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 "qapi/qapi-types-migration.h" 18 #include "qemu/thread.h" 19 #include "exec/cpu-common.h" 20 #include "qemu/coroutine_int.h" 21 #include "hw/qdev.h" 22 #include "io/channel.h" 23 #include "net/announce.h" 24 25 struct PostcopyBlocktimeContext; 26 27 #define MIGRATION_RESUME_ACK_VALUE (1) 28 29 /* State for the incoming migration */ 30 struct MigrationIncomingState { 31 QEMUFile *from_src_file; 32 33 /* 34 * Free at the start of the main state load, set as the main thread finishes 35 * loading state. 36 */ 37 QemuEvent main_thread_load_event; 38 39 /* For network announces */ 40 AnnounceTimer announce_timer; 41 42 size_t largest_page_size; 43 bool have_fault_thread; 44 QemuThread fault_thread; 45 QemuSemaphore fault_thread_sem; 46 /* Set this when we want the fault thread to quit */ 47 bool fault_thread_quit; 48 49 bool have_listen_thread; 50 QemuThread listen_thread; 51 QemuSemaphore listen_thread_sem; 52 53 /* For the kernel to send us notifications */ 54 int userfault_fd; 55 /* To notify the fault_thread to wake, e.g., when need to quit */ 56 int userfault_event_fd; 57 QEMUFile *to_src_file; 58 QemuMutex rp_mutex; /* We send replies from multiple threads */ 59 /* RAMBlock of last request sent to source */ 60 RAMBlock *last_rb; 61 void *postcopy_tmp_page; 62 void *postcopy_tmp_zero_page; 63 /* PostCopyFD's for external userfaultfds & handlers of shared memory */ 64 GArray *postcopy_remote_fds; 65 66 QEMUBH *bh; 67 68 int state; 69 70 bool have_colo_incoming_thread; 71 QemuThread colo_incoming_thread; 72 /* The coroutine we should enter (back) after failover */ 73 Coroutine *migration_incoming_co; 74 QemuSemaphore colo_incoming_sem; 75 76 /* 77 * PostcopyBlocktimeContext to keep information for postcopy 78 * live migration, to calculate vCPU block time 79 * */ 80 struct PostcopyBlocktimeContext *blocktime_ctx; 81 82 /* notify PAUSED postcopy incoming migrations to try to continue */ 83 bool postcopy_recover_triggered; 84 QemuSemaphore postcopy_pause_sem_dst; 85 QemuSemaphore postcopy_pause_sem_fault; 86 87 /* List of listening socket addresses */ 88 SocketAddressList *socket_address_list; 89 }; 90 91 MigrationIncomingState *migration_incoming_get_current(void); 92 void migration_incoming_state_destroy(void); 93 /* 94 * Functions to work with blocktime context 95 */ 96 void fill_destination_postcopy_migration_info(MigrationInfo *info); 97 98 #define TYPE_MIGRATION "migration" 99 100 #define MIGRATION_CLASS(klass) \ 101 OBJECT_CLASS_CHECK(MigrationClass, (klass), TYPE_MIGRATION) 102 #define MIGRATION_OBJ(obj) \ 103 OBJECT_CHECK(MigrationState, (obj), TYPE_MIGRATION) 104 #define MIGRATION_GET_CLASS(obj) \ 105 OBJECT_GET_CLASS(MigrationClass, (obj), TYPE_MIGRATION) 106 107 typedef struct MigrationClass { 108 /*< private >*/ 109 DeviceClass parent_class; 110 } MigrationClass; 111 112 struct MigrationState 113 { 114 /*< private >*/ 115 DeviceState parent_obj; 116 117 /*< public >*/ 118 size_t bytes_xfer; 119 QemuThread thread; 120 QEMUBH *cleanup_bh; 121 QEMUFile *to_dst_file; 122 /* 123 * Protects to_dst_file pointer. We need to make sure we won't 124 * yield or hang during the critical section, since this lock will 125 * be used in OOB command handler. 126 */ 127 QemuMutex qemu_file_lock; 128 129 /* 130 * Used to allow urgent requests to override rate limiting. 131 */ 132 QemuSemaphore rate_limit_sem; 133 134 /* pages already send at the beginning of current iteration */ 135 uint64_t iteration_initial_pages; 136 137 /* pages transferred per second */ 138 double pages_per_second; 139 140 /* bytes already send at the beginning of current iteration */ 141 uint64_t iteration_initial_bytes; 142 /* time at the start of current iteration */ 143 int64_t iteration_start_time; 144 /* 145 * The final stage happens when the remaining data is smaller than 146 * this threshold; it's calculated from the requested downtime and 147 * measured bandwidth 148 */ 149 int64_t threshold_size; 150 151 /* params from 'migrate-set-parameters' */ 152 MigrationParameters parameters; 153 154 int state; 155 156 /* State related to return path */ 157 struct { 158 QEMUFile *from_dst_file; 159 QemuThread rp_thread; 160 bool error; 161 QemuSemaphore rp_sem; 162 } rp_state; 163 164 double mbps; 165 /* Timestamp when recent migration starts (ms) */ 166 int64_t start_time; 167 /* Total time used by latest migration (ms) */ 168 int64_t total_time; 169 /* Timestamp when VM is down (ms) to migrate the last stuff */ 170 int64_t downtime_start; 171 int64_t downtime; 172 int64_t expected_downtime; 173 bool enabled_capabilities[MIGRATION_CAPABILITY__MAX]; 174 int64_t setup_time; 175 /* 176 * Whether guest was running when we enter the completion stage. 177 * If migration is interrupted by any reason, we need to continue 178 * running the guest on source. 179 */ 180 bool vm_was_running; 181 182 /* Flag set once the migration has been asked to enter postcopy */ 183 bool start_postcopy; 184 /* Flag set after postcopy has sent the device state */ 185 bool postcopy_after_devices; 186 187 /* Flag set once the migration thread is running (and needs joining) */ 188 bool migration_thread_running; 189 190 /* Flag set once the migration thread called bdrv_inactivate_all */ 191 bool block_inactive; 192 193 /* Migration is paused due to pause-before-switchover */ 194 QemuSemaphore pause_sem; 195 196 /* The semaphore is used to notify COLO thread that failover is finished */ 197 QemuSemaphore colo_exit_sem; 198 199 /* The semaphore is used to notify COLO thread to do checkpoint */ 200 QemuSemaphore colo_checkpoint_sem; 201 int64_t colo_checkpoint_time; 202 QEMUTimer *colo_delay_timer; 203 204 /* The first error that has occurred. 205 We used the mutex to be able to return the 1st error message */ 206 Error *error; 207 /* mutex to protect errp */ 208 QemuMutex error_mutex; 209 210 /* Do we have to clean up -b/-i from old migrate parameters */ 211 /* This feature is deprecated and will be removed */ 212 bool must_remove_block_options; 213 214 /* 215 * Global switch on whether we need to store the global state 216 * during migration. 217 */ 218 bool store_global_state; 219 220 /* Whether we send QEMU_VM_CONFIGURATION during migration */ 221 bool send_configuration; 222 /* Whether we send section footer during migration */ 223 bool send_section_footer; 224 225 /* Needed by postcopy-pause state */ 226 QemuSemaphore postcopy_pause_sem; 227 QemuSemaphore postcopy_pause_rp_sem; 228 /* 229 * Whether we abort the migration if decompression errors are 230 * detected at the destination. It is left at false for qemu 231 * older than 3.0, since only newer qemu sends streams that 232 * do not trigger spurious decompression errors. 233 */ 234 bool decompress_error_check; 235 }; 236 237 void migrate_set_state(int *state, int old_state, int new_state); 238 239 void migration_fd_process_incoming(QEMUFile *f); 240 void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp); 241 void migration_incoming_process(void); 242 243 bool migration_has_all_channels(void); 244 245 uint64_t migrate_max_downtime(void); 246 247 void migrate_set_error(MigrationState *s, const Error *error); 248 void migrate_fd_error(MigrationState *s, const Error *error); 249 250 void migrate_fd_connect(MigrationState *s, Error *error_in); 251 252 bool migration_is_setup_or_active(int state); 253 254 void migrate_init(MigrationState *s); 255 bool migration_is_blocked(Error **errp); 256 /* True if outgoing migration has entered postcopy phase */ 257 bool migration_in_postcopy(void); 258 MigrationState *migrate_get_current(void); 259 260 bool migrate_postcopy(void); 261 262 bool migrate_release_ram(void); 263 bool migrate_postcopy_ram(void); 264 bool migrate_zero_blocks(void); 265 bool migrate_dirty_bitmaps(void); 266 bool migrate_ignore_shared(void); 267 268 bool migrate_auto_converge(void); 269 bool migrate_use_multifd(void); 270 bool migrate_pause_before_switchover(void); 271 int migrate_multifd_channels(void); 272 273 int migrate_use_xbzrle(void); 274 int64_t migrate_xbzrle_cache_size(void); 275 bool migrate_colo_enabled(void); 276 277 bool migrate_use_block(void); 278 bool migrate_use_block_incremental(void); 279 int migrate_max_cpu_throttle(void); 280 bool migrate_use_return_path(void); 281 282 uint64_t ram_get_total_transferred_pages(void); 283 284 bool migrate_use_compression(void); 285 int migrate_compress_level(void); 286 int migrate_compress_threads(void); 287 int migrate_compress_wait_thread(void); 288 int migrate_decompress_threads(void); 289 bool migrate_use_events(void); 290 bool migrate_postcopy_blocktime(void); 291 292 /* Sending on the return path - generic and then for each message type */ 293 void migrate_send_rp_shut(MigrationIncomingState *mis, 294 uint32_t value); 295 void migrate_send_rp_pong(MigrationIncomingState *mis, 296 uint32_t value); 297 int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname, 298 ram_addr_t start, size_t len); 299 void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis, 300 char *block_name); 301 void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value); 302 303 void dirty_bitmap_mig_before_vm_start(void); 304 void init_dirty_bitmap_incoming_migration(void); 305 void migrate_add_address(SocketAddress *address); 306 307 int foreach_not_ignored_block(RAMBlockIterFunc func, void *opaque); 308 309 #define qemu_ram_foreach_block \ 310 #warning "Use foreach_not_ignored_block in migration code" 311 312 void migration_make_urgent_request(void); 313 void migration_consume_urgent_request(void); 314 315 #endif 316