5 QEMU has code to load/save the state of the guest that it is running.
6 These are two complementary operations. Saving the state just does
7 that, saves the state for each device that the guest is running.
8 Restoring a guest is just the opposite operation: we need to load the
11 For this to work, QEMU has to be launched with the same arguments the
12 two times. I.e. it can only restore the state in one guest that has
13 the same devices that the one it was saved (this last requirement can
14 be relaxed a bit, but for now we can consider that configuration has
15 to be exactly the same).
17 Once that we are able to save/restore a guest, a new functionality is
18 requested: migration. This means that QEMU is able to start in one
19 machine and being "migrated" to another machine. I.e. being moved to
22 Next was the "live migration" functionality. This is important
23 because some guests run with a lot of state (specially RAM), and it
24 can take a while to move all state from one machine to another. Live
25 migration allows the guest to continue running while the state is
26 transferred. Only while the last part of the state is transferred has
27 the guest to be stopped. Typically the time that the guest is
28 unresponsive during live migration is the low hundred of milliseconds
29 (notice that this depends on a lot of things).
34 The migration stream is normally just a byte stream that can be passed
37 - tcp migration: do the migration using tcp sockets
38 - unix migration: do the migration using unix sockets
39 - exec migration: do the migration using the stdin/stdout through a process.
40 - fd migration: do the migration using a file descriptor that is
41 passed to QEMU. QEMU doesn't care how this file descriptor is opened.
43 In addition, support is included for migration using RDMA, which
44 transports the page data using ``RDMA``, where the hardware takes care of
45 transporting the pages, and the load on the CPU is much lower. While the
46 internals of RDMA migration are a bit different, this isn't really visible
47 outside the RAM migration code.
49 All these migration protocols use the same infrastructure to
50 save/restore state devices. This infrastructure is shared with the
51 savevm/loadvm functionality.
56 The files, sockets or fd's that carry the migration stream are abstracted by
57 the ``QEMUFile`` type (see `migration/qemu-file.h`). In most cases this
58 is connected to a subtype of ``QIOChannel`` (see `io/`).
61 Saving the state of one device
62 ==============================
64 For most devices, the state is saved in a single call to the migration
65 infrastructure; these are *non-iterative* devices. The data for these
66 devices is sent at the end of precopy migration, when the CPUs are paused.
67 There are also *iterative* devices, which contain a very large amount of
68 data (e.g. RAM or large tables). See the iterative device section below.
70 General advice for device developers
71 ------------------------------------
73 - The migration state saved should reflect the device being modelled rather
74 than the way your implementation works. That way if you change the implementation
75 later the migration stream will stay compatible. That model may include
76 internal state that's not directly visible in a register.
78 - When saving a migration stream the device code may walk and check
79 the state of the device. These checks might fail in various ways (e.g.
80 discovering internal state is corrupt or that the guest has done something bad).
81 Consider carefully before asserting/aborting at this point, since the
82 normal response from users is that *migration broke their VM* since it had
83 apparently been running fine until then. In these error cases, the device
84 should log a message indicating the cause of error, and should consider
85 putting the device into an error state, allowing the rest of the VM to
88 - The migration might happen at an inconvenient point,
89 e.g. right in the middle of the guest reprogramming the device, during
90 guest reboot or shutdown or while the device is waiting for external IO.
91 It's strongly preferred that migrations do not fail in this situation,
92 since in the cloud environment migrations might happen automatically to
93 VMs that the administrator doesn't directly control.
95 - If you do need to fail a migration, ensure that sufficient information
96 is logged to identify what went wrong.
98 - The destination should treat an incoming migration stream as hostile
99 (which we do to varying degrees in the existing code). Check that offsets
100 into buffers and the like can't cause overruns. Fail the incoming migration
101 in the case of a corrupted stream like this.
103 - Take care with internal device state or behaviour that might become
104 migration version dependent. For example, the order of PCI capabilities
105 is required to stay constant across migration. Another example would
106 be that a special case handled by subsections (see below) might become
107 much more common if a default behaviour is changed.
109 - The state of the source should not be changed or destroyed by the
110 outgoing migration. Migrations timing out or being failed by
111 higher levels of management, or failures of the destination host are
112 not unusual, and in that case the VM is restarted on the source.
113 Note that the management layer can validly revert the migration
114 even though the QEMU level of migration has succeeded as long as it
115 does it before starting execution on the destination.
117 - Buses and devices should be able to explicitly specify addresses when
118 instantiated, and management tools should use those. For example,
119 when hot adding USB devices it's important to specify the ports
120 and addresses, since implicit ordering based on the command line order
121 may be different on the destination. This can result in the
122 device state being loaded into the wrong device.
127 Most device data can be described using the ``VMSTATE`` macros (mostly defined
128 in ``include/migration/vmstate.h``).
130 An example (from hw/input/pckbd.c)
134 static const VMStateDescription vmstate_kbd = {
137 .minimum_version_id = 3,
138 .fields = (VMStateField[]) {
139 VMSTATE_UINT8(write_cmd, KBDState),
140 VMSTATE_UINT8(status, KBDState),
141 VMSTATE_UINT8(mode, KBDState),
142 VMSTATE_UINT8(pending, KBDState),
143 VMSTATE_END_OF_LIST()
147 We are declaring the state with name "pckbd".
148 The `version_id` is 3, and the fields are 4 uint8_t in a KBDState structure.
149 We registered this with:
153 vmstate_register(NULL, 0, &vmstate_kbd, s);
155 For devices that are `qdev` based, we can register the device in the class
160 dc->vmsd = &vmstate_kbd_isa;
162 The VMState macros take care of ensuring that the device data section
163 is formatted portably (normally big endian) and make some compile time checks
164 against the types of the fields in the structures.
166 VMState macros can include other VMStateDescriptions to store substructures
167 (see ``VMSTATE_STRUCT_``), arrays (``VMSTATE_ARRAY_``) and variable length
168 arrays (``VMSTATE_VARRAY_``). Various other macros exist for special
171 Note that the format on the wire is still very raw; i.e. a VMSTATE_UINT32
172 ends up with a 4 byte bigendian representation on the wire; in the future
173 it might be possible to use a more structured format.
178 This way is going to disappear as soon as all current users are ported to VMSTATE;
179 although converting existing code can be tricky, and thus 'soon' is relative.
181 Each device has to register two functions, one to save the state and
182 another to load the state back.
186 int register_savevm_live(DeviceState *dev,
193 Two functions in the ``ops`` structure are the `save_state`
194 and `load_state` functions. Notice that `load_state` receives a version_id
195 parameter to know what state format is receiving. `save_state` doesn't
196 have a version_id parameter because it always uses the latest version.
198 Note that because the VMState macros still save the data in a raw
199 format, in many cases it's possible to replace legacy code
200 with a carefully constructed VMState description that matches the
201 byte layout of the existing code.
203 Changing migration data structures
204 ----------------------------------
206 When we migrate a device, we save/load the state as a series
207 of fields. Sometimes, due to bugs or new functionality, we need to
208 change the state to store more/different information. Changing the migration
209 state saved for a device can break migration compatibility unless
210 care is taken to use the appropriate techniques. In general QEMU tries
211 to maintain forward migration compatibility (i.e. migrating from
212 QEMU n->n+1) and there are users who benefit from backward compatibility
218 The most common structure change is adding new data, e.g. when adding
219 a newer form of device, or adding that state that you previously
220 forgot to migrate. This is best solved using a subsection.
222 A subsection is "like" a device vmstate, but with a particularity, it
223 has a Boolean function that tells if that values are needed to be sent
224 or not. If this functions returns false, the subsection is not sent.
225 Subsections have a unique name, that is looked for on the receiving
228 On the receiving side, if we found a subsection for a device that we
229 don't understand, we just fail the migration. If we understand all
230 the subsections, then we load the state with success. There's no check
231 that a subsection is loaded, so a newer QEMU that knows about a subsection
232 can (with care) load a stream from an older QEMU that didn't send
235 If the new data is only needed in a rare case, then the subsection
236 can be made conditional on that case and the migration will still
237 succeed to older QEMUs in most cases. This is OK for data that's
238 critical, but in some use cases it's preferred that the migration
239 should succeed even with the data missing. To support this the
240 subsection can be connected to a device property and from there
241 to a versioned machine type.
243 The 'pre_load' and 'post_load' functions on subsections are only
244 called if the subsection is loaded.
246 One important note is that the outer post_load() function is called "after"
247 loading all subsections, because a newer subsection could change the same
248 value that it uses. A flag, and the combination of outer pre_load and
249 post_load can be used to detect whether a subsection was loaded, and to
250 fall back on default behaviour when the subsection isn't present.
256 static bool ide_drive_pio_state_needed(void *opaque)
258 IDEState *s = opaque;
260 return ((s->status & DRQ_STAT) != 0)
261 || (s->bus->error_status & BM_STATUS_PIO_RETRY);
264 const VMStateDescription vmstate_ide_drive_pio_state = {
265 .name = "ide_drive/pio_state",
267 .minimum_version_id = 1,
268 .pre_save = ide_drive_pio_pre_save,
269 .post_load = ide_drive_pio_post_load,
270 .needed = ide_drive_pio_state_needed,
271 .fields = (VMStateField[]) {
272 VMSTATE_INT32(req_nb_sectors, IDEState),
273 VMSTATE_VARRAY_INT32(io_buffer, IDEState, io_buffer_total_len, 1,
274 vmstate_info_uint8, uint8_t),
275 VMSTATE_INT32(cur_io_buffer_offset, IDEState),
276 VMSTATE_INT32(cur_io_buffer_len, IDEState),
277 VMSTATE_UINT8(end_transfer_fn_idx, IDEState),
278 VMSTATE_INT32(elementary_transfer_size, IDEState),
279 VMSTATE_INT32(packet_transfer_size, IDEState),
280 VMSTATE_END_OF_LIST()
284 const VMStateDescription vmstate_ide_drive = {
287 .minimum_version_id = 0,
288 .post_load = ide_drive_post_load,
289 .fields = (VMStateField[]) {
290 .... several fields ....
291 VMSTATE_END_OF_LIST()
293 .subsections = (const VMStateDescription*[]) {
294 &vmstate_ide_drive_pio_state,
299 Here we have a subsection for the pio state. We only need to
300 save/send this state when we are in the middle of a pio operation
301 (that is what ``ide_drive_pio_state_needed()`` checks). If DRQ_STAT is
302 not enabled, the values on that fields are garbage and don't need to
305 Connecting subsections to properties
306 ------------------------------------
308 Using a condition function that checks a 'property' to determine whether
309 to send a subsection allows backward migration compatibility when
310 new subsections are added, especially when combined with versioned
315 a) Add a new property using ``DEFINE_PROP_BOOL`` - e.g. support-foo and
317 b) Add an entry to the ``HW_COMPAT_`` for the previous version that sets
318 the property to false.
319 c) Add a static bool support_foo function that tests the property.
320 d) Add a subsection with a .needed set to the support_foo function
321 e) (potentially) Add an outer pre_load that sets up a default value
322 for 'foo' to be used if the subsection isn't loaded.
324 Now that subsection will not be generated when using an older
325 machine type and the migration stream will be accepted by older
328 Not sending existing elements
329 -----------------------------
331 Sometimes members of the VMState are no longer needed:
333 - removing them will break migration compatibility
335 - making them version dependent and bumping the version will break backward migration
338 Adding a dummy field into the migration stream is normally the best way to preserve
341 If the field really does need to be removed then:
343 a) Add a new property/compatibility/function in the same way for subsections above.
344 b) replace the VMSTATE macro with the _TEST version of the macro, e.g.:
346 ``VMSTATE_UINT32(foo, barstruct)``
350 ``VMSTATE_UINT32_TEST(foo, barstruct, pre_version_baz)``
352 Sometime in the future when we no longer care about the ancient versions these can be killed off.
353 Note that for backward compatibility it's important to fill in the structure with
354 data that the destination will understand.
356 Any difference in the predicates on the source and destination will end up
357 with different fields being enabled and data being loaded into the wrong
358 fields; for this reason conditional fields like this are very fragile.
363 Version numbers are intended for major incompatible changes to the
364 migration of a device, and using them breaks backward-migration
365 compatibility; in general most changes can be made by adding Subsections
366 (see above) or _TEST macros (see above) which won't break compatibility.
368 Each version is associated with a series of fields saved. The `save_state` always saves
369 the state as the newer version. But `load_state` sometimes is able to
370 load state from an older version.
372 You can see that there are several version fields:
374 - `version_id`: the maximum version_id supported by VMState for that device.
375 - `minimum_version_id`: the minimum version_id that VMState is able to understand
377 - `minimum_version_id_old`: For devices that were not able to port to vmstate, we can
378 assign a function that knows how to read this old state. This field is
379 ignored if there is no `load_state_old` handler.
381 VMState is able to read versions from minimum_version_id to
382 version_id. And the function ``load_state_old()`` (if present) is able to
383 load state from minimum_version_id_old to minimum_version_id. This
384 function is deprecated and will be removed when no more users are left.
386 There are *_V* forms of many ``VMSTATE_`` macros to load fields for version dependent fields,
391 VMSTATE_UINT16_V(ip_id, Slirp, 2),
393 only loads that field for versions 2 and newer.
395 Saving state will always create a section with the 'version_id' value
396 and thus can't be loaded by any older QEMU.
401 Sometimes, it is not enough to be able to save the state directly
402 from one structure, we need to fill the correct values there. One
403 example is when we are using kvm. Before saving the cpu state, we
404 need to ask kvm to copy to QEMU the state that it is using. And the
405 opposite when we are loading the state, we need a way to tell kvm to
406 load the state for the cpu that we have just loaded from the QEMUFile.
408 The functions to do that are inside a vmstate definition, and are called:
410 - ``int (*pre_load)(void *opaque);``
412 This function is called before we load the state of one device.
414 - ``int (*post_load)(void *opaque, int version_id);``
416 This function is called after we load the state of one device.
418 - ``int (*pre_save)(void *opaque);``
420 This function is called before we save the state of one device.
422 Example: You can look at hpet.c, that uses the three function to
423 massage the state that is transferred.
425 The ``VMSTATE_WITH_TMP`` macro may be useful when the migration
426 data doesn't match the stored device data well; it allows an
427 intermediate temporary structure to be populated with migration
428 data and then transferred to the main structure.
430 If you use memory API functions that update memory layout outside
431 initialization (i.e., in response to a guest action), this is a strong
432 indication that you need to call these functions in a `post_load` callback.
433 Examples of such memory API functions are:
435 - memory_region_add_subregion()
436 - memory_region_del_subregion()
437 - memory_region_set_readonly()
438 - memory_region_set_nonvolatile()
439 - memory_region_set_enabled()
440 - memory_region_set_address()
441 - memory_region_set_alias_offset()
443 Iterative device migration
444 --------------------------
446 Some devices, such as RAM, Block storage or certain platform devices,
447 have large amounts of data that would mean that the CPUs would be
448 paused for too long if they were sent in one section. For these
449 devices an *iterative* approach is taken.
451 The iterative devices generally don't use VMState macros
452 (although it may be possible in some cases) and instead use
453 qemu_put_*/qemu_get_* macros to read/write data to the stream. Specialist
454 versions exist for high bandwidth IO.
457 An iterative device must provide:
459 - A ``save_setup`` function that initialises the data structures and
460 transmits a first section containing information on the device. In the
461 case of RAM this transmits a list of RAMBlocks and sizes.
463 - A ``load_setup`` function that initialises the data structures on the
466 - A ``save_live_pending`` function that is called repeatedly and must
467 indicate how much more data the iterative data must save. The core
468 migration code will use this to determine when to pause the CPUs
469 and complete the migration.
471 - A ``save_live_iterate`` function (called after ``save_live_pending``
472 when there is significant data still to be sent). It should send
473 a chunk of data until the point that stream bandwidth limits tell it
474 to stop. Each call generates one section.
476 - A ``save_live_complete_precopy`` function that must transmit the
477 last section for the device containing any remaining data.
479 - A ``load_state`` function used to load sections generated by
480 any of the save functions that generate sections.
482 - ``cleanup`` functions for both save and load that are called
483 at the end of migration.
485 Note that the contents of the sections for iterative migration tend
486 to be open-coded by the devices; care should be taken in parsing
487 the results and structuring the stream to make them easy to validate.
492 There are cases in which the ordering of device loading matters; for
493 example in some systems where a device may assert an interrupt during loading,
494 if the interrupt controller is loaded later then it might lose the state.
496 Some ordering is implicitly provided by the order in which the machine
497 definition creates devices, however this is somewhat fragile.
499 The ``MigrationPriority`` enum provides a means of explicitly enforcing
500 ordering. Numerically higher priorities are loaded earlier.
501 The priority is set by setting the ``priority`` field of the top level
502 ``VMStateDescription`` for the device.
507 The stream tries to be word and endian agnostic, allowing migration between hosts
508 of different characteristics running the same VM.
514 - VM configuration section
519 Each section contains a device, or one iteration of a device save.
523 - ID string (First section of each device)
524 - instance id (First section of each device)
525 - version id (First section of each device)
529 - VM Description structure
530 Consisting of a JSON description of the contents for analysis only
532 The ``device data`` in each section consists of the data produced
533 by the code described above. For non-iterative devices they have a single
534 section; iterative devices have an initial and last section and a set
536 Note that there is very little checking by the common code of the integrity
537 of the ``device data`` contents, that's up to the devices themselves.
538 The ``footer mark`` provides a little bit of protection for the case where
539 the receiving side reads more or less data than expected.
541 The ``ID string`` is normally unique, having been formed from a bus name
542 and device address, PCI devices and storage devices hung off PCI controllers
543 fit this pattern well. Some devices are fixed single instances (e.g. "pc-ram").
544 Others (especially either older devices or system devices which for
545 some reason don't have a bus concept) make use of the ``instance id``
546 for otherwise identically named devices.
551 Only a unidirectional stream is required for normal migration, however a
552 ``return path`` can be created when bidirectional communication is desired.
553 This is primarily used by postcopy, but is also used to return a success
554 flag to the source at the end of migration.
556 ``qemu_file_get_return_path(QEMUFile* fwdpath)`` gives the QEMUFile* for the return
561 Forward path - written by migration thread
562 Return path - opened by main thread, read by return-path thread
566 Forward path - read by main thread
567 Return path - opened by main thread, written by main thread AND postcopy
568 thread (protected by rp_mutex)
573 'Postcopy' migration is a way to deal with migrations that refuse to converge
574 (or take too long to converge) its plus side is that there is an upper bound on
575 the amount of migration traffic and time it takes, the down side is that during
576 the postcopy phase, a failure of *either* side or the network connection causes
577 the guest to be lost.
579 In postcopy the destination CPUs are started before all the memory has been
580 transferred, and accesses to pages that are yet to be transferred cause
581 a fault that's translated by QEMU into a request to the source QEMU.
583 Postcopy can be combined with precopy (i.e. normal migration) so that if precopy
584 doesn't finish in a given time the switch is made to postcopy.
589 To enable postcopy, issue this command on the monitor (both source and
590 destination) prior to the start of migration:
592 ``migrate_set_capability postcopy-ram on``
594 The normal commands are then used to start a migration, which is still
595 started in precopy mode. Issuing:
597 ``migrate_start_postcopy``
599 will now cause the transition from precopy to postcopy.
600 It can be issued immediately after migration is started or any
601 time later on. Issuing it after the end of a migration is harmless.
603 Blocktime is a postcopy live migration metric, intended to show how
604 long the vCPU was in state of interruptable sleep due to pagefault.
605 That metric is calculated both for all vCPUs as overlapped value, and
606 separately for each vCPU. These values are calculated on destination
607 side. To enable postcopy blocktime calculation, enter following
608 command on destination monitor:
610 ``migrate_set_capability postcopy-blocktime on``
612 Postcopy blocktime can be retrieved by query-migrate qmp command.
613 postcopy-blocktime value of qmp command will show overlapped blocking
614 time for all vCPU, postcopy-vcpu-blocktime will show list of blocking
618 During the postcopy phase, the bandwidth limits set using
619 ``migrate_set_speed`` is ignored (to avoid delaying requested pages that
620 the destination is waiting for).
622 Postcopy device transfer
623 ------------------------
625 Loading of device data may cause the device emulation to access guest RAM
626 that may trigger faults that have to be resolved by the source, as such
627 the migration stream has to be able to respond with page data *during* the
628 device load, and hence the device data has to be read from the stream completely
629 before the device load begins to free the stream up. This is achieved by
630 'packaging' the device data into a blob that's read in one go.
635 Until postcopy is entered the migration stream is identical to normal
636 precopy, except for the addition of a 'postcopy advise' command at
637 the beginning, to tell the destination that postcopy might happen.
638 When postcopy starts the source sends the page discard data and then
639 forms the 'package' containing:
641 - Command: 'postcopy listen'
644 A series of sections, identical to the precopy streams device state stream
645 containing everything except postcopiable devices (i.e. RAM)
646 - Command: 'postcopy run'
648 The 'package' is sent as the data part of a Command: ``CMD_PACKAGED``, and the
649 contents are formatted in the same way as the main migration stream.
651 During postcopy the source scans the list of dirty pages and sends them
652 to the destination without being requested (in much the same way as precopy),
653 however when a page request is received from the destination, the dirty page
654 scanning restarts from the requested location. This causes requested pages
655 to be sent quickly, and also causes pages directly after the requested page
656 to be sent quickly in the hope that those pages are likely to be used
657 by the destination soon.
659 Destination behaviour
660 ---------------------
662 Initially the destination looks the same as precopy, with a single thread
663 reading the migration stream; the 'postcopy advise' and 'discard' commands
664 are processed to change the way RAM is managed, but don't affect the stream
669 ------------------------------------------------------------------------------
671 main -----DISCARD-CMD_PACKAGED ( LISTEN DEVICE DEVICE DEVICE RUN )
676 listen thread: --- page -- page -- page -- page -- page --
679 ------------------------------------------------------------------------------
681 - On receipt of ``CMD_PACKAGED`` (1)
683 All the data associated with the package - the ( ... ) section in the diagram -
684 is read into memory, and the main thread recurses into qemu_loadvm_state_main
685 to process the contents of the package (2) which contains commands (3,6) and
688 - On receipt of 'postcopy listen' - 3 -(i.e. the 1st command in the package)
690 a new thread (a) is started that takes over servicing the migration stream,
691 while the main thread carries on loading the package. It loads normal
692 background page data (b) but if during a device load a fault happens (5)
693 the returned page (c) is loaded by the listen thread allowing the main
694 threads device load to carry on.
696 - The last thing in the ``CMD_PACKAGED`` is a 'RUN' command (6)
698 letting the destination CPUs start running. At the end of the
699 ``CMD_PACKAGED`` (7) the main thread returns to normal running behaviour and
700 is no longer used by migration, while the listen thread carries on servicing
701 page data until the end of migration.
706 Postcopy moves through a series of states (see postcopy_state) from
707 ADVISE->DISCARD->LISTEN->RUNNING->END
711 Set at the start of migration if postcopy is enabled, even
712 if it hasn't had the start command; here the destination
713 checks that its OS has the support needed for postcopy, and performs
714 setup to ensure the RAM mappings are suitable for later postcopy.
715 The destination will fail early in migration at this point if the
716 required OS support is not present.
717 (Triggered by reception of POSTCOPY_ADVISE command)
721 Entered on receipt of the first 'discard' command; prior to
722 the first Discard being performed, hugepages are switched off
723 (using madvise) to ensure that no new huge pages are created
724 during the postcopy phase, and to cause any huge pages that
725 have discards on them to be broken.
729 The first command in the package, POSTCOPY_LISTEN, switches
730 the destination state to Listen, and starts a new thread
731 (the 'listen thread') which takes over the job of receiving
732 pages off the migration stream, while the main thread carries
733 on processing the blob. With this thread able to process page
734 reception, the destination now 'sensitises' the RAM to detect
735 any access to missing pages (on Linux using the 'userfault'
740 POSTCOPY_RUN causes the destination to synchronise all
741 state and start the CPUs and IO devices running. The main
742 thread now finishes processing the migration package and
743 now carries on as it would for normal precopy migration
744 (although it can't do the cleanup it would do as it
745 finishes a normal migration).
749 The listen thread can now quit, and perform the cleanup of migration
750 state, the migration is now complete.
752 Source side page maps
753 ---------------------
755 The source side keeps two bitmaps during postcopy; 'the migration bitmap'
756 and 'unsent map'. The 'migration bitmap' is basically the same as in
757 the precopy case, and holds a bit to indicate that page is 'dirty' -
758 i.e. needs sending. During the precopy phase this is updated as the CPU
759 dirties pages, however during postcopy the CPUs are stopped and nothing
760 should dirty anything any more.
762 The 'unsent map' is used for the transition to postcopy. It is a bitmap that
763 has a bit cleared whenever a page is sent to the destination, however during
764 the transition to postcopy mode it is combined with the migration bitmap
765 to form a set of pages that:
767 a) Have been sent but then redirtied (which must be discarded)
768 b) Have not yet been sent - which also must be discarded to cause any
769 transparent huge pages built during precopy to be broken.
771 Note that the contents of the unsentmap are sacrificed during the calculation
772 of the discard set and thus aren't valid once in postcopy. The dirtymap
773 is still valid and is used to ensure that no page is sent more than once. Any
774 request for a page that has already been sent is ignored. Duplicate requests
775 such as this can happen as a page is sent at about the same time the
776 destination accesses it.
778 Postcopy with hugepages
779 -----------------------
781 Postcopy now works with hugetlbfs backed memory:
783 a) The linux kernel on the destination must support userfault on hugepages.
784 b) The huge-page configuration on the source and destination VMs must be
785 identical; i.e. RAMBlocks on both sides must use the same page size.
786 c) Note that ``-mem-path /dev/hugepages`` will fall back to allocating normal
787 RAM if it doesn't have enough hugepages, triggering (b) to fail.
788 Using ``-mem-prealloc`` enforces the allocation using hugepages.
789 d) Care should be taken with the size of hugepage used; postcopy with 2MB
790 hugepages works well, however 1GB hugepages are likely to be problematic
791 since it takes ~1 second to transfer a 1GB hugepage across a 10Gbps link,
792 and until the full page is transferred the destination thread is blocked.
794 Postcopy with shared memory
795 ---------------------------
797 Postcopy migration with shared memory needs explicit support from the other
798 processes that share memory and from QEMU. There are restrictions on the type of
799 memory that userfault can support shared.
801 The Linux kernel userfault support works on `/dev/shm` memory and on `hugetlbfs`
802 (although the kernel doesn't provide an equivalent to `madvise(MADV_DONTNEED)`
803 for hugetlbfs which may be a problem in some configurations).
805 The vhost-user code in QEMU supports clients that have Postcopy support,
806 and the `vhost-user-bridge` (in `tests/`) and the DPDK package have changes
809 The client needs to open a userfaultfd and register the areas
810 of memory that it maps with userfault. The client must then pass the
811 userfaultfd back to QEMU together with a mapping table that allows
812 fault addresses in the clients address space to be converted back to
813 RAMBlock/offsets. The client's userfaultfd is added to the postcopy
814 fault-thread and page requests are made on behalf of the client by QEMU.
815 QEMU performs 'wake' operations on the client's userfaultfd to allow it
816 to continue after a page has arrived.
819 There are two future improvements that would be nice:
820 a) Some way to make QEMU ignorant of the addresses in the clients
822 b) Avoiding the need for QEMU to perform ufd-wake calls after the
825 Retro-fitting postcopy to existing clients is possible:
826 a) A mechanism is needed for the registration with userfault as above,
827 and the registration needs to be coordinated with the phases of
828 postcopy. In vhost-user extra messages are added to the existing
830 b) Any thread that can block due to guest memory accesses must be
831 identified and the implication understood; for example if the
832 guest memory access is made while holding a lock then all other
833 threads waiting for that lock will also be blocked.
838 Migration migrates the copies of RAM and ROM, and thus when running
839 on the destination it includes the firmware from the source. Even after
840 resetting a VM, the old firmware is used. Only once QEMU has been restarted
841 is the new firmware in use.
843 - Changes in firmware size can cause changes in the required RAMBlock size
844 to hold the firmware and thus migration can fail. In practice it's best
845 to pad firmware images to convenient powers of 2 with plenty of space
848 - Care should be taken with device emulation code so that newer
849 emulation code can work with older firmware to allow forward migration.
851 - Care should be taken with newer firmware so that backward migration
852 to older systems with older device emulation code will work.
854 In some cases it may be best to tie specific firmware versions to specific
855 versioned machine types to cut down on the combinations that will need
856 support. This is also useful when newer versions of firmware outgrow