]>
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" |
88c16567 | 34 | #include "hw/xen/xen.h" |
1422e32d | 35 | #include "net/net.h" |
9c17d615 | 36 | #include "sysemu/sysemu.h" |
1de7afc9 | 37 | #include "qemu/timer.h" |
caf71f86 | 38 | #include "migration/migration.h" |
40014d81 | 39 | #include "qemu-file-channel.h" |
be07b0ac | 40 | #include "postcopy-ram.h" |
cc7a8ea7 | 41 | #include "qapi/qmp/qerror.h" |
d49b6836 | 42 | #include "qemu/error-report.h" |
1de7afc9 | 43 | #include "qemu/queue.h" |
9c17d615 | 44 | #include "sysemu/cpus.h" |
022c62cb | 45 | #include "exec/memory.h" |
51180423 | 46 | #include "exec/target_page.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" |
f348b6d1 | 52 | #include "qemu/cutils.h" |
61b67d47 | 53 | #include "io/channel-buffer.h" |
8925839f | 54 | #include "io/channel-file.h" |
a672b469 | 55 | |
18995b98 | 56 | #ifndef ETH_P_RARP |
f8778a77 | 57 | #define ETH_P_RARP 0x8035 |
18995b98 N |
58 | #endif |
59 | #define ARP_HTYPE_ETH 0x0001 | |
60 | #define ARP_PTYPE_IP 0x0800 | |
61 | #define ARP_OP_REQUEST_REV 0x3 | |
62 | ||
093e3c42 DDAG |
63 | const unsigned int postcopy_ram_discard_version = 0; |
64 | ||
37fb569c DDAG |
65 | static bool skip_section_footers; |
66 | ||
c76ca188 DDAG |
67 | static struct mig_cmd_args { |
68 | ssize_t len; /* -1 = variable */ | |
69 | const char *name; | |
70 | } mig_cmd_args[] = { | |
71 | [MIG_CMD_INVALID] = { .len = -1, .name = "INVALID" }, | |
2e37701e DDAG |
72 | [MIG_CMD_OPEN_RETURN_PATH] = { .len = 0, .name = "OPEN_RETURN_PATH" }, |
73 | [MIG_CMD_PING] = { .len = sizeof(uint32_t), .name = "PING" }, | |
093e3c42 DDAG |
74 | [MIG_CMD_POSTCOPY_ADVISE] = { .len = 16, .name = "POSTCOPY_ADVISE" }, |
75 | [MIG_CMD_POSTCOPY_LISTEN] = { .len = 0, .name = "POSTCOPY_LISTEN" }, | |
76 | [MIG_CMD_POSTCOPY_RUN] = { .len = 0, .name = "POSTCOPY_RUN" }, | |
77 | [MIG_CMD_POSTCOPY_RAM_DISCARD] = { | |
78 | .len = -1, .name = "POSTCOPY_RAM_DISCARD" }, | |
11cf1d98 | 79 | [MIG_CMD_PACKAGED] = { .len = 4, .name = "PACKAGED" }, |
c76ca188 DDAG |
80 | [MIG_CMD_MAX] = { .len = -1, .name = "MAX" }, |
81 | }; | |
82 | ||
18995b98 | 83 | static int announce_self_create(uint8_t *buf, |
5cecf414 | 84 | uint8_t *mac_addr) |
a672b469 | 85 | { |
18995b98 N |
86 | /* Ethernet header. */ |
87 | memset(buf, 0xff, 6); /* destination MAC addr */ | |
88 | memcpy(buf + 6, mac_addr, 6); /* source MAC addr */ | |
89 | *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */ | |
90 | ||
91 | /* RARP header. */ | |
92 | *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */ | |
93 | *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */ | |
94 | *(buf + 18) = 6; /* hardware addr length (ethernet) */ | |
95 | *(buf + 19) = 4; /* protocol addr length (IPv4) */ | |
96 | *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */ | |
97 | memcpy(buf + 22, mac_addr, 6); /* source hw addr */ | |
98 | memset(buf + 28, 0x00, 4); /* source protocol addr */ | |
99 | memcpy(buf + 32, mac_addr, 6); /* target hw addr */ | |
100 | memset(buf + 38, 0x00, 4); /* target protocol addr */ | |
101 | ||
102 | /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */ | |
103 | memset(buf + 42, 0x00, 18); | |
104 | ||
105 | return 60; /* len (FCS will be added by hardware) */ | |
a672b469 AL |
106 | } |
107 | ||
f401ca22 | 108 | static void qemu_announce_self_iter(NICState *nic, void *opaque) |
a672b469 | 109 | { |
18995b98 | 110 | uint8_t buf[60]; |
f401ca22 MM |
111 | int len; |
112 | ||
9013dca5 | 113 | trace_qemu_announce_self_iter(qemu_ether_ntoa(&nic->conf->macaddr)); |
f401ca22 MM |
114 | len = announce_self_create(buf, nic->conf->macaddr.a); |
115 | ||
b356f76d | 116 | qemu_send_packet_raw(qemu_get_queue(nic), buf, len); |
f401ca22 MM |
117 | } |
118 | ||
119 | ||
120 | static void qemu_announce_self_once(void *opaque) | |
121 | { | |
ed8b330b GN |
122 | static int count = SELF_ANNOUNCE_ROUNDS; |
123 | QEMUTimer *timer = *(QEMUTimer **)opaque; | |
a672b469 | 124 | |
f401ca22 MM |
125 | qemu_foreach_nic(qemu_announce_self_iter, NULL); |
126 | ||
18995b98 N |
127 | if (--count) { |
128 | /* delay 50ms, 150ms, 250ms, ... */ | |
bc72ad67 | 129 | timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + |
508e1180 | 130 | self_announce_delay(count)); |
ed8b330b | 131 | } else { |
5cecf414 EH |
132 | timer_del(timer); |
133 | timer_free(timer); | |
ed8b330b GN |
134 | } |
135 | } | |
136 | ||
137 | void qemu_announce_self(void) | |
138 | { | |
5cecf414 EH |
139 | static QEMUTimer *timer; |
140 | timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer); | |
141 | qemu_announce_self_once(&timer); | |
a672b469 AL |
142 | } |
143 | ||
144 | /***********************************************************/ | |
145 | /* savevm/loadvm support */ | |
146 | ||
05fcc848 KW |
147 | static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt, |
148 | int64_t pos) | |
149 | { | |
150 | int ret; | |
151 | QEMUIOVector qiov; | |
152 | ||
153 | qemu_iovec_init_external(&qiov, iov, iovcnt); | |
154 | ret = bdrv_writev_vmstate(opaque, &qiov, pos); | |
155 | if (ret < 0) { | |
156 | return ret; | |
157 | } | |
158 | ||
159 | return qiov.size; | |
160 | } | |
161 | ||
a202a4c0 DDAG |
162 | static ssize_t block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, |
163 | size_t size) | |
a672b469 | 164 | { |
45566e9c | 165 | return bdrv_load_vmstate(opaque, buf, pos, size); |
a672b469 AL |
166 | } |
167 | ||
168 | static int bdrv_fclose(void *opaque) | |
169 | { | |
ad492c92 | 170 | return bdrv_flush(opaque); |
a672b469 AL |
171 | } |
172 | ||
9229bf3c PB |
173 | static const QEMUFileOps bdrv_read_ops = { |
174 | .get_buffer = block_get_buffer, | |
175 | .close = bdrv_fclose | |
176 | }; | |
177 | ||
178 | static const QEMUFileOps bdrv_write_ops = { | |
05fcc848 KW |
179 | .writev_buffer = block_writev_buffer, |
180 | .close = bdrv_fclose | |
9229bf3c PB |
181 | }; |
182 | ||
45566e9c | 183 | static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable) |
a672b469 | 184 | { |
38ff78d3 | 185 | if (is_writable) { |
9229bf3c | 186 | return qemu_fopen_ops(bs, &bdrv_write_ops); |
38ff78d3 | 187 | } |
9229bf3c | 188 | return qemu_fopen_ops(bs, &bdrv_read_ops); |
a672b469 AL |
189 | } |
190 | ||
2ff68d07 | 191 | |
bb1a6d8c EH |
192 | /* QEMUFile timer support. |
193 | * Not in qemu-file.c to not add qemu-timer.c as dependency to qemu-file.c | |
194 | */ | |
2ff68d07 | 195 | |
40daca54 | 196 | void timer_put(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
197 | { |
198 | uint64_t expire_time; | |
199 | ||
e93379b0 | 200 | expire_time = timer_expire_time_ns(ts); |
2ff68d07 PB |
201 | qemu_put_be64(f, expire_time); |
202 | } | |
203 | ||
40daca54 | 204 | void timer_get(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
205 | { |
206 | uint64_t expire_time; | |
207 | ||
208 | expire_time = qemu_get_be64(f); | |
209 | if (expire_time != -1) { | |
bc72ad67 | 210 | timer_mod_ns(ts, expire_time); |
2ff68d07 | 211 | } else { |
bc72ad67 | 212 | timer_del(ts); |
2ff68d07 PB |
213 | } |
214 | } | |
215 | ||
216 | ||
bb1a6d8c EH |
217 | /* VMState timer support. |
218 | * Not in vmstate.c to not add qemu-timer.c as dependency to vmstate.c | |
219 | */ | |
dde0463b | 220 | |
2c21ee76 | 221 | static int get_timer(QEMUFile *f, void *pv, size_t size, VMStateField *field) |
dde0463b JQ |
222 | { |
223 | QEMUTimer *v = pv; | |
40daca54 | 224 | timer_get(f, v); |
dde0463b JQ |
225 | return 0; |
226 | } | |
227 | ||
2c21ee76 JD |
228 | static int put_timer(QEMUFile *f, void *pv, size_t size, VMStateField *field, |
229 | QJSON *vmdesc) | |
dde0463b | 230 | { |
84e2e3eb | 231 | QEMUTimer *v = pv; |
40daca54 | 232 | timer_put(f, v); |
2c21ee76 JD |
233 | |
234 | return 0; | |
dde0463b JQ |
235 | } |
236 | ||
237 | const VMStateInfo vmstate_info_timer = { | |
238 | .name = "timer", | |
239 | .get = get_timer, | |
240 | .put = put_timer, | |
241 | }; | |
242 | ||
08e99e29 | 243 | |
7685ee6a AW |
244 | typedef struct CompatEntry { |
245 | char idstr[256]; | |
246 | int instance_id; | |
247 | } CompatEntry; | |
248 | ||
a672b469 | 249 | typedef struct SaveStateEntry { |
72cf2d4f | 250 | QTAILQ_ENTRY(SaveStateEntry) entry; |
a672b469 AL |
251 | char idstr[256]; |
252 | int instance_id; | |
4d2ffa08 | 253 | int alias_id; |
a672b469 AL |
254 | int version_id; |
255 | int section_id; | |
22ea40f4 | 256 | SaveVMHandlers *ops; |
9ed7d6ae | 257 | const VMStateDescription *vmsd; |
a672b469 | 258 | void *opaque; |
7685ee6a | 259 | CompatEntry *compat; |
a7ae8355 | 260 | int is_ram; |
a672b469 AL |
261 | } SaveStateEntry; |
262 | ||
0163a2e0 JQ |
263 | typedef struct SaveState { |
264 | QTAILQ_HEAD(, SaveStateEntry) handlers; | |
265 | int global_section_id; | |
61964c23 JQ |
266 | bool skip_configuration; |
267 | uint32_t len; | |
268 | const char *name; | |
59811a32 | 269 | uint32_t target_page_bits; |
0163a2e0 JQ |
270 | } SaveState; |
271 | ||
272 | static SaveState savevm_state = { | |
273 | .handlers = QTAILQ_HEAD_INITIALIZER(savevm_state.handlers), | |
274 | .global_section_id = 0, | |
61964c23 JQ |
275 | .skip_configuration = false, |
276 | }; | |
277 | ||
278 | void savevm_skip_configuration(void) | |
279 | { | |
280 | savevm_state.skip_configuration = true; | |
281 | } | |
282 | ||
283 | ||
284 | static void configuration_pre_save(void *opaque) | |
285 | { | |
286 | SaveState *state = opaque; | |
287 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
288 | ||
289 | state->len = strlen(current_name); | |
290 | state->name = current_name; | |
59811a32 PM |
291 | state->target_page_bits = TARGET_PAGE_BITS; |
292 | } | |
293 | ||
294 | static int configuration_pre_load(void *opaque) | |
295 | { | |
296 | SaveState *state = opaque; | |
297 | ||
298 | /* If there is no target-page-bits subsection it means the source | |
299 | * predates the variable-target-page-bits support and is using the | |
300 | * minimum possible value for this CPU. | |
301 | */ | |
302 | state->target_page_bits = TARGET_PAGE_BITS_MIN; | |
303 | return 0; | |
61964c23 JQ |
304 | } |
305 | ||
306 | static int configuration_post_load(void *opaque, int version_id) | |
307 | { | |
308 | SaveState *state = opaque; | |
309 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
310 | ||
311 | if (strncmp(state->name, current_name, state->len) != 0) { | |
15d61692 GK |
312 | error_report("Machine type received is '%.*s' and local is '%s'", |
313 | (int) state->len, state->name, current_name); | |
61964c23 JQ |
314 | return -EINVAL; |
315 | } | |
59811a32 PM |
316 | |
317 | if (state->target_page_bits != TARGET_PAGE_BITS) { | |
318 | error_report("Received TARGET_PAGE_BITS is %d but local is %d", | |
319 | state->target_page_bits, TARGET_PAGE_BITS); | |
320 | return -EINVAL; | |
321 | } | |
322 | ||
61964c23 JQ |
323 | return 0; |
324 | } | |
325 | ||
59811a32 PM |
326 | /* The target-page-bits subsection is present only if the |
327 | * target page size is not the same as the default (ie the | |
328 | * minimum page size for a variable-page-size guest CPU). | |
329 | * If it is present then it contains the actual target page | |
330 | * bits for the machine, and migration will fail if the | |
331 | * two ends don't agree about it. | |
332 | */ | |
333 | static bool vmstate_target_page_bits_needed(void *opaque) | |
334 | { | |
335 | return TARGET_PAGE_BITS > TARGET_PAGE_BITS_MIN; | |
336 | } | |
337 | ||
338 | static const VMStateDescription vmstate_target_page_bits = { | |
339 | .name = "configuration/target-page-bits", | |
340 | .version_id = 1, | |
341 | .minimum_version_id = 1, | |
342 | .needed = vmstate_target_page_bits_needed, | |
343 | .fields = (VMStateField[]) { | |
344 | VMSTATE_UINT32(target_page_bits, SaveState), | |
345 | VMSTATE_END_OF_LIST() | |
346 | } | |
347 | }; | |
348 | ||
61964c23 JQ |
349 | static const VMStateDescription vmstate_configuration = { |
350 | .name = "configuration", | |
351 | .version_id = 1, | |
59811a32 | 352 | .pre_load = configuration_pre_load, |
61964c23 JQ |
353 | .post_load = configuration_post_load, |
354 | .pre_save = configuration_pre_save, | |
355 | .fields = (VMStateField[]) { | |
356 | VMSTATE_UINT32(len, SaveState), | |
59046ec2 | 357 | VMSTATE_VBUFFER_ALLOC_UINT32(name, SaveState, 0, NULL, len), |
61964c23 JQ |
358 | VMSTATE_END_OF_LIST() |
359 | }, | |
59811a32 PM |
360 | .subsections = (const VMStateDescription*[]) { |
361 | &vmstate_target_page_bits, | |
362 | NULL | |
363 | } | |
0163a2e0 | 364 | }; |
a672b469 | 365 | |
abfd9ce3 AS |
366 | static void dump_vmstate_vmsd(FILE *out_file, |
367 | const VMStateDescription *vmsd, int indent, | |
368 | bool is_subsection); | |
369 | ||
370 | static void dump_vmstate_vmsf(FILE *out_file, const VMStateField *field, | |
371 | int indent) | |
372 | { | |
373 | fprintf(out_file, "%*s{\n", indent, ""); | |
374 | indent += 2; | |
375 | fprintf(out_file, "%*s\"field\": \"%s\",\n", indent, "", field->name); | |
376 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
377 | field->version_id); | |
378 | fprintf(out_file, "%*s\"field_exists\": %s,\n", indent, "", | |
379 | field->field_exists ? "true" : "false"); | |
380 | fprintf(out_file, "%*s\"size\": %zu", indent, "", field->size); | |
381 | if (field->vmsd != NULL) { | |
382 | fprintf(out_file, ",\n"); | |
383 | dump_vmstate_vmsd(out_file, field->vmsd, indent, false); | |
384 | } | |
385 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
386 | } | |
387 | ||
388 | static void dump_vmstate_vmss(FILE *out_file, | |
5cd8cada | 389 | const VMStateDescription **subsection, |
abfd9ce3 AS |
390 | int indent) |
391 | { | |
5cd8cada JQ |
392 | if (*subsection != NULL) { |
393 | dump_vmstate_vmsd(out_file, *subsection, indent, true); | |
abfd9ce3 AS |
394 | } |
395 | } | |
396 | ||
397 | static void dump_vmstate_vmsd(FILE *out_file, | |
398 | const VMStateDescription *vmsd, int indent, | |
399 | bool is_subsection) | |
400 | { | |
401 | if (is_subsection) { | |
402 | fprintf(out_file, "%*s{\n", indent, ""); | |
403 | } else { | |
404 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", "Description"); | |
405 | } | |
406 | indent += 2; | |
407 | fprintf(out_file, "%*s\"name\": \"%s\",\n", indent, "", vmsd->name); | |
408 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
409 | vmsd->version_id); | |
410 | fprintf(out_file, "%*s\"minimum_version_id\": %d", indent, "", | |
411 | vmsd->minimum_version_id); | |
412 | if (vmsd->fields != NULL) { | |
413 | const VMStateField *field = vmsd->fields; | |
414 | bool first; | |
415 | ||
416 | fprintf(out_file, ",\n%*s\"Fields\": [\n", indent, ""); | |
417 | first = true; | |
418 | while (field->name != NULL) { | |
419 | if (field->flags & VMS_MUST_EXIST) { | |
420 | /* Ignore VMSTATE_VALIDATE bits; these don't get migrated */ | |
421 | field++; | |
422 | continue; | |
423 | } | |
424 | if (!first) { | |
425 | fprintf(out_file, ",\n"); | |
426 | } | |
427 | dump_vmstate_vmsf(out_file, field, indent + 2); | |
428 | field++; | |
429 | first = false; | |
430 | } | |
431 | fprintf(out_file, "\n%*s]", indent, ""); | |
432 | } | |
433 | if (vmsd->subsections != NULL) { | |
5cd8cada | 434 | const VMStateDescription **subsection = vmsd->subsections; |
abfd9ce3 AS |
435 | bool first; |
436 | ||
437 | fprintf(out_file, ",\n%*s\"Subsections\": [\n", indent, ""); | |
438 | first = true; | |
5cd8cada | 439 | while (*subsection != NULL) { |
abfd9ce3 AS |
440 | if (!first) { |
441 | fprintf(out_file, ",\n"); | |
442 | } | |
443 | dump_vmstate_vmss(out_file, subsection, indent + 2); | |
444 | subsection++; | |
445 | first = false; | |
446 | } | |
447 | fprintf(out_file, "\n%*s]", indent, ""); | |
448 | } | |
449 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
450 | } | |
451 | ||
452 | static void dump_machine_type(FILE *out_file) | |
453 | { | |
454 | MachineClass *mc; | |
455 | ||
456 | mc = MACHINE_GET_CLASS(current_machine); | |
457 | ||
458 | fprintf(out_file, " \"vmschkmachine\": {\n"); | |
459 | fprintf(out_file, " \"Name\": \"%s\"\n", mc->name); | |
460 | fprintf(out_file, " },\n"); | |
461 | } | |
462 | ||
463 | void dump_vmstate_json_to_file(FILE *out_file) | |
464 | { | |
465 | GSList *list, *elt; | |
466 | bool first; | |
467 | ||
468 | fprintf(out_file, "{\n"); | |
469 | dump_machine_type(out_file); | |
470 | ||
471 | first = true; | |
472 | list = object_class_get_list(TYPE_DEVICE, true); | |
473 | for (elt = list; elt; elt = elt->next) { | |
474 | DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, elt->data, | |
475 | TYPE_DEVICE); | |
476 | const char *name; | |
477 | int indent = 2; | |
478 | ||
479 | if (!dc->vmsd) { | |
480 | continue; | |
481 | } | |
482 | ||
483 | if (!first) { | |
484 | fprintf(out_file, ",\n"); | |
485 | } | |
486 | name = object_class_get_name(OBJECT_CLASS(dc)); | |
487 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", name); | |
488 | indent += 2; | |
489 | fprintf(out_file, "%*s\"Name\": \"%s\",\n", indent, "", name); | |
490 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
491 | dc->vmsd->version_id); | |
492 | fprintf(out_file, "%*s\"minimum_version_id\": %d,\n", indent, "", | |
493 | dc->vmsd->minimum_version_id); | |
494 | ||
495 | dump_vmstate_vmsd(out_file, dc->vmsd, indent, false); | |
496 | ||
497 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
498 | first = false; | |
499 | } | |
500 | fprintf(out_file, "\n}\n"); | |
501 | fclose(out_file); | |
502 | } | |
503 | ||
8718e999 JQ |
504 | static int calculate_new_instance_id(const char *idstr) |
505 | { | |
506 | SaveStateEntry *se; | |
507 | int instance_id = 0; | |
508 | ||
0163a2e0 | 509 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
8718e999 JQ |
510 | if (strcmp(idstr, se->idstr) == 0 |
511 | && instance_id <= se->instance_id) { | |
512 | instance_id = se->instance_id + 1; | |
513 | } | |
514 | } | |
515 | return instance_id; | |
516 | } | |
517 | ||
7685ee6a AW |
518 | static int calculate_compat_instance_id(const char *idstr) |
519 | { | |
520 | SaveStateEntry *se; | |
521 | int instance_id = 0; | |
522 | ||
0163a2e0 | 523 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
38ff78d3 | 524 | if (!se->compat) { |
7685ee6a | 525 | continue; |
38ff78d3 | 526 | } |
7685ee6a AW |
527 | |
528 | if (strcmp(idstr, se->compat->idstr) == 0 | |
529 | && instance_id <= se->compat->instance_id) { | |
530 | instance_id = se->compat->instance_id + 1; | |
531 | } | |
532 | } | |
533 | return instance_id; | |
534 | } | |
535 | ||
f37bc036 PX |
536 | static inline MigrationPriority save_state_priority(SaveStateEntry *se) |
537 | { | |
538 | if (se->vmsd) { | |
539 | return se->vmsd->priority; | |
540 | } | |
541 | return MIG_PRI_DEFAULT; | |
542 | } | |
543 | ||
544 | static void savevm_state_handler_insert(SaveStateEntry *nse) | |
545 | { | |
546 | MigrationPriority priority = save_state_priority(nse); | |
547 | SaveStateEntry *se; | |
548 | ||
549 | assert(priority <= MIG_PRI_MAX); | |
550 | ||
551 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
552 | if (save_state_priority(se) < priority) { | |
553 | break; | |
554 | } | |
555 | } | |
556 | ||
557 | if (se) { | |
558 | QTAILQ_INSERT_BEFORE(se, nse, entry); | |
559 | } else { | |
560 | QTAILQ_INSERT_TAIL(&savevm_state.handlers, nse, entry); | |
561 | } | |
562 | } | |
563 | ||
a672b469 AL |
564 | /* TODO: Individual devices generally have very little idea about the rest |
565 | of the system, so instance_id should be removed/replaced. | |
566 | Meanwhile pass -1 as instance_id if you do not already have a clearly | |
567 | distinguishing id for all instances of your device class. */ | |
0be71e32 AW |
568 | int register_savevm_live(DeviceState *dev, |
569 | const char *idstr, | |
a672b469 AL |
570 | int instance_id, |
571 | int version_id, | |
7908c78d | 572 | SaveVMHandlers *ops, |
a672b469 AL |
573 | void *opaque) |
574 | { | |
8718e999 | 575 | SaveStateEntry *se; |
a672b469 | 576 | |
97f3ad35 | 577 | se = g_new0(SaveStateEntry, 1); |
a672b469 | 578 | se->version_id = version_id; |
0163a2e0 | 579 | se->section_id = savevm_state.global_section_id++; |
7908c78d | 580 | se->ops = ops; |
a672b469 | 581 | se->opaque = opaque; |
9ed7d6ae | 582 | se->vmsd = NULL; |
a7ae8355 | 583 | /* if this is a live_savem then set is_ram */ |
16310a3c | 584 | if (ops->save_live_setup != NULL) { |
a7ae8355 SS |
585 | se->is_ram = 1; |
586 | } | |
a672b469 | 587 | |
09e5ab63 AL |
588 | if (dev) { |
589 | char *id = qdev_get_dev_path(dev); | |
7685ee6a | 590 | if (id) { |
581f08ba DDAG |
591 | if (snprintf(se->idstr, sizeof(se->idstr), "%s/", id) >= |
592 | sizeof(se->idstr)) { | |
593 | error_report("Path too long for VMState (%s)", id); | |
594 | g_free(id); | |
595 | g_free(se); | |
596 | ||
597 | return -1; | |
598 | } | |
7267c094 | 599 | g_free(id); |
7685ee6a | 600 | |
97f3ad35 | 601 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
602 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr); |
603 | se->compat->instance_id = instance_id == -1 ? | |
604 | calculate_compat_instance_id(idstr) : instance_id; | |
605 | instance_id = -1; | |
606 | } | |
607 | } | |
608 | pstrcat(se->idstr, sizeof(se->idstr), idstr); | |
609 | ||
8718e999 | 610 | if (instance_id == -1) { |
7685ee6a | 611 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
612 | } else { |
613 | se->instance_id = instance_id; | |
a672b469 | 614 | } |
7685ee6a | 615 | assert(!se->compat || se->instance_id == 0); |
f37bc036 | 616 | savevm_state_handler_insert(se); |
a672b469 AL |
617 | return 0; |
618 | } | |
619 | ||
0be71e32 AW |
620 | int register_savevm(DeviceState *dev, |
621 | const char *idstr, | |
a672b469 AL |
622 | int instance_id, |
623 | int version_id, | |
624 | SaveStateHandler *save_state, | |
625 | LoadStateHandler *load_state, | |
626 | void *opaque) | |
627 | { | |
97f3ad35 | 628 | SaveVMHandlers *ops = g_new0(SaveVMHandlers, 1); |
7908c78d JQ |
629 | ops->save_state = save_state; |
630 | ops->load_state = load_state; | |
0be71e32 | 631 | return register_savevm_live(dev, idstr, instance_id, version_id, |
7908c78d | 632 | ops, opaque); |
a672b469 AL |
633 | } |
634 | ||
0be71e32 | 635 | void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque) |
41bd13af | 636 | { |
8718e999 | 637 | SaveStateEntry *se, *new_se; |
7685ee6a AW |
638 | char id[256] = ""; |
639 | ||
09e5ab63 AL |
640 | if (dev) { |
641 | char *path = qdev_get_dev_path(dev); | |
7685ee6a AW |
642 | if (path) { |
643 | pstrcpy(id, sizeof(id), path); | |
644 | pstrcat(id, sizeof(id), "/"); | |
7267c094 | 645 | g_free(path); |
7685ee6a AW |
646 | } |
647 | } | |
648 | pstrcat(id, sizeof(id), idstr); | |
41bd13af | 649 | |
0163a2e0 | 650 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
7685ee6a | 651 | if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) { |
0163a2e0 | 652 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 653 | g_free(se->compat); |
22ea40f4 | 654 | g_free(se->ops); |
7267c094 | 655 | g_free(se); |
41bd13af | 656 | } |
41bd13af AL |
657 | } |
658 | } | |
659 | ||
0be71e32 | 660 | int vmstate_register_with_alias_id(DeviceState *dev, int instance_id, |
4d2ffa08 JK |
661 | const VMStateDescription *vmsd, |
662 | void *opaque, int alias_id, | |
bc5c4f21 DDAG |
663 | int required_for_version, |
664 | Error **errp) | |
9ed7d6ae | 665 | { |
8718e999 | 666 | SaveStateEntry *se; |
9ed7d6ae | 667 | |
4d2ffa08 JK |
668 | /* If this triggers, alias support can be dropped for the vmsd. */ |
669 | assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id); | |
670 | ||
97f3ad35 | 671 | se = g_new0(SaveStateEntry, 1); |
9ed7d6ae | 672 | se->version_id = vmsd->version_id; |
0163a2e0 | 673 | se->section_id = savevm_state.global_section_id++; |
9ed7d6ae JQ |
674 | se->opaque = opaque; |
675 | se->vmsd = vmsd; | |
4d2ffa08 | 676 | se->alias_id = alias_id; |
9ed7d6ae | 677 | |
09e5ab63 AL |
678 | if (dev) { |
679 | char *id = qdev_get_dev_path(dev); | |
7685ee6a | 680 | if (id) { |
581f08ba DDAG |
681 | if (snprintf(se->idstr, sizeof(se->idstr), "%s/", id) >= |
682 | sizeof(se->idstr)) { | |
683 | error_setg(errp, "Path too long for VMState (%s)", id); | |
684 | g_free(id); | |
685 | g_free(se); | |
686 | ||
687 | return -1; | |
688 | } | |
128e4e10 | 689 | g_free(id); |
7685ee6a | 690 | |
97f3ad35 | 691 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
692 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name); |
693 | se->compat->instance_id = instance_id == -1 ? | |
694 | calculate_compat_instance_id(vmsd->name) : instance_id; | |
695 | instance_id = -1; | |
696 | } | |
697 | } | |
698 | pstrcat(se->idstr, sizeof(se->idstr), vmsd->name); | |
699 | ||
8718e999 | 700 | if (instance_id == -1) { |
7685ee6a | 701 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
702 | } else { |
703 | se->instance_id = instance_id; | |
9ed7d6ae | 704 | } |
7685ee6a | 705 | assert(!se->compat || se->instance_id == 0); |
f37bc036 | 706 | savevm_state_handler_insert(se); |
9ed7d6ae JQ |
707 | return 0; |
708 | } | |
709 | ||
0be71e32 AW |
710 | void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd, |
711 | void *opaque) | |
9ed7d6ae | 712 | { |
1eb7538b JQ |
713 | SaveStateEntry *se, *new_se; |
714 | ||
0163a2e0 | 715 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
1eb7538b | 716 | if (se->vmsd == vmsd && se->opaque == opaque) { |
0163a2e0 | 717 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 718 | g_free(se->compat); |
7267c094 | 719 | g_free(se); |
1eb7538b JQ |
720 | } |
721 | } | |
9ed7d6ae JQ |
722 | } |
723 | ||
4082be4d JQ |
724 | static int vmstate_load(QEMUFile *f, SaveStateEntry *se, int version_id) |
725 | { | |
9013dca5 | 726 | trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
9ed7d6ae | 727 | if (!se->vmsd) { /* Old style */ |
22ea40f4 | 728 | return se->ops->load_state(f, se->opaque, version_id); |
9ed7d6ae JQ |
729 | } |
730 | return vmstate_load_state(f, se->vmsd, se->opaque, version_id); | |
4082be4d JQ |
731 | } |
732 | ||
8118f095 AG |
733 | static void vmstate_save_old_style(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) |
734 | { | |
735 | int64_t old_offset, size; | |
736 | ||
737 | old_offset = qemu_ftell_fast(f); | |
738 | se->ops->save_state(f, se->opaque); | |
739 | size = qemu_ftell_fast(f) - old_offset; | |
740 | ||
741 | if (vmdesc) { | |
742 | json_prop_int(vmdesc, "size", size); | |
743 | json_start_array(vmdesc, "fields"); | |
744 | json_start_object(vmdesc, NULL); | |
745 | json_prop_str(vmdesc, "name", "data"); | |
746 | json_prop_int(vmdesc, "size", size); | |
747 | json_prop_str(vmdesc, "type", "buffer"); | |
748 | json_end_object(vmdesc); | |
749 | json_end_array(vmdesc); | |
750 | } | |
751 | } | |
752 | ||
753 | static void vmstate_save(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) | |
4082be4d | 754 | { |
9013dca5 | 755 | trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
8118f095 AG |
756 | if (!se->vmsd) { |
757 | vmstate_save_old_style(f, se, vmdesc); | |
dc912121 | 758 | return; |
9ed7d6ae | 759 | } |
8118f095 | 760 | vmstate_save_state(f, se->vmsd, se->opaque, vmdesc); |
4082be4d JQ |
761 | } |
762 | ||
37fb569c DDAG |
763 | void savevm_skip_section_footers(void) |
764 | { | |
765 | skip_section_footers = true; | |
766 | } | |
767 | ||
ce39bfc9 DDAG |
768 | /* |
769 | * Write the header for device section (QEMU_VM_SECTION START/END/PART/FULL) | |
770 | */ | |
771 | static void save_section_header(QEMUFile *f, SaveStateEntry *se, | |
772 | uint8_t section_type) | |
773 | { | |
774 | qemu_put_byte(f, section_type); | |
775 | qemu_put_be32(f, se->section_id); | |
776 | ||
777 | if (section_type == QEMU_VM_SECTION_FULL || | |
778 | section_type == QEMU_VM_SECTION_START) { | |
779 | /* ID string */ | |
780 | size_t len = strlen(se->idstr); | |
781 | qemu_put_byte(f, len); | |
782 | qemu_put_buffer(f, (uint8_t *)se->idstr, len); | |
783 | ||
784 | qemu_put_be32(f, se->instance_id); | |
785 | qemu_put_be32(f, se->version_id); | |
786 | } | |
787 | } | |
788 | ||
f68945d4 DDAG |
789 | /* |
790 | * Write a footer onto device sections that catches cases misformatted device | |
791 | * sections. | |
792 | */ | |
793 | static void save_section_footer(QEMUFile *f, SaveStateEntry *se) | |
794 | { | |
795 | if (!skip_section_footers) { | |
796 | qemu_put_byte(f, QEMU_VM_SECTION_FOOTER); | |
797 | qemu_put_be32(f, se->section_id); | |
798 | } | |
799 | } | |
800 | ||
c76ca188 DDAG |
801 | /** |
802 | * qemu_savevm_command_send: Send a 'QEMU_VM_COMMAND' type element with the | |
803 | * command and associated data. | |
804 | * | |
805 | * @f: File to send command on | |
806 | * @command: Command type to send | |
807 | * @len: Length of associated data | |
808 | * @data: Data associated with command. | |
809 | */ | |
810 | void qemu_savevm_command_send(QEMUFile *f, | |
811 | enum qemu_vm_cmd command, | |
812 | uint16_t len, | |
813 | uint8_t *data) | |
814 | { | |
815 | trace_savevm_command_send(command, len); | |
816 | qemu_put_byte(f, QEMU_VM_COMMAND); | |
817 | qemu_put_be16(f, (uint16_t)command); | |
818 | qemu_put_be16(f, len); | |
819 | qemu_put_buffer(f, data, len); | |
820 | qemu_fflush(f); | |
821 | } | |
822 | ||
2e37701e DDAG |
823 | void qemu_savevm_send_ping(QEMUFile *f, uint32_t value) |
824 | { | |
825 | uint32_t buf; | |
826 | ||
827 | trace_savevm_send_ping(value); | |
828 | buf = cpu_to_be32(value); | |
829 | qemu_savevm_command_send(f, MIG_CMD_PING, sizeof(value), (uint8_t *)&buf); | |
830 | } | |
831 | ||
832 | void qemu_savevm_send_open_return_path(QEMUFile *f) | |
833 | { | |
834 | trace_savevm_send_open_return_path(); | |
835 | qemu_savevm_command_send(f, MIG_CMD_OPEN_RETURN_PATH, 0, NULL); | |
836 | } | |
837 | ||
11cf1d98 DDAG |
838 | /* We have a buffer of data to send; we don't want that all to be loaded |
839 | * by the command itself, so the command contains just the length of the | |
840 | * extra buffer that we then send straight after it. | |
841 | * TODO: Must be a better way to organise that | |
842 | * | |
843 | * Returns: | |
844 | * 0 on success | |
845 | * -ve on error | |
846 | */ | |
61b67d47 | 847 | int qemu_savevm_send_packaged(QEMUFile *f, const uint8_t *buf, size_t len) |
11cf1d98 | 848 | { |
11cf1d98 DDAG |
849 | uint32_t tmp; |
850 | ||
851 | if (len > MAX_VM_CMD_PACKAGED_SIZE) { | |
852 | error_report("%s: Unreasonably large packaged state: %zu", | |
853 | __func__, len); | |
854 | return -1; | |
855 | } | |
856 | ||
857 | tmp = cpu_to_be32(len); | |
858 | ||
859 | trace_qemu_savevm_send_packaged(); | |
860 | qemu_savevm_command_send(f, MIG_CMD_PACKAGED, 4, (uint8_t *)&tmp); | |
861 | ||
61b67d47 | 862 | qemu_put_buffer(f, buf, len); |
11cf1d98 DDAG |
863 | |
864 | return 0; | |
865 | } | |
866 | ||
093e3c42 DDAG |
867 | /* Send prior to any postcopy transfer */ |
868 | void qemu_savevm_send_postcopy_advise(QEMUFile *f) | |
869 | { | |
870 | uint64_t tmp[2]; | |
e8ca1db2 | 871 | tmp[0] = cpu_to_be64(ram_pagesize_summary()); |
20afaed9 | 872 | tmp[1] = cpu_to_be64(qemu_target_page_size()); |
093e3c42 DDAG |
873 | |
874 | trace_qemu_savevm_send_postcopy_advise(); | |
875 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_ADVISE, 16, (uint8_t *)tmp); | |
876 | } | |
877 | ||
878 | /* Sent prior to starting the destination running in postcopy, discard pages | |
879 | * that have already been sent but redirtied on the source. | |
880 | * CMD_POSTCOPY_RAM_DISCARD consist of: | |
881 | * byte version (0) | |
882 | * byte Length of name field (not including 0) | |
883 | * n x byte RAM block name | |
884 | * byte 0 terminator (just for safety) | |
885 | * n x Byte ranges within the named RAMBlock | |
886 | * be64 Start of the range | |
887 | * be64 Length | |
888 | * | |
889 | * name: RAMBlock name that these entries are part of | |
890 | * len: Number of page entries | |
891 | * start_list: 'len' addresses | |
892 | * length_list: 'len' addresses | |
893 | * | |
894 | */ | |
895 | void qemu_savevm_send_postcopy_ram_discard(QEMUFile *f, const char *name, | |
896 | uint16_t len, | |
897 | uint64_t *start_list, | |
898 | uint64_t *length_list) | |
899 | { | |
900 | uint8_t *buf; | |
901 | uint16_t tmplen; | |
902 | uint16_t t; | |
903 | size_t name_len = strlen(name); | |
904 | ||
905 | trace_qemu_savevm_send_postcopy_ram_discard(name, len); | |
906 | assert(name_len < 256); | |
907 | buf = g_malloc0(1 + 1 + name_len + 1 + (8 + 8) * len); | |
908 | buf[0] = postcopy_ram_discard_version; | |
909 | buf[1] = name_len; | |
910 | memcpy(buf + 2, name, name_len); | |
911 | tmplen = 2 + name_len; | |
912 | buf[tmplen++] = '\0'; | |
913 | ||
914 | for (t = 0; t < len; t++) { | |
4d885131 | 915 | stq_be_p(buf + tmplen, start_list[t]); |
093e3c42 | 916 | tmplen += 8; |
4d885131 | 917 | stq_be_p(buf + tmplen, length_list[t]); |
093e3c42 DDAG |
918 | tmplen += 8; |
919 | } | |
920 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RAM_DISCARD, tmplen, buf); | |
921 | g_free(buf); | |
922 | } | |
923 | ||
924 | /* Get the destination into a state where it can receive postcopy data. */ | |
925 | void qemu_savevm_send_postcopy_listen(QEMUFile *f) | |
926 | { | |
927 | trace_savevm_send_postcopy_listen(); | |
928 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_LISTEN, 0, NULL); | |
929 | } | |
930 | ||
931 | /* Kick the destination into running */ | |
932 | void qemu_savevm_send_postcopy_run(QEMUFile *f) | |
933 | { | |
934 | trace_savevm_send_postcopy_run(); | |
935 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RUN, 0, NULL); | |
936 | } | |
937 | ||
e1c37d0e | 938 | bool qemu_savevm_state_blocked(Error **errp) |
dc912121 AW |
939 | { |
940 | SaveStateEntry *se; | |
941 | ||
0163a2e0 | 942 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
7d854c47 | 943 | if (se->vmsd && se->vmsd->unmigratable) { |
f231b88d CR |
944 | error_setg(errp, "State blocked by non-migratable device '%s'", |
945 | se->idstr); | |
dc912121 AW |
946 | return true; |
947 | } | |
948 | } | |
949 | return false; | |
950 | } | |
951 | ||
902c053d GK |
952 | static bool enforce_config_section(void) |
953 | { | |
954 | MachineState *machine = MACHINE(qdev_get_machine()); | |
955 | return machine->enforce_config_section; | |
956 | } | |
957 | ||
f796baa1 DDAG |
958 | void qemu_savevm_state_header(QEMUFile *f) |
959 | { | |
960 | trace_savevm_state_header(); | |
961 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
962 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
172dfd4f | 963 | |
902c053d | 964 | if (!savevm_state.skip_configuration || enforce_config_section()) { |
172dfd4f DDAG |
965 | qemu_put_byte(f, QEMU_VM_CONFIGURATION); |
966 | vmstate_save_state(f, &vmstate_configuration, &savevm_state, 0); | |
967 | } | |
968 | ||
f796baa1 DDAG |
969 | } |
970 | ||
a0762d9e | 971 | void qemu_savevm_state_begin(QEMUFile *f) |
a672b469 AL |
972 | { |
973 | SaveStateEntry *se; | |
39346385 | 974 | int ret; |
a672b469 | 975 | |
9013dca5 | 976 | trace_savevm_state_begin(); |
0163a2e0 | 977 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
d1315aac | 978 | if (!se->ops || !se->ops->save_live_setup) { |
a672b469 | 979 | continue; |
22ea40f4 | 980 | } |
6bd68781 JQ |
981 | if (se->ops && se->ops->is_active) { |
982 | if (!se->ops->is_active(se->opaque)) { | |
983 | continue; | |
984 | } | |
985 | } | |
ce39bfc9 | 986 | save_section_header(f, se, QEMU_VM_SECTION_START); |
a672b469 | 987 | |
d1315aac | 988 | ret = se->ops->save_live_setup(f, se->opaque); |
f68945d4 | 989 | save_section_footer(f, se); |
2975725f | 990 | if (ret < 0) { |
47c8c17a PB |
991 | qemu_file_set_error(f, ret); |
992 | break; | |
2975725f | 993 | } |
a672b469 | 994 | } |
a672b469 AL |
995 | } |
996 | ||
39346385 | 997 | /* |
07f35073 | 998 | * this function has three return values: |
39346385 JQ |
999 | * negative: there was one error, and we have -errno. |
1000 | * 0 : We haven't finished, caller have to go again | |
1001 | * 1 : We have finished, we can go to complete phase | |
1002 | */ | |
35ecd943 | 1003 | int qemu_savevm_state_iterate(QEMUFile *f, bool postcopy) |
a672b469 AL |
1004 | { |
1005 | SaveStateEntry *se; | |
1006 | int ret = 1; | |
1007 | ||
9013dca5 | 1008 | trace_savevm_state_iterate(); |
0163a2e0 | 1009 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
16310a3c | 1010 | if (!se->ops || !se->ops->save_live_iterate) { |
a672b469 | 1011 | continue; |
22ea40f4 | 1012 | } |
6bd68781 JQ |
1013 | if (se->ops && se->ops->is_active) { |
1014 | if (!se->ops->is_active(se->opaque)) { | |
1015 | continue; | |
1016 | } | |
1017 | } | |
35ecd943 DDAG |
1018 | /* |
1019 | * In the postcopy phase, any device that doesn't know how to | |
1020 | * do postcopy should have saved it's state in the _complete | |
1021 | * call that's already run, it might get confused if we call | |
1022 | * iterate afterwards. | |
1023 | */ | |
1024 | if (postcopy && !se->ops->save_live_complete_postcopy) { | |
1025 | continue; | |
1026 | } | |
aac844ed JQ |
1027 | if (qemu_file_rate_limit(f)) { |
1028 | return 0; | |
1029 | } | |
464400f6 | 1030 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
1031 | |
1032 | save_section_header(f, se, QEMU_VM_SECTION_PART); | |
a672b469 | 1033 | |
16310a3c | 1034 | ret = se->ops->save_live_iterate(f, se->opaque); |
a5df2a02 | 1035 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 1036 | save_section_footer(f, se); |
517a13c9 | 1037 | |
47c8c17a PB |
1038 | if (ret < 0) { |
1039 | qemu_file_set_error(f, ret); | |
1040 | } | |
2975725f | 1041 | if (ret <= 0) { |
90697be8 JK |
1042 | /* Do not proceed to the next vmstate before this one reported |
1043 | completion of the current stage. This serializes the migration | |
1044 | and reduces the probability that a faster changing state is | |
1045 | synchronized over and over again. */ | |
1046 | break; | |
1047 | } | |
a672b469 | 1048 | } |
39346385 | 1049 | return ret; |
a672b469 AL |
1050 | } |
1051 | ||
9850c604 AG |
1052 | static bool should_send_vmdesc(void) |
1053 | { | |
1054 | MachineState *machine = MACHINE(qdev_get_machine()); | |
5727309d | 1055 | bool in_postcopy = migration_in_postcopy(); |
8421b205 | 1056 | return !machine->suppress_vmdesc && !in_postcopy; |
9850c604 AG |
1057 | } |
1058 | ||
763c906b DDAG |
1059 | /* |
1060 | * Calls the save_live_complete_postcopy methods | |
1061 | * causing the last few pages to be sent immediately and doing any associated | |
1062 | * cleanup. | |
1063 | * Note postcopy also calls qemu_savevm_state_complete_precopy to complete | |
1064 | * all the other devices, but that happens at the point we switch to postcopy. | |
1065 | */ | |
1066 | void qemu_savevm_state_complete_postcopy(QEMUFile *f) | |
1067 | { | |
1068 | SaveStateEntry *se; | |
1069 | int ret; | |
1070 | ||
1071 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
1072 | if (!se->ops || !se->ops->save_live_complete_postcopy) { | |
1073 | continue; | |
1074 | } | |
1075 | if (se->ops && se->ops->is_active) { | |
1076 | if (!se->ops->is_active(se->opaque)) { | |
1077 | continue; | |
1078 | } | |
1079 | } | |
1080 | trace_savevm_section_start(se->idstr, se->section_id); | |
1081 | /* Section type */ | |
1082 | qemu_put_byte(f, QEMU_VM_SECTION_END); | |
1083 | qemu_put_be32(f, se->section_id); | |
1084 | ||
1085 | ret = se->ops->save_live_complete_postcopy(f, se->opaque); | |
1086 | trace_savevm_section_end(se->idstr, se->section_id, ret); | |
1087 | save_section_footer(f, se); | |
1088 | if (ret < 0) { | |
1089 | qemu_file_set_error(f, ret); | |
1090 | return; | |
1091 | } | |
1092 | } | |
1093 | ||
1094 | qemu_put_byte(f, QEMU_VM_EOF); | |
1095 | qemu_fflush(f); | |
1096 | } | |
1097 | ||
1c0d249d | 1098 | void qemu_savevm_state_complete_precopy(QEMUFile *f, bool iterable_only) |
a672b469 | 1099 | { |
8118f095 AG |
1100 | QJSON *vmdesc; |
1101 | int vmdesc_len; | |
a672b469 | 1102 | SaveStateEntry *se; |
2975725f | 1103 | int ret; |
5727309d | 1104 | bool in_postcopy = migration_in_postcopy(); |
a672b469 | 1105 | |
a3e06c3d | 1106 | trace_savevm_state_complete_precopy(); |
9013dca5 | 1107 | |
ea375f9a JK |
1108 | cpu_synchronize_all_states(); |
1109 | ||
0163a2e0 | 1110 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
763c906b DDAG |
1111 | if (!se->ops || |
1112 | (in_postcopy && se->ops->save_live_complete_postcopy) || | |
1c0d249d | 1113 | (in_postcopy && !iterable_only) || |
763c906b | 1114 | !se->ops->save_live_complete_precopy) { |
a672b469 | 1115 | continue; |
22ea40f4 | 1116 | } |
1c0d249d | 1117 | |
6bd68781 JQ |
1118 | if (se->ops && se->ops->is_active) { |
1119 | if (!se->ops->is_active(se->opaque)) { | |
1120 | continue; | |
1121 | } | |
1122 | } | |
464400f6 | 1123 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
1124 | |
1125 | save_section_header(f, se, QEMU_VM_SECTION_END); | |
a672b469 | 1126 | |
a3e06c3d | 1127 | ret = se->ops->save_live_complete_precopy(f, se->opaque); |
a5df2a02 | 1128 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 1129 | save_section_footer(f, se); |
2975725f | 1130 | if (ret < 0) { |
47c8c17a PB |
1131 | qemu_file_set_error(f, ret); |
1132 | return; | |
2975725f | 1133 | } |
a672b469 AL |
1134 | } |
1135 | ||
1c0d249d DDAG |
1136 | if (iterable_only) { |
1137 | return; | |
1138 | } | |
1139 | ||
8118f095 AG |
1140 | vmdesc = qjson_new(); |
1141 | json_prop_int(vmdesc, "page_size", TARGET_PAGE_SIZE); | |
1142 | json_start_array(vmdesc, "devices"); | |
0163a2e0 | 1143 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1144 | |
22ea40f4 | 1145 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
5cecf414 | 1146 | continue; |
22ea40f4 | 1147 | } |
df896152 JQ |
1148 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1149 | trace_savevm_section_skip(se->idstr, se->section_id); | |
1150 | continue; | |
1151 | } | |
1152 | ||
464400f6 | 1153 | trace_savevm_section_start(se->idstr, se->section_id); |
8118f095 AG |
1154 | |
1155 | json_start_object(vmdesc, NULL); | |
1156 | json_prop_str(vmdesc, "name", se->idstr); | |
1157 | json_prop_int(vmdesc, "instance_id", se->instance_id); | |
1158 | ||
ce39bfc9 | 1159 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
8118f095 | 1160 | vmstate_save(f, se, vmdesc); |
a5df2a02 | 1161 | trace_savevm_section_end(se->idstr, se->section_id, 0); |
f68945d4 | 1162 | save_section_footer(f, se); |
bdf46d64 WY |
1163 | |
1164 | json_end_object(vmdesc); | |
a672b469 AL |
1165 | } |
1166 | ||
763c906b DDAG |
1167 | if (!in_postcopy) { |
1168 | /* Postcopy stream will still be going */ | |
1169 | qemu_put_byte(f, QEMU_VM_EOF); | |
1170 | } | |
8118f095 AG |
1171 | |
1172 | json_end_array(vmdesc); | |
1173 | qjson_finish(vmdesc); | |
1174 | vmdesc_len = strlen(qjson_get_str(vmdesc)); | |
1175 | ||
9850c604 AG |
1176 | if (should_send_vmdesc()) { |
1177 | qemu_put_byte(f, QEMU_VM_VMDESCRIPTION); | |
1178 | qemu_put_be32(f, vmdesc_len); | |
1179 | qemu_put_buffer(f, (uint8_t *)qjson_get_str(vmdesc), vmdesc_len); | |
1180 | } | |
b72fe9e6 | 1181 | qjson_destroy(vmdesc); |
8118f095 | 1182 | |
edaae611 | 1183 | qemu_fflush(f); |
a672b469 AL |
1184 | } |
1185 | ||
c31b098f DDAG |
1186 | /* Give an estimate of the amount left to be transferred, |
1187 | * the result is split into the amount for units that can and | |
1188 | * for units that can't do postcopy. | |
1189 | */ | |
faec066a | 1190 | void qemu_savevm_state_pending(QEMUFile *f, uint64_t threshold_size, |
c31b098f DDAG |
1191 | uint64_t *res_non_postcopiable, |
1192 | uint64_t *res_postcopiable) | |
e4ed1541 JQ |
1193 | { |
1194 | SaveStateEntry *se; | |
c31b098f DDAG |
1195 | |
1196 | *res_non_postcopiable = 0; | |
1197 | *res_postcopiable = 0; | |
1198 | ||
e4ed1541 | 1199 | |
0163a2e0 | 1200 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
e4ed1541 JQ |
1201 | if (!se->ops || !se->ops->save_live_pending) { |
1202 | continue; | |
1203 | } | |
1204 | if (se->ops && se->ops->is_active) { | |
1205 | if (!se->ops->is_active(se->opaque)) { | |
1206 | continue; | |
1207 | } | |
1208 | } | |
faec066a | 1209 | se->ops->save_live_pending(f, se->opaque, threshold_size, |
c31b098f | 1210 | res_non_postcopiable, res_postcopiable); |
e4ed1541 | 1211 | } |
e4ed1541 JQ |
1212 | } |
1213 | ||
ea7415fa | 1214 | void qemu_savevm_state_cleanup(void) |
4ec7fcc7 JK |
1215 | { |
1216 | SaveStateEntry *se; | |
1217 | ||
ea7415fa | 1218 | trace_savevm_state_cleanup(); |
0163a2e0 | 1219 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
d1a8548c LL |
1220 | if (se->ops && se->ops->cleanup) { |
1221 | se->ops->cleanup(se->opaque); | |
4ec7fcc7 JK |
1222 | } |
1223 | } | |
1224 | } | |
1225 | ||
5d80448c | 1226 | static int qemu_savevm_state(QEMUFile *f, Error **errp) |
a672b469 | 1227 | { |
a672b469 | 1228 | int ret; |
a0762d9e | 1229 | MigrationState *ms = migrate_init(); |
6dcf6668 | 1230 | MigrationStatus status; |
89a02a9f | 1231 | ms->to_dst_file = f; |
a672b469 | 1232 | |
24f3902b | 1233 | if (migration_is_blocked(errp)) { |
6dcf6668 DL |
1234 | ret = -EINVAL; |
1235 | goto done; | |
dc912121 AW |
1236 | } |
1237 | ||
ce7c817c JQ |
1238 | if (migrate_use_block()) { |
1239 | error_setg(errp, "Block migration and snapshots are incompatible"); | |
1240 | ret = -EINVAL; | |
1241 | goto done; | |
1242 | } | |
1243 | ||
9b095037 | 1244 | qemu_mutex_unlock_iothread(); |
f796baa1 | 1245 | qemu_savevm_state_header(f); |
a0762d9e | 1246 | qemu_savevm_state_begin(f); |
9b095037 PB |
1247 | qemu_mutex_lock_iothread(); |
1248 | ||
47c8c17a | 1249 | while (qemu_file_get_error(f) == 0) { |
35ecd943 | 1250 | if (qemu_savevm_state_iterate(f, false) > 0) { |
47c8c17a PB |
1251 | break; |
1252 | } | |
1253 | } | |
a672b469 | 1254 | |
47c8c17a | 1255 | ret = qemu_file_get_error(f); |
39346385 | 1256 | if (ret == 0) { |
1c0d249d | 1257 | qemu_savevm_state_complete_precopy(f, false); |
624b9cc2 | 1258 | ret = qemu_file_get_error(f); |
39346385 | 1259 | } |
15b3b8ea | 1260 | qemu_savevm_state_cleanup(); |
04943eba | 1261 | if (ret != 0) { |
5d80448c | 1262 | error_setg_errno(errp, -ret, "Error while writing VM state"); |
04943eba | 1263 | } |
6dcf6668 DL |
1264 | |
1265 | done: | |
1266 | if (ret != 0) { | |
1267 | status = MIGRATION_STATUS_FAILED; | |
1268 | } else { | |
1269 | status = MIGRATION_STATUS_COMPLETED; | |
1270 | } | |
1271 | migrate_set_state(&ms->state, MIGRATION_STATUS_SETUP, status); | |
f9c8caa0 VSO |
1272 | |
1273 | /* f is outer parameter, it should not stay in global migration state after | |
1274 | * this function finished */ | |
1275 | ms->to_dst_file = NULL; | |
1276 | ||
a672b469 AL |
1277 | return ret; |
1278 | } | |
1279 | ||
a7ae8355 SS |
1280 | static int qemu_save_device_state(QEMUFile *f) |
1281 | { | |
1282 | SaveStateEntry *se; | |
1283 | ||
1284 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
1285 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
1286 | ||
1287 | cpu_synchronize_all_states(); | |
1288 | ||
0163a2e0 | 1289 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a7ae8355 SS |
1290 | if (se->is_ram) { |
1291 | continue; | |
1292 | } | |
22ea40f4 | 1293 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
a7ae8355 SS |
1294 | continue; |
1295 | } | |
df896152 JQ |
1296 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1297 | continue; | |
1298 | } | |
a7ae8355 | 1299 | |
ce39bfc9 | 1300 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
a7ae8355 | 1301 | |
8118f095 | 1302 | vmstate_save(f, se, NULL); |
f68945d4 DDAG |
1303 | |
1304 | save_section_footer(f, se); | |
a7ae8355 SS |
1305 | } |
1306 | ||
1307 | qemu_put_byte(f, QEMU_VM_EOF); | |
1308 | ||
1309 | return qemu_file_get_error(f); | |
1310 | } | |
1311 | ||
a672b469 AL |
1312 | static SaveStateEntry *find_se(const char *idstr, int instance_id) |
1313 | { | |
1314 | SaveStateEntry *se; | |
1315 | ||
0163a2e0 | 1316 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1317 | if (!strcmp(se->idstr, idstr) && |
4d2ffa08 JK |
1318 | (instance_id == se->instance_id || |
1319 | instance_id == se->alias_id)) | |
a672b469 | 1320 | return se; |
7685ee6a AW |
1321 | /* Migrating from an older version? */ |
1322 | if (strstr(se->idstr, idstr) && se->compat) { | |
1323 | if (!strcmp(se->compat->idstr, idstr) && | |
1324 | (instance_id == se->compat->instance_id || | |
1325 | instance_id == se->alias_id)) | |
1326 | return se; | |
1327 | } | |
a672b469 AL |
1328 | } |
1329 | return NULL; | |
1330 | } | |
1331 | ||
7b89bf27 DDAG |
1332 | enum LoadVMExitCodes { |
1333 | /* Allow a command to quit all layers of nested loadvm loops */ | |
1334 | LOADVM_QUIT = 1, | |
1335 | }; | |
1336 | ||
1337 | static int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis); | |
093e3c42 DDAG |
1338 | |
1339 | /* ------ incoming postcopy messages ------ */ | |
1340 | /* 'advise' arrives before any transfers just to tell us that a postcopy | |
1341 | * *might* happen - it might be skipped if precopy transferred everything | |
1342 | * quickly. | |
1343 | */ | |
1344 | static int loadvm_postcopy_handle_advise(MigrationIncomingState *mis) | |
1345 | { | |
1346 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
e8ca1db2 | 1347 | uint64_t remote_pagesize_summary, local_pagesize_summary, remote_tps; |
093e3c42 DDAG |
1348 | |
1349 | trace_loadvm_postcopy_handle_advise(); | |
1350 | if (ps != POSTCOPY_INCOMING_NONE) { | |
1351 | error_report("CMD_POSTCOPY_ADVISE in wrong postcopy state (%d)", ps); | |
1352 | return -1; | |
1353 | } | |
1354 | ||
eb59db53 | 1355 | if (!postcopy_ram_supported_by_host()) { |
328d4d85 | 1356 | postcopy_state_set(POSTCOPY_INCOMING_NONE); |
eb59db53 DDAG |
1357 | return -1; |
1358 | } | |
1359 | ||
e8ca1db2 DDAG |
1360 | remote_pagesize_summary = qemu_get_be64(mis->from_src_file); |
1361 | local_pagesize_summary = ram_pagesize_summary(); | |
1362 | ||
1363 | if (remote_pagesize_summary != local_pagesize_summary) { | |
093e3c42 | 1364 | /* |
e8ca1db2 DDAG |
1365 | * This detects two potential causes of mismatch: |
1366 | * a) A mismatch in host page sizes | |
1367 | * Some combinations of mismatch are probably possible but it gets | |
1368 | * a bit more complicated. In particular we need to place whole | |
1369 | * host pages on the dest at once, and we need to ensure that we | |
1370 | * handle dirtying to make sure we never end up sending part of | |
1371 | * a hostpage on it's own. | |
1372 | * b) The use of different huge page sizes on source/destination | |
1373 | * a more fine grain test is performed during RAM block migration | |
1374 | * but this test here causes a nice early clear failure, and | |
1375 | * also fails when passed to an older qemu that doesn't | |
1376 | * do huge pages. | |
093e3c42 | 1377 | */ |
e8ca1db2 DDAG |
1378 | error_report("Postcopy needs matching RAM page sizes (s=%" PRIx64 |
1379 | " d=%" PRIx64 ")", | |
1380 | remote_pagesize_summary, local_pagesize_summary); | |
093e3c42 DDAG |
1381 | return -1; |
1382 | } | |
1383 | ||
1384 | remote_tps = qemu_get_be64(mis->from_src_file); | |
20afaed9 | 1385 | if (remote_tps != qemu_target_page_size()) { |
093e3c42 DDAG |
1386 | /* |
1387 | * Again, some differences could be dealt with, but for now keep it | |
1388 | * simple. | |
1389 | */ | |
20afaed9 JQ |
1390 | error_report("Postcopy needs matching target page sizes (s=%d d=%zd)", |
1391 | (int)remote_tps, qemu_target_page_size()); | |
093e3c42 DDAG |
1392 | return -1; |
1393 | } | |
1394 | ||
1caddf8a DDAG |
1395 | if (ram_postcopy_incoming_init(mis)) { |
1396 | return -1; | |
1397 | } | |
1398 | ||
1399 | postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
1400 | ||
093e3c42 DDAG |
1401 | return 0; |
1402 | } | |
1403 | ||
1404 | /* After postcopy we will be told to throw some pages away since they're | |
1405 | * dirty and will have to be demand fetched. Must happen before CPU is | |
1406 | * started. | |
1407 | * There can be 0..many of these messages, each encoding multiple pages. | |
1408 | */ | |
1409 | static int loadvm_postcopy_ram_handle_discard(MigrationIncomingState *mis, | |
1410 | uint16_t len) | |
1411 | { | |
1412 | int tmp; | |
1413 | char ramid[256]; | |
1414 | PostcopyState ps = postcopy_state_get(); | |
1415 | ||
1416 | trace_loadvm_postcopy_ram_handle_discard(); | |
1417 | ||
1418 | switch (ps) { | |
1419 | case POSTCOPY_INCOMING_ADVISE: | |
1420 | /* 1st discard */ | |
f9527107 | 1421 | tmp = postcopy_ram_prepare_discard(mis); |
093e3c42 DDAG |
1422 | if (tmp) { |
1423 | return tmp; | |
1424 | } | |
1425 | break; | |
1426 | ||
1427 | case POSTCOPY_INCOMING_DISCARD: | |
1428 | /* Expected state */ | |
1429 | break; | |
1430 | ||
1431 | default: | |
1432 | error_report("CMD_POSTCOPY_RAM_DISCARD in wrong postcopy state (%d)", | |
1433 | ps); | |
1434 | return -1; | |
1435 | } | |
1436 | /* We're expecting a | |
1437 | * Version (0) | |
1438 | * a RAM ID string (length byte, name, 0 term) | |
1439 | * then at least 1 16 byte chunk | |
1440 | */ | |
1441 | if (len < (1 + 1 + 1 + 1 + 2 * 8)) { | |
1442 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1443 | return -1; | |
1444 | } | |
1445 | ||
1446 | tmp = qemu_get_byte(mis->from_src_file); | |
1447 | if (tmp != postcopy_ram_discard_version) { | |
1448 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid version (%d)", tmp); | |
1449 | return -1; | |
1450 | } | |
1451 | ||
1452 | if (!qemu_get_counted_string(mis->from_src_file, ramid)) { | |
1453 | error_report("CMD_POSTCOPY_RAM_DISCARD Failed to read RAMBlock ID"); | |
1454 | return -1; | |
1455 | } | |
1456 | tmp = qemu_get_byte(mis->from_src_file); | |
1457 | if (tmp != 0) { | |
1458 | error_report("CMD_POSTCOPY_RAM_DISCARD missing nil (%d)", tmp); | |
1459 | return -1; | |
1460 | } | |
1461 | ||
1462 | len -= 3 + strlen(ramid); | |
1463 | if (len % 16) { | |
1464 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1465 | return -1; | |
1466 | } | |
1467 | trace_loadvm_postcopy_ram_handle_discard_header(ramid, len); | |
1468 | while (len) { | |
093e3c42 DDAG |
1469 | uint64_t start_addr, block_length; |
1470 | start_addr = qemu_get_be64(mis->from_src_file); | |
1471 | block_length = qemu_get_be64(mis->from_src_file); | |
1472 | ||
1473 | len -= 16; | |
aaa2064c | 1474 | int ret = ram_discard_range(ramid, start_addr, block_length); |
093e3c42 DDAG |
1475 | if (ret) { |
1476 | return ret; | |
1477 | } | |
093e3c42 DDAG |
1478 | } |
1479 | trace_loadvm_postcopy_ram_handle_discard_end(); | |
1480 | ||
1481 | return 0; | |
1482 | } | |
1483 | ||
c76201ab DDAG |
1484 | /* |
1485 | * Triggered by a postcopy_listen command; this thread takes over reading | |
1486 | * the input stream, leaving the main thread free to carry on loading the rest | |
1487 | * of the device state (from RAM). | |
1488 | * (TODO:This could do with being in a postcopy file - but there again it's | |
1489 | * just another input loop, not that postcopy specific) | |
1490 | */ | |
1491 | static void *postcopy_ram_listen_thread(void *opaque) | |
1492 | { | |
1493 | QEMUFile *f = opaque; | |
1494 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
1495 | int load_res; | |
1496 | ||
6ba996bb DDAG |
1497 | migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE, |
1498 | MIGRATION_STATUS_POSTCOPY_ACTIVE); | |
c76201ab DDAG |
1499 | qemu_sem_post(&mis->listen_thread_sem); |
1500 | trace_postcopy_ram_listen_thread_start(); | |
1501 | ||
1502 | /* | |
1503 | * Because we're a thread and not a coroutine we can't yield | |
1504 | * in qemu_file, and thus we must be blocking now. | |
1505 | */ | |
1506 | qemu_file_set_blocking(f, true); | |
1507 | load_res = qemu_loadvm_state_main(f, mis); | |
1508 | /* And non-blocking again so we don't block in any cleanup */ | |
1509 | qemu_file_set_blocking(f, false); | |
1510 | ||
1511 | trace_postcopy_ram_listen_thread_exit(); | |
1512 | if (load_res < 0) { | |
1513 | error_report("%s: loadvm failed: %d", __func__, load_res); | |
1514 | qemu_file_set_error(f, load_res); | |
6ba996bb DDAG |
1515 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1516 | MIGRATION_STATUS_FAILED); | |
c76201ab DDAG |
1517 | } else { |
1518 | /* | |
1519 | * This looks good, but it's possible that the device loading in the | |
1520 | * main thread hasn't finished yet, and so we might not be in 'RUN' | |
1521 | * state yet; wait for the end of the main thread. | |
1522 | */ | |
1523 | qemu_event_wait(&mis->main_thread_load_event); | |
1524 | } | |
1525 | postcopy_ram_incoming_cleanup(mis); | |
c76201ab DDAG |
1526 | |
1527 | if (load_res < 0) { | |
1528 | /* | |
1529 | * If something went wrong then we have a bad state so exit; | |
1530 | * depending how far we got it might be possible at this point | |
1531 | * to leave the guest running and fire MCEs for pages that never | |
1532 | * arrived as a desperate recovery step. | |
1533 | */ | |
1534 | exit(EXIT_FAILURE); | |
1535 | } | |
1536 | ||
6ba996bb DDAG |
1537 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1538 | MIGRATION_STATUS_COMPLETED); | |
1539 | /* | |
1540 | * If everything has worked fine, then the main thread has waited | |
1541 | * for us to start, and we're the last use of the mis. | |
1542 | * (If something broke then qemu will have to exit anyway since it's | |
1543 | * got a bad migration state). | |
1544 | */ | |
1545 | migration_incoming_state_destroy(); | |
1546 | ||
1547 | ||
c76201ab DDAG |
1548 | return NULL; |
1549 | } | |
1550 | ||
093e3c42 DDAG |
1551 | /* After this message we must be able to immediately receive postcopy data */ |
1552 | static int loadvm_postcopy_handle_listen(MigrationIncomingState *mis) | |
1553 | { | |
1554 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_LISTENING); | |
1555 | trace_loadvm_postcopy_handle_listen(); | |
1556 | if (ps != POSTCOPY_INCOMING_ADVISE && ps != POSTCOPY_INCOMING_DISCARD) { | |
1557 | error_report("CMD_POSTCOPY_LISTEN in wrong postcopy state (%d)", ps); | |
1558 | return -1; | |
1559 | } | |
f9527107 DDAG |
1560 | if (ps == POSTCOPY_INCOMING_ADVISE) { |
1561 | /* | |
1562 | * A rare case, we entered listen without having to do any discards, | |
1563 | * so do the setup that's normally done at the time of the 1st discard. | |
1564 | */ | |
1565 | postcopy_ram_prepare_discard(mis); | |
1566 | } | |
093e3c42 | 1567 | |
f0a227ad DDAG |
1568 | /* |
1569 | * Sensitise RAM - can now generate requests for blocks that don't exist | |
1570 | * However, at this point the CPU shouldn't be running, and the IO | |
1571 | * shouldn't be doing anything yet so don't actually expect requests | |
1572 | */ | |
1573 | if (postcopy_ram_enable_notify(mis)) { | |
1574 | return -1; | |
1575 | } | |
1576 | ||
c76201ab DDAG |
1577 | if (mis->have_listen_thread) { |
1578 | error_report("CMD_POSTCOPY_RAM_LISTEN already has a listen thread"); | |
1579 | return -1; | |
1580 | } | |
1581 | ||
1582 | mis->have_listen_thread = true; | |
1583 | /* Start up the listening thread and wait for it to signal ready */ | |
1584 | qemu_sem_init(&mis->listen_thread_sem, 0); | |
1585 | qemu_thread_create(&mis->listen_thread, "postcopy/listen", | |
1586 | postcopy_ram_listen_thread, mis->from_src_file, | |
a587a3fe | 1587 | QEMU_THREAD_DETACHED); |
c76201ab DDAG |
1588 | qemu_sem_wait(&mis->listen_thread_sem); |
1589 | qemu_sem_destroy(&mis->listen_thread_sem); | |
1590 | ||
093e3c42 DDAG |
1591 | return 0; |
1592 | } | |
1593 | ||
86469922 DL |
1594 | |
1595 | typedef struct { | |
1596 | QEMUBH *bh; | |
1597 | } HandleRunBhData; | |
1598 | ||
ea6a55bc | 1599 | static void loadvm_postcopy_handle_run_bh(void *opaque) |
093e3c42 | 1600 | { |
27c6825b | 1601 | Error *local_err = NULL; |
86469922 | 1602 | HandleRunBhData *data = opaque; |
093e3c42 | 1603 | |
27c6825b DDAG |
1604 | /* TODO we should move all of this lot into postcopy_ram.c or a shared code |
1605 | * in migration.c | |
1606 | */ | |
1607 | cpu_synchronize_all_post_init(); | |
1608 | ||
1609 | qemu_announce_self(); | |
1610 | ||
ace21a58 KW |
1611 | /* Make sure all file formats flush their mutable metadata. |
1612 | * If we get an error here, just don't restart the VM yet. */ | |
27c6825b | 1613 | bdrv_invalidate_cache_all(&local_err); |
0042fd36 | 1614 | if (local_err) { |
ace21a58 | 1615 | error_report_err(local_err); |
0042fd36 KW |
1616 | local_err = NULL; |
1617 | autostart = false; | |
1618 | } | |
1619 | ||
27c6825b DDAG |
1620 | trace_loadvm_postcopy_handle_run_cpu_sync(); |
1621 | cpu_synchronize_all_post_init(); | |
1622 | ||
1623 | trace_loadvm_postcopy_handle_run_vmstart(); | |
1624 | ||
093e3c42 DDAG |
1625 | if (autostart) { |
1626 | /* Hold onto your hats, starting the CPU */ | |
1627 | vm_start(); | |
1628 | } else { | |
1629 | /* leave it paused and let management decide when to start the CPU */ | |
1630 | runstate_set(RUN_STATE_PAUSED); | |
1631 | } | |
1632 | ||
86469922 DL |
1633 | qemu_bh_delete(data->bh); |
1634 | g_free(data); | |
ea6a55bc DL |
1635 | } |
1636 | ||
1637 | /* After all discards we can start running and asking for pages */ | |
1638 | static int loadvm_postcopy_handle_run(MigrationIncomingState *mis) | |
1639 | { | |
1640 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_RUNNING); | |
86469922 | 1641 | HandleRunBhData *data; |
ea6a55bc DL |
1642 | |
1643 | trace_loadvm_postcopy_handle_run(); | |
1644 | if (ps != POSTCOPY_INCOMING_LISTENING) { | |
1645 | error_report("CMD_POSTCOPY_RUN in wrong postcopy state (%d)", ps); | |
1646 | return -1; | |
1647 | } | |
1648 | ||
86469922 DL |
1649 | data = g_new(HandleRunBhData, 1); |
1650 | data->bh = qemu_bh_new(loadvm_postcopy_handle_run_bh, data); | |
1651 | qemu_bh_schedule(data->bh); | |
ea6a55bc | 1652 | |
27c6825b DDAG |
1653 | /* We need to finish reading the stream from the package |
1654 | * and also stop reading anything more from the stream that loaded the | |
1655 | * package (since it's now being read by the listener thread). | |
1656 | * LOADVM_QUIT will quit all the layers of nested loadvm loops. | |
1657 | */ | |
1658 | return LOADVM_QUIT; | |
093e3c42 DDAG |
1659 | } |
1660 | ||
c76ca188 | 1661 | /** |
11cf1d98 DDAG |
1662 | * Immediately following this command is a blob of data containing an embedded |
1663 | * chunk of migration stream; read it and load it. | |
1664 | * | |
1665 | * @mis: Incoming state | |
1666 | * @length: Length of packaged data to read | |
c76ca188 | 1667 | * |
11cf1d98 DDAG |
1668 | * Returns: Negative values on error |
1669 | * | |
1670 | */ | |
1671 | static int loadvm_handle_cmd_packaged(MigrationIncomingState *mis) | |
1672 | { | |
1673 | int ret; | |
61b67d47 DB |
1674 | size_t length; |
1675 | QIOChannelBuffer *bioc; | |
11cf1d98 DDAG |
1676 | |
1677 | length = qemu_get_be32(mis->from_src_file); | |
1678 | trace_loadvm_handle_cmd_packaged(length); | |
1679 | ||
1680 | if (length > MAX_VM_CMD_PACKAGED_SIZE) { | |
61b67d47 | 1681 | error_report("Unreasonably large packaged state: %zu", length); |
11cf1d98 DDAG |
1682 | return -1; |
1683 | } | |
61b67d47 DB |
1684 | |
1685 | bioc = qio_channel_buffer_new(length); | |
6f01f136 | 1686 | qio_channel_set_name(QIO_CHANNEL(bioc), "migration-loadvm-buffer"); |
61b67d47 DB |
1687 | ret = qemu_get_buffer(mis->from_src_file, |
1688 | bioc->data, | |
1689 | length); | |
11cf1d98 | 1690 | if (ret != length) { |
61b67d47 DB |
1691 | object_unref(OBJECT(bioc)); |
1692 | error_report("CMD_PACKAGED: Buffer receive fail ret=%d length=%zu", | |
9af9e0fe | 1693 | ret, length); |
11cf1d98 DDAG |
1694 | return (ret < 0) ? ret : -EAGAIN; |
1695 | } | |
61b67d47 | 1696 | bioc->usage += length; |
11cf1d98 DDAG |
1697 | trace_loadvm_handle_cmd_packaged_received(ret); |
1698 | ||
61b67d47 | 1699 | QEMUFile *packf = qemu_fopen_channel_input(QIO_CHANNEL(bioc)); |
11cf1d98 DDAG |
1700 | |
1701 | ret = qemu_loadvm_state_main(packf, mis); | |
1702 | trace_loadvm_handle_cmd_packaged_main(ret); | |
1703 | qemu_fclose(packf); | |
61b67d47 | 1704 | object_unref(OBJECT(bioc)); |
11cf1d98 DDAG |
1705 | |
1706 | return ret; | |
1707 | } | |
1708 | ||
1709 | /* | |
1710 | * Process an incoming 'QEMU_VM_COMMAND' | |
1711 | * 0 just a normal return | |
1712 | * LOADVM_QUIT All good, but exit the loop | |
1713 | * <0 Error | |
c76ca188 DDAG |
1714 | */ |
1715 | static int loadvm_process_command(QEMUFile *f) | |
1716 | { | |
2e37701e | 1717 | MigrationIncomingState *mis = migration_incoming_get_current(); |
c76ca188 DDAG |
1718 | uint16_t cmd; |
1719 | uint16_t len; | |
2e37701e | 1720 | uint32_t tmp32; |
c76ca188 DDAG |
1721 | |
1722 | cmd = qemu_get_be16(f); | |
1723 | len = qemu_get_be16(f); | |
1724 | ||
1725 | trace_loadvm_process_command(cmd, len); | |
1726 | if (cmd >= MIG_CMD_MAX || cmd == MIG_CMD_INVALID) { | |
1727 | error_report("MIG_CMD 0x%x unknown (len 0x%x)", cmd, len); | |
1728 | return -EINVAL; | |
1729 | } | |
1730 | ||
1731 | if (mig_cmd_args[cmd].len != -1 && mig_cmd_args[cmd].len != len) { | |
1732 | error_report("%s received with bad length - expecting %zu, got %d", | |
1733 | mig_cmd_args[cmd].name, | |
1734 | (size_t)mig_cmd_args[cmd].len, len); | |
1735 | return -ERANGE; | |
1736 | } | |
1737 | ||
1738 | switch (cmd) { | |
2e37701e DDAG |
1739 | case MIG_CMD_OPEN_RETURN_PATH: |
1740 | if (mis->to_src_file) { | |
1741 | error_report("CMD_OPEN_RETURN_PATH called when RP already open"); | |
1742 | /* Not really a problem, so don't give up */ | |
1743 | return 0; | |
1744 | } | |
1745 | mis->to_src_file = qemu_file_get_return_path(f); | |
1746 | if (!mis->to_src_file) { | |
1747 | error_report("CMD_OPEN_RETURN_PATH failed"); | |
1748 | return -1; | |
1749 | } | |
1750 | break; | |
1751 | ||
1752 | case MIG_CMD_PING: | |
1753 | tmp32 = qemu_get_be32(f); | |
1754 | trace_loadvm_process_command_ping(tmp32); | |
1755 | if (!mis->to_src_file) { | |
1756 | error_report("CMD_PING (0x%x) received with no return path", | |
1757 | tmp32); | |
1758 | return -1; | |
1759 | } | |
6decec93 | 1760 | migrate_send_rp_pong(mis, tmp32); |
2e37701e | 1761 | break; |
093e3c42 | 1762 | |
11cf1d98 DDAG |
1763 | case MIG_CMD_PACKAGED: |
1764 | return loadvm_handle_cmd_packaged(mis); | |
1765 | ||
093e3c42 DDAG |
1766 | case MIG_CMD_POSTCOPY_ADVISE: |
1767 | return loadvm_postcopy_handle_advise(mis); | |
1768 | ||
1769 | case MIG_CMD_POSTCOPY_LISTEN: | |
1770 | return loadvm_postcopy_handle_listen(mis); | |
1771 | ||
1772 | case MIG_CMD_POSTCOPY_RUN: | |
1773 | return loadvm_postcopy_handle_run(mis); | |
1774 | ||
1775 | case MIG_CMD_POSTCOPY_RAM_DISCARD: | |
1776 | return loadvm_postcopy_ram_handle_discard(mis, len); | |
c76ca188 DDAG |
1777 | } |
1778 | ||
1779 | return 0; | |
1780 | } | |
1781 | ||
1a8f46f8 | 1782 | struct LoadStateEntry { |
72cf2d4f | 1783 | QLIST_ENTRY(LoadStateEntry) entry; |
a672b469 AL |
1784 | SaveStateEntry *se; |
1785 | int section_id; | |
1786 | int version_id; | |
1a8f46f8 | 1787 | }; |
a672b469 | 1788 | |
59f39a47 DDAG |
1789 | /* |
1790 | * Read a footer off the wire and check that it matches the expected section | |
1791 | * | |
1792 | * Returns: true if the footer was good | |
1793 | * false if there is a problem (and calls error_report to say why) | |
1794 | */ | |
1795 | static bool check_section_footer(QEMUFile *f, LoadStateEntry *le) | |
1796 | { | |
1797 | uint8_t read_mark; | |
1798 | uint32_t read_section_id; | |
1799 | ||
1800 | if (skip_section_footers) { | |
1801 | /* No footer to check */ | |
1802 | return true; | |
1803 | } | |
1804 | ||
1805 | read_mark = qemu_get_byte(f); | |
1806 | ||
1807 | if (read_mark != QEMU_VM_SECTION_FOOTER) { | |
1808 | error_report("Missing section footer for %s", le->se->idstr); | |
1809 | return false; | |
1810 | } | |
1811 | ||
1812 | read_section_id = qemu_get_be32(f); | |
1813 | if (read_section_id != le->section_id) { | |
1814 | error_report("Mismatched section id in footer for %s -" | |
1815 | " read 0x%x expected 0x%x", | |
1816 | le->se->idstr, read_section_id, le->section_id); | |
1817 | return false; | |
1818 | } | |
1819 | ||
1820 | /* All good */ | |
1821 | return true; | |
1822 | } | |
1823 | ||
1a8f46f8 | 1824 | void loadvm_free_handlers(MigrationIncomingState *mis) |
a672b469 | 1825 | { |
f4dbb8dd | 1826 | LoadStateEntry *le, *new_le; |
1a8f46f8 DDAG |
1827 | |
1828 | QLIST_FOREACH_SAFE(le, &mis->loadvm_handlers, entry, new_le) { | |
1829 | QLIST_REMOVE(le, entry); | |
1830 | g_free(le); | |
1831 | } | |
1832 | } | |
1833 | ||
fb3520a8 HZ |
1834 | static int |
1835 | qemu_loadvm_section_start_full(QEMUFile *f, MigrationIncomingState *mis) | |
1836 | { | |
1837 | uint32_t instance_id, version_id, section_id; | |
1838 | SaveStateEntry *se; | |
1839 | LoadStateEntry *le; | |
1840 | char idstr[256]; | |
1841 | int ret; | |
1842 | ||
1843 | /* Read section start */ | |
1844 | section_id = qemu_get_be32(f); | |
1845 | if (!qemu_get_counted_string(f, idstr)) { | |
1846 | error_report("Unable to read ID string for section %u", | |
1847 | section_id); | |
1848 | return -EINVAL; | |
1849 | } | |
1850 | instance_id = qemu_get_be32(f); | |
1851 | version_id = qemu_get_be32(f); | |
1852 | ||
1853 | trace_qemu_loadvm_state_section_startfull(section_id, idstr, | |
1854 | instance_id, version_id); | |
1855 | /* Find savevm section */ | |
1856 | se = find_se(idstr, instance_id); | |
1857 | if (se == NULL) { | |
1858 | error_report("Unknown savevm section or instance '%s' %d", | |
1859 | idstr, instance_id); | |
1860 | return -EINVAL; | |
1861 | } | |
1862 | ||
1863 | /* Validate version */ | |
1864 | if (version_id > se->version_id) { | |
1865 | error_report("savevm: unsupported version %d for '%s' v%d", | |
1866 | version_id, idstr, se->version_id); | |
1867 | return -EINVAL; | |
1868 | } | |
1869 | ||
88c16567 WC |
1870 | /* Validate if it is a device's state */ |
1871 | if (xen_enabled() && se->is_ram) { | |
1872 | error_report("loadvm: %s RAM loading not allowed on Xen", idstr); | |
1873 | return -EINVAL; | |
1874 | } | |
1875 | ||
fb3520a8 HZ |
1876 | /* Add entry */ |
1877 | le = g_malloc0(sizeof(*le)); | |
1878 | ||
1879 | le->se = se; | |
1880 | le->section_id = section_id; | |
1881 | le->version_id = version_id; | |
1882 | QLIST_INSERT_HEAD(&mis->loadvm_handlers, le, entry); | |
1883 | ||
1884 | ret = vmstate_load(f, le->se, le->version_id); | |
1885 | if (ret < 0) { | |
1886 | error_report("error while loading state for instance 0x%x of" | |
1887 | " device '%s'", instance_id, idstr); | |
1888 | return ret; | |
1889 | } | |
1890 | if (!check_section_footer(f, le)) { | |
1891 | return -EINVAL; | |
1892 | } | |
1893 | ||
1894 | return 0; | |
1895 | } | |
1896 | ||
1897 | static int | |
1898 | qemu_loadvm_section_part_end(QEMUFile *f, MigrationIncomingState *mis) | |
1899 | { | |
1900 | uint32_t section_id; | |
1901 | LoadStateEntry *le; | |
1902 | int ret; | |
1903 | ||
1904 | section_id = qemu_get_be32(f); | |
1905 | ||
1906 | trace_qemu_loadvm_state_section_partend(section_id); | |
1907 | QLIST_FOREACH(le, &mis->loadvm_handlers, entry) { | |
1908 | if (le->section_id == section_id) { | |
1909 | break; | |
1910 | } | |
1911 | } | |
1912 | if (le == NULL) { | |
1913 | error_report("Unknown savevm section %d", section_id); | |
1914 | return -EINVAL; | |
1915 | } | |
1916 | ||
1917 | ret = vmstate_load(f, le->se, le->version_id); | |
1918 | if (ret < 0) { | |
1919 | error_report("error while loading state section id %d(%s)", | |
1920 | section_id, le->se->idstr); | |
1921 | return ret; | |
1922 | } | |
1923 | if (!check_section_footer(f, le)) { | |
1924 | return -EINVAL; | |
1925 | } | |
1926 | ||
1927 | return 0; | |
1928 | } | |
1929 | ||
7b89bf27 | 1930 | static int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis) |
1a8f46f8 | 1931 | { |
a672b469 | 1932 | uint8_t section_type; |
ccb783c3 | 1933 | int ret = 0; |
61964c23 | 1934 | |
a672b469 | 1935 | while ((section_type = qemu_get_byte(f)) != QEMU_VM_EOF) { |
ccb783c3 | 1936 | ret = 0; |
a5df2a02 | 1937 | trace_qemu_loadvm_state_section(section_type); |
a672b469 AL |
1938 | switch (section_type) { |
1939 | case QEMU_VM_SECTION_START: | |
1940 | case QEMU_VM_SECTION_FULL: | |
fb3520a8 | 1941 | ret = qemu_loadvm_section_start_full(f, mis); |
b5a22e4a | 1942 | if (ret < 0) { |
ccb783c3 | 1943 | goto out; |
b5a22e4a | 1944 | } |
a672b469 AL |
1945 | break; |
1946 | case QEMU_VM_SECTION_PART: | |
1947 | case QEMU_VM_SECTION_END: | |
fb3520a8 | 1948 | ret = qemu_loadvm_section_part_end(f, mis); |
b5a22e4a | 1949 | if (ret < 0) { |
ccb783c3 | 1950 | goto out; |
b5a22e4a | 1951 | } |
a672b469 | 1952 | break; |
c76ca188 DDAG |
1953 | case QEMU_VM_COMMAND: |
1954 | ret = loadvm_process_command(f); | |
7b89bf27 DDAG |
1955 | trace_qemu_loadvm_state_section_command(ret); |
1956 | if ((ret < 0) || (ret & LOADVM_QUIT)) { | |
ccb783c3 | 1957 | goto out; |
c76ca188 DDAG |
1958 | } |
1959 | break; | |
a672b469 | 1960 | default: |
6a64b644 | 1961 | error_report("Unknown savevm section type %d", section_type); |
ccb783c3 DDAG |
1962 | ret = -EINVAL; |
1963 | goto out; | |
a672b469 AL |
1964 | } |
1965 | } | |
1966 | ||
ccb783c3 DDAG |
1967 | out: |
1968 | if (ret < 0) { | |
1969 | qemu_file_set_error(f, ret); | |
1970 | } | |
1971 | return ret; | |
7b89bf27 DDAG |
1972 | } |
1973 | ||
1974 | int qemu_loadvm_state(QEMUFile *f) | |
1975 | { | |
1976 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
1977 | Error *local_err = NULL; | |
1978 | unsigned int v; | |
1979 | int ret; | |
1980 | ||
1981 | if (qemu_savevm_state_blocked(&local_err)) { | |
1982 | error_report_err(local_err); | |
1983 | return -EINVAL; | |
1984 | } | |
1985 | ||
1986 | v = qemu_get_be32(f); | |
1987 | if (v != QEMU_VM_FILE_MAGIC) { | |
1988 | error_report("Not a migration stream"); | |
1989 | return -EINVAL; | |
1990 | } | |
1991 | ||
1992 | v = qemu_get_be32(f); | |
1993 | if (v == QEMU_VM_FILE_VERSION_COMPAT) { | |
1994 | error_report("SaveVM v2 format is obsolete and don't work anymore"); | |
1995 | return -ENOTSUP; | |
1996 | } | |
1997 | if (v != QEMU_VM_FILE_VERSION) { | |
1998 | error_report("Unsupported migration stream version"); | |
1999 | return -ENOTSUP; | |
2000 | } | |
2001 | ||
902c053d | 2002 | if (!savevm_state.skip_configuration || enforce_config_section()) { |
7b89bf27 DDAG |
2003 | if (qemu_get_byte(f) != QEMU_VM_CONFIGURATION) { |
2004 | error_report("Configuration section missing"); | |
2005 | return -EINVAL; | |
2006 | } | |
2007 | ret = vmstate_load_state(f, &vmstate_configuration, &savevm_state, 0); | |
2008 | ||
2009 | if (ret) { | |
2010 | return ret; | |
2011 | } | |
2012 | } | |
2013 | ||
2014 | ret = qemu_loadvm_state_main(f, mis); | |
2015 | qemu_event_set(&mis->main_thread_load_event); | |
2016 | ||
2017 | trace_qemu_loadvm_state_post_main(ret); | |
2018 | ||
c76201ab DDAG |
2019 | if (mis->have_listen_thread) { |
2020 | /* Listen thread still going, can't clean up yet */ | |
2021 | return ret; | |
2022 | } | |
2023 | ||
7b89bf27 DDAG |
2024 | if (ret == 0) { |
2025 | ret = qemu_file_get_error(f); | |
2026 | } | |
1925cebc AG |
2027 | |
2028 | /* | |
2029 | * Try to read in the VMDESC section as well, so that dumping tools that | |
2030 | * intercept our migration stream have the chance to see it. | |
2031 | */ | |
1aca9a5f DDAG |
2032 | |
2033 | /* We've got to be careful; if we don't read the data and just shut the fd | |
2034 | * then the sender can error if we close while it's still sending. | |
2035 | * We also mustn't read data that isn't there; some transports (RDMA) | |
2036 | * will stall waiting for that data when the source has already closed. | |
2037 | */ | |
7b89bf27 | 2038 | if (ret == 0 && should_send_vmdesc()) { |
1aca9a5f DDAG |
2039 | uint8_t *buf; |
2040 | uint32_t size; | |
7b89bf27 | 2041 | uint8_t section_type = qemu_get_byte(f); |
1aca9a5f DDAG |
2042 | |
2043 | if (section_type != QEMU_VM_VMDESCRIPTION) { | |
2044 | error_report("Expected vmdescription section, but got %d", | |
2045 | section_type); | |
2046 | /* | |
2047 | * It doesn't seem worth failing at this point since | |
2048 | * we apparently have an otherwise valid VM state | |
2049 | */ | |
2050 | } else { | |
2051 | buf = g_malloc(0x1000); | |
2052 | size = qemu_get_be32(f); | |
2053 | ||
2054 | while (size > 0) { | |
2055 | uint32_t read_chunk = MIN(size, 0x1000); | |
2056 | qemu_get_buffer(f, buf, read_chunk); | |
2057 | size -= read_chunk; | |
2058 | } | |
2059 | g_free(buf); | |
1925cebc | 2060 | } |
1925cebc AG |
2061 | } |
2062 | ||
ea375f9a JK |
2063 | cpu_synchronize_all_post_init(); |
2064 | ||
a672b469 AL |
2065 | return ret; |
2066 | } | |
2067 | ||
927d6638 | 2068 | int save_vmstate(const char *name, Error **errp) |
a672b469 AL |
2069 | { |
2070 | BlockDriverState *bs, *bs1; | |
2071 | QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1; | |
ac8c19ba | 2072 | int ret = -1; |
a672b469 AL |
2073 | QEMUFile *f; |
2074 | int saved_vm_running; | |
c2c9a466 | 2075 | uint64_t vm_state_size; |
68b891ec | 2076 | qemu_timeval tv; |
7d631a11 | 2077 | struct tm tm; |
79b3c12a | 2078 | AioContext *aio_context; |
a672b469 | 2079 | |
e9ff957a | 2080 | if (!bdrv_all_can_snapshot(&bs)) { |
927d6638 JQ |
2081 | error_setg(errp, "Device '%s' is writable but does not support " |
2082 | "snapshots", bdrv_get_device_name(bs)); | |
ac8c19ba | 2083 | return ret; |
feeee5ac MDCF |
2084 | } |
2085 | ||
0b461605 | 2086 | /* Delete old snapshots of the same name */ |
ac8c19ba | 2087 | if (name) { |
927d6638 | 2088 | ret = bdrv_all_delete_snapshot(name, &bs1, errp); |
ac8c19ba | 2089 | if (ret < 0) { |
927d6638 JQ |
2090 | error_prepend(errp, "Error while deleting snapshot on device " |
2091 | "'%s': ", bdrv_get_device_name(bs1)); | |
ac8c19ba PD |
2092 | return ret; |
2093 | } | |
0b461605 DL |
2094 | } |
2095 | ||
7cb14481 DL |
2096 | bs = bdrv_all_find_vmstate_bs(); |
2097 | if (bs == NULL) { | |
927d6638 | 2098 | error_setg(errp, "No block device can accept snapshots"); |
ac8c19ba | 2099 | return ret; |
a672b469 | 2100 | } |
79b3c12a | 2101 | aio_context = bdrv_get_aio_context(bs); |
a672b469 | 2102 | |
1354869c | 2103 | saved_vm_running = runstate_is_running(); |
560d027b JQ |
2104 | |
2105 | ret = global_state_store(); | |
2106 | if (ret) { | |
927d6638 | 2107 | error_setg(errp, "Error saving global state"); |
ac8c19ba | 2108 | return ret; |
560d027b | 2109 | } |
0461d5a6 | 2110 | vm_stop(RUN_STATE_SAVE_VM); |
a672b469 | 2111 | |
79b3c12a DL |
2112 | aio_context_acquire(aio_context); |
2113 | ||
cb499fb2 | 2114 | memset(sn, 0, sizeof(*sn)); |
a672b469 AL |
2115 | |
2116 | /* fill auxiliary fields */ | |
68b891ec | 2117 | qemu_gettimeofday(&tv); |
a672b469 AL |
2118 | sn->date_sec = tv.tv_sec; |
2119 | sn->date_nsec = tv.tv_usec * 1000; | |
bc72ad67 | 2120 | sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL); |
a672b469 | 2121 | |
7d631a11 MDCF |
2122 | if (name) { |
2123 | ret = bdrv_snapshot_find(bs, old_sn, name); | |
2124 | if (ret >= 0) { | |
2125 | pstrcpy(sn->name, sizeof(sn->name), old_sn->name); | |
2126 | pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str); | |
2127 | } else { | |
2128 | pstrcpy(sn->name, sizeof(sn->name), name); | |
2129 | } | |
2130 | } else { | |
d7d9b528 BS |
2131 | /* cast below needed for OpenBSD where tv_sec is still 'long' */ |
2132 | localtime_r((const time_t *)&tv.tv_sec, &tm); | |
7d631a11 | 2133 | strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm); |
7d631a11 MDCF |
2134 | } |
2135 | ||
a672b469 | 2136 | /* save the VM state */ |
45566e9c | 2137 | f = qemu_fopen_bdrv(bs, 1); |
a672b469 | 2138 | if (!f) { |
927d6638 | 2139 | error_setg(errp, "Could not open VM state file"); |
a672b469 AL |
2140 | goto the_end; |
2141 | } | |
927d6638 | 2142 | ret = qemu_savevm_state(f, errp); |
2d22b18f | 2143 | vm_state_size = qemu_ftell(f); |
a672b469 AL |
2144 | qemu_fclose(f); |
2145 | if (ret < 0) { | |
a672b469 AL |
2146 | goto the_end; |
2147 | } | |
2148 | ||
a9085f9b DL |
2149 | ret = bdrv_all_create_snapshot(sn, bs, vm_state_size, &bs); |
2150 | if (ret < 0) { | |
927d6638 JQ |
2151 | error_setg(errp, "Error while creating snapshot on '%s'", |
2152 | bdrv_get_device_name(bs)); | |
ac8c19ba | 2153 | goto the_end; |
a672b469 AL |
2154 | } |
2155 | ||
ac8c19ba PD |
2156 | ret = 0; |
2157 | ||
a672b469 | 2158 | the_end: |
79b3c12a | 2159 | aio_context_release(aio_context); |
38ff78d3 | 2160 | if (saved_vm_running) { |
a672b469 | 2161 | vm_start(); |
38ff78d3 | 2162 | } |
ac8c19ba PD |
2163 | return ret; |
2164 | } | |
2165 | ||
a7ae8355 SS |
2166 | void qmp_xen_save_devices_state(const char *filename, Error **errp) |
2167 | { | |
2168 | QEMUFile *f; | |
8925839f | 2169 | QIOChannelFile *ioc; |
a7ae8355 SS |
2170 | int saved_vm_running; |
2171 | int ret; | |
2172 | ||
2173 | saved_vm_running = runstate_is_running(); | |
2174 | vm_stop(RUN_STATE_SAVE_VM); | |
c69adea4 | 2175 | global_state_store_running(); |
a7ae8355 | 2176 | |
8925839f DB |
2177 | ioc = qio_channel_file_new_path(filename, O_WRONLY | O_CREAT, 0660, errp); |
2178 | if (!ioc) { | |
a7ae8355 SS |
2179 | goto the_end; |
2180 | } | |
6f01f136 | 2181 | qio_channel_set_name(QIO_CHANNEL(ioc), "migration-xen-save-state"); |
8925839f | 2182 | f = qemu_fopen_channel_output(QIO_CHANNEL(ioc)); |
a7ae8355 SS |
2183 | ret = qemu_save_device_state(f); |
2184 | qemu_fclose(f); | |
2185 | if (ret < 0) { | |
c6bd8c70 | 2186 | error_setg(errp, QERR_IO_ERROR); |
a7ae8355 SS |
2187 | } |
2188 | ||
2189 | the_end: | |
38ff78d3 | 2190 | if (saved_vm_running) { |
a7ae8355 | 2191 | vm_start(); |
38ff78d3 | 2192 | } |
a7ae8355 SS |
2193 | } |
2194 | ||
88c16567 WC |
2195 | void qmp_xen_load_devices_state(const char *filename, Error **errp) |
2196 | { | |
2197 | QEMUFile *f; | |
2198 | QIOChannelFile *ioc; | |
2199 | int ret; | |
2200 | ||
2201 | /* Guest must be paused before loading the device state; the RAM state | |
2202 | * will already have been loaded by xc | |
2203 | */ | |
2204 | if (runstate_is_running()) { | |
2205 | error_setg(errp, "Cannot update device state while vm is running"); | |
2206 | return; | |
2207 | } | |
2208 | vm_stop(RUN_STATE_RESTORE_VM); | |
2209 | ||
2210 | ioc = qio_channel_file_new_path(filename, O_RDONLY | O_BINARY, 0, errp); | |
2211 | if (!ioc) { | |
2212 | return; | |
2213 | } | |
6f01f136 | 2214 | qio_channel_set_name(QIO_CHANNEL(ioc), "migration-xen-load-state"); |
88c16567 WC |
2215 | f = qemu_fopen_channel_input(QIO_CHANNEL(ioc)); |
2216 | ||
88c16567 WC |
2217 | ret = qemu_loadvm_state(f); |
2218 | qemu_fclose(f); | |
2219 | if (ret < 0) { | |
2220 | error_setg(errp, QERR_IO_ERROR); | |
2221 | } | |
2222 | migration_incoming_state_destroy(); | |
2223 | } | |
2224 | ||
927d6638 | 2225 | int load_vmstate(const char *name, Error **errp) |
a672b469 | 2226 | { |
f0aa7a8b | 2227 | BlockDriverState *bs, *bs_vm_state; |
2d22b18f | 2228 | QEMUSnapshotInfo sn; |
a672b469 | 2229 | QEMUFile *f; |
751c6a17 | 2230 | int ret; |
79b3c12a | 2231 | AioContext *aio_context; |
b4b076da | 2232 | MigrationIncomingState *mis = migration_incoming_get_current(); |
a672b469 | 2233 | |
849f96e2 | 2234 | if (!bdrv_all_can_snapshot(&bs)) { |
927d6638 JQ |
2235 | error_setg(errp, |
2236 | "Device '%s' is writable but does not support snapshots", | |
2237 | bdrv_get_device_name(bs)); | |
849f96e2 DL |
2238 | return -ENOTSUP; |
2239 | } | |
723ccda1 DL |
2240 | ret = bdrv_all_find_snapshot(name, &bs); |
2241 | if (ret < 0) { | |
927d6638 JQ |
2242 | error_setg(errp, |
2243 | "Device '%s' does not have the requested snapshot '%s'", | |
2244 | bdrv_get_device_name(bs), name); | |
723ccda1 DL |
2245 | return ret; |
2246 | } | |
849f96e2 | 2247 | |
7cb14481 | 2248 | bs_vm_state = bdrv_all_find_vmstate_bs(); |
f0aa7a8b | 2249 | if (!bs_vm_state) { |
927d6638 | 2250 | error_setg(errp, "No block device supports snapshots"); |
f0aa7a8b MDCF |
2251 | return -ENOTSUP; |
2252 | } | |
79b3c12a | 2253 | aio_context = bdrv_get_aio_context(bs_vm_state); |
f0aa7a8b MDCF |
2254 | |
2255 | /* Don't even try to load empty VM states */ | |
79b3c12a | 2256 | aio_context_acquire(aio_context); |
f0aa7a8b | 2257 | ret = bdrv_snapshot_find(bs_vm_state, &sn, name); |
79b3c12a | 2258 | aio_context_release(aio_context); |
f0aa7a8b MDCF |
2259 | if (ret < 0) { |
2260 | return ret; | |
2261 | } else if (sn.vm_state_size == 0) { | |
927d6638 JQ |
2262 | error_setg(errp, "This is a disk-only snapshot. Revert to it " |
2263 | " offline using qemu-img"); | |
f0aa7a8b MDCF |
2264 | return -EINVAL; |
2265 | } | |
2266 | ||
a672b469 | 2267 | /* Flush all IO requests so they don't interfere with the new state. */ |
922453bc | 2268 | bdrv_drain_all(); |
a672b469 | 2269 | |
4c1cdbaa DL |
2270 | ret = bdrv_all_goto_snapshot(name, &bs); |
2271 | if (ret < 0) { | |
927d6638 | 2272 | error_setg(errp, "Error %d while activating snapshot '%s' on '%s'", |
4c1cdbaa DL |
2273 | ret, name, bdrv_get_device_name(bs)); |
2274 | return ret; | |
a672b469 AL |
2275 | } |
2276 | ||
a672b469 | 2277 | /* restore the VM state */ |
f0aa7a8b | 2278 | f = qemu_fopen_bdrv(bs_vm_state, 0); |
a672b469 | 2279 | if (!f) { |
927d6638 | 2280 | error_setg(errp, "Could not open VM state file"); |
05f2401e | 2281 | return -EINVAL; |
a672b469 | 2282 | } |
f0aa7a8b | 2283 | |
5a8a49d7 | 2284 | qemu_system_reset(VMRESET_SILENT); |
b4b076da | 2285 | mis->from_src_file = f; |
f0aa7a8b | 2286 | |
79b3c12a DL |
2287 | aio_context_acquire(aio_context); |
2288 | ret = qemu_loadvm_state(f); | |
a672b469 | 2289 | qemu_fclose(f); |
79b3c12a DL |
2290 | aio_context_release(aio_context); |
2291 | ||
bca7856a | 2292 | migration_incoming_state_destroy(); |
a672b469 | 2293 | if (ret < 0) { |
927d6638 | 2294 | error_setg(errp, "Error %d while loading VM state", ret); |
05f2401e | 2295 | return ret; |
a672b469 | 2296 | } |
f0aa7a8b | 2297 | |
05f2401e | 2298 | return 0; |
7b630349 JQ |
2299 | } |
2300 | ||
c5705a77 AK |
2301 | void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev) |
2302 | { | |
fa53a0e5 | 2303 | qemu_ram_set_idstr(mr->ram_block, |
c5705a77 AK |
2304 | memory_region_name(mr), dev); |
2305 | } | |
2306 | ||
2307 | void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev) | |
2308 | { | |
fa53a0e5 | 2309 | qemu_ram_unset_idstr(mr->ram_block); |
c5705a77 AK |
2310 | } |
2311 | ||
2312 | void vmstate_register_ram_global(MemoryRegion *mr) | |
2313 | { | |
2314 | vmstate_register_ram(mr, NULL); | |
2315 | } | |
1bfe5f05 JQ |
2316 | |
2317 | bool vmstate_check_only_migratable(const VMStateDescription *vmsd) | |
2318 | { | |
2319 | /* check needed if --only-migratable is specified */ | |
2320 | if (!only_migratable) { | |
2321 | return true; | |
2322 | } | |
2323 | ||
2324 | return !(vmsd && vmsd->unmigratable); | |
2325 | } |