Merge remote-tracking branch 'remotes/ericb/tags/pull-bitmaps-2020-07-27' into staging

bitmaps patches for 2020-07-27

- Improve handling of various post-copy bitmap migration scenarios. A lost
bitmap should merely mean that the next backup must be full rather than
incremental, rather than abruptly breaking the entire guest migration.
- Associated iotest improvements

# gpg: Signature made Mon 27 Jul 2020 21:46:17 BST
# gpg:                using RSA key 71C2CC22B1C4602927D2F3AAA7A16B4A2527436A
# gpg: Good signature from "Eric Blake <eblake@redhat.com>" [full]
# gpg:                 aka "Eric Blake (Free Software Programmer) <ebb9@byu.net>" [full]
# gpg:                 aka "[jpeg image of size 6874]" [full]
# Primary key fingerprint: 71C2 CC22 B1C4 6029 27D2  F3AA A7A1 6B4A 2527 436A

* remotes/ericb/tags/pull-bitmaps-2020-07-27: (24 commits)
  migration: Fix typos in bitmap migration comments
  iotests: Adjust which migration tests are quick
  qemu-iotests/199: add source-killed case to bitmaps postcopy
  qemu-iotests/199: add early shutdown case to bitmaps postcopy
  qemu-iotests/199: check persistent bitmaps
  qemu-iotests/199: prepare for new test-cases addition
  migration/savevm: don't worry if bitmap migration postcopy failed
  migration/block-dirty-bitmap: cancel migration on shutdown
  migration/block-dirty-bitmap: relax error handling in incoming part
  migration/block-dirty-bitmap: keep bitmap state for all bitmaps
  migration/block-dirty-bitmap: simplify dirty_bitmap_load_complete
  migration/block-dirty-bitmap: rename finish_lock to just lock
  migration/block-dirty-bitmap: refactor state global variables
  migration/block-dirty-bitmap: move mutex init to dirty_bitmap_mig_init
  migration/block-dirty-bitmap: rename dirty_bitmap_mig_cleanup
  migration/block-dirty-bitmap: rename state structure types
  migration/block-dirty-bitmap: fix dirty_bitmap_mig_before_vm_start
  qemu-iotests/199: increase postcopy period
  qemu-iotests/199: change discard patterns
  qemu-iotests/199: improve performance: set bitmap by discard
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This commit is contained in:
Peter Maydell
2020-07-28 14:38:17 +01:00
9 changed files with 555 additions and 244 deletions
+1 -1
View File
@@ -66,7 +66,7 @@ typedef struct {
} QEMU_PACKED QCowExtension;
#define QCOW2_EXT_MAGIC_END 0
#define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA
#define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xe2792aca
#define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857
#define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77
#define QCOW2_EXT_MAGIC_BITMAPS 0x23852875
+1 -1
View File
@@ -231,7 +231,7 @@ be stored. Each extension has a structure like the following:
Byte 0 - 3: Header extension type:
0x00000000 - End of the header extension area
0xE2792ACA - Backing file format name string
0xe2792aca - Backing file format name string
0x6803f857 - Feature name table
0x23852875 - Bitmaps extension
0x0537be77 - Full disk encryption header pointer
File diff suppressed because it is too large Load Diff
+13 -2
View File
@@ -165,8 +165,6 @@ void migration_object_init(void)
qemu_sem_init(&current_incoming->postcopy_pause_sem_dst, 0);
qemu_sem_init(&current_incoming->postcopy_pause_sem_fault, 0);
init_dirty_bitmap_incoming_migration();
if (!migration_object_check(current_migration, &err)) {
error_report_err(err);
exit(1);
@@ -190,6 +188,19 @@ void migration_shutdown(void)
*/
migrate_fd_cancel(current_migration);
object_unref(OBJECT(current_migration));
/*
* Cancel outgoing migration of dirty bitmaps. It should
* at least unref used block nodes.
*/
dirty_bitmap_mig_cancel_outgoing();
/*
* Cancel incoming migration of dirty bitmaps. Dirty bitmaps
* are non-critical data, and their loss never considered as
* something serious.
*/
dirty_bitmap_mig_cancel_incoming();
}
/* For outgoing */
+2 -1
View File
@@ -335,7 +335,8 @@ void migrate_send_rp_recv_bitmap(MigrationIncomingState *mis,
void migrate_send_rp_resume_ack(MigrationIncomingState *mis, uint32_t value);
void dirty_bitmap_mig_before_vm_start(void);
void init_dirty_bitmap_incoming_migration(void);
void dirty_bitmap_mig_cancel_outgoing(void);
void dirty_bitmap_mig_cancel_incoming(void);
void migrate_add_address(SocketAddress *address);
int foreach_not_ignored_block(RAMBlockIterFunc func, void *opaque);
+32 -5
View File
@@ -1813,6 +1813,9 @@ static void *postcopy_ram_listen_thread(void *opaque)
MigrationIncomingState *mis = migration_incoming_get_current();
QEMUFile *f = mis->from_src_file;
int load_res;
MigrationState *migr = migrate_get_current();
object_ref(OBJECT(migr));
migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
MIGRATION_STATUS_POSTCOPY_ACTIVE);
@@ -1839,11 +1842,24 @@ static void *postcopy_ram_listen_thread(void *opaque)
trace_postcopy_ram_listen_thread_exit();
if (load_res < 0) {
error_report("%s: loadvm failed: %d", __func__, load_res);
qemu_file_set_error(f, load_res);
migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE,
MIGRATION_STATUS_FAILED);
} else {
dirty_bitmap_mig_cancel_incoming();
if (postcopy_state_get() == POSTCOPY_INCOMING_RUNNING &&
!migrate_postcopy_ram() && migrate_dirty_bitmaps())
{
error_report("%s: loadvm failed during postcopy: %d. All states "
"are migrated except dirty bitmaps. Some dirty "
"bitmaps may be lost, and present migrated dirty "
"bitmaps are correctly migrated and valid.",
__func__, load_res);
load_res = 0; /* prevent further exit() */
} else {
error_report("%s: loadvm failed: %d", __func__, load_res);
migrate_set_state(&mis->state, MIGRATION_STATUS_POSTCOPY_ACTIVE,
MIGRATION_STATUS_FAILED);
}
}
if (load_res >= 0) {
/*
* This looks good, but it's possible that the device loading in the
* main thread hasn't finished yet, and so we might not be in 'RUN'
@@ -1879,6 +1895,8 @@ static void *postcopy_ram_listen_thread(void *opaque)
mis->have_listen_thread = false;
postcopy_state_set(POSTCOPY_INCOMING_END);
object_unref(OBJECT(migr));
return NULL;
}
@@ -2437,6 +2455,8 @@ static bool postcopy_pause_incoming(MigrationIncomingState *mis)
{
trace_postcopy_pause_incoming();
assert(migrate_postcopy_ram());
/* Clear the triggered bit to allow one recovery */
mis->postcopy_recover_triggered = false;
@@ -2521,15 +2541,22 @@ out:
if (ret < 0) {
qemu_file_set_error(f, ret);
/* Cancel bitmaps incoming regardless of recovery */
dirty_bitmap_mig_cancel_incoming();
/*
* If we are during an active postcopy, then we pause instead
* of bail out to at least keep the VM's dirty data. Note
* that POSTCOPY_INCOMING_LISTENING stage is still not enough,
* during which we're still receiving device states and we
* still haven't yet started the VM on destination.
*
* Only RAM postcopy supports recovery. Still, if RAM postcopy is
* enabled, canceled bitmaps postcopy will not affect RAM postcopy
* recovering.
*/
if (postcopy_state_get() == POSTCOPY_INCOMING_RUNNING &&
postcopy_pause_incoming(mis)) {
migrate_postcopy_ram() && postcopy_pause_incoming(mis)) {
/* Reset f to point to the newly created channel */
f = mis->from_src_file;
goto retry;
+194 -52
View File
@@ -20,17 +20,76 @@
import os
import iotests
import time
from iotests import qemu_img
debug = False
disk_a = os.path.join(iotests.test_dir, 'disk_a')
disk_b = os.path.join(iotests.test_dir, 'disk_b')
size = '256G'
fifo = os.path.join(iotests.test_dir, 'mig_fifo')
class TestDirtyBitmapPostcopyMigration(iotests.QMPTestCase):
granularity = 512
nb_bitmaps = 15
GiB = 1024 * 1024 * 1024
discards1 = (
(0, GiB),
(2 * GiB + 512 * 5, 512),
(3 * GiB + 512 * 5, 512),
(100 * GiB, GiB)
)
discards2 = (
(3 * GiB + 512 * 8, 512),
(4 * GiB + 512 * 8, 512),
(50 * GiB, GiB),
(100 * GiB + GiB // 2, GiB)
)
def apply_discards(vm, discards):
for d in discards:
vm.hmp_qemu_io('drive0', 'discard {} {}'.format(*d))
def event_seconds(event):
return event['timestamp']['seconds'] + \
event['timestamp']['microseconds'] / 1000000.0
def event_dist(e1, e2):
return event_seconds(e2) - event_seconds(e1)
def check_bitmaps(vm, count):
result = vm.qmp('query-block')
if count == 0:
assert 'dirty-bitmaps' not in result['return'][0]
else:
assert len(result['return'][0]['dirty-bitmaps']) == count
class TestDirtyBitmapPostcopyMigration(iotests.QMPTestCase):
def tearDown(self):
if debug:
self.vm_a_events += self.vm_a.get_qmp_events()
self.vm_b_events += self.vm_b.get_qmp_events()
for e in self.vm_a_events:
e['vm'] = 'SRC'
for e in self.vm_b_events:
e['vm'] = 'DST'
events = (self.vm_a_events + self.vm_b_events)
events = [(e['timestamp']['seconds'],
e['timestamp']['microseconds'],
e['vm'],
e['event'],
e.get('data', '')) for e in events]
for e in sorted(events):
print('{}.{:06} {} {} {}'.format(*e))
self.vm_a.shutdown()
self.vm_b.shutdown()
os.remove(disk_a)
@@ -41,51 +100,66 @@ class TestDirtyBitmapPostcopyMigration(iotests.QMPTestCase):
os.mkfifo(fifo)
qemu_img('create', '-f', iotests.imgfmt, disk_a, size)
qemu_img('create', '-f', iotests.imgfmt, disk_b, size)
self.vm_a = iotests.VM(path_suffix='a').add_drive(disk_a)
self.vm_b = iotests.VM(path_suffix='b').add_drive(disk_b)
self.vm_a = iotests.VM(path_suffix='a').add_drive(disk_a,
'discard=unmap')
self.vm_b = iotests.VM(path_suffix='b').add_drive(disk_b,
'discard=unmap')
self.vm_b.add_incoming("exec: cat '" + fifo + "'")
self.vm_a.launch()
self.vm_b.launch()
def test_postcopy(self):
write_size = 0x40000000
granularity = 512
chunk = 4096
# collect received events for debug
self.vm_a_events = []
self.vm_b_events = []
result = self.vm_a.qmp('block-dirty-bitmap-add', node='drive0',
name='bitmap', granularity=granularity)
self.assert_qmp(result, 'return', {});
s = 0
while s < write_size:
self.vm_a.hmp_qemu_io('drive0', 'write %d %d' % (s, chunk))
s += 0x10000
s = 0x8000
while s < write_size:
self.vm_a.hmp_qemu_io('drive0', 'write %d %d' % (s, chunk))
s += 0x10000
def start_postcopy(self):
""" Run migration until RESUME event on target. Return this event. """
for i in range(nb_bitmaps):
result = self.vm_a.qmp('block-dirty-bitmap-add', node='drive0',
name='bitmap{}'.format(i),
granularity=granularity,
persistent=True)
self.assert_qmp(result, 'return', {})
result = self.vm_a.qmp('x-debug-block-dirty-bitmap-sha256',
node='drive0', name='bitmap')
sha256 = result['return']['sha256']
node='drive0', name='bitmap0')
empty_sha256 = result['return']['sha256']
result = self.vm_a.qmp('block-dirty-bitmap-clear', node='drive0',
name='bitmap')
self.assert_qmp(result, 'return', {});
s = 0
while s < write_size:
self.vm_a.hmp_qemu_io('drive0', 'write %d %d' % (s, chunk))
s += 0x10000
apply_discards(self.vm_a, discards1)
bitmaps_cap = {'capability': 'dirty-bitmaps', 'state': True}
events_cap = {'capability': 'events', 'state': True}
result = self.vm_a.qmp('x-debug-block-dirty-bitmap-sha256',
node='drive0', name='bitmap0')
self.discards1_sha256 = result['return']['sha256']
result = self.vm_a.qmp('migrate-set-capabilities',
capabilities=[bitmaps_cap, events_cap])
# Check, that updating the bitmap by discards works
assert self.discards1_sha256 != empty_sha256
# We want to calculate resulting sha256. Do it in bitmap0, so, disable
# other bitmaps
for i in range(1, nb_bitmaps):
result = self.vm_a.qmp('block-dirty-bitmap-disable', node='drive0',
name='bitmap{}'.format(i))
self.assert_qmp(result, 'return', {})
apply_discards(self.vm_a, discards2)
result = self.vm_a.qmp('x-debug-block-dirty-bitmap-sha256',
node='drive0', name='bitmap0')
self.all_discards_sha256 = result['return']['sha256']
# Now, enable some bitmaps, to be updated during migration
for i in range(2, nb_bitmaps, 2):
result = self.vm_a.qmp('block-dirty-bitmap-enable', node='drive0',
name='bitmap{}'.format(i))
self.assert_qmp(result, 'return', {})
caps = [{'capability': 'dirty-bitmaps', 'state': True},
{'capability': 'events', 'state': True}]
result = self.vm_a.qmp('migrate-set-capabilities', capabilities=caps)
self.assert_qmp(result, 'return', {})
result = self.vm_b.qmp('migrate-set-capabilities',
capabilities=[bitmaps_cap])
result = self.vm_b.qmp('migrate-set-capabilities', capabilities=caps)
self.assert_qmp(result, 'return', {})
result = self.vm_a.qmp('migrate', uri='exec:cat>' + fifo)
@@ -94,26 +168,94 @@ class TestDirtyBitmapPostcopyMigration(iotests.QMPTestCase):
result = self.vm_a.qmp('migrate-start-postcopy')
self.assert_qmp(result, 'return', {})
while True:
event = self.vm_a.event_wait('MIGRATION')
if event['data']['status'] == 'completed':
break
event_resume = self.vm_b.event_wait('RESUME')
self.vm_b_events.append(event_resume)
return event_resume
s = 0x8000
while s < write_size:
self.vm_b.hmp_qemu_io('drive0', 'write %d %d' % (s, chunk))
s += 0x10000
def test_postcopy_success(self):
event_resume = self.start_postcopy()
result = self.vm_b.qmp('query-block');
while len(result['return'][0]['dirty-bitmaps']) > 1:
time.sleep(2)
result = self.vm_b.qmp('query-block');
# enabled bitmaps should be updated
apply_discards(self.vm_b, discards2)
result = self.vm_b.qmp('x-debug-block-dirty-bitmap-sha256',
node='drive0', name='bitmap')
match = {'data': {'status': 'completed'}}
event_complete = self.vm_b.event_wait('MIGRATION', match=match)
self.vm_b_events.append(event_complete)
# take queued event, should already been happened
event_stop = self.vm_a.event_wait('STOP')
self.vm_a_events.append(event_stop)
downtime = event_dist(event_stop, event_resume)
postcopy_time = event_dist(event_resume, event_complete)
assert downtime * 10 < postcopy_time
if debug:
print('downtime:', downtime)
print('postcopy_time:', postcopy_time)
# check that there are no bitmaps stored on source
self.vm_a_events += self.vm_a.get_qmp_events()
self.vm_a.shutdown()
self.vm_a.launch()
check_bitmaps(self.vm_a, 0)
# check that bitmaps are migrated and persistence works
check_bitmaps(self.vm_b, nb_bitmaps)
self.vm_b.shutdown()
# recreate vm_b, so there is no incoming option, which prevents
# loading bitmaps from disk
self.vm_b = iotests.VM(path_suffix='b').add_drive(disk_b)
self.vm_b.launch()
check_bitmaps(self.vm_b, nb_bitmaps)
# Check content of migrated bitmaps. Still, don't waste time checking
# every bitmap
for i in range(0, nb_bitmaps, 5):
result = self.vm_b.qmp('x-debug-block-dirty-bitmap-sha256',
node='drive0', name='bitmap{}'.format(i))
sha = self.discards1_sha256 if i % 2 else self.all_discards_sha256
self.assert_qmp(result, 'return/sha256', sha)
def test_early_shutdown_destination(self):
self.start_postcopy()
self.vm_b_events += self.vm_b.get_qmp_events()
self.vm_b.shutdown()
# recreate vm_b, so there is no incoming option, which prevents
# loading bitmaps from disk
self.vm_b = iotests.VM(path_suffix='b').add_drive(disk_b)
self.vm_b.launch()
check_bitmaps(self.vm_b, 0)
# Bitmaps will be lost if we just shutdown the vm, as they are marked
# to skip storing to disk when prepared for migration. And that's
# correct, as actual data may be modified in target vm, so we play
# safe.
# Still, this mark would be taken away if we do 'cont', and bitmaps
# become persistent again. (see iotest 169 for such behavior case)
result = self.vm_a.qmp('query-status')
assert not result['return']['running']
self.vm_a_events += self.vm_a.get_qmp_events()
self.vm_a.shutdown()
self.vm_a.launch()
check_bitmaps(self.vm_a, 0)
def test_early_kill_source(self):
self.start_postcopy()
self.vm_a_events = self.vm_a.get_qmp_events()
self.vm_a.kill()
self.vm_a.launch()
match = {'data': {'status': 'completed'}}
e_complete = self.vm_b.event_wait('MIGRATION', match=match)
self.vm_b_events.append(e_complete)
check_bitmaps(self.vm_a, 0)
check_bitmaps(self.vm_b, 0)
self.assert_qmp(result, 'return/sha256', sha256);
if __name__ == '__main__':
iotests.main(supported_fmts=['qcow2'], supported_cache_modes=['none'],
supported_protocols=['file'])
iotests.main(supported_fmts=['qcow2'])
+2 -2
View File
@@ -1,5 +1,5 @@
.
...
----------------------------------------------------------------------
Ran 1 tests
Ran 3 tests
OK
+6 -6
View File
@@ -112,7 +112,7 @@
088 rw quick
089 rw auto quick
090 rw auto quick
091 rw migration
091 rw migration quick
092 rw quick
093 throttle
094 rw quick
@@ -186,7 +186,7 @@
162 quick
163 rw
165 rw quick
169 rw quick migration
169 rw migration
170 rw auto quick
171 rw quick
172 auto
@@ -197,9 +197,9 @@
177 rw auto quick
178 img
179 rw auto quick
181 rw auto migration
181 rw auto migration quick
182 rw quick
183 rw migration
183 rw migration quick
184 rw auto quick
185 rw
186 rw auto
@@ -216,9 +216,9 @@
198 rw
199 rw migration
200 rw
201 rw migration
201 rw migration quick
202 rw quick
203 rw auto migration
203 rw auto migration quick
204 rw quick
205 rw quick
206 rw