]>
Commit | Line | Data |
---|---|---|
5bb7910a AL |
1 | /* |
2 | * QEMU live migration | |
3 | * | |
4 | * Copyright IBM, Corp. 2008 | |
5 | * | |
6 | * Authors: | |
7 | * Anthony Liguori <[email protected]> | |
8 | * | |
9 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
10 | * the COPYING file in the top-level directory. | |
11 | * | |
12 | */ | |
13 | ||
14 | #ifndef QEMU_MIGRATION_H | |
15 | #define QEMU_MIGRATION_H | |
16 | ||
376253ec | 17 | #include "qemu-common.h" |
9af23989 | 18 | #include "qapi/qapi-types-migration.h" |
9848a404 | 19 | #include "qemu/thread.h" |
43487c67 | 20 | #include "exec/cpu-common.h" |
25d0c16f | 21 | #include "qemu/coroutine_int.h" |
e5cb7e76 | 22 | #include "hw/qdev.h" |
4f0fae7f | 23 | #include "io/channel.h" |
7659505c | 24 | #include "net/announce.h" |
376253ec | 25 | |
2a4c42f1 AP |
26 | struct PostcopyBlocktimeContext; |
27 | ||
13955b89 PX |
28 | #define MIGRATION_RESUME_ACK_VALUE (1) |
29 | ||
bca7856a DDAG |
30 | /* State for the incoming migration */ |
31 | struct MigrationIncomingState { | |
42e2aa56 | 32 | QEMUFile *from_src_file; |
1a8f46f8 | 33 | |
7b89bf27 DDAG |
34 | /* |
35 | * Free at the start of the main state load, set as the main thread finishes | |
36 | * loading state. | |
37 | */ | |
38 | QemuEvent main_thread_load_event; | |
39 | ||
7659505c DDAG |
40 | /* For network announces */ |
41 | AnnounceTimer announce_timer; | |
42 | ||
67f11b5c | 43 | size_t largest_page_size; |
c4faeed2 | 44 | bool have_fault_thread; |
f0a227ad DDAG |
45 | QemuThread fault_thread; |
46 | QemuSemaphore fault_thread_sem; | |
64f615fe PX |
47 | /* Set this when we want the fault thread to quit */ |
48 | bool fault_thread_quit; | |
f0a227ad | 49 | |
c76201ab DDAG |
50 | bool have_listen_thread; |
51 | QemuThread listen_thread; | |
52 | QemuSemaphore listen_thread_sem; | |
53 | ||
1caddf8a DDAG |
54 | /* For the kernel to send us notifications */ |
55 | int userfault_fd; | |
64f615fe PX |
56 | /* To notify the fault_thread to wake, e.g., when need to quit */ |
57 | int userfault_event_fd; | |
2e37701e | 58 | QEMUFile *to_src_file; |
6decec93 | 59 | QemuMutex rp_mutex; /* We send replies from multiple threads */ |
096bf4c8 DDAG |
60 | /* RAMBlock of last request sent to source */ |
61 | RAMBlock *last_rb; | |
696ed9a9 | 62 | void *postcopy_tmp_page; |
41d84210 | 63 | void *postcopy_tmp_zero_page; |
00fa4fc8 DDAG |
64 | /* PostCopyFD's for external userfaultfds & handlers of shared memory */ |
65 | GArray *postcopy_remote_fds; | |
2e37701e | 66 | |
0aa6aefc DL |
67 | QEMUBH *bh; |
68 | ||
93d7af6f | 69 | int state; |
25d0c16f HZ |
70 | |
71 | bool have_colo_incoming_thread; | |
72 | QemuThread colo_incoming_thread; | |
73 | /* The coroutine we should enter (back) after failover */ | |
74 | Coroutine *migration_incoming_co; | |
c937b9a6 | 75 | QemuSemaphore colo_incoming_sem; |
2a4c42f1 AP |
76 | |
77 | /* | |
78 | * PostcopyBlocktimeContext to keep information for postcopy | |
79 | * live migration, to calculate vCPU block time | |
80 | * */ | |
81 | struct PostcopyBlocktimeContext *blocktime_ctx; | |
b411b844 PX |
82 | |
83 | /* notify PAUSED postcopy incoming migrations to try to continue */ | |
02affd41 | 84 | bool postcopy_recover_triggered; |
b411b844 | 85 | QemuSemaphore postcopy_pause_sem_dst; |
3a7804c3 | 86 | QemuSemaphore postcopy_pause_sem_fault; |
9aca82ba JQ |
87 | |
88 | /* List of listening socket addresses */ | |
89 | SocketAddressList *socket_address_list; | |
bca7856a DDAG |
90 | }; |
91 | ||
92 | MigrationIncomingState *migration_incoming_get_current(void); | |
bca7856a | 93 | void migration_incoming_state_destroy(void); |
65ace060 AP |
94 | /* |
95 | * Functions to work with blocktime context | |
96 | */ | |
97 | void fill_destination_postcopy_migration_info(MigrationInfo *info); | |
bca7856a | 98 | |
e5cb7e76 PX |
99 | #define TYPE_MIGRATION "migration" |
100 | ||
101 | #define MIGRATION_CLASS(klass) \ | |
102 | OBJECT_CLASS_CHECK(MigrationClass, (klass), TYPE_MIGRATION) | |
103 | #define MIGRATION_OBJ(obj) \ | |
104 | OBJECT_CHECK(MigrationState, (obj), TYPE_MIGRATION) | |
105 | #define MIGRATION_GET_CLASS(obj) \ | |
106 | OBJECT_GET_CLASS(MigrationClass, (obj), TYPE_MIGRATION) | |
107 | ||
108 | typedef struct MigrationClass { | |
109 | /*< private >*/ | |
110 | DeviceClass parent_class; | |
111 | } MigrationClass; | |
112 | ||
22f00a44 | 113 | struct MigrationState |
065e2813 | 114 | { |
e5cb7e76 PX |
115 | /*< private >*/ |
116 | DeviceState parent_obj; | |
117 | ||
118 | /*< public >*/ | |
9848a404 JQ |
119 | size_t bytes_xfer; |
120 | size_t xfer_limit; | |
9848a404 | 121 | QemuThread thread; |
bb1fadc4 | 122 | QEMUBH *cleanup_bh; |
89a02a9f | 123 | QEMUFile *to_dst_file; |
62df066f PX |
124 | /* |
125 | * Protects to_dst_file pointer. We need to make sure we won't | |
126 | * yield or hang during the critical section, since this lock will | |
127 | * be used in OOB command handler. | |
128 | */ | |
129 | QemuMutex qemu_file_lock; | |
2594f56d | 130 | |
ad767bed DDAG |
131 | /* |
132 | * Used to allow urgent requests to override rate limiting. | |
133 | */ | |
134 | QemuSemaphore rate_limit_sem; | |
135 | ||
aecbfe9c XG |
136 | /* pages already send at the beginning of current iteration */ |
137 | uint64_t iteration_initial_pages; | |
138 | ||
139 | /* pages transferred per second */ | |
140 | double pages_per_second; | |
141 | ||
142 | /* bytes already send at the beginning of current iteration */ | |
b15df1ae PX |
143 | uint64_t iteration_initial_bytes; |
144 | /* time at the start of current iteration */ | |
145 | int64_t iteration_start_time; | |
146 | /* | |
147 | * The final stage happens when the remaining data is smaller than | |
148 | * this threshold; it's calculated from the requested downtime and | |
149 | * measured bandwidth | |
150 | */ | |
151 | int64_t threshold_size; | |
152 | ||
a0762d9e | 153 | /* params from 'migrate-set-parameters' */ |
2594f56d | 154 | MigrationParameters parameters; |
f8bbc128 | 155 | |
f8bbc128 | 156 | int state; |
70b20477 DDAG |
157 | |
158 | /* State related to return path */ | |
159 | struct { | |
160 | QEMUFile *from_dst_file; | |
161 | QemuThread rp_thread; | |
162 | bool error; | |
edd090c7 | 163 | QemuSemaphore rp_sem; |
70b20477 DDAG |
164 | } rp_state; |
165 | ||
7e114f8c | 166 | double mbps; |
4af246a3 PX |
167 | /* Timestamp when recent migration starts (ms) */ |
168 | int64_t start_time; | |
169 | /* Total time used by latest migration (ms) */ | |
d5f8a570 | 170 | int64_t total_time; |
64909f97 PX |
171 | /* Timestamp when VM is down (ms) to migrate the last stuff */ |
172 | int64_t downtime_start; | |
9c5a9fcf | 173 | int64_t downtime; |
2c52ddf1 | 174 | int64_t expected_downtime; |
7fb1cf16 | 175 | bool enabled_capabilities[MIGRATION_CAPABILITY__MAX]; |
ed4fbd10 | 176 | int64_t setup_time; |
7287cbd4 PX |
177 | /* |
178 | * Whether guest was running when we enter the completion stage. | |
179 | * If migration is interrupted by any reason, we need to continue | |
180 | * running the guest on source. | |
181 | */ | |
182 | bool vm_was_running; | |
4886a1bc DDAG |
183 | |
184 | /* Flag set once the migration has been asked to enter postcopy */ | |
185 | bool start_postcopy; | |
b82fc321 DDAG |
186 | /* Flag set after postcopy has sent the device state */ |
187 | bool postcopy_after_devices; | |
1d34e4bf DDAG |
188 | |
189 | /* Flag set once the migration thread is running (and needs joining) */ | |
190 | bool migration_thread_running; | |
6c595cde | 191 | |
1d2acc31 HZ |
192 | /* Flag set once the migration thread called bdrv_inactivate_all */ |
193 | bool block_inactive; | |
194 | ||
e91d8951 DDAG |
195 | /* Migration is paused due to pause-before-switchover */ |
196 | QemuSemaphore pause_sem; | |
197 | ||
c937b9a6 HZ |
198 | /* The semaphore is used to notify COLO thread that failover is finished */ |
199 | QemuSemaphore colo_exit_sem; | |
d59ce6f3 | 200 | |
479125d5 HZ |
201 | /* The semaphore is used to notify COLO thread to do checkpoint */ |
202 | QemuSemaphore colo_checkpoint_sem; | |
203 | int64_t colo_checkpoint_time; | |
204 | QEMUTimer *colo_delay_timer; | |
205 | ||
87db1a7d JQ |
206 | /* The first error that has occurred. |
207 | We used the mutex to be able to return the 1st error message */ | |
d59ce6f3 | 208 | Error *error; |
87db1a7d JQ |
209 | /* mutex to protect errp */ |
210 | QemuMutex error_mutex; | |
211 | ||
2833c59b JQ |
212 | /* Do we have to clean up -b/-i from old migrate parameters */ |
213 | /* This feature is deprecated and will be removed */ | |
214 | bool must_remove_block_options; | |
5272298c PX |
215 | |
216 | /* | |
217 | * Global switch on whether we need to store the global state | |
218 | * during migration. | |
219 | */ | |
220 | bool store_global_state; | |
3df663e5 | 221 | |
71dd4c1a PX |
222 | /* Whether we send QEMU_VM_CONFIGURATION during migration */ |
223 | bool send_configuration; | |
15c38503 PX |
224 | /* Whether we send section footer during migration */ |
225 | bool send_section_footer; | |
b23c2ade PX |
226 | |
227 | /* Needed by postcopy-pause state */ | |
228 | QemuSemaphore postcopy_pause_sem; | |
14b1742e | 229 | QemuSemaphore postcopy_pause_rp_sem; |
f548222c XG |
230 | /* |
231 | * Whether we abort the migration if decompression errors are | |
232 | * detected at the destination. It is left at false for qemu | |
233 | * older than 3.0, since only newer qemu sends streams that | |
234 | * do not trigger spurious decompression errors. | |
235 | */ | |
236 | bool decompress_error_check; | |
065e2813 AL |
237 | }; |
238 | ||
48781e5b HZ |
239 | void migrate_set_state(int *state, int old_state, int new_state); |
240 | ||
22724f49 | 241 | void migration_fd_process_incoming(QEMUFile *f); |
49ed0d24 | 242 | void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp); |
36c2f8be | 243 | void migration_incoming_process(void); |
511c0231 | 244 | |
428d8908 JQ |
245 | bool migration_has_all_channels(void); |
246 | ||
a0a3fd60 GC |
247 | uint64_t migrate_max_downtime(void); |
248 | ||
87db1a7d | 249 | void migrate_set_error(MigrationState *s, const Error *error); |
d59ce6f3 | 250 | void migrate_fd_error(MigrationState *s, const Error *error); |
065e2813 | 251 | |
cce8040b | 252 | void migrate_fd_connect(MigrationState *s, Error *error_in); |
065e2813 | 253 | |
3d63da16 JL |
254 | bool migration_is_setup_or_active(int state); |
255 | ||
3e0c8050 | 256 | void migrate_init(MigrationState *s); |
24f3902b | 257 | bool migration_is_blocked(Error **errp); |
9ec055ae | 258 | /* True if outgoing migration has entered postcopy phase */ |
5727309d | 259 | bool migration_in_postcopy(void); |
859bc756 | 260 | MigrationState *migrate_get_current(void); |
99a0db9b | 261 | |
58110f0a VSO |
262 | bool migrate_postcopy(void); |
263 | ||
53f09a10 | 264 | bool migrate_release_ram(void); |
53dd370c | 265 | bool migrate_postcopy_ram(void); |
323004a3 | 266 | bool migrate_zero_blocks(void); |
55efc8c2 | 267 | bool migrate_dirty_bitmaps(void); |
18269069 | 268 | bool migrate_ignore_shared(void); |
60d9222c | 269 | |
bde1e2ec | 270 | bool migrate_auto_converge(void); |
30126bbf | 271 | bool migrate_use_multifd(void); |
93fbd031 | 272 | bool migrate_pause_before_switchover(void); |
4075fb1c | 273 | int migrate_multifd_channels(void); |
bde1e2ec | 274 | |
17ad9b35 OW |
275 | int migrate_use_xbzrle(void); |
276 | int64_t migrate_xbzrle_cache_size(void); | |
35a6ed4f | 277 | bool migrate_colo_enabled(void); |
17ad9b35 | 278 | |
2833c59b JQ |
279 | bool migrate_use_block(void); |
280 | bool migrate_use_block_incremental(void); | |
4cbc9c7f | 281 | int migrate_max_cpu_throttle(void); |
c788ada8 | 282 | bool migrate_use_return_path(void); |
2833c59b | 283 | |
aecbfe9c XG |
284 | uint64_t ram_get_total_transferred_pages(void); |
285 | ||
8706d2d5 LL |
286 | bool migrate_use_compression(void); |
287 | int migrate_compress_level(void); | |
288 | int migrate_compress_threads(void); | |
1d58872a | 289 | int migrate_compress_wait_thread(void); |
3fcb38c2 | 290 | int migrate_decompress_threads(void); |
b05dc723 | 291 | bool migrate_use_events(void); |
f22f928e | 292 | bool migrate_postcopy_blocktime(void); |
8706d2d5 | 293 | |
6decec93 | 294 | /* Sending on the return path - generic and then for each message type */ |
6decec93 DDAG |
295 | void migrate_send_rp_shut(MigrationIncomingState *mis, |
296 | uint32_t value); | |
297 | void migrate_send_rp_pong(MigrationIncomingState *mis, | |
298 | uint32_t value); | |
d6208e35 | 299 | int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname, |
1e2d90eb | 300 | ram_addr_t start, size_t len); |
a335debb PX |
301 | void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis, |
302 | char *block_name); | |
13955b89 | 303 | void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value); |
6decec93 | 304 | |
b35ebdf0 VSO |
305 | void dirty_bitmap_mig_before_vm_start(void); |
306 | void init_dirty_bitmap_incoming_migration(void); | |
9aca82ba | 307 | void migrate_add_address(SocketAddress *address); |
b35ebdf0 | 308 | |
fbd162e6 YK |
309 | int foreach_not_ignored_block(RAMBlockIterFunc func, void *opaque); |
310 | ||
343f632c | 311 | #define qemu_ram_foreach_block \ |
fbd162e6 | 312 | #warning "Use foreach_not_ignored_block in migration code" |
343f632c | 313 | |
ad767bed DDAG |
314 | void migration_make_urgent_request(void); |
315 | void migration_consume_urgent_request(void); | |
316 | ||
5bb7910a | 317 | #endif |