xref: /openbmc/qemu/migration/migration.h (revision ca693d1c)
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     QemuThread thread;
121     QEMUBH *cleanup_bh;
122     QEMUFile *to_dst_file;
123     /*
124      * Protects to_dst_file pointer.  We need to make sure we won't
125      * yield or hang during the critical section, since this lock will
126      * be used in OOB command handler.
127      */
128     QemuMutex qemu_file_lock;
129 
130     /*
131      * Used to allow urgent requests to override rate limiting.
132      */
133     QemuSemaphore rate_limit_sem;
134 
135     /* pages already send at the beginning of current iteration */
136     uint64_t iteration_initial_pages;
137 
138     /* pages transferred per second */
139     double pages_per_second;
140 
141     /* bytes already send at the beginning of current iteration */
142     uint64_t iteration_initial_bytes;
143     /* time at the start of current iteration */
144     int64_t iteration_start_time;
145     /*
146      * The final stage happens when the remaining data is smaller than
147      * this threshold; it's calculated from the requested downtime and
148      * measured bandwidth
149      */
150     int64_t threshold_size;
151 
152     /* params from 'migrate-set-parameters' */
153     MigrationParameters parameters;
154 
155     int state;
156 
157     /* State related to return path */
158     struct {
159         QEMUFile     *from_dst_file;
160         QemuThread    rp_thread;
161         bool          error;
162         QemuSemaphore rp_sem;
163     } rp_state;
164 
165     double mbps;
166     /* Timestamp when recent migration starts (ms) */
167     int64_t start_time;
168     /* Total time used by latest migration (ms) */
169     int64_t total_time;
170     /* Timestamp when VM is down (ms) to migrate the last stuff */
171     int64_t downtime_start;
172     int64_t downtime;
173     int64_t expected_downtime;
174     bool enabled_capabilities[MIGRATION_CAPABILITY__MAX];
175     int64_t setup_time;
176     /*
177      * Whether guest was running when we enter the completion stage.
178      * If migration is interrupted by any reason, we need to continue
179      * running the guest on source.
180      */
181     bool vm_was_running;
182 
183     /* Flag set once the migration has been asked to enter postcopy */
184     bool start_postcopy;
185     /* Flag set after postcopy has sent the device state */
186     bool postcopy_after_devices;
187 
188     /* Flag set once the migration thread is running (and needs joining) */
189     bool migration_thread_running;
190 
191     /* Flag set once the migration thread called bdrv_inactivate_all */
192     bool block_inactive;
193 
194     /* Migration is paused due to pause-before-switchover */
195     QemuSemaphore pause_sem;
196 
197     /* The semaphore is used to notify COLO thread that failover is finished */
198     QemuSemaphore colo_exit_sem;
199 
200     /* The semaphore is used to notify COLO thread to do checkpoint */
201     QemuSemaphore colo_checkpoint_sem;
202     int64_t colo_checkpoint_time;
203     QEMUTimer *colo_delay_timer;
204 
205     /* The first error that has occurred.
206        We used the mutex to be able to return the 1st error message */
207     Error *error;
208     /* mutex to protect errp */
209     QemuMutex error_mutex;
210 
211     /* Do we have to clean up -b/-i from old migrate parameters */
212     /* This feature is deprecated and will be removed */
213     bool must_remove_block_options;
214 
215     /*
216      * Global switch on whether we need to store the global state
217      * during migration.
218      */
219     bool store_global_state;
220 
221     /* Whether we send QEMU_VM_CONFIGURATION during migration */
222     bool send_configuration;
223     /* Whether we send section footer during migration */
224     bool send_section_footer;
225 
226     /* Needed by postcopy-pause state */
227     QemuSemaphore postcopy_pause_sem;
228     QemuSemaphore postcopy_pause_rp_sem;
229     /*
230      * Whether we abort the migration if decompression errors are
231      * detected at the destination. It is left at false for qemu
232      * older than 3.0, since only newer qemu sends streams that
233      * do not trigger spurious decompression errors.
234      */
235     bool decompress_error_check;
236 };
237 
238 void migrate_set_state(int *state, int old_state, int new_state);
239 
240 void migration_fd_process_incoming(QEMUFile *f);
241 void migration_ioc_process_incoming(QIOChannel *ioc, Error **errp);
242 void migration_incoming_process(void);
243 
244 bool  migration_has_all_channels(void);
245 
246 uint64_t migrate_max_downtime(void);
247 
248 void migrate_set_error(MigrationState *s, const Error *error);
249 void migrate_fd_error(MigrationState *s, const Error *error);
250 
251 void migrate_fd_connect(MigrationState *s, Error *error_in);
252 
253 bool migration_is_setup_or_active(int state);
254 
255 void migrate_init(MigrationState *s);
256 bool migration_is_blocked(Error **errp);
257 /* True if outgoing migration has entered postcopy phase */
258 bool migration_in_postcopy(void);
259 MigrationState *migrate_get_current(void);
260 
261 bool migrate_postcopy(void);
262 
263 bool migrate_release_ram(void);
264 bool migrate_postcopy_ram(void);
265 bool migrate_zero_blocks(void);
266 bool migrate_dirty_bitmaps(void);
267 bool migrate_ignore_shared(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 
274 int migrate_use_xbzrle(void);
275 int64_t migrate_xbzrle_cache_size(void);
276 bool migrate_colo_enabled(void);
277 
278 bool migrate_use_block(void);
279 bool migrate_use_block_incremental(void);
280 int migrate_max_cpu_throttle(void);
281 bool migrate_use_return_path(void);
282 
283 uint64_t ram_get_total_transferred_pages(void);
284 
285 bool migrate_use_compression(void);
286 int migrate_compress_level(void);
287 int migrate_compress_threads(void);
288 int migrate_compress_wait_thread(void);
289 int migrate_decompress_threads(void);
290 bool migrate_use_events(void);
291 bool migrate_postcopy_blocktime(void);
292 
293 /* Sending on the return path - generic and then for each message type */
294 void migrate_send_rp_shut(MigrationIncomingState *mis,
295                           uint32_t value);
296 void migrate_send_rp_pong(MigrationIncomingState *mis,
297                           uint32_t value);
298 int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname,
299                               ram_addr_t start, size_t len);
300 void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis,
301                                  char *block_name);
302 void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value);
303 
304 void dirty_bitmap_mig_before_vm_start(void);
305 void init_dirty_bitmap_incoming_migration(void);
306 void migrate_add_address(SocketAddress *address);
307 
308 int foreach_not_ignored_block(RAMBlockIterFunc func, void *opaque);
309 
310 #define qemu_ram_foreach_block \
311   #warning "Use foreach_not_ignored_block in migration code"
312 
313 void migration_make_urgent_request(void);
314 void migration_consume_urgent_request(void);
315 
316 #endif
317