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