qemu-e2k/tests/qemu-iotests/261

#!/usr/bin/env bash
# group: rw
#
# Test case for qcow2's handling of extra data in snapshot table entries
# (and more generally, how certain cases of broken snapshot tables are
# handled)
#
# Copyright (C) 2019 Red Hat, Inc.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#

# creator
owner=mreitz@redhat.com

seq=$(basename $0)
echo "QA output created by $seq"

status=1	# failure is the default!

_cleanup()
{
    _cleanup_test_img
    rm -f "$TEST_IMG".v{2,3}.orig
    rm -f "$TEST_DIR"/sn{0,1,2}{,-pre,-extra,-post}
}
trap "_cleanup; exit \$status" 0 1 2 3 15

# get standard environment, filters and checks
. ./common.rc
. ./common.filter

# This tests qcow2-specific low-level functionality
_supported_fmt qcow2
_supported_proto file
_supported_os Linux
# (1) We create a v2 image that supports nothing but refcount_bits=16
# (2) We do some refcount management on our own which expects
#     refcount_bits=16
# As for data files, they do not support snapshots at all.
_unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)' data_file

# Parameters:
#   $1: image filename
#   $2: snapshot table entry offset in the image
snapshot_table_entry_size()
{
    id_len=$(peek_file_be "$1" $(($2 + 12)) 2)
    name_len=$(peek_file_be "$1" $(($2 + 14)) 2)
    extra_len=$(peek_file_be "$1" $(($2 + 36)) 4)

    full_len=$((40 + extra_len + id_len + name_len))
    echo $(((full_len + 7) / 8 * 8))
}

# Parameter:
#   $1: image filename
print_snapshot_table()
{
    nb_entries=$(peek_file_be "$1" 60 4)
    offset=$(peek_file_be "$1" 64 8)

    echo "Snapshots in $1:" | _filter_testdir | _filter_imgfmt

    for ((i = 0; i < nb_entries; i++)); do
        id_len=$(peek_file_be "$1" $((offset + 12)) 2)
        name_len=$(peek_file_be "$1" $((offset + 14)) 2)
        extra_len=$(peek_file_be "$1" $((offset + 36)) 4)

        extra_ofs=$((offset + 40))
        id_ofs=$((extra_ofs + extra_len))
        name_ofs=$((id_ofs + id_len))

        echo "  [$i]"
        echo "    ID: $(peek_file_raw "$1" $id_ofs $id_len)"
        echo "    Name: $(peek_file_raw "$1" $name_ofs $name_len)"
        echo "    Extra data size: $extra_len"
        if [ $extra_len -ge 8 ]; then
            echo "    VM state size: $(peek_file_be "$1" $extra_ofs 8)"
        fi
        if [ $extra_len -ge 16 ]; then
            echo "    Disk size: $(peek_file_be "$1" $((extra_ofs + 8)) 8)"
        fi
        if [ $extra_len -ge 24 ]; then
            echo "    Icount: $(peek_file_be "$1" $((extra_ofs + 16)) 8)"
        fi
        if [ $extra_len -gt 24 ]; then
            echo '    Unknown extra data:' \
                "$(peek_file_raw "$1" $((extra_ofs + 16)) $((extra_len - 16)) \
                   | tr -d '\0')"
        fi

        offset=$((offset + $(snapshot_table_entry_size "$1" $offset)))
    done
}

# Mark clusters as allocated; works only in refblock 0 (i.e. before
# cluster #32768).
# Parameters:
#   $1: Start offset of what to allocate
#   $2: End offset (exclusive)
refblock0_allocate()
{
    reftable_ofs=$(peek_file_be "$TEST_IMG" 48 8)
    refblock_ofs=$(peek_file_be "$TEST_IMG" $reftable_ofs 8)

    cluster=$(($1 / 65536))
    ecluster=$((($2 + 65535) / 65536))

    while [ $cluster -lt $ecluster ]; do
        if [ $cluster -ge 32768 ]; then
            echo "*** Abort: Cluster $cluster exceeds refblock 0 ***"
            exit 1
        fi
        poke_file "$TEST_IMG" $((refblock_ofs + cluster * 2)) '\x00\x01'
        cluster=$((cluster + 1))
    done
}


echo
echo '=== Create v2 template ==='
echo

# Create v2 image with a snapshot table with three entries:
# [0]: No extra data (valid with v2, not valid with v3)
# [1]: Has extra data unknown to qemu
# [2]: Has the 64-bit VM state size, but not the disk size (again,
#      valid with v2, not valid with v3)

TEST_IMG="$TEST_IMG.v2.orig" IMGOPTS='compat=0.10' _make_test_img 64M
$QEMU_IMG snapshot -c sn0 "$TEST_IMG.v2.orig"
$QEMU_IMG snapshot -c sn1 "$TEST_IMG.v2.orig"
$QEMU_IMG snapshot -c sn2 "$TEST_IMG.v2.orig"

# Copy out all existing snapshot table entries
sn_table_ofs=$(peek_file_be "$TEST_IMG.v2.orig" 64 8)

# ofs: Snapshot table entry offset
# eds: Extra data size
# ids: Name + ID size
# len: Total entry length
sn0_ofs=$sn_table_ofs
sn0_eds=$(peek_file_be "$TEST_IMG.v2.orig" $((sn0_ofs + 36)) 4)
sn0_ids=$(($(peek_file_be "$TEST_IMG.v2.orig" $((sn0_ofs + 12)) 2) +
           $(peek_file_be "$TEST_IMG.v2.orig" $((sn0_ofs + 14)) 2)))
sn0_len=$(snapshot_table_entry_size "$TEST_IMG.v2.orig" $sn0_ofs)
sn1_ofs=$((sn0_ofs + sn0_len))
sn1_eds=$(peek_file_be "$TEST_IMG.v2.orig" $((sn1_ofs + 36)) 4)
sn1_ids=$(($(peek_file_be "$TEST_IMG.v2.orig" $((sn1_ofs + 12)) 2) +
           $(peek_file_be "$TEST_IMG.v2.orig" $((sn1_ofs + 14)) 2)))
sn1_len=$(snapshot_table_entry_size "$TEST_IMG.v2.orig" $sn1_ofs)
sn2_ofs=$((sn1_ofs + sn1_len))
sn2_eds=$(peek_file_be "$TEST_IMG.v2.orig" $((sn2_ofs + 36)) 4)
sn2_ids=$(($(peek_file_be "$TEST_IMG.v2.orig" $((sn2_ofs + 12)) 2) +
           $(peek_file_be "$TEST_IMG.v2.orig" $((sn2_ofs + 14)) 2)))
sn2_len=$(snapshot_table_entry_size "$TEST_IMG.v2.orig" $sn2_ofs)

# Data before extra data
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn0-pre" bs=1 skip=$sn0_ofs count=40 \
    &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn1-pre" bs=1 skip=$sn1_ofs count=40 \
    &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn2-pre" bs=1 skip=$sn2_ofs count=40 \
    &> /dev/null

# Extra data
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn0-extra" bs=1 \
    skip=$((sn0_ofs + 40)) count=$sn0_eds &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn1-extra" bs=1 \
    skip=$((sn1_ofs + 40)) count=$sn1_eds &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn2-extra" bs=1 \
    skip=$((sn2_ofs + 40)) count=$sn2_eds &> /dev/null

# Data after extra data
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn0-post" bs=1 \
    skip=$((sn0_ofs + 40 + sn0_eds)) count=$sn0_ids \
    &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn1-post" bs=1 \
    skip=$((sn1_ofs + 40 + sn1_eds)) count=$sn1_ids \
    &> /dev/null
dd if="$TEST_IMG.v2.orig" of="$TEST_DIR/sn2-post" bs=1 \
    skip=$((sn2_ofs + 40 + sn2_eds)) count=$sn2_ids \
    &> /dev/null

# Amend them, one by one
# Set sn0's extra data size to 0
poke_file "$TEST_DIR/sn0-pre" 36 '\x00\x00\x00\x00'
truncate -s 0 "$TEST_DIR/sn0-extra"
# Grow sn0-post to pad
truncate -s $(($(snapshot_table_entry_size "$TEST_DIR/sn0-pre") - 40)) \
    "$TEST_DIR/sn0-post"

# Set sn1's extra data size to 50
poke_file "$TEST_DIR/sn1-pre" 36 '\x00\x00\x00\x32'
truncate -s 50 "$TEST_DIR/sn1-extra"
poke_file "$TEST_DIR/sn1-extra" 24 'very important data'
# Grow sn1-post to pad
truncate -s $(($(snapshot_table_entry_size "$TEST_DIR/sn1-pre") - 90)) \
    "$TEST_DIR/sn1-post"

# Set sn2's extra data size to 8
poke_file "$TEST_DIR/sn2-pre" 36 '\x00\x00\x00\x08'
truncate -s 8 "$TEST_DIR/sn2-extra"
# Grow sn2-post to pad
truncate -s $(($(snapshot_table_entry_size "$TEST_DIR/sn2-pre") - 48)) \
    "$TEST_DIR/sn2-post"

# Construct snapshot table
cat "$TEST_DIR"/sn0-{pre,extra,post} \
    "$TEST_DIR"/sn1-{pre,extra,post} \
    "$TEST_DIR"/sn2-{pre,extra,post} \
    | dd of="$TEST_IMG.v2.orig" bs=1 seek=$sn_table_ofs conv=notrunc \
          &> /dev/null

# Done!
TEST_IMG="$TEST_IMG.v2.orig" _check_test_img
print_snapshot_table "$TEST_IMG.v2.orig"

echo
echo '=== Upgrade to v3 ==='
echo

cp "$TEST_IMG.v2.orig" "$TEST_IMG.v3.orig"
$QEMU_IMG amend -o compat=1.1 "$TEST_IMG.v3.orig"
TEST_IMG="$TEST_IMG.v3.orig" _check_test_img
print_snapshot_table "$TEST_IMG.v3.orig"

echo
echo '=== Repair botched v3 ==='
echo

# Force the v2 file to be v3.  v3 requires each snapshot table entry
# to have at least 16 bytes of extra data, so it will not comply to
# the qcow2 v3 specification; but we can fix that.
cp "$TEST_IMG.v2.orig" "$TEST_IMG"

# Set version
poke_file "$TEST_IMG" 4 '\x00\x00\x00\x03'
# Increase header length (necessary for v3)
poke_file "$TEST_IMG" 100 '\x00\x00\x00\x68'
# Set refcount order (necessary for v3)
poke_file "$TEST_IMG" 96 '\x00\x00\x00\x04'

_check_test_img -r all
print_snapshot_table "$TEST_IMG"


# From now on, just test the qcow2 version we are supposed to test.
# (v3 by default, v2 by choice through $IMGOPTS.)
# That works because we always write all known extra data when
# updating the snapshot table, independent of the version.

if echo "$IMGOPTS" | grep -q 'compat=\(0\.10\|v2\)' 2> /dev/null; then
    subver=v2
else
    subver=v3
fi

echo
echo '=== Add new snapshot ==='
echo

cp "$TEST_IMG.$subver.orig" "$TEST_IMG"
$QEMU_IMG snapshot -c sn3 "$TEST_IMG"
_check_test_img
print_snapshot_table "$TEST_IMG"

echo
echo '=== Remove different snapshots ==='

for sn in sn0 sn1 sn2; do
    echo
    echo "--- $sn ---"

    cp "$TEST_IMG.$subver.orig" "$TEST_IMG"
    $QEMU_IMG snapshot -d $sn "$TEST_IMG"
    _check_test_img
    print_snapshot_table "$TEST_IMG"
done

echo
echo '=== Reject too much unknown extra data ==='
echo

cp "$TEST_IMG.$subver.orig" "$TEST_IMG"
$QEMU_IMG snapshot -c sn3 "$TEST_IMG"

sn_table_ofs=$(peek_file_be "$TEST_IMG" 64 8)
sn0_ofs=$sn_table_ofs
sn1_ofs=$((sn0_ofs + $(snapshot_table_entry_size "$TEST_IMG" $sn0_ofs)))
sn2_ofs=$((sn1_ofs + $(snapshot_table_entry_size "$TEST_IMG" $sn1_ofs)))
sn3_ofs=$((sn2_ofs + $(snapshot_table_entry_size "$TEST_IMG" $sn2_ofs)))

# 64 kB of extra data should be rejected
# (Note that this also induces a refcount error, because it spills
# over to the next cluster.  That's a good way to test that we can
# handle simultaneous snapshot table and refcount errors.)
poke_file "$TEST_IMG" $((sn3_ofs + 36)) '\x00\x01\x00\x00'

# Print error
_img_info
echo
_check_test_img
echo

# Should be repairable
_check_test_img -r all

echo
echo '=== Snapshot table too big ==='
echo

sn_table_ofs=$(peek_file_be "$TEST_IMG.v3.orig" 64 8)

# Fill a snapshot with 1 kB of extra data, a 65535-char ID, and a
# 65535-char name, and repeat it as many times as necessary to fill
# 64 MB (the maximum supported by qemu)

touch "$TEST_DIR/sn0"

# Full size (fixed + extra + ID + name + padding)
sn_size=$((40 + 1024 + 65535 + 65535 + 2))

# We only need the fixed part, though.
truncate -s 40 "$TEST_DIR/sn0"

# 65535-char ID string
poke_file "$TEST_DIR/sn0" 12 '\xff\xff'
# 65535-char name
poke_file "$TEST_DIR/sn0" 14 '\xff\xff'
# 1 kB of extra data
poke_file "$TEST_DIR/sn0" 36 '\x00\x00\x04\x00'

# Create test image
_make_test_img 64M

# Hook up snapshot table somewhere safe (at 1 MB)
poke_file "$TEST_IMG" 64 '\x00\x00\x00\x00\x00\x10\x00\x00'

offset=1048576
size_written=0
sn_count=0
while [ $size_written -le $((64 * 1048576)) ]; do
    dd if="$TEST_DIR/sn0" of="$TEST_IMG" bs=1 seek=$offset conv=notrunc \
        &> /dev/null
    offset=$((offset + sn_size))
    size_written=$((size_written + sn_size))
    sn_count=$((sn_count + 1))
done
truncate -s "$offset" "$TEST_IMG"

# Give the last snapshot (the one to be removed) an L1 table so we can
# see how that is handled when repairing the image
# (Put it two clusters before 1 MB, and one L2 table one cluster
# before 1 MB)
poke_file "$TEST_IMG" $((offset - sn_size + 0)) \
    '\x00\x00\x00\x00\x00\x0e\x00\x00'
poke_file "$TEST_IMG" $((offset - sn_size + 8)) \
    '\x00\x00\x00\x01'

# Hook up the L2 table
poke_file "$TEST_IMG" $((1048576 - 2 * 65536)) \
    '\x80\x00\x00\x00\x00\x0f\x00\x00'

# Make sure all of the clusters we just hooked up are allocated:
# - The snapshot table
# - The last snapshot's L1 and L2 table
refblock0_allocate $((1048576 - 2 * 65536)) $offset

poke_file "$TEST_IMG" 60 \
    "$(printf '%08x' $sn_count | sed -e 's/\(..\)/\\x\1/g')"

# Print error
_img_info
echo
_check_test_img
echo

# Should be repairable
_check_test_img -r all

echo
echo "$((sn_count - 1)) snapshots should remain:"
echo "  qemu-img info reports $(_img_info | grep -c '^ \{32\}') snapshots"
echo "  Image header reports $(peek_file_be "$TEST_IMG" 60 4) snapshots"

echo
echo '=== Snapshot table too big with one entry with too much extra data ==='
echo

# For this test, we reuse the image from the previous case, which has
# a snapshot table that is right at the limit.
# Our layout looks like this:
# - (a number of snapshot table entries)
# - One snapshot with $extra_data_size extra data
# - One normal snapshot that breaks the 64 MB boundary
# - One normal snapshot beyond the 64 MB boundary
#
# $extra_data_size is calculated so that simply by virtue of it
# decreasing to 1 kB, the penultimate snapshot will fit into 64 MB
# limit again.  The final snapshot will always be beyond the limit, so
# that we can see that the repair algorithm does still determine the
# limit to be somewhere, even when truncating one snapshot's extra
# data.

# The last case has removed the last snapshot, so calculate
# $old_offset to get the current image's real length
old_offset=$((offset - sn_size))

# The layout from the previous test had one snapshot beyond the 64 MB
# limit; we want the same (after the oversized extra data has been
# truncated to 1 kB), so we drop the last three snapshots and
# construct them from scratch.
offset=$((offset - 3 * sn_size))
sn_count=$((sn_count - 3))

# Assuming we had already written one of the three snapshots
# (necessary so we can calculate $extra_data_size next).
size_written=$((size_written - 2 * sn_size))

# Increase the extra data size so we go past the limit
# (The -1024 comes from the 1 kB of extra data we already have)
extra_data_size=$((64 * 1048576 + 8 - sn_size - (size_written - 1024)))

poke_file "$TEST_IMG" $((offset + 36)) \
    "$(printf '%08x' $extra_data_size | sed -e 's/\(..\)/\\x\1/g')"

offset=$((offset + sn_size - 1024 + extra_data_size))
size_written=$((size_written - 1024 + extra_data_size))
sn_count=$((sn_count + 1))

# Write the two normal snapshots
for ((i = 0; i < 2; i++)); do
    dd if="$TEST_DIR/sn0" of="$TEST_IMG" bs=1 seek=$offset conv=notrunc \
        &> /dev/null
    offset=$((offset + sn_size))
    size_written=$((size_written + sn_size))
    sn_count=$((sn_count + 1))

    if [ $i = 0 ]; then
        # Check that the penultimate snapshot is beyond the 64 MB limit
        echo "Snapshot table size should equal $((64 * 1048576 + 8)):" \
            $size_written
        echo
    fi
done

truncate -s $offset "$TEST_IMG"
refblock0_allocate $old_offset $offset

poke_file "$TEST_IMG" 60 \
    "$(printf '%08x' $sn_count | sed -e 's/\(..\)/\\x\1/g')"

# Print error
_img_info
echo
_check_test_img
echo

# Just truncating the extra data should be sufficient to shorten the
# snapshot table so only one snapshot exceeds the extra size
_check_test_img -r all

echo
echo '=== Too many snapshots ==='
echo

# Create a v2 image, for speeds' sake: All-zero snapshot table entries
# are only valid in v2.
IMGOPTS='compat=0.10' _make_test_img 64M

# Hook up snapshot table somewhere safe (at 1 MB)
poke_file "$TEST_IMG" 64 '\x00\x00\x00\x00\x00\x10\x00\x00'
# "Create" more than 65536 snapshots (twice that many here)
poke_file "$TEST_IMG" 60 '\x00\x02\x00\x00'

# 40-byte all-zero snapshot table entries are valid snapshots, but
# only in v2 (v3 needs 16 bytes of extra data, so we would have to
# write 131072x '\x10').
truncate -s $((1048576 + 40 * 131072)) "$TEST_IMG"

# But let us give one of the snapshots to be removed an L1 table so
# we can see how that is handled when repairing the image.
# (Put it two clusters before 1 MB, and one L2 table one cluster
# before 1 MB)
poke_file "$TEST_IMG" $((1048576 + 40 * 65536 + 0)) \
    '\x00\x00\x00\x00\x00\x0e\x00\x00'
poke_file "$TEST_IMG" $((1048576 + 40 * 65536 + 8)) \
    '\x00\x00\x00\x01'

# Hook up the L2 table
poke_file "$TEST_IMG" $((1048576 - 2 * 65536)) \
    '\x80\x00\x00\x00\x00\x0f\x00\x00'

# Make sure all of the clusters we just hooked up are allocated:
# - The snapshot table
# - The last snapshot's L1 and L2 table
refblock0_allocate $((1048576 - 2 * 65536)) $((1048576 + 40 * 131072))

# Print error
_img_info
echo
_check_test_img
echo

# Should be repairable
_check_test_img -r all

echo
echo '65536 snapshots should remain:'
echo "  qemu-img info reports $(_img_info | grep -c '^ \{32\}') snapshots"
echo "  Image header reports $(peek_file_be "$TEST_IMG" 60 4) snapshots"

# success, all done
echo "*** done"
status=0