]>
Commit | Line | Data |
---|---|---|
a672b469 AL |
1 | /* |
2 | * QEMU System Emulator | |
3 | * | |
4 | * Copyright (c) 2003-2008 Fabrice Bellard | |
76cc7b58 JQ |
5 | * Copyright (c) 2009-2015 Red Hat Inc |
6 | * | |
7 | * Authors: | |
8 | * Juan Quintela <quintela@redhat.com> | |
a672b469 AL |
9 | * |
10 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
11 | * of this software and associated documentation files (the "Software"), to deal | |
12 | * in the Software without restriction, including without limitation the rights | |
13 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
14 | * copies of the Software, and to permit persons to whom the Software is | |
15 | * furnished to do so, subject to the following conditions: | |
16 | * | |
17 | * The above copyright notice and this permission notice shall be included in | |
18 | * all copies or substantial portions of the Software. | |
19 | * | |
20 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
21 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
23 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
24 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
25 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
26 | * THE SOFTWARE. | |
27 | */ | |
a672b469 | 28 | |
1393a485 | 29 | #include "qemu/osdep.h" |
33c11879 | 30 | #include "cpu.h" |
abfd9ce3 | 31 | #include "hw/boards.h" |
511d2b14 | 32 | #include "hw/hw.h" |
7685ee6a | 33 | #include "hw/qdev.h" |
1422e32d | 34 | #include "net/net.h" |
83c9089e | 35 | #include "monitor/monitor.h" |
9c17d615 | 36 | #include "sysemu/sysemu.h" |
1de7afc9 | 37 | #include "qemu/timer.h" |
511d2b14 | 38 | #include "audio/audio.h" |
caf71f86 | 39 | #include "migration/migration.h" |
eb59db53 | 40 | #include "migration/postcopy-ram.h" |
cc7a8ea7 | 41 | #include "qapi/qmp/qerror.h" |
d49b6836 | 42 | #include "qemu/error-report.h" |
1de7afc9 PB |
43 | #include "qemu/sockets.h" |
44 | #include "qemu/queue.h" | |
9c17d615 | 45 | #include "sysemu/cpus.h" |
022c62cb | 46 | #include "exec/memory.h" |
a7ae8355 | 47 | #include "qmp-commands.h" |
517a13c9 | 48 | #include "trace.h" |
093e3c42 | 49 | #include "qemu/bitops.h" |
28085f7b | 50 | #include "qemu/iov.h" |
de08c606 | 51 | #include "block/snapshot.h" |
f364ec65 | 52 | #include "block/qapi.h" |
f348b6d1 | 53 | #include "qemu/cutils.h" |
61b67d47 | 54 | #include "io/channel-buffer.h" |
8925839f | 55 | #include "io/channel-file.h" |
a672b469 | 56 | |
18995b98 | 57 | #ifndef ETH_P_RARP |
f8778a77 | 58 | #define ETH_P_RARP 0x8035 |
18995b98 N |
59 | #endif |
60 | #define ARP_HTYPE_ETH 0x0001 | |
61 | #define ARP_PTYPE_IP 0x0800 | |
62 | #define ARP_OP_REQUEST_REV 0x3 | |
63 | ||
093e3c42 DDAG |
64 | const unsigned int postcopy_ram_discard_version = 0; |
65 | ||
37fb569c DDAG |
66 | static bool skip_section_footers; |
67 | ||
c76ca188 DDAG |
68 | static struct mig_cmd_args { |
69 | ssize_t len; /* -1 = variable */ | |
70 | const char *name; | |
71 | } mig_cmd_args[] = { | |
72 | [MIG_CMD_INVALID] = { .len = -1, .name = "INVALID" }, | |
2e37701e DDAG |
73 | [MIG_CMD_OPEN_RETURN_PATH] = { .len = 0, .name = "OPEN_RETURN_PATH" }, |
74 | [MIG_CMD_PING] = { .len = sizeof(uint32_t), .name = "PING" }, | |
093e3c42 DDAG |
75 | [MIG_CMD_POSTCOPY_ADVISE] = { .len = 16, .name = "POSTCOPY_ADVISE" }, |
76 | [MIG_CMD_POSTCOPY_LISTEN] = { .len = 0, .name = "POSTCOPY_LISTEN" }, | |
77 | [MIG_CMD_POSTCOPY_RUN] = { .len = 0, .name = "POSTCOPY_RUN" }, | |
78 | [MIG_CMD_POSTCOPY_RAM_DISCARD] = { | |
79 | .len = -1, .name = "POSTCOPY_RAM_DISCARD" }, | |
11cf1d98 | 80 | [MIG_CMD_PACKAGED] = { .len = 4, .name = "PACKAGED" }, |
c76ca188 DDAG |
81 | [MIG_CMD_MAX] = { .len = -1, .name = "MAX" }, |
82 | }; | |
83 | ||
18995b98 | 84 | static int announce_self_create(uint8_t *buf, |
5cecf414 | 85 | uint8_t *mac_addr) |
a672b469 | 86 | { |
18995b98 N |
87 | /* Ethernet header. */ |
88 | memset(buf, 0xff, 6); /* destination MAC addr */ | |
89 | memcpy(buf + 6, mac_addr, 6); /* source MAC addr */ | |
90 | *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */ | |
91 | ||
92 | /* RARP header. */ | |
93 | *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */ | |
94 | *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */ | |
95 | *(buf + 18) = 6; /* hardware addr length (ethernet) */ | |
96 | *(buf + 19) = 4; /* protocol addr length (IPv4) */ | |
97 | *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */ | |
98 | memcpy(buf + 22, mac_addr, 6); /* source hw addr */ | |
99 | memset(buf + 28, 0x00, 4); /* source protocol addr */ | |
100 | memcpy(buf + 32, mac_addr, 6); /* target hw addr */ | |
101 | memset(buf + 38, 0x00, 4); /* target protocol addr */ | |
102 | ||
103 | /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */ | |
104 | memset(buf + 42, 0x00, 18); | |
105 | ||
106 | return 60; /* len (FCS will be added by hardware) */ | |
a672b469 AL |
107 | } |
108 | ||
f401ca22 | 109 | static void qemu_announce_self_iter(NICState *nic, void *opaque) |
a672b469 | 110 | { |
18995b98 | 111 | uint8_t buf[60]; |
f401ca22 MM |
112 | int len; |
113 | ||
9013dca5 | 114 | trace_qemu_announce_self_iter(qemu_ether_ntoa(&nic->conf->macaddr)); |
f401ca22 MM |
115 | len = announce_self_create(buf, nic->conf->macaddr.a); |
116 | ||
b356f76d | 117 | qemu_send_packet_raw(qemu_get_queue(nic), buf, len); |
f401ca22 MM |
118 | } |
119 | ||
120 | ||
121 | static void qemu_announce_self_once(void *opaque) | |
122 | { | |
ed8b330b GN |
123 | static int count = SELF_ANNOUNCE_ROUNDS; |
124 | QEMUTimer *timer = *(QEMUTimer **)opaque; | |
a672b469 | 125 | |
f401ca22 MM |
126 | qemu_foreach_nic(qemu_announce_self_iter, NULL); |
127 | ||
18995b98 N |
128 | if (--count) { |
129 | /* delay 50ms, 150ms, 250ms, ... */ | |
bc72ad67 | 130 | timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + |
508e1180 | 131 | self_announce_delay(count)); |
ed8b330b | 132 | } else { |
5cecf414 EH |
133 | timer_del(timer); |
134 | timer_free(timer); | |
ed8b330b GN |
135 | } |
136 | } | |
137 | ||
138 | void qemu_announce_self(void) | |
139 | { | |
5cecf414 EH |
140 | static QEMUTimer *timer; |
141 | timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer); | |
142 | qemu_announce_self_once(&timer); | |
a672b469 AL |
143 | } |
144 | ||
145 | /***********************************************************/ | |
146 | /* savevm/loadvm support */ | |
147 | ||
05fcc848 KW |
148 | static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt, |
149 | int64_t pos) | |
150 | { | |
151 | int ret; | |
152 | QEMUIOVector qiov; | |
153 | ||
154 | qemu_iovec_init_external(&qiov, iov, iovcnt); | |
155 | ret = bdrv_writev_vmstate(opaque, &qiov, pos); | |
156 | if (ret < 0) { | |
157 | return ret; | |
158 | } | |
159 | ||
160 | return qiov.size; | |
161 | } | |
162 | ||
a202a4c0 DDAG |
163 | static ssize_t block_put_buffer(void *opaque, const uint8_t *buf, |
164 | int64_t pos, size_t size) | |
a672b469 | 165 | { |
45566e9c | 166 | bdrv_save_vmstate(opaque, buf, pos, size); |
a672b469 AL |
167 | return size; |
168 | } | |
169 | ||
a202a4c0 DDAG |
170 | static ssize_t block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, |
171 | size_t size) | |
a672b469 | 172 | { |
45566e9c | 173 | return bdrv_load_vmstate(opaque, buf, pos, size); |
a672b469 AL |
174 | } |
175 | ||
176 | static int bdrv_fclose(void *opaque) | |
177 | { | |
ad492c92 | 178 | return bdrv_flush(opaque); |
a672b469 AL |
179 | } |
180 | ||
9229bf3c PB |
181 | static const QEMUFileOps bdrv_read_ops = { |
182 | .get_buffer = block_get_buffer, | |
183 | .close = bdrv_fclose | |
184 | }; | |
185 | ||
186 | static const QEMUFileOps bdrv_write_ops = { | |
05fcc848 KW |
187 | .put_buffer = block_put_buffer, |
188 | .writev_buffer = block_writev_buffer, | |
189 | .close = bdrv_fclose | |
9229bf3c PB |
190 | }; |
191 | ||
45566e9c | 192 | static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable) |
a672b469 | 193 | { |
38ff78d3 | 194 | if (is_writable) { |
9229bf3c | 195 | return qemu_fopen_ops(bs, &bdrv_write_ops); |
38ff78d3 | 196 | } |
9229bf3c | 197 | return qemu_fopen_ops(bs, &bdrv_read_ops); |
a672b469 AL |
198 | } |
199 | ||
2ff68d07 | 200 | |
bb1a6d8c EH |
201 | /* QEMUFile timer support. |
202 | * Not in qemu-file.c to not add qemu-timer.c as dependency to qemu-file.c | |
203 | */ | |
2ff68d07 | 204 | |
40daca54 | 205 | void timer_put(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
206 | { |
207 | uint64_t expire_time; | |
208 | ||
e93379b0 | 209 | expire_time = timer_expire_time_ns(ts); |
2ff68d07 PB |
210 | qemu_put_be64(f, expire_time); |
211 | } | |
212 | ||
40daca54 | 213 | void timer_get(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
214 | { |
215 | uint64_t expire_time; | |
216 | ||
217 | expire_time = qemu_get_be64(f); | |
218 | if (expire_time != -1) { | |
bc72ad67 | 219 | timer_mod_ns(ts, expire_time); |
2ff68d07 | 220 | } else { |
bc72ad67 | 221 | timer_del(ts); |
2ff68d07 PB |
222 | } |
223 | } | |
224 | ||
225 | ||
bb1a6d8c EH |
226 | /* VMState timer support. |
227 | * Not in vmstate.c to not add qemu-timer.c as dependency to vmstate.c | |
228 | */ | |
dde0463b JQ |
229 | |
230 | static int get_timer(QEMUFile *f, void *pv, size_t size) | |
231 | { | |
232 | QEMUTimer *v = pv; | |
40daca54 | 233 | timer_get(f, v); |
dde0463b JQ |
234 | return 0; |
235 | } | |
236 | ||
84e2e3eb | 237 | static void put_timer(QEMUFile *f, void *pv, size_t size) |
dde0463b | 238 | { |
84e2e3eb | 239 | QEMUTimer *v = pv; |
40daca54 | 240 | timer_put(f, v); |
dde0463b JQ |
241 | } |
242 | ||
243 | const VMStateInfo vmstate_info_timer = { | |
244 | .name = "timer", | |
245 | .get = get_timer, | |
246 | .put = put_timer, | |
247 | }; | |
248 | ||
08e99e29 | 249 | |
7685ee6a AW |
250 | typedef struct CompatEntry { |
251 | char idstr[256]; | |
252 | int instance_id; | |
253 | } CompatEntry; | |
254 | ||
a672b469 | 255 | typedef struct SaveStateEntry { |
72cf2d4f | 256 | QTAILQ_ENTRY(SaveStateEntry) entry; |
a672b469 AL |
257 | char idstr[256]; |
258 | int instance_id; | |
4d2ffa08 | 259 | int alias_id; |
a672b469 AL |
260 | int version_id; |
261 | int section_id; | |
22ea40f4 | 262 | SaveVMHandlers *ops; |
9ed7d6ae | 263 | const VMStateDescription *vmsd; |
a672b469 | 264 | void *opaque; |
7685ee6a | 265 | CompatEntry *compat; |
a7ae8355 | 266 | int is_ram; |
a672b469 AL |
267 | } SaveStateEntry; |
268 | ||
0163a2e0 JQ |
269 | typedef struct SaveState { |
270 | QTAILQ_HEAD(, SaveStateEntry) handlers; | |
271 | int global_section_id; | |
61964c23 JQ |
272 | bool skip_configuration; |
273 | uint32_t len; | |
274 | const char *name; | |
0163a2e0 JQ |
275 | } SaveState; |
276 | ||
277 | static SaveState savevm_state = { | |
278 | .handlers = QTAILQ_HEAD_INITIALIZER(savevm_state.handlers), | |
279 | .global_section_id = 0, | |
61964c23 JQ |
280 | .skip_configuration = false, |
281 | }; | |
282 | ||
283 | void savevm_skip_configuration(void) | |
284 | { | |
285 | savevm_state.skip_configuration = true; | |
286 | } | |
287 | ||
288 | ||
289 | static void configuration_pre_save(void *opaque) | |
290 | { | |
291 | SaveState *state = opaque; | |
292 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
293 | ||
294 | state->len = strlen(current_name); | |
295 | state->name = current_name; | |
296 | } | |
297 | ||
298 | static int configuration_post_load(void *opaque, int version_id) | |
299 | { | |
300 | SaveState *state = opaque; | |
301 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
302 | ||
303 | if (strncmp(state->name, current_name, state->len) != 0) { | |
15d61692 GK |
304 | error_report("Machine type received is '%.*s' and local is '%s'", |
305 | (int) state->len, state->name, current_name); | |
61964c23 JQ |
306 | return -EINVAL; |
307 | } | |
308 | return 0; | |
309 | } | |
310 | ||
311 | static const VMStateDescription vmstate_configuration = { | |
312 | .name = "configuration", | |
313 | .version_id = 1, | |
314 | .post_load = configuration_post_load, | |
315 | .pre_save = configuration_pre_save, | |
316 | .fields = (VMStateField[]) { | |
317 | VMSTATE_UINT32(len, SaveState), | |
318 | VMSTATE_VBUFFER_ALLOC_UINT32(name, SaveState, 0, NULL, 0, len), | |
319 | VMSTATE_END_OF_LIST() | |
320 | }, | |
0163a2e0 | 321 | }; |
a672b469 | 322 | |
abfd9ce3 AS |
323 | static void dump_vmstate_vmsd(FILE *out_file, |
324 | const VMStateDescription *vmsd, int indent, | |
325 | bool is_subsection); | |
326 | ||
327 | static void dump_vmstate_vmsf(FILE *out_file, const VMStateField *field, | |
328 | int indent) | |
329 | { | |
330 | fprintf(out_file, "%*s{\n", indent, ""); | |
331 | indent += 2; | |
332 | fprintf(out_file, "%*s\"field\": \"%s\",\n", indent, "", field->name); | |
333 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
334 | field->version_id); | |
335 | fprintf(out_file, "%*s\"field_exists\": %s,\n", indent, "", | |
336 | field->field_exists ? "true" : "false"); | |
337 | fprintf(out_file, "%*s\"size\": %zu", indent, "", field->size); | |
338 | if (field->vmsd != NULL) { | |
339 | fprintf(out_file, ",\n"); | |
340 | dump_vmstate_vmsd(out_file, field->vmsd, indent, false); | |
341 | } | |
342 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
343 | } | |
344 | ||
345 | static void dump_vmstate_vmss(FILE *out_file, | |
5cd8cada | 346 | const VMStateDescription **subsection, |
abfd9ce3 AS |
347 | int indent) |
348 | { | |
5cd8cada JQ |
349 | if (*subsection != NULL) { |
350 | dump_vmstate_vmsd(out_file, *subsection, indent, true); | |
abfd9ce3 AS |
351 | } |
352 | } | |
353 | ||
354 | static void dump_vmstate_vmsd(FILE *out_file, | |
355 | const VMStateDescription *vmsd, int indent, | |
356 | bool is_subsection) | |
357 | { | |
358 | if (is_subsection) { | |
359 | fprintf(out_file, "%*s{\n", indent, ""); | |
360 | } else { | |
361 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", "Description"); | |
362 | } | |
363 | indent += 2; | |
364 | fprintf(out_file, "%*s\"name\": \"%s\",\n", indent, "", vmsd->name); | |
365 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
366 | vmsd->version_id); | |
367 | fprintf(out_file, "%*s\"minimum_version_id\": %d", indent, "", | |
368 | vmsd->minimum_version_id); | |
369 | if (vmsd->fields != NULL) { | |
370 | const VMStateField *field = vmsd->fields; | |
371 | bool first; | |
372 | ||
373 | fprintf(out_file, ",\n%*s\"Fields\": [\n", indent, ""); | |
374 | first = true; | |
375 | while (field->name != NULL) { | |
376 | if (field->flags & VMS_MUST_EXIST) { | |
377 | /* Ignore VMSTATE_VALIDATE bits; these don't get migrated */ | |
378 | field++; | |
379 | continue; | |
380 | } | |
381 | if (!first) { | |
382 | fprintf(out_file, ",\n"); | |
383 | } | |
384 | dump_vmstate_vmsf(out_file, field, indent + 2); | |
385 | field++; | |
386 | first = false; | |
387 | } | |
388 | fprintf(out_file, "\n%*s]", indent, ""); | |
389 | } | |
390 | if (vmsd->subsections != NULL) { | |
5cd8cada | 391 | const VMStateDescription **subsection = vmsd->subsections; |
abfd9ce3 AS |
392 | bool first; |
393 | ||
394 | fprintf(out_file, ",\n%*s\"Subsections\": [\n", indent, ""); | |
395 | first = true; | |
5cd8cada | 396 | while (*subsection != NULL) { |
abfd9ce3 AS |
397 | if (!first) { |
398 | fprintf(out_file, ",\n"); | |
399 | } | |
400 | dump_vmstate_vmss(out_file, subsection, indent + 2); | |
401 | subsection++; | |
402 | first = false; | |
403 | } | |
404 | fprintf(out_file, "\n%*s]", indent, ""); | |
405 | } | |
406 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
407 | } | |
408 | ||
409 | static void dump_machine_type(FILE *out_file) | |
410 | { | |
411 | MachineClass *mc; | |
412 | ||
413 | mc = MACHINE_GET_CLASS(current_machine); | |
414 | ||
415 | fprintf(out_file, " \"vmschkmachine\": {\n"); | |
416 | fprintf(out_file, " \"Name\": \"%s\"\n", mc->name); | |
417 | fprintf(out_file, " },\n"); | |
418 | } | |
419 | ||
420 | void dump_vmstate_json_to_file(FILE *out_file) | |
421 | { | |
422 | GSList *list, *elt; | |
423 | bool first; | |
424 | ||
425 | fprintf(out_file, "{\n"); | |
426 | dump_machine_type(out_file); | |
427 | ||
428 | first = true; | |
429 | list = object_class_get_list(TYPE_DEVICE, true); | |
430 | for (elt = list; elt; elt = elt->next) { | |
431 | DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, elt->data, | |
432 | TYPE_DEVICE); | |
433 | const char *name; | |
434 | int indent = 2; | |
435 | ||
436 | if (!dc->vmsd) { | |
437 | continue; | |
438 | } | |
439 | ||
440 | if (!first) { | |
441 | fprintf(out_file, ",\n"); | |
442 | } | |
443 | name = object_class_get_name(OBJECT_CLASS(dc)); | |
444 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", name); | |
445 | indent += 2; | |
446 | fprintf(out_file, "%*s\"Name\": \"%s\",\n", indent, "", name); | |
447 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
448 | dc->vmsd->version_id); | |
449 | fprintf(out_file, "%*s\"minimum_version_id\": %d,\n", indent, "", | |
450 | dc->vmsd->minimum_version_id); | |
451 | ||
452 | dump_vmstate_vmsd(out_file, dc->vmsd, indent, false); | |
453 | ||
454 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
455 | first = false; | |
456 | } | |
457 | fprintf(out_file, "\n}\n"); | |
458 | fclose(out_file); | |
459 | } | |
460 | ||
8718e999 JQ |
461 | static int calculate_new_instance_id(const char *idstr) |
462 | { | |
463 | SaveStateEntry *se; | |
464 | int instance_id = 0; | |
465 | ||
0163a2e0 | 466 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
8718e999 JQ |
467 | if (strcmp(idstr, se->idstr) == 0 |
468 | && instance_id <= se->instance_id) { | |
469 | instance_id = se->instance_id + 1; | |
470 | } | |
471 | } | |
472 | return instance_id; | |
473 | } | |
474 | ||
7685ee6a AW |
475 | static int calculate_compat_instance_id(const char *idstr) |
476 | { | |
477 | SaveStateEntry *se; | |
478 | int instance_id = 0; | |
479 | ||
0163a2e0 | 480 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
38ff78d3 | 481 | if (!se->compat) { |
7685ee6a | 482 | continue; |
38ff78d3 | 483 | } |
7685ee6a AW |
484 | |
485 | if (strcmp(idstr, se->compat->idstr) == 0 | |
486 | && instance_id <= se->compat->instance_id) { | |
487 | instance_id = se->compat->instance_id + 1; | |
488 | } | |
489 | } | |
490 | return instance_id; | |
491 | } | |
492 | ||
a672b469 AL |
493 | /* TODO: Individual devices generally have very little idea about the rest |
494 | of the system, so instance_id should be removed/replaced. | |
495 | Meanwhile pass -1 as instance_id if you do not already have a clearly | |
496 | distinguishing id for all instances of your device class. */ | |
0be71e32 AW |
497 | int register_savevm_live(DeviceState *dev, |
498 | const char *idstr, | |
a672b469 AL |
499 | int instance_id, |
500 | int version_id, | |
7908c78d | 501 | SaveVMHandlers *ops, |
a672b469 AL |
502 | void *opaque) |
503 | { | |
8718e999 | 504 | SaveStateEntry *se; |
a672b469 | 505 | |
97f3ad35 | 506 | se = g_new0(SaveStateEntry, 1); |
a672b469 | 507 | se->version_id = version_id; |
0163a2e0 | 508 | se->section_id = savevm_state.global_section_id++; |
7908c78d | 509 | se->ops = ops; |
a672b469 | 510 | se->opaque = opaque; |
9ed7d6ae | 511 | se->vmsd = NULL; |
a7ae8355 | 512 | /* if this is a live_savem then set is_ram */ |
16310a3c | 513 | if (ops->save_live_setup != NULL) { |
a7ae8355 SS |
514 | se->is_ram = 1; |
515 | } | |
a672b469 | 516 | |
09e5ab63 AL |
517 | if (dev) { |
518 | char *id = qdev_get_dev_path(dev); | |
7685ee6a AW |
519 | if (id) { |
520 | pstrcpy(se->idstr, sizeof(se->idstr), id); | |
521 | pstrcat(se->idstr, sizeof(se->idstr), "/"); | |
7267c094 | 522 | g_free(id); |
7685ee6a | 523 | |
97f3ad35 | 524 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
525 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr); |
526 | se->compat->instance_id = instance_id == -1 ? | |
527 | calculate_compat_instance_id(idstr) : instance_id; | |
528 | instance_id = -1; | |
529 | } | |
530 | } | |
531 | pstrcat(se->idstr, sizeof(se->idstr), idstr); | |
532 | ||
8718e999 | 533 | if (instance_id == -1) { |
7685ee6a | 534 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
535 | } else { |
536 | se->instance_id = instance_id; | |
a672b469 | 537 | } |
7685ee6a | 538 | assert(!se->compat || se->instance_id == 0); |
8718e999 | 539 | /* add at the end of list */ |
0163a2e0 | 540 | QTAILQ_INSERT_TAIL(&savevm_state.handlers, se, entry); |
a672b469 AL |
541 | return 0; |
542 | } | |
543 | ||
0be71e32 AW |
544 | int register_savevm(DeviceState *dev, |
545 | const char *idstr, | |
a672b469 AL |
546 | int instance_id, |
547 | int version_id, | |
548 | SaveStateHandler *save_state, | |
549 | LoadStateHandler *load_state, | |
550 | void *opaque) | |
551 | { | |
97f3ad35 | 552 | SaveVMHandlers *ops = g_new0(SaveVMHandlers, 1); |
7908c78d JQ |
553 | ops->save_state = save_state; |
554 | ops->load_state = load_state; | |
0be71e32 | 555 | return register_savevm_live(dev, idstr, instance_id, version_id, |
7908c78d | 556 | ops, opaque); |
a672b469 AL |
557 | } |
558 | ||
0be71e32 | 559 | void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque) |
41bd13af | 560 | { |
8718e999 | 561 | SaveStateEntry *se, *new_se; |
7685ee6a AW |
562 | char id[256] = ""; |
563 | ||
09e5ab63 AL |
564 | if (dev) { |
565 | char *path = qdev_get_dev_path(dev); | |
7685ee6a AW |
566 | if (path) { |
567 | pstrcpy(id, sizeof(id), path); | |
568 | pstrcat(id, sizeof(id), "/"); | |
7267c094 | 569 | g_free(path); |
7685ee6a AW |
570 | } |
571 | } | |
572 | pstrcat(id, sizeof(id), idstr); | |
41bd13af | 573 | |
0163a2e0 | 574 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
7685ee6a | 575 | if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) { |
0163a2e0 | 576 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 577 | g_free(se->compat); |
22ea40f4 | 578 | g_free(se->ops); |
7267c094 | 579 | g_free(se); |
41bd13af | 580 | } |
41bd13af AL |
581 | } |
582 | } | |
583 | ||
0be71e32 | 584 | int vmstate_register_with_alias_id(DeviceState *dev, int instance_id, |
4d2ffa08 JK |
585 | const VMStateDescription *vmsd, |
586 | void *opaque, int alias_id, | |
587 | int required_for_version) | |
9ed7d6ae | 588 | { |
8718e999 | 589 | SaveStateEntry *se; |
9ed7d6ae | 590 | |
4d2ffa08 JK |
591 | /* If this triggers, alias support can be dropped for the vmsd. */ |
592 | assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id); | |
593 | ||
97f3ad35 | 594 | se = g_new0(SaveStateEntry, 1); |
9ed7d6ae | 595 | se->version_id = vmsd->version_id; |
0163a2e0 | 596 | se->section_id = savevm_state.global_section_id++; |
9ed7d6ae JQ |
597 | se->opaque = opaque; |
598 | se->vmsd = vmsd; | |
4d2ffa08 | 599 | se->alias_id = alias_id; |
9ed7d6ae | 600 | |
09e5ab63 AL |
601 | if (dev) { |
602 | char *id = qdev_get_dev_path(dev); | |
7685ee6a AW |
603 | if (id) { |
604 | pstrcpy(se->idstr, sizeof(se->idstr), id); | |
605 | pstrcat(se->idstr, sizeof(se->idstr), "/"); | |
7267c094 | 606 | g_free(id); |
7685ee6a | 607 | |
97f3ad35 | 608 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
609 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name); |
610 | se->compat->instance_id = instance_id == -1 ? | |
611 | calculate_compat_instance_id(vmsd->name) : instance_id; | |
612 | instance_id = -1; | |
613 | } | |
614 | } | |
615 | pstrcat(se->idstr, sizeof(se->idstr), vmsd->name); | |
616 | ||
8718e999 | 617 | if (instance_id == -1) { |
7685ee6a | 618 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
619 | } else { |
620 | se->instance_id = instance_id; | |
9ed7d6ae | 621 | } |
7685ee6a | 622 | assert(!se->compat || se->instance_id == 0); |
8718e999 | 623 | /* add at the end of list */ |
0163a2e0 | 624 | QTAILQ_INSERT_TAIL(&savevm_state.handlers, se, entry); |
9ed7d6ae JQ |
625 | return 0; |
626 | } | |
627 | ||
0be71e32 AW |
628 | void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd, |
629 | void *opaque) | |
9ed7d6ae | 630 | { |
1eb7538b JQ |
631 | SaveStateEntry *se, *new_se; |
632 | ||
0163a2e0 | 633 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
1eb7538b | 634 | if (se->vmsd == vmsd && se->opaque == opaque) { |
0163a2e0 | 635 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 636 | g_free(se->compat); |
7267c094 | 637 | g_free(se); |
1eb7538b JQ |
638 | } |
639 | } | |
9ed7d6ae JQ |
640 | } |
641 | ||
4082be4d JQ |
642 | static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id) |
643 | { | |
9013dca5 | 644 | trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
9ed7d6ae | 645 | if (!se->vmsd) { /* Old style */ |
22ea40f4 | 646 | return se->ops->load_state(f, se->opaque, version_id); |
9ed7d6ae JQ |
647 | } |
648 | return vmstate_load_state(f, se->vmsd, se->opaque, version_id); | |
4082be4d JQ |
649 | } |
650 | ||
8118f095 AG |
651 | static void vmstate_save_old_style(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) |
652 | { | |
653 | int64_t old_offset, size; | |
654 | ||
655 | old_offset = qemu_ftell_fast(f); | |
656 | se->ops->save_state(f, se->opaque); | |
657 | size = qemu_ftell_fast(f) - old_offset; | |
658 | ||
659 | if (vmdesc) { | |
660 | json_prop_int(vmdesc, "size", size); | |
661 | json_start_array(vmdesc, "fields"); | |
662 | json_start_object(vmdesc, NULL); | |
663 | json_prop_str(vmdesc, "name", "data"); | |
664 | json_prop_int(vmdesc, "size", size); | |
665 | json_prop_str(vmdesc, "type", "buffer"); | |
666 | json_end_object(vmdesc); | |
667 | json_end_array(vmdesc); | |
668 | } | |
669 | } | |
670 | ||
671 | static void vmstate_save(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) | |
4082be4d | 672 | { |
9013dca5 | 673 | trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
8118f095 AG |
674 | if (!se->vmsd) { |
675 | vmstate_save_old_style(f, se, vmdesc); | |
dc912121 | 676 | return; |
9ed7d6ae | 677 | } |
8118f095 | 678 | vmstate_save_state(f, se->vmsd, se->opaque, vmdesc); |
4082be4d JQ |
679 | } |
680 | ||
37fb569c DDAG |
681 | void savevm_skip_section_footers(void) |
682 | { | |
683 | skip_section_footers = true; | |
684 | } | |
685 | ||
ce39bfc9 DDAG |
686 | /* |
687 | * Write the header for device section (QEMU_VM_SECTION START/END/PART/FULL) | |
688 | */ | |
689 | static void save_section_header(QEMUFile *f, SaveStateEntry *se, | |
690 | uint8_t section_type) | |
691 | { | |
692 | qemu_put_byte(f, section_type); | |
693 | qemu_put_be32(f, se->section_id); | |
694 | ||
695 | if (section_type == QEMU_VM_SECTION_FULL || | |
696 | section_type == QEMU_VM_SECTION_START) { | |
697 | /* ID string */ | |
698 | size_t len = strlen(se->idstr); | |
699 | qemu_put_byte(f, len); | |
700 | qemu_put_buffer(f, (uint8_t *)se->idstr, len); | |
701 | ||
702 | qemu_put_be32(f, se->instance_id); | |
703 | qemu_put_be32(f, se->version_id); | |
704 | } | |
705 | } | |
706 | ||
f68945d4 DDAG |
707 | /* |
708 | * Write a footer onto device sections that catches cases misformatted device | |
709 | * sections. | |
710 | */ | |
711 | static void save_section_footer(QEMUFile *f, SaveStateEntry *se) | |
712 | { | |
713 | if (!skip_section_footers) { | |
714 | qemu_put_byte(f, QEMU_VM_SECTION_FOOTER); | |
715 | qemu_put_be32(f, se->section_id); | |
716 | } | |
717 | } | |
718 | ||
c76ca188 DDAG |
719 | /** |
720 | * qemu_savevm_command_send: Send a 'QEMU_VM_COMMAND' type element with the | |
721 | * command and associated data. | |
722 | * | |
723 | * @f: File to send command on | |
724 | * @command: Command type to send | |
725 | * @len: Length of associated data | |
726 | * @data: Data associated with command. | |
727 | */ | |
728 | void qemu_savevm_command_send(QEMUFile *f, | |
729 | enum qemu_vm_cmd command, | |
730 | uint16_t len, | |
731 | uint8_t *data) | |
732 | { | |
733 | trace_savevm_command_send(command, len); | |
734 | qemu_put_byte(f, QEMU_VM_COMMAND); | |
735 | qemu_put_be16(f, (uint16_t)command); | |
736 | qemu_put_be16(f, len); | |
737 | qemu_put_buffer(f, data, len); | |
738 | qemu_fflush(f); | |
739 | } | |
740 | ||
2e37701e DDAG |
741 | void qemu_savevm_send_ping(QEMUFile *f, uint32_t value) |
742 | { | |
743 | uint32_t buf; | |
744 | ||
745 | trace_savevm_send_ping(value); | |
746 | buf = cpu_to_be32(value); | |
747 | qemu_savevm_command_send(f, MIG_CMD_PING, sizeof(value), (uint8_t *)&buf); | |
748 | } | |
749 | ||
750 | void qemu_savevm_send_open_return_path(QEMUFile *f) | |
751 | { | |
752 | trace_savevm_send_open_return_path(); | |
753 | qemu_savevm_command_send(f, MIG_CMD_OPEN_RETURN_PATH, 0, NULL); | |
754 | } | |
755 | ||
11cf1d98 DDAG |
756 | /* We have a buffer of data to send; we don't want that all to be loaded |
757 | * by the command itself, so the command contains just the length of the | |
758 | * extra buffer that we then send straight after it. | |
759 | * TODO: Must be a better way to organise that | |
760 | * | |
761 | * Returns: | |
762 | * 0 on success | |
763 | * -ve on error | |
764 | */ | |
61b67d47 | 765 | int qemu_savevm_send_packaged(QEMUFile *f, const uint8_t *buf, size_t len) |
11cf1d98 | 766 | { |
11cf1d98 DDAG |
767 | uint32_t tmp; |
768 | ||
769 | if (len > MAX_VM_CMD_PACKAGED_SIZE) { | |
770 | error_report("%s: Unreasonably large packaged state: %zu", | |
771 | __func__, len); | |
772 | return -1; | |
773 | } | |
774 | ||
775 | tmp = cpu_to_be32(len); | |
776 | ||
777 | trace_qemu_savevm_send_packaged(); | |
778 | qemu_savevm_command_send(f, MIG_CMD_PACKAGED, 4, (uint8_t *)&tmp); | |
779 | ||
61b67d47 | 780 | qemu_put_buffer(f, buf, len); |
11cf1d98 DDAG |
781 | |
782 | return 0; | |
783 | } | |
784 | ||
093e3c42 DDAG |
785 | /* Send prior to any postcopy transfer */ |
786 | void qemu_savevm_send_postcopy_advise(QEMUFile *f) | |
787 | { | |
788 | uint64_t tmp[2]; | |
789 | tmp[0] = cpu_to_be64(getpagesize()); | |
790 | tmp[1] = cpu_to_be64(1ul << qemu_target_page_bits()); | |
791 | ||
792 | trace_qemu_savevm_send_postcopy_advise(); | |
793 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_ADVISE, 16, (uint8_t *)tmp); | |
794 | } | |
795 | ||
796 | /* Sent prior to starting the destination running in postcopy, discard pages | |
797 | * that have already been sent but redirtied on the source. | |
798 | * CMD_POSTCOPY_RAM_DISCARD consist of: | |
799 | * byte version (0) | |
800 | * byte Length of name field (not including 0) | |
801 | * n x byte RAM block name | |
802 | * byte 0 terminator (just for safety) | |
803 | * n x Byte ranges within the named RAMBlock | |
804 | * be64 Start of the range | |
805 | * be64 Length | |
806 | * | |
807 | * name: RAMBlock name that these entries are part of | |
808 | * len: Number of page entries | |
809 | * start_list: 'len' addresses | |
810 | * length_list: 'len' addresses | |
811 | * | |
812 | */ | |
813 | void qemu_savevm_send_postcopy_ram_discard(QEMUFile *f, const char *name, | |
814 | uint16_t len, | |
815 | uint64_t *start_list, | |
816 | uint64_t *length_list) | |
817 | { | |
818 | uint8_t *buf; | |
819 | uint16_t tmplen; | |
820 | uint16_t t; | |
821 | size_t name_len = strlen(name); | |
822 | ||
823 | trace_qemu_savevm_send_postcopy_ram_discard(name, len); | |
824 | assert(name_len < 256); | |
825 | buf = g_malloc0(1 + 1 + name_len + 1 + (8 + 8) * len); | |
826 | buf[0] = postcopy_ram_discard_version; | |
827 | buf[1] = name_len; | |
828 | memcpy(buf + 2, name, name_len); | |
829 | tmplen = 2 + name_len; | |
830 | buf[tmplen++] = '\0'; | |
831 | ||
832 | for (t = 0; t < len; t++) { | |
833 | cpu_to_be64w((uint64_t *)(buf + tmplen), start_list[t]); | |
834 | tmplen += 8; | |
835 | cpu_to_be64w((uint64_t *)(buf + tmplen), length_list[t]); | |
836 | tmplen += 8; | |
837 | } | |
838 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RAM_DISCARD, tmplen, buf); | |
839 | g_free(buf); | |
840 | } | |
841 | ||
842 | /* Get the destination into a state where it can receive postcopy data. */ | |
843 | void qemu_savevm_send_postcopy_listen(QEMUFile *f) | |
844 | { | |
845 | trace_savevm_send_postcopy_listen(); | |
846 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_LISTEN, 0, NULL); | |
847 | } | |
848 | ||
849 | /* Kick the destination into running */ | |
850 | void qemu_savevm_send_postcopy_run(QEMUFile *f) | |
851 | { | |
852 | trace_savevm_send_postcopy_run(); | |
853 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RUN, 0, NULL); | |
854 | } | |
855 | ||
e1c37d0e | 856 | bool qemu_savevm_state_blocked(Error **errp) |
dc912121 AW |
857 | { |
858 | SaveStateEntry *se; | |
859 | ||
0163a2e0 | 860 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
7d854c47 | 861 | if (se->vmsd && se->vmsd->unmigratable) { |
f231b88d CR |
862 | error_setg(errp, "State blocked by non-migratable device '%s'", |
863 | se->idstr); | |
dc912121 AW |
864 | return true; |
865 | } | |
866 | } | |
867 | return false; | |
868 | } | |
869 | ||
902c053d GK |
870 | static bool enforce_config_section(void) |
871 | { | |
872 | MachineState *machine = MACHINE(qdev_get_machine()); | |
873 | return machine->enforce_config_section; | |
874 | } | |
875 | ||
f796baa1 DDAG |
876 | void qemu_savevm_state_header(QEMUFile *f) |
877 | { | |
878 | trace_savevm_state_header(); | |
879 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
880 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
172dfd4f | 881 | |
902c053d | 882 | if (!savevm_state.skip_configuration || enforce_config_section()) { |
172dfd4f DDAG |
883 | qemu_put_byte(f, QEMU_VM_CONFIGURATION); |
884 | vmstate_save_state(f, &vmstate_configuration, &savevm_state, 0); | |
885 | } | |
886 | ||
f796baa1 DDAG |
887 | } |
888 | ||
47c8c17a PB |
889 | void qemu_savevm_state_begin(QEMUFile *f, |
890 | const MigrationParams *params) | |
a672b469 AL |
891 | { |
892 | SaveStateEntry *se; | |
39346385 | 893 | int ret; |
a672b469 | 894 | |
9013dca5 | 895 | trace_savevm_state_begin(); |
0163a2e0 | 896 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
22ea40f4 | 897 | if (!se->ops || !se->ops->set_params) { |
c163b5ca | 898 | continue; |
6607ae23 | 899 | } |
22ea40f4 | 900 | se->ops->set_params(params, se->opaque); |
c163b5ca | 901 | } |
38ff78d3 | 902 | |
0163a2e0 | 903 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
d1315aac | 904 | if (!se->ops || !se->ops->save_live_setup) { |
a672b469 | 905 | continue; |
22ea40f4 | 906 | } |
6bd68781 JQ |
907 | if (se->ops && se->ops->is_active) { |
908 | if (!se->ops->is_active(se->opaque)) { | |
909 | continue; | |
910 | } | |
911 | } | |
ce39bfc9 | 912 | save_section_header(f, se, QEMU_VM_SECTION_START); |
a672b469 | 913 | |
d1315aac | 914 | ret = se->ops->save_live_setup(f, se->opaque); |
f68945d4 | 915 | save_section_footer(f, se); |
2975725f | 916 | if (ret < 0) { |
47c8c17a PB |
917 | qemu_file_set_error(f, ret); |
918 | break; | |
2975725f | 919 | } |
a672b469 | 920 | } |
a672b469 AL |
921 | } |
922 | ||
39346385 | 923 | /* |
07f35073 | 924 | * this function has three return values: |
39346385 JQ |
925 | * negative: there was one error, and we have -errno. |
926 | * 0 : We haven't finished, caller have to go again | |
927 | * 1 : We have finished, we can go to complete phase | |
928 | */ | |
35ecd943 | 929 | int qemu_savevm_state_iterate(QEMUFile *f, bool postcopy) |
a672b469 AL |
930 | { |
931 | SaveStateEntry *se; | |
932 | int ret = 1; | |
933 | ||
9013dca5 | 934 | trace_savevm_state_iterate(); |
0163a2e0 | 935 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
16310a3c | 936 | if (!se->ops || !se->ops->save_live_iterate) { |
a672b469 | 937 | continue; |
22ea40f4 | 938 | } |
6bd68781 JQ |
939 | if (se->ops && se->ops->is_active) { |
940 | if (!se->ops->is_active(se->opaque)) { | |
941 | continue; | |
942 | } | |
943 | } | |
35ecd943 DDAG |
944 | /* |
945 | * In the postcopy phase, any device that doesn't know how to | |
946 | * do postcopy should have saved it's state in the _complete | |
947 | * call that's already run, it might get confused if we call | |
948 | * iterate afterwards. | |
949 | */ | |
950 | if (postcopy && !se->ops->save_live_complete_postcopy) { | |
951 | continue; | |
952 | } | |
aac844ed JQ |
953 | if (qemu_file_rate_limit(f)) { |
954 | return 0; | |
955 | } | |
464400f6 | 956 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
957 | |
958 | save_section_header(f, se, QEMU_VM_SECTION_PART); | |
a672b469 | 959 | |
16310a3c | 960 | ret = se->ops->save_live_iterate(f, se->opaque); |
a5df2a02 | 961 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 962 | save_section_footer(f, se); |
517a13c9 | 963 | |
47c8c17a PB |
964 | if (ret < 0) { |
965 | qemu_file_set_error(f, ret); | |
966 | } | |
2975725f | 967 | if (ret <= 0) { |
90697be8 JK |
968 | /* Do not proceed to the next vmstate before this one reported |
969 | completion of the current stage. This serializes the migration | |
970 | and reduces the probability that a faster changing state is | |
971 | synchronized over and over again. */ | |
972 | break; | |
973 | } | |
a672b469 | 974 | } |
39346385 | 975 | return ret; |
a672b469 AL |
976 | } |
977 | ||
9850c604 AG |
978 | static bool should_send_vmdesc(void) |
979 | { | |
980 | MachineState *machine = MACHINE(qdev_get_machine()); | |
8421b205 DDAG |
981 | bool in_postcopy = migration_in_postcopy(migrate_get_current()); |
982 | return !machine->suppress_vmdesc && !in_postcopy; | |
9850c604 AG |
983 | } |
984 | ||
763c906b DDAG |
985 | /* |
986 | * Calls the save_live_complete_postcopy methods | |
987 | * causing the last few pages to be sent immediately and doing any associated | |
988 | * cleanup. | |
989 | * Note postcopy also calls qemu_savevm_state_complete_precopy to complete | |
990 | * all the other devices, but that happens at the point we switch to postcopy. | |
991 | */ | |
992 | void qemu_savevm_state_complete_postcopy(QEMUFile *f) | |
993 | { | |
994 | SaveStateEntry *se; | |
995 | int ret; | |
996 | ||
997 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
998 | if (!se->ops || !se->ops->save_live_complete_postcopy) { | |
999 | continue; | |
1000 | } | |
1001 | if (se->ops && se->ops->is_active) { | |
1002 | if (!se->ops->is_active(se->opaque)) { | |
1003 | continue; | |
1004 | } | |
1005 | } | |
1006 | trace_savevm_section_start(se->idstr, se->section_id); | |
1007 | /* Section type */ | |
1008 | qemu_put_byte(f, QEMU_VM_SECTION_END); | |
1009 | qemu_put_be32(f, se->section_id); | |
1010 | ||
1011 | ret = se->ops->save_live_complete_postcopy(f, se->opaque); | |
1012 | trace_savevm_section_end(se->idstr, se->section_id, ret); | |
1013 | save_section_footer(f, se); | |
1014 | if (ret < 0) { | |
1015 | qemu_file_set_error(f, ret); | |
1016 | return; | |
1017 | } | |
1018 | } | |
1019 | ||
1020 | qemu_put_byte(f, QEMU_VM_EOF); | |
1021 | qemu_fflush(f); | |
1022 | } | |
1023 | ||
1c0d249d | 1024 | void qemu_savevm_state_complete_precopy(QEMUFile *f, bool iterable_only) |
a672b469 | 1025 | { |
8118f095 AG |
1026 | QJSON *vmdesc; |
1027 | int vmdesc_len; | |
a672b469 | 1028 | SaveStateEntry *se; |
2975725f | 1029 | int ret; |
763c906b | 1030 | bool in_postcopy = migration_in_postcopy(migrate_get_current()); |
a672b469 | 1031 | |
a3e06c3d | 1032 | trace_savevm_state_complete_precopy(); |
9013dca5 | 1033 | |
ea375f9a JK |
1034 | cpu_synchronize_all_states(); |
1035 | ||
0163a2e0 | 1036 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
763c906b DDAG |
1037 | if (!se->ops || |
1038 | (in_postcopy && se->ops->save_live_complete_postcopy) || | |
1c0d249d | 1039 | (in_postcopy && !iterable_only) || |
763c906b | 1040 | !se->ops->save_live_complete_precopy) { |
a672b469 | 1041 | continue; |
22ea40f4 | 1042 | } |
1c0d249d | 1043 | |
6bd68781 JQ |
1044 | if (se->ops && se->ops->is_active) { |
1045 | if (!se->ops->is_active(se->opaque)) { | |
1046 | continue; | |
1047 | } | |
1048 | } | |
464400f6 | 1049 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
1050 | |
1051 | save_section_header(f, se, QEMU_VM_SECTION_END); | |
a672b469 | 1052 | |
a3e06c3d | 1053 | ret = se->ops->save_live_complete_precopy(f, se->opaque); |
a5df2a02 | 1054 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 1055 | save_section_footer(f, se); |
2975725f | 1056 | if (ret < 0) { |
47c8c17a PB |
1057 | qemu_file_set_error(f, ret); |
1058 | return; | |
2975725f | 1059 | } |
a672b469 AL |
1060 | } |
1061 | ||
1c0d249d DDAG |
1062 | if (iterable_only) { |
1063 | return; | |
1064 | } | |
1065 | ||
8118f095 AG |
1066 | vmdesc = qjson_new(); |
1067 | json_prop_int(vmdesc, "page_size", TARGET_PAGE_SIZE); | |
1068 | json_start_array(vmdesc, "devices"); | |
0163a2e0 | 1069 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1070 | |
22ea40f4 | 1071 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
5cecf414 | 1072 | continue; |
22ea40f4 | 1073 | } |
df896152 JQ |
1074 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1075 | trace_savevm_section_skip(se->idstr, se->section_id); | |
1076 | continue; | |
1077 | } | |
1078 | ||
464400f6 | 1079 | trace_savevm_section_start(se->idstr, se->section_id); |
8118f095 AG |
1080 | |
1081 | json_start_object(vmdesc, NULL); | |
1082 | json_prop_str(vmdesc, "name", se->idstr); | |
1083 | json_prop_int(vmdesc, "instance_id", se->instance_id); | |
1084 | ||
ce39bfc9 | 1085 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
8118f095 | 1086 | vmstate_save(f, se, vmdesc); |
a5df2a02 | 1087 | trace_savevm_section_end(se->idstr, se->section_id, 0); |
f68945d4 | 1088 | save_section_footer(f, se); |
bdf46d64 WY |
1089 | |
1090 | json_end_object(vmdesc); | |
a672b469 AL |
1091 | } |
1092 | ||
763c906b DDAG |
1093 | if (!in_postcopy) { |
1094 | /* Postcopy stream will still be going */ | |
1095 | qemu_put_byte(f, QEMU_VM_EOF); | |
1096 | } | |
8118f095 AG |
1097 | |
1098 | json_end_array(vmdesc); | |
1099 | qjson_finish(vmdesc); | |
1100 | vmdesc_len = strlen(qjson_get_str(vmdesc)); | |
1101 | ||
9850c604 AG |
1102 | if (should_send_vmdesc()) { |
1103 | qemu_put_byte(f, QEMU_VM_VMDESCRIPTION); | |
1104 | qemu_put_be32(f, vmdesc_len); | |
1105 | qemu_put_buffer(f, (uint8_t *)qjson_get_str(vmdesc), vmdesc_len); | |
1106 | } | |
b72fe9e6 | 1107 | qjson_destroy(vmdesc); |
8118f095 | 1108 | |
edaae611 | 1109 | qemu_fflush(f); |
a672b469 AL |
1110 | } |
1111 | ||
c31b098f DDAG |
1112 | /* Give an estimate of the amount left to be transferred, |
1113 | * the result is split into the amount for units that can and | |
1114 | * for units that can't do postcopy. | |
1115 | */ | |
1116 | void qemu_savevm_state_pending(QEMUFile *f, uint64_t max_size, | |
1117 | uint64_t *res_non_postcopiable, | |
1118 | uint64_t *res_postcopiable) | |
e4ed1541 JQ |
1119 | { |
1120 | SaveStateEntry *se; | |
c31b098f DDAG |
1121 | |
1122 | *res_non_postcopiable = 0; | |
1123 | *res_postcopiable = 0; | |
1124 | ||
e4ed1541 | 1125 | |
0163a2e0 | 1126 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
e4ed1541 JQ |
1127 | if (!se->ops || !se->ops->save_live_pending) { |
1128 | continue; | |
1129 | } | |
1130 | if (se->ops && se->ops->is_active) { | |
1131 | if (!se->ops->is_active(se->opaque)) { | |
1132 | continue; | |
1133 | } | |
1134 | } | |
c31b098f DDAG |
1135 | se->ops->save_live_pending(f, se->opaque, max_size, |
1136 | res_non_postcopiable, res_postcopiable); | |
e4ed1541 | 1137 | } |
e4ed1541 JQ |
1138 | } |
1139 | ||
ea7415fa | 1140 | void qemu_savevm_state_cleanup(void) |
4ec7fcc7 JK |
1141 | { |
1142 | SaveStateEntry *se; | |
1143 | ||
ea7415fa | 1144 | trace_savevm_state_cleanup(); |
0163a2e0 | 1145 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
d1a8548c LL |
1146 | if (se->ops && se->ops->cleanup) { |
1147 | se->ops->cleanup(se->opaque); | |
4ec7fcc7 JK |
1148 | } |
1149 | } | |
1150 | } | |
1151 | ||
5d80448c | 1152 | static int qemu_savevm_state(QEMUFile *f, Error **errp) |
a672b469 | 1153 | { |
a672b469 | 1154 | int ret; |
6607ae23 IY |
1155 | MigrationParams params = { |
1156 | .blk = 0, | |
1157 | .shared = 0 | |
1158 | }; | |
aefeb18b | 1159 | MigrationState *ms = migrate_init(¶ms); |
89a02a9f | 1160 | ms->to_dst_file = f; |
a672b469 | 1161 | |
24f3902b | 1162 | if (migration_is_blocked(errp)) { |
04943eba | 1163 | return -EINVAL; |
dc912121 AW |
1164 | } |
1165 | ||
9b095037 | 1166 | qemu_mutex_unlock_iothread(); |
f796baa1 | 1167 | qemu_savevm_state_header(f); |
47c8c17a | 1168 | qemu_savevm_state_begin(f, ¶ms); |
9b095037 PB |
1169 | qemu_mutex_lock_iothread(); |
1170 | ||
47c8c17a | 1171 | while (qemu_file_get_error(f) == 0) { |
35ecd943 | 1172 | if (qemu_savevm_state_iterate(f, false) > 0) { |
47c8c17a PB |
1173 | break; |
1174 | } | |
1175 | } | |
a672b469 | 1176 | |
47c8c17a | 1177 | ret = qemu_file_get_error(f); |
39346385 | 1178 | if (ret == 0) { |
1c0d249d | 1179 | qemu_savevm_state_complete_precopy(f, false); |
624b9cc2 | 1180 | ret = qemu_file_get_error(f); |
39346385 | 1181 | } |
15b3b8ea | 1182 | qemu_savevm_state_cleanup(); |
04943eba | 1183 | if (ret != 0) { |
5d80448c | 1184 | error_setg_errno(errp, -ret, "Error while writing VM state"); |
04943eba | 1185 | } |
a672b469 AL |
1186 | return ret; |
1187 | } | |
1188 | ||
a7ae8355 SS |
1189 | static int qemu_save_device_state(QEMUFile *f) |
1190 | { | |
1191 | SaveStateEntry *se; | |
1192 | ||
1193 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
1194 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
1195 | ||
1196 | cpu_synchronize_all_states(); | |
1197 | ||
0163a2e0 | 1198 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a7ae8355 SS |
1199 | if (se->is_ram) { |
1200 | continue; | |
1201 | } | |
22ea40f4 | 1202 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
a7ae8355 SS |
1203 | continue; |
1204 | } | |
df896152 JQ |
1205 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1206 | continue; | |
1207 | } | |
a7ae8355 | 1208 | |
ce39bfc9 | 1209 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
a7ae8355 | 1210 | |
8118f095 | 1211 | vmstate_save(f, se, NULL); |
f68945d4 DDAG |
1212 | |
1213 | save_section_footer(f, se); | |
a7ae8355 SS |
1214 | } |
1215 | ||
1216 | qemu_put_byte(f, QEMU_VM_EOF); | |
1217 | ||
1218 | return qemu_file_get_error(f); | |
1219 | } | |
1220 | ||
a672b469 AL |
1221 | static SaveStateEntry *find_se(const char *idstr, int instance_id) |
1222 | { | |
1223 | SaveStateEntry *se; | |
1224 | ||
0163a2e0 | 1225 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1226 | if (!strcmp(se->idstr, idstr) && |
4d2ffa08 JK |
1227 | (instance_id == se->instance_id || |
1228 | instance_id == se->alias_id)) | |
a672b469 | 1229 | return se; |
7685ee6a AW |
1230 | /* Migrating from an older version? */ |
1231 | if (strstr(se->idstr, idstr) && se->compat) { | |
1232 | if (!strcmp(se->compat->idstr, idstr) && | |
1233 | (instance_id == se->compat->instance_id || | |
1234 | instance_id == se->alias_id)) | |
1235 | return se; | |
1236 | } | |
a672b469 AL |
1237 | } |
1238 | return NULL; | |
1239 | } | |
1240 | ||
7b89bf27 DDAG |
1241 | enum LoadVMExitCodes { |
1242 | /* Allow a command to quit all layers of nested loadvm loops */ | |
1243 | LOADVM_QUIT = 1, | |
1244 | }; | |
1245 | ||
1246 | static int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis); | |
093e3c42 DDAG |
1247 | |
1248 | /* ------ incoming postcopy messages ------ */ | |
1249 | /* 'advise' arrives before any transfers just to tell us that a postcopy | |
1250 | * *might* happen - it might be skipped if precopy transferred everything | |
1251 | * quickly. | |
1252 | */ | |
1253 | static int loadvm_postcopy_handle_advise(MigrationIncomingState *mis) | |
1254 | { | |
1255 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
1256 | uint64_t remote_hps, remote_tps; | |
1257 | ||
1258 | trace_loadvm_postcopy_handle_advise(); | |
1259 | if (ps != POSTCOPY_INCOMING_NONE) { | |
1260 | error_report("CMD_POSTCOPY_ADVISE in wrong postcopy state (%d)", ps); | |
1261 | return -1; | |
1262 | } | |
1263 | ||
eb59db53 DDAG |
1264 | if (!postcopy_ram_supported_by_host()) { |
1265 | return -1; | |
1266 | } | |
1267 | ||
093e3c42 DDAG |
1268 | remote_hps = qemu_get_be64(mis->from_src_file); |
1269 | if (remote_hps != getpagesize()) { | |
1270 | /* | |
1271 | * Some combinations of mismatch are probably possible but it gets | |
1272 | * a bit more complicated. In particular we need to place whole | |
1273 | * host pages on the dest at once, and we need to ensure that we | |
1274 | * handle dirtying to make sure we never end up sending part of | |
1275 | * a hostpage on it's own. | |
1276 | */ | |
1277 | error_report("Postcopy needs matching host page sizes (s=%d d=%d)", | |
1278 | (int)remote_hps, getpagesize()); | |
1279 | return -1; | |
1280 | } | |
1281 | ||
1282 | remote_tps = qemu_get_be64(mis->from_src_file); | |
1283 | if (remote_tps != (1ul << qemu_target_page_bits())) { | |
1284 | /* | |
1285 | * Again, some differences could be dealt with, but for now keep it | |
1286 | * simple. | |
1287 | */ | |
1288 | error_report("Postcopy needs matching target page sizes (s=%d d=%d)", | |
1289 | (int)remote_tps, 1 << qemu_target_page_bits()); | |
1290 | return -1; | |
1291 | } | |
1292 | ||
1caddf8a DDAG |
1293 | if (ram_postcopy_incoming_init(mis)) { |
1294 | return -1; | |
1295 | } | |
1296 | ||
1297 | postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
1298 | ||
093e3c42 DDAG |
1299 | return 0; |
1300 | } | |
1301 | ||
1302 | /* After postcopy we will be told to throw some pages away since they're | |
1303 | * dirty and will have to be demand fetched. Must happen before CPU is | |
1304 | * started. | |
1305 | * There can be 0..many of these messages, each encoding multiple pages. | |
1306 | */ | |
1307 | static int loadvm_postcopy_ram_handle_discard(MigrationIncomingState *mis, | |
1308 | uint16_t len) | |
1309 | { | |
1310 | int tmp; | |
1311 | char ramid[256]; | |
1312 | PostcopyState ps = postcopy_state_get(); | |
1313 | ||
1314 | trace_loadvm_postcopy_ram_handle_discard(); | |
1315 | ||
1316 | switch (ps) { | |
1317 | case POSTCOPY_INCOMING_ADVISE: | |
1318 | /* 1st discard */ | |
f9527107 | 1319 | tmp = postcopy_ram_prepare_discard(mis); |
093e3c42 DDAG |
1320 | if (tmp) { |
1321 | return tmp; | |
1322 | } | |
1323 | break; | |
1324 | ||
1325 | case POSTCOPY_INCOMING_DISCARD: | |
1326 | /* Expected state */ | |
1327 | break; | |
1328 | ||
1329 | default: | |
1330 | error_report("CMD_POSTCOPY_RAM_DISCARD in wrong postcopy state (%d)", | |
1331 | ps); | |
1332 | return -1; | |
1333 | } | |
1334 | /* We're expecting a | |
1335 | * Version (0) | |
1336 | * a RAM ID string (length byte, name, 0 term) | |
1337 | * then at least 1 16 byte chunk | |
1338 | */ | |
1339 | if (len < (1 + 1 + 1 + 1 + 2 * 8)) { | |
1340 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1341 | return -1; | |
1342 | } | |
1343 | ||
1344 | tmp = qemu_get_byte(mis->from_src_file); | |
1345 | if (tmp != postcopy_ram_discard_version) { | |
1346 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid version (%d)", tmp); | |
1347 | return -1; | |
1348 | } | |
1349 | ||
1350 | if (!qemu_get_counted_string(mis->from_src_file, ramid)) { | |
1351 | error_report("CMD_POSTCOPY_RAM_DISCARD Failed to read RAMBlock ID"); | |
1352 | return -1; | |
1353 | } | |
1354 | tmp = qemu_get_byte(mis->from_src_file); | |
1355 | if (tmp != 0) { | |
1356 | error_report("CMD_POSTCOPY_RAM_DISCARD missing nil (%d)", tmp); | |
1357 | return -1; | |
1358 | } | |
1359 | ||
1360 | len -= 3 + strlen(ramid); | |
1361 | if (len % 16) { | |
1362 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1363 | return -1; | |
1364 | } | |
1365 | trace_loadvm_postcopy_ram_handle_discard_header(ramid, len); | |
1366 | while (len) { | |
093e3c42 DDAG |
1367 | uint64_t start_addr, block_length; |
1368 | start_addr = qemu_get_be64(mis->from_src_file); | |
1369 | block_length = qemu_get_be64(mis->from_src_file); | |
1370 | ||
1371 | len -= 16; | |
1372 | int ret = ram_discard_range(mis, ramid, start_addr, | |
1373 | block_length); | |
1374 | if (ret) { | |
1375 | return ret; | |
1376 | } | |
093e3c42 DDAG |
1377 | } |
1378 | trace_loadvm_postcopy_ram_handle_discard_end(); | |
1379 | ||
1380 | return 0; | |
1381 | } | |
1382 | ||
c76201ab DDAG |
1383 | /* |
1384 | * Triggered by a postcopy_listen command; this thread takes over reading | |
1385 | * the input stream, leaving the main thread free to carry on loading the rest | |
1386 | * of the device state (from RAM). | |
1387 | * (TODO:This could do with being in a postcopy file - but there again it's | |
1388 | * just another input loop, not that postcopy specific) | |
1389 | */ | |
1390 | static void *postcopy_ram_listen_thread(void *opaque) | |
1391 | { | |
1392 | QEMUFile *f = opaque; | |
1393 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
1394 | int load_res; | |
1395 | ||
6ba996bb DDAG |
1396 | migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE, |
1397 | MIGRATION_STATUS_POSTCOPY_ACTIVE); | |
c76201ab DDAG |
1398 | qemu_sem_post(&mis->listen_thread_sem); |
1399 | trace_postcopy_ram_listen_thread_start(); | |
1400 | ||
1401 | /* | |
1402 | * Because we're a thread and not a coroutine we can't yield | |
1403 | * in qemu_file, and thus we must be blocking now. | |
1404 | */ | |
1405 | qemu_file_set_blocking(f, true); | |
1406 | load_res = qemu_loadvm_state_main(f, mis); | |
1407 | /* And non-blocking again so we don't block in any cleanup */ | |
1408 | qemu_file_set_blocking(f, false); | |
1409 | ||
1410 | trace_postcopy_ram_listen_thread_exit(); | |
1411 | if (load_res < 0) { | |
1412 | error_report("%s: loadvm failed: %d", __func__, load_res); | |
1413 | qemu_file_set_error(f, load_res); | |
6ba996bb DDAG |
1414 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1415 | MIGRATION_STATUS_FAILED); | |
c76201ab DDAG |
1416 | } else { |
1417 | /* | |
1418 | * This looks good, but it's possible that the device loading in the | |
1419 | * main thread hasn't finished yet, and so we might not be in 'RUN' | |
1420 | * state yet; wait for the end of the main thread. | |
1421 | */ | |
1422 | qemu_event_wait(&mis->main_thread_load_event); | |
1423 | } | |
1424 | postcopy_ram_incoming_cleanup(mis); | |
c76201ab DDAG |
1425 | |
1426 | if (load_res < 0) { | |
1427 | /* | |
1428 | * If something went wrong then we have a bad state so exit; | |
1429 | * depending how far we got it might be possible at this point | |
1430 | * to leave the guest running and fire MCEs for pages that never | |
1431 | * arrived as a desperate recovery step. | |
1432 | */ | |
1433 | exit(EXIT_FAILURE); | |
1434 | } | |
1435 | ||
6ba996bb DDAG |
1436 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1437 | MIGRATION_STATUS_COMPLETED); | |
1438 | /* | |
1439 | * If everything has worked fine, then the main thread has waited | |
1440 | * for us to start, and we're the last use of the mis. | |
1441 | * (If something broke then qemu will have to exit anyway since it's | |
1442 | * got a bad migration state). | |
1443 | */ | |
1444 | migration_incoming_state_destroy(); | |
1445 | ||
1446 | ||
c76201ab DDAG |
1447 | return NULL; |
1448 | } | |
1449 | ||
093e3c42 DDAG |
1450 | /* After this message we must be able to immediately receive postcopy data */ |
1451 | static int loadvm_postcopy_handle_listen(MigrationIncomingState *mis) | |
1452 | { | |
1453 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_LISTENING); | |
1454 | trace_loadvm_postcopy_handle_listen(); | |
1455 | if (ps != POSTCOPY_INCOMING_ADVISE && ps != POSTCOPY_INCOMING_DISCARD) { | |
1456 | error_report("CMD_POSTCOPY_LISTEN in wrong postcopy state (%d)", ps); | |
1457 | return -1; | |
1458 | } | |
f9527107 DDAG |
1459 | if (ps == POSTCOPY_INCOMING_ADVISE) { |
1460 | /* | |
1461 | * A rare case, we entered listen without having to do any discards, | |
1462 | * so do the setup that's normally done at the time of the 1st discard. | |
1463 | */ | |
1464 | postcopy_ram_prepare_discard(mis); | |
1465 | } | |
093e3c42 | 1466 | |
f0a227ad DDAG |
1467 | /* |
1468 | * Sensitise RAM - can now generate requests for blocks that don't exist | |
1469 | * However, at this point the CPU shouldn't be running, and the IO | |
1470 | * shouldn't be doing anything yet so don't actually expect requests | |
1471 | */ | |
1472 | if (postcopy_ram_enable_notify(mis)) { | |
1473 | return -1; | |
1474 | } | |
1475 | ||
c76201ab DDAG |
1476 | if (mis->have_listen_thread) { |
1477 | error_report("CMD_POSTCOPY_RAM_LISTEN already has a listen thread"); | |
1478 | return -1; | |
1479 | } | |
1480 | ||
1481 | mis->have_listen_thread = true; | |
1482 | /* Start up the listening thread and wait for it to signal ready */ | |
1483 | qemu_sem_init(&mis->listen_thread_sem, 0); | |
1484 | qemu_thread_create(&mis->listen_thread, "postcopy/listen", | |
1485 | postcopy_ram_listen_thread, mis->from_src_file, | |
a587a3fe | 1486 | QEMU_THREAD_DETACHED); |
c76201ab DDAG |
1487 | qemu_sem_wait(&mis->listen_thread_sem); |
1488 | qemu_sem_destroy(&mis->listen_thread_sem); | |
1489 | ||
093e3c42 DDAG |
1490 | return 0; |
1491 | } | |
1492 | ||
86469922 DL |
1493 | |
1494 | typedef struct { | |
1495 | QEMUBH *bh; | |
1496 | } HandleRunBhData; | |
1497 | ||
ea6a55bc | 1498 | static void loadvm_postcopy_handle_run_bh(void *opaque) |
093e3c42 | 1499 | { |
27c6825b | 1500 | Error *local_err = NULL; |
86469922 | 1501 | HandleRunBhData *data = opaque; |
093e3c42 | 1502 | |
27c6825b DDAG |
1503 | /* TODO we should move all of this lot into postcopy_ram.c or a shared code |
1504 | * in migration.c | |
1505 | */ | |
1506 | cpu_synchronize_all_post_init(); | |
1507 | ||
1508 | qemu_announce_self(); | |
1509 | ||
1510 | /* Make sure all file formats flush their mutable metadata */ | |
1511 | bdrv_invalidate_cache_all(&local_err); | |
1512 | if (local_err) { | |
1513 | error_report_err(local_err); | |
27c6825b DDAG |
1514 | } |
1515 | ||
1516 | trace_loadvm_postcopy_handle_run_cpu_sync(); | |
1517 | cpu_synchronize_all_post_init(); | |
1518 | ||
1519 | trace_loadvm_postcopy_handle_run_vmstart(); | |
1520 | ||
093e3c42 DDAG |
1521 | if (autostart) { |
1522 | /* Hold onto your hats, starting the CPU */ | |
1523 | vm_start(); | |
1524 | } else { | |
1525 | /* leave it paused and let management decide when to start the CPU */ | |
1526 | runstate_set(RUN_STATE_PAUSED); | |
1527 | } | |
1528 | ||
86469922 DL |
1529 | qemu_bh_delete(data->bh); |
1530 | g_free(data); | |
ea6a55bc DL |
1531 | } |
1532 | ||
1533 | /* After all discards we can start running and asking for pages */ | |
1534 | static int loadvm_postcopy_handle_run(MigrationIncomingState *mis) | |
1535 | { | |
1536 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_RUNNING); | |
86469922 | 1537 | HandleRunBhData *data; |
ea6a55bc DL |
1538 | |
1539 | trace_loadvm_postcopy_handle_run(); | |
1540 | if (ps != POSTCOPY_INCOMING_LISTENING) { | |
1541 | error_report("CMD_POSTCOPY_RUN in wrong postcopy state (%d)", ps); | |
1542 | return -1; | |
1543 | } | |
1544 | ||
86469922 DL |
1545 | data = g_new(HandleRunBhData, 1); |
1546 | data->bh = qemu_bh_new(loadvm_postcopy_handle_run_bh, data); | |
1547 | qemu_bh_schedule(data->bh); | |
ea6a55bc | 1548 | |
27c6825b DDAG |
1549 | /* We need to finish reading the stream from the package |
1550 | * and also stop reading anything more from the stream that loaded the | |
1551 | * package (since it's now being read by the listener thread). | |
1552 | * LOADVM_QUIT will quit all the layers of nested loadvm loops. | |
1553 | */ | |
1554 | return LOADVM_QUIT; | |
093e3c42 DDAG |
1555 | } |
1556 | ||
c76ca188 | 1557 | /** |
11cf1d98 DDAG |
1558 | * Immediately following this command is a blob of data containing an embedded |
1559 | * chunk of migration stream; read it and load it. | |
1560 | * | |
1561 | * @mis: Incoming state | |
1562 | * @length: Length of packaged data to read | |
c76ca188 | 1563 | * |
11cf1d98 DDAG |
1564 | * Returns: Negative values on error |
1565 | * | |
1566 | */ | |
1567 | static int loadvm_handle_cmd_packaged(MigrationIncomingState *mis) | |
1568 | { | |
1569 | int ret; | |
61b67d47 DB |
1570 | size_t length; |
1571 | QIOChannelBuffer *bioc; | |
11cf1d98 DDAG |
1572 | |
1573 | length = qemu_get_be32(mis->from_src_file); | |
1574 | trace_loadvm_handle_cmd_packaged(length); | |
1575 | ||
1576 | if (length > MAX_VM_CMD_PACKAGED_SIZE) { | |
61b67d47 | 1577 | error_report("Unreasonably large packaged state: %zu", length); |
11cf1d98 DDAG |
1578 | return -1; |
1579 | } | |
61b67d47 DB |
1580 | |
1581 | bioc = qio_channel_buffer_new(length); | |
1582 | ret = qemu_get_buffer(mis->from_src_file, | |
1583 | bioc->data, | |
1584 | length); | |
11cf1d98 | 1585 | if (ret != length) { |
61b67d47 DB |
1586 | object_unref(OBJECT(bioc)); |
1587 | error_report("CMD_PACKAGED: Buffer receive fail ret=%d length=%zu", | |
9af9e0fe | 1588 | ret, length); |
11cf1d98 DDAG |
1589 | return (ret < 0) ? ret : -EAGAIN; |
1590 | } | |
61b67d47 | 1591 | bioc->usage += length; |
11cf1d98 DDAG |
1592 | trace_loadvm_handle_cmd_packaged_received(ret); |
1593 | ||
61b67d47 | 1594 | QEMUFile *packf = qemu_fopen_channel_input(QIO_CHANNEL(bioc)); |
11cf1d98 DDAG |
1595 | |
1596 | ret = qemu_loadvm_state_main(packf, mis); | |
1597 | trace_loadvm_handle_cmd_packaged_main(ret); | |
1598 | qemu_fclose(packf); | |
61b67d47 | 1599 | object_unref(OBJECT(bioc)); |
11cf1d98 DDAG |
1600 | |
1601 | return ret; | |
1602 | } | |
1603 | ||
1604 | /* | |
1605 | * Process an incoming 'QEMU_VM_COMMAND' | |
1606 | * 0 just a normal return | |
1607 | * LOADVM_QUIT All good, but exit the loop | |
1608 | * <0 Error | |
c76ca188 DDAG |
1609 | */ |
1610 | static int loadvm_process_command(QEMUFile *f) | |
1611 | { | |
2e37701e | 1612 | MigrationIncomingState *mis = migration_incoming_get_current(); |
c76ca188 DDAG |
1613 | uint16_t cmd; |
1614 | uint16_t len; | |
2e37701e | 1615 | uint32_t tmp32; |
c76ca188 DDAG |
1616 | |
1617 | cmd = qemu_get_be16(f); | |
1618 | len = qemu_get_be16(f); | |
1619 | ||
1620 | trace_loadvm_process_command(cmd, len); | |
1621 | if (cmd >= MIG_CMD_MAX || cmd == MIG_CMD_INVALID) { | |
1622 | error_report("MIG_CMD 0x%x unknown (len 0x%x)", cmd, len); | |
1623 | return -EINVAL; | |
1624 | } | |
1625 | ||
1626 | if (mig_cmd_args[cmd].len != -1 && mig_cmd_args[cmd].len != len) { | |
1627 | error_report("%s received with bad length - expecting %zu, got %d", | |
1628 | mig_cmd_args[cmd].name, | |
1629 | (size_t)mig_cmd_args[cmd].len, len); | |
1630 | return -ERANGE; | |
1631 | } | |
1632 | ||
1633 | switch (cmd) { | |
2e37701e DDAG |
1634 | case MIG_CMD_OPEN_RETURN_PATH: |
1635 | if (mis->to_src_file) { | |
1636 | error_report("CMD_OPEN_RETURN_PATH called when RP already open"); | |
1637 | /* Not really a problem, so don't give up */ | |
1638 | return 0; | |
1639 | } | |
1640 | mis->to_src_file = qemu_file_get_return_path(f); | |
1641 | if (!mis->to_src_file) { | |
1642 | error_report("CMD_OPEN_RETURN_PATH failed"); | |
1643 | return -1; | |
1644 | } | |
1645 | break; | |
1646 | ||
1647 | case MIG_CMD_PING: | |
1648 | tmp32 = qemu_get_be32(f); | |
1649 | trace_loadvm_process_command_ping(tmp32); | |
1650 | if (!mis->to_src_file) { | |
1651 | error_report("CMD_PING (0x%x) received with no return path", | |
1652 | tmp32); | |
1653 | return -1; | |
1654 | } | |
6decec93 | 1655 | migrate_send_rp_pong(mis, tmp32); |
2e37701e | 1656 | break; |
093e3c42 | 1657 | |
11cf1d98 DDAG |
1658 | case MIG_CMD_PACKAGED: |
1659 | return loadvm_handle_cmd_packaged(mis); | |
1660 | ||
093e3c42 DDAG |
1661 | case MIG_CMD_POSTCOPY_ADVISE: |
1662 | return loadvm_postcopy_handle_advise(mis); | |
1663 | ||
1664 | case MIG_CMD_POSTCOPY_LISTEN: | |
1665 | return loadvm_postcopy_handle_listen(mis); | |
1666 | ||
1667 | case MIG_CMD_POSTCOPY_RUN: | |
1668 | return loadvm_postcopy_handle_run(mis); | |
1669 | ||
1670 | case MIG_CMD_POSTCOPY_RAM_DISCARD: | |
1671 | return loadvm_postcopy_ram_handle_discard(mis, len); | |
c76ca188 DDAG |
1672 | } |
1673 | ||
1674 | return 0; | |
1675 | } | |
1676 | ||
1a8f46f8 | 1677 | struct LoadStateEntry { |
72cf2d4f | 1678 | QLIST_ENTRY(LoadStateEntry) entry; |
a672b469 AL |
1679 | SaveStateEntry *se; |
1680 | int section_id; | |
1681 | int version_id; | |
1a8f46f8 | 1682 | }; |
a672b469 | 1683 | |
59f39a47 DDAG |
1684 | /* |
1685 | * Read a footer off the wire and check that it matches the expected section | |
1686 | * | |
1687 | * Returns: true if the footer was good | |
1688 | * false if there is a problem (and calls error_report to say why) | |
1689 | */ | |
1690 | static bool check_section_footer(QEMUFile *f, LoadStateEntry *le) | |
1691 | { | |
1692 | uint8_t read_mark; | |
1693 | uint32_t read_section_id; | |
1694 | ||
1695 | if (skip_section_footers) { | |
1696 | /* No footer to check */ | |
1697 | return true; | |
1698 | } | |
1699 | ||
1700 | read_mark = qemu_get_byte(f); | |
1701 | ||
1702 | if (read_mark != QEMU_VM_SECTION_FOOTER) { | |
1703 | error_report("Missing section footer for %s", le->se->idstr); | |
1704 | return false; | |
1705 | } | |
1706 | ||
1707 | read_section_id = qemu_get_be32(f); | |
1708 | if (read_section_id != le->section_id) { | |
1709 | error_report("Mismatched section id in footer for %s -" | |
1710 | " read 0x%x expected 0x%x", | |
1711 | le->se->idstr, read_section_id, le->section_id); | |
1712 | return false; | |
1713 | } | |
1714 | ||
1715 | /* All good */ | |
1716 | return true; | |
1717 | } | |
1718 | ||
1a8f46f8 | 1719 | void loadvm_free_handlers(MigrationIncomingState *mis) |
a672b469 | 1720 | { |
f4dbb8dd | 1721 | LoadStateEntry *le, *new_le; |
1a8f46f8 DDAG |
1722 | |
1723 | QLIST_FOREACH_SAFE(le, &mis->loadvm_handlers, entry, new_le) { | |
1724 | QLIST_REMOVE(le, entry); | |
1725 | g_free(le); | |
1726 | } | |
1727 | } | |
1728 | ||
fb3520a8 HZ |
1729 | static int |
1730 | qemu_loadvm_section_start_full(QEMUFile *f, MigrationIncomingState *mis) | |
1731 | { | |
1732 | uint32_t instance_id, version_id, section_id; | |
1733 | SaveStateEntry *se; | |
1734 | LoadStateEntry *le; | |
1735 | char idstr[256]; | |
1736 | int ret; | |
1737 | ||
1738 | /* Read section start */ | |
1739 | section_id = qemu_get_be32(f); | |
1740 | if (!qemu_get_counted_string(f, idstr)) { | |
1741 | error_report("Unable to read ID string for section %u", | |
1742 | section_id); | |
1743 | return -EINVAL; | |
1744 | } | |
1745 | instance_id = qemu_get_be32(f); | |
1746 | version_id = qemu_get_be32(f); | |
1747 | ||
1748 | trace_qemu_loadvm_state_section_startfull(section_id, idstr, | |
1749 | instance_id, version_id); | |
1750 | /* Find savevm section */ | |
1751 | se = find_se(idstr, instance_id); | |
1752 | if (se == NULL) { | |
1753 | error_report("Unknown savevm section or instance '%s' %d", | |
1754 | idstr, instance_id); | |
1755 | return -EINVAL; | |
1756 | } | |
1757 | ||
1758 | /* Validate version */ | |
1759 | if (version_id > se->version_id) { | |
1760 | error_report("savevm: unsupported version %d for '%s' v%d", | |
1761 | version_id, idstr, se->version_id); | |
1762 | return -EINVAL; | |
1763 | } | |
1764 | ||
1765 | /* Add entry */ | |
1766 | le = g_malloc0(sizeof(*le)); | |
1767 | ||
1768 | le->se = se; | |
1769 | le->section_id = section_id; | |
1770 | le->version_id = version_id; | |
1771 | QLIST_INSERT_HEAD(&mis->loadvm_handlers, le, entry); | |
1772 | ||
1773 | ret = vmstate_load(f, le->se, le->version_id); | |
1774 | if (ret < 0) { | |
1775 | error_report("error while loading state for instance 0x%x of" | |
1776 | " device '%s'", instance_id, idstr); | |
1777 | return ret; | |
1778 | } | |
1779 | if (!check_section_footer(f, le)) { | |
1780 | return -EINVAL; | |
1781 | } | |
1782 | ||
1783 | return 0; | |
1784 | } | |
1785 | ||
1786 | static int | |
1787 | qemu_loadvm_section_part_end(QEMUFile *f, MigrationIncomingState *mis) | |
1788 | { | |
1789 | uint32_t section_id; | |
1790 | LoadStateEntry *le; | |
1791 | int ret; | |
1792 | ||
1793 | section_id = qemu_get_be32(f); | |
1794 | ||
1795 | trace_qemu_loadvm_state_section_partend(section_id); | |
1796 | QLIST_FOREACH(le, &mis->loadvm_handlers, entry) { | |
1797 | if (le->section_id == section_id) { | |
1798 | break; | |
1799 | } | |
1800 | } | |
1801 | if (le == NULL) { | |
1802 | error_report("Unknown savevm section %d", section_id); | |
1803 | return -EINVAL; | |
1804 | } | |
1805 | ||
1806 | ret = vmstate_load(f, le->se, le->version_id); | |
1807 | if (ret < 0) { | |
1808 | error_report("error while loading state section id %d(%s)", | |
1809 | section_id, le->se->idstr); | |
1810 | return ret; | |
1811 | } | |
1812 | if (!check_section_footer(f, le)) { | |
1813 | return -EINVAL; | |
1814 | } | |
1815 | ||
1816 | return 0; | |
1817 | } | |
1818 | ||
7b89bf27 | 1819 | static int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis) |
1a8f46f8 | 1820 | { |
a672b469 | 1821 | uint8_t section_type; |
a672b469 | 1822 | int ret; |
61964c23 | 1823 | |
a672b469 | 1824 | while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) { |
a672b469 | 1825 | |
a5df2a02 | 1826 | trace_qemu_loadvm_state_section(section_type); |
a672b469 AL |
1827 | switch (section_type) { |
1828 | case QEMU_VM_SECTION_START: | |
1829 | case QEMU_VM_SECTION_FULL: | |
fb3520a8 | 1830 | ret = qemu_loadvm_section_start_full(f, mis); |
b5a22e4a | 1831 | if (ret < 0) { |
7b89bf27 | 1832 | return ret; |
b5a22e4a | 1833 | } |
a672b469 AL |
1834 | break; |
1835 | case QEMU_VM_SECTION_PART: | |
1836 | case QEMU_VM_SECTION_END: | |
fb3520a8 | 1837 | ret = qemu_loadvm_section_part_end(f, mis); |
b5a22e4a | 1838 | if (ret < 0) { |
7b89bf27 | 1839 | return ret; |
b5a22e4a | 1840 | } |
a672b469 | 1841 | break; |
c76ca188 DDAG |
1842 | case QEMU_VM_COMMAND: |
1843 | ret = loadvm_process_command(f); | |
7b89bf27 DDAG |
1844 | trace_qemu_loadvm_state_section_command(ret); |
1845 | if ((ret < 0) || (ret & LOADVM_QUIT)) { | |
1846 | return ret; | |
c76ca188 DDAG |
1847 | } |
1848 | break; | |
a672b469 | 1849 | default: |
6a64b644 | 1850 | error_report("Unknown savevm section type %d", section_type); |
7b89bf27 | 1851 | return -EINVAL; |
a672b469 AL |
1852 | } |
1853 | } | |
1854 | ||
7b89bf27 DDAG |
1855 | return 0; |
1856 | } | |
1857 | ||
1858 | int qemu_loadvm_state(QEMUFile *f) | |
1859 | { | |
1860 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
1861 | Error *local_err = NULL; | |
1862 | unsigned int v; | |
1863 | int ret; | |
1864 | ||
1865 | if (qemu_savevm_state_blocked(&local_err)) { | |
1866 | error_report_err(local_err); | |
1867 | return -EINVAL; | |
1868 | } | |
1869 | ||
1870 | v = qemu_get_be32(f); | |
1871 | if (v != QEMU_VM_FILE_MAGIC) { | |
1872 | error_report("Not a migration stream"); | |
1873 | return -EINVAL; | |
1874 | } | |
1875 | ||
1876 | v = qemu_get_be32(f); | |
1877 | if (v == QEMU_VM_FILE_VERSION_COMPAT) { | |
1878 | error_report("SaveVM v2 format is obsolete and don't work anymore"); | |
1879 | return -ENOTSUP; | |
1880 | } | |
1881 | if (v != QEMU_VM_FILE_VERSION) { | |
1882 | error_report("Unsupported migration stream version"); | |
1883 | return -ENOTSUP; | |
1884 | } | |
1885 | ||
902c053d | 1886 | if (!savevm_state.skip_configuration || enforce_config_section()) { |
7b89bf27 DDAG |
1887 | if (qemu_get_byte(f) != QEMU_VM_CONFIGURATION) { |
1888 | error_report("Configuration section missing"); | |
1889 | return -EINVAL; | |
1890 | } | |
1891 | ret = vmstate_load_state(f, &vmstate_configuration, &savevm_state, 0); | |
1892 | ||
1893 | if (ret) { | |
1894 | return ret; | |
1895 | } | |
1896 | } | |
1897 | ||
1898 | ret = qemu_loadvm_state_main(f, mis); | |
1899 | qemu_event_set(&mis->main_thread_load_event); | |
1900 | ||
1901 | trace_qemu_loadvm_state_post_main(ret); | |
1902 | ||
c76201ab DDAG |
1903 | if (mis->have_listen_thread) { |
1904 | /* Listen thread still going, can't clean up yet */ | |
1905 | return ret; | |
1906 | } | |
1907 | ||
7b89bf27 DDAG |
1908 | if (ret == 0) { |
1909 | ret = qemu_file_get_error(f); | |
1910 | } | |
1925cebc AG |
1911 | |
1912 | /* | |
1913 | * Try to read in the VMDESC section as well, so that dumping tools that | |
1914 | * intercept our migration stream have the chance to see it. | |
1915 | */ | |
1aca9a5f DDAG |
1916 | |
1917 | /* We've got to be careful; if we don't read the data and just shut the fd | |
1918 | * then the sender can error if we close while it's still sending. | |
1919 | * We also mustn't read data that isn't there; some transports (RDMA) | |
1920 | * will stall waiting for that data when the source has already closed. | |
1921 | */ | |
7b89bf27 | 1922 | if (ret == 0 && should_send_vmdesc()) { |
1aca9a5f DDAG |
1923 | uint8_t *buf; |
1924 | uint32_t size; | |
7b89bf27 | 1925 | uint8_t section_type = qemu_get_byte(f); |
1aca9a5f DDAG |
1926 | |
1927 | if (section_type != QEMU_VM_VMDESCRIPTION) { | |
1928 | error_report("Expected vmdescription section, but got %d", | |
1929 | section_type); | |
1930 | /* | |
1931 | * It doesn't seem worth failing at this point since | |
1932 | * we apparently have an otherwise valid VM state | |
1933 | */ | |
1934 | } else { | |
1935 | buf = g_malloc(0x1000); | |
1936 | size = qemu_get_be32(f); | |
1937 | ||
1938 | while (size > 0) { | |
1939 | uint32_t read_chunk = MIN(size, 0x1000); | |
1940 | qemu_get_buffer(f, buf, read_chunk); | |
1941 | size -= read_chunk; | |
1942 | } | |
1943 | g_free(buf); | |
1925cebc | 1944 | } |
1925cebc AG |
1945 | } |
1946 | ||
ea375f9a JK |
1947 | cpu_synchronize_all_post_init(); |
1948 | ||
a672b469 AL |
1949 | return ret; |
1950 | } | |
1951 | ||
3e5a50d6 | 1952 | void hmp_savevm(Monitor *mon, const QDict *qdict) |
a672b469 AL |
1953 | { |
1954 | BlockDriverState *bs, *bs1; | |
1955 | QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1; | |
cb499fb2 | 1956 | int ret; |
a672b469 AL |
1957 | QEMUFile *f; |
1958 | int saved_vm_running; | |
c2c9a466 | 1959 | uint64_t vm_state_size; |
68b891ec | 1960 | qemu_timeval tv; |
7d631a11 | 1961 | struct tm tm; |
d54908a5 | 1962 | const char *name = qdict_get_try_str(qdict, "name"); |
5d80448c | 1963 | Error *local_err = NULL; |
79b3c12a | 1964 | AioContext *aio_context; |
a672b469 | 1965 | |
e9ff957a DL |
1966 | if (!bdrv_all_can_snapshot(&bs)) { |
1967 | monitor_printf(mon, "Device '%s' is writable but does not " | |
1968 | "support snapshots.\n", bdrv_get_device_name(bs)); | |
1969 | return; | |
feeee5ac MDCF |
1970 | } |
1971 | ||
0b461605 DL |
1972 | /* Delete old snapshots of the same name */ |
1973 | if (name && bdrv_all_delete_snapshot(name, &bs1, &local_err) < 0) { | |
d410fe14 MA |
1974 | error_reportf_err(local_err, |
1975 | "Error while deleting snapshot on device '%s': ", | |
1976 | bdrv_get_device_name(bs1)); | |
0b461605 DL |
1977 | return; |
1978 | } | |
1979 | ||
7cb14481 DL |
1980 | bs = bdrv_all_find_vmstate_bs(); |
1981 | if (bs == NULL) { | |
376253ec | 1982 | monitor_printf(mon, "No block device can accept snapshots\n"); |
a672b469 AL |
1983 | return; |
1984 | } | |
79b3c12a | 1985 | aio_context = bdrv_get_aio_context(bs); |
a672b469 | 1986 | |
1354869c | 1987 | saved_vm_running = runstate_is_running(); |
560d027b JQ |
1988 | |
1989 | ret = global_state_store(); | |
1990 | if (ret) { | |
1991 | monitor_printf(mon, "Error saving global state\n"); | |
1992 | return; | |
1993 | } | |
0461d5a6 | 1994 | vm_stop(RUN_STATE_SAVE_VM); |
a672b469 | 1995 | |
79b3c12a DL |
1996 | aio_context_acquire(aio_context); |
1997 | ||
cb499fb2 | 1998 | memset(sn, 0, sizeof(*sn)); |
a672b469 AL |
1999 | |
2000 | /* fill auxiliary fields */ | |
68b891ec | 2001 | qemu_gettimeofday(&tv); |
a672b469 AL |
2002 | sn->date_sec = tv.tv_sec; |
2003 | sn->date_nsec = tv.tv_usec * 1000; | |
bc72ad67 | 2004 | sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL); |
a672b469 | 2005 | |
7d631a11 MDCF |
2006 | if (name) { |
2007 | ret = bdrv_snapshot_find(bs, old_sn, name); | |
2008 | if (ret >= 0) { | |
2009 | pstrcpy(sn->name, sizeof(sn->name), old_sn->name); | |
2010 | pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str); | |
2011 | } else { | |
2012 | pstrcpy(sn->name, sizeof(sn->name), name); | |
2013 | } | |
2014 | } else { | |
d7d9b528 BS |
2015 | /* cast below needed for OpenBSD where tv_sec is still 'long' */ |
2016 | localtime_r((const time_t *)&tv.tv_sec, &tm); | |
7d631a11 | 2017 | strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm); |
7d631a11 MDCF |
2018 | } |
2019 | ||
a672b469 | 2020 | /* save the VM state */ |
45566e9c | 2021 | f = qemu_fopen_bdrv(bs, 1); |
a672b469 | 2022 | if (!f) { |
376253ec | 2023 | monitor_printf(mon, "Could not open VM state file\n"); |
a672b469 AL |
2024 | goto the_end; |
2025 | } | |
5d80448c | 2026 | ret = qemu_savevm_state(f, &local_err); |
2d22b18f | 2027 | vm_state_size = qemu_ftell(f); |
a672b469 AL |
2028 | qemu_fclose(f); |
2029 | if (ret < 0) { | |
193227f9 | 2030 | error_report_err(local_err); |
a672b469 AL |
2031 | goto the_end; |
2032 | } | |
2033 | ||
a9085f9b DL |
2034 | ret = bdrv_all_create_snapshot(sn, bs, vm_state_size, &bs); |
2035 | if (ret < 0) { | |
2036 | monitor_printf(mon, "Error while creating snapshot on '%s'\n", | |
2037 | bdrv_get_device_name(bs)); | |
a672b469 AL |
2038 | } |
2039 | ||
2040 | the_end: | |
79b3c12a | 2041 | aio_context_release(aio_context); |
38ff78d3 | 2042 | if (saved_vm_running) { |
a672b469 | 2043 | vm_start(); |
38ff78d3 | 2044 | } |
a672b469 AL |
2045 | } |
2046 | ||
a7ae8355 SS |
2047 | void qmp_xen_save_devices_state(const char *filename, Error **errp) |
2048 | { | |
2049 | QEMUFile *f; | |
8925839f | 2050 | QIOChannelFile *ioc; |
a7ae8355 SS |
2051 | int saved_vm_running; |
2052 | int ret; | |
2053 | ||
2054 | saved_vm_running = runstate_is_running(); | |
2055 | vm_stop(RUN_STATE_SAVE_VM); | |
c69adea4 | 2056 | global_state_store_running(); |
a7ae8355 | 2057 | |
8925839f DB |
2058 | ioc = qio_channel_file_new_path(filename, O_WRONLY | O_CREAT, 0660, errp); |
2059 | if (!ioc) { | |
a7ae8355 SS |
2060 | goto the_end; |
2061 | } | |
8925839f | 2062 | f = qemu_fopen_channel_output(QIO_CHANNEL(ioc)); |
a7ae8355 SS |
2063 | ret = qemu_save_device_state(f); |
2064 | qemu_fclose(f); | |
2065 | if (ret < 0) { | |
c6bd8c70 | 2066 | error_setg(errp, QERR_IO_ERROR); |
a7ae8355 SS |
2067 | } |
2068 | ||
2069 | the_end: | |
38ff78d3 | 2070 | if (saved_vm_running) { |
a7ae8355 | 2071 | vm_start(); |
38ff78d3 | 2072 | } |
a7ae8355 SS |
2073 | } |
2074 | ||
03cd4655 | 2075 | int load_vmstate(const char *name) |
a672b469 | 2076 | { |
f0aa7a8b | 2077 | BlockDriverState *bs, *bs_vm_state; |
2d22b18f | 2078 | QEMUSnapshotInfo sn; |
a672b469 | 2079 | QEMUFile *f; |
751c6a17 | 2080 | int ret; |
79b3c12a | 2081 | AioContext *aio_context; |
a672b469 | 2082 | |
849f96e2 DL |
2083 | if (!bdrv_all_can_snapshot(&bs)) { |
2084 | error_report("Device '%s' is writable but does not support snapshots.", | |
2085 | bdrv_get_device_name(bs)); | |
2086 | return -ENOTSUP; | |
2087 | } | |
723ccda1 DL |
2088 | ret = bdrv_all_find_snapshot(name, &bs); |
2089 | if (ret < 0) { | |
2090 | error_report("Device '%s' does not have the requested snapshot '%s'", | |
2091 | bdrv_get_device_name(bs), name); | |
2092 | return ret; | |
2093 | } | |
849f96e2 | 2094 | |
7cb14481 | 2095 | bs_vm_state = bdrv_all_find_vmstate_bs(); |
f0aa7a8b MDCF |
2096 | if (!bs_vm_state) { |
2097 | error_report("No block device supports snapshots"); | |
2098 | return -ENOTSUP; | |
2099 | } | |
79b3c12a | 2100 | aio_context = bdrv_get_aio_context(bs_vm_state); |
f0aa7a8b MDCF |
2101 | |
2102 | /* Don't even try to load empty VM states */ | |
79b3c12a | 2103 | aio_context_acquire(aio_context); |
f0aa7a8b | 2104 | ret = bdrv_snapshot_find(bs_vm_state, &sn, name); |
79b3c12a | 2105 | aio_context_release(aio_context); |
f0aa7a8b MDCF |
2106 | if (ret < 0) { |
2107 | return ret; | |
2108 | } else if (sn.vm_state_size == 0) { | |
e11480db KW |
2109 | error_report("This is a disk-only snapshot. Revert to it offline " |
2110 | "using qemu-img."); | |
f0aa7a8b MDCF |
2111 | return -EINVAL; |
2112 | } | |
2113 | ||
a672b469 | 2114 | /* Flush all IO requests so they don't interfere with the new state. */ |
922453bc | 2115 | bdrv_drain_all(); |
a672b469 | 2116 | |
4c1cdbaa DL |
2117 | ret = bdrv_all_goto_snapshot(name, &bs); |
2118 | if (ret < 0) { | |
2119 | error_report("Error %d while activating snapshot '%s' on '%s'", | |
2120 | ret, name, bdrv_get_device_name(bs)); | |
2121 | return ret; | |
a672b469 AL |
2122 | } |
2123 | ||
a672b469 | 2124 | /* restore the VM state */ |
f0aa7a8b | 2125 | f = qemu_fopen_bdrv(bs_vm_state, 0); |
a672b469 | 2126 | if (!f) { |
1ecda02b | 2127 | error_report("Could not open VM state file"); |
05f2401e | 2128 | return -EINVAL; |
a672b469 | 2129 | } |
f0aa7a8b | 2130 | |
5a8a49d7 | 2131 | qemu_system_reset(VMRESET_SILENT); |
bca7856a | 2132 | migration_incoming_state_new(f); |
f0aa7a8b | 2133 | |
79b3c12a DL |
2134 | aio_context_acquire(aio_context); |
2135 | ret = qemu_loadvm_state(f); | |
a672b469 | 2136 | qemu_fclose(f); |
79b3c12a DL |
2137 | aio_context_release(aio_context); |
2138 | ||
bca7856a | 2139 | migration_incoming_state_destroy(); |
a672b469 | 2140 | if (ret < 0) { |
1ecda02b | 2141 | error_report("Error %d while loading VM state", ret); |
05f2401e | 2142 | return ret; |
a672b469 | 2143 | } |
f0aa7a8b | 2144 | |
05f2401e | 2145 | return 0; |
7b630349 JQ |
2146 | } |
2147 | ||
3e5a50d6 | 2148 | void hmp_delvm(Monitor *mon, const QDict *qdict) |
a672b469 | 2149 | { |
af957387 | 2150 | BlockDriverState *bs; |
ba2b2288 | 2151 | Error *err; |
d54908a5 | 2152 | const char *name = qdict_get_str(qdict, "name"); |
a672b469 | 2153 | |
9b00ea37 | 2154 | if (bdrv_all_delete_snapshot(name, &bs, &err) < 0) { |
d410fe14 MA |
2155 | error_reportf_err(err, |
2156 | "Error while deleting snapshot on device '%s': ", | |
2157 | bdrv_get_device_name(bs)); | |
a672b469 AL |
2158 | } |
2159 | } | |
2160 | ||
1ce6be24 | 2161 | void hmp_info_snapshots(Monitor *mon, const QDict *qdict) |
a672b469 AL |
2162 | { |
2163 | BlockDriverState *bs, *bs1; | |
723ccda1 DL |
2164 | QEMUSnapshotInfo *sn_tab, *sn; |
2165 | int nb_sns, i; | |
f9209915 MDCF |
2166 | int total; |
2167 | int *available_snapshots; | |
79b3c12a | 2168 | AioContext *aio_context; |
a672b469 | 2169 | |
7cb14481 | 2170 | bs = bdrv_all_find_vmstate_bs(); |
a672b469 | 2171 | if (!bs) { |
376253ec | 2172 | monitor_printf(mon, "No available block device supports snapshots\n"); |
a672b469 AL |
2173 | return; |
2174 | } | |
79b3c12a | 2175 | aio_context = bdrv_get_aio_context(bs); |
a672b469 | 2176 | |
79b3c12a | 2177 | aio_context_acquire(aio_context); |
a672b469 | 2178 | nb_sns = bdrv_snapshot_list(bs, &sn_tab); |
79b3c12a DL |
2179 | aio_context_release(aio_context); |
2180 | ||
a672b469 | 2181 | if (nb_sns < 0) { |
376253ec | 2182 | monitor_printf(mon, "bdrv_snapshot_list: error %d\n", nb_sns); |
a672b469 AL |
2183 | return; |
2184 | } | |
f9209915 MDCF |
2185 | |
2186 | if (nb_sns == 0) { | |
2187 | monitor_printf(mon, "There is no snapshot available.\n"); | |
2188 | return; | |
2189 | } | |
2190 | ||
97f3ad35 | 2191 | available_snapshots = g_new0(int, nb_sns); |
f9209915 MDCF |
2192 | total = 0; |
2193 | for (i = 0; i < nb_sns; i++) { | |
723ccda1 | 2194 | if (bdrv_all_find_snapshot(sn_tab[i].id_str, &bs1) == 0) { |
f9209915 MDCF |
2195 | available_snapshots[total] = i; |
2196 | total++; | |
2197 | } | |
a672b469 | 2198 | } |
f9209915 MDCF |
2199 | |
2200 | if (total > 0) { | |
5b917044 WX |
2201 | bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, NULL); |
2202 | monitor_printf(mon, "\n"); | |
f9209915 MDCF |
2203 | for (i = 0; i < total; i++) { |
2204 | sn = &sn_tab[available_snapshots[i]]; | |
5b917044 WX |
2205 | bdrv_snapshot_dump((fprintf_function)monitor_printf, mon, sn); |
2206 | monitor_printf(mon, "\n"); | |
f9209915 MDCF |
2207 | } |
2208 | } else { | |
2209 | monitor_printf(mon, "There is no suitable snapshot available\n"); | |
2210 | } | |
2211 | ||
7267c094 AL |
2212 | g_free(sn_tab); |
2213 | g_free(available_snapshots); | |
f9209915 | 2214 | |
a672b469 | 2215 | } |
c5705a77 AK |
2216 | |
2217 | void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev) | |
2218 | { | |
fa53a0e5 | 2219 | qemu_ram_set_idstr(mr->ram_block, |
c5705a77 AK |
2220 | memory_region_name(mr), dev); |
2221 | } | |
2222 | ||
2223 | void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev) | |
2224 | { | |
fa53a0e5 | 2225 | qemu_ram_unset_idstr(mr->ram_block); |
c5705a77 AK |
2226 | } |
2227 | ||
2228 | void vmstate_register_ram_global(MemoryRegion *mr) | |
2229 | { | |
2230 | vmstate_register_ram(mr, NULL); | |
2231 | } |