]> git.proxmox.com Git - mirror_qemu.git/blob - include/migration/migration.h
COLO: Shutdown related socket fd while do failover
[mirror_qemu.git] / include / migration / migration.h
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/qmp/qdict.h"
18 #include "qemu-common.h"
19 #include "qemu/thread.h"
20 #include "qemu/notify.h"
21 #include "migration/vmstate.h"
22 #include "qapi-types.h"
23 #include "exec/cpu-common.h"
24 #include "qemu/coroutine_int.h"
25
26 #define QEMU_VM_FILE_MAGIC 0x5145564d
27 #define QEMU_VM_FILE_VERSION_COMPAT 0x00000002
28 #define QEMU_VM_FILE_VERSION 0x00000003
29
30 #define QEMU_VM_EOF 0x00
31 #define QEMU_VM_SECTION_START 0x01
32 #define QEMU_VM_SECTION_PART 0x02
33 #define QEMU_VM_SECTION_END 0x03
34 #define QEMU_VM_SECTION_FULL 0x04
35 #define QEMU_VM_SUBSECTION 0x05
36 #define QEMU_VM_VMDESCRIPTION 0x06
37 #define QEMU_VM_CONFIGURATION 0x07
38 #define QEMU_VM_COMMAND 0x08
39 #define QEMU_VM_SECTION_FOOTER 0x7e
40
41 /* for vl.c */
42 extern int only_migratable;
43
44 struct MigrationParams {
45 bool blk;
46 bool shared;
47 };
48
49 /* Messages sent on the return path from destination to source */
50 enum mig_rp_message_type {
51 MIG_RP_MSG_INVALID = 0, /* Must be 0 */
52 MIG_RP_MSG_SHUT, /* sibling will not send any more RP messages */
53 MIG_RP_MSG_PONG, /* Response to a PING; data (seq: be32 ) */
54
55 MIG_RP_MSG_REQ_PAGES_ID, /* data (start: be64, len: be32, id: string) */
56 MIG_RP_MSG_REQ_PAGES, /* data (start: be64, len: be32) */
57
58 MIG_RP_MSG_MAX
59 };
60
61 typedef QLIST_HEAD(, LoadStateEntry) LoadStateEntry_Head;
62
63 /* The current postcopy state is read/set by postcopy_state_get/set
64 * which update it atomically.
65 * The state is updated as postcopy messages are received, and
66 * in general only one thread should be writing to the state at any one
67 * time, initially the main thread and then the listen thread;
68 * Corner cases are where either thread finishes early and/or errors.
69 * The state is checked as messages are received to ensure that
70 * the source is sending us messages in the correct order.
71 * The state is also used by the RAM reception code to know if it
72 * has to place pages atomically, and the cleanup code at the end of
73 * the main thread to know if it has to delay cleanup until the end
74 * of postcopy.
75 */
76 typedef enum {
77 POSTCOPY_INCOMING_NONE = 0, /* Initial state - no postcopy */
78 POSTCOPY_INCOMING_ADVISE,
79 POSTCOPY_INCOMING_DISCARD,
80 POSTCOPY_INCOMING_LISTENING,
81 POSTCOPY_INCOMING_RUNNING,
82 POSTCOPY_INCOMING_END
83 } PostcopyState;
84
85 /* State for the incoming migration */
86 struct MigrationIncomingState {
87 QEMUFile *from_src_file;
88
89 /*
90 * Free at the start of the main state load, set as the main thread finishes
91 * loading state.
92 */
93 QemuEvent main_thread_load_event;
94
95 bool have_fault_thread;
96 QemuThread fault_thread;
97 QemuSemaphore fault_thread_sem;
98
99 bool have_listen_thread;
100 QemuThread listen_thread;
101 QemuSemaphore listen_thread_sem;
102
103 /* For the kernel to send us notifications */
104 int userfault_fd;
105 /* To tell the fault_thread to quit */
106 int userfault_quit_fd;
107 QEMUFile *to_src_file;
108 QemuMutex rp_mutex; /* We send replies from multiple threads */
109 void *postcopy_tmp_page;
110
111 QEMUBH *bh;
112
113 int state;
114
115 bool have_colo_incoming_thread;
116 QemuThread colo_incoming_thread;
117 /* The coroutine we should enter (back) after failover */
118 Coroutine *migration_incoming_co;
119 QemuSemaphore colo_incoming_sem;
120
121 /* See savevm.c */
122 LoadStateEntry_Head loadvm_handlers;
123 };
124
125 MigrationIncomingState *migration_incoming_get_current(void);
126 void migration_incoming_state_destroy(void);
127
128 /*
129 * An outstanding page request, on the source, having been received
130 * and queued
131 */
132 struct MigrationSrcPageRequest {
133 RAMBlock *rb;
134 hwaddr offset;
135 hwaddr len;
136
137 QSIMPLEQ_ENTRY(MigrationSrcPageRequest) next_req;
138 };
139
140 struct MigrationState
141 {
142 size_t bytes_xfer;
143 size_t xfer_limit;
144 QemuThread thread;
145 QEMUBH *cleanup_bh;
146 QEMUFile *to_dst_file;
147
148 /* New style params from 'migrate-set-parameters' */
149 MigrationParameters parameters;
150
151 int state;
152 /* Old style params from 'migrate' command */
153 MigrationParams params;
154
155 /* State related to return path */
156 struct {
157 QEMUFile *from_dst_file;
158 QemuThread rp_thread;
159 bool error;
160 } rp_state;
161
162 double mbps;
163 int64_t total_time;
164 int64_t downtime;
165 int64_t expected_downtime;
166 int64_t dirty_pages_rate;
167 int64_t dirty_bytes_rate;
168 bool enabled_capabilities[MIGRATION_CAPABILITY__MAX];
169 int64_t xbzrle_cache_size;
170 int64_t setup_time;
171 int64_t dirty_sync_count;
172 /* Count of requests incoming from destination */
173 int64_t postcopy_requests;
174
175 /* Flag set once the migration has been asked to enter postcopy */
176 bool start_postcopy;
177 /* Flag set after postcopy has sent the device state */
178 bool postcopy_after_devices;
179
180 /* Flag set once the migration thread is running (and needs joining) */
181 bool migration_thread_running;
182
183 /* Flag set once the migration thread called bdrv_inactivate_all */
184 bool block_inactive;
185
186 /* Queue of outstanding page requests from the destination */
187 QemuMutex src_page_req_mutex;
188 QSIMPLEQ_HEAD(src_page_requests, MigrationSrcPageRequest) src_page_requests;
189 /* The RAMBlock used in the last src_page_request */
190 RAMBlock *last_req_rb;
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 last error that occurred */
200 Error *error;
201 };
202
203 void migrate_set_state(int *state, int old_state, int new_state);
204
205 void migration_fd_process_incoming(QEMUFile *f);
206
207 void qemu_start_incoming_migration(const char *uri, Error **errp);
208
209 void migration_channel_process_incoming(MigrationState *s,
210 QIOChannel *ioc);
211
212 void migration_tls_channel_process_incoming(MigrationState *s,
213 QIOChannel *ioc,
214 Error **errp);
215
216 void migration_channel_connect(MigrationState *s,
217 QIOChannel *ioc,
218 const char *hostname);
219
220 void migration_tls_channel_connect(MigrationState *s,
221 QIOChannel *ioc,
222 const char *hostname,
223 Error **errp);
224
225 uint64_t migrate_max_downtime(void);
226
227 void exec_start_incoming_migration(const char *host_port, Error **errp);
228
229 void exec_start_outgoing_migration(MigrationState *s, const char *host_port, Error **errp);
230
231 void tcp_start_incoming_migration(const char *host_port, Error **errp);
232
233 void tcp_start_outgoing_migration(MigrationState *s, const char *host_port, Error **errp);
234
235 void unix_start_incoming_migration(const char *path, Error **errp);
236
237 void unix_start_outgoing_migration(MigrationState *s, const char *path, Error **errp);
238
239 void fd_start_incoming_migration(const char *path, Error **errp);
240
241 void fd_start_outgoing_migration(MigrationState *s, const char *fdname, Error **errp);
242
243 void rdma_start_outgoing_migration(void *opaque, const char *host_port, Error **errp);
244
245 void rdma_start_incoming_migration(const char *host_port, Error **errp);
246
247 void migrate_fd_error(MigrationState *s, const Error *error);
248
249 void migrate_fd_connect(MigrationState *s);
250
251 void add_migration_state_change_notifier(Notifier *notify);
252 void remove_migration_state_change_notifier(Notifier *notify);
253 MigrationState *migrate_init(const MigrationParams *params);
254 bool migration_is_blocked(Error **errp);
255 bool migration_in_setup(MigrationState *);
256 bool migration_is_idle(MigrationState *s);
257 bool migration_has_finished(MigrationState *);
258 bool migration_has_failed(MigrationState *);
259 /* True if outgoing migration has entered postcopy phase */
260 bool migration_in_postcopy(MigrationState *);
261 /* ...and after the device transmission */
262 bool migration_in_postcopy_after_devices(MigrationState *);
263 MigrationState *migrate_get_current(void);
264
265 void migrate_compress_threads_create(void);
266 void migrate_compress_threads_join(void);
267 void migrate_decompress_threads_create(void);
268 void migrate_decompress_threads_join(void);
269 uint64_t ram_bytes_remaining(void);
270 uint64_t ram_bytes_transferred(void);
271 uint64_t ram_bytes_total(void);
272 void free_xbzrle_decoded_buf(void);
273
274 void acct_update_position(QEMUFile *f, size_t size, bool zero);
275
276 uint64_t dup_mig_bytes_transferred(void);
277 uint64_t dup_mig_pages_transferred(void);
278 uint64_t skipped_mig_bytes_transferred(void);
279 uint64_t skipped_mig_pages_transferred(void);
280 uint64_t norm_mig_bytes_transferred(void);
281 uint64_t norm_mig_pages_transferred(void);
282 uint64_t xbzrle_mig_bytes_transferred(void);
283 uint64_t xbzrle_mig_pages_transferred(void);
284 uint64_t xbzrle_mig_pages_overflow(void);
285 uint64_t xbzrle_mig_pages_cache_miss(void);
286 double xbzrle_mig_cache_miss_rate(void);
287
288 void ram_handle_compressed(void *host, uint8_t ch, uint64_t size);
289 void ram_debug_dump_bitmap(unsigned long *todump, bool expected);
290 /* For outgoing discard bitmap */
291 int ram_postcopy_send_discard_bitmap(MigrationState *ms);
292 /* For incoming postcopy discard */
293 int ram_discard_range(MigrationIncomingState *mis, const char *block_name,
294 uint64_t start, size_t length);
295 int ram_postcopy_incoming_init(MigrationIncomingState *mis);
296 void ram_postcopy_migrated_memory_release(MigrationState *ms);
297
298 /**
299 * @migrate_add_blocker - prevent migration from proceeding
300 *
301 * @reason - an error to be returned whenever migration is attempted
302 *
303 * @errp - [out] The reason (if any) we cannot block migration right now.
304 *
305 * @returns - 0 on success, -EBUSY/-EACCES on failure, with errp set.
306 */
307 int migrate_add_blocker(Error *reason, Error **errp);
308
309 /**
310 * @migrate_del_blocker - remove a blocking error from migration
311 *
312 * @reason - the error blocking migration
313 */
314 void migrate_del_blocker(Error *reason);
315
316 bool migrate_release_ram(void);
317 bool migrate_postcopy_ram(void);
318 bool migrate_zero_blocks(void);
319
320 bool migrate_auto_converge(void);
321
322 int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
323 uint8_t *dst, int dlen);
324 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen);
325
326 int migrate_use_xbzrle(void);
327 int64_t migrate_xbzrle_cache_size(void);
328 bool migrate_colo_enabled(void);
329
330 int64_t xbzrle_cache_resize(int64_t new_size);
331
332 bool migrate_use_compression(void);
333 int migrate_compress_level(void);
334 int migrate_compress_threads(void);
335 int migrate_decompress_threads(void);
336 bool migrate_use_events(void);
337
338 /* Sending on the return path - generic and then for each message type */
339 void migrate_send_rp_message(MigrationIncomingState *mis,
340 enum mig_rp_message_type message_type,
341 uint16_t len, void *data);
342 void migrate_send_rp_shut(MigrationIncomingState *mis,
343 uint32_t value);
344 void migrate_send_rp_pong(MigrationIncomingState *mis,
345 uint32_t value);
346 void migrate_send_rp_req_pages(MigrationIncomingState *mis, const char* rbname,
347 ram_addr_t start, size_t len);
348
349 void ram_control_before_iterate(QEMUFile *f, uint64_t flags);
350 void ram_control_after_iterate(QEMUFile *f, uint64_t flags);
351 void ram_control_load_hook(QEMUFile *f, uint64_t flags, void *data);
352
353 /* Whenever this is found in the data stream, the flags
354 * will be passed to ram_control_load_hook in the incoming-migration
355 * side. This lets before_ram_iterate/after_ram_iterate add
356 * transport-specific sections to the RAM migration data.
357 */
358 #define RAM_SAVE_FLAG_HOOK 0x80
359
360 #define RAM_SAVE_CONTROL_NOT_SUPP -1000
361 #define RAM_SAVE_CONTROL_DELAYED -2000
362
363 size_t ram_control_save_page(QEMUFile *f, ram_addr_t block_offset,
364 ram_addr_t offset, size_t size,
365 uint64_t *bytes_sent);
366
367 void ram_mig_init(void);
368 void savevm_skip_section_footers(void);
369 void register_global_state(void);
370 void global_state_set_optional(void);
371 void savevm_skip_configuration(void);
372 int global_state_store(void);
373 void global_state_store_running(void);
374
375 void flush_page_queue(MigrationState *ms);
376 int ram_save_queue_pages(MigrationState *ms, const char *rbname,
377 ram_addr_t start, ram_addr_t len);
378
379 PostcopyState postcopy_state_get(void);
380 /* Set the state and return the old state */
381 PostcopyState postcopy_state_set(PostcopyState new_state);
382 #endif