tests/qemu-iotests/060

   1 #!/usr/bin/env bash
   2 #
   3 # Test case for image corruption (overlapping data structures) in qcow2
   4 #
   5 # Copyright (C) 2013 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
  21 # creator
  22 owner=mreitz@redhat.com
  23
  24 seq="$(basename $0)"
  25 echo "QA output created by $seq"
  26
  27 status=1        # failure is the default!
  28
  29 _cleanup()
  30 {
  31         _cleanup_test_img
  32 }
  33 trap "_cleanup; exit \$status" 0 1 2 3 15
  34
  35 # Sometimes the error line might be dumped before/after an event
  36 # randomly.  Mask it out for specific test that may trigger this
  37 # uncertainty for current test for now.
  38 _filter_io_error()
  39 {
  40     sed '/Input\/output error/d'
  41 }
  42
  43 # get standard environment, filters and checks
  44 . ./common.rc
  45 . ./common.filter
  46
  47 # This tests qcow2-specific low-level functionality
  48 _supported_fmt qcow2
  49 _supported_proto file
  50 _supported_os Linux
  51 # These tests only work for compat=1.1 images without an external
  52 # data file with refcount_bits=16
  53 _unsupported_imgopts 'compat=0.10' data_file \
  54     'refcount_bits=\([^1]\|.\([^6]\|$\)\)'
  55
  56 # The repair process will create a large file - so check for availability first
  57 _require_large_file 64G
  58
  59 rt_offset=65536  # 0x10000 (XXX: just an assumption)
  60 rb_offset=131072 # 0x20000 (XXX: just an assumption)
  61 l1_offset=196608 # 0x30000 (XXX: just an assumption)
  62 l2_offset=262144 # 0x40000 (XXX: just an assumption)
  63 l2_offset_after_snapshot=524288 # 0x80000 (XXX: just an assumption)
  64
  65 OPEN_RW="open -o overlap-check=all $TEST_IMG"
  66 # Overlap checks are done before write operations only, therefore opening an
  67 # image read-only makes the overlap-check option irrelevant
  68 OPEN_RO="open -r $TEST_IMG"
  69
  70 echo
  71 echo "=== Testing L2 reference into L1 ==="
  72 echo
  73 _make_test_img 64M
  74 # Link first L1 entry (first L2 table) onto itself
  75 # (Note the MSb in the L1 entry is set, ensuring the refcount is one - else any
  76 # later write will result in a COW operation, effectively ruining this attempt
  77 # on image corruption)
  78 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x03\x00\x00"
  79 _check_test_img
  80
  81 # The corrupt bit should not be set anyway
  82 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
  83
  84 # Try to write something, thereby forcing the corrupt bit to be set
  85 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
  86
  87 # The corrupt bit must now be set
  88 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
  89
  90 # This information should be available through qemu-img info
  91 _img_info --format-specific
  92
  93 # Try to open the image R/W (which should fail)
  94 $QEMU_IO -c "$OPEN_RW" -c "read 0 512" 2>&1 | _filter_qemu_io \
  95                                             | _filter_testdir \
  96                                             | _filter_imgfmt
  97
  98 # Try to open it RO (which should succeed)
  99 $QEMU_IO -c "$OPEN_RO" -c "read 0 512" | _filter_qemu_io
 100
 101 # We could now try to fix the image, but this would probably fail (how should an
 102 # L2 table linked onto the L1 table be fixed?)
 103
 104 echo
 105 echo "=== Testing cluster data reference into refcount block ==="
 106 echo
 107 _make_test_img 64M
 108 # Allocate L2 table
 109 truncate -s "$(($l2_offset+65536))" "$TEST_IMG"
 110 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x00\x00"
 111 # Mark cluster as used
 112 poke_file "$TEST_IMG" "$(($rb_offset+8))" "\x00\x01"
 113 # Redirect new data cluster onto refcount block
 114 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x02\x00\x00"
 115 _check_test_img
 116 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 117 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
 118 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 119
 120 # Try to fix it
 121 _check_test_img -r all
 122
 123 # The corrupt bit should be cleared
 124 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 125
 126 # Look if it's really really fixed
 127 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
 128 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 129
 130 echo
 131 echo "=== Testing cluster data reference into inactive L2 table ==="
 132 echo
 133 _make_test_img 64M
 134 $QEMU_IO -c "$OPEN_RW" -c "write -P 1 0 512" | _filter_qemu_io
 135 $QEMU_IMG snapshot -c foo "$TEST_IMG"
 136 $QEMU_IO -c "$OPEN_RW" -c "write -P 2 0 512" | _filter_qemu_io
 137 # The inactive L2 table remains at its old offset
 138 poke_file "$TEST_IMG" "$l2_offset_after_snapshot" \
 139                       "\x80\x00\x00\x00\x00\x04\x00\x00"
 140 _check_test_img
 141 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 142 $QEMU_IO -c "$OPEN_RW" -c "write -P 3 0 512" | _filter_qemu_io
 143 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 144 _check_test_img -r all
 145 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 146 $QEMU_IO -c "$OPEN_RW" -c "write -P 4 0 512" | _filter_qemu_io
 147 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 148
 149 # Check data
 150 $QEMU_IO -c "$OPEN_RO" -c "read -P 4 0 512" | _filter_qemu_io
 151 $QEMU_IMG snapshot -a foo "$TEST_IMG"
 152 _check_test_img
 153 $QEMU_IO -c "$OPEN_RO" -c "read -P 1 0 512" | _filter_qemu_io
 154
 155 echo
 156 echo "=== Testing overlap while COW is in flight ==="
 157 echo
 158 BACKING_IMG=$TEST_IMG.base
 159 TEST_IMG=$BACKING_IMG _make_test_img 1G
 160
 161 $QEMU_IO -c 'write 0k 64k' "$BACKING_IMG" | _filter_qemu_io
 162
 163 # compat=0.10 is required in order to make the following discard actually
 164 # unallocate the sector rather than make it a zero sector - we want COW, after
 165 # all.
 166 _make_test_img -o 'compat=0.10' -b "$BACKING_IMG" 1G
 167 # Write two clusters, the second one enforces creation of an L2 table after
 168 # the first data cluster.
 169 $QEMU_IO -c 'write 0k 64k' -c 'write 512M 64k' "$TEST_IMG" | _filter_qemu_io
 170 # Discard the first cluster. This cluster will soon enough be reallocated and
 171 # used for COW.
 172 $QEMU_IO -c 'discard 0k 64k' "$TEST_IMG" | _filter_qemu_io
 173 # Now, corrupt the image by marking the second L2 table cluster as free.
 174 poke_file "$TEST_IMG" '131084' "\x00\x00" # 0x2000c
 175 # Start a write operation requiring COW on the image stopping it right before
 176 # doing the read; then, trigger the corruption prevention by writing anything to
 177 # any unallocated cluster, leading to an attempt to overwrite the second L2
 178 # table. Finally, resume the COW write and see it fail (but not crash).
 179 echo "open -o file.driver=blkdebug $TEST_IMG
 180 break cow_read 0
 181 aio_write 0k 1k
 182 wait_break 0
 183 write 64k 64k
 184 resume 0" | $QEMU_IO | _filter_qemu_io
 185
 186 echo
 187 echo "=== Testing unallocated image header ==="
 188 echo
 189 _make_test_img 64M
 190 # Create L1/L2
 191 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 192 poke_file "$TEST_IMG" "$rb_offset" "\x00\x00"
 193 $QEMU_IO -c "write 64k 64k" "$TEST_IMG" | _filter_qemu_io
 194
 195 echo
 196 echo "=== Testing unaligned L1 entry ==="
 197 echo
 198 _make_test_img 64M
 199 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 200 # This will be masked with ~(512 - 1) = ~0x1ff, so whether the lower 9 bits are
 201 # aligned or not does not matter
 202 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x2a\x00"
 203 $QEMU_IO -c "read 0 64k" "$TEST_IMG" | _filter_qemu_io
 204
 205 # Test how well zero cluster expansion can cope with this
 206 _make_test_img 64M
 207 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 208 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x2a\x00"
 209 $QEMU_IMG amend -o compat=0.10 "$TEST_IMG"
 210
 211 echo
 212 echo "=== Testing unaligned L2 entry ==="
 213 echo
 214 _make_test_img 64M
 215 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 216 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 217 $QEMU_IO -c "read 0 64k" "$TEST_IMG" | _filter_qemu_io
 218
 219 echo
 220 echo "=== Testing unaligned pre-allocated zero cluster ==="
 221 echo
 222 _make_test_img 64M
 223 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 224 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x01"
 225 # zero cluster expansion
 226 $QEMU_IMG amend -o compat=0.10 "$TEST_IMG"
 227
 228 echo
 229 echo "=== Testing unaligned reftable entry ==="
 230 echo
 231 _make_test_img 64M
 232 poke_file "$TEST_IMG" "$rt_offset" "\x00\x00\x00\x00\x00\x02\x2a\x00"
 233 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 234
 235 echo
 236 echo "=== Testing non-fatal corruption on freeing ==="
 237 echo
 238 _make_test_img 64M
 239 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 240 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 241 $QEMU_IO -c "discard 0 64k" "$TEST_IMG" | _filter_qemu_io
 242
 243 echo
 244 echo "=== Testing read-only corruption report ==="
 245 echo
 246 _make_test_img 64M
 247 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 248 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 249 # Should only emit a single error message
 250 $QEMU_IO -c "$OPEN_RO" -c "read 0 64k" -c "read 0 64k" | _filter_qemu_io
 251
 252 echo
 253 echo "=== Testing non-fatal and then fatal corruption report ==="
 254 echo
 255 _make_test_img 64M
 256 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 257 poke_file "$TEST_IMG" "$l2_offset"        "\x80\x00\x00\x00\x00\x05\x2a\x00"
 258 poke_file "$TEST_IMG" "$(($l2_offset+8))" "\x80\x00\x00\x00\x00\x06\x2a\x00"
 259 # Should emit two error messages
 260 $QEMU_IO -c "discard 0 64k" -c "read 64k 64k" "$TEST_IMG" | _filter_qemu_io
 261
 262 echo
 263 echo "=== Testing empty refcount table ==="
 264 echo
 265 _make_test_img 64M
 266 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 267 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 268 # Repair the image
 269 _check_test_img -r all
 270
 271 echo
 272 echo "=== Testing empty refcount table with valid L1 and L2 tables ==="
 273 echo
 274 _make_test_img 64M
 275 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 276 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 277 # Since the first data cluster is already allocated this triggers an
 278 # allocation with an explicit offset (using qcow2_alloc_clusters_at())
 279 # causing a refcount block to be allocated at offset 0
 280 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 281 # Repair the image
 282 _check_test_img -r all
 283
 284 echo
 285 echo "=== Testing empty refcount block ==="
 286 echo
 287 _make_test_img 64M
 288 poke_file "$TEST_IMG" "$rb_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 289 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 290 # Repair the image
 291 _check_test_img -r all
 292
 293 echo
 294 echo "=== Testing empty refcount block with compressed write ==="
 295 echo
 296 _make_test_img 64M
 297 $QEMU_IO -c "write 64k 64k" "$TEST_IMG" | _filter_qemu_io
 298 poke_file "$TEST_IMG" "$rb_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 299 # The previous write already allocated an L2 table, so now this new
 300 # write will try to allocate a compressed data cluster at offset 0.
 301 $QEMU_IO -c "write -c 0k 64k" "$TEST_IMG" | _filter_qemu_io
 302 # Repair the image
 303 _check_test_img -r all
 304
 305 echo
 306 echo "=== Testing zero refcount table size ==="
 307 echo
 308 _make_test_img 64M
 309 poke_file "$TEST_IMG" "56"                "\x00\x00\x00\x00"
 310 $QEMU_IO -c "write 0 64k" "$TEST_IMG" 2>&1 | _filter_testdir | _filter_imgfmt
 311 # Repair the image
 312 _check_test_img -r all
 313
 314 echo
 315 echo "=== Testing incorrect refcount table offset ==="
 316 echo
 317 _make_test_img 64M
 318 poke_file "$TEST_IMG" "48"                "\x00\x00\x00\x00\x00\x00\x00\x00"
 319 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 320
 321 echo
 322 echo "=== Testing dirty corrupt image ==="
 323 echo
 324
 325 _make_test_img 64M
 326
 327 # Let the refblock appear unaligned
 328 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\xff\xff\x2a\x00"
 329 # Mark the image dirty, thus forcing an automatic check when opening it
 330 poke_file "$TEST_IMG" 72 "\x00\x00\x00\x00\x00\x00\x00\x01"
 331 # Open the image (qemu should refuse to do so)
 332 $QEMU_IO -c close "$TEST_IMG" 2>&1 | _filter_testdir | _filter_imgfmt
 333
 334 echo '--- Repairing ---'
 335
 336 # The actual repair should have happened (because of the dirty bit),
 337 # but some cleanup may have failed (like freeing the old reftable)
 338 # because the image was already marked corrupt by that point
 339 _check_test_img -r all
 340
 341 echo
 342 echo "=== Writing to an unaligned preallocated zero cluster ==="
 343 echo
 344
 345 _make_test_img 64M
 346
 347 # Allocate the L2 table
 348 $QEMU_IO -c "write 0 64k" -c "discard 0 64k" "$TEST_IMG" | _filter_qemu_io
 349 # Pretend there is a preallocated zero cluster somewhere inside the
 350 # image header
 351 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x00\x2a\x01"
 352 # Let's write to it!
 353 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 354
 355 echo '--- Repairing ---'
 356 _check_test_img -r all
 357
 358 echo
 359 echo '=== Discarding with an unaligned refblock ==='
 360 echo
 361
 362 _make_test_img 64M
 363
 364 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 365 # Make our refblock unaligned
 366 poke_file "$TEST_IMG" "$(($rt_offset))" "\x00\x00\x00\x00\x00\x00\x2a\x00"
 367 # Now try to discard something that will be submitted as two requests
 368 # (main part + tail)
 369 $QEMU_IO -c "discard 0 65537" "$TEST_IMG"
 370
 371 echo '--- Repairing ---'
 372 # Fails the first repair because the corruption prevents the check
 373 # function from double-checking
 374 # (Using -q for the first invocation, because otherwise the
 375 #  double-check error message appears above the summary for some
 376 #  reason -- so let's just hide the summary)
 377 _check_test_img -q -r all
 378 _check_test_img -r all
 379
 380 echo
 381 echo "=== Discarding an out-of-bounds refblock ==="
 382 echo
 383
 384 _make_test_img 64M
 385
 386 # Pretend there's a refblock really up high
 387 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\xff\xff\xff\x00\x00\x00\x00"
 388 # Let's try to shrink the qcow2 image so that the block driver tries
 389 # to discard that refblock (and see what happens!)
 390 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 391
 392 echo '--- Checking and retrying ---'
 393 # Image should not be resized
 394 _img_info | grep 'virtual size'
 395 # But it should pass this check, because the "partial" resize has
 396 # already overwritten refblocks past the end
 397 _check_test_img -r all
 398 # So let's try again
 399 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 400 _img_info | grep 'virtual size'
 401
 402 echo
 403 echo "=== Discarding a non-covered in-bounds refblock ==="
 404 echo
 405
 406 _make_test_img -o 'refcount_bits=1' 64M
 407
 408 # Pretend there's a refblock somewhere where there is no refblock to
 409 # cover it (but the covering refblock has a valid index in the
 410 # reftable)
 411 # Every refblock covers 65536 * 8 * 65536 = 32 GB, so we have to point
 412 # to 0x10_0000_0000 (64G) to point to the third refblock
 413 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\x00\x00\x10\x00\x00\x00\x00"
 414 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 415
 416 echo '--- Checking and retrying ---'
 417 # Image should not be resized
 418 _img_info | grep 'virtual size'
 419 # But it should pass this check, because the "partial" resize has
 420 # already overwritten refblocks past the end
 421 _check_test_img -r all
 422 # So let's try again
 423 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 424 _img_info | grep 'virtual size'
 425
 426 echo
 427 echo "=== Discarding a refblock covered by an unaligned refblock ==="
 428 echo
 429
 430 _make_test_img -o 'refcount_bits=1' 64M
 431
 432 # Same as above
 433 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\x00\x00\x10\x00\x00\x00\x00"
 434 # But now we actually "create" an unaligned third refblock
 435 poke_file "$TEST_IMG" "$(($rt_offset+16))" "\x00\x00\x00\x00\x00\x00\x02\x00"
 436 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 437
 438 echo '--- Repairing ---'
 439 # Fails the first repair because the corruption prevents the check
 440 # function from double-checking
 441 # (Using -q for the first invocation, because otherwise the
 442 #  double-check error message appears above the summary for some
 443 #  reason -- so let's just hide the summary)
 444 _check_test_img -q -r all
 445 _check_test_img -r all
 446
 447 echo
 448 echo "=== Testing the QEMU shutdown with a corrupted image ==="
 449 echo
 450 _make_test_img 64M
 451 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 452 echo "{'execute': 'qmp_capabilities'}
 453       {'execute': 'human-monitor-command',
 454        'arguments': {'command-line': 'qemu-io drive \"write 0 512\"'}}
 455       {'execute': 'quit'}" \
 456     | $QEMU -qmp stdio -nographic -nodefaults \
 457             -drive if=none,node-name=drive,file="$TEST_IMG",driver=qcow2 \
 458     | _filter_qmp | _filter_qemu_io
 459
 460 echo
 461 echo "=== Testing incoming inactive corrupted image ==="
 462 echo
 463
 464 _make_test_img 64M
 465 # Create an unaligned L1 entry, so qemu will signal a corruption when
 466 # reading from the covered area
 467 poke_file "$TEST_IMG" "$l1_offset" "\x00\x00\x00\x00\x2a\x2a\x2a\x2a"
 468
 469 # Inactive images are effectively read-only images, so this should be a
 470 # non-fatal corruption (which does not modify the image)
 471 echo "{'execute': 'qmp_capabilities'}
 472       {'execute': 'human-monitor-command',
 473        'arguments': {'command-line': 'qemu-io drive \"read 0 512\"'}}
 474       {'execute': 'quit'}" \
 475     | $QEMU -qmp stdio -nographic -nodefaults \
 476             -blockdev "{'node-name': 'drive',
 477                         'driver': 'qcow2',
 478                         'file': {
 479                             'driver': 'file',
 480                             'filename': '$TEST_IMG'
 481                         }}" \
 482             -incoming exec:'cat /dev/null' \
 483             2>&1 \
 484     | _filter_qmp | _filter_qemu_io | _filter_io_error
 485
 486 echo
 487 # Image should not have been marked corrupt
 488 _img_info --format-specific | grep 'corrupt:'
 489
 490 # success, all done
 491 echo "*** done"
 492 rm -f $seq.full
 493 status=0