xref: /openbmc/qemu/migration/migration.h (revision fe7f9b8e)
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     /* bytes already send at the beggining of current interation */
125     uint64_t iteration_initial_bytes;
126     /* time at the start of current iteration */
127     int64_t iteration_start_time;
128     /*
129      * The final stage happens when the remaining data is smaller than
130      * this threshold; it's calculated from the requested downtime and
131      * measured bandwidth
132      */
133     int64_t threshold_size;
134 
135     /* params from 'migrate-set-parameters' */
136     MigrationParameters parameters;
137 
138     int state;
139 
140     /* State related to return path */
141     struct {
142         QEMUFile     *from_dst_file;
143         QemuThread    rp_thread;
144         bool          error;
145         QemuSemaphore rp_sem;
146     } rp_state;
147 
148     double mbps;
149     /* Timestamp when recent migration starts (ms) */
150     int64_t start_time;
151     /* Total time used by latest migration (ms) */
152     int64_t total_time;
153     /* Timestamp when VM is down (ms) to migrate the last stuff */
154     int64_t downtime_start;
155     int64_t downtime;
156     int64_t expected_downtime;
157     bool enabled_capabilities[MIGRATION_CAPABILITY__MAX];
158     int64_t setup_time;
159     /*
160      * Whether guest was running when we enter the completion stage.
161      * If migration is interrupted by any reason, we need to continue
162      * running the guest on source.
163      */
164     bool vm_was_running;
165 
166     /* Flag set once the migration has been asked to enter postcopy */
167     bool start_postcopy;
168     /* Flag set after postcopy has sent the device state */
169     bool postcopy_after_devices;
170 
171     /* Flag set once the migration thread is running (and needs joining) */
172     bool migration_thread_running;
173 
174     /* Flag set once the migration thread called bdrv_inactivate_all */
175     bool block_inactive;
176 
177     /* Migration is paused due to pause-before-switchover */
178     QemuSemaphore pause_sem;
179 
180     /* The semaphore is used to notify COLO thread that failover is finished */
181     QemuSemaphore colo_exit_sem;
182 
183     /* The semaphore is used to notify COLO thread to do checkpoint */
184     QemuSemaphore colo_checkpoint_sem;
185     int64_t colo_checkpoint_time;
186     QEMUTimer *colo_delay_timer;
187 
188     /* The first error that has occurred.
189        We used the mutex to be able to return the 1st error message */
190     Error *error;
191     /* mutex to protect errp */
192     QemuMutex error_mutex;
193 
194     /* Do we have to clean up -b/-i from old migrate parameters */
195     /* This feature is deprecated and will be removed */
196     bool must_remove_block_options;
197 
198     /*
199      * Global switch on whether we need to store the global state
200      * during migration.
201      */
202     bool store_global_state;
203 
204     /* Whether the VM is only allowing for migratable devices */
205     bool only_migratable;
206 
207     /* Whether we send QEMU_VM_CONFIGURATION during migration */
208     bool send_configuration;
209     /* Whether we send section footer during migration */
210     bool send_section_footer;
211 
212     /* Needed by postcopy-pause state */
213     QemuSemaphore postcopy_pause_sem;
214     QemuSemaphore postcopy_pause_rp_sem;
215 };
216 
217 void migrate_set_state(int *state, int old_state, int new_state);
218 
219 void migration_fd_process_incoming(QEMUFile *f);
220 void migration_ioc_process_incoming(QIOChannel *ioc);
221 void migration_incoming_process(void);
222 
223 bool  migration_has_all_channels(void);
224 
225 uint64_t migrate_max_downtime(void);
226 
227 void migrate_set_error(MigrationState *s, const Error *error);
228 void migrate_fd_error(MigrationState *s, const Error *error);
229 
230 void migrate_fd_connect(MigrationState *s, Error *error_in);
231 
232 void migrate_init(MigrationState *s);
233 bool migration_is_blocked(Error **errp);
234 /* True if outgoing migration has entered postcopy phase */
235 bool migration_in_postcopy(void);
236 MigrationState *migrate_get_current(void);
237 
238 bool migrate_postcopy(void);
239 
240 bool migrate_release_ram(void);
241 bool migrate_postcopy_ram(void);
242 bool migrate_zero_blocks(void);
243 bool migrate_dirty_bitmaps(void);
244 
245 bool migrate_auto_converge(void);
246 bool migrate_use_multifd(void);
247 bool migrate_pause_before_switchover(void);
248 int migrate_multifd_channels(void);
249 int migrate_multifd_page_count(void);
250 
251 int migrate_use_xbzrle(void);
252 int64_t migrate_xbzrle_cache_size(void);
253 bool migrate_colo_enabled(void);
254 
255 bool migrate_use_block(void);
256 bool migrate_use_block_incremental(void);
257 bool migrate_use_return_path(void);
258 
259 bool migrate_use_compression(void);
260 int migrate_compress_level(void);
261 int migrate_compress_threads(void);
262 int migrate_decompress_threads(void);
263 bool migrate_use_events(void);
264 bool migrate_postcopy_blocktime(void);
265 
266 /* Sending on the return path - generic and then for each message type */
267 void migrate_send_rp_shut(MigrationIncomingState *mis,
268                           uint32_t value);
269 void migrate_send_rp_pong(MigrationIncomingState *mis,
270                           uint32_t value);
271 int migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname,
272                               ram_addr_t start, size_t len);
273 void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis,
274                                  char *block_name);
275 void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value);
276 
277 void dirty_bitmap_mig_before_vm_start(void);
278 void init_dirty_bitmap_incoming_migration(void);
279 
280 #endif
281