1#!/usr/bin/env python3 2# group: rw 3# 4# Test bitmap-sync backups (incremental, differential, and partials) 5# 6# Copyright (c) 2019 John Snow for Red Hat, Inc. 7# 8# This program is free software; you can redistribute it and/or modify 9# it under the terms of the GNU General Public License as published by 10# the Free Software Foundation; either version 2 of the License, or 11# (at your option) any later version. 12# 13# This program is distributed in the hope that it will be useful, 14# but WITHOUT ANY WARRANTY; without even the implied warranty of 15# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16# GNU General Public License for more details. 17# 18# You should have received a copy of the GNU General Public License 19# along with this program. If not, see <http://www.gnu.org/licenses/>. 20# 21# owner=jsnow@redhat.com 22 23import math 24import os 25 26import iotests 27from iotests import log, qemu_img 28 29SIZE = 64 * 1024 * 1024 30GRANULARITY = 64 * 1024 31 32 33class Pattern: 34 def __init__(self, byte, offset, size=GRANULARITY): 35 self.byte = byte 36 self.offset = offset 37 self.size = size 38 39 def bits(self, granularity): 40 lower = self.offset // granularity 41 upper = (self.offset + self.size - 1) // granularity 42 return set(range(lower, upper + 1)) 43 44 45class PatternGroup: 46 """Grouping of Pattern objects. Initialize with an iterable of Patterns.""" 47 def __init__(self, patterns): 48 self.patterns = patterns 49 50 def bits(self, granularity): 51 """Calculate the unique bits dirtied by this pattern grouping""" 52 res = set() 53 for pattern in self.patterns: 54 res |= pattern.bits(granularity) 55 return res 56 57 58GROUPS = [ 59 PatternGroup([ 60 # Batch 0: 4 clusters 61 Pattern('0x49', 0x0000000), 62 Pattern('0x6c', 0x0100000), # 1M 63 Pattern('0x6f', 0x2000000), # 32M 64 Pattern('0x76', 0x3ff0000)]), # 64M - 64K 65 PatternGroup([ 66 # Batch 1: 6 clusters (3 new) 67 Pattern('0x65', 0x0000000), # Full overwrite 68 Pattern('0x77', 0x00f8000), # Partial-left (1M-32K) 69 Pattern('0x72', 0x2008000), # Partial-right (32M+32K) 70 Pattern('0x69', 0x3fe0000)]), # Adjacent-left (64M - 128K) 71 PatternGroup([ 72 # Batch 2: 7 clusters (3 new) 73 Pattern('0x74', 0x0010000), # Adjacent-right 74 Pattern('0x69', 0x00e8000), # Partial-left (1M-96K) 75 Pattern('0x6e', 0x2018000), # Partial-right (32M+96K) 76 Pattern('0x67', 0x3fe0000, 77 2*GRANULARITY)]), # Overwrite [(64M-128K)-64M) 78 PatternGroup([ 79 # Batch 3: 8 clusters (5 new) 80 # Carefully chosen such that nothing re-dirties the one cluster 81 # that copies out successfully before failure in Group #1. 82 Pattern('0xaa', 0x0010000, 83 3*GRANULARITY), # Overwrite and 2x Adjacent-right 84 Pattern('0xbb', 0x00d8000), # Partial-left (1M-160K) 85 Pattern('0xcc', 0x2028000), # Partial-right (32M+160K) 86 Pattern('0xdd', 0x3fc0000)]), # New; leaving a gap to the right 87] 88 89 90class EmulatedBitmap: 91 def __init__(self, granularity=GRANULARITY): 92 self._bits = set() 93 self.granularity = granularity 94 95 def dirty_bits(self, bits): 96 self._bits |= set(bits) 97 98 def dirty_group(self, n): 99 self.dirty_bits(GROUPS[n].bits(self.granularity)) 100 101 def clear(self): 102 self._bits = set() 103 104 def clear_bits(self, bits): 105 self._bits -= set(bits) 106 107 def clear_bit(self, bit): 108 self.clear_bits({bit}) 109 110 def clear_group(self, n): 111 self.clear_bits(GROUPS[n].bits(self.granularity)) 112 113 @property 114 def first_bit(self): 115 return sorted(self.bits)[0] 116 117 @property 118 def bits(self): 119 return self._bits 120 121 @property 122 def count(self): 123 return len(self.bits) 124 125 def compare(self, qmp_bitmap): 126 """ 127 Print a nice human-readable message checking that a bitmap as reported 128 by the QMP interface has as many bits set as we expect it to. 129 """ 130 131 name = qmp_bitmap.get('name', '(anonymous)') 132 log("= Checking Bitmap {:s} =".format(name)) 133 134 want = self.count 135 have = qmp_bitmap['count'] // qmp_bitmap['granularity'] 136 137 log("expecting {:d} dirty sectors; have {:d}. {:s}".format( 138 want, have, "OK!" if want == have else "ERROR!")) 139 log('') 140 141 142class Drive: 143 """Represents, vaguely, a drive attached to a VM. 144 Includes format, graph, and device information.""" 145 146 def __init__(self, path, vm=None): 147 self.path = path 148 self.vm = vm 149 self.fmt = None 150 self.size = None 151 self.node = None 152 153 def img_create(self, fmt, size): 154 self.fmt = fmt 155 self.size = size 156 iotests.qemu_img_create('-f', self.fmt, self.path, str(self.size)) 157 158 def create_target(self, name, fmt, size): 159 basename = os.path.basename(self.path) 160 file_node_name = "file_{}".format(basename) 161 vm = self.vm 162 163 log(vm.command('blockdev-create', job_id='bdc-file-job', 164 options={ 165 'driver': 'file', 166 'filename': self.path, 167 'size': 0, 168 })) 169 vm.run_job('bdc-file-job') 170 log(vm.command('blockdev-add', driver='file', 171 node_name=file_node_name, filename=self.path)) 172 173 log(vm.command('blockdev-create', job_id='bdc-fmt-job', 174 options={ 175 'driver': fmt, 176 'file': file_node_name, 177 'size': size, 178 })) 179 vm.run_job('bdc-fmt-job') 180 log(vm.command('blockdev-add', driver=fmt, 181 node_name=name, 182 file=file_node_name)) 183 self.fmt = fmt 184 self.size = size 185 self.node = name 186 187def blockdev_backup(vm, device, target, sync, **kwargs): 188 # Strip any arguments explicitly nulled by the caller: 189 kwargs = {key: val for key, val in kwargs.items() if val is not None} 190 result = vm.qmp_log('blockdev-backup', 191 device=device, 192 target=target, 193 sync=sync, 194 filter_node_name='backup-top', 195 **kwargs) 196 return result 197 198def blockdev_backup_mktarget(drive, target_id, filepath, sync, **kwargs): 199 target_drive = Drive(filepath, vm=drive.vm) 200 target_drive.create_target(target_id, drive.fmt, drive.size) 201 blockdev_backup(drive.vm, drive.node, target_id, sync, **kwargs) 202 203def reference_backup(drive, n, filepath): 204 log("--- Reference Backup #{:d} ---\n".format(n)) 205 target_id = "ref_target_{:d}".format(n) 206 job_id = "ref_backup_{:d}".format(n) 207 blockdev_backup_mktarget(drive, target_id, filepath, "full", 208 job_id=job_id) 209 drive.vm.run_job(job_id, auto_dismiss=True) 210 log('') 211 212def backup(drive, n, filepath, sync, **kwargs): 213 log("--- Test Backup #{:d} ---\n".format(n)) 214 target_id = "backup_target_{:d}".format(n) 215 job_id = "backup_{:d}".format(n) 216 kwargs.setdefault('auto-finalize', False) 217 blockdev_backup_mktarget(drive, target_id, filepath, sync, 218 job_id=job_id, **kwargs) 219 return job_id 220 221def perform_writes(drive, n, filter_node_name=None): 222 log("--- Write #{:d} ---\n".format(n)) 223 for pattern in GROUPS[n].patterns: 224 cmd = "write -P{:s} 0x{:07x} 0x{:x}".format( 225 pattern.byte, 226 pattern.offset, 227 pattern.size) 228 log(cmd) 229 log(drive.vm.hmp_qemu_io(filter_node_name or drive.node, cmd)) 230 bitmaps = drive.vm.query_bitmaps() 231 log({'bitmaps': bitmaps}, indent=2) 232 log('') 233 return bitmaps 234 235 236def compare_images(image, reference, baseimg=None, expected_match=True): 237 """ 238 Print a nice human-readable message comparing these images. 239 """ 240 expected_ret = 0 if expected_match else 1 241 if baseimg: 242 assert qemu_img("rebase", "-u", "-b", baseimg, '-F', iotests.imgfmt, 243 image) == 0 244 ret = qemu_img("compare", image, reference) 245 log('qemu_img compare "{:s}" "{:s}" ==> {:s}, {:s}'.format( 246 image, reference, 247 "Identical" if ret == 0 else "Mismatch", 248 "OK!" if ret == expected_ret else "ERROR!"), 249 filters=[iotests.filter_testfiles]) 250 251def test_bitmap_sync(bsync_mode, msync_mode='bitmap', failure=None): 252 """ 253 Test bitmap backup routines. 254 255 :param bsync_mode: Is the Bitmap Sync mode, and can be any of: 256 - on-success: This is the "incremental" style mode. Bitmaps are 257 synchronized to what was copied out only on success. 258 (Partial images must be discarded.) 259 - never: This is the "differential" style mode. 260 Bitmaps are never synchronized. 261 - always: This is a "best effort" style mode. 262 Bitmaps are always synchronized, regardless of failure. 263 (Partial images must be kept.) 264 265 :param msync_mode: The mirror sync mode to use for the first backup. 266 Can be any one of: 267 - bitmap: Backups based on bitmap manifest. 268 - full: Full backups. 269 - top: Full backups of the top layer only. 270 271 :param failure: Is the (optional) failure mode, and can be any of: 272 - None: No failure. Test the normative path. Default. 273 - simulated: Cancel the job right before it completes. 274 This also tests writes "during" the job. 275 - intermediate: This tests a job that fails mid-process and produces 276 an incomplete backup. Testing limitations prevent 277 testing competing writes. 278 """ 279 with iotests.FilePath( 280 'img', 'bsync1', 'bsync2', 'fbackup0', 'fbackup1', 'fbackup2') as \ 281 (img_path, bsync1, bsync2, fbackup0, fbackup1, fbackup2), \ 282 iotests.VM() as vm: 283 284 mode = "Mode {:s}; Bitmap Sync {:s}".format(msync_mode, bsync_mode) 285 preposition = "with" if failure else "without" 286 cond = "{:s} {:s}".format(preposition, 287 "{:s} failure".format(failure) if failure 288 else "failure") 289 log("\n=== {:s} {:s} ===\n".format(mode, cond)) 290 291 log('--- Preparing image & VM ---\n') 292 drive0 = Drive(img_path, vm=vm) 293 drive0.img_create(iotests.imgfmt, SIZE) 294 vm.add_device("{},id=scsi0".format(iotests.get_virtio_scsi_device())) 295 vm.launch() 296 297 file_config = { 298 'driver': 'file', 299 'filename': drive0.path 300 } 301 302 if failure == 'intermediate': 303 file_config = { 304 'driver': 'blkdebug', 305 'image': file_config, 306 'set-state': [{ 307 'event': 'flush_to_disk', 308 'state': 1, 309 'new_state': 2 310 }, { 311 'event': 'read_aio', 312 'state': 2, 313 'new_state': 3 314 }], 315 'inject-error': [{ 316 'event': 'read_aio', 317 'errno': 5, 318 'state': 3, 319 'immediately': False, 320 'once': True 321 }] 322 } 323 324 drive0.node = 'drive0' 325 vm.qmp_log('blockdev-add', 326 filters=[iotests.filter_qmp_testfiles], 327 node_name=drive0.node, 328 driver=drive0.fmt, 329 file=file_config) 330 log('') 331 332 # 0 - Writes and Reference Backup 333 perform_writes(drive0, 0) 334 reference_backup(drive0, 0, fbackup0) 335 log('--- Add Bitmap ---\n') 336 vm.qmp_log("block-dirty-bitmap-add", node=drive0.node, 337 name="bitmap0", granularity=GRANULARITY) 338 log('') 339 ebitmap = EmulatedBitmap() 340 341 # 1 - Writes and Reference Backup 342 bitmaps = perform_writes(drive0, 1) 343 ebitmap.dirty_group(1) 344 bitmap = vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps) 345 ebitmap.compare(bitmap) 346 reference_backup(drive0, 1, fbackup1) 347 348 # 1 - Test Backup (w/ Optional induced failure) 349 if failure == 'intermediate': 350 # Activate blkdebug induced failure for second-to-next read 351 log(vm.hmp_qemu_io(drive0.node, 'flush')) 352 log('') 353 job = backup(drive0, 1, bsync1, msync_mode, 354 bitmap="bitmap0", bitmap_mode=bsync_mode) 355 356 def _callback(): 357 """Issue writes while the job is open to test bitmap divergence.""" 358 # Note: when `failure` is 'intermediate', this isn't called. 359 log('') 360 bitmaps = perform_writes(drive0, 2, filter_node_name='backup-top') 361 # Named bitmap (static, should be unchanged) 362 ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', 363 bitmaps=bitmaps)) 364 # Anonymous bitmap (dynamic, shows new writes) 365 anonymous = EmulatedBitmap() 366 anonymous.dirty_group(2) 367 anonymous.compare(vm.get_bitmap(drive0.node, '', recording=True, 368 bitmaps=bitmaps)) 369 370 # Simulate the order in which this will happen: 371 # group 1 gets cleared first, then group two gets written. 372 if ((bsync_mode == 'on-success' and not failure) or 373 (bsync_mode == 'always')): 374 ebitmap.clear() 375 ebitmap.dirty_group(2) 376 377 vm.run_job(job, auto_dismiss=True, auto_finalize=False, 378 pre_finalize=_callback, 379 cancel=(failure == 'simulated')) 380 bitmaps = vm.query_bitmaps() 381 log({'bitmaps': bitmaps}, indent=2) 382 log('') 383 384 if bsync_mode == 'always' and failure == 'intermediate': 385 # TOP treats anything allocated as dirty, expect to see: 386 if msync_mode == 'top': 387 ebitmap.dirty_group(0) 388 389 # We manage to copy one sector (one bit) before the error. 390 ebitmap.clear_bit(ebitmap.first_bit) 391 392 # Full returns all bits set except what was copied/skipped 393 if msync_mode == 'full': 394 fail_bit = ebitmap.first_bit 395 ebitmap.clear() 396 ebitmap.dirty_bits(range(fail_bit, SIZE // GRANULARITY)) 397 398 ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps)) 399 400 # 2 - Writes and Reference Backup 401 bitmaps = perform_writes(drive0, 3) 402 ebitmap.dirty_group(3) 403 ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps)) 404 reference_backup(drive0, 2, fbackup2) 405 406 # 2 - Bitmap Backup (In failure modes, this is a recovery.) 407 job = backup(drive0, 2, bsync2, "bitmap", 408 bitmap="bitmap0", bitmap_mode=bsync_mode) 409 vm.run_job(job, auto_dismiss=True, auto_finalize=False) 410 bitmaps = vm.query_bitmaps() 411 log({'bitmaps': bitmaps}, indent=2) 412 log('') 413 if bsync_mode != 'never': 414 ebitmap.clear() 415 ebitmap.compare(vm.get_bitmap(drive0.node, 'bitmap0', bitmaps=bitmaps)) 416 417 log('--- Cleanup ---\n') 418 vm.qmp_log("block-dirty-bitmap-remove", 419 node=drive0.node, name="bitmap0") 420 bitmaps = vm.query_bitmaps() 421 log({'bitmaps': bitmaps}, indent=2) 422 vm.shutdown() 423 log('') 424 425 log('--- Verification ---\n') 426 # 'simulated' failures will actually all pass here because we canceled 427 # while "pending". This is actually undefined behavior, 428 # don't rely on this to be true! 429 compare_images(bsync1, fbackup1, baseimg=fbackup0, 430 expected_match=failure != 'intermediate') 431 if not failure or bsync_mode == 'always': 432 # Always keep the last backup on success or when using 'always' 433 base = bsync1 434 else: 435 base = fbackup0 436 compare_images(bsync2, fbackup2, baseimg=base) 437 compare_images(img_path, fbackup2) 438 log('') 439 440def test_backup_api(): 441 """ 442 Test malformed and prohibited invocations of the backup API. 443 """ 444 with iotests.FilePath('img', 'bsync1') as (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 516def 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 528if __name__ == '__main__': 529 iotests.script_main(main, supported_fmts=['qcow2'], 530 supported_protocols=['file']) 531