]>
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" |
abfd9ce3 | 30 | #include "hw/boards.h" |
88c16567 | 31 | #include "hw/xen/xen.h" |
1422e32d | 32 | #include "net/net.h" |
6666c96a | 33 | #include "migration.h" |
5e22479a | 34 | #include "migration/snapshot.h" |
f8d806c9 | 35 | #include "migration/misc.h" |
f2a8f0a6 | 36 | #include "migration/register.h" |
84a899de | 37 | #include "migration/global_state.h" |
7b1e1a22 | 38 | #include "ram.h" |
40014d81 | 39 | #include "qemu-file-channel.h" |
08a0aee1 | 40 | #include "qemu-file.h" |
20a519a0 | 41 | #include "savevm.h" |
be07b0ac | 42 | #include "postcopy-ram.h" |
e688df6b | 43 | #include "qapi/error.h" |
112ed241 MA |
44 | #include "qapi/qapi-commands-migration.h" |
45 | #include "qapi/qapi-commands-misc.h" | |
cc7a8ea7 | 46 | #include "qapi/qmp/qerror.h" |
d49b6836 | 47 | #include "qemu/error-report.h" |
9c17d615 | 48 | #include "sysemu/cpus.h" |
022c62cb | 49 | #include "exec/memory.h" |
51180423 | 50 | #include "exec/target_page.h" |
517a13c9 | 51 | #include "trace.h" |
28085f7b | 52 | #include "qemu/iov.h" |
de08c606 | 53 | #include "block/snapshot.h" |
f348b6d1 | 54 | #include "qemu/cutils.h" |
61b67d47 | 55 | #include "io/channel-buffer.h" |
8925839f | 56 | #include "io/channel-file.h" |
377b21cc | 57 | #include "sysemu/replay.h" |
83ee768d | 58 | #include "qjson.h" |
aad555c2 | 59 | #include "migration/colo.h" |
a672b469 | 60 | |
18995b98 | 61 | #ifndef ETH_P_RARP |
f8778a77 | 62 | #define ETH_P_RARP 0x8035 |
18995b98 N |
63 | #endif |
64 | #define ARP_HTYPE_ETH 0x0001 | |
65 | #define ARP_PTYPE_IP 0x0800 | |
66 | #define ARP_OP_REQUEST_REV 0x3 | |
67 | ||
093e3c42 DDAG |
68 | const unsigned int postcopy_ram_discard_version = 0; |
69 | ||
20a519a0 JQ |
70 | /* Subcommands for QEMU_VM_COMMAND */ |
71 | enum qemu_vm_cmd { | |
72 | MIG_CMD_INVALID = 0, /* Must be 0 */ | |
73 | MIG_CMD_OPEN_RETURN_PATH, /* Tell the dest to open the Return path */ | |
74 | MIG_CMD_PING, /* Request a PONG on the RP */ | |
75 | ||
76 | MIG_CMD_POSTCOPY_ADVISE, /* Prior to any page transfers, just | |
77 | warn we might want to do PC */ | |
78 | MIG_CMD_POSTCOPY_LISTEN, /* Start listening for incoming | |
79 | pages as it's running. */ | |
80 | MIG_CMD_POSTCOPY_RUN, /* Start execution */ | |
81 | ||
82 | MIG_CMD_POSTCOPY_RAM_DISCARD, /* A list of pages to discard that | |
83 | were previously sent during | |
84 | precopy but are dirty. */ | |
85 | MIG_CMD_PACKAGED, /* Send a wrapped stream within this stream */ | |
aad555c2 | 86 | MIG_CMD_ENABLE_COLO, /* Enable COLO */ |
858b6d62 | 87 | MIG_CMD_POSTCOPY_RESUME, /* resume postcopy on dest */ |
f25d4225 | 88 | MIG_CMD_RECV_BITMAP, /* Request for recved bitmap on dst */ |
20a519a0 JQ |
89 | MIG_CMD_MAX |
90 | }; | |
91 | ||
ee555cdf | 92 | #define MAX_VM_CMD_PACKAGED_SIZE UINT32_MAX |
c76ca188 DDAG |
93 | static struct mig_cmd_args { |
94 | ssize_t len; /* -1 = variable */ | |
95 | const char *name; | |
96 | } mig_cmd_args[] = { | |
97 | [MIG_CMD_INVALID] = { .len = -1, .name = "INVALID" }, | |
2e37701e DDAG |
98 | [MIG_CMD_OPEN_RETURN_PATH] = { .len = 0, .name = "OPEN_RETURN_PATH" }, |
99 | [MIG_CMD_PING] = { .len = sizeof(uint32_t), .name = "PING" }, | |
58110f0a | 100 | [MIG_CMD_POSTCOPY_ADVISE] = { .len = -1, .name = "POSTCOPY_ADVISE" }, |
093e3c42 DDAG |
101 | [MIG_CMD_POSTCOPY_LISTEN] = { .len = 0, .name = "POSTCOPY_LISTEN" }, |
102 | [MIG_CMD_POSTCOPY_RUN] = { .len = 0, .name = "POSTCOPY_RUN" }, | |
103 | [MIG_CMD_POSTCOPY_RAM_DISCARD] = { | |
104 | .len = -1, .name = "POSTCOPY_RAM_DISCARD" }, | |
3f5875ec | 105 | [MIG_CMD_POSTCOPY_RESUME] = { .len = 0, .name = "POSTCOPY_RESUME" }, |
11cf1d98 | 106 | [MIG_CMD_PACKAGED] = { .len = 4, .name = "PACKAGED" }, |
f25d4225 | 107 | [MIG_CMD_RECV_BITMAP] = { .len = -1, .name = "RECV_BITMAP" }, |
c76ca188 DDAG |
108 | [MIG_CMD_MAX] = { .len = -1, .name = "MAX" }, |
109 | }; | |
110 | ||
58110f0a VSO |
111 | /* Note for MIG_CMD_POSTCOPY_ADVISE: |
112 | * The format of arguments is depending on postcopy mode: | |
113 | * - postcopy RAM only | |
114 | * uint64_t host page size | |
115 | * uint64_t taget page size | |
116 | * | |
117 | * - postcopy RAM and postcopy dirty bitmaps | |
118 | * format is the same as for postcopy RAM only | |
119 | * | |
120 | * - postcopy dirty bitmaps only | |
121 | * Nothing. Command length field is 0. | |
122 | * | |
123 | * Be careful: adding a new postcopy entity with some other parameters should | |
124 | * not break format self-description ability. Good way is to introduce some | |
125 | * generic extendable format with an exception for two old entities. | |
126 | */ | |
127 | ||
18995b98 | 128 | static int announce_self_create(uint8_t *buf, |
5cecf414 | 129 | uint8_t *mac_addr) |
a672b469 | 130 | { |
18995b98 N |
131 | /* Ethernet header. */ |
132 | memset(buf, 0xff, 6); /* destination MAC addr */ | |
133 | memcpy(buf + 6, mac_addr, 6); /* source MAC addr */ | |
134 | *(uint16_t *)(buf + 12) = htons(ETH_P_RARP); /* ethertype */ | |
135 | ||
136 | /* RARP header. */ | |
137 | *(uint16_t *)(buf + 14) = htons(ARP_HTYPE_ETH); /* hardware addr space */ | |
138 | *(uint16_t *)(buf + 16) = htons(ARP_PTYPE_IP); /* protocol addr space */ | |
139 | *(buf + 18) = 6; /* hardware addr length (ethernet) */ | |
140 | *(buf + 19) = 4; /* protocol addr length (IPv4) */ | |
141 | *(uint16_t *)(buf + 20) = htons(ARP_OP_REQUEST_REV); /* opcode */ | |
142 | memcpy(buf + 22, mac_addr, 6); /* source hw addr */ | |
143 | memset(buf + 28, 0x00, 4); /* source protocol addr */ | |
144 | memcpy(buf + 32, mac_addr, 6); /* target hw addr */ | |
145 | memset(buf + 38, 0x00, 4); /* target protocol addr */ | |
146 | ||
147 | /* Padding to get up to 60 bytes (ethernet min packet size, minus FCS). */ | |
148 | memset(buf + 42, 0x00, 18); | |
149 | ||
150 | return 60; /* len (FCS will be added by hardware) */ | |
a672b469 AL |
151 | } |
152 | ||
f401ca22 | 153 | static void qemu_announce_self_iter(NICState *nic, void *opaque) |
a672b469 | 154 | { |
18995b98 | 155 | uint8_t buf[60]; |
f401ca22 MM |
156 | int len; |
157 | ||
9013dca5 | 158 | trace_qemu_announce_self_iter(qemu_ether_ntoa(&nic->conf->macaddr)); |
f401ca22 MM |
159 | len = announce_self_create(buf, nic->conf->macaddr.a); |
160 | ||
b356f76d | 161 | qemu_send_packet_raw(qemu_get_queue(nic), buf, len); |
f401ca22 MM |
162 | } |
163 | ||
164 | ||
165 | static void qemu_announce_self_once(void *opaque) | |
166 | { | |
ed8b330b GN |
167 | static int count = SELF_ANNOUNCE_ROUNDS; |
168 | QEMUTimer *timer = *(QEMUTimer **)opaque; | |
a672b469 | 169 | |
f401ca22 MM |
170 | qemu_foreach_nic(qemu_announce_self_iter, NULL); |
171 | ||
18995b98 N |
172 | if (--count) { |
173 | /* delay 50ms, 150ms, 250ms, ... */ | |
bc72ad67 | 174 | timer_mod(timer, qemu_clock_get_ms(QEMU_CLOCK_REALTIME) + |
508e1180 | 175 | self_announce_delay(count)); |
ed8b330b | 176 | } else { |
5cecf414 EH |
177 | timer_del(timer); |
178 | timer_free(timer); | |
ed8b330b GN |
179 | } |
180 | } | |
181 | ||
182 | void qemu_announce_self(void) | |
183 | { | |
5cecf414 EH |
184 | static QEMUTimer *timer; |
185 | timer = timer_new_ms(QEMU_CLOCK_REALTIME, qemu_announce_self_once, &timer); | |
186 | qemu_announce_self_once(&timer); | |
a672b469 AL |
187 | } |
188 | ||
189 | /***********************************************************/ | |
190 | /* savevm/loadvm support */ | |
191 | ||
05fcc848 KW |
192 | static ssize_t block_writev_buffer(void *opaque, struct iovec *iov, int iovcnt, |
193 | int64_t pos) | |
194 | { | |
195 | int ret; | |
196 | QEMUIOVector qiov; | |
197 | ||
198 | qemu_iovec_init_external(&qiov, iov, iovcnt); | |
199 | ret = bdrv_writev_vmstate(opaque, &qiov, pos); | |
200 | if (ret < 0) { | |
201 | return ret; | |
202 | } | |
203 | ||
204 | return qiov.size; | |
205 | } | |
206 | ||
a202a4c0 DDAG |
207 | static ssize_t block_get_buffer(void *opaque, uint8_t *buf, int64_t pos, |
208 | size_t size) | |
a672b469 | 209 | { |
45566e9c | 210 | return bdrv_load_vmstate(opaque, buf, pos, size); |
a672b469 AL |
211 | } |
212 | ||
213 | static int bdrv_fclose(void *opaque) | |
214 | { | |
ad492c92 | 215 | return bdrv_flush(opaque); |
a672b469 AL |
216 | } |
217 | ||
9229bf3c PB |
218 | static const QEMUFileOps bdrv_read_ops = { |
219 | .get_buffer = block_get_buffer, | |
220 | .close = bdrv_fclose | |
221 | }; | |
222 | ||
223 | static const QEMUFileOps bdrv_write_ops = { | |
05fcc848 KW |
224 | .writev_buffer = block_writev_buffer, |
225 | .close = bdrv_fclose | |
9229bf3c PB |
226 | }; |
227 | ||
45566e9c | 228 | static QEMUFile *qemu_fopen_bdrv(BlockDriverState *bs, int is_writable) |
a672b469 | 229 | { |
38ff78d3 | 230 | if (is_writable) { |
9229bf3c | 231 | return qemu_fopen_ops(bs, &bdrv_write_ops); |
38ff78d3 | 232 | } |
9229bf3c | 233 | return qemu_fopen_ops(bs, &bdrv_read_ops); |
a672b469 AL |
234 | } |
235 | ||
2ff68d07 | 236 | |
bb1a6d8c EH |
237 | /* QEMUFile timer support. |
238 | * Not in qemu-file.c to not add qemu-timer.c as dependency to qemu-file.c | |
239 | */ | |
2ff68d07 | 240 | |
40daca54 | 241 | void timer_put(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
242 | { |
243 | uint64_t expire_time; | |
244 | ||
e93379b0 | 245 | expire_time = timer_expire_time_ns(ts); |
2ff68d07 PB |
246 | qemu_put_be64(f, expire_time); |
247 | } | |
248 | ||
40daca54 | 249 | void timer_get(QEMUFile *f, QEMUTimer *ts) |
2ff68d07 PB |
250 | { |
251 | uint64_t expire_time; | |
252 | ||
253 | expire_time = qemu_get_be64(f); | |
254 | if (expire_time != -1) { | |
bc72ad67 | 255 | timer_mod_ns(ts, expire_time); |
2ff68d07 | 256 | } else { |
bc72ad67 | 257 | timer_del(ts); |
2ff68d07 PB |
258 | } |
259 | } | |
260 | ||
261 | ||
bb1a6d8c EH |
262 | /* VMState timer support. |
263 | * Not in vmstate.c to not add qemu-timer.c as dependency to vmstate.c | |
264 | */ | |
dde0463b | 265 | |
2c21ee76 | 266 | static int get_timer(QEMUFile *f, void *pv, size_t size, VMStateField *field) |
dde0463b JQ |
267 | { |
268 | QEMUTimer *v = pv; | |
40daca54 | 269 | timer_get(f, v); |
dde0463b JQ |
270 | return 0; |
271 | } | |
272 | ||
2c21ee76 JD |
273 | static int put_timer(QEMUFile *f, void *pv, size_t size, VMStateField *field, |
274 | QJSON *vmdesc) | |
dde0463b | 275 | { |
84e2e3eb | 276 | QEMUTimer *v = pv; |
40daca54 | 277 | timer_put(f, v); |
2c21ee76 JD |
278 | |
279 | return 0; | |
dde0463b JQ |
280 | } |
281 | ||
282 | const VMStateInfo vmstate_info_timer = { | |
283 | .name = "timer", | |
284 | .get = get_timer, | |
285 | .put = put_timer, | |
286 | }; | |
287 | ||
08e99e29 | 288 | |
7685ee6a AW |
289 | typedef struct CompatEntry { |
290 | char idstr[256]; | |
291 | int instance_id; | |
292 | } CompatEntry; | |
293 | ||
a672b469 | 294 | typedef struct SaveStateEntry { |
72cf2d4f | 295 | QTAILQ_ENTRY(SaveStateEntry) entry; |
a672b469 AL |
296 | char idstr[256]; |
297 | int instance_id; | |
4d2ffa08 | 298 | int alias_id; |
a672b469 | 299 | int version_id; |
0f42f657 JQ |
300 | /* version id read from the stream */ |
301 | int load_version_id; | |
a672b469 | 302 | int section_id; |
0f42f657 JQ |
303 | /* section id read from the stream */ |
304 | int load_section_id; | |
22ea40f4 | 305 | SaveVMHandlers *ops; |
9ed7d6ae | 306 | const VMStateDescription *vmsd; |
a672b469 | 307 | void *opaque; |
7685ee6a | 308 | CompatEntry *compat; |
a7ae8355 | 309 | int is_ram; |
a672b469 AL |
310 | } SaveStateEntry; |
311 | ||
0163a2e0 JQ |
312 | typedef struct SaveState { |
313 | QTAILQ_HEAD(, SaveStateEntry) handlers; | |
314 | int global_section_id; | |
61964c23 JQ |
315 | uint32_t len; |
316 | const char *name; | |
59811a32 | 317 | uint32_t target_page_bits; |
0163a2e0 JQ |
318 | } SaveState; |
319 | ||
320 | static SaveState savevm_state = { | |
321 | .handlers = QTAILQ_HEAD_INITIALIZER(savevm_state.handlers), | |
322 | .global_section_id = 0, | |
61964c23 JQ |
323 | }; |
324 | ||
44b1ff31 | 325 | static int configuration_pre_save(void *opaque) |
61964c23 JQ |
326 | { |
327 | SaveState *state = opaque; | |
328 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
329 | ||
330 | state->len = strlen(current_name); | |
331 | state->name = current_name; | |
46d702b1 | 332 | state->target_page_bits = qemu_target_page_bits(); |
44b1ff31 DDAG |
333 | |
334 | return 0; | |
59811a32 PM |
335 | } |
336 | ||
337 | static int configuration_pre_load(void *opaque) | |
338 | { | |
339 | SaveState *state = opaque; | |
340 | ||
341 | /* If there is no target-page-bits subsection it means the source | |
342 | * predates the variable-target-page-bits support and is using the | |
343 | * minimum possible value for this CPU. | |
344 | */ | |
46d702b1 | 345 | state->target_page_bits = qemu_target_page_bits_min(); |
59811a32 | 346 | return 0; |
61964c23 JQ |
347 | } |
348 | ||
349 | static int configuration_post_load(void *opaque, int version_id) | |
350 | { | |
351 | SaveState *state = opaque; | |
352 | const char *current_name = MACHINE_GET_CLASS(current_machine)->name; | |
353 | ||
354 | if (strncmp(state->name, current_name, state->len) != 0) { | |
15d61692 GK |
355 | error_report("Machine type received is '%.*s' and local is '%s'", |
356 | (int) state->len, state->name, current_name); | |
61964c23 JQ |
357 | return -EINVAL; |
358 | } | |
59811a32 | 359 | |
46d702b1 | 360 | if (state->target_page_bits != qemu_target_page_bits()) { |
59811a32 | 361 | error_report("Received TARGET_PAGE_BITS is %d but local is %d", |
46d702b1 | 362 | state->target_page_bits, qemu_target_page_bits()); |
59811a32 PM |
363 | return -EINVAL; |
364 | } | |
365 | ||
61964c23 JQ |
366 | return 0; |
367 | } | |
368 | ||
59811a32 PM |
369 | /* The target-page-bits subsection is present only if the |
370 | * target page size is not the same as the default (ie the | |
371 | * minimum page size for a variable-page-size guest CPU). | |
372 | * If it is present then it contains the actual target page | |
373 | * bits for the machine, and migration will fail if the | |
374 | * two ends don't agree about it. | |
375 | */ | |
376 | static bool vmstate_target_page_bits_needed(void *opaque) | |
377 | { | |
46d702b1 JQ |
378 | return qemu_target_page_bits() |
379 | > qemu_target_page_bits_min(); | |
59811a32 PM |
380 | } |
381 | ||
382 | static const VMStateDescription vmstate_target_page_bits = { | |
383 | .name = "configuration/target-page-bits", | |
384 | .version_id = 1, | |
385 | .minimum_version_id = 1, | |
386 | .needed = vmstate_target_page_bits_needed, | |
387 | .fields = (VMStateField[]) { | |
388 | VMSTATE_UINT32(target_page_bits, SaveState), | |
389 | VMSTATE_END_OF_LIST() | |
390 | } | |
391 | }; | |
392 | ||
61964c23 JQ |
393 | static const VMStateDescription vmstate_configuration = { |
394 | .name = "configuration", | |
395 | .version_id = 1, | |
59811a32 | 396 | .pre_load = configuration_pre_load, |
61964c23 JQ |
397 | .post_load = configuration_post_load, |
398 | .pre_save = configuration_pre_save, | |
399 | .fields = (VMStateField[]) { | |
400 | VMSTATE_UINT32(len, SaveState), | |
59046ec2 | 401 | VMSTATE_VBUFFER_ALLOC_UINT32(name, SaveState, 0, NULL, len), |
61964c23 JQ |
402 | VMSTATE_END_OF_LIST() |
403 | }, | |
59811a32 PM |
404 | .subsections = (const VMStateDescription*[]) { |
405 | &vmstate_target_page_bits, | |
406 | NULL | |
407 | } | |
0163a2e0 | 408 | }; |
a672b469 | 409 | |
abfd9ce3 AS |
410 | static void dump_vmstate_vmsd(FILE *out_file, |
411 | const VMStateDescription *vmsd, int indent, | |
412 | bool is_subsection); | |
413 | ||
414 | static void dump_vmstate_vmsf(FILE *out_file, const VMStateField *field, | |
415 | int indent) | |
416 | { | |
417 | fprintf(out_file, "%*s{\n", indent, ""); | |
418 | indent += 2; | |
419 | fprintf(out_file, "%*s\"field\": \"%s\",\n", indent, "", field->name); | |
420 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
421 | field->version_id); | |
422 | fprintf(out_file, "%*s\"field_exists\": %s,\n", indent, "", | |
423 | field->field_exists ? "true" : "false"); | |
424 | fprintf(out_file, "%*s\"size\": %zu", indent, "", field->size); | |
425 | if (field->vmsd != NULL) { | |
426 | fprintf(out_file, ",\n"); | |
427 | dump_vmstate_vmsd(out_file, field->vmsd, indent, false); | |
428 | } | |
429 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
430 | } | |
431 | ||
432 | static void dump_vmstate_vmss(FILE *out_file, | |
5cd8cada | 433 | const VMStateDescription **subsection, |
abfd9ce3 AS |
434 | int indent) |
435 | { | |
5cd8cada JQ |
436 | if (*subsection != NULL) { |
437 | dump_vmstate_vmsd(out_file, *subsection, indent, true); | |
abfd9ce3 AS |
438 | } |
439 | } | |
440 | ||
441 | static void dump_vmstate_vmsd(FILE *out_file, | |
442 | const VMStateDescription *vmsd, int indent, | |
443 | bool is_subsection) | |
444 | { | |
445 | if (is_subsection) { | |
446 | fprintf(out_file, "%*s{\n", indent, ""); | |
447 | } else { | |
448 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", "Description"); | |
449 | } | |
450 | indent += 2; | |
451 | fprintf(out_file, "%*s\"name\": \"%s\",\n", indent, "", vmsd->name); | |
452 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
453 | vmsd->version_id); | |
454 | fprintf(out_file, "%*s\"minimum_version_id\": %d", indent, "", | |
455 | vmsd->minimum_version_id); | |
456 | if (vmsd->fields != NULL) { | |
457 | const VMStateField *field = vmsd->fields; | |
458 | bool first; | |
459 | ||
460 | fprintf(out_file, ",\n%*s\"Fields\": [\n", indent, ""); | |
461 | first = true; | |
462 | while (field->name != NULL) { | |
463 | if (field->flags & VMS_MUST_EXIST) { | |
464 | /* Ignore VMSTATE_VALIDATE bits; these don't get migrated */ | |
465 | field++; | |
466 | continue; | |
467 | } | |
468 | if (!first) { | |
469 | fprintf(out_file, ",\n"); | |
470 | } | |
471 | dump_vmstate_vmsf(out_file, field, indent + 2); | |
472 | field++; | |
473 | first = false; | |
474 | } | |
475 | fprintf(out_file, "\n%*s]", indent, ""); | |
476 | } | |
477 | if (vmsd->subsections != NULL) { | |
5cd8cada | 478 | const VMStateDescription **subsection = vmsd->subsections; |
abfd9ce3 AS |
479 | bool first; |
480 | ||
481 | fprintf(out_file, ",\n%*s\"Subsections\": [\n", indent, ""); | |
482 | first = true; | |
5cd8cada | 483 | while (*subsection != NULL) { |
abfd9ce3 AS |
484 | if (!first) { |
485 | fprintf(out_file, ",\n"); | |
486 | } | |
487 | dump_vmstate_vmss(out_file, subsection, indent + 2); | |
488 | subsection++; | |
489 | first = false; | |
490 | } | |
491 | fprintf(out_file, "\n%*s]", indent, ""); | |
492 | } | |
493 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
494 | } | |
495 | ||
496 | static void dump_machine_type(FILE *out_file) | |
497 | { | |
498 | MachineClass *mc; | |
499 | ||
500 | mc = MACHINE_GET_CLASS(current_machine); | |
501 | ||
502 | fprintf(out_file, " \"vmschkmachine\": {\n"); | |
503 | fprintf(out_file, " \"Name\": \"%s\"\n", mc->name); | |
504 | fprintf(out_file, " },\n"); | |
505 | } | |
506 | ||
507 | void dump_vmstate_json_to_file(FILE *out_file) | |
508 | { | |
509 | GSList *list, *elt; | |
510 | bool first; | |
511 | ||
512 | fprintf(out_file, "{\n"); | |
513 | dump_machine_type(out_file); | |
514 | ||
515 | first = true; | |
516 | list = object_class_get_list(TYPE_DEVICE, true); | |
517 | for (elt = list; elt; elt = elt->next) { | |
518 | DeviceClass *dc = OBJECT_CLASS_CHECK(DeviceClass, elt->data, | |
519 | TYPE_DEVICE); | |
520 | const char *name; | |
521 | int indent = 2; | |
522 | ||
523 | if (!dc->vmsd) { | |
524 | continue; | |
525 | } | |
526 | ||
527 | if (!first) { | |
528 | fprintf(out_file, ",\n"); | |
529 | } | |
530 | name = object_class_get_name(OBJECT_CLASS(dc)); | |
531 | fprintf(out_file, "%*s\"%s\": {\n", indent, "", name); | |
532 | indent += 2; | |
533 | fprintf(out_file, "%*s\"Name\": \"%s\",\n", indent, "", name); | |
534 | fprintf(out_file, "%*s\"version_id\": %d,\n", indent, "", | |
535 | dc->vmsd->version_id); | |
536 | fprintf(out_file, "%*s\"minimum_version_id\": %d,\n", indent, "", | |
537 | dc->vmsd->minimum_version_id); | |
538 | ||
539 | dump_vmstate_vmsd(out_file, dc->vmsd, indent, false); | |
540 | ||
541 | fprintf(out_file, "\n%*s}", indent - 2, ""); | |
542 | first = false; | |
543 | } | |
544 | fprintf(out_file, "\n}\n"); | |
545 | fclose(out_file); | |
546 | } | |
547 | ||
8718e999 JQ |
548 | static int calculate_new_instance_id(const char *idstr) |
549 | { | |
550 | SaveStateEntry *se; | |
551 | int instance_id = 0; | |
552 | ||
0163a2e0 | 553 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
8718e999 JQ |
554 | if (strcmp(idstr, se->idstr) == 0 |
555 | && instance_id <= se->instance_id) { | |
556 | instance_id = se->instance_id + 1; | |
557 | } | |
558 | } | |
559 | return instance_id; | |
560 | } | |
561 | ||
7685ee6a AW |
562 | static int calculate_compat_instance_id(const char *idstr) |
563 | { | |
564 | SaveStateEntry *se; | |
565 | int instance_id = 0; | |
566 | ||
0163a2e0 | 567 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
38ff78d3 | 568 | if (!se->compat) { |
7685ee6a | 569 | continue; |
38ff78d3 | 570 | } |
7685ee6a AW |
571 | |
572 | if (strcmp(idstr, se->compat->idstr) == 0 | |
573 | && instance_id <= se->compat->instance_id) { | |
574 | instance_id = se->compat->instance_id + 1; | |
575 | } | |
576 | } | |
577 | return instance_id; | |
578 | } | |
579 | ||
f37bc036 PX |
580 | static inline MigrationPriority save_state_priority(SaveStateEntry *se) |
581 | { | |
582 | if (se->vmsd) { | |
583 | return se->vmsd->priority; | |
584 | } | |
585 | return MIG_PRI_DEFAULT; | |
586 | } | |
587 | ||
588 | static void savevm_state_handler_insert(SaveStateEntry *nse) | |
589 | { | |
590 | MigrationPriority priority = save_state_priority(nse); | |
591 | SaveStateEntry *se; | |
592 | ||
593 | assert(priority <= MIG_PRI_MAX); | |
594 | ||
595 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
596 | if (save_state_priority(se) < priority) { | |
597 | break; | |
598 | } | |
599 | } | |
600 | ||
601 | if (se) { | |
602 | QTAILQ_INSERT_BEFORE(se, nse, entry); | |
603 | } else { | |
604 | QTAILQ_INSERT_TAIL(&savevm_state.handlers, nse, entry); | |
605 | } | |
606 | } | |
607 | ||
a672b469 AL |
608 | /* TODO: Individual devices generally have very little idea about the rest |
609 | of the system, so instance_id should be removed/replaced. | |
610 | Meanwhile pass -1 as instance_id if you do not already have a clearly | |
611 | distinguishing id for all instances of your device class. */ | |
0be71e32 AW |
612 | int register_savevm_live(DeviceState *dev, |
613 | const char *idstr, | |
a672b469 AL |
614 | int instance_id, |
615 | int version_id, | |
7908c78d | 616 | SaveVMHandlers *ops, |
a672b469 AL |
617 | void *opaque) |
618 | { | |
8718e999 | 619 | SaveStateEntry *se; |
a672b469 | 620 | |
97f3ad35 | 621 | se = g_new0(SaveStateEntry, 1); |
a672b469 | 622 | se->version_id = version_id; |
0163a2e0 | 623 | se->section_id = savevm_state.global_section_id++; |
7908c78d | 624 | se->ops = ops; |
a672b469 | 625 | se->opaque = opaque; |
9ed7d6ae | 626 | se->vmsd = NULL; |
a7ae8355 | 627 | /* if this is a live_savem then set is_ram */ |
9907e842 | 628 | if (ops->save_setup != NULL) { |
a7ae8355 SS |
629 | se->is_ram = 1; |
630 | } | |
a672b469 | 631 | |
09e5ab63 AL |
632 | if (dev) { |
633 | char *id = qdev_get_dev_path(dev); | |
7685ee6a | 634 | if (id) { |
581f08ba DDAG |
635 | if (snprintf(se->idstr, sizeof(se->idstr), "%s/", id) >= |
636 | sizeof(se->idstr)) { | |
637 | error_report("Path too long for VMState (%s)", id); | |
638 | g_free(id); | |
639 | g_free(se); | |
640 | ||
641 | return -1; | |
642 | } | |
7267c094 | 643 | g_free(id); |
7685ee6a | 644 | |
97f3ad35 | 645 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
646 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), idstr); |
647 | se->compat->instance_id = instance_id == -1 ? | |
648 | calculate_compat_instance_id(idstr) : instance_id; | |
649 | instance_id = -1; | |
650 | } | |
651 | } | |
652 | pstrcat(se->idstr, sizeof(se->idstr), idstr); | |
653 | ||
8718e999 | 654 | if (instance_id == -1) { |
7685ee6a | 655 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
656 | } else { |
657 | se->instance_id = instance_id; | |
a672b469 | 658 | } |
7685ee6a | 659 | assert(!se->compat || se->instance_id == 0); |
f37bc036 | 660 | savevm_state_handler_insert(se); |
a672b469 AL |
661 | return 0; |
662 | } | |
663 | ||
0be71e32 | 664 | void unregister_savevm(DeviceState *dev, const char *idstr, void *opaque) |
41bd13af | 665 | { |
8718e999 | 666 | SaveStateEntry *se, *new_se; |
7685ee6a AW |
667 | char id[256] = ""; |
668 | ||
09e5ab63 AL |
669 | if (dev) { |
670 | char *path = qdev_get_dev_path(dev); | |
7685ee6a AW |
671 | if (path) { |
672 | pstrcpy(id, sizeof(id), path); | |
673 | pstrcat(id, sizeof(id), "/"); | |
7267c094 | 674 | g_free(path); |
7685ee6a AW |
675 | } |
676 | } | |
677 | pstrcat(id, sizeof(id), idstr); | |
41bd13af | 678 | |
0163a2e0 | 679 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
7685ee6a | 680 | if (strcmp(se->idstr, id) == 0 && se->opaque == opaque) { |
0163a2e0 | 681 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 682 | g_free(se->compat); |
7267c094 | 683 | g_free(se); |
41bd13af | 684 | } |
41bd13af AL |
685 | } |
686 | } | |
687 | ||
0be71e32 | 688 | int vmstate_register_with_alias_id(DeviceState *dev, int instance_id, |
4d2ffa08 JK |
689 | const VMStateDescription *vmsd, |
690 | void *opaque, int alias_id, | |
bc5c4f21 DDAG |
691 | int required_for_version, |
692 | Error **errp) | |
9ed7d6ae | 693 | { |
8718e999 | 694 | SaveStateEntry *se; |
9ed7d6ae | 695 | |
4d2ffa08 JK |
696 | /* If this triggers, alias support can be dropped for the vmsd. */ |
697 | assert(alias_id == -1 || required_for_version >= vmsd->minimum_version_id); | |
698 | ||
97f3ad35 | 699 | se = g_new0(SaveStateEntry, 1); |
9ed7d6ae | 700 | se->version_id = vmsd->version_id; |
0163a2e0 | 701 | se->section_id = savevm_state.global_section_id++; |
9ed7d6ae JQ |
702 | se->opaque = opaque; |
703 | se->vmsd = vmsd; | |
4d2ffa08 | 704 | se->alias_id = alias_id; |
9ed7d6ae | 705 | |
09e5ab63 AL |
706 | if (dev) { |
707 | char *id = qdev_get_dev_path(dev); | |
7685ee6a | 708 | if (id) { |
581f08ba DDAG |
709 | if (snprintf(se->idstr, sizeof(se->idstr), "%s/", id) >= |
710 | sizeof(se->idstr)) { | |
711 | error_setg(errp, "Path too long for VMState (%s)", id); | |
712 | g_free(id); | |
713 | g_free(se); | |
714 | ||
715 | return -1; | |
716 | } | |
128e4e10 | 717 | g_free(id); |
7685ee6a | 718 | |
97f3ad35 | 719 | se->compat = g_new0(CompatEntry, 1); |
7685ee6a AW |
720 | pstrcpy(se->compat->idstr, sizeof(se->compat->idstr), vmsd->name); |
721 | se->compat->instance_id = instance_id == -1 ? | |
722 | calculate_compat_instance_id(vmsd->name) : instance_id; | |
723 | instance_id = -1; | |
724 | } | |
725 | } | |
726 | pstrcat(se->idstr, sizeof(se->idstr), vmsd->name); | |
727 | ||
8718e999 | 728 | if (instance_id == -1) { |
7685ee6a | 729 | se->instance_id = calculate_new_instance_id(se->idstr); |
8718e999 JQ |
730 | } else { |
731 | se->instance_id = instance_id; | |
9ed7d6ae | 732 | } |
7685ee6a | 733 | assert(!se->compat || se->instance_id == 0); |
f37bc036 | 734 | savevm_state_handler_insert(se); |
9ed7d6ae JQ |
735 | return 0; |
736 | } | |
737 | ||
0be71e32 AW |
738 | void vmstate_unregister(DeviceState *dev, const VMStateDescription *vmsd, |
739 | void *opaque) | |
9ed7d6ae | 740 | { |
1eb7538b JQ |
741 | SaveStateEntry *se, *new_se; |
742 | ||
0163a2e0 | 743 | QTAILQ_FOREACH_SAFE(se, &savevm_state.handlers, entry, new_se) { |
1eb7538b | 744 | if (se->vmsd == vmsd && se->opaque == opaque) { |
0163a2e0 | 745 | QTAILQ_REMOVE(&savevm_state.handlers, se, entry); |
ef1e1e07 | 746 | g_free(se->compat); |
7267c094 | 747 | g_free(se); |
1eb7538b JQ |
748 | } |
749 | } | |
9ed7d6ae JQ |
750 | } |
751 | ||
3a011c26 | 752 | static int vmstate_load(QEMUFile *f, SaveStateEntry *se) |
4082be4d | 753 | { |
9013dca5 | 754 | trace_vmstate_load(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
9ed7d6ae | 755 | if (!se->vmsd) { /* Old style */ |
3a011c26 | 756 | return se->ops->load_state(f, se->opaque, se->load_version_id); |
9ed7d6ae | 757 | } |
3a011c26 | 758 | return vmstate_load_state(f, se->vmsd, se->opaque, se->load_version_id); |
4082be4d JQ |
759 | } |
760 | ||
8118f095 AG |
761 | static void vmstate_save_old_style(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) |
762 | { | |
763 | int64_t old_offset, size; | |
764 | ||
765 | old_offset = qemu_ftell_fast(f); | |
766 | se->ops->save_state(f, se->opaque); | |
767 | size = qemu_ftell_fast(f) - old_offset; | |
768 | ||
769 | if (vmdesc) { | |
770 | json_prop_int(vmdesc, "size", size); | |
771 | json_start_array(vmdesc, "fields"); | |
772 | json_start_object(vmdesc, NULL); | |
773 | json_prop_str(vmdesc, "name", "data"); | |
774 | json_prop_int(vmdesc, "size", size); | |
775 | json_prop_str(vmdesc, "type", "buffer"); | |
776 | json_end_object(vmdesc); | |
777 | json_end_array(vmdesc); | |
778 | } | |
779 | } | |
780 | ||
687433f6 | 781 | static int vmstate_save(QEMUFile *f, SaveStateEntry *se, QJSON *vmdesc) |
4082be4d | 782 | { |
9013dca5 | 783 | trace_vmstate_save(se->idstr, se->vmsd ? se->vmsd->name : "(old)"); |
8118f095 AG |
784 | if (!se->vmsd) { |
785 | vmstate_save_old_style(f, se, vmdesc); | |
687433f6 | 786 | return 0; |
9ed7d6ae | 787 | } |
687433f6 | 788 | return vmstate_save_state(f, se->vmsd, se->opaque, vmdesc); |
4082be4d JQ |
789 | } |
790 | ||
ce39bfc9 DDAG |
791 | /* |
792 | * Write the header for device section (QEMU_VM_SECTION START/END/PART/FULL) | |
793 | */ | |
794 | static void save_section_header(QEMUFile *f, SaveStateEntry *se, | |
795 | uint8_t section_type) | |
796 | { | |
797 | qemu_put_byte(f, section_type); | |
798 | qemu_put_be32(f, se->section_id); | |
799 | ||
800 | if (section_type == QEMU_VM_SECTION_FULL || | |
801 | section_type == QEMU_VM_SECTION_START) { | |
802 | /* ID string */ | |
803 | size_t len = strlen(se->idstr); | |
804 | qemu_put_byte(f, len); | |
805 | qemu_put_buffer(f, (uint8_t *)se->idstr, len); | |
806 | ||
807 | qemu_put_be32(f, se->instance_id); | |
808 | qemu_put_be32(f, se->version_id); | |
809 | } | |
810 | } | |
811 | ||
f68945d4 DDAG |
812 | /* |
813 | * Write a footer onto device sections that catches cases misformatted device | |
814 | * sections. | |
815 | */ | |
816 | static void save_section_footer(QEMUFile *f, SaveStateEntry *se) | |
817 | { | |
15c38503 | 818 | if (migrate_get_current()->send_section_footer) { |
f68945d4 DDAG |
819 | qemu_put_byte(f, QEMU_VM_SECTION_FOOTER); |
820 | qemu_put_be32(f, se->section_id); | |
821 | } | |
822 | } | |
823 | ||
c76ca188 DDAG |
824 | /** |
825 | * qemu_savevm_command_send: Send a 'QEMU_VM_COMMAND' type element with the | |
826 | * command and associated data. | |
827 | * | |
828 | * @f: File to send command on | |
829 | * @command: Command type to send | |
830 | * @len: Length of associated data | |
831 | * @data: Data associated with command. | |
832 | */ | |
20a519a0 JQ |
833 | static void qemu_savevm_command_send(QEMUFile *f, |
834 | enum qemu_vm_cmd command, | |
835 | uint16_t len, | |
836 | uint8_t *data) | |
c76ca188 DDAG |
837 | { |
838 | trace_savevm_command_send(command, len); | |
839 | qemu_put_byte(f, QEMU_VM_COMMAND); | |
840 | qemu_put_be16(f, (uint16_t)command); | |
841 | qemu_put_be16(f, len); | |
842 | qemu_put_buffer(f, data, len); | |
843 | qemu_fflush(f); | |
844 | } | |
845 | ||
aad555c2 ZC |
846 | void qemu_savevm_send_colo_enable(QEMUFile *f) |
847 | { | |
848 | trace_savevm_send_colo_enable(); | |
849 | qemu_savevm_command_send(f, MIG_CMD_ENABLE_COLO, 0, NULL); | |
850 | } | |
851 | ||
2e37701e DDAG |
852 | void qemu_savevm_send_ping(QEMUFile *f, uint32_t value) |
853 | { | |
854 | uint32_t buf; | |
855 | ||
856 | trace_savevm_send_ping(value); | |
857 | buf = cpu_to_be32(value); | |
858 | qemu_savevm_command_send(f, MIG_CMD_PING, sizeof(value), (uint8_t *)&buf); | |
859 | } | |
860 | ||
861 | void qemu_savevm_send_open_return_path(QEMUFile *f) | |
862 | { | |
863 | trace_savevm_send_open_return_path(); | |
864 | qemu_savevm_command_send(f, MIG_CMD_OPEN_RETURN_PATH, 0, NULL); | |
865 | } | |
866 | ||
11cf1d98 DDAG |
867 | /* We have a buffer of data to send; we don't want that all to be loaded |
868 | * by the command itself, so the command contains just the length of the | |
869 | * extra buffer that we then send straight after it. | |
870 | * TODO: Must be a better way to organise that | |
871 | * | |
872 | * Returns: | |
873 | * 0 on success | |
874 | * -ve on error | |
875 | */ | |
61b67d47 | 876 | int qemu_savevm_send_packaged(QEMUFile *f, const uint8_t *buf, size_t len) |
11cf1d98 | 877 | { |
11cf1d98 DDAG |
878 | uint32_t tmp; |
879 | ||
880 | if (len > MAX_VM_CMD_PACKAGED_SIZE) { | |
881 | error_report("%s: Unreasonably large packaged state: %zu", | |
882 | __func__, len); | |
883 | return -1; | |
884 | } | |
885 | ||
886 | tmp = cpu_to_be32(len); | |
887 | ||
888 | trace_qemu_savevm_send_packaged(); | |
889 | qemu_savevm_command_send(f, MIG_CMD_PACKAGED, 4, (uint8_t *)&tmp); | |
890 | ||
61b67d47 | 891 | qemu_put_buffer(f, buf, len); |
11cf1d98 DDAG |
892 | |
893 | return 0; | |
894 | } | |
895 | ||
093e3c42 DDAG |
896 | /* Send prior to any postcopy transfer */ |
897 | void qemu_savevm_send_postcopy_advise(QEMUFile *f) | |
898 | { | |
58110f0a VSO |
899 | if (migrate_postcopy_ram()) { |
900 | uint64_t tmp[2]; | |
901 | tmp[0] = cpu_to_be64(ram_pagesize_summary()); | |
902 | tmp[1] = cpu_to_be64(qemu_target_page_size()); | |
903 | ||
904 | trace_qemu_savevm_send_postcopy_advise(); | |
905 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_ADVISE, | |
906 | 16, (uint8_t *)tmp); | |
907 | } else { | |
908 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_ADVISE, 0, NULL); | |
909 | } | |
093e3c42 DDAG |
910 | } |
911 | ||
912 | /* Sent prior to starting the destination running in postcopy, discard pages | |
913 | * that have already been sent but redirtied on the source. | |
914 | * CMD_POSTCOPY_RAM_DISCARD consist of: | |
915 | * byte version (0) | |
916 | * byte Length of name field (not including 0) | |
917 | * n x byte RAM block name | |
918 | * byte 0 terminator (just for safety) | |
919 | * n x Byte ranges within the named RAMBlock | |
920 | * be64 Start of the range | |
921 | * be64 Length | |
922 | * | |
923 | * name: RAMBlock name that these entries are part of | |
924 | * len: Number of page entries | |
925 | * start_list: 'len' addresses | |
926 | * length_list: 'len' addresses | |
927 | * | |
928 | */ | |
929 | void qemu_savevm_send_postcopy_ram_discard(QEMUFile *f, const char *name, | |
930 | uint16_t len, | |
931 | uint64_t *start_list, | |
932 | uint64_t *length_list) | |
933 | { | |
934 | uint8_t *buf; | |
935 | uint16_t tmplen; | |
936 | uint16_t t; | |
937 | size_t name_len = strlen(name); | |
938 | ||
939 | trace_qemu_savevm_send_postcopy_ram_discard(name, len); | |
940 | assert(name_len < 256); | |
941 | buf = g_malloc0(1 + 1 + name_len + 1 + (8 + 8) * len); | |
942 | buf[0] = postcopy_ram_discard_version; | |
943 | buf[1] = name_len; | |
944 | memcpy(buf + 2, name, name_len); | |
945 | tmplen = 2 + name_len; | |
946 | buf[tmplen++] = '\0'; | |
947 | ||
948 | for (t = 0; t < len; t++) { | |
4d885131 | 949 | stq_be_p(buf + tmplen, start_list[t]); |
093e3c42 | 950 | tmplen += 8; |
4d885131 | 951 | stq_be_p(buf + tmplen, length_list[t]); |
093e3c42 DDAG |
952 | tmplen += 8; |
953 | } | |
954 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RAM_DISCARD, tmplen, buf); | |
955 | g_free(buf); | |
956 | } | |
957 | ||
958 | /* Get the destination into a state where it can receive postcopy data. */ | |
959 | void qemu_savevm_send_postcopy_listen(QEMUFile *f) | |
960 | { | |
961 | trace_savevm_send_postcopy_listen(); | |
962 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_LISTEN, 0, NULL); | |
963 | } | |
964 | ||
965 | /* Kick the destination into running */ | |
966 | void qemu_savevm_send_postcopy_run(QEMUFile *f) | |
967 | { | |
968 | trace_savevm_send_postcopy_run(); | |
969 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RUN, 0, NULL); | |
970 | } | |
971 | ||
3f5875ec PX |
972 | void qemu_savevm_send_postcopy_resume(QEMUFile *f) |
973 | { | |
974 | trace_savevm_send_postcopy_resume(); | |
975 | qemu_savevm_command_send(f, MIG_CMD_POSTCOPY_RESUME, 0, NULL); | |
976 | } | |
977 | ||
f25d4225 PX |
978 | void qemu_savevm_send_recv_bitmap(QEMUFile *f, char *block_name) |
979 | { | |
980 | size_t len; | |
981 | char buf[256]; | |
982 | ||
983 | trace_savevm_send_recv_bitmap(block_name); | |
984 | ||
985 | buf[0] = len = strlen(block_name); | |
986 | memcpy(buf + 1, block_name, len); | |
987 | ||
988 | qemu_savevm_command_send(f, MIG_CMD_RECV_BITMAP, len + 1, (uint8_t *)buf); | |
989 | } | |
990 | ||
e1c37d0e | 991 | bool qemu_savevm_state_blocked(Error **errp) |
dc912121 AW |
992 | { |
993 | SaveStateEntry *se; | |
994 | ||
0163a2e0 | 995 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
7d854c47 | 996 | if (se->vmsd && se->vmsd->unmigratable) { |
f231b88d CR |
997 | error_setg(errp, "State blocked by non-migratable device '%s'", |
998 | se->idstr); | |
dc912121 AW |
999 | return true; |
1000 | } | |
1001 | } | |
1002 | return false; | |
1003 | } | |
1004 | ||
f796baa1 DDAG |
1005 | void qemu_savevm_state_header(QEMUFile *f) |
1006 | { | |
1007 | trace_savevm_state_header(); | |
1008 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
1009 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
172dfd4f | 1010 | |
4ffdb337 | 1011 | if (migrate_get_current()->send_configuration) { |
172dfd4f DDAG |
1012 | qemu_put_byte(f, QEMU_VM_CONFIGURATION); |
1013 | vmstate_save_state(f, &vmstate_configuration, &savevm_state, 0); | |
1014 | } | |
f796baa1 DDAG |
1015 | } |
1016 | ||
9907e842 | 1017 | void qemu_savevm_state_setup(QEMUFile *f) |
a672b469 AL |
1018 | { |
1019 | SaveStateEntry *se; | |
39346385 | 1020 | int ret; |
a672b469 | 1021 | |
9907e842 | 1022 | trace_savevm_state_setup(); |
0163a2e0 | 1023 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
9907e842 | 1024 | if (!se->ops || !se->ops->save_setup) { |
a672b469 | 1025 | continue; |
22ea40f4 | 1026 | } |
6bd68781 JQ |
1027 | if (se->ops && se->ops->is_active) { |
1028 | if (!se->ops->is_active(se->opaque)) { | |
1029 | continue; | |
1030 | } | |
1031 | } | |
ce39bfc9 | 1032 | save_section_header(f, se, QEMU_VM_SECTION_START); |
a672b469 | 1033 | |
9907e842 | 1034 | ret = se->ops->save_setup(f, se->opaque); |
f68945d4 | 1035 | save_section_footer(f, se); |
2975725f | 1036 | if (ret < 0) { |
47c8c17a PB |
1037 | qemu_file_set_error(f, ret); |
1038 | break; | |
2975725f | 1039 | } |
a672b469 | 1040 | } |
a672b469 AL |
1041 | } |
1042 | ||
d1b8eadb PX |
1043 | int qemu_savevm_state_resume_prepare(MigrationState *s) |
1044 | { | |
1045 | SaveStateEntry *se; | |
1046 | int ret; | |
1047 | ||
1048 | trace_savevm_state_resume_prepare(); | |
1049 | ||
1050 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
1051 | if (!se->ops || !se->ops->resume_prepare) { | |
1052 | continue; | |
1053 | } | |
1054 | if (se->ops && se->ops->is_active) { | |
1055 | if (!se->ops->is_active(se->opaque)) { | |
1056 | continue; | |
1057 | } | |
1058 | } | |
1059 | ret = se->ops->resume_prepare(s, se->opaque); | |
1060 | if (ret < 0) { | |
1061 | return ret; | |
1062 | } | |
1063 | } | |
1064 | ||
1065 | return 0; | |
1066 | } | |
1067 | ||
39346385 | 1068 | /* |
07f35073 | 1069 | * this function has three return values: |
39346385 JQ |
1070 | * negative: there was one error, and we have -errno. |
1071 | * 0 : We haven't finished, caller have to go again | |
1072 | * 1 : We have finished, we can go to complete phase | |
1073 | */ | |
35ecd943 | 1074 | int qemu_savevm_state_iterate(QEMUFile *f, bool postcopy) |
a672b469 AL |
1075 | { |
1076 | SaveStateEntry *se; | |
1077 | int ret = 1; | |
1078 | ||
9013dca5 | 1079 | trace_savevm_state_iterate(); |
0163a2e0 | 1080 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
16310a3c | 1081 | if (!se->ops || !se->ops->save_live_iterate) { |
a672b469 | 1082 | continue; |
22ea40f4 | 1083 | } |
6bd68781 JQ |
1084 | if (se->ops && se->ops->is_active) { |
1085 | if (!se->ops->is_active(se->opaque)) { | |
1086 | continue; | |
1087 | } | |
1088 | } | |
c865d848 VSO |
1089 | if (se->ops && se->ops->is_active_iterate) { |
1090 | if (!se->ops->is_active_iterate(se->opaque)) { | |
1091 | continue; | |
1092 | } | |
1093 | } | |
35ecd943 DDAG |
1094 | /* |
1095 | * In the postcopy phase, any device that doesn't know how to | |
1096 | * do postcopy should have saved it's state in the _complete | |
1097 | * call that's already run, it might get confused if we call | |
1098 | * iterate afterwards. | |
1099 | */ | |
c6467627 VSO |
1100 | if (postcopy && |
1101 | !(se->ops->has_postcopy && se->ops->has_postcopy(se->opaque))) { | |
35ecd943 DDAG |
1102 | continue; |
1103 | } | |
aac844ed JQ |
1104 | if (qemu_file_rate_limit(f)) { |
1105 | return 0; | |
1106 | } | |
464400f6 | 1107 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
1108 | |
1109 | save_section_header(f, se, QEMU_VM_SECTION_PART); | |
a672b469 | 1110 | |
16310a3c | 1111 | ret = se->ops->save_live_iterate(f, se->opaque); |
a5df2a02 | 1112 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 1113 | save_section_footer(f, se); |
517a13c9 | 1114 | |
47c8c17a PB |
1115 | if (ret < 0) { |
1116 | qemu_file_set_error(f, ret); | |
1117 | } | |
2975725f | 1118 | if (ret <= 0) { |
90697be8 JK |
1119 | /* Do not proceed to the next vmstate before this one reported |
1120 | completion of the current stage. This serializes the migration | |
1121 | and reduces the probability that a faster changing state is | |
1122 | synchronized over and over again. */ | |
1123 | break; | |
1124 | } | |
a672b469 | 1125 | } |
39346385 | 1126 | return ret; |
a672b469 AL |
1127 | } |
1128 | ||
9850c604 AG |
1129 | static bool should_send_vmdesc(void) |
1130 | { | |
1131 | MachineState *machine = MACHINE(qdev_get_machine()); | |
5727309d | 1132 | bool in_postcopy = migration_in_postcopy(); |
8421b205 | 1133 | return !machine->suppress_vmdesc && !in_postcopy; |
9850c604 AG |
1134 | } |
1135 | ||
763c906b DDAG |
1136 | /* |
1137 | * Calls the save_live_complete_postcopy methods | |
1138 | * causing the last few pages to be sent immediately and doing any associated | |
1139 | * cleanup. | |
1140 | * Note postcopy also calls qemu_savevm_state_complete_precopy to complete | |
1141 | * all the other devices, but that happens at the point we switch to postcopy. | |
1142 | */ | |
1143 | void qemu_savevm_state_complete_postcopy(QEMUFile *f) | |
1144 | { | |
1145 | SaveStateEntry *se; | |
1146 | int ret; | |
1147 | ||
1148 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
1149 | if (!se->ops || !se->ops->save_live_complete_postcopy) { | |
1150 | continue; | |
1151 | } | |
1152 | if (se->ops && se->ops->is_active) { | |
1153 | if (!se->ops->is_active(se->opaque)) { | |
1154 | continue; | |
1155 | } | |
1156 | } | |
1157 | trace_savevm_section_start(se->idstr, se->section_id); | |
1158 | /* Section type */ | |
1159 | qemu_put_byte(f, QEMU_VM_SECTION_END); | |
1160 | qemu_put_be32(f, se->section_id); | |
1161 | ||
1162 | ret = se->ops->save_live_complete_postcopy(f, se->opaque); | |
1163 | trace_savevm_section_end(se->idstr, se->section_id, ret); | |
1164 | save_section_footer(f, se); | |
1165 | if (ret < 0) { | |
1166 | qemu_file_set_error(f, ret); | |
1167 | return; | |
1168 | } | |
1169 | } | |
1170 | ||
1171 | qemu_put_byte(f, QEMU_VM_EOF); | |
1172 | qemu_fflush(f); | |
1173 | } | |
1174 | ||
a1fbe750 FZ |
1175 | int qemu_savevm_state_complete_precopy(QEMUFile *f, bool iterable_only, |
1176 | bool inactivate_disks) | |
a672b469 | 1177 | { |
8118f095 AG |
1178 | QJSON *vmdesc; |
1179 | int vmdesc_len; | |
a672b469 | 1180 | SaveStateEntry *se; |
2975725f | 1181 | int ret; |
5727309d | 1182 | bool in_postcopy = migration_in_postcopy(); |
a672b469 | 1183 | |
a3e06c3d | 1184 | trace_savevm_state_complete_precopy(); |
9013dca5 | 1185 | |
ea375f9a JK |
1186 | cpu_synchronize_all_states(); |
1187 | ||
0163a2e0 | 1188 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
763c906b | 1189 | if (!se->ops || |
c6467627 VSO |
1190 | (in_postcopy && se->ops->has_postcopy && |
1191 | se->ops->has_postcopy(se->opaque)) || | |
1c0d249d | 1192 | (in_postcopy && !iterable_only) || |
763c906b | 1193 | !se->ops->save_live_complete_precopy) { |
a672b469 | 1194 | continue; |
22ea40f4 | 1195 | } |
1c0d249d | 1196 | |
6bd68781 JQ |
1197 | if (se->ops && se->ops->is_active) { |
1198 | if (!se->ops->is_active(se->opaque)) { | |
1199 | continue; | |
1200 | } | |
1201 | } | |
464400f6 | 1202 | trace_savevm_section_start(se->idstr, se->section_id); |
ce39bfc9 DDAG |
1203 | |
1204 | save_section_header(f, se, QEMU_VM_SECTION_END); | |
a672b469 | 1205 | |
a3e06c3d | 1206 | ret = se->ops->save_live_complete_precopy(f, se->opaque); |
a5df2a02 | 1207 | trace_savevm_section_end(se->idstr, se->section_id, ret); |
f68945d4 | 1208 | save_section_footer(f, se); |
2975725f | 1209 | if (ret < 0) { |
47c8c17a | 1210 | qemu_file_set_error(f, ret); |
a1fbe750 | 1211 | return -1; |
2975725f | 1212 | } |
a672b469 AL |
1213 | } |
1214 | ||
1c0d249d | 1215 | if (iterable_only) { |
a1fbe750 | 1216 | return 0; |
1c0d249d DDAG |
1217 | } |
1218 | ||
8118f095 | 1219 | vmdesc = qjson_new(); |
46d702b1 | 1220 | json_prop_int(vmdesc, "page_size", qemu_target_page_size()); |
8118f095 | 1221 | json_start_array(vmdesc, "devices"); |
0163a2e0 | 1222 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1223 | |
22ea40f4 | 1224 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
5cecf414 | 1225 | continue; |
22ea40f4 | 1226 | } |
df896152 JQ |
1227 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1228 | trace_savevm_section_skip(se->idstr, se->section_id); | |
1229 | continue; | |
1230 | } | |
1231 | ||
464400f6 | 1232 | trace_savevm_section_start(se->idstr, se->section_id); |
8118f095 AG |
1233 | |
1234 | json_start_object(vmdesc, NULL); | |
1235 | json_prop_str(vmdesc, "name", se->idstr); | |
1236 | json_prop_int(vmdesc, "instance_id", se->instance_id); | |
1237 | ||
ce39bfc9 | 1238 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
687433f6 DDAG |
1239 | ret = vmstate_save(f, se, vmdesc); |
1240 | if (ret) { | |
1241 | qemu_file_set_error(f, ret); | |
1242 | return ret; | |
1243 | } | |
a5df2a02 | 1244 | trace_savevm_section_end(se->idstr, se->section_id, 0); |
f68945d4 | 1245 | save_section_footer(f, se); |
bdf46d64 WY |
1246 | |
1247 | json_end_object(vmdesc); | |
a672b469 AL |
1248 | } |
1249 | ||
a1fbe750 FZ |
1250 | if (inactivate_disks) { |
1251 | /* Inactivate before sending QEMU_VM_EOF so that the | |
1252 | * bdrv_invalidate_cache_all() on the other end won't fail. */ | |
1253 | ret = bdrv_inactivate_all(); | |
1254 | if (ret) { | |
c232bf58 DDAG |
1255 | error_report("%s: bdrv_inactivate_all() failed (%d)", |
1256 | __func__, ret); | |
a1fbe750 FZ |
1257 | qemu_file_set_error(f, ret); |
1258 | return ret; | |
1259 | } | |
1260 | } | |
763c906b DDAG |
1261 | if (!in_postcopy) { |
1262 | /* Postcopy stream will still be going */ | |
1263 | qemu_put_byte(f, QEMU_VM_EOF); | |
1264 | } | |
8118f095 AG |
1265 | |
1266 | json_end_array(vmdesc); | |
1267 | qjson_finish(vmdesc); | |
1268 | vmdesc_len = strlen(qjson_get_str(vmdesc)); | |
1269 | ||
9850c604 AG |
1270 | if (should_send_vmdesc()) { |
1271 | qemu_put_byte(f, QEMU_VM_VMDESCRIPTION); | |
1272 | qemu_put_be32(f, vmdesc_len); | |
1273 | qemu_put_buffer(f, (uint8_t *)qjson_get_str(vmdesc), vmdesc_len); | |
1274 | } | |
b72fe9e6 | 1275 | qjson_destroy(vmdesc); |
8118f095 | 1276 | |
edaae611 | 1277 | qemu_fflush(f); |
a1fbe750 | 1278 | return 0; |
a672b469 AL |
1279 | } |
1280 | ||
c31b098f DDAG |
1281 | /* Give an estimate of the amount left to be transferred, |
1282 | * the result is split into the amount for units that can and | |
1283 | * for units that can't do postcopy. | |
1284 | */ | |
faec066a | 1285 | void qemu_savevm_state_pending(QEMUFile *f, uint64_t threshold_size, |
47995026 VSO |
1286 | uint64_t *res_precopy_only, |
1287 | uint64_t *res_compatible, | |
1288 | uint64_t *res_postcopy_only) | |
e4ed1541 JQ |
1289 | { |
1290 | SaveStateEntry *se; | |
c31b098f | 1291 | |
47995026 VSO |
1292 | *res_precopy_only = 0; |
1293 | *res_compatible = 0; | |
1294 | *res_postcopy_only = 0; | |
c31b098f | 1295 | |
e4ed1541 | 1296 | |
0163a2e0 | 1297 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
e4ed1541 JQ |
1298 | if (!se->ops || !se->ops->save_live_pending) { |
1299 | continue; | |
1300 | } | |
1301 | if (se->ops && se->ops->is_active) { | |
1302 | if (!se->ops->is_active(se->opaque)) { | |
1303 | continue; | |
1304 | } | |
1305 | } | |
faec066a | 1306 | se->ops->save_live_pending(f, se->opaque, threshold_size, |
47995026 VSO |
1307 | res_precopy_only, res_compatible, |
1308 | res_postcopy_only); | |
e4ed1541 | 1309 | } |
e4ed1541 JQ |
1310 | } |
1311 | ||
ea7415fa | 1312 | void qemu_savevm_state_cleanup(void) |
4ec7fcc7 JK |
1313 | { |
1314 | SaveStateEntry *se; | |
1315 | ||
ea7415fa | 1316 | trace_savevm_state_cleanup(); |
0163a2e0 | 1317 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
70f794fc JQ |
1318 | if (se->ops && se->ops->save_cleanup) { |
1319 | se->ops->save_cleanup(se->opaque); | |
4ec7fcc7 JK |
1320 | } |
1321 | } | |
1322 | } | |
1323 | ||
5d80448c | 1324 | static int qemu_savevm_state(QEMUFile *f, Error **errp) |
a672b469 | 1325 | { |
a672b469 | 1326 | int ret; |
3e0c8050 | 1327 | MigrationState *ms = migrate_get_current(); |
6dcf6668 | 1328 | MigrationStatus status; |
3e0c8050 PX |
1329 | |
1330 | migrate_init(ms); | |
1331 | ||
89a02a9f | 1332 | ms->to_dst_file = f; |
a672b469 | 1333 | |
24f3902b | 1334 | if (migration_is_blocked(errp)) { |
6dcf6668 DL |
1335 | ret = -EINVAL; |
1336 | goto done; | |
dc912121 AW |
1337 | } |
1338 | ||
ce7c817c JQ |
1339 | if (migrate_use_block()) { |
1340 | error_setg(errp, "Block migration and snapshots are incompatible"); | |
1341 | ret = -EINVAL; | |
1342 | goto done; | |
1343 | } | |
1344 | ||
9b095037 | 1345 | qemu_mutex_unlock_iothread(); |
f796baa1 | 1346 | qemu_savevm_state_header(f); |
9907e842 | 1347 | qemu_savevm_state_setup(f); |
9b095037 PB |
1348 | qemu_mutex_lock_iothread(); |
1349 | ||
47c8c17a | 1350 | while (qemu_file_get_error(f) == 0) { |
35ecd943 | 1351 | if (qemu_savevm_state_iterate(f, false) > 0) { |
47c8c17a PB |
1352 | break; |
1353 | } | |
1354 | } | |
a672b469 | 1355 | |
47c8c17a | 1356 | ret = qemu_file_get_error(f); |
39346385 | 1357 | if (ret == 0) { |
a1fbe750 | 1358 | qemu_savevm_state_complete_precopy(f, false, false); |
624b9cc2 | 1359 | ret = qemu_file_get_error(f); |
39346385 | 1360 | } |
15b3b8ea | 1361 | qemu_savevm_state_cleanup(); |
04943eba | 1362 | if (ret != 0) { |
5d80448c | 1363 | error_setg_errno(errp, -ret, "Error while writing VM state"); |
04943eba | 1364 | } |
6dcf6668 DL |
1365 | |
1366 | done: | |
1367 | if (ret != 0) { | |
1368 | status = MIGRATION_STATUS_FAILED; | |
1369 | } else { | |
1370 | status = MIGRATION_STATUS_COMPLETED; | |
1371 | } | |
1372 | migrate_set_state(&ms->state, MIGRATION_STATUS_SETUP, status); | |
f9c8caa0 VSO |
1373 | |
1374 | /* f is outer parameter, it should not stay in global migration state after | |
1375 | * this function finished */ | |
1376 | ms->to_dst_file = NULL; | |
1377 | ||
a672b469 AL |
1378 | return ret; |
1379 | } | |
1380 | ||
3f6df99d | 1381 | void qemu_savevm_live_state(QEMUFile *f) |
a7ae8355 | 1382 | { |
3f6df99d ZC |
1383 | /* save QEMU_VM_SECTION_END section */ |
1384 | qemu_savevm_state_complete_precopy(f, true, false); | |
1385 | qemu_put_byte(f, QEMU_VM_EOF); | |
1386 | } | |
a7ae8355 | 1387 | |
3f6df99d ZC |
1388 | int qemu_save_device_state(QEMUFile *f) |
1389 | { | |
1390 | SaveStateEntry *se; | |
a7ae8355 | 1391 | |
3f6df99d ZC |
1392 | if (!migration_in_colo_state()) { |
1393 | qemu_put_be32(f, QEMU_VM_FILE_MAGIC); | |
1394 | qemu_put_be32(f, QEMU_VM_FILE_VERSION); | |
1395 | } | |
a7ae8355 SS |
1396 | cpu_synchronize_all_states(); |
1397 | ||
0163a2e0 | 1398 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
687433f6 DDAG |
1399 | int ret; |
1400 | ||
a7ae8355 SS |
1401 | if (se->is_ram) { |
1402 | continue; | |
1403 | } | |
22ea40f4 | 1404 | if ((!se->ops || !se->ops->save_state) && !se->vmsd) { |
a7ae8355 SS |
1405 | continue; |
1406 | } | |
df896152 JQ |
1407 | if (se->vmsd && !vmstate_save_needed(se->vmsd, se->opaque)) { |
1408 | continue; | |
1409 | } | |
a7ae8355 | 1410 | |
ce39bfc9 | 1411 | save_section_header(f, se, QEMU_VM_SECTION_FULL); |
a7ae8355 | 1412 | |
687433f6 DDAG |
1413 | ret = vmstate_save(f, se, NULL); |
1414 | if (ret) { | |
1415 | return ret; | |
1416 | } | |
f68945d4 DDAG |
1417 | |
1418 | save_section_footer(f, se); | |
a7ae8355 SS |
1419 | } |
1420 | ||
1421 | qemu_put_byte(f, QEMU_VM_EOF); | |
1422 | ||
1423 | return qemu_file_get_error(f); | |
1424 | } | |
1425 | ||
a672b469 AL |
1426 | static SaveStateEntry *find_se(const char *idstr, int instance_id) |
1427 | { | |
1428 | SaveStateEntry *se; | |
1429 | ||
0163a2e0 | 1430 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
a672b469 | 1431 | if (!strcmp(se->idstr, idstr) && |
4d2ffa08 JK |
1432 | (instance_id == se->instance_id || |
1433 | instance_id == se->alias_id)) | |
a672b469 | 1434 | return se; |
7685ee6a AW |
1435 | /* Migrating from an older version? */ |
1436 | if (strstr(se->idstr, idstr) && se->compat) { | |
1437 | if (!strcmp(se->compat->idstr, idstr) && | |
1438 | (instance_id == se->compat->instance_id || | |
1439 | instance_id == se->alias_id)) | |
1440 | return se; | |
1441 | } | |
a672b469 AL |
1442 | } |
1443 | return NULL; | |
1444 | } | |
1445 | ||
7b89bf27 DDAG |
1446 | enum LoadVMExitCodes { |
1447 | /* Allow a command to quit all layers of nested loadvm loops */ | |
1448 | LOADVM_QUIT = 1, | |
1449 | }; | |
1450 | ||
093e3c42 DDAG |
1451 | /* ------ incoming postcopy messages ------ */ |
1452 | /* 'advise' arrives before any transfers just to tell us that a postcopy | |
1453 | * *might* happen - it might be skipped if precopy transferred everything | |
1454 | * quickly. | |
1455 | */ | |
875fcd01 GK |
1456 | static int loadvm_postcopy_handle_advise(MigrationIncomingState *mis, |
1457 | uint16_t len) | |
093e3c42 DDAG |
1458 | { |
1459 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
e8ca1db2 | 1460 | uint64_t remote_pagesize_summary, local_pagesize_summary, remote_tps; |
d3dff7a5 | 1461 | Error *local_err = NULL; |
093e3c42 DDAG |
1462 | |
1463 | trace_loadvm_postcopy_handle_advise(); | |
1464 | if (ps != POSTCOPY_INCOMING_NONE) { | |
1465 | error_report("CMD_POSTCOPY_ADVISE in wrong postcopy state (%d)", ps); | |
1466 | return -1; | |
1467 | } | |
1468 | ||
875fcd01 GK |
1469 | switch (len) { |
1470 | case 0: | |
1471 | if (migrate_postcopy_ram()) { | |
1472 | error_report("RAM postcopy is enabled but have 0 byte advise"); | |
1473 | return -EINVAL; | |
1474 | } | |
58110f0a | 1475 | return 0; |
875fcd01 GK |
1476 | case 8 + 8: |
1477 | if (!migrate_postcopy_ram()) { | |
1478 | error_report("RAM postcopy is disabled but have 16 byte advise"); | |
1479 | return -EINVAL; | |
1480 | } | |
1481 | break; | |
1482 | default: | |
1483 | error_report("CMD_POSTCOPY_ADVISE invalid length (%d)", len); | |
1484 | return -EINVAL; | |
58110f0a VSO |
1485 | } |
1486 | ||
d7651f15 | 1487 | if (!postcopy_ram_supported_by_host(mis)) { |
328d4d85 | 1488 | postcopy_state_set(POSTCOPY_INCOMING_NONE); |
eb59db53 DDAG |
1489 | return -1; |
1490 | } | |
1491 | ||
e8ca1db2 DDAG |
1492 | remote_pagesize_summary = qemu_get_be64(mis->from_src_file); |
1493 | local_pagesize_summary = ram_pagesize_summary(); | |
1494 | ||
1495 | if (remote_pagesize_summary != local_pagesize_summary) { | |
093e3c42 | 1496 | /* |
e8ca1db2 DDAG |
1497 | * This detects two potential causes of mismatch: |
1498 | * a) A mismatch in host page sizes | |
1499 | * Some combinations of mismatch are probably possible but it gets | |
1500 | * a bit more complicated. In particular we need to place whole | |
1501 | * host pages on the dest at once, and we need to ensure that we | |
1502 | * handle dirtying to make sure we never end up sending part of | |
1503 | * a hostpage on it's own. | |
1504 | * b) The use of different huge page sizes on source/destination | |
1505 | * a more fine grain test is performed during RAM block migration | |
1506 | * but this test here causes a nice early clear failure, and | |
1507 | * also fails when passed to an older qemu that doesn't | |
1508 | * do huge pages. | |
093e3c42 | 1509 | */ |
e8ca1db2 DDAG |
1510 | error_report("Postcopy needs matching RAM page sizes (s=%" PRIx64 |
1511 | " d=%" PRIx64 ")", | |
1512 | remote_pagesize_summary, local_pagesize_summary); | |
093e3c42 DDAG |
1513 | return -1; |
1514 | } | |
1515 | ||
1516 | remote_tps = qemu_get_be64(mis->from_src_file); | |
20afaed9 | 1517 | if (remote_tps != qemu_target_page_size()) { |
093e3c42 DDAG |
1518 | /* |
1519 | * Again, some differences could be dealt with, but for now keep it | |
1520 | * simple. | |
1521 | */ | |
20afaed9 JQ |
1522 | error_report("Postcopy needs matching target page sizes (s=%d d=%zd)", |
1523 | (int)remote_tps, qemu_target_page_size()); | |
093e3c42 DDAG |
1524 | return -1; |
1525 | } | |
1526 | ||
d3dff7a5 DDAG |
1527 | if (postcopy_notify(POSTCOPY_NOTIFY_INBOUND_ADVISE, &local_err)) { |
1528 | error_report_err(local_err); | |
1529 | return -1; | |
1530 | } | |
1531 | ||
1caddf8a DDAG |
1532 | if (ram_postcopy_incoming_init(mis)) { |
1533 | return -1; | |
1534 | } | |
1535 | ||
1536 | postcopy_state_set(POSTCOPY_INCOMING_ADVISE); | |
1537 | ||
093e3c42 DDAG |
1538 | return 0; |
1539 | } | |
1540 | ||
1541 | /* After postcopy we will be told to throw some pages away since they're | |
1542 | * dirty and will have to be demand fetched. Must happen before CPU is | |
1543 | * started. | |
1544 | * There can be 0..many of these messages, each encoding multiple pages. | |
1545 | */ | |
1546 | static int loadvm_postcopy_ram_handle_discard(MigrationIncomingState *mis, | |
1547 | uint16_t len) | |
1548 | { | |
1549 | int tmp; | |
1550 | char ramid[256]; | |
1551 | PostcopyState ps = postcopy_state_get(); | |
1552 | ||
1553 | trace_loadvm_postcopy_ram_handle_discard(); | |
1554 | ||
1555 | switch (ps) { | |
1556 | case POSTCOPY_INCOMING_ADVISE: | |
1557 | /* 1st discard */ | |
f9527107 | 1558 | tmp = postcopy_ram_prepare_discard(mis); |
093e3c42 DDAG |
1559 | if (tmp) { |
1560 | return tmp; | |
1561 | } | |
1562 | break; | |
1563 | ||
1564 | case POSTCOPY_INCOMING_DISCARD: | |
1565 | /* Expected state */ | |
1566 | break; | |
1567 | ||
1568 | default: | |
1569 | error_report("CMD_POSTCOPY_RAM_DISCARD in wrong postcopy state (%d)", | |
1570 | ps); | |
1571 | return -1; | |
1572 | } | |
1573 | /* We're expecting a | |
1574 | * Version (0) | |
1575 | * a RAM ID string (length byte, name, 0 term) | |
1576 | * then at least 1 16 byte chunk | |
1577 | */ | |
1578 | if (len < (1 + 1 + 1 + 1 + 2 * 8)) { | |
1579 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1580 | return -1; | |
1581 | } | |
1582 | ||
1583 | tmp = qemu_get_byte(mis->from_src_file); | |
1584 | if (tmp != postcopy_ram_discard_version) { | |
1585 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid version (%d)", tmp); | |
1586 | return -1; | |
1587 | } | |
1588 | ||
1589 | if (!qemu_get_counted_string(mis->from_src_file, ramid)) { | |
1590 | error_report("CMD_POSTCOPY_RAM_DISCARD Failed to read RAMBlock ID"); | |
1591 | return -1; | |
1592 | } | |
1593 | tmp = qemu_get_byte(mis->from_src_file); | |
1594 | if (tmp != 0) { | |
1595 | error_report("CMD_POSTCOPY_RAM_DISCARD missing nil (%d)", tmp); | |
1596 | return -1; | |
1597 | } | |
1598 | ||
1599 | len -= 3 + strlen(ramid); | |
1600 | if (len % 16) { | |
1601 | error_report("CMD_POSTCOPY_RAM_DISCARD invalid length (%d)", len); | |
1602 | return -1; | |
1603 | } | |
1604 | trace_loadvm_postcopy_ram_handle_discard_header(ramid, len); | |
1605 | while (len) { | |
093e3c42 DDAG |
1606 | uint64_t start_addr, block_length; |
1607 | start_addr = qemu_get_be64(mis->from_src_file); | |
1608 | block_length = qemu_get_be64(mis->from_src_file); | |
1609 | ||
1610 | len -= 16; | |
aaa2064c | 1611 | int ret = ram_discard_range(ramid, start_addr, block_length); |
093e3c42 DDAG |
1612 | if (ret) { |
1613 | return ret; | |
1614 | } | |
093e3c42 DDAG |
1615 | } |
1616 | trace_loadvm_postcopy_ram_handle_discard_end(); | |
1617 | ||
1618 | return 0; | |
1619 | } | |
1620 | ||
c76201ab DDAG |
1621 | /* |
1622 | * Triggered by a postcopy_listen command; this thread takes over reading | |
1623 | * the input stream, leaving the main thread free to carry on loading the rest | |
1624 | * of the device state (from RAM). | |
1625 | * (TODO:This could do with being in a postcopy file - but there again it's | |
1626 | * just another input loop, not that postcopy specific) | |
1627 | */ | |
1628 | static void *postcopy_ram_listen_thread(void *opaque) | |
1629 | { | |
c76201ab | 1630 | MigrationIncomingState *mis = migration_incoming_get_current(); |
b411b844 | 1631 | QEMUFile *f = mis->from_src_file; |
c76201ab DDAG |
1632 | int load_res; |
1633 | ||
6ba996bb DDAG |
1634 | migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE, |
1635 | MIGRATION_STATUS_POSTCOPY_ACTIVE); | |
c76201ab DDAG |
1636 | qemu_sem_post(&mis->listen_thread_sem); |
1637 | trace_postcopy_ram_listen_thread_start(); | |
1638 | ||
74637e6f | 1639 | rcu_register_thread(); |
c76201ab DDAG |
1640 | /* |
1641 | * Because we're a thread and not a coroutine we can't yield | |
1642 | * in qemu_file, and thus we must be blocking now. | |
1643 | */ | |
1644 | qemu_file_set_blocking(f, true); | |
1645 | load_res = qemu_loadvm_state_main(f, mis); | |
b411b844 PX |
1646 | |
1647 | /* | |
1648 | * This is tricky, but, mis->from_src_file can change after it | |
1649 | * returns, when postcopy recovery happened. In the future, we may | |
1650 | * want a wrapper for the QEMUFile handle. | |
1651 | */ | |
1652 | f = mis->from_src_file; | |
1653 | ||
c76201ab DDAG |
1654 | /* And non-blocking again so we don't block in any cleanup */ |
1655 | qemu_file_set_blocking(f, false); | |
1656 | ||
1657 | trace_postcopy_ram_listen_thread_exit(); | |
1658 | if (load_res < 0) { | |
1659 | error_report("%s: loadvm failed: %d", __func__, load_res); | |
1660 | qemu_file_set_error(f, load_res); | |
6ba996bb DDAG |
1661 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1662 | MIGRATION_STATUS_FAILED); | |
c76201ab DDAG |
1663 | } else { |
1664 | /* | |
1665 | * This looks good, but it's possible that the device loading in the | |
1666 | * main thread hasn't finished yet, and so we might not be in 'RUN' | |
1667 | * state yet; wait for the end of the main thread. | |
1668 | */ | |
1669 | qemu_event_wait(&mis->main_thread_load_event); | |
1670 | } | |
1671 | postcopy_ram_incoming_cleanup(mis); | |
c76201ab DDAG |
1672 | |
1673 | if (load_res < 0) { | |
1674 | /* | |
1675 | * If something went wrong then we have a bad state so exit; | |
1676 | * depending how far we got it might be possible at this point | |
1677 | * to leave the guest running and fire MCEs for pages that never | |
1678 | * arrived as a desperate recovery step. | |
1679 | */ | |
74637e6f | 1680 | rcu_unregister_thread(); |
c76201ab DDAG |
1681 | exit(EXIT_FAILURE); |
1682 | } | |
1683 | ||
6ba996bb DDAG |
1684 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
1685 | MIGRATION_STATUS_COMPLETED); | |
1686 | /* | |
1687 | * If everything has worked fine, then the main thread has waited | |
1688 | * for us to start, and we're the last use of the mis. | |
1689 | * (If something broke then qemu will have to exit anyway since it's | |
1690 | * got a bad migration state). | |
1691 | */ | |
1692 | migration_incoming_state_destroy(); | |
acb5ea86 | 1693 | qemu_loadvm_state_cleanup(); |
6ba996bb | 1694 | |
74637e6f | 1695 | rcu_unregister_thread(); |
9cf4bb87 | 1696 | mis->have_listen_thread = false; |
c76201ab DDAG |
1697 | return NULL; |
1698 | } | |
1699 | ||
093e3c42 DDAG |
1700 | /* After this message we must be able to immediately receive postcopy data */ |
1701 | static int loadvm_postcopy_handle_listen(MigrationIncomingState *mis) | |
1702 | { | |
1703 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_LISTENING); | |
1704 | trace_loadvm_postcopy_handle_listen(); | |
6864a7b5 DDAG |
1705 | Error *local_err = NULL; |
1706 | ||
093e3c42 DDAG |
1707 | if (ps != POSTCOPY_INCOMING_ADVISE && ps != POSTCOPY_INCOMING_DISCARD) { |
1708 | error_report("CMD_POSTCOPY_LISTEN in wrong postcopy state (%d)", ps); | |
1709 | return -1; | |
1710 | } | |
f9527107 DDAG |
1711 | if (ps == POSTCOPY_INCOMING_ADVISE) { |
1712 | /* | |
1713 | * A rare case, we entered listen without having to do any discards, | |
1714 | * so do the setup that's normally done at the time of the 1st discard. | |
1715 | */ | |
58110f0a VSO |
1716 | if (migrate_postcopy_ram()) { |
1717 | postcopy_ram_prepare_discard(mis); | |
1718 | } | |
f9527107 | 1719 | } |
093e3c42 | 1720 | |
f0a227ad DDAG |
1721 | /* |
1722 | * Sensitise RAM - can now generate requests for blocks that don't exist | |
1723 | * However, at this point the CPU shouldn't be running, and the IO | |
1724 | * shouldn't be doing anything yet so don't actually expect requests | |
1725 | */ | |
58110f0a VSO |
1726 | if (migrate_postcopy_ram()) { |
1727 | if (postcopy_ram_enable_notify(mis)) { | |
1728 | return -1; | |
1729 | } | |
f0a227ad DDAG |
1730 | } |
1731 | ||
6864a7b5 DDAG |
1732 | if (postcopy_notify(POSTCOPY_NOTIFY_INBOUND_LISTEN, &local_err)) { |
1733 | error_report_err(local_err); | |
1734 | return -1; | |
1735 | } | |
1736 | ||
c76201ab DDAG |
1737 | if (mis->have_listen_thread) { |
1738 | error_report("CMD_POSTCOPY_RAM_LISTEN already has a listen thread"); | |
1739 | return -1; | |
1740 | } | |
1741 | ||
1742 | mis->have_listen_thread = true; | |
1743 | /* Start up the listening thread and wait for it to signal ready */ | |
1744 | qemu_sem_init(&mis->listen_thread_sem, 0); | |
1745 | qemu_thread_create(&mis->listen_thread, "postcopy/listen", | |
b411b844 | 1746 | postcopy_ram_listen_thread, NULL, |
a587a3fe | 1747 | QEMU_THREAD_DETACHED); |
c76201ab DDAG |
1748 | qemu_sem_wait(&mis->listen_thread_sem); |
1749 | qemu_sem_destroy(&mis->listen_thread_sem); | |
1750 | ||
093e3c42 DDAG |
1751 | return 0; |
1752 | } | |
1753 | ||
86469922 DL |
1754 | |
1755 | typedef struct { | |
1756 | QEMUBH *bh; | |
1757 | } HandleRunBhData; | |
1758 | ||
ea6a55bc | 1759 | static void loadvm_postcopy_handle_run_bh(void *opaque) |
093e3c42 | 1760 | { |
27c6825b | 1761 | Error *local_err = NULL; |
86469922 | 1762 | HandleRunBhData *data = opaque; |
093e3c42 | 1763 | |
27c6825b DDAG |
1764 | /* TODO we should move all of this lot into postcopy_ram.c or a shared code |
1765 | * in migration.c | |
1766 | */ | |
1767 | cpu_synchronize_all_post_init(); | |
1768 | ||
1769 | qemu_announce_self(); | |
1770 | ||
ace21a58 KW |
1771 | /* Make sure all file formats flush their mutable metadata. |
1772 | * If we get an error here, just don't restart the VM yet. */ | |
27c6825b | 1773 | bdrv_invalidate_cache_all(&local_err); |
0042fd36 | 1774 | if (local_err) { |
ace21a58 | 1775 | error_report_err(local_err); |
0042fd36 KW |
1776 | local_err = NULL; |
1777 | autostart = false; | |
1778 | } | |
1779 | ||
27c6825b DDAG |
1780 | trace_loadvm_postcopy_handle_run_cpu_sync(); |
1781 | cpu_synchronize_all_post_init(); | |
1782 | ||
1783 | trace_loadvm_postcopy_handle_run_vmstart(); | |
1784 | ||
b35ebdf0 VSO |
1785 | dirty_bitmap_mig_before_vm_start(); |
1786 | ||
093e3c42 DDAG |
1787 | if (autostart) { |
1788 | /* Hold onto your hats, starting the CPU */ | |
1789 | vm_start(); | |
1790 | } else { | |
1791 | /* leave it paused and let management decide when to start the CPU */ | |
1792 | runstate_set(RUN_STATE_PAUSED); | |
1793 | } | |
1794 | ||
86469922 DL |
1795 | qemu_bh_delete(data->bh); |
1796 | g_free(data); | |
ea6a55bc DL |
1797 | } |
1798 | ||
1799 | /* After all discards we can start running and asking for pages */ | |
1800 | static int loadvm_postcopy_handle_run(MigrationIncomingState *mis) | |
1801 | { | |
1802 | PostcopyState ps = postcopy_state_set(POSTCOPY_INCOMING_RUNNING); | |
86469922 | 1803 | HandleRunBhData *data; |
ea6a55bc DL |
1804 | |
1805 | trace_loadvm_postcopy_handle_run(); | |
1806 | if (ps != POSTCOPY_INCOMING_LISTENING) { | |
1807 | error_report("CMD_POSTCOPY_RUN in wrong postcopy state (%d)", ps); | |
1808 | return -1; | |
1809 | } | |
1810 | ||
86469922 DL |
1811 | data = g_new(HandleRunBhData, 1); |
1812 | data->bh = qemu_bh_new(loadvm_postcopy_handle_run_bh, data); | |
1813 | qemu_bh_schedule(data->bh); | |
ea6a55bc | 1814 | |
27c6825b DDAG |
1815 | /* We need to finish reading the stream from the package |
1816 | * and also stop reading anything more from the stream that loaded the | |
1817 | * package (since it's now being read by the listener thread). | |
1818 | * LOADVM_QUIT will quit all the layers of nested loadvm loops. | |
1819 | */ | |
1820 | return LOADVM_QUIT; | |
093e3c42 DDAG |
1821 | } |
1822 | ||
3f5875ec PX |
1823 | static int loadvm_postcopy_handle_resume(MigrationIncomingState *mis) |
1824 | { | |
1825 | if (mis->state != MIGRATION_STATUS_POSTCOPY_RECOVER) { | |
1826 | error_report("%s: illegal resume received", __func__); | |
1827 | /* Don't fail the load, only for this. */ | |
1828 | return 0; | |
1829 | } | |
1830 | ||
1831 | /* | |
1832 | * This means source VM is ready to resume the postcopy migration. | |
1833 | * It's time to switch state and release the fault thread to | |
1834 | * continue service page faults. | |
1835 | */ | |
1836 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_RECOVER, | |
1837 | MIGRATION_STATUS_POSTCOPY_ACTIVE); | |
1838 | qemu_sem_post(&mis->postcopy_pause_sem_fault); | |
1839 | ||
1840 | trace_loadvm_postcopy_handle_resume(); | |
1841 | ||
13955b89 PX |
1842 | /* Tell source that "we are ready" */ |
1843 | migrate_send_rp_resume_ack(mis, MIGRATION_RESUME_ACK_VALUE); | |
3f5875ec PX |
1844 | |
1845 | return 0; | |
1846 | } | |
1847 | ||
c76ca188 | 1848 | /** |
11cf1d98 DDAG |
1849 | * Immediately following this command is a blob of data containing an embedded |
1850 | * chunk of migration stream; read it and load it. | |
1851 | * | |
1852 | * @mis: Incoming state | |
1853 | * @length: Length of packaged data to read | |
c76ca188 | 1854 | * |
11cf1d98 DDAG |
1855 | * Returns: Negative values on error |
1856 | * | |
1857 | */ | |
1858 | static int loadvm_handle_cmd_packaged(MigrationIncomingState *mis) | |
1859 | { | |
1860 | int ret; | |
61b67d47 DB |
1861 | size_t length; |
1862 | QIOChannelBuffer *bioc; | |
11cf1d98 DDAG |
1863 | |
1864 | length = qemu_get_be32(mis->from_src_file); | |
1865 | trace_loadvm_handle_cmd_packaged(length); | |
1866 | ||
1867 | if (length > MAX_VM_CMD_PACKAGED_SIZE) { | |
61b67d47 | 1868 | error_report("Unreasonably large packaged state: %zu", length); |
11cf1d98 DDAG |
1869 | return -1; |
1870 | } | |
61b67d47 DB |
1871 | |
1872 | bioc = qio_channel_buffer_new(length); | |
6f01f136 | 1873 | qio_channel_set_name(QIO_CHANNEL(bioc), "migration-loadvm-buffer"); |
61b67d47 DB |
1874 | ret = qemu_get_buffer(mis->from_src_file, |
1875 | bioc->data, | |
1876 | length); | |
11cf1d98 | 1877 | if (ret != length) { |
61b67d47 DB |
1878 | object_unref(OBJECT(bioc)); |
1879 | error_report("CMD_PACKAGED: Buffer receive fail ret=%d length=%zu", | |
9af9e0fe | 1880 | ret, length); |
11cf1d98 DDAG |
1881 | return (ret < 0) ? ret : -EAGAIN; |
1882 | } | |
61b67d47 | 1883 | bioc->usage += length; |
11cf1d98 DDAG |
1884 | trace_loadvm_handle_cmd_packaged_received(ret); |
1885 | ||
61b67d47 | 1886 | QEMUFile *packf = qemu_fopen_channel_input(QIO_CHANNEL(bioc)); |
11cf1d98 DDAG |
1887 | |
1888 | ret = qemu_loadvm_state_main(packf, mis); | |
1889 | trace_loadvm_handle_cmd_packaged_main(ret); | |
1890 | qemu_fclose(packf); | |
61b67d47 | 1891 | object_unref(OBJECT(bioc)); |
11cf1d98 DDAG |
1892 | |
1893 | return ret; | |
1894 | } | |
1895 | ||
f25d4225 PX |
1896 | /* |
1897 | * Handle request that source requests for recved_bitmap on | |
1898 | * destination. Payload format: | |
1899 | * | |
1900 | * len (1 byte) + ramblock_name (<255 bytes) | |
1901 | */ | |
1902 | static int loadvm_handle_recv_bitmap(MigrationIncomingState *mis, | |
1903 | uint16_t len) | |
1904 | { | |
1905 | QEMUFile *file = mis->from_src_file; | |
1906 | RAMBlock *rb; | |
1907 | char block_name[256]; | |
1908 | size_t cnt; | |
1909 | ||
1910 | cnt = qemu_get_counted_string(file, block_name); | |
1911 | if (!cnt) { | |
1912 | error_report("%s: failed to read block name", __func__); | |
1913 | return -EINVAL; | |
1914 | } | |
1915 | ||
1916 | /* Validate before using the data */ | |
1917 | if (qemu_file_get_error(file)) { | |
1918 | return qemu_file_get_error(file); | |
1919 | } | |
1920 | ||
1921 | if (len != cnt + 1) { | |
1922 | error_report("%s: invalid payload length (%d)", __func__, len); | |
1923 | return -EINVAL; | |
1924 | } | |
1925 | ||
1926 | rb = qemu_ram_block_by_name(block_name); | |
1927 | if (!rb) { | |
1928 | error_report("%s: block '%s' not found", __func__, block_name); | |
1929 | return -EINVAL; | |
1930 | } | |
1931 | ||
a335debb | 1932 | migrate_send_rp_recv_bitmap(mis, block_name); |
f25d4225 PX |
1933 | |
1934 | trace_loadvm_handle_recv_bitmap(block_name); | |
1935 | ||
1936 | return 0; | |
1937 | } | |
1938 | ||
aad555c2 ZC |
1939 | static int loadvm_process_enable_colo(MigrationIncomingState *mis) |
1940 | { | |
1941 | migration_incoming_enable_colo(); | |
13af18f2 | 1942 | return colo_init_ram_cache(); |
aad555c2 ZC |
1943 | } |
1944 | ||
11cf1d98 DDAG |
1945 | /* |
1946 | * Process an incoming 'QEMU_VM_COMMAND' | |
1947 | * 0 just a normal return | |
1948 | * LOADVM_QUIT All good, but exit the loop | |
1949 | * <0 Error | |
c76ca188 DDAG |
1950 | */ |
1951 | static int loadvm_process_command(QEMUFile *f) | |
1952 | { | |
2e37701e | 1953 | MigrationIncomingState *mis = migration_incoming_get_current(); |
c76ca188 DDAG |
1954 | uint16_t cmd; |
1955 | uint16_t len; | |
2e37701e | 1956 | uint32_t tmp32; |
c76ca188 DDAG |
1957 | |
1958 | cmd = qemu_get_be16(f); | |
1959 | len = qemu_get_be16(f); | |
1960 | ||
7a9ddfbf PX |
1961 | /* Check validity before continue processing of cmds */ |
1962 | if (qemu_file_get_error(f)) { | |
1963 | return qemu_file_get_error(f); | |
1964 | } | |
1965 | ||
c76ca188 DDAG |
1966 | trace_loadvm_process_command(cmd, len); |
1967 | if (cmd >= MIG_CMD_MAX || cmd == MIG_CMD_INVALID) { | |
1968 | error_report("MIG_CMD 0x%x unknown (len 0x%x)", cmd, len); | |
1969 | return -EINVAL; | |
1970 | } | |
1971 | ||
1972 | if (mig_cmd_args[cmd].len != -1 && mig_cmd_args[cmd].len != len) { | |
1973 | error_report("%s received with bad length - expecting %zu, got %d", | |
1974 | mig_cmd_args[cmd].name, | |
1975 | (size_t)mig_cmd_args[cmd].len, len); | |
1976 | return -ERANGE; | |
1977 | } | |
1978 | ||
1979 | switch (cmd) { | |
2e37701e DDAG |
1980 | case MIG_CMD_OPEN_RETURN_PATH: |
1981 | if (mis->to_src_file) { | |
1982 | error_report("CMD_OPEN_RETURN_PATH called when RP already open"); | |
1983 | /* Not really a problem, so don't give up */ | |
1984 | return 0; | |
1985 | } | |
1986 | mis->to_src_file = qemu_file_get_return_path(f); | |
1987 | if (!mis->to_src_file) { | |
1988 | error_report("CMD_OPEN_RETURN_PATH failed"); | |
1989 | return -1; | |
1990 | } | |
1991 | break; | |
1992 | ||
1993 | case MIG_CMD_PING: | |
1994 | tmp32 = qemu_get_be32(f); | |
1995 | trace_loadvm_process_command_ping(tmp32); | |
1996 | if (!mis->to_src_file) { | |
1997 | error_report("CMD_PING (0x%x) received with no return path", | |
1998 | tmp32); | |
1999 | return -1; | |
2000 | } | |
6decec93 | 2001 | migrate_send_rp_pong(mis, tmp32); |
2e37701e | 2002 | break; |
093e3c42 | 2003 | |
11cf1d98 DDAG |
2004 | case MIG_CMD_PACKAGED: |
2005 | return loadvm_handle_cmd_packaged(mis); | |
2006 | ||
093e3c42 | 2007 | case MIG_CMD_POSTCOPY_ADVISE: |
875fcd01 | 2008 | return loadvm_postcopy_handle_advise(mis, len); |
093e3c42 DDAG |
2009 | |
2010 | case MIG_CMD_POSTCOPY_LISTEN: | |
2011 | return loadvm_postcopy_handle_listen(mis); | |
2012 | ||
2013 | case MIG_CMD_POSTCOPY_RUN: | |
2014 | return loadvm_postcopy_handle_run(mis); | |
2015 | ||
2016 | case MIG_CMD_POSTCOPY_RAM_DISCARD: | |
2017 | return loadvm_postcopy_ram_handle_discard(mis, len); | |
f25d4225 | 2018 | |
3f5875ec PX |
2019 | case MIG_CMD_POSTCOPY_RESUME: |
2020 | return loadvm_postcopy_handle_resume(mis); | |
2021 | ||
f25d4225 PX |
2022 | case MIG_CMD_RECV_BITMAP: |
2023 | return loadvm_handle_recv_bitmap(mis, len); | |
aad555c2 ZC |
2024 | |
2025 | case MIG_CMD_ENABLE_COLO: | |
2026 | return loadvm_process_enable_colo(mis); | |
c76ca188 DDAG |
2027 | } |
2028 | ||
2029 | return 0; | |
2030 | } | |
2031 | ||
59f39a47 DDAG |
2032 | /* |
2033 | * Read a footer off the wire and check that it matches the expected section | |
2034 | * | |
2035 | * Returns: true if the footer was good | |
2036 | * false if there is a problem (and calls error_report to say why) | |
2037 | */ | |
0f42f657 | 2038 | static bool check_section_footer(QEMUFile *f, SaveStateEntry *se) |
59f39a47 | 2039 | { |
7a9ddfbf | 2040 | int ret; |
59f39a47 DDAG |
2041 | uint8_t read_mark; |
2042 | uint32_t read_section_id; | |
2043 | ||
15c38503 | 2044 | if (!migrate_get_current()->send_section_footer) { |
59f39a47 DDAG |
2045 | /* No footer to check */ |
2046 | return true; | |
2047 | } | |
2048 | ||
2049 | read_mark = qemu_get_byte(f); | |
2050 | ||
7a9ddfbf PX |
2051 | ret = qemu_file_get_error(f); |
2052 | if (ret) { | |
2053 | error_report("%s: Read section footer failed: %d", | |
2054 | __func__, ret); | |
2055 | return false; | |
2056 | } | |
2057 | ||
59f39a47 | 2058 | if (read_mark != QEMU_VM_SECTION_FOOTER) { |
0f42f657 | 2059 | error_report("Missing section footer for %s", se->idstr); |
59f39a47 DDAG |
2060 | return false; |
2061 | } | |
2062 | ||
2063 | read_section_id = qemu_get_be32(f); | |
0f42f657 | 2064 | if (read_section_id != se->load_section_id) { |
59f39a47 DDAG |
2065 | error_report("Mismatched section id in footer for %s -" |
2066 | " read 0x%x expected 0x%x", | |
0f42f657 | 2067 | se->idstr, read_section_id, se->load_section_id); |
59f39a47 DDAG |
2068 | return false; |
2069 | } | |
2070 | ||
2071 | /* All good */ | |
2072 | return true; | |
2073 | } | |
2074 | ||
fb3520a8 HZ |
2075 | static int |
2076 | qemu_loadvm_section_start_full(QEMUFile *f, MigrationIncomingState *mis) | |
2077 | { | |
2078 | uint32_t instance_id, version_id, section_id; | |
2079 | SaveStateEntry *se; | |
fb3520a8 HZ |
2080 | char idstr[256]; |
2081 | int ret; | |
2082 | ||
2083 | /* Read section start */ | |
2084 | section_id = qemu_get_be32(f); | |
2085 | if (!qemu_get_counted_string(f, idstr)) { | |
2086 | error_report("Unable to read ID string for section %u", | |
2087 | section_id); | |
2088 | return -EINVAL; | |
2089 | } | |
2090 | instance_id = qemu_get_be32(f); | |
2091 | version_id = qemu_get_be32(f); | |
2092 | ||
7a9ddfbf PX |
2093 | ret = qemu_file_get_error(f); |
2094 | if (ret) { | |
2095 | error_report("%s: Failed to read instance/version ID: %d", | |
2096 | __func__, ret); | |
2097 | return ret; | |
2098 | } | |
2099 | ||
fb3520a8 HZ |
2100 | trace_qemu_loadvm_state_section_startfull(section_id, idstr, |
2101 | instance_id, version_id); | |
2102 | /* Find savevm section */ | |
2103 | se = find_se(idstr, instance_id); | |
2104 | if (se == NULL) { | |
827beacb JRZ |
2105 | error_report("Unknown savevm section or instance '%s' %d. " |
2106 | "Make sure that your current VM setup matches your " | |
2107 | "saved VM setup, including any hotplugged devices", | |
fb3520a8 HZ |
2108 | idstr, instance_id); |
2109 | return -EINVAL; | |
2110 | } | |
2111 | ||
2112 | /* Validate version */ | |
2113 | if (version_id > se->version_id) { | |
2114 | error_report("savevm: unsupported version %d for '%s' v%d", | |
2115 | version_id, idstr, se->version_id); | |
2116 | return -EINVAL; | |
2117 | } | |
0f42f657 JQ |
2118 | se->load_version_id = version_id; |
2119 | se->load_section_id = section_id; | |
fb3520a8 | 2120 | |
88c16567 WC |
2121 | /* Validate if it is a device's state */ |
2122 | if (xen_enabled() && se->is_ram) { | |
2123 | error_report("loadvm: %s RAM loading not allowed on Xen", idstr); | |
2124 | return -EINVAL; | |
2125 | } | |
2126 | ||
3a011c26 | 2127 | ret = vmstate_load(f, se); |
fb3520a8 HZ |
2128 | if (ret < 0) { |
2129 | error_report("error while loading state for instance 0x%x of" | |
2130 | " device '%s'", instance_id, idstr); | |
2131 | return ret; | |
2132 | } | |
0f42f657 | 2133 | if (!check_section_footer(f, se)) { |
fb3520a8 HZ |
2134 | return -EINVAL; |
2135 | } | |
2136 | ||
2137 | return 0; | |
2138 | } | |
2139 | ||
2140 | static int | |
2141 | qemu_loadvm_section_part_end(QEMUFile *f, MigrationIncomingState *mis) | |
2142 | { | |
2143 | uint32_t section_id; | |
0f42f657 | 2144 | SaveStateEntry *se; |
fb3520a8 HZ |
2145 | int ret; |
2146 | ||
2147 | section_id = qemu_get_be32(f); | |
2148 | ||
7a9ddfbf PX |
2149 | ret = qemu_file_get_error(f); |
2150 | if (ret) { | |
2151 | error_report("%s: Failed to read section ID: %d", | |
2152 | __func__, ret); | |
2153 | return ret; | |
2154 | } | |
2155 | ||
fb3520a8 | 2156 | trace_qemu_loadvm_state_section_partend(section_id); |
0f42f657 JQ |
2157 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { |
2158 | if (se->load_section_id == section_id) { | |
fb3520a8 HZ |
2159 | break; |
2160 | } | |
2161 | } | |
0f42f657 | 2162 | if (se == NULL) { |
fb3520a8 HZ |
2163 | error_report("Unknown savevm section %d", section_id); |
2164 | return -EINVAL; | |
2165 | } | |
2166 | ||
3a011c26 | 2167 | ret = vmstate_load(f, se); |
fb3520a8 HZ |
2168 | if (ret < 0) { |
2169 | error_report("error while loading state section id %d(%s)", | |
0f42f657 | 2170 | section_id, se->idstr); |
fb3520a8 HZ |
2171 | return ret; |
2172 | } | |
0f42f657 | 2173 | if (!check_section_footer(f, se)) { |
fb3520a8 HZ |
2174 | return -EINVAL; |
2175 | } | |
2176 | ||
2177 | return 0; | |
2178 | } | |
2179 | ||
acb5ea86 JQ |
2180 | static int qemu_loadvm_state_setup(QEMUFile *f) |
2181 | { | |
2182 | SaveStateEntry *se; | |
2183 | int ret; | |
2184 | ||
2185 | trace_loadvm_state_setup(); | |
2186 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
2187 | if (!se->ops || !se->ops->load_setup) { | |
2188 | continue; | |
2189 | } | |
2190 | if (se->ops && se->ops->is_active) { | |
2191 | if (!se->ops->is_active(se->opaque)) { | |
2192 | continue; | |
2193 | } | |
2194 | } | |
2195 | ||
2196 | ret = se->ops->load_setup(f, se->opaque); | |
2197 | if (ret < 0) { | |
2198 | qemu_file_set_error(f, ret); | |
2199 | error_report("Load state of device %s failed", se->idstr); | |
2200 | return ret; | |
2201 | } | |
2202 | } | |
2203 | return 0; | |
2204 | } | |
2205 | ||
2206 | void qemu_loadvm_state_cleanup(void) | |
2207 | { | |
2208 | SaveStateEntry *se; | |
2209 | ||
2210 | trace_loadvm_state_cleanup(); | |
2211 | QTAILQ_FOREACH(se, &savevm_state.handlers, entry) { | |
2212 | if (se->ops && se->ops->load_cleanup) { | |
2213 | se->ops->load_cleanup(se->opaque); | |
2214 | } | |
2215 | } | |
2216 | } | |
2217 | ||
b411b844 PX |
2218 | /* Return true if we should continue the migration, or false. */ |
2219 | static bool postcopy_pause_incoming(MigrationIncomingState *mis) | |
2220 | { | |
2221 | trace_postcopy_pause_incoming(); | |
2222 | ||
02affd41 PX |
2223 | /* Clear the triggered bit to allow one recovery */ |
2224 | mis->postcopy_recover_triggered = false; | |
2225 | ||
b411b844 PX |
2226 | assert(mis->from_src_file); |
2227 | qemu_file_shutdown(mis->from_src_file); | |
2228 | qemu_fclose(mis->from_src_file); | |
2229 | mis->from_src_file = NULL; | |
2230 | ||
2231 | assert(mis->to_src_file); | |
2232 | qemu_file_shutdown(mis->to_src_file); | |
2233 | qemu_mutex_lock(&mis->rp_mutex); | |
2234 | qemu_fclose(mis->to_src_file); | |
2235 | mis->to_src_file = NULL; | |
2236 | qemu_mutex_unlock(&mis->rp_mutex); | |
2237 | ||
eed1cc78 PX |
2238 | migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE, |
2239 | MIGRATION_STATUS_POSTCOPY_PAUSED); | |
2240 | ||
3a7804c3 PX |
2241 | /* Notify the fault thread for the invalidated file handle */ |
2242 | postcopy_fault_thread_notify(mis); | |
2243 | ||
b411b844 PX |
2244 | error_report("Detected IO failure for postcopy. " |
2245 | "Migration paused."); | |
2246 | ||
2247 | while (mis->state == MIGRATION_STATUS_POSTCOPY_PAUSED) { | |
2248 | qemu_sem_wait(&mis->postcopy_pause_sem_dst); | |
2249 | } | |
2250 | ||
2251 | trace_postcopy_pause_incoming_continued(); | |
2252 | ||
2253 | return true; | |
2254 | } | |
2255 | ||
3f6df99d | 2256 | int qemu_loadvm_state_main(QEMUFile *f, MigrationIncomingState *mis) |
1a8f46f8 | 2257 | { |
a672b469 | 2258 | uint8_t section_type; |
ccb783c3 | 2259 | int ret = 0; |
61964c23 | 2260 | |
b411b844 | 2261 | retry: |
7a9ddfbf PX |
2262 | while (true) { |
2263 | section_type = qemu_get_byte(f); | |
2264 | ||
2265 | if (qemu_file_get_error(f)) { | |
2266 | ret = qemu_file_get_error(f); | |
2267 | break; | |
2268 | } | |
2269 | ||
a5df2a02 | 2270 | trace_qemu_loadvm_state_section(section_type); |
a672b469 AL |
2271 | switch (section_type) { |
2272 | case QEMU_VM_SECTION_START: | |
2273 | case QEMU_VM_SECTION_FULL: | |
fb3520a8 | 2274 | ret = qemu_loadvm_section_start_full(f, mis); |
b5a22e4a | 2275 | if (ret < 0) { |
ccb783c3 | 2276 | goto out; |
b5a22e4a | 2277 | } |
a672b469 AL |
2278 | break; |
2279 | case QEMU_VM_SECTION_PART: | |
2280 | case QEMU_VM_SECTION_END: | |
fb3520a8 | 2281 | ret = qemu_loadvm_section_part_end(f, mis); |
b5a22e4a | 2282 | if (ret < 0) { |
ccb783c3 | 2283 | goto out; |
b5a22e4a | 2284 | } |
a672b469 | 2285 | break; |
c76ca188 DDAG |
2286 | case QEMU_VM_COMMAND: |
2287 | ret = loadvm_process_command(f); | |
7b89bf27 DDAG |
2288 | trace_qemu_loadvm_state_section_command(ret); |
2289 | if ((ret < 0) || (ret & LOADVM_QUIT)) { | |
ccb783c3 | 2290 | goto out; |
c76ca188 DDAG |
2291 | } |
2292 | break; | |
7a9ddfbf PX |
2293 | case QEMU_VM_EOF: |
2294 | /* This is the end of migration */ | |
2295 | goto out; | |
a672b469 | 2296 | default: |
6a64b644 | 2297 | error_report("Unknown savevm section type %d", section_type); |
ccb783c3 DDAG |
2298 | ret = -EINVAL; |
2299 | goto out; | |
a672b469 AL |
2300 | } |
2301 | } | |
2302 | ||
ccb783c3 DDAG |
2303 | out: |
2304 | if (ret < 0) { | |
2305 | qemu_file_set_error(f, ret); | |
b411b844 PX |
2306 | |
2307 | /* | |
fd037a65 PX |
2308 | * If we are during an active postcopy, then we pause instead |
2309 | * of bail out to at least keep the VM's dirty data. Note | |
2310 | * that POSTCOPY_INCOMING_LISTENING stage is still not enough, | |
2311 | * during which we're still receiving device states and we | |
2312 | * still haven't yet started the VM on destination. | |
b411b844 PX |
2313 | */ |
2314 | if (postcopy_state_get() == POSTCOPY_INCOMING_RUNNING && | |
fd037a65 | 2315 | postcopy_pause_incoming(mis)) { |
b411b844 PX |
2316 | /* Reset f to point to the newly created channel */ |
2317 | f = mis->from_src_file; | |
2318 | goto retry; | |
2319 | } | |
ccb783c3 DDAG |
2320 | } |
2321 | return ret; | |
7b89bf27 DDAG |
2322 | } |
2323 | ||
2324 | int qemu_loadvm_state(QEMUFile *f) | |
2325 | { | |
2326 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
2327 | Error *local_err = NULL; | |
2328 | unsigned int v; | |
2329 | int ret; | |
2330 | ||
2331 | if (qemu_savevm_state_blocked(&local_err)) { | |
2332 | error_report_err(local_err); | |
2333 | return -EINVAL; | |
2334 | } | |
2335 | ||
2336 | v = qemu_get_be32(f); | |
2337 | if (v != QEMU_VM_FILE_MAGIC) { | |
2338 | error_report("Not a migration stream"); | |
2339 | return -EINVAL; | |
2340 | } | |
2341 | ||
2342 | v = qemu_get_be32(f); | |
2343 | if (v == QEMU_VM_FILE_VERSION_COMPAT) { | |
2344 | error_report("SaveVM v2 format is obsolete and don't work anymore"); | |
2345 | return -ENOTSUP; | |
2346 | } | |
2347 | if (v != QEMU_VM_FILE_VERSION) { | |
2348 | error_report("Unsupported migration stream version"); | |
2349 | return -ENOTSUP; | |
2350 | } | |
2351 | ||
acb5ea86 JQ |
2352 | if (qemu_loadvm_state_setup(f) != 0) { |
2353 | return -EINVAL; | |
2354 | } | |
2355 | ||
4ffdb337 | 2356 | if (migrate_get_current()->send_configuration) { |
7b89bf27 DDAG |
2357 | if (qemu_get_byte(f) != QEMU_VM_CONFIGURATION) { |
2358 | error_report("Configuration section missing"); | |
096c83b7 | 2359 | qemu_loadvm_state_cleanup(); |
7b89bf27 DDAG |
2360 | return -EINVAL; |
2361 | } | |
2362 | ret = vmstate_load_state(f, &vmstate_configuration, &savevm_state, 0); | |
2363 | ||
2364 | if (ret) { | |
096c83b7 | 2365 | qemu_loadvm_state_cleanup(); |
7b89bf27 DDAG |
2366 | return ret; |
2367 | } | |
2368 | } | |
2369 | ||
75e972da DG |
2370 | cpu_synchronize_all_pre_loadvm(); |
2371 | ||
7b89bf27 DDAG |
2372 | ret = qemu_loadvm_state_main(f, mis); |
2373 | qemu_event_set(&mis->main_thread_load_event); | |
2374 | ||
2375 | trace_qemu_loadvm_state_post_main(ret); | |
2376 | ||
c76201ab DDAG |
2377 | if (mis->have_listen_thread) { |
2378 | /* Listen thread still going, can't clean up yet */ | |
2379 | return ret; | |
2380 | } | |
2381 | ||
7b89bf27 DDAG |
2382 | if (ret == 0) { |
2383 | ret = qemu_file_get_error(f); | |
2384 | } | |
1925cebc AG |
2385 | |
2386 | /* | |
2387 | * Try to read in the VMDESC section as well, so that dumping tools that | |
2388 | * intercept our migration stream have the chance to see it. | |
2389 | */ | |
1aca9a5f DDAG |
2390 | |
2391 | /* We've got to be careful; if we don't read the data and just shut the fd | |
2392 | * then the sender can error if we close while it's still sending. | |
2393 | * We also mustn't read data that isn't there; some transports (RDMA) | |
2394 | * will stall waiting for that data when the source has already closed. | |
2395 | */ | |
7b89bf27 | 2396 | if (ret == 0 && should_send_vmdesc()) { |
1aca9a5f DDAG |
2397 | uint8_t *buf; |
2398 | uint32_t size; | |
7b89bf27 | 2399 | uint8_t section_type = qemu_get_byte(f); |
1aca9a5f DDAG |
2400 | |
2401 | if (section_type != QEMU_VM_VMDESCRIPTION) { | |
2402 | error_report("Expected vmdescription section, but got %d", | |
2403 | section_type); | |
2404 | /* | |
2405 | * It doesn't seem worth failing at this point since | |
2406 | * we apparently have an otherwise valid VM state | |
2407 | */ | |
2408 | } else { | |
2409 | buf = g_malloc(0x1000); | |
2410 | size = qemu_get_be32(f); | |
2411 | ||
2412 | while (size > 0) { | |
2413 | uint32_t read_chunk = MIN(size, 0x1000); | |
2414 | qemu_get_buffer(f, buf, read_chunk); | |
2415 | size -= read_chunk; | |
2416 | } | |
2417 | g_free(buf); | |
1925cebc | 2418 | } |
1925cebc AG |
2419 | } |
2420 | ||
acb5ea86 | 2421 | qemu_loadvm_state_cleanup(); |
ea375f9a JK |
2422 | cpu_synchronize_all_post_init(); |
2423 | ||
a672b469 AL |
2424 | return ret; |
2425 | } | |
2426 | ||
3f6df99d ZC |
2427 | int qemu_load_device_state(QEMUFile *f) |
2428 | { | |
2429 | MigrationIncomingState *mis = migration_incoming_get_current(); | |
2430 | int ret; | |
2431 | ||
2432 | /* Load QEMU_VM_SECTION_FULL section */ | |
2433 | ret = qemu_loadvm_state_main(f, mis); | |
2434 | if (ret < 0) { | |
2435 | error_report("Failed to load device state: %d", ret); | |
2436 | return ret; | |
2437 | } | |
2438 | ||
2439 | cpu_synchronize_all_post_init(); | |
2440 | return 0; | |
2441 | } | |
2442 | ||
5e22479a | 2443 | int save_snapshot(const char *name, Error **errp) |
a672b469 AL |
2444 | { |
2445 | BlockDriverState *bs, *bs1; | |
2446 | QEMUSnapshotInfo sn1, *sn = &sn1, old_sn1, *old_sn = &old_sn1; | |
ac8c19ba | 2447 | int ret = -1; |
a672b469 AL |
2448 | QEMUFile *f; |
2449 | int saved_vm_running; | |
c2c9a466 | 2450 | uint64_t vm_state_size; |
68b891ec | 2451 | qemu_timeval tv; |
7d631a11 | 2452 | struct tm tm; |
79b3c12a | 2453 | AioContext *aio_context; |
a672b469 | 2454 | |
377b21cc | 2455 | if (!replay_can_snapshot()) { |
4dd32b3d MA |
2456 | error_setg(errp, "Record/replay does not allow making snapshot " |
2457 | "right now. Try once more later."); | |
377b21cc PD |
2458 | return ret; |
2459 | } | |
2460 | ||
e9ff957a | 2461 | if (!bdrv_all_can_snapshot(&bs)) { |
927d6638 JQ |
2462 | error_setg(errp, "Device '%s' is writable but does not support " |
2463 | "snapshots", bdrv_get_device_name(bs)); | |
ac8c19ba | 2464 | return ret; |
feeee5ac MDCF |
2465 | } |
2466 | ||
0b461605 | 2467 | /* Delete old snapshots of the same name */ |
ac8c19ba | 2468 | if (name) { |
927d6638 | 2469 | ret = bdrv_all_delete_snapshot(name, &bs1, errp); |
ac8c19ba | 2470 | if (ret < 0) { |
927d6638 JQ |
2471 | error_prepend(errp, "Error while deleting snapshot on device " |
2472 | "'%s': ", bdrv_get_device_name(bs1)); | |
ac8c19ba PD |
2473 | return ret; |
2474 | } | |
0b461605 DL |
2475 | } |
2476 | ||
7cb14481 DL |
2477 | bs = bdrv_all_find_vmstate_bs(); |
2478 | if (bs == NULL) { | |
927d6638 | 2479 | error_setg(errp, "No block device can accept snapshots"); |
ac8c19ba | 2480 | return ret; |
a672b469 | 2481 | } |
79b3c12a | 2482 | aio_context = bdrv_get_aio_context(bs); |
a672b469 | 2483 | |
1354869c | 2484 | saved_vm_running = runstate_is_running(); |
560d027b JQ |
2485 | |
2486 | ret = global_state_store(); | |
2487 | if (ret) { | |
927d6638 | 2488 | error_setg(errp, "Error saving global state"); |
ac8c19ba | 2489 | return ret; |
560d027b | 2490 | } |
0461d5a6 | 2491 | vm_stop(RUN_STATE_SAVE_VM); |
a672b469 | 2492 | |
8649f2f9 SH |
2493 | bdrv_drain_all_begin(); |
2494 | ||
79b3c12a DL |
2495 | aio_context_acquire(aio_context); |
2496 | ||
cb499fb2 | 2497 | memset(sn, 0, sizeof(*sn)); |
a672b469 AL |
2498 | |
2499 | /* fill auxiliary fields */ | |
68b891ec | 2500 | qemu_gettimeofday(&tv); |
a672b469 AL |
2501 | sn->date_sec = tv.tv_sec; |
2502 | sn->date_nsec = tv.tv_usec * 1000; | |
bc72ad67 | 2503 | sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL); |
a672b469 | 2504 | |
7d631a11 MDCF |
2505 | if (name) { |
2506 | ret = bdrv_snapshot_find(bs, old_sn, name); | |
2507 | if (ret >= 0) { | |
2508 | pstrcpy(sn->name, sizeof(sn->name), old_sn->name); | |
2509 | pstrcpy(sn->id_str, sizeof(sn->id_str), old_sn->id_str); | |
2510 | } else { | |
2511 | pstrcpy(sn->name, sizeof(sn->name), name); | |
2512 | } | |
2513 | } else { | |
d7d9b528 BS |
2514 | /* cast below needed for OpenBSD where tv_sec is still 'long' */ |
2515 | localtime_r((const time_t *)&tv.tv_sec, &tm); | |
7d631a11 | 2516 | strftime(sn->name, sizeof(sn->name), "vm-%Y%m%d%H%M%S", &tm); |
7d631a11 MDCF |
2517 | } |
2518 | ||
a672b469 | 2519 | /* save the VM state */ |
45566e9c | 2520 | f = qemu_fopen_bdrv(bs, 1); |
a672b469 | 2521 | if (!f) { |
927d6638 | 2522 | error_setg(errp, "Could not open VM state file"); |
a672b469 AL |
2523 | goto the_end; |
2524 | } | |
927d6638 | 2525 | ret = qemu_savevm_state(f, errp); |
2d22b18f | 2526 | vm_state_size = qemu_ftell(f); |
a672b469 AL |
2527 | qemu_fclose(f); |
2528 | if (ret < 0) { | |
a672b469 AL |
2529 | goto the_end; |
2530 | } | |
2531 | ||
17e2a4a4 SH |
2532 | /* The bdrv_all_create_snapshot() call that follows acquires the AioContext |
2533 | * for itself. BDRV_POLL_WHILE() does not support nested locking because | |
2534 | * it only releases the lock once. Therefore synchronous I/O will deadlock | |
2535 | * unless we release the AioContext before bdrv_all_create_snapshot(). | |
2536 | */ | |
2537 | aio_context_release(aio_context); | |
2538 | aio_context = NULL; | |
2539 | ||
a9085f9b DL |
2540 | ret = bdrv_all_create_snapshot(sn, bs, vm_state_size, &bs); |
2541 | if (ret < 0) { | |
927d6638 JQ |
2542 | error_setg(errp, "Error while creating snapshot on '%s'", |
2543 | bdrv_get_device_name(bs)); | |
ac8c19ba | 2544 | goto the_end; |
a672b469 AL |
2545 | } |
2546 | ||
ac8c19ba PD |
2547 | ret = 0; |
2548 | ||
a672b469 | 2549 | the_end: |
17e2a4a4 SH |
2550 | if (aio_context) { |
2551 | aio_context_release(aio_context); | |
2552 | } | |
8649f2f9 SH |
2553 | |
2554 | bdrv_drain_all_end(); | |
2555 | ||
38ff78d3 | 2556 | if (saved_vm_running) { |
a672b469 | 2557 | vm_start(); |
38ff78d3 | 2558 | } |
ac8c19ba PD |
2559 | return ret; |
2560 | } | |
2561 | ||
5d6c599f AP |
2562 | void qmp_xen_save_devices_state(const char *filename, bool has_live, bool live, |
2563 | Error **errp) | |
a7ae8355 SS |
2564 | { |
2565 | QEMUFile *f; | |
8925839f | 2566 | QIOChannelFile *ioc; |
a7ae8355 SS |
2567 | int saved_vm_running; |
2568 | int ret; | |
2569 | ||
5d6c599f AP |
2570 | if (!has_live) { |
2571 | /* live default to true so old version of Xen tool stack can have a | |
2572 | * successfull live migration */ | |
2573 | live = true; | |
2574 | } | |
2575 | ||
a7ae8355 SS |
2576 | saved_vm_running = runstate_is_running(); |
2577 | vm_stop(RUN_STATE_SAVE_VM); | |
c69adea4 | 2578 | global_state_store_running(); |
a7ae8355 | 2579 | |
8925839f DB |
2580 | ioc = qio_channel_file_new_path(filename, O_WRONLY | O_CREAT, 0660, errp); |
2581 | if (!ioc) { | |
a7ae8355 SS |
2582 | goto the_end; |
2583 | } | |
6f01f136 | 2584 | qio_channel_set_name(QIO_CHANNEL(ioc), "migration-xen-save-state"); |
8925839f | 2585 | f = qemu_fopen_channel_output(QIO_CHANNEL(ioc)); |
032b79f7 | 2586 | object_unref(OBJECT(ioc)); |
a7ae8355 | 2587 | ret = qemu_save_device_state(f); |
96994fd1 | 2588 | if (ret < 0 || qemu_fclose(f) < 0) { |
c6bd8c70 | 2589 | error_setg(errp, QERR_IO_ERROR); |
5d6c599f AP |
2590 | } else { |
2591 | /* libxl calls the QMP command "stop" before calling | |
2592 | * "xen-save-devices-state" and in case of migration failure, libxl | |
2593 | * would call "cont". | |
2594 | * So call bdrv_inactivate_all (release locks) here to let the other | |
2595 | * side of the migration take controle of the images. | |
2596 | */ | |
2597 | if (live && !saved_vm_running) { | |
2598 | ret = bdrv_inactivate_all(); | |
2599 | if (ret) { | |
2600 | error_setg(errp, "%s: bdrv_inactivate_all() failed (%d)", | |
2601 | __func__, ret); | |
2602 | } | |
2603 | } | |
a7ae8355 SS |
2604 | } |
2605 | ||
2606 | the_end: | |
38ff78d3 | 2607 | if (saved_vm_running) { |
a7ae8355 | 2608 | vm_start(); |
38ff78d3 | 2609 | } |
a7ae8355 SS |
2610 | } |
2611 | ||
88c16567 WC |
2612 | void qmp_xen_load_devices_state(const char *filename, Error **errp) |
2613 | { | |
2614 | QEMUFile *f; | |
2615 | QIOChannelFile *ioc; | |
2616 | int ret; | |
2617 | ||
2618 | /* Guest must be paused before loading the device state; the RAM state | |
2619 | * will already have been loaded by xc | |
2620 | */ | |
2621 | if (runstate_is_running()) { | |
2622 | error_setg(errp, "Cannot update device state while vm is running"); | |
2623 | return; | |
2624 | } | |
2625 | vm_stop(RUN_STATE_RESTORE_VM); | |
2626 | ||
2627 | ioc = qio_channel_file_new_path(filename, O_RDONLY | O_BINARY, 0, errp); | |
2628 | if (!ioc) { | |
2629 | return; | |
2630 | } | |
6f01f136 | 2631 | qio_channel_set_name(QIO_CHANNEL(ioc), "migration-xen-load-state"); |
88c16567 | 2632 | f = qemu_fopen_channel_input(QIO_CHANNEL(ioc)); |
032b79f7 | 2633 | object_unref(OBJECT(ioc)); |
88c16567 | 2634 | |
88c16567 WC |
2635 | ret = qemu_loadvm_state(f); |
2636 | qemu_fclose(f); | |
2637 | if (ret < 0) { | |
2638 | error_setg(errp, QERR_IO_ERROR); | |
2639 | } | |
2640 | migration_incoming_state_destroy(); | |
2641 | } | |
2642 | ||
5e22479a | 2643 | int load_snapshot(const char *name, Error **errp) |
a672b469 | 2644 | { |
f0aa7a8b | 2645 | BlockDriverState *bs, *bs_vm_state; |
2d22b18f | 2646 | QEMUSnapshotInfo sn; |
a672b469 | 2647 | QEMUFile *f; |
751c6a17 | 2648 | int ret; |
79b3c12a | 2649 | AioContext *aio_context; |
b4b076da | 2650 | MigrationIncomingState *mis = migration_incoming_get_current(); |
a672b469 | 2651 | |
377b21cc | 2652 | if (!replay_can_snapshot()) { |
4dd32b3d MA |
2653 | error_setg(errp, "Record/replay does not allow loading snapshot " |
2654 | "right now. Try once more later."); | |
377b21cc PD |
2655 | return -EINVAL; |
2656 | } | |
2657 | ||
849f96e2 | 2658 | if (!bdrv_all_can_snapshot(&bs)) { |
927d6638 JQ |
2659 | error_setg(errp, |
2660 | "Device '%s' is writable but does not support snapshots", | |
2661 | bdrv_get_device_name(bs)); | |
849f96e2 DL |
2662 | return -ENOTSUP; |
2663 | } | |
723ccda1 DL |
2664 | ret = bdrv_all_find_snapshot(name, &bs); |
2665 | if (ret < 0) { | |
927d6638 JQ |
2666 | error_setg(errp, |
2667 | "Device '%s' does not have the requested snapshot '%s'", | |
2668 | bdrv_get_device_name(bs), name); | |
723ccda1 DL |
2669 | return ret; |
2670 | } | |
849f96e2 | 2671 | |
7cb14481 | 2672 | bs_vm_state = bdrv_all_find_vmstate_bs(); |
f0aa7a8b | 2673 | if (!bs_vm_state) { |
927d6638 | 2674 | error_setg(errp, "No block device supports snapshots"); |
f0aa7a8b MDCF |
2675 | return -ENOTSUP; |
2676 | } | |
79b3c12a | 2677 | aio_context = bdrv_get_aio_context(bs_vm_state); |
f0aa7a8b MDCF |
2678 | |
2679 | /* Don't even try to load empty VM states */ | |
79b3c12a | 2680 | aio_context_acquire(aio_context); |
f0aa7a8b | 2681 | ret = bdrv_snapshot_find(bs_vm_state, &sn, name); |
79b3c12a | 2682 | aio_context_release(aio_context); |
f0aa7a8b MDCF |
2683 | if (ret < 0) { |
2684 | return ret; | |
2685 | } else if (sn.vm_state_size == 0) { | |
927d6638 JQ |
2686 | error_setg(errp, "This is a disk-only snapshot. Revert to it " |
2687 | " offline using qemu-img"); | |
f0aa7a8b MDCF |
2688 | return -EINVAL; |
2689 | } | |
2690 | ||
a672b469 | 2691 | /* Flush all IO requests so they don't interfere with the new state. */ |
8649f2f9 | 2692 | bdrv_drain_all_begin(); |
a672b469 | 2693 | |
2b624fe0 | 2694 | ret = bdrv_all_goto_snapshot(name, &bs, errp); |
4c1cdbaa | 2695 | if (ret < 0) { |
2b624fe0 KW |
2696 | error_prepend(errp, "Could not load snapshot '%s' on '%s': ", |
2697 | name, bdrv_get_device_name(bs)); | |
8649f2f9 | 2698 | goto err_drain; |
a672b469 AL |
2699 | } |
2700 | ||
a672b469 | 2701 | /* restore the VM state */ |
f0aa7a8b | 2702 | f = qemu_fopen_bdrv(bs_vm_state, 0); |
a672b469 | 2703 | if (!f) { |
927d6638 | 2704 | error_setg(errp, "Could not open VM state file"); |
8649f2f9 SH |
2705 | ret = -EINVAL; |
2706 | goto err_drain; | |
a672b469 | 2707 | } |
f0aa7a8b | 2708 | |
aedbe192 | 2709 | qemu_system_reset(SHUTDOWN_CAUSE_NONE); |
b4b076da | 2710 | mis->from_src_file = f; |
f0aa7a8b | 2711 | |
79b3c12a DL |
2712 | aio_context_acquire(aio_context); |
2713 | ret = qemu_loadvm_state(f); | |
1575829d | 2714 | migration_incoming_state_destroy(); |
79b3c12a DL |
2715 | aio_context_release(aio_context); |
2716 | ||
8649f2f9 SH |
2717 | bdrv_drain_all_end(); |
2718 | ||
a672b469 | 2719 | if (ret < 0) { |
927d6638 | 2720 | error_setg(errp, "Error %d while loading VM state", ret); |
05f2401e | 2721 | return ret; |
a672b469 | 2722 | } |
f0aa7a8b | 2723 | |
05f2401e | 2724 | return 0; |
8649f2f9 SH |
2725 | |
2726 | err_drain: | |
2727 | bdrv_drain_all_end(); | |
2728 | return ret; | |
7b630349 JQ |
2729 | } |
2730 | ||
c5705a77 AK |
2731 | void vmstate_register_ram(MemoryRegion *mr, DeviceState *dev) |
2732 | { | |
fa53a0e5 | 2733 | qemu_ram_set_idstr(mr->ram_block, |
c5705a77 | 2734 | memory_region_name(mr), dev); |
b895de50 | 2735 | qemu_ram_set_migratable(mr->ram_block); |
c5705a77 AK |
2736 | } |
2737 | ||
2738 | void vmstate_unregister_ram(MemoryRegion *mr, DeviceState *dev) | |
2739 | { | |
fa53a0e5 | 2740 | qemu_ram_unset_idstr(mr->ram_block); |
b895de50 | 2741 | qemu_ram_unset_migratable(mr->ram_block); |
c5705a77 AK |
2742 | } |
2743 | ||
2744 | void vmstate_register_ram_global(MemoryRegion *mr) | |
2745 | { | |
2746 | vmstate_register_ram(mr, NULL); | |
2747 | } | |
1bfe5f05 JQ |
2748 | |
2749 | bool vmstate_check_only_migratable(const VMStateDescription *vmsd) | |
2750 | { | |
2751 | /* check needed if --only-migratable is specified */ | |
3df663e5 | 2752 | if (!migrate_get_current()->only_migratable) { |
1bfe5f05 JQ |
2753 | return true; |
2754 | } | |
2755 | ||
2756 | return !(vmsd && vmsd->unmigratable); | |
2757 | } |