1 /* 2 * QEMU live migration 3 * 4 * Copyright IBM, Corp. 2008 5 * 6 * Authors: 7 * Anthony Liguori <aliguori@us.ibm.com> 8 * 9 * This work is licensed under the terms of the GNU GPL, version 2. See 10 * the COPYING file in the top-level directory. 11 * 12 */ 13 14 #ifndef QEMU_MIGRATION_H 15 #define QEMU_MIGRATION_H 16 17 #include "qemu-common.h" 18 #include "qapi/qapi-types-migration.h" 19 #include "qemu/thread.h" 20 #include "exec/cpu-common.h" 21 #include "qemu/coroutine_int.h" 22 #include "hw/qdev.h" 23 #include "io/channel.h" 24 #include "net/announce.h" 25 26 struct PostcopyBlocktimeContext; 27 28 #define MIGRATION_RESUME_ACK_VALUE (1) 29 30 /* State for the incoming migration */ 31 struct MigrationIncomingState { 32 QEMUFile *from_src_file; 33 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 40 /* For network announces */ 41 AnnounceTimer announce_timer; 42 43 size_t largest_page_size; 44 bool have_fault_thread; 45 QemuThread fault_thread; 46 QemuSemaphore fault_thread_sem; 47 /* Set this when we want the fault thread to quit */ 48 bool fault_thread_quit; 49 50 bool have_listen_thread; 51 QemuThread listen_thread; 52 QemuSemaphore listen_thread_sem; 53 54 /* For the kernel to send us notifications */ 55 int userfault_fd; 56 /* To notify the fault_thread to wake, e.g., when need to quit */ 57 int userfault_event_fd; 58 QEMUFile *to_src_file; 59 QemuMutex rp_mutex; /* We send replies from multiple threads */ 60 /* RAMBlock of last request sent to source */ 61 RAMBlock *last_rb; 62 void *postcopy_tmp_page; 63 void *postcopy_tmp_zero_page; 64 /* PostCopyFD's for external userfaultfds & handlers of shared memory */ 65 GArray *postcopy_remote_fds; 66 67 QEMUBH *bh; 68 69 int state; 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; 75 QemuSemaphore colo_incoming_sem; 76 77 /* 78 * PostcopyBlocktimeContext to keep information for postcopy 79 * live migration, to calculate vCPU block time 80 * */ 81 struct PostcopyBlocktimeContext *blocktime_ctx; 82 83 /* notify PAUSED postcopy incoming migrations to try to continue */ 84 bool postcopy_recover_triggered; 85 QemuSemaphore postcopy_pause_sem_dst; 86 QemuSemaphore postcopy_pause_sem_fault; 87 88 /* List of listening socket addresses */ 89 SocketAddressList *socket_address_list; 90 }; 91 92 MigrationIncomingState *migration_incoming_get_current(void); 93 void migration_incoming_state_destroy(void); 94 /* 95 * Functions to work with blocktime context 96 */ 97 void fill_destination_postcopy_migration_info(MigrationInfo *info); 98 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; 112 113 struct MigrationState 114 { 115 /*< private >*/ 116 DeviceState parent_obj; 117 118 /*< public >*/ 119 size_t bytes_xfer; 120 size_t xfer_limit; 121 QemuThread thread; 122 QEMUBH *cleanup_bh; 123 QEMUFile *to_dst_file; 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; 130 131 /* 132 * Used to allow urgent requests to override rate limiting. 133 */ 134 QemuSemaphore rate_limit_sem; 135 136 /* 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 */ 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; 152 153 /* params from 'migrate-set-parameters' */ 154 MigrationParameters parameters; 155 156 int state; 157 158 /* State related to return path */ 159 struct { 160 QEMUFile *from_dst_file; 161 QemuThread rp_thread; 162 bool error; 163 QemuSemaphore rp_sem; 164 } rp_state; 165 166 double mbps; 167 /* Timestamp when recent migration starts (ms) */ 168 int64_t start_time; 169 /* Total time used by latest migration (ms) */ 170 int64_t total_time; 171 /* Timestamp when VM is down (ms) to migrate the last stuff */ 172 int64_t downtime_start; 173 int64_t downtime; 174 int64_t expected_downtime; 175 bool enabled_capabilities[MIGRATION_CAPABILITY__MAX]; 176 int64_t setup_time; 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; 183 184 /* Flag set once the migration has been asked to enter postcopy */ 185 bool start_postcopy; 186 /* Flag set after postcopy has sent the device state */ 187 bool postcopy_after_devices; 188 189 /* Flag set once the migration thread is running (and needs joining) */ 190 bool migration_thread_running; 191 192 /* Flag set once the migration thread called bdrv_inactivate_all */ 193 bool block_inactive; 194 195 /* Migration is paused due to pause-before-switchover */ 196 QemuSemaphore pause_sem; 197 198 /* The semaphore is used to notify COLO thread that failover is finished */ 199 QemuSemaphore colo_exit_sem; 200 201 /* 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; 205 206 /* The first error that has occurred. 207 We used the mutex to be able to return the 1st error message */ 208 Error *error; 209 /* mutex to protect errp */ 210 QemuMutex error_mutex; 211 212 /* 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; 215 216 /* 217 * Global switch on whether we need to store the global state 218 * during migration. 219 */ 220 bool store_global_state; 221 222 /* Whether the VM is only allowing for migratable devices */ 223 bool only_migratable; 224 225 /* Whether we send QEMU_VM_CONFIGURATION during migration */ 226 bool send_configuration; 227 /* Whether we send section footer during migration */ 228 bool send_section_footer; 229 230 /* Needed by postcopy-pause state */ 231 QemuSemaphore postcopy_pause_sem; 232 QemuSemaphore postcopy_pause_rp_sem; 233 /* 234 * Whether we abort the migration if decompression errors are 235 * detected at the destination. It is left at false for qemu 236 * older than 3.0, since only newer qemu sends streams that 237 * do not trigger spurious decompression errors. 238 */ 239 bool decompress_error_check; 240 }; 241 242 void migrate_set_state(int *state, int old_state, int new_state); 243 244 void migration_fd_process_incoming(QEMUFile *f); 245 void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp); 246 void migration_incoming_process(void); 247 248 bool migration_has_all_channels(void); 249 250 uint64_t migrate_max_downtime(void); 251 252 void migrate_set_error(MigrationState *s, const Error *error); 253 void migrate_fd_error(MigrationState *s, const Error *error); 254 255 void migrate_fd_connect(MigrationState *s, Error *error_in); 256 257 bool migration_is_setup_or_active(int state); 258 259 void migrate_init(MigrationState *s); 260 bool migration_is_blocked(Error **errp); 261 /* True if outgoing migration has entered postcopy phase */ 262 bool migration_in_postcopy(void); 263 MigrationState *migrate_get_current(void); 264 265 bool migrate_postcopy(void); 266 267 bool migrate_release_ram(void); 268 bool migrate_postcopy_ram(void); 269 bool migrate_zero_blocks(void); 270 bool migrate_dirty_bitmaps(void); 271 bool migrate_ignore_shared(void); 272 273 bool migrate_auto_converge(void); 274 bool migrate_use_multifd(void); 275 bool migrate_pause_before_switchover(void); 276 int migrate_multifd_channels(void); 277 int migrate_multifd_page_count(void); 278 279 int migrate_use_xbzrle(void); 280 int64_t migrate_xbzrle_cache_size(void); 281 bool migrate_colo_enabled(void); 282 283 bool migrate_use_block(void); 284 bool migrate_use_block_incremental(void); 285 int migrate_max_cpu_throttle(void); 286 bool migrate_use_return_path(void); 287 288 uint64_t ram_get_total_transferred_pages(void); 289 290 bool migrate_use_compression(void); 291 int migrate_compress_level(void); 292 int migrate_compress_threads(void); 293 int migrate_compress_wait_thread(void); 294 int migrate_decompress_threads(void); 295 bool migrate_use_events(void); 296 bool migrate_postcopy_blocktime(void); 297 298 /* Sending on the return path - generic and then for each message type */ 299 void migrate_send_rp_shut(MigrationIncomingState *mis, 300 uint32_t value); 301 void migrate_send_rp_pong(MigrationIncomingState *mis, 302 uint32_t value); 303 int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname, 304 ram_addr_t start, size_t len); 305 void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis, 306 char *block_name); 307 void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value); 308 309 void dirty_bitmap_mig_before_vm_start(void); 310 void init_dirty_bitmap_incoming_migration(void); 311 void migrate_add_address(SocketAddress *address); 312 313 int foreach_not_ignored_block(RAMBlockIterFunc func, void *opaque); 314 315 #define qemu_ram_foreach_block \ 316 #warning "Use foreach_not_ignored_block in migration code" 317 318 void migration_make_urgent_request(void); 319 void migration_consume_urgent_request(void); 320 321 #endif 322