tests/qemu-iotests/257

   1 #!/usr/bin/env python
   2 #
   3 # Test bitmap-sync backups (incremental, differential, and partials)
   4 #
   5 # Copyright (c) 2019 John Snow for Red Hat, Inc.
   6 #
   7 # This program is free software; you can redistribute it and/or modify
   8 # it under the terms of the GNU General Public License as published by
   9 # the Free Software Foundation; either version 2 of the License, or
  10 # (at your option) any later version.
  11 #
  12 # This program is distributed in the hope that it will be useful,
  13 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  14 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15 # GNU General Public License for more details.
  16 #
  17 # You should have received a copy of the GNU General Public License
  18 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
  19 #
  20 # owner=jsnow@redhat.com
  21
  22 import math
  23 import os
  24
  25 import iotests
  26 from iotests import log, qemu_img
  27
  28 SIZE = 64 * 1024 * 1024
  29 GRANULARITY = 64 * 1024
  30
  31
  32 class Pattern:
  33     def __init__(self, byte, offset, size=GRANULARITY):
  34         self.byte = byte
  35         self.offset = offset
  36         self.size = size
  37
  38     def bits(self, granularity):
  39         lower = self.offset // granularity
  40         upper = (self.offset + self.size - 1) // granularity
  41         return set(range(lower, upper + 1))
  42
  43
  44 class PatternGroup:
  45     """Grouping of Pattern objects. Initialize with an iterable of Patterns."""
  46     def __init__(self, patterns):
  47         self.patterns = patterns
  48
  49     def bits(self, granularity):
  50         """Calculate the unique bits dirtied by this pattern grouping"""
  51         res = set()
  52         for pattern in self.patterns:
  53             res |= pattern.bits(granularity)
  54         return res
  55
  56
  57 GROUPS = [
  58     PatternGroup([
  59         # Batch 0: 4 clusters
  60         Pattern('0x49', 0x0000000),
  61         Pattern('0x6c', 0x0100000),   # 1M
  62         Pattern('0x6f', 0x2000000),   # 32M
  63         Pattern('0x76', 0x3ff0000)]), # 64M - 64K
  64     PatternGroup([
  65         # Batch 1: 6 clusters (3 new)
  66         Pattern('0x65', 0x0000000),   # Full overwrite
  67         Pattern('0x77', 0x00f8000),   # Partial-left (1M-32K)
  68         Pattern('0x72', 0x2008000),   # Partial-right (32M+32K)
  69         Pattern('0x69', 0x3fe0000)]), # Adjacent-left (64M - 128K)
  70     PatternGroup([
  71         # Batch 2: 7 clusters (3 new)
  72         Pattern('0x74', 0x0010000),   # Adjacent-right
  73         Pattern('0x69', 0x00e8000),   # Partial-left  (1M-96K)
  74         Pattern('0x6e', 0x2018000),   # Partial-right (32M+96K)
  75         Pattern('0x67', 0x3fe0000,
  76                 2*GRANULARITY)]),     # Overwrite [(64M-128K)-64M)
  77     PatternGroup([
  78         # Batch 3: 8 clusters (5 new)
  79         # Carefully chosen such that nothing re-dirties the one cluster
  80         # that copies out successfully before failure in Group #1.
  81         Pattern('0xaa', 0x0010000,
  82                 3*GRANULARITY),       # Overwrite and 2x Adjacent-right
  83         Pattern('0xbb', 0x00d8000),   # Partial-left (1M-160K)
  84         Pattern('0xcc', 0x2028000),   # Partial-right (32M+160K)
  85         Pattern('0xdd', 0x3fc0000)]), # New; leaving a gap to the right
  86 ]
  87
  88
  89 class EmulatedBitmap:
  90     def __init__(self, granularity=GRANULARITY):
  91         self._bits = set()
  92         self.granularity = granularity
  93
  94     def dirty_bits(self, bits):
  95         self._bits |= set(bits)
  96
  97     def dirty_group(self, n):
  98         self.dirty_bits(GROUPS[n].bits(self.granularity))
  99
 100     def clear(self):
 101         self._bits = set()
 102
 103     def clear_bits(self, bits):
 104         self._bits -= set(bits)
 105
 106     def clear_bit(self, bit):
 107         self.clear_bits({bit})
 108
 109     def clear_group(self, n):
 110         self.clear_bits(GROUPS[n].bits(self.granularity))
 111
 112     @property
 113     def first_bit(self):
 114         return sorted(self.bits)[0]
 115
 116     @property
 117     def bits(self):
 118         return self._bits
 119
 120     @property
 121     def count(self):
 122         return len(self.bits)
 123
 124     def compare(self, qmp_bitmap):
 125         """
 126         Print a nice human-readable message checking that a bitmap as reported
 127         by the QMP interface has as many bits set as we expect it to.
 128         """
 129
 130         name = qmp_bitmap.get('name', '(anonymous)')
 131         log("= Checking Bitmap {:s} =".format(name))
 132
 133         want = self.count
 134         have = qmp_bitmap['count'] // qmp_bitmap['granularity']
 135
 136         log("expecting {:d} dirty sectors; have {:d}. {:s}".format(
 137             want, have, "OK!" if want == have else "ERROR!"))
 138         log('')
 139
 140
 141 class Drive:
 142     """Represents, vaguely, a drive attached to a VM.
 143     Includes format, graph, and device information."""
 144
 145     def __init__(self, path, vm=None):
 146         self.path = path
 147         self.vm = vm
 148         self.fmt = None
 149         self.size = None
 150         self.node = None
 151
 152     def img_create(self, fmt, size):
 153         self.fmt = fmt
 154         self.size = size
 155         iotests.qemu_img_create('-f', self.fmt, self.path, str(self.size))
 156
 157     def create_target(self, name, fmt, size):
 158         basename = os.path.basename(self.path)
 159         file_node_name = "file_{}".format(basename)
 160         vm = self.vm
 161
 162         log(vm.command('blockdev-create', job_id='bdc-file-job',
 163                        options={
 164                            'driver': 'file',
 165                            'filename': self.path,
 166                            'size': 0,
 167                        }))
 168         vm.run_job('bdc-file-job')
 169         log(vm.command('blockdev-add', driver='file',
 170                        node_name=file_node_name, filename=self.path))
 171
 172         log(vm.command('blockdev-create', job_id='bdc-fmt-job',
 173                        options={
 174                            'driver': fmt,
 175                            'file': file_node_name,
 176                            'size': size,
 177                        }))
 178         vm.run_job('bdc-fmt-job')
 179         log(vm.command('blockdev-add', driver=fmt,
 180                        node_name=name,
 181                        file=file_node_name))
 182         self.fmt = fmt
 183         self.size = size
 184         self.node = name
 185
 186 def blockdev_backup(vm, device, target, sync, **kwargs):
 187     # Strip any arguments explicitly nulled by the caller:
 188     kwargs = {key: val for key, val in kwargs.items() if val is not None}
 189     result = vm.qmp_log('blockdev-backup',
 190                         device=device,
 191                         target=target,
 192                         sync=sync,
 193                         filter_node_name='backup-top',
 194                         **kwargs)
 195     return result
 196
 197 def blockdev_backup_mktarget(drive, target_id, filepath, sync, **kwargs):
 198     target_drive = Drive(filepath, vm=drive.vm)
 199     target_drive.create_target(target_id, drive.fmt, drive.size)
 200     blockdev_backup(drive.vm, drive.node, target_id, sync, **kwargs)
 201
 202 def reference_backup(drive, n, filepath):
 203     log("--- Reference Backup #{:d} ---\n".format(n))
 204     target_id = "ref_target_{:d}".format(n)
 205     job_id = "ref_backup_{:d}".format(n)
 206     blockdev_backup_mktarget(drive, target_id, filepath, "full",
 207                              job_id=job_id)
 208     drive.vm.run_job(job_id, auto_dismiss=True)
 209     log('')
 210
 211 def backup(drive, n, filepath, sync, **kwargs):
 212     log("--- Test Backup #{:d} ---\n".format(n))
 213     target_id = "backup_target_{:d}".format(n)
 214     job_id = "backup_{:d}".format(n)
 215     kwargs.setdefault('auto-finalize', False)
 216     blockdev_backup_mktarget(drive, target_id, filepath, sync,
 217                              job_id=job_id, **kwargs)
 218     return job_id
 219
 220 def perform_writes(drive, n, filter_node_name=None):
 221     log("--- Write #{:d} ---\n".format(n))
 222     for pattern in GROUPS[n].patterns:
 223         cmd = "write -P{:s} 0x{:07x} 0x{:x}".format(
 224             pattern.byte,
 225             pattern.offset,
 226             pattern.size)
 227         log(cmd)
 228         log(drive.vm.hmp_qemu_io(filter_node_name or drive.node, cmd))
 229     bitmaps = drive.vm.query_bitmaps()
 230     log({'bitmaps': bitmaps}, indent=2)
 231     log('')
 232     return bitmaps
 233
 234
 235 def compare_images(image, reference, baseimg=None, expected_match=True):
 236     """
 237     Print a nice human-readable message comparing these images.
 238     """
 239     expected_ret = 0 if expected_match else 1
 240     if baseimg:
 241         assert qemu_img("rebase", "-u", "-b", baseimg, image) == 0
 242     ret = qemu_img("compare", image, reference)
 243     log('qemu_img compare "{:s}" "{:s}" ==> {:s}, {:s}'.format(
 244         image, reference,
 245         "Identical" if ret == 0 else "Mismatch",
 246         "OK!" if ret == expected_ret else "ERROR!"),
 247         filters=[iotests.filter_testfiles])
 248
 249 def test_bitmap_sync(bsync_mode, msync_mode='bitmap', failure=None):
 250     """
 251     Test bitmap backup routines.
 252
 253     :param bsync_mode: Is the Bitmap Sync mode, and can be any of:
 254         - on-success: This is the "incremental" style mode. Bitmaps are
 255                       synchronized to what was copied out only on success.
 256                       (Partial images must be discarded.)
 257         - never:      This is the "differential" style mode.
 258                       Bitmaps are never synchronized.
 259         - always:     This is a "best effort" style mode.
 260                       Bitmaps are always synchronized, regardless of failure.
 261                       (Partial images must be kept.)
 262
 263     :param msync_mode: The mirror sync mode to use for the first backup.
 264                        Can be any one of:
 265         - bitmap: Backups based on bitmap manifest.
 266         - full:   Full backups.
 267         - top:    Full backups of the top layer only.
 268
 269     :param failure: Is the (optional) failure mode, and can be any of:
 270         - None:         No failure. Test the normative path. Default.
 271         - simulated:    Cancel the job right before it completes.
 272                         This also tests writes "during" the job.
 273         - intermediate: This tests a job that fails mid-process and produces
 274                         an incomplete backup. Testing limitations prevent
 275                         testing competing writes.
 276     """
 277     with iotests.FilePaths(['img', 'bsync1', 'bsync2',
 278                             'fbackup0', 'fbackup1', 'fbackup2']) as \
 279                             (img_path, bsync1, bsync2,
 280                              fbackup0, fbackup1, fbackup2), \
 281          iotests.VM() as vm:
 282
 283         mode = "Mode {:s}; Bitmap Sync {:s}".format(msync_mode, bsync_mode)
 284         preposition = "with" if failure else "without"
 285         cond = "{:s} {:s}".format(preposition,
 286                                   "{:s} failure".format(failure) if failure
 287                                   else "failure")
 288         log("\n=== {:s} {:s} ===\n".format(mode, cond))
 289
 290         log('--- Preparing image & VM ---\n')
 291         drive0 = Drive(img_path, vm=vm)
 292         drive0.img_create(iotests.imgfmt, SIZE)
 293         vm.add_device("{},id=scsi0".format(iotests.get_virtio_scsi_device()))
 294         vm.launch()
 295
 296         file_config = {
 297             'driver': 'file',
 298             'filename': drive0.path
 299         }
 300
 301         if failure == 'intermediate':
 302             file_config = {
 303                 'driver': 'blkdebug',
 304                 'image': file_config,
 305                 'set-state': [{
 306                     'event': 'flush_to_disk',
 307                     'state': 1,
 308                     'new_state': 2
 309                 }, {
 310                     'event': 'read_aio',
 311                     'state': 2,
 312                     'new_state': 3
 313                 }],
 314                 'inject-error': [{
 315                     'event': 'read_aio',
 316                     'errno': 5,
 317                     'state': 3,
 318                     'immediately': False,
 319                     'once': True
 320                 }]
 321             }
 322
 323         drive0.node = 'drive0'
 324         vm.qmp_log('blockdev-add',
 325                    filters=[iotests.filter_qmp_testfiles],
 326                    node_name=drive0.node,
 327                    driver=drive0.fmt,
 328                    file=file_config)
 329         log('')
 330
 331         # 0 - Writes and Reference Backup
 332         perform_writes(drive0, 0)
 333         reference_backup(drive0, 0, fbackup0)
 334         log('--- Add Bitmap ---\n')
 335         vm.qmp_log("block-dirty-bitmap-add", node=drive0.node,
 336                    name="bitmap0", granularity=GRANULARITY)
 337         log('')
 338         ebitmap = EmulatedBitmap()
 339
 340         # 1 - Writes and Reference Backup
 341         bitmaps = perform_writes(drive0, 1)
 342         ebitmap.dirty_group(1)
 343         bitmap = vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps)
 344         ebitmap.compare(bitmap)
 345         reference_backup(drive0, 1, fbackup1)
 346
 347         # 1 - Test Backup (w/ Optional induced failure)
 348         if failure == 'intermediate':
 349             # Activate blkdebug induced failure for second-to-next read
 350             log(vm.hmp_qemu_io(drive0.node, 'flush'))
 351             log('')
 352         job = backup(drive0, 1, bsync1, msync_mode,
 353                      bitmap="bitmap0", bitmap_mode=bsync_mode)
 354
 355         def _callback():
 356             """Issue writes while the job is open to test bitmap divergence."""
 357             # Note: when `failure` is 'intermediate', this isn't called.
 358             log('')
 359             bitmaps = perform_writes(drive0, 2, filter_node_name='backup-top')
 360             # Named bitmap (static, should be unchanged)
 361             ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0',
 362                                           bitmaps=bitmaps))
 363             # Anonymous bitmap (dynamic, shows new writes)
 364             anonymous = EmulatedBitmap()
 365             anonymous.dirty_group(2)
 366             anonymous.compare(vm.get_bitmap(drive0.node, '', recording=True,
 367                                             bitmaps=bitmaps))
 368
 369             # Simulate the order in which this will happen:
 370             # group 1 gets cleared first, then group two gets written.
 371             if ((bsync_mode == 'on-success' and not failure) or
 372                 (bsync_mode == 'always')):
 373                 ebitmap.clear()
 374             ebitmap.dirty_group(2)
 375
 376         vm.run_job(job, auto_dismiss=True, auto_finalize=False,
 377                    pre_finalize=_callback,
 378                    cancel=(failure == 'simulated'))
 379         bitmaps = vm.query_bitmaps()
 380         log({'bitmaps': bitmaps}, indent=2)
 381         log('')
 382
 383         if bsync_mode == 'always' and failure == 'intermediate':
 384             # TOP treats anything allocated as dirty, expect to see:
 385             if msync_mode == 'top':
 386                 ebitmap.dirty_group(0)
 387
 388             # We manage to copy one sector (one bit) before the error.
 389             ebitmap.clear_bit(ebitmap.first_bit)
 390
 391             # Full returns all bits set except what was copied/skipped
 392             if msync_mode == 'full':
 393                 fail_bit = ebitmap.first_bit
 394                 ebitmap.clear()
 395                 ebitmap.dirty_bits(range(fail_bit, SIZE // GRANULARITY))
 396
 397         ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps))
 398
 399         # 2 - Writes and Reference Backup
 400         bitmaps = perform_writes(drive0, 3)
 401         ebitmap.dirty_group(3)
 402         ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps))
 403         reference_backup(drive0, 2, fbackup2)
 404
 405         # 2 - Bitmap Backup (In failure modes, this is a recovery.)
 406         job = backup(drive0, 2, bsync2, "bitmap",
 407                      bitmap="bitmap0", bitmap_mode=bsync_mode)
 408         vm.run_job(job, auto_dismiss=True, auto_finalize=False)
 409         bitmaps = vm.query_bitmaps()
 410         log({'bitmaps': bitmaps}, indent=2)
 411         log('')
 412         if bsync_mode != 'never':
 413             ebitmap.clear()
 414         ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps))
 415
 416         log('--- Cleanup ---\n')
 417         vm.qmp_log("block-dirty-bitmap-remove",
 418                    node=drive0.node, name="bitmap0")
 419         bitmaps = vm.query_bitmaps()
 420         log({'bitmaps': bitmaps}, indent=2)
 421         vm.shutdown()
 422         log('')
 423
 424         log('--- Verification ---\n')
 425         # 'simulated' failures will actually all pass here because we canceled
 426         # while "pending". This is actually undefined behavior,
 427         # don't rely on this to be true!
 428         compare_images(bsync1, fbackup1, baseimg=fbackup0,
 429                        expected_match=failure != 'intermediate')
 430         if not failure or bsync_mode == 'always':
 431             # Always keep the last backup on success or when using 'always'
 432             base = bsync1
 433         else:
 434             base = fbackup0
 435         compare_images(bsync2, fbackup2, baseimg=base)
 436         compare_images(img_path, fbackup2)
 437         log('')
 438
 439 def test_backup_api():
 440     """
 441     Test malformed and prohibited invocations of the backup API.
 442     """
 443     with iotests.FilePaths(['img', 'bsync1']) as \
 444          (img_path, backup_path), \
 445          iotests.VM() as vm:
 446
 447         log("\n=== API failure tests ===\n")
 448         log('--- Preparing image & VM ---\n')
 449         drive0 = Drive(img_path, vm=vm)
 450         drive0.img_create(iotests.imgfmt, SIZE)
 451         vm.add_device("{},id=scsi0".format(iotests.get_virtio_scsi_device()))
 452         vm.launch()
 453
 454         file_config = {
 455             'driver': 'file',
 456             'filename': drive0.path
 457         }
 458
 459         drive0.node = 'drive0'
 460         vm.qmp_log('blockdev-add',
 461                    filters=[iotests.filter_qmp_testfiles],
 462                    node_name=drive0.node,
 463                    driver=drive0.fmt,
 464                    file=file_config)
 465         log('')
 466
 467         target0 = Drive(backup_path, vm=vm)
 468         target0.create_target("backup_target", drive0.fmt, drive0.size)
 469         log('')
 470
 471         vm.qmp_log("block-dirty-bitmap-add", node=drive0.node,
 472                    name="bitmap0", granularity=GRANULARITY)
 473         log('')
 474
 475         log('-- Testing invalid QMP commands --\n')
 476
 477         error_cases = {
 478             'incremental': {
 479                 None:        ['on-success', 'always', 'never', None],
 480                 'bitmap404': ['on-success', 'always', 'never', None],
 481                 'bitmap0':   ['always', 'never']
 482             },
 483             'bitmap': {
 484                 None:        ['on-success', 'always', 'never', None],
 485                 'bitmap404': ['on-success', 'always', 'never', None],
 486                 'bitmap0':   [None],
 487             },
 488             'full': {
 489                 None:        ['on-success', 'always', 'never'],
 490                 'bitmap404': ['on-success', 'always', 'never', None],
 491                 'bitmap0':   ['never', None],
 492             },
 493             'top': {
 494                 None:        ['on-success', 'always', 'never'],
 495                 'bitmap404': ['on-success', 'always', 'never', None],
 496                 'bitmap0':   ['never', None],
 497             },
 498             'none': {
 499                 None:        ['on-success', 'always', 'never'],
 500                 'bitmap404': ['on-success', 'always', 'never', None],
 501                 'bitmap0':   ['on-success', 'always', 'never', None],
 502             }
 503         }
 504
 505         # Dicts, as always, are not stably-ordered prior to 3.7, so use tuples:
 506         for sync_mode in ('incremental', 'bitmap', 'full', 'top', 'none'):
 507             log("-- Sync mode {:s} tests --\n".format(sync_mode))
 508             for bitmap in (None, 'bitmap404', 'bitmap0'):
 509                 for policy in error_cases[sync_mode][bitmap]:
 510                     blockdev_backup(drive0.vm, drive0.node, "backup_target",
 511                                     sync_mode, job_id='api_job',
 512                                     bitmap=bitmap, bitmap_mode=policy)
 513                     log('')
 514
 515
 516 def main():
 517     for bsync_mode in ("never", "on-success", "always"):
 518         for failure in ("simulated", "intermediate", None):
 519             test_bitmap_sync(bsync_mode, "bitmap", failure)
 520
 521     for sync_mode in ('full', 'top'):
 522         for bsync_mode in ('on-success', 'always'):
 523             for failure in ('simulated', 'intermediate', None):
 524                 test_bitmap_sync(bsync_mode, sync_mode, failure)
 525
 526     test_backup_api()
 527
 528 if __name__ == '__main__':
 529     iotests.script_main(main, supported_fmts=['qcow2'],
 530                         supported_protocols=['file'])