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