tests/qemu-iotests/060

   1 #!/bin/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 $seq"
  26
  27 here="$PWD"
  28 status=1        # failure is the default!
  29
  30 _cleanup()
  31 {
  32         _cleanup_test_img
  33 }
  34 trap "_cleanup; exit \$status" 0 1 2 3 15
  35
  36 # get standard environment, filters and checks
  37 . ./common.rc
  38 . ./common.filter
  39
  40 # This tests qocw2-specific low-level functionality
  41 _supported_fmt qcow2
  42 _supported_proto file
  43 _supported_os Linux
  44
  45 rt_offset=65536  # 0x10000 (XXX: just an assumption)
  46 rb_offset=131072 # 0x20000 (XXX: just an assumption)
  47 l1_offset=196608 # 0x30000 (XXX: just an assumption)
  48 l2_offset=262144 # 0x40000 (XXX: just an assumption)
  49 l2_offset_after_snapshot=524288 # 0x80000 (XXX: just an assumption)
  50
  51 IMGOPTS="compat=1.1"
  52
  53 OPEN_RW="open -o overlap-check=all $TEST_IMG"
  54 # Overlap checks are done before write operations only, therefore opening an
  55 # image read-only makes the overlap-check option irrelevant
  56 OPEN_RO="open -r $TEST_IMG"
  57
  58 echo
  59 echo "=== Testing L2 reference into L1 ==="
  60 echo
  61 _make_test_img 64M
  62 # Link first L1 entry (first L2 table) onto itself
  63 # (Note the MSb in the L1 entry is set, ensuring the refcount is one - else any
  64 # later write will result in a COW operation, effectively ruining this attempt
  65 # on image corruption)
  66 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x03\x00\x00"
  67 _check_test_img
  68
  69 # The corrupt bit should not be set anyway
  70 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
  71
  72 # Try to write something, thereby forcing the corrupt bit to be set
  73 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
  74
  75 # The corrupt bit must now be set
  76 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
  77
  78 # This information should be available through qemu-img info
  79 _img_info --format-specific
  80
  81 # Try to open the image R/W (which should fail)
  82 $QEMU_IO -c "$OPEN_RW" -c "read 0 512" 2>&1 | _filter_qemu_io \
  83                                             | _filter_testdir \
  84                                             | _filter_imgfmt
  85
  86 # Try to open it RO (which should succeed)
  87 $QEMU_IO -c "$OPEN_RO" -c "read 0 512" | _filter_qemu_io
  88
  89 # We could now try to fix the image, but this would probably fail (how should an
  90 # L2 table linked onto the L1 table be fixed?)
  91
  92 echo
  93 echo "=== Testing cluster data reference into refcount block ==="
  94 echo
  95 _make_test_img 64M
  96 # Allocate L2 table
  97 truncate -s "$(($l2_offset+65536))" "$TEST_IMG"
  98 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x00\x00"
  99 # Mark cluster as used
 100 poke_file "$TEST_IMG" "$(($rb_offset+8))" "\x00\x01"
 101 # Redirect new data cluster onto refcount block
 102 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x02\x00\x00"
 103 _check_test_img
 104 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 105 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
 106 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 107
 108 # Try to fix it
 109 _check_test_img -r all
 110
 111 # The corrupt bit should be cleared
 112 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 113
 114 # Look if it's really really fixed
 115 $QEMU_IO -c "$OPEN_RW" -c "write -P 0x2a 0 512" | _filter_qemu_io
 116 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 117
 118 echo
 119 echo "=== Testing cluster data reference into inactive L2 table ==="
 120 echo
 121 _make_test_img 64M
 122 $QEMU_IO -c "$OPEN_RW" -c "write -P 1 0 512" | _filter_qemu_io
 123 $QEMU_IMG snapshot -c foo "$TEST_IMG"
 124 $QEMU_IO -c "$OPEN_RW" -c "write -P 2 0 512" | _filter_qemu_io
 125 # The inactive L2 table remains at its old offset
 126 poke_file "$TEST_IMG" "$l2_offset_after_snapshot" \
 127                       "\x80\x00\x00\x00\x00\x04\x00\x00"
 128 _check_test_img
 129 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 130 $QEMU_IO -c "$OPEN_RW" -c "write -P 3 0 512" | _filter_qemu_io
 131 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 132 _check_test_img -r all
 133 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 134 $QEMU_IO -c "$OPEN_RW" -c "write -P 4 0 512" | _filter_qemu_io
 135 $PYTHON qcow2.py "$TEST_IMG" dump-header | grep incompatible_features
 136
 137 # Check data
 138 $QEMU_IO -c "$OPEN_RO" -c "read -P 4 0 512" | _filter_qemu_io
 139 $QEMU_IMG snapshot -a foo "$TEST_IMG"
 140 _check_test_img
 141 $QEMU_IO -c "$OPEN_RO" -c "read -P 1 0 512" | _filter_qemu_io
 142
 143 echo
 144 echo "=== Testing overlap while COW is in flight ==="
 145 echo
 146 # compat=0.10 is required in order to make the following discard actually
 147 # unallocate the sector rather than make it a zero sector - we want COW, after
 148 # all.
 149 IMGOPTS='compat=0.10' _make_test_img 1G
 150 # Write two clusters, the second one enforces creation of an L2 table after
 151 # the first data cluster.
 152 $QEMU_IO -c 'write 0k 64k' -c 'write 512M 64k' "$TEST_IMG" | _filter_qemu_io
 153 # Discard the first cluster. This cluster will soon enough be reallocated and
 154 # used for COW.
 155 $QEMU_IO -c 'discard 0k 64k' "$TEST_IMG" | _filter_qemu_io
 156 # Now, corrupt the image by marking the second L2 table cluster as free.
 157 poke_file "$TEST_IMG" '131084' "\x00\x00" # 0x2000c
 158 # Start a write operation requiring COW on the image stopping it right before
 159 # doing the read; then, trigger the corruption prevention by writing anything to
 160 # any unallocated cluster, leading to an attempt to overwrite the second L2
 161 # table. Finally, resume the COW write and see it fail (but not crash).
 162 echo "open -o file.driver=blkdebug $TEST_IMG
 163 break cow_read 0
 164 aio_write 0k 1k
 165 wait_break 0
 166 write 64k 64k
 167 resume 0" | $QEMU_IO | _filter_qemu_io
 168
 169 echo
 170 echo "=== Testing unallocated image header ==="
 171 echo
 172 _make_test_img 64M
 173 # Create L1/L2
 174 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 175 poke_file "$TEST_IMG" "$rb_offset" "\x00\x00"
 176 $QEMU_IO -c "write 64k 64k" "$TEST_IMG" | _filter_qemu_io
 177
 178 echo
 179 echo "=== Testing unaligned L1 entry ==="
 180 echo
 181 _make_test_img 64M
 182 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 183 # This will be masked with ~(512 - 1) = ~0x1ff, so whether the lower 9 bits are
 184 # aligned or not does not matter
 185 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x2a\x00"
 186 $QEMU_IO -c "read 0 64k" "$TEST_IMG" | _filter_qemu_io
 187
 188 # Test how well zero cluster expansion can cope with this
 189 _make_test_img 64M
 190 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 191 poke_file "$TEST_IMG" "$l1_offset" "\x80\x00\x00\x00\x00\x04\x2a\x00"
 192 $QEMU_IMG amend -o compat=0.10 "$TEST_IMG"
 193
 194 echo
 195 echo "=== Testing unaligned L2 entry ==="
 196 echo
 197 _make_test_img 64M
 198 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 199 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 200 $QEMU_IO -c "read 0 64k" "$TEST_IMG" | _filter_qemu_io
 201
 202 echo
 203 echo "=== Testing unaligned pre-allocated zero cluster ==="
 204 echo
 205 _make_test_img 64M
 206 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 207 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x01"
 208 # zero cluster expansion
 209 $QEMU_IMG amend -o compat=0.10 "$TEST_IMG"
 210
 211 echo
 212 echo "=== Testing unaligned reftable entry ==="
 213 echo
 214 _make_test_img 64M
 215 poke_file "$TEST_IMG" "$rt_offset" "\x00\x00\x00\x00\x00\x02\x2a\x00"
 216 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 217
 218 echo
 219 echo "=== Testing non-fatal corruption on freeing ==="
 220 echo
 221 _make_test_img 64M
 222 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 223 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 224 $QEMU_IO -c "discard 0 64k" "$TEST_IMG" | _filter_qemu_io
 225
 226 echo
 227 echo "=== Testing read-only corruption report ==="
 228 echo
 229 _make_test_img 64M
 230 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 231 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x05\x2a\x00"
 232 # Should only emit a single error message
 233 $QEMU_IO -c "$OPEN_RO" -c "read 0 64k" -c "read 0 64k" | _filter_qemu_io
 234
 235 echo
 236 echo "=== Testing non-fatal and then fatal corruption report ==="
 237 echo
 238 _make_test_img 64M
 239 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 240 poke_file "$TEST_IMG" "$l2_offset"        "\x80\x00\x00\x00\x00\x05\x2a\x00"
 241 poke_file "$TEST_IMG" "$(($l2_offset+8))" "\x80\x00\x00\x00\x00\x06\x2a\x00"
 242 # Should emit two error messages
 243 $QEMU_IO -c "discard 0 64k" -c "read 64k 64k" "$TEST_IMG" | _filter_qemu_io
 244
 245 echo
 246 echo "=== Testing empty refcount table ==="
 247 echo
 248 _make_test_img 64M
 249 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 250 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 251 # Repair the image
 252 _check_test_img -r all
 253
 254 echo
 255 echo "=== Testing empty refcount table with valid L1 and L2 tables ==="
 256 echo
 257 _make_test_img 64M
 258 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 259 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 260 # Since the first data cluster is already allocated this triggers an
 261 # allocation with an explicit offset (using qcow2_alloc_clusters_at())
 262 # causing a refcount block to be allocated at offset 0
 263 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 264 # Repair the image
 265 _check_test_img -r all
 266
 267 echo
 268 echo "=== Testing empty refcount block ==="
 269 echo
 270 _make_test_img 64M
 271 poke_file "$TEST_IMG" "$rb_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 272 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 273 # Repair the image
 274 _check_test_img -r all
 275
 276 echo
 277 echo "=== Testing empty refcount block with compressed write ==="
 278 echo
 279 _make_test_img 64M
 280 $QEMU_IO -c "write 64k 64k" "$TEST_IMG" | _filter_qemu_io
 281 poke_file "$TEST_IMG" "$rb_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 282 # The previous write already allocated an L2 table, so now this new
 283 # write will try to allocate a compressed data cluster at offset 0.
 284 $QEMU_IO -c "write -c 0k 64k" "$TEST_IMG" | _filter_qemu_io
 285 # Repair the image
 286 _check_test_img -r all
 287
 288 echo
 289 echo "=== Testing zero refcount table size ==="
 290 echo
 291 _make_test_img 64M
 292 poke_file "$TEST_IMG" "56"                "\x00\x00\x00\x00"
 293 $QEMU_IO -c "write 0 64k" "$TEST_IMG" 2>&1 | _filter_testdir | _filter_imgfmt
 294 # Repair the image
 295 _check_test_img -r all
 296
 297 echo
 298 echo "=== Testing incorrect refcount table offset ==="
 299 echo
 300 _make_test_img 64M
 301 poke_file "$TEST_IMG" "48"                "\x00\x00\x00\x00\x00\x00\x00\x00"
 302 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 303
 304 echo
 305 echo "=== Testing dirty corrupt image ==="
 306 echo
 307
 308 _make_test_img 64M
 309
 310 # Let the refblock appear unaligned
 311 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\xff\xff\x2a\x00"
 312 # Mark the image dirty, thus forcing an automatic check when opening it
 313 poke_file "$TEST_IMG" 72 "\x00\x00\x00\x00\x00\x00\x00\x01"
 314 # Open the image (qemu should refuse to do so)
 315 $QEMU_IO -c close "$TEST_IMG" 2>&1 | _filter_testdir | _filter_imgfmt
 316
 317 echo '--- Repairing ---'
 318
 319 # The actual repair should have happened (because of the dirty bit),
 320 # but some cleanup may have failed (like freeing the old reftable)
 321 # because the image was already marked corrupt by that point
 322 _check_test_img -r all
 323
 324 echo
 325 echo "=== Writing to an unaligned preallocated zero cluster ==="
 326 echo
 327
 328 _make_test_img 64M
 329
 330 # Allocate the L2 table
 331 $QEMU_IO -c "write 0 64k" -c "discard 0 64k" "$TEST_IMG" | _filter_qemu_io
 332 # Pretend there is a preallocated zero cluster somewhere inside the
 333 # image header
 334 poke_file "$TEST_IMG" "$l2_offset" "\x80\x00\x00\x00\x00\x00\x2a\x01"
 335 # Let's write to it!
 336 $QEMU_IO -c "write 0 64k" "$TEST_IMG" | _filter_qemu_io
 337
 338 # Can't repair this yet (TODO: We can just deallocate the cluster)
 339
 340 echo
 341 echo '=== Discarding with an unaligned refblock ==='
 342 echo
 343
 344 _make_test_img 64M
 345
 346 $QEMU_IO -c "write 0 128k" "$TEST_IMG" | _filter_qemu_io
 347 # Make our refblock unaligned
 348 poke_file "$TEST_IMG" "$(($rt_offset))" "\x00\x00\x00\x00\x00\x00\x2a\x00"
 349 # Now try to discard something that will be submitted as two requests
 350 # (main part + tail)
 351 $QEMU_IO -c "discard 0 65537" "$TEST_IMG"
 352
 353 echo '--- Repairing ---'
 354 # Fails the first repair because the corruption prevents the check
 355 # function from double-checking
 356 # (Using -q for the first invocation, because otherwise the
 357 #  double-check error message appears above the summary for some
 358 #  reason -- so let's just hide the summary)
 359 _check_test_img -q -r all
 360 _check_test_img -r all
 361
 362 echo
 363 echo "=== Discarding an out-of-bounds refblock ==="
 364 echo
 365
 366 _make_test_img 64M
 367
 368 # Pretend there's a refblock really up high
 369 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\xff\xff\xff\x00\x00\x00\x00"
 370 # Let's try to shrink the qcow2 image so that the block driver tries
 371 # to discard that refblock (and see what happens!)
 372 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 373
 374 echo '--- Checking and retrying ---'
 375 # Image should not be resized
 376 _img_info | grep 'virtual size'
 377 # But it should pass this check, because the "partial" resize has
 378 # already overwritten refblocks past the end
 379 _check_test_img -r all
 380 # So let's try again
 381 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 382 _img_info | grep 'virtual size'
 383
 384 echo
 385 echo "=== Discarding a non-covered in-bounds refblock ==="
 386 echo
 387
 388 IMGOPTS='refcount_bits=1' _make_test_img 64M
 389
 390 # Pretend there's a refblock somewhere where there is no refblock to
 391 # cover it (but the covering refblock has a valid index in the
 392 # reftable)
 393 # Every refblock covers 65536 * 8 * 65536 = 32 GB, so we have to point
 394 # to 0x10_0000_0000 (64G) to point to the third refblock
 395 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\x00\x00\x10\x00\x00\x00\x00"
 396 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 397
 398 echo '--- Checking and retrying ---'
 399 # Image should not be resized
 400 _img_info | grep 'virtual size'
 401 # But it should pass this check, because the "partial" resize has
 402 # already overwritten refblocks past the end
 403 _check_test_img -r all
 404 # So let's try again
 405 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 406 _img_info | grep 'virtual size'
 407
 408 echo
 409 echo "=== Discarding a refblock covered by an unaligned refblock ==="
 410 echo
 411
 412 IMGOPTS='refcount_bits=1' _make_test_img 64M
 413
 414 # Same as above
 415 poke_file "$TEST_IMG" "$(($rt_offset+8))" "\x00\x00\x00\x10\x00\x00\x00\x00"
 416 # But now we actually "create" an unaligned third refblock
 417 poke_file "$TEST_IMG" "$(($rt_offset+16))" "\x00\x00\x00\x00\x00\x00\x02\x00"
 418 $QEMU_IMG resize --shrink "$TEST_IMG" 32M
 419
 420 echo '--- Repairing ---'
 421 # Fails the first repair because the corruption prevents the check
 422 # function from double-checking
 423 # (Using -q for the first invocation, because otherwise the
 424 #  double-check error message appears above the summary for some
 425 #  reason -- so let's just hide the summary)
 426 _check_test_img -q -r all
 427 _check_test_img -r all
 428
 429 echo
 430 echo "=== Testing the QEMU shutdown with a corrupted image ==="
 431 echo
 432 _make_test_img 64M
 433 poke_file "$TEST_IMG" "$rt_offset"        "\x00\x00\x00\x00\x00\x00\x00\x00"
 434 echo "{'execute': 'qmp_capabilities'}
 435       {'execute': 'human-monitor-command',
 436        'arguments': {'command-line': 'qemu-io drive \"write 0 512\"'}}
 437       {'execute': 'quit'}" \
 438     | $QEMU -qmp stdio -nographic -nodefaults \
 439             -drive if=none,node-name=drive,file="$TEST_IMG",driver=qcow2 \
 440     | _filter_qmp | _filter_qemu_io
 441
 442 # success, all done
 443 echo "*** done"
 444 rm -f $seq.full
 445 status=0