2019-05-12 12:27:43

by Eryu Guan

[permalink] [raw]
Subject: Re: [RFC PATCH 4/7] common/encrypt: add helper for ciphertext verification tests

On Fri, Apr 26, 2019 at 01:41:50PM -0700, Eric Biggers wrote:
> From: Eric Biggers <[email protected]>
>
> Introduce a function _verify_ciphertext_for_encryption_policy() which
> verifies the correctness of encryption with the specified settings.
>
> Basically, it does the following:
>
> 1. If missing any prerequisites, skip the test.
>
> 2. Create files in encrypted directories on the scratch device.
>
> 3. Unmount the scratch device and compare the actual ciphertext stored
> on-disk to the ciphertext computed by the fscrypt-crypt-util program.
>
> Both file contents and names are verified, and non-default encryption
> modes are supported. Previously, non-default encryption modes were
> untested by xfstests. Also, while there's an existing test generic/399
> that checks that encrypted contents seem random, it doesn't actually
> test for correctness, nor does it test filenames encryption.
>
> Signed-off-by: Eric Biggers <[email protected]>
> ---
> common/encrypt | 390 +++++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 390 insertions(+)
>
> diff --git a/common/encrypt b/common/encrypt
> index 37f16b94..3e48abc0 100644
> --- a/common/encrypt
> +++ b/common/encrypt
> @@ -263,3 +263,393 @@ _get_encpolicy()
>
> $XFS_IO_PROG -c "get_encpolicy $*" "$file"
> }
> +
> +# Retrieve the encryption nonce of the given inode as a hex string. The nonce
> +# was randomly generated by the filesystem and isn't exposed directly to
> +# userspace. But it can be read using the filesystem's debugging tools.
> +_get_encryption_nonce()
> +{
> + local device=$1
> + local inode=$2
> +
> + case $FSTYP in
> + ext4)
> + # Use debugfs to dump the special xattr named "c", which is the
> + # file's fscrypt_context. This produces a line like:
> + #
> + # c (28) = 01 01 04 02 00 00 00 00 00 00 00 00 ef bd 18 76 5d f6 41 4e c0 a2 cd 5f 91 29 7e 12
> + #
> + # Then filter it to get just the 16-byte 'nonce' field at the end:
> + #
> + # efbd18765df6414ec0a2cd5f91297e12
> + #
> + $DEBUGFS_PROG $device -R "ea_get <$inode> c" 2>>$seqres.full \
> + | grep '^c' | sed 's/^.*=//' | tr -d ' \n' | tail -c 32
> + ;;
> + f2fs)
> + # dump.f2fs prints the fscrypt_context like:
> + #
> + # xattr: e_name_index:9 e_name:c e_name_len:1 e_value_size:28 e_value:
> + # format: 1
> + # contents_encryption_mode: 0x1
> + # filenames_encryption_mode: 0x4
> + # flags: 0x2
> + # master_key_descriptor: 0000000000000000
> + # nonce: EFBD18765DF6414EC0A2CD5F91297E12
> + $DUMP_F2FS_PROG -i $inode $device | awk '
> + /\<e_name:c\>/ { found = 1 }
> + /^nonce:/ && found {
> + print substr($0, length($0) - 31, 32);
> + found = 0;
> + }'
> + ;;
> + *)
> + _notrun "_get_encryption_nonce() isn't implemented on $FSTYP"
> + ;;

No need to _notrun here, _require_get_encryption_nonce_support() should
be called first and already does the check in test. Perhaps just echoing
an error message to indicate the failure here is fine.

> + esac
> +}
> +
> +# Require support for _get_encryption_nonce()
> +_require_get_encryption_nonce_support()
> +{
> + echo "Checking for _get_encryption_nonce() support for $FSTYP" >> $seqres.full
> + case $FSTYP in
> + ext4)
> + _require_command "$DEBUGFS_PROG" debugfs
> + ;;
> + f2fs)
> + _require_command "$DUMP_F2FS_PROG" dump.f2fs
> + ;;
> + *)
> + _notrun "_get_encryption_nonce() isn't implemented on $FSTYP"
> + ;;
> + esac
> +}
> +
> +# Retrieve the filename stored on-disk for the given file.
> +# The name is printed to stdout in binary.
> +_get_on_disk_filename()
> +{
> + local device=$1
> + local inode=$2
> + local dir_inode=$3
> +
> + case $FSTYP in
> + ext4)
> + # Extract the filename from the debugfs output line like:
> + #
> + # 131075 100644 (1) 0 0 0 22-Apr-2019 16:54 \xa2\x85\xb0z\x13\xe9\x09\x86R\xed\xdc\xce\xad\x14d\x19
> + #
> + $DEBUGFS_PROG $device -R "ls -l -r <$dir_inode>" \
> + 2>>$seqres.full | perl -ne '
> + next if not /^\s*'$inode'\s+/;
> + s/.*?\d\d:\d\d //;
> + chomp;
> + s/\\x([[:xdigit:]]{2})/chr hex $1/eg;
> + print;'
> + ;;
> + f2fs)
> + # Extract the filename from the dump.f2fs output line like:
> + #
> + # i_name [UpkzIPuts9by1oDmE+Ivfw]
> + #
> + # The name is base64-encoded, so we have to decode it here.
> + #
> + $DUMP_F2FS_PROG $device -i $inode | perl -ne '
> + next if not /^i_name\s+\[([A-Za-z0-9+,]+)\]/;
> + chomp $1;
> + my @chars = split //, $1;
> + my $ac = 0;
> + my $bits = 0;
> + my $table = join "", (A..Z, a..z, 0..9, "+", ",");
> + foreach (@chars) {
> + $ac += index($table, $_) << $bits;
> + $bits += 6;
> + if ($bits >= 8) {
> + print chr($ac & 0xff);
> + $ac >>= 8;
> + $bits -= 8;
> + }
> + }
> + if ($ac != 0) {
> + print STDERR "Invalid base64-encoded string!\n";
> + }'
> + ;;
> + *)
> + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"

Same here, the get_on_disk_filename support is checked by
_require_get_on_disk_filename_support(). And looks like this function
has nothing to do with fs encryption, move it to common/rc?

> + ;;
> + esac
> +}
> +
> +# Require support for _get_on_disk_filename()
> +_require_get_on_disk_filename_support()
> +{
> + echo "Checking for _get_on_disk_filename() support for $FSTYP" >> $seqres.full
> + case $FSTYP in
> + ext4)
> + # Verify that the "ls -l -r" debugfs command is supported and
> + # hex-encodes non-ASCII characters, rather than using an
> + # ambiguous escaping method. This requires the e2fsprogs patch
> + # "debugfs: avoid ambiguity when printing filenames"
> + # (https://marc.info/?l=linux-ext4&m=155596495624232&w=2).
> + # TODO: once merged, list the minimum e2fsprogs version here.
> + _require_command "$DEBUGFS_PROG" debugfs
> + _scratch_mount
> + touch $SCRATCH_MNT/$'\xc1'
> + _scratch_unmount
> + if ! $DEBUGFS_PROG $SCRATCH_DEV -R "ls -l -r /" 2>&1 \
> + | tee -a $seqres.full | grep -E -q '\s+\\xc1\s*$'; then
> + _notrun "debugfs (e2fsprogs) is too old; doesn't support showing unambiguous on-disk filenames"
> + fi
> + ;;
> + f2fs)
> + # Verify that dump.f2fs shows encrypted filenames in full. This
> + # requires the patch "f2fs-tools: improve filename printing"
> + # (https://sourceforge.net/p/linux-f2fs/mailman/message/36648641/).
> + # TODO: once merged, list the minimum f2fs-tools version here.
> +
> + _require_command "$DUMP_F2FS_PROG" dump.f2fs
> + _require_command "$KEYCTL_PROG" keyctl
> + _scratch_mount
> + _new_session_keyring
> +
> + local keydesc=$(_generate_encryption_key)
> + local dir=$SCRATCH_MNT/test.${FUNCNAME[0]}
> + local file=$dir/$(perl -e 'print "A" x 255')
> + mkdir $dir
> + _set_encpolicy $dir $keydesc
> + touch $file
> + local inode=$(stat -c %i $file)
> +
> + _scratch_unmount
> + $KEYCTL_PROG clear @s
> +
> + # 255-character filename should result in 340 base64 characters.
> + if ! $DUMP_F2FS_PROG -i $inode $SCRATCH_DEV \
> + | grep -E -q '^i_name[[:space:]]+\[[A-Za-z0-9+,]{340}\]'; then
> + _notrun "dump.f2fs (f2fs-tools) is too old; doesn't support showing unambiguous on-disk filenames"
> + fi
> + ;;
> + *)
> + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"
> + ;;
> + esac
> +}
> +
> +# Get the file's list of on-disk blocks as a comma-separated list of block
> +# offsets from the start of the device. "Blocks" are 512 bytes each here.
> +_get_file_block_list()
> +{
> + local file=$1
> +
> + sync
> + $XFS_IO_PROG -c fiemap $file | perl -ne '
> + next if not /^\s*\d+: \[\d+\.\.\d+\]: (\d+)\.\.(\d+)/;
> + print $_ . "," foreach $1..$2;' | sed 's/,$//'
> +}
> +
> +# Dump a block list that was previously saved by _get_file_block_list().
> +_dump_file_blocks()
> +{
> + local device=$1
> + local blocklist=$2
> + local block
> +
> + for block in $(tr ',' ' ' <<< $blocklist); do
> + dd if=$device bs=512 count=1 skip=$block status=none
> + done
> +}

Above two functions seem generic enough to be moved to common/rc

Thanks,
Eryu

> +
> +_do_verify_ciphertext_for_encryption_policy()
> +{
> + local contents_encryption_mode=$1
> + local filenames_encryption_mode=$2
> + local policy_flags=$3
> + local set_encpolicy_args=$4
> + local keydesc=$5
> + local raw_key_hex=$6
> + local crypt_cmd="src/fscrypt-crypt-util $7"
> +
> + local blocksize=$(_get_block_size $SCRATCH_MNT)
> + local test_contents_files=()
> + local test_filenames_files=()
> + local i src dir dst inode blocklist \
> + padding_flag padding dir_inode len name f nonce decrypted_name
> +
> + # Create files whose encrypted contents we'll verify. For each, save
> + # the information: (copy of original file, inode number of encrypted
> + # file, comma-separated block list) into test_contents_files[].
> + echo "Creating files for contents verification" >> $seqres.full
> + i=1
> + rm -f $tmp.testfile_*
> + for src in /dev/zero /dev/urandom; do
> + head -c $((4 * blocksize)) $src > $tmp.testfile_$i
> + (( i++ ))
> + done
> + dir=$SCRATCH_MNT/encdir
> + mkdir $dir
> + _set_encpolicy $dir $keydesc $set_encpolicy_args -f $policy_flags
> + for src in $tmp.testfile_*; do
> + dst=$dir/${src##*.}
> + cp $src $dst
> + inode=$(stat -c %i $dst)
> + blocklist=$(_get_file_block_list $dst)
> + test_contents_files+=("$src $inode $blocklist")
> + done
> +
> + # Create files whose encrypted names we'll verify. For each, save the
> + # information: (original filename, inode number of encrypted file, inode
> + # of parent directory, padding amount) into test_filenames_files[]. Try
> + # each padding amount: 4, 8, 16, or 32 bytes. Also try various filename
> + # lengths, including boundary cases. Assume NAME_MAX == 255.
> + echo "Creating files for filenames verification" >> $seqres.full
> + for padding_flag in 0 1 2 3; do
> + padding=$((4 << padding_flag))
> + dir=$SCRATCH_MNT/encdir.pad$padding
> + mkdir $dir
> + dir_inode=$(stat -c %i $dir)
> + _set_encpolicy $dir $keydesc $set_encpolicy_args \
> + -f $((policy_flags | padding_flag))
> + for len in 1 3 15 16 17 32 100 254 255; do
> + name=$(tr -d -C a-zA-Z0-9 < /dev/urandom | head -c $len)
> + touch $dir/$name
> + inode=$(stat -c %i $dir/$name)
> + test_filenames_files+=("$name $inode $dir_inode $padding")
> + done
> + done
> +
> + # Now unmount the filesystem and verify the ciphertext we just wrote.
> + _scratch_unmount
> +
> + echo "Verifying encrypted file contents" >> $seqres.full
> + for f in "${test_contents_files[@]}"; do
> + read -r src inode blocklist <<< "$f"
> + nonce=$(_get_encryption_nonce $SCRATCH_DEV $inode)
> + _dump_file_blocks $SCRATCH_DEV $blocklist > $tmp.actual_contents
> + $crypt_cmd $contents_encryption_mode $raw_key_hex \
> + --file-nonce=$nonce --block-size=$blocksize \
> + < $src > $tmp.expected_contents
> + if ! cmp $tmp.expected_contents $tmp.actual_contents; then
> + _fail "Expected encrypted contents != actual encrypted contents. File: $f"
> + fi
> + $crypt_cmd $contents_encryption_mode $raw_key_hex --decrypt \
> + --file-nonce=$nonce --block-size=$blocksize \
> + < $tmp.actual_contents > $tmp.decrypted_contents
> + if ! cmp $src $tmp.decrypted_contents; then
> + _fail "Contents decryption sanity check failed. File: $f"
> + fi
> + done
> +
> + echo "Verifying encrypted file names" >> $seqres.full
> + for f in "${test_filenames_files[@]}"; do
> + read -r name inode dir_inode padding <<< "$f"
> + nonce=$(_get_encryption_nonce $SCRATCH_DEV $dir_inode)
> + _get_on_disk_filename $SCRATCH_DEV $inode $dir_inode \
> + > $tmp.actual_name
> + echo -n "$name" | \
> + $crypt_cmd $filenames_encryption_mode $raw_key_hex \
> + --file-nonce=$nonce --padding=$padding \
> + --block-size=255 > $tmp.expected_name
> + if ! cmp $tmp.expected_name $tmp.actual_name; then
> + _fail "Expected encrypted filename != actual encrypted filename. File: $f"
> + fi
> + $crypt_cmd $filenames_encryption_mode $raw_key_hex --decrypt \
> + --file-nonce=$nonce --padding=$padding \
> + --block-size=255 < $tmp.actual_name \
> + > $tmp.decrypted_name
> + decrypted_name=$(tr -d '\0' < $tmp.decrypted_name)
> + if [ "$name" != "$decrypted_name" ]; then
> + _fail "Filename decryption sanity check failed ($name != $decrypted_name). File: $f"
> + fi
> + done
> +}
> +
> +_fscrypt_mode_name_to_num()
> +{
> + local name=$1
> +
> + case "$name" in
> + AES-256-XTS) echo 1 ;; # FS_ENCRYPTION_MODE_AES_256_XTS
> + AES-256-CTS-CBC) echo 4 ;; # FS_ENCRYPTION_MODE_AES_256_CTS
> + AES-128-CBC-ESSIV) echo 5 ;; # FS_ENCRYPTION_MODE_AES_128_CBC
> + AES-128-CTS-CBC) echo 6 ;; # FS_ENCRYPTION_MODE_AES_128_CTS
> + Adiantum) echo 9 ;; # FS_ENCRYPTION_MODE_ADIANTUM
> + *) _fail "Unknown fscrypt mode: $name" ;;
> + esac
> +}
> +
> +# Verify that file contents and names are encrypted correctly when an encryption
> +# policy of the specified type is used.
> +#
> +# The first two parameters are the contents and filenames encryption modes to
> +# test. Optionally, also specify 'direct' to test the DIRECT_KEY flag.
> +_verify_ciphertext_for_encryption_policy()
> +{
> + local contents_encryption_mode=$1
> + local filenames_encryption_mode=$2
> + local opt
> + local policy_flags=0
> + local set_encpolicy_args=""
> + local crypt_util_args=""
> +
> + shift 2
> + for opt; do
> + case "$opt" in
> + direct)
> + if [ $contents_encryption_mode != \
> + $filenames_encryption_mode ]; then
> + _fail "For direct key mode, contents and filenames modes must match"
> + fi
> + (( policy_flags |= 0x04 )) # FS_POLICY_FLAG_DIRECT_KEY
> + ;;
> + *)
> + _fail "Unknown option '$opt' passed to ${FUNCNAME[0]}"
> + ;;
> + esac
> + done
> + local contents_mode_num=$(_fscrypt_mode_name_to_num $contents_encryption_mode)
> + local filenames_mode_num=$(_fscrypt_mode_name_to_num $filenames_encryption_mode)
> +
> + set_encpolicy_args+=" -c $contents_mode_num"
> + set_encpolicy_args+=" -n $filenames_mode_num"
> +
> + if (( policy_flags & 0x04 )); then
> + crypt_util_args+=" --kdf=none"
> + else
> + crypt_util_args+=" --kdf=AES-128-ECB"
> + fi
> + set_encpolicy_args=${set_encpolicy_args# }
> +
> + _require_scratch_encryption $set_encpolicy_args
> + _require_test_program "fscrypt-crypt-util"
> + _require_xfs_io_command "fiemap"
> + _require_get_encryption_nonce_support
> + _require_get_on_disk_filename_support
> + _require_command "$KEYCTL_PROG" keyctl
> +
> + echo "Creating encryption-capable filesystem" >> $seqres.full
> + _scratch_mkfs_encrypted &>> $seqres.full
> + _scratch_mount
> +
> + echo "Generating encryption key" >> $seqres.full
> + local raw_key=$(_generate_raw_encryption_key)
> + local keydesc=$(_generate_key_descriptor)
> + _new_session_keyring
> + _add_encryption_key $keydesc $raw_key
> + local raw_key_hex=$(echo "$raw_key" | tr -d '\\x')
> +
> + echo
> + echo -e "Verifying ciphertext with parameters:"
> + echo -e "\tcontents_encryption_mode: $contents_encryption_mode"
> + echo -e "\tfilenames_encryption_mode: $filenames_encryption_mode"
> + [ $# -ne 0 ] && echo -e "\toptions: $*"
> +
> + _do_verify_ciphertext_for_encryption_policy \
> + "$contents_encryption_mode" \
> + "$filenames_encryption_mode" \
> + "$policy_flags" \
> + "$set_encpolicy_args" \
> + "$keydesc" \
> + "$raw_key_hex" \
> + "$crypt_util_args"
> +}
> --
> 2.21.0.593.g511ec345e18-goog
>


2019-05-13 19:53:09

by Eric Biggers

[permalink] [raw]
Subject: Re: [RFC PATCH 4/7] common/encrypt: add helper for ciphertext verification tests

Hi Eryu,

On Sun, May 12, 2019 at 08:27:03PM +0800, Eryu Guan wrote:
> > +# Retrieve the filename stored on-disk for the given file.
> > +# The name is printed to stdout in binary.
> > +_get_on_disk_filename()
> > +{
> > + local device=$1
> > + local inode=$2
> > + local dir_inode=$3
> > +
> > + case $FSTYP in
> > + ext4)
> > + # Extract the filename from the debugfs output line like:
> > + #
> > + # 131075 100644 (1) 0 0 0 22-Apr-2019 16:54 \xa2\x85\xb0z\x13\xe9\x09\x86R\xed\xdc\xce\xad\x14d\x19
> > + #
> > + $DEBUGFS_PROG $device -R "ls -l -r <$dir_inode>" \
> > + 2>>$seqres.full | perl -ne '
> > + next if not /^\s*'$inode'\s+/;
> > + s/.*?\d\d:\d\d //;
> > + chomp;
> > + s/\\x([[:xdigit:]]{2})/chr hex $1/eg;
> > + print;'
> > + ;;
> > + f2fs)
> > + # Extract the filename from the dump.f2fs output line like:
> > + #
> > + # i_name [UpkzIPuts9by1oDmE+Ivfw]
> > + #
> > + # The name is base64-encoded, so we have to decode it here.
> > + #
> > + $DUMP_F2FS_PROG $device -i $inode | perl -ne '
> > + next if not /^i_name\s+\[([A-Za-z0-9+,]+)\]/;
> > + chomp $1;
> > + my @chars = split //, $1;
> > + my $ac = 0;
> > + my $bits = 0;
> > + my $table = join "", (A..Z, a..z, 0..9, "+", ",");
> > + foreach (@chars) {
> > + $ac += index($table, $_) << $bits;
> > + $bits += 6;
> > + if ($bits >= 8) {
> > + print chr($ac & 0xff);
> > + $ac >>= 8;
> > + $bits -= 8;
> > + }
> > + }
> > + if ($ac != 0) {
> > + print STDERR "Invalid base64-encoded string!\n";
> > + }'
> > + ;;
> > + *)
> > + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"
>
> And looks like this function has nothing to do with fs encryption, move it to
> common/rc?

For ext4 that's true, but for f2fs the name is assumed to be base64-encoded,
which f2fs-tools only does for encrypted filenames. I'll update the comment to
clarify that the function assumes the directory is encrypted.

>
> > + ;;
> > + esac
> > +}
> > +
> > +# Require support for _get_on_disk_filename()
> > +_require_get_on_disk_filename_support()
> > +{
> > + echo "Checking for _get_on_disk_filename() support for $FSTYP" >> $seqres.full
> > + case $FSTYP in
> > + ext4)
> > + # Verify that the "ls -l -r" debugfs command is supported and
> > + # hex-encodes non-ASCII characters, rather than using an
> > + # ambiguous escaping method. This requires the e2fsprogs patch
> > + # "debugfs: avoid ambiguity when printing filenames"
> > + # (https://marc.info/?l=linux-ext4&m=155596495624232&w=2).
> > + # TODO: once merged, list the minimum e2fsprogs version here.
> > + _require_command "$DEBUGFS_PROG" debugfs
> > + _scratch_mount
> > + touch $SCRATCH_MNT/$'\xc1'
> > + _scratch_unmount
> > + if ! $DEBUGFS_PROG $SCRATCH_DEV -R "ls -l -r /" 2>&1 \
> > + | tee -a $seqres.full | grep -E -q '\s+\\xc1\s*$'; then
> > + _notrun "debugfs (e2fsprogs) is too old; doesn't support showing unambiguous on-disk filenames"
> > + fi
> > + ;;
> > + f2fs)
> > + # Verify that dump.f2fs shows encrypted filenames in full. This
> > + # requires the patch "f2fs-tools: improve filename printing"
> > + # (https://sourceforge.net/p/linux-f2fs/mailman/message/36648641/).
> > + # TODO: once merged, list the minimum f2fs-tools version here.
> > +
> > + _require_command "$DUMP_F2FS_PROG" dump.f2fs
> > + _require_command "$KEYCTL_PROG" keyctl
> > + _scratch_mount
> > + _new_session_keyring
> > +
> > + local keydesc=$(_generate_encryption_key)
> > + local dir=$SCRATCH_MNT/test.${FUNCNAME[0]}
> > + local file=$dir/$(perl -e 'print "A" x 255')
> > + mkdir $dir
> > + _set_encpolicy $dir $keydesc
> > + touch $file
> > + local inode=$(stat -c %i $file)
> > +
> > + _scratch_unmount
> > + $KEYCTL_PROG clear @s
> > +
> > + # 255-character filename should result in 340 base64 characters.
> > + if ! $DUMP_F2FS_PROG -i $inode $SCRATCH_DEV \
> > + | grep -E -q '^i_name[[:space:]]+\[[A-Za-z0-9+,]{340}\]'; then
> > + _notrun "dump.f2fs (f2fs-tools) is too old; doesn't support showing unambiguous on-disk filenames"
> > + fi
> > + ;;
> > + *)
> > + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"
> > + ;;
> > + esac
> > +}
> > +
> > +# Get the file's list of on-disk blocks as a comma-separated list of block
> > +# offsets from the start of the device. "Blocks" are 512 bytes each here.
> > +_get_file_block_list()
> > +{
> > + local file=$1
> > +
> > + sync
> > + $XFS_IO_PROG -c fiemap $file | perl -ne '
> > + next if not /^\s*\d+: \[\d+\.\.\d+\]: (\d+)\.\.(\d+)/;
> > + print $_ . "," foreach $1..$2;' | sed 's/,$//'
> > +}
> > +
> > +# Dump a block list that was previously saved by _get_file_block_list().
> > +_dump_file_blocks()
> > +{
> > + local device=$1
> > + local blocklist=$2
> > + local block
> > +
> > + for block in $(tr ',' ' ' <<< $blocklist); do
> > + dd if=$device bs=512 count=1 skip=$block status=none
> > + done
> > +}
>
> Above two functions seem generic enough to be moved to common/rc

I feel that would be premature because common/rc is kind of bloated, and there's
a good chance these functions will only ever be used for encryption tests.
Normally, xfstests only test for user-visible behavior, so tests just 'cat' the
file contents, or 'ls' the filenames. The encryption tests are somewhat special
in that they really care about what's *actually* stored on-disk.

So I think that common/encrypt is the most logical location for now. But I
don't feel too strongly, and I can move it if you prefer.

Thanks for the review!

- Eric

2019-05-14 02:21:19

by Eryu Guan

[permalink] [raw]
Subject: Re: [RFC PATCH 4/7] common/encrypt: add helper for ciphertext verification tests

On Mon, May 13, 2019 at 12:12:05PM -0700, Eric Biggers wrote:
> Hi Eryu,
>
> On Sun, May 12, 2019 at 08:27:03PM +0800, Eryu Guan wrote:
> > > +# Retrieve the filename stored on-disk for the given file.
> > > +# The name is printed to stdout in binary.
> > > +_get_on_disk_filename()
> > > +{
> > > + local device=$1
> > > + local inode=$2
> > > + local dir_inode=$3
> > > +
> > > + case $FSTYP in
> > > + ext4)
> > > + # Extract the filename from the debugfs output line like:
> > > + #
> > > + # 131075 100644 (1) 0 0 0 22-Apr-2019 16:54 \xa2\x85\xb0z\x13\xe9\x09\x86R\xed\xdc\xce\xad\x14d\x19
> > > + #
> > > + $DEBUGFS_PROG $device -R "ls -l -r <$dir_inode>" \
> > > + 2>>$seqres.full | perl -ne '
> > > + next if not /^\s*'$inode'\s+/;
> > > + s/.*?\d\d:\d\d //;
> > > + chomp;
> > > + s/\\x([[:xdigit:]]{2})/chr hex $1/eg;
> > > + print;'
> > > + ;;
> > > + f2fs)
> > > + # Extract the filename from the dump.f2fs output line like:
> > > + #
> > > + # i_name [UpkzIPuts9by1oDmE+Ivfw]
> > > + #
> > > + # The name is base64-encoded, so we have to decode it here.
> > > + #
> > > + $DUMP_F2FS_PROG $device -i $inode | perl -ne '
> > > + next if not /^i_name\s+\[([A-Za-z0-9+,]+)\]/;
> > > + chomp $1;
> > > + my @chars = split //, $1;
> > > + my $ac = 0;
> > > + my $bits = 0;
> > > + my $table = join "", (A..Z, a..z, 0..9, "+", ",");
> > > + foreach (@chars) {
> > > + $ac += index($table, $_) << $bits;
> > > + $bits += 6;
> > > + if ($bits >= 8) {
> > > + print chr($ac & 0xff);
> > > + $ac >>= 8;
> > > + $bits -= 8;
> > > + }
> > > + }
> > > + if ($ac != 0) {
> > > + print STDERR "Invalid base64-encoded string!\n";
> > > + }'
> > > + ;;
> > > + *)
> > > + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"
> >
> > And looks like this function has nothing to do with fs encryption, move it to
> > common/rc?
>
> For ext4 that's true, but for f2fs the name is assumed to be base64-encoded,
> which f2fs-tools only does for encrypted filenames. I'll update the comment to
> clarify that the function assumes the directory is encrypted.
>
> >
> > > + ;;
> > > + esac
> > > +}
> > > +
> > > +# Require support for _get_on_disk_filename()
> > > +_require_get_on_disk_filename_support()
> > > +{
> > > + echo "Checking for _get_on_disk_filename() support for $FSTYP" >> $seqres.full
> > > + case $FSTYP in
> > > + ext4)
> > > + # Verify that the "ls -l -r" debugfs command is supported and
> > > + # hex-encodes non-ASCII characters, rather than using an
> > > + # ambiguous escaping method. This requires the e2fsprogs patch
> > > + # "debugfs: avoid ambiguity when printing filenames"
> > > + # (https://marc.info/?l=linux-ext4&m=155596495624232&w=2).
> > > + # TODO: once merged, list the minimum e2fsprogs version here.
> > > + _require_command "$DEBUGFS_PROG" debugfs
> > > + _scratch_mount
> > > + touch $SCRATCH_MNT/$'\xc1'
> > > + _scratch_unmount
> > > + if ! $DEBUGFS_PROG $SCRATCH_DEV -R "ls -l -r /" 2>&1 \
> > > + | tee -a $seqres.full | grep -E -q '\s+\\xc1\s*$'; then
> > > + _notrun "debugfs (e2fsprogs) is too old; doesn't support showing unambiguous on-disk filenames"
> > > + fi
> > > + ;;
> > > + f2fs)
> > > + # Verify that dump.f2fs shows encrypted filenames in full. This
> > > + # requires the patch "f2fs-tools: improve filename printing"
> > > + # (https://sourceforge.net/p/linux-f2fs/mailman/message/36648641/).
> > > + # TODO: once merged, list the minimum f2fs-tools version here.
> > > +
> > > + _require_command "$DUMP_F2FS_PROG" dump.f2fs
> > > + _require_command "$KEYCTL_PROG" keyctl
> > > + _scratch_mount
> > > + _new_session_keyring
> > > +
> > > + local keydesc=$(_generate_encryption_key)
> > > + local dir=$SCRATCH_MNT/test.${FUNCNAME[0]}
> > > + local file=$dir/$(perl -e 'print "A" x 255')
> > > + mkdir $dir
> > > + _set_encpolicy $dir $keydesc
> > > + touch $file
> > > + local inode=$(stat -c %i $file)
> > > +
> > > + _scratch_unmount
> > > + $KEYCTL_PROG clear @s
> > > +
> > > + # 255-character filename should result in 340 base64 characters.
> > > + if ! $DUMP_F2FS_PROG -i $inode $SCRATCH_DEV \
> > > + | grep -E -q '^i_name[[:space:]]+\[[A-Za-z0-9+,]{340}\]'; then
> > > + _notrun "dump.f2fs (f2fs-tools) is too old; doesn't support showing unambiguous on-disk filenames"
> > > + fi
> > > + ;;
> > > + *)
> > > + _notrun "_get_on_disk_filename() isn't implemented on $FSTYP"
> > > + ;;
> > > + esac
> > > +}
> > > +
> > > +# Get the file's list of on-disk blocks as a comma-separated list of block
> > > +# offsets from the start of the device. "Blocks" are 512 bytes each here.
> > > +_get_file_block_list()
> > > +{
> > > + local file=$1
> > > +
> > > + sync
> > > + $XFS_IO_PROG -c fiemap $file | perl -ne '
> > > + next if not /^\s*\d+: \[\d+\.\.\d+\]: (\d+)\.\.(\d+)/;
> > > + print $_ . "," foreach $1..$2;' | sed 's/,$//'
> > > +}
> > > +
> > > +# Dump a block list that was previously saved by _get_file_block_list().
> > > +_dump_file_blocks()
> > > +{
> > > + local device=$1
> > > + local blocklist=$2
> > > + local block
> > > +
> > > + for block in $(tr ',' ' ' <<< $blocklist); do
> > > + dd if=$device bs=512 count=1 skip=$block status=none
> > > + done
> > > +}
> >
> > Above two functions seem generic enough to be moved to common/rc
>
> I feel that would be premature because common/rc is kind of bloated, and there's
> a good chance these functions will only ever be used for encryption tests.
> Normally, xfstests only test for user-visible behavior, so tests just 'cat' the
> file contents, or 'ls' the filenames. The encryption tests are somewhat special
> in that they really care about what's *actually* stored on-disk.
>
> So I think that common/encrypt is the most logical location for now. But I

Yeah, that makes sense to me. Perhaps the functions should be renamed to
reflect that they're encryption-related?

Thanks,
Eryu

> don't feel too strongly, and I can move it if you prefer.
>
> Thanks for the review!
>
> - Eric