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 class Drive:
  89     """Represents, vaguely, a drive attached to a VM.
  90     Includes format, graph, and device information."""
  91
  92     def __init__(self, path, vm=None):
  93         self.path = path
  94         self.vm = vm
  95         self.fmt = None
  96         self.size = None
  97         self.node = None
  98         self.device = None
  99
 100     @property
 101     def name(self):
 102         return self.node or self.device
 103
 104     def img_create(self, fmt, size):
 105         self.fmt = fmt
 106         self.size = size
 107         iotests.qemu_img_create('-f', self.fmt, self.path, str(self.size))
 108
 109     def create_target(self, name, fmt, size):
 110         basename = os.path.basename(self.path)
 111         file_node_name = "file_{}".format(basename)
 112         vm = self.vm
 113
 114         log(vm.command('blockdev-create', job_id='bdc-file-job',
 115                        options={
 116                            'driver': 'file',
 117                            'filename': self.path,
 118                            'size': 0,
 119                        }))
 120         vm.run_job('bdc-file-job')
 121         log(vm.command('blockdev-add', driver='file',
 122                        node_name=file_node_name, filename=self.path))
 123
 124         log(vm.command('blockdev-create', job_id='bdc-fmt-job',
 125                        options={
 126                            'driver': fmt,
 127                            'file': file_node_name,
 128                            'size': size,
 129                        }))
 130         vm.run_job('bdc-fmt-job')
 131         log(vm.command('blockdev-add', driver=fmt,
 132                        node_name=name,
 133                        file=file_node_name))
 134         self.fmt = fmt
 135         self.size = size
 136         self.node = name
 137
 138 def query_bitmaps(vm):
 139     res = vm.qmp("query-block")
 140     return {"bitmaps": {device['device'] or device['qdev']:
 141                         device.get('dirty-bitmaps', []) for
 142                         device in res['return']}}
 143
 144 def get_bitmap(bitmaps, drivename, name, recording=None):
 145     """
 146     get a specific bitmap from the object returned by query_bitmaps.
 147     :param recording: If specified, filter results by the specified value.
 148     """
 149     for bitmap in bitmaps['bitmaps'][drivename]:
 150         if bitmap.get('name', '') == name:
 151             if recording is None:
 152                 return bitmap
 153             elif bitmap.get('recording') == recording:
 154                 return bitmap
 155     return None
 156
 157 def reference_backup(drive, n, filepath):
 158     log("--- Reference Backup #{:d} ---\n".format(n))
 159     target_id = "ref_target_{:d}".format(n)
 160     job_id = "ref_backup_{:d}".format(n)
 161     target_drive = Drive(filepath, vm=drive.vm)
 162
 163     target_drive.create_target(target_id, drive.fmt, drive.size)
 164     drive.vm.qmp_log("blockdev-backup",
 165                      job_id=job_id, device=drive.name,
 166                      target=target_id, sync="full")
 167     drive.vm.run_job(job_id, auto_dismiss=True)
 168     log('')
 169
 170 def bitmap_backup(drive, n, filepath, bitmap, bitmap_mode):
 171     log("--- Bitmap Backup #{:d} ---\n".format(n))
 172     target_id = "bitmap_target_{:d}".format(n)
 173     job_id = "bitmap_backup_{:d}".format(n)
 174     target_drive = Drive(filepath, vm=drive.vm)
 175
 176     target_drive.create_target(target_id, drive.fmt, drive.size)
 177     drive.vm.qmp_log("blockdev-backup", job_id=job_id, device=drive.name,
 178                      target=target_id, sync="bitmap",
 179                      bitmap_mode=bitmap_mode,
 180                      bitmap=bitmap,
 181                      auto_finalize=False)
 182     return job_id
 183
 184 def perform_writes(drive, n):
 185     log("--- Write #{:d} ---\n".format(n))
 186     for pattern in GROUPS[n].patterns:
 187         cmd = "write -P{:s} 0x{:07x} 0x{:x}".format(
 188             pattern.byte,
 189             pattern.offset,
 190             pattern.size)
 191         log(cmd)
 192         log(drive.vm.hmp_qemu_io(drive.name, cmd))
 193     bitmaps = query_bitmaps(drive.vm)
 194     log(bitmaps, indent=2)
 195     log('')
 196     return bitmaps
 197
 198 def calculate_bits(groups=None):
 199     """Calculate how many bits we expect to see dirtied."""
 200     if groups:
 201         bits = set.union(*(GROUPS[group].bits(GRANULARITY) for group in groups))
 202         return len(bits)
 203     return 0
 204
 205 def bitmap_comparison(bitmap, groups=None, want=0):
 206     """
 207     Print a nice human-readable message checking that this bitmap has as
 208     many bits set as we expect it to.
 209     """
 210     log("= Checking Bitmap {:s} =".format(bitmap.get('name', '(anonymous)')))
 211
 212     if groups:
 213         want = calculate_bits(groups)
 214     have = bitmap['count'] // bitmap['granularity']
 215
 216     log("expecting {:d} dirty sectors; have {:d}. {:s}".format(
 217         want, have, "OK!" if want == have else "ERROR!"))
 218     log('')
 219
 220 def compare_images(image, reference, baseimg=None, expected_match=True):
 221     """
 222     Print a nice human-readable message comparing these images.
 223     """
 224     expected_ret = 0 if expected_match else 1
 225     if baseimg:
 226         assert qemu_img("rebase", "-u", "-b", baseimg, image) == 0
 227     ret = qemu_img("compare", image, reference)
 228     log('qemu_img compare "{:s}" "{:s}" ==> {:s}, {:s}'.format(
 229         image, reference,
 230         "Identical" if ret == 0 else "Mismatch",
 231         "OK!" if ret == expected_ret else "ERROR!"),
 232         filters=[iotests.filter_testfiles])
 233
 234 def test_bitmap_sync(bsync_mode, failure=None):
 235     """
 236     Test bitmap backup routines.
 237
 238     :param bsync_mode: Is the Bitmap Sync mode, and can be any of:
 239         - on-success: This is the "incremental" style mode. Bitmaps are
 240                       synchronized to what was copied out only on success.
 241                       (Partial images must be discarded.)
 242         - never:      This is the "differential" style mode.
 243                       Bitmaps are never synchronized.
 244         - always:     This is a "best effort" style mode.
 245                       Bitmaps are always synchronized, regardless of failure.
 246                       (Partial images must be kept.)
 247
 248     :param failure: Is the (optional) failure mode, and can be any of:
 249         - None:         No failure. Test the normative path. Default.
 250         - simulated:    Cancel the job right before it completes.
 251                         This also tests writes "during" the job.
 252         - intermediate: This tests a job that fails mid-process and produces
 253                         an incomplete backup. Testing limitations prevent
 254                         testing competing writes.
 255     """
 256     with iotests.FilePaths(['img', 'bsync1', 'bsync2',
 257                             'fbackup0', 'fbackup1', 'fbackup2']) as \
 258                             (img_path, bsync1, bsync2,
 259                              fbackup0, fbackup1, fbackup2), \
 260          iotests.VM() as vm:
 261
 262         mode = "Bitmap Sync Mode {:s}".format(bsync_mode)
 263         preposition = "with" if failure else "without"
 264         cond = "{:s} {:s}".format(preposition,
 265                                   "{:s} failure".format(failure) if failure
 266                                   else "failure")
 267         log("\n=== {:s} {:s} ===\n".format(mode, cond))
 268
 269         log('--- Preparing image & VM ---\n')
 270         drive0 = Drive(img_path, vm=vm)
 271         drive0.img_create(iotests.imgfmt, SIZE)
 272         vm.add_device("{},id=scsi0".format(iotests.get_virtio_scsi_device()))
 273         vm.launch()
 274
 275         file_config = {
 276             'driver': 'file',
 277             'filename': drive0.path
 278         }
 279
 280         if failure == 'intermediate':
 281             file_config = {
 282                 'driver': 'blkdebug',
 283                 'image': file_config,
 284                 'set-state': [{
 285                     'event': 'flush_to_disk',
 286                     'state': 1,
 287                     'new_state': 2
 288                 }, {
 289                     'event': 'read_aio',
 290                     'state': 2,
 291                     'new_state': 3
 292                 }],
 293                 'inject-error': [{
 294                     'event': 'read_aio',
 295                     'errno': 5,
 296                     'state': 3,
 297                     'immediately': False,
 298                     'once': True
 299                 }]
 300             }
 301
 302         vm.qmp_log('blockdev-add',
 303                    filters=[iotests.filter_qmp_testfiles],
 304                    node_name="drive0",
 305                    driver=drive0.fmt,
 306                    file=file_config)
 307         drive0.node = 'drive0'
 308         drive0.device = 'device0'
 309         # Use share-rw to allow writes directly to the node;
 310         # The anonymous block-backend for this configuration prevents us
 311         # from using HMP's qemu-io commands to address the device.
 312         vm.qmp_log("device_add", id=drive0.device,
 313                    drive=drive0.name, driver="scsi-hd",
 314                    share_rw=True)
 315         log('')
 316
 317         # 0 - Writes and Reference Backup
 318         perform_writes(drive0, 0)
 319         reference_backup(drive0, 0, fbackup0)
 320         log('--- Add Bitmap ---\n')
 321         vm.qmp_log("block-dirty-bitmap-add", node=drive0.name,
 322                    name="bitmap0", granularity=GRANULARITY)
 323         log('')
 324
 325         # 1 - Writes and Reference Backup
 326         bitmaps = perform_writes(drive0, 1)
 327         dirty_groups = {1}
 328         bitmap = get_bitmap(bitmaps, drive0.device, 'bitmap0')
 329         bitmap_comparison(bitmap, groups=dirty_groups)
 330         reference_backup(drive0, 1, fbackup1)
 331
 332         # 1 - Bitmap Backup (Optional induced failure)
 333         if failure == 'intermediate':
 334             # Activate blkdebug induced failure for second-to-next read
 335             log(vm.hmp_qemu_io(drive0.name, 'flush'))
 336             log('')
 337         job = bitmap_backup(drive0, 1, bsync1, "bitmap0", bsync_mode)
 338
 339         def _callback():
 340             """Issue writes while the job is open to test bitmap divergence."""
 341             # Note: when `failure` is 'intermediate', this isn't called.
 342             log('')
 343             bitmaps = perform_writes(drive0, 2)
 344             # Named bitmap (static, should be unchanged)
 345             bitmap_comparison(get_bitmap(bitmaps, drive0.device, 'bitmap0'),
 346                               groups=dirty_groups)
 347             # Anonymous bitmap (dynamic, shows new writes)
 348             bitmap_comparison(get_bitmap(bitmaps, drive0.device, '',
 349                                          recording=True), groups={2})
 350             dirty_groups.add(2)
 351
 352         vm.run_job(job, auto_dismiss=True, auto_finalize=False,
 353                    pre_finalize=_callback,
 354                    cancel=(failure == 'simulated'))
 355         bitmaps = query_bitmaps(vm)
 356         bitmap = get_bitmap(bitmaps, drive0.device, 'bitmap0')
 357         log(bitmaps, indent=2)
 358         log('')
 359
 360         if ((bsync_mode == 'on-success' and not failure) or
 361                 (bsync_mode == 'always' and failure != 'intermediate')):
 362             dirty_groups.remove(1)
 363
 364         if bsync_mode == 'always' and failure == 'intermediate':
 365             # We manage to copy one sector (one bit) before the error.
 366             bitmap_comparison(bitmap,
 367                               want=calculate_bits(groups=dirty_groups) - 1)
 368         else:
 369             bitmap_comparison(bitmap, groups=dirty_groups)
 370
 371         # 2 - Writes and Reference Backup
 372         bitmaps = perform_writes(drive0, 3)
 373         dirty_groups.add(3)
 374         bitmap = get_bitmap(bitmaps, drive0.device, 'bitmap0')
 375         if bsync_mode == 'always' and failure == 'intermediate':
 376             # We're one bit short, still.
 377             bitmap_comparison(bitmap,
 378                               want=calculate_bits(groups=dirty_groups) - 1)
 379         else:
 380             bitmap_comparison(bitmap, groups=dirty_groups)
 381         reference_backup(drive0, 2, fbackup2)
 382
 383         # 2 - Bitmap Backup (In failure modes, this is a recovery.)
 384         job = bitmap_backup(drive0, 2, bsync2, "bitmap0", bsync_mode)
 385         vm.run_job(job, auto_dismiss=True, auto_finalize=False)
 386         bitmaps = query_bitmaps(vm)
 387         bitmap = get_bitmap(bitmaps, drive0.device, 'bitmap0')
 388         log(bitmaps, indent=2)
 389         log('')
 390         bitmap_comparison(bitmap, groups={}
 391                           if bsync_mode != 'never'
 392                           else dirty_groups)
 393
 394         log('--- Cleanup ---\n')
 395         vm.qmp_log("block-dirty-bitmap-remove",
 396                    node=drive0.name, name="bitmap0")
 397         log(query_bitmaps(vm), indent=2)
 398         vm.shutdown()
 399         log('')
 400
 401         log('--- Verification ---\n')
 402         # 'simulated' failures will actually all pass here because we canceled
 403         # while "pending". This is actually undefined behavior,
 404         # don't rely on this to be true!
 405         compare_images(bsync1, fbackup1, baseimg=fbackup0,
 406                        expected_match=failure != 'intermediate')
 407         if not failure or bsync_mode == 'always':
 408             # Always keep the last backup on success or when using 'always'
 409             base = bsync1
 410         else:
 411             base = fbackup0
 412         compare_images(bsync2, fbackup2, baseimg=base)
 413         compare_images(img_path, fbackup2)
 414         log('')
 415
 416 def main():
 417     for bsync_mode in ("never", "on-success", "always"):
 418         for failure in ("simulated", "intermediate", None):
 419             test_bitmap_sync(bsync_mode, failure)
 420
 421 if __name__ == '__main__':
 422     iotests.script_main(main, supported_fmts=['qcow2'])