't appeared in an upstream release of | |
541 | * dhclient yet. | |
542 | * | |
543 | * To avoid breaking existing guests, we catch udp packets and add | |
544 | * checksums. This is terrible but it's better than hacking the guest | |
545 | * kernels. | |
546 | * | |
547 | * N.B. if we introduce a zero-copy API, this operation is no longer free so | |
548 | * we should provide a mechanism to disable it to avoid polluting the host | |
549 | * cache. | |
550 | */ | |
551 | static void work_around_broken_dhclient(struct virtio_net_hdr *hdr, | |
22cc84db | 552 | uint8_t *buf, size_t size) |
1d41b0c1 AL |
553 | { |
554 | if ((hdr->flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) && /* missing csum */ | |
555 | (size > 27 && size < 1500) && /* normal sized MTU */ | |
556 | (buf[12] == 0x08 && buf[13] == 0x00) && /* ethertype == IPv4 */ | |
557 | (buf[23] == 17) && /* ip.protocol == UDP */ | |
558 | (buf[34] == 0 && buf[35] == 67)) { /* udp.srcport == bootps */ | |
22cc84db | 559 | net_checksum_calculate(buf, size); |
1d41b0c1 AL |
560 | hdr->flags &= ~VIRTIO_NET_HDR_F_NEEDS_CSUM; |
561 | } | |
562 | } | |
563 | ||
280598b7 MT |
564 | static void receive_header(VirtIONet *n, const struct iovec *iov, int iov_cnt, |
565 | const void *buf, size_t size) | |
fbe78f4f | 566 | { |
3a330134 | 567 | if (n->has_vnet_hdr) { |
22cc84db MT |
568 | /* FIXME this cast is evil */ |
569 | void *wbuf = (void *)buf; | |
280598b7 MT |
570 | work_around_broken_dhclient(wbuf, wbuf + n->host_hdr_len, |
571 | size - n->host_hdr_len); | |
572 | iov_from_buf(iov, iov_cnt, 0, buf, sizeof(struct virtio_net_hdr)); | |
22cc84db MT |
573 | } else { |
574 | struct virtio_net_hdr hdr = { | |
575 | .flags = 0, | |
576 | .gso_type = VIRTIO_NET_HDR_GSO_NONE | |
577 | }; | |
578 | iov_from_buf(iov, iov_cnt, 0, &hdr, sizeof hdr); | |
3a330134 | 579 | } |
fbe78f4f AL |
580 | } |
581 | ||
3831ab20 AL |
582 | static int receive_filter(VirtIONet *n, const uint8_t *buf, int size) |
583 | { | |
584 | static const uint8_t bcast[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}; | |
f21c0ed9 | 585 | static const uint8_t vlan[] = {0x81, 0x00}; |
3831ab20 | 586 | uint8_t *ptr = (uint8_t *)buf; |
b6503ed9 | 587 | int i; |
3831ab20 AL |
588 | |
589 | if (n->promisc) | |
590 | return 1; | |
591 | ||
e043ebc6 | 592 | ptr += n->host_hdr_len; |
3a330134 | 593 | |
f21c0ed9 AL |
594 | if (!memcmp(&ptr[12], vlan, sizeof(vlan))) { |
595 | int vid = be16_to_cpup((uint16_t *)(ptr + 14)) & 0xfff; | |
596 | if (!(n->vlans[vid >> 5] & (1U << (vid & 0x1f)))) | |
597 | return 0; | |
598 | } | |
599 | ||
bbe2f399 AW |
600 | if (ptr[0] & 1) { // multicast |
601 | if (!memcmp(ptr, bcast, sizeof(bcast))) { | |
015cb166 AW |
602 | return !n->nobcast; |
603 | } else if (n->nomulti) { | |
604 | return 0; | |
8fd2a2f1 | 605 | } else if (n->allmulti || n->mac_table.multi_overflow) { |
bbe2f399 AW |
606 | return 1; |
607 | } | |
2d9aba39 AW |
608 | |
609 | for (i = n->mac_table.first_multi; i < n->mac_table.in_use; i++) { | |
610 | if (!memcmp(ptr, &n->mac_table.macs[i * ETH_ALEN], ETH_ALEN)) { | |
611 | return 1; | |
612 | } | |
613 | } | |
bbe2f399 | 614 | } else { // unicast |
015cb166 AW |
615 | if (n->nouni) { |
616 | return 0; | |
617 | } else if (n->alluni || n->mac_table.uni_overflow) { | |
8fd2a2f1 AW |
618 | return 1; |
619 | } else if (!memcmp(ptr, n->mac, ETH_ALEN)) { | |
bbe2f399 AW |
620 | return 1; |
621 | } | |
3831ab20 | 622 | |
2d9aba39 AW |
623 | for (i = 0; i < n->mac_table.first_multi; i++) { |
624 | if (!memcmp(ptr, &n->mac_table.macs[i * ETH_ALEN], ETH_ALEN)) { | |
625 | return 1; | |
626 | } | |
627 | } | |
b6503ed9 AL |
628 | } |
629 | ||
3831ab20 AL |
630 | return 0; |
631 | } | |
632 | ||
4e68f7a0 | 633 | static ssize_t virtio_net_receive(NetClientState *nc, const uint8_t *buf, size_t size) |
fbe78f4f | 634 | { |
eb6b6c12 | 635 | VirtIONet *n = DO_UPCAST(NICState, nc, nc)->opaque; |
63c58728 MT |
636 | struct iovec mhdr_sg[VIRTQUEUE_MAX_SIZE]; |
637 | struct virtio_net_hdr_mrg_rxbuf mhdr; | |
638 | unsigned mhdr_cnt = 0; | |
22cc84db | 639 | size_t offset, i, guest_offset; |
fbe78f4f | 640 | |
eb6b6c12 | 641 | if (!virtio_net_can_receive(&n->nic->nc)) |
cdd5cc12 MM |
642 | return -1; |
643 | ||
940cda94 | 644 | /* hdr_len refers to the header we supply to the guest */ |
e35e23f6 | 645 | if (!virtio_net_has_buffers(n, size + n->guest_hdr_len - n->host_hdr_len)) |
8aeff62d | 646 | return 0; |
fbe78f4f | 647 | |
3831ab20 | 648 | if (!receive_filter(n, buf, size)) |
4f1c942b | 649 | return size; |
3831ab20 | 650 | |
fbe78f4f AL |
651 | offset = i = 0; |
652 | ||
653 | while (offset < size) { | |
654 | VirtQueueElement elem; | |
655 | int len, total; | |
22cc84db | 656 | const struct iovec *sg = elem.in_sg; |
fbe78f4f | 657 | |
22c253d9 | 658 | total = 0; |
fbe78f4f | 659 | |
279a4253 | 660 | if (virtqueue_pop(n->rx_vq, &elem) == 0) { |
fbe78f4f | 661 | if (i == 0) |
4f1c942b | 662 | return -1; |
e7b43f7e | 663 | error_report("virtio-net unexpected empty queue: " |
279a4253 | 664 | "i %zd mergeable %d offset %zd, size %zd, " |
e7b43f7e | 665 | "guest hdr len %zd, host hdr len %zd guest features 0x%x", |
279a4253 | 666 | i, n->mergeable_rx_bufs, offset, size, |
e35e23f6 | 667 | n->guest_hdr_len, n->host_hdr_len, n->vdev.guest_features); |
fbe78f4f AL |
668 | exit(1); |
669 | } | |
670 | ||
671 | if (elem.in_num < 1) { | |
e7b43f7e | 672 | error_report("virtio-net receive queue contains no in buffers"); |
fbe78f4f AL |
673 | exit(1); |
674 | } | |
675 | ||
fbe78f4f | 676 | if (i == 0) { |
c8d28e7e | 677 | assert(offset == 0); |
63c58728 MT |
678 | if (n->mergeable_rx_bufs) { |
679 | mhdr_cnt = iov_copy(mhdr_sg, ARRAY_SIZE(mhdr_sg), | |
680 | sg, elem.in_num, | |
681 | offsetof(typeof(mhdr), num_buffers), | |
682 | sizeof(mhdr.num_buffers)); | |
683 | } | |
fbe78f4f | 684 | |
c8d28e7e MT |
685 | receive_header(n, sg, elem.in_num, buf, size); |
686 | offset = n->host_hdr_len; | |
e35e23f6 | 687 | total += n->guest_hdr_len; |
22cc84db MT |
688 | guest_offset = n->guest_hdr_len; |
689 | } else { | |
690 | guest_offset = 0; | |
fbe78f4f AL |
691 | } |
692 | ||
693 | /* copy in packet. ugh */ | |
22cc84db | 694 | len = iov_from_buf(sg, elem.in_num, guest_offset, |
dcf6f5e1 | 695 | buf + offset, size - offset); |
fbe78f4f | 696 | total += len; |
279a4253 MT |
697 | offset += len; |
698 | /* If buffers can't be merged, at this point we | |
699 | * must have consumed the complete packet. | |
700 | * Otherwise, drop it. */ | |
701 | if (!n->mergeable_rx_bufs && offset < size) { | |
702 | #if 0 | |
e7b43f7e SH |
703 | error_report("virtio-net truncated non-mergeable packet: " |
704 | "i %zd mergeable %d offset %zd, size %zd, " | |
705 | "guest hdr len %zd, host hdr len %zd", | |
706 | i, n->mergeable_rx_bufs, | |
e35e23f6 | 707 | offset, size, n->guest_hdr_len, n->host_hdr_len); |
279a4253 MT |
708 | #endif |
709 | return size; | |
710 | } | |
fbe78f4f AL |
711 | |
712 | /* signal other side */ | |
713 | virtqueue_fill(n->rx_vq, &elem, total, i++); | |
fbe78f4f AL |
714 | } |
715 | ||
63c58728 MT |
716 | if (mhdr_cnt) { |
717 | stw_p(&mhdr.num_buffers, i); | |
718 | iov_from_buf(mhdr_sg, mhdr_cnt, | |
719 | 0, | |
720 | &mhdr.num_buffers, sizeof mhdr.num_buffers); | |
44b15bc5 | 721 | } |
fbe78f4f AL |
722 | |
723 | virtqueue_flush(n->rx_vq, i); | |
724 | virtio_notify(&n->vdev, n->rx_vq); | |
4f1c942b MM |
725 | |
726 | return size; | |
fbe78f4f AL |
727 | } |
728 | ||
e3f30488 | 729 | static int32_t virtio_net_flush_tx(VirtIONet *n, VirtQueue *vq); |
6243375f | 730 | |
4e68f7a0 | 731 | static void virtio_net_tx_complete(NetClientState *nc, ssize_t len) |
6243375f | 732 | { |
eb6b6c12 | 733 | VirtIONet *n = DO_UPCAST(NICState, nc, nc)->opaque; |
6243375f | 734 | |
40bad8f3 | 735 | virtqueue_push(n->tx_vq, &n->async_tx.elem, 0); |
6243375f MM |
736 | virtio_notify(&n->vdev, n->tx_vq); |
737 | ||
738 | n->async_tx.elem.out_num = n->async_tx.len = 0; | |
739 | ||
740 | virtio_queue_set_notification(n->tx_vq, 1); | |
741 | virtio_net_flush_tx(n, n->tx_vq); | |
742 | } | |
743 | ||
fbe78f4f | 744 | /* TX */ |
e3f30488 | 745 | static int32_t virtio_net_flush_tx(VirtIONet *n, VirtQueue *vq) |
fbe78f4f AL |
746 | { |
747 | VirtQueueElement elem; | |
e3f30488 | 748 | int32_t num_packets = 0; |
e3f30488 AW |
749 | if (!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK)) { |
750 | return num_packets; | |
751 | } | |
fbe78f4f | 752 | |
85cf2a8d | 753 | assert(n->vdev.vm_running); |
783e7706 | 754 | |
6243375f MM |
755 | if (n->async_tx.elem.out_num) { |
756 | virtio_queue_set_notification(n->tx_vq, 0); | |
e3f30488 | 757 | return num_packets; |
6243375f MM |
758 | } |
759 | ||
fbe78f4f | 760 | while (virtqueue_pop(vq, &elem)) { |
14761f9c | 761 | ssize_t ret, len; |
fbe78f4f AL |
762 | unsigned int out_num = elem.out_num; |
763 | struct iovec *out_sg = &elem.out_sg[0]; | |
14761f9c | 764 | struct iovec sg[VIRTQUEUE_MAX_SIZE]; |
fbe78f4f | 765 | |
7b80d08e | 766 | if (out_num < 1) { |
e7b43f7e | 767 | error_report("virtio-net header not in first element"); |
fbe78f4f AL |
768 | exit(1); |
769 | } | |
770 | ||
14761f9c MT |
771 | /* |
772 | * If host wants to see the guest header as is, we can | |
773 | * pass it on unchanged. Otherwise, copy just the parts | |
774 | * that host is interested in. | |
775 | */ | |
776 | assert(n->host_hdr_len <= n->guest_hdr_len); | |
777 | if (n->host_hdr_len != n->guest_hdr_len) { | |
778 | unsigned sg_num = iov_copy(sg, ARRAY_SIZE(sg), | |
779 | out_sg, out_num, | |
780 | 0, n->host_hdr_len); | |
781 | sg_num += iov_copy(sg + sg_num, ARRAY_SIZE(sg) - sg_num, | |
782 | out_sg, out_num, | |
783 | n->guest_hdr_len, -1); | |
784 | out_num = sg_num; | |
785 | out_sg = sg; | |
fbe78f4f AL |
786 | } |
787 | ||
7b80d08e | 788 | len = n->guest_hdr_len; |
14761f9c | 789 | |
eb6b6c12 | 790 | ret = qemu_sendv_packet_async(&n->nic->nc, out_sg, out_num, |
6243375f MM |
791 | virtio_net_tx_complete); |
792 | if (ret == 0) { | |
793 | virtio_queue_set_notification(n->tx_vq, 0); | |
794 | n->async_tx.elem = elem; | |
795 | n->async_tx.len = len; | |
e3f30488 | 796 | return -EBUSY; |
6243375f MM |
797 | } |
798 | ||
799 | len += ret; | |
fbe78f4f | 800 | |
40bad8f3 | 801 | virtqueue_push(vq, &elem, 0); |
fbe78f4f | 802 | virtio_notify(&n->vdev, vq); |
e3f30488 AW |
803 | |
804 | if (++num_packets >= n->tx_burst) { | |
805 | break; | |
806 | } | |
fbe78f4f | 807 | } |
e3f30488 | 808 | return num_packets; |
fbe78f4f AL |
809 | } |
810 | ||
a697a334 | 811 | static void virtio_net_handle_tx_timer(VirtIODevice *vdev, VirtQueue *vq) |
fbe78f4f AL |
812 | { |
813 | VirtIONet *n = to_virtio_net(vdev); | |
814 | ||
783e7706 | 815 | /* This happens when device was stopped but VCPU wasn't. */ |
85cf2a8d | 816 | if (!n->vdev.vm_running) { |
783e7706 MT |
817 | n->tx_waiting = 1; |
818 | return; | |
819 | } | |
820 | ||
4b4b8d36 | 821 | if (n->tx_waiting) { |
fbe78f4f AL |
822 | virtio_queue_set_notification(vq, 1); |
823 | qemu_del_timer(n->tx_timer); | |
4b4b8d36 | 824 | n->tx_waiting = 0; |
fbe78f4f AL |
825 | virtio_net_flush_tx(n, vq); |
826 | } else { | |
827 | qemu_mod_timer(n->tx_timer, | |
74475455 | 828 | qemu_get_clock_ns(vm_clock) + n->tx_timeout); |
4b4b8d36 | 829 | n->tx_waiting = 1; |
fbe78f4f AL |
830 | virtio_queue_set_notification(vq, 0); |
831 | } | |
832 | } | |
833 | ||
a697a334 AW |
834 | static void virtio_net_handle_tx_bh(VirtIODevice *vdev, VirtQueue *vq) |
835 | { | |
836 | VirtIONet *n = to_virtio_net(vdev); | |
837 | ||
838 | if (unlikely(n->tx_waiting)) { | |
839 | return; | |
840 | } | |
783e7706 MT |
841 | n->tx_waiting = 1; |
842 | /* This happens when device was stopped but VCPU wasn't. */ | |
85cf2a8d | 843 | if (!n->vdev.vm_running) { |
783e7706 MT |
844 | return; |
845 | } | |
a697a334 AW |
846 | virtio_queue_set_notification(vq, 0); |
847 | qemu_bh_schedule(n->tx_bh); | |
a697a334 AW |
848 | } |
849 | ||
fbe78f4f AL |
850 | static void virtio_net_tx_timer(void *opaque) |
851 | { | |
852 | VirtIONet *n = opaque; | |
85cf2a8d | 853 | assert(n->vdev.vm_running); |
fbe78f4f | 854 | |
4b4b8d36 | 855 | n->tx_waiting = 0; |
fbe78f4f AL |
856 | |
857 | /* Just in case the driver is not ready on more */ | |
858 | if (!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK)) | |
859 | return; | |
860 | ||
861 | virtio_queue_set_notification(n->tx_vq, 1); | |
862 | virtio_net_flush_tx(n, n->tx_vq); | |
863 | } | |
864 | ||
a697a334 AW |
865 | static void virtio_net_tx_bh(void *opaque) |
866 | { | |
867 | VirtIONet *n = opaque; | |
868 | int32_t ret; | |
869 | ||
85cf2a8d | 870 | assert(n->vdev.vm_running); |
783e7706 | 871 | |
a697a334 AW |
872 | n->tx_waiting = 0; |
873 | ||
874 | /* Just in case the driver is not ready on more */ | |
875 | if (unlikely(!(n->vdev.status & VIRTIO_CONFIG_S_DRIVER_OK))) | |
876 | return; | |
877 | ||
878 | ret = virtio_net_flush_tx(n, n->tx_vq); | |
879 | if (ret == -EBUSY) { | |
880 | return; /* Notification re-enable handled by tx_complete */ | |
881 | } | |
882 | ||
883 | /* If we flush a full burst of packets, assume there are | |
884 | * more coming and immediately reschedule */ | |
885 | if (ret >= n->tx_burst) { | |
886 | qemu_bh_schedule(n->tx_bh); | |
887 | n->tx_waiting = 1; | |
888 | return; | |
889 | } | |
890 | ||
891 | /* If less than a full burst, re-enable notification and flush | |
892 | * anything that may have come in while we weren't looking. If | |
893 | * we find something, assume the guest is still active and reschedule */ | |
894 | virtio_queue_set_notification(n->tx_vq, 1); | |
895 | if (virtio_net_flush_tx(n, n->tx_vq) > 0) { | |
896 | virtio_queue_set_notification(n->tx_vq, 0); | |
897 | qemu_bh_schedule(n->tx_bh); | |
898 | n->tx_waiting = 1; | |
899 | } | |
900 | } | |
901 | ||
fbe78f4f AL |
902 | static void virtio_net_save(QEMUFile *f, void *opaque) |
903 | { | |
904 | VirtIONet *n = opaque; | |
905 | ||
afbaa7b4 MT |
906 | /* At this point, backend must be stopped, otherwise |
907 | * it might keep writing to memory. */ | |
908 | assert(!n->vhost_started); | |
fbe78f4f AL |
909 | virtio_save(&n->vdev, f); |
910 | ||
79674068 | 911 | qemu_put_buffer(f, n->mac, ETH_ALEN); |
4b4b8d36 | 912 | qemu_put_be32(f, n->tx_waiting); |
e46cb38f | 913 | qemu_put_be32(f, n->mergeable_rx_bufs); |
9d6271b8 | 914 | qemu_put_be16(f, n->status); |
f10c592e AW |
915 | qemu_put_byte(f, n->promisc); |
916 | qemu_put_byte(f, n->allmulti); | |
b6503ed9 AL |
917 | qemu_put_be32(f, n->mac_table.in_use); |
918 | qemu_put_buffer(f, n->mac_table.macs, n->mac_table.in_use * ETH_ALEN); | |
f21c0ed9 | 919 | qemu_put_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3); |
3a330134 | 920 | qemu_put_be32(f, n->has_vnet_hdr); |
8fd2a2f1 AW |
921 | qemu_put_byte(f, n->mac_table.multi_overflow); |
922 | qemu_put_byte(f, n->mac_table.uni_overflow); | |
015cb166 AW |
923 | qemu_put_byte(f, n->alluni); |
924 | qemu_put_byte(f, n->nomulti); | |
925 | qemu_put_byte(f, n->nouni); | |
926 | qemu_put_byte(f, n->nobcast); | |
0ce0e8f4 | 927 | qemu_put_byte(f, n->has_ufo); |
fbe78f4f AL |
928 | } |
929 | ||
930 | static int virtio_net_load(QEMUFile *f, void *opaque, int version_id) | |
931 | { | |
932 | VirtIONet *n = opaque; | |
2d9aba39 | 933 | int i; |
2a633c46 | 934 | int ret; |
fbe78f4f | 935 | |
9d6271b8 | 936 | if (version_id < 2 || version_id > VIRTIO_NET_VM_VERSION) |
fbe78f4f AL |
937 | return -EINVAL; |
938 | ||
2a633c46 OW |
939 | ret = virtio_load(&n->vdev, f); |
940 | if (ret) { | |
941 | return ret; | |
942 | } | |
fbe78f4f | 943 | |
79674068 | 944 | qemu_get_buffer(f, n->mac, ETH_ALEN); |
4b4b8d36 | 945 | n->tx_waiting = qemu_get_be32(f); |
ff3a8066 MT |
946 | |
947 | virtio_net_set_mrg_rx_bufs(n, qemu_get_be32(f)); | |
fbe78f4f | 948 | |
9d6271b8 AL |
949 | if (version_id >= 3) |
950 | n->status = qemu_get_be16(f); | |
951 | ||
002437cd | 952 | if (version_id >= 4) { |
f10c592e AW |
953 | if (version_id < 8) { |
954 | n->promisc = qemu_get_be32(f); | |
955 | n->allmulti = qemu_get_be32(f); | |
956 | } else { | |
957 | n->promisc = qemu_get_byte(f); | |
958 | n->allmulti = qemu_get_byte(f); | |
959 | } | |
002437cd AL |
960 | } |
961 | ||
b6503ed9 AL |
962 | if (version_id >= 5) { |
963 | n->mac_table.in_use = qemu_get_be32(f); | |
964 | /* MAC_TABLE_ENTRIES may be different from the saved image */ | |
965 | if (n->mac_table.in_use <= MAC_TABLE_ENTRIES) { | |
966 | qemu_get_buffer(f, n->mac_table.macs, | |
967 | n->mac_table.in_use * ETH_ALEN); | |
968 | } else if (n->mac_table.in_use) { | |
e398d61b JQ |
969 | uint8_t *buf = g_malloc0(n->mac_table.in_use); |
970 | qemu_get_buffer(f, buf, n->mac_table.in_use * ETH_ALEN); | |
971 | g_free(buf); | |
8fd2a2f1 | 972 | n->mac_table.multi_overflow = n->mac_table.uni_overflow = 1; |
b6503ed9 AL |
973 | n->mac_table.in_use = 0; |
974 | } | |
975 | } | |
976 | ||
f21c0ed9 AL |
977 | if (version_id >= 6) |
978 | qemu_get_buffer(f, (uint8_t *)n->vlans, MAX_VLAN >> 3); | |
979 | ||
3a330134 MM |
980 | if (version_id >= 7) { |
981 | if (qemu_get_be32(f) && !peer_has_vnet_hdr(n)) { | |
1ecda02b | 982 | error_report("virtio-net: saved image requires vnet_hdr=on"); |
3a330134 MM |
983 | return -1; |
984 | } | |
985 | ||
986 | if (n->has_vnet_hdr) { | |
eb6b6c12 | 987 | tap_set_offload(n->nic->nc.peer, |
704a76fc MT |
988 | (n->vdev.guest_features >> VIRTIO_NET_F_GUEST_CSUM) & 1, |
989 | (n->vdev.guest_features >> VIRTIO_NET_F_GUEST_TSO4) & 1, | |
990 | (n->vdev.guest_features >> VIRTIO_NET_F_GUEST_TSO6) & 1, | |
991 | (n->vdev.guest_features >> VIRTIO_NET_F_GUEST_ECN) & 1, | |
992 | (n->vdev.guest_features >> VIRTIO_NET_F_GUEST_UFO) & 1); | |
3a330134 | 993 | } |
6c042c16 AW |
994 | } |
995 | ||
8fd2a2f1 AW |
996 | if (version_id >= 9) { |
997 | n->mac_table.multi_overflow = qemu_get_byte(f); | |
998 | n->mac_table.uni_overflow = qemu_get_byte(f); | |
999 | } | |
1000 | ||
015cb166 AW |
1001 | if (version_id >= 10) { |
1002 | n->alluni = qemu_get_byte(f); | |
1003 | n->nomulti = qemu_get_byte(f); | |
1004 | n->nouni = qemu_get_byte(f); | |
1005 | n->nobcast = qemu_get_byte(f); | |
1006 | } | |
1007 | ||
0ce0e8f4 MM |
1008 | if (version_id >= 11) { |
1009 | if (qemu_get_byte(f) && !peer_has_ufo(n)) { | |
1ecda02b | 1010 | error_report("virtio-net: saved image requires TUN_F_UFO support"); |
0ce0e8f4 MM |
1011 | return -1; |
1012 | } | |
1013 | } | |
1014 | ||
2d9aba39 AW |
1015 | /* Find the first multicast entry in the saved MAC filter */ |
1016 | for (i = 0; i < n->mac_table.in_use; i++) { | |
1017 | if (n->mac_table.macs[i * ETH_ALEN] & 1) { | |
1018 | break; | |
1019 | } | |
1020 | } | |
1021 | n->mac_table.first_multi = i; | |
98991481 AK |
1022 | |
1023 | /* nc.link_down can't be migrated, so infer link_down according | |
1024 | * to link status bit in n->status */ | |
1025 | n->nic->nc.link_down = (n->status & VIRTIO_NET_S_LINK_UP) == 0; | |
1026 | ||
fbe78f4f AL |
1027 | return 0; |
1028 | } | |
1029 | ||
4e68f7a0 | 1030 | static void virtio_net_cleanup(NetClientState *nc) |
b946a153 | 1031 | { |
eb6b6c12 | 1032 | VirtIONet *n = DO_UPCAST(NICState, nc, nc)->opaque; |
b946a153 | 1033 | |
eb6b6c12 | 1034 | n->nic = NULL; |
b946a153 AL |
1035 | } |
1036 | ||
eb6b6c12 | 1037 | static NetClientInfo net_virtio_info = { |
2be64a68 | 1038 | .type = NET_CLIENT_OPTIONS_KIND_NIC, |
eb6b6c12 MM |
1039 | .size = sizeof(NICState), |
1040 | .can_receive = virtio_net_can_receive, | |
1041 | .receive = virtio_net_receive, | |
1042 | .cleanup = virtio_net_cleanup, | |
1043 | .link_status_changed = virtio_net_set_link_status, | |
1044 | }; | |
1045 | ||
f56a1247 MT |
1046 | static bool virtio_net_guest_notifier_pending(VirtIODevice *vdev, int idx) |
1047 | { | |
1048 | VirtIONet *n = to_virtio_net(vdev); | |
1049 | assert(n->vhost_started); | |
1050 | return vhost_net_virtqueue_pending(tap_get_vhost_net(n->nic->nc.peer), idx); | |
1051 | } | |
1052 | ||
1053 | static void virtio_net_guest_notifier_mask(VirtIODevice *vdev, int idx, | |
1054 | bool mask) | |
1055 | { | |
1056 | VirtIONet *n = to_virtio_net(vdev); | |
1057 | assert(n->vhost_started); | |
1058 | vhost_net_virtqueue_mask(tap_get_vhost_net(n->nic->nc.peer), | |
1059 | vdev, idx, mask); | |
1060 | } | |
1061 | ||
f0c07c7c AW |
1062 | VirtIODevice *virtio_net_init(DeviceState *dev, NICConf *conf, |
1063 | virtio_net_conf *net) | |
fbe78f4f AL |
1064 | { |
1065 | VirtIONet *n; | |
fbe78f4f | 1066 | |
53c25cea PB |
1067 | n = (VirtIONet *)virtio_common_init("virtio-net", VIRTIO_ID_NET, |
1068 | sizeof(struct virtio_net_config), | |
1069 | sizeof(VirtIONet)); | |
fbe78f4f | 1070 | |
0f03eca6 AL |
1071 | n->vdev.get_config = virtio_net_get_config; |
1072 | n->vdev.set_config = virtio_net_set_config; | |
fbe78f4f AL |
1073 | n->vdev.get_features = virtio_net_get_features; |
1074 | n->vdev.set_features = virtio_net_set_features; | |
8eca6b1b | 1075 | n->vdev.bad_features = virtio_net_bad_features; |
002437cd | 1076 | n->vdev.reset = virtio_net_reset; |
9bc6304c | 1077 | n->vdev.set_status = virtio_net_set_status; |
f56a1247 MT |
1078 | n->vdev.guest_notifier_mask = virtio_net_guest_notifier_mask; |
1079 | n->vdev.guest_notifier_pending = virtio_net_guest_notifier_pending; | |
fbe78f4f | 1080 | n->rx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_rx); |
a697a334 AW |
1081 | |
1082 | if (net->tx && strcmp(net->tx, "timer") && strcmp(net->tx, "bh")) { | |
e7b43f7e SH |
1083 | error_report("virtio-net: " |
1084 | "Unknown option tx=%s, valid options: \"timer\" \"bh\"", | |
1085 | net->tx); | |
1086 | error_report("Defaulting to \"bh\""); | |
a697a334 AW |
1087 | } |
1088 | ||
1089 | if (net->tx && !strcmp(net->tx, "timer")) { | |
1090 | n->tx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_tx_timer); | |
74475455 | 1091 | n->tx_timer = qemu_new_timer_ns(vm_clock, virtio_net_tx_timer, n); |
a697a334 AW |
1092 | n->tx_timeout = net->txtimer; |
1093 | } else { | |
1094 | n->tx_vq = virtio_add_queue(&n->vdev, 256, virtio_net_handle_tx_bh); | |
1095 | n->tx_bh = qemu_bh_new(virtio_net_tx_bh, n); | |
1096 | } | |
4ffb17f5 | 1097 | n->ctrl_vq = virtio_add_queue(&n->vdev, 64, virtio_net_handle_ctrl); |
97b15621 | 1098 | qemu_macaddr_default_if_unset(&conf->macaddr); |
3cbe04c4 | 1099 | memcpy(&n->mac[0], &conf->macaddr, sizeof(n->mac)); |
554c97dd | 1100 | n->status = VIRTIO_NET_S_LINK_UP; |
fbe78f4f | 1101 | |
f79f2bfc | 1102 | n->nic = qemu_new_nic(&net_virtio_info, conf, object_get_typename(OBJECT(dev)), dev->id, n); |
6e371ab8 MT |
1103 | peer_test_vnet_hdr(n); |
1104 | if (peer_has_vnet_hdr(n)) { | |
1105 | tap_using_vnet_hdr(n->nic->nc.peer, 1); | |
1106 | n->host_hdr_len = sizeof(struct virtio_net_hdr); | |
1107 | } else { | |
1108 | n->host_hdr_len = 0; | |
1109 | } | |
eb6b6c12 MM |
1110 | |
1111 | qemu_format_nic_info_str(&n->nic->nc, conf->macaddr.a); | |
96d5e201 | 1112 | |
4b4b8d36 | 1113 | n->tx_waiting = 0; |
e3f30488 | 1114 | n->tx_burst = net->txburst; |
ff3a8066 | 1115 | virtio_net_set_mrg_rx_bufs(n, 0); |
002437cd | 1116 | n->promisc = 1; /* for compatibility */ |
fbe78f4f | 1117 | |
7267c094 | 1118 | n->mac_table.macs = g_malloc0(MAC_TABLE_ENTRIES * ETH_ALEN); |
b6503ed9 | 1119 | |
7267c094 | 1120 | n->vlans = g_malloc0(MAX_VLAN >> 3); |
f21c0ed9 | 1121 | |
01657c86 AW |
1122 | n->qdev = dev; |
1123 | register_savevm(dev, "virtio-net", -1, VIRTIO_NET_VM_VERSION, | |
fbe78f4f | 1124 | virtio_net_save, virtio_net_load, n); |
cf21e106 | 1125 | |
1ca4d09a GN |
1126 | add_boot_device_path(conf->bootindex, dev, "/ethernet-phy@0"); |
1127 | ||
53c25cea | 1128 | return &n->vdev; |
cf21e106 | 1129 | } |
97b15621 GH |
1130 | |
1131 | void virtio_net_exit(VirtIODevice *vdev) | |
1132 | { | |
1133 | VirtIONet *n = DO_UPCAST(VirtIONet, vdev, vdev); | |
9bc6304c | 1134 | |
afbaa7b4 MT |
1135 | /* This will stop vhost backend if appropriate. */ |
1136 | virtio_net_set_status(vdev, 0); | |
97b15621 | 1137 | |
eb6b6c12 | 1138 | qemu_purge_queued_packets(&n->nic->nc); |
97b15621 | 1139 | |
01657c86 | 1140 | unregister_savevm(n->qdev, "virtio-net", n); |
97b15621 | 1141 | |
7267c094 AL |
1142 | g_free(n->mac_table.macs); |
1143 | g_free(n->vlans); | |
97b15621 | 1144 | |
a697a334 AW |
1145 | if (n->tx_timer) { |
1146 | qemu_del_timer(n->tx_timer); | |
1147 | qemu_free_timer(n->tx_timer); | |
1148 | } else { | |
1149 | qemu_bh_delete(n->tx_bh); | |
1150 | } | |
97b15621 | 1151 | |
b20c6b9e | 1152 | qemu_del_net_client(&n->nic->nc); |
b52dfd71 | 1153 | virtio_cleanup(&n->vdev); |
97b15621 | 1154 | } |