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 89 MigrationIncomingState *migration_incoming_get_current(void); 90 void migration_incoming_state_destroy(void); 91 /* 92 * Functions to work with blocktime context 93 */ 94 void fill_destination_postcopy_migration_info(MigrationInfo *info); 95 96 #define TYPE_MIGRATION "migration" 97 98 #define MIGRATION_CLASS(klass) \ 99 OBJECT_CLASS_CHECK(MigrationClass, (klass), TYPE_MIGRATION) 100 #define MIGRATION_OBJ(obj) \ 101 OBJECT_CHECK(MigrationState, (obj), TYPE_MIGRATION) 102 #define MIGRATION_GET_CLASS(obj) \ 103 OBJECT_GET_CLASS(MigrationClass, (obj), TYPE_MIGRATION) 104 105 typedef struct MigrationClass { 106 /*< private >*/ 107 DeviceClass parent_class; 108 } MigrationClass; 109 110 struct MigrationState 111 { 112 /*< private >*/ 113 DeviceState parent_obj; 114 115 /*< public >*/ 116 size_t bytes_xfer; 117 size_t xfer_limit; 118 QemuThread thread; 119 QEMUBH *cleanup_bh; 120 QEMUFile *to_dst_file; 121 /* 122 * Protects to_dst_file pointer. We need to make sure we won't 123 * yield or hang during the critical section, since this lock will 124 * be used in OOB command handler. 125 */ 126 QemuMutex qemu_file_lock; 127 128 /* 129 * Used to allow urgent requests to override rate limiting. 130 */ 131 QemuSemaphore rate_limit_sem; 132 133 /* pages already send at the beginning of current iteration */ 134 uint64_t iteration_initial_pages; 135 136 /* pages transferred per second */ 137 double pages_per_second; 138 139 /* bytes already send at the beginning of current iteration */ 140 uint64_t iteration_initial_bytes; 141 /* time at the start of current iteration */ 142 int64_t iteration_start_time; 143 /* 144 * The final stage happens when the remaining data is smaller than 145 * this threshold; it's calculated from the requested downtime and 146 * measured bandwidth 147 */ 148 int64_t threshold_size; 149 150 /* params from 'migrate-set-parameters' */ 151 MigrationParameters parameters; 152 153 int state; 154 155 /* State related to return path */ 156 struct { 157 QEMUFile *from_dst_file; 158 QemuThread rp_thread; 159 bool error; 160 QemuSemaphore rp_sem; 161 } rp_state; 162 163 double mbps; 164 /* Timestamp when recent migration starts (ms) */ 165 int64_t start_time; 166 /* Total time used by latest migration (ms) */ 167 int64_t total_time; 168 /* Timestamp when VM is down (ms) to migrate the last stuff */ 169 int64_t downtime_start; 170 int64_t downtime; 171 int64_t expected_downtime; 172 bool enabled_capabilities[MIGRATION_CAPABILITY__MAX]; 173 int64_t setup_time; 174 /* 175 * Whether guest was running when we enter the completion stage. 176 * If migration is interrupted by any reason, we need to continue 177 * running the guest on source. 178 */ 179 bool vm_was_running; 180 181 /* Flag set once the migration has been asked to enter postcopy */ 182 bool start_postcopy; 183 /* Flag set after postcopy has sent the device state */ 184 bool postcopy_after_devices; 185 186 /* Flag set once the migration thread is running (and needs joining) */ 187 bool migration_thread_running; 188 189 /* Flag set once the migration thread called bdrv_inactivate_all */ 190 bool block_inactive; 191 192 /* Migration is paused due to pause-before-switchover */ 193 QemuSemaphore pause_sem; 194 195 /* The semaphore is used to notify COLO thread that failover is finished */ 196 QemuSemaphore colo_exit_sem; 197 198 /* The semaphore is used to notify COLO thread to do checkpoint */ 199 QemuSemaphore colo_checkpoint_sem; 200 int64_t colo_checkpoint_time; 201 QEMUTimer *colo_delay_timer; 202 203 /* The first error that has occurred. 204 We used the mutex to be able to return the 1st error message */ 205 Error *error; 206 /* mutex to protect errp */ 207 QemuMutex error_mutex; 208 209 /* Do we have to clean up -b/-i from old migrate parameters */ 210 /* This feature is deprecated and will be removed */ 211 bool must_remove_block_options; 212 213 /* 214 * Global switch on whether we need to store the global state 215 * during migration. 216 */ 217 bool store_global_state; 218 219 /* Whether the VM is only allowing for migratable devices */ 220 bool only_migratable; 221 222 /* Whether we send QEMU_VM_CONFIGURATION during migration */ 223 bool send_configuration; 224 /* Whether we send section footer during migration */ 225 bool send_section_footer; 226 227 /* Needed by postcopy-pause state */ 228 QemuSemaphore postcopy_pause_sem; 229 QemuSemaphore postcopy_pause_rp_sem; 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; 237 }; 238 239 void migrate_set_state(int *state, int old_state, int new_state); 240 241 void migration_fd_process_incoming(QEMUFile *f); 242 void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp); 243 void migration_incoming_process(void); 244 245 bool migration_has_all_channels(void); 246 247 uint64_t migrate_max_downtime(void); 248 249 void migrate_set_error(MigrationState *s, const Error *error); 250 void migrate_fd_error(MigrationState *s, const Error *error); 251 252 void migrate_fd_connect(MigrationState *s, Error *error_in); 253 254 bool migration_is_setup_or_active(int state); 255 256 void migrate_init(MigrationState *s); 257 bool migration_is_blocked(Error **errp); 258 /* True if outgoing migration has entered postcopy phase */ 259 bool migration_in_postcopy(void); 260 MigrationState *migrate_get_current(void); 261 262 bool migrate_postcopy(void); 263 264 bool migrate_release_ram(void); 265 bool migrate_postcopy_ram(void); 266 bool migrate_zero_blocks(void); 267 bool migrate_dirty_bitmaps(void); 268 269 bool migrate_auto_converge(void); 270 bool migrate_use_multifd(void); 271 bool migrate_pause_before_switchover(void); 272 int migrate_multifd_channels(void); 273 int migrate_multifd_page_count(void); 274 275 int migrate_use_xbzrle(void); 276 int64_t migrate_xbzrle_cache_size(void); 277 bool migrate_colo_enabled(void); 278 279 bool migrate_use_block(void); 280 bool migrate_use_block_incremental(void); 281 int migrate_max_cpu_throttle(void); 282 bool migrate_use_return_path(void); 283 284 uint64_t ram_get_total_transferred_pages(void); 285 286 bool migrate_use_compression(void); 287 int migrate_compress_level(void); 288 int migrate_compress_threads(void); 289 int migrate_compress_wait_thread(void); 290 int migrate_decompress_threads(void); 291 bool migrate_use_events(void); 292 bool migrate_postcopy_blocktime(void); 293 294 /* Sending on the return path - generic and then for each message type */ 295 void migrate_send_rp_shut(MigrationIncomingState *mis, 296 uint32_t value); 297 void migrate_send_rp_pong(MigrationIncomingState *mis, 298 uint32_t value); 299 int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname, 300 ram_addr_t start, size_t len); 301 void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis, 302 char *block_name); 303 void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value); 304 305 void dirty_bitmap_mig_before_vm_start(void); 306 void init_dirty_bitmap_incoming_migration(void); 307 308 #define qemu_ram_foreach_block \ 309 #warning "Use qemu_ram_foreach_block_migratable in migration code" 310 311 void migration_make_urgent_request(void); 312 void migration_consume_urgent_request(void); 313 314 #endif 315