diff mbox series

[v2] common/encrypt: allow the use of 'fscrypt:' as key prefix

Message ID 20220404102554.6616-1-lhenriques@suse.de (mailing list archive)
State New, archived
Headers show
Series [v2] common/encrypt: allow the use of 'fscrypt:' as key prefix | expand

Commit Message

Luis Henriques April 4, 2022, 10:25 a.m. UTC
fscrypt keys have used the $FSTYP as prefix.  However this format is being
deprecated -- newer kernels already allow the usage of the generic
'fscrypt:' prefix for ext4 and f2fs.  This patch allows the usage of this
new prefix for testing filesystems that have never supported the old
format, but keeping the $FSTYP prefix for filesystems that support it, so
that old kernels can be tested.

Signed-off-by: Luís Henriques <lhenriques@suse.de>
---
 common/encrypt | 35 ++++++++++++++++++++++++-----------
 1 file changed, 24 insertions(+), 11 deletions(-)

Changes since v1:
- ubifs now follows into the default case (i.e. to use the 'fscrypt' key
  prefix)
- dropped local variable from _get_fs_keyprefix()

Comments

Eric Biggers April 5, 2022, 12:23 a.m. UTC | #1
The code looks fine, but the explanation needs some tweaks:

On Mon, Apr 04, 2022 at 11:25:54AM +0100, Luís Henriques wrote:
> fscrypt keys have used the $FSTYP as prefix.  However this format is being
> deprecated -- newer kernels already allow the usage of the generic
> 'fscrypt:' prefix for ext4 and f2fs.  This patch allows the usage of this
> new prefix for testing filesystems that have never supported the old
> format, but keeping the $FSTYP prefix for filesystems that support it, so
> that old kernels can be tested.

This explanation is inconsistent with the code, which uses FSTYP for only ext4
and f2fs, and fscrypt for everything else including ubifs.

A better explanation would be something like "Only use $FSTYP on filesystems
that never supported the 'fscrypt' prefix, i.e. ext4 and f2fs."

> +# Keys are named $FSTYP:KEYDESC where KEYDESC is the 16-character key descriptor
> +# hex string.  Newer kernels (ext4 4.8 and later, f2fs 4.6 and later) also allow
> +# the common key prefix "fscrypt:" in addition to their filesystem-specific key
> +# prefix ("ext4:", "f2fs:").  It would be nice to use the common key prefix, but
> +# for now use the filesystem- specific prefix for these 2 filesystems to make it
> +# possible to test older kernels, and the "fscrypt" prefix for anything else.
> +_get_fs_keyprefix()

The first part of this comment sort of implies that FSTYP is the default and
"fscrypt" is the exception, but it should be the other way around.

How about:

# When fscrypt keys are added using the legacy mechanism (process-subscribed
# keyrings rather than filesystem keyrings), they are normally named
# "fscrypt:KEYDESC" where KEYDESC is the 16-character key descriptor hex string.
# However, ext4 and f2fs didn't add support for the "fscrypt" prefix until
# kernel v4.8 and v4.6, respectively.  Before that, they used "ext4" and "f2fs",
# respectively.  To allow testing ext4 and f2fs encryption on kernels older than
# this, we use these filesystem-specific prefixes for ext4 and f2fs.
Luis Henriques April 5, 2022, 8:53 a.m. UTC | #2
Eric Biggers <ebiggers@kernel.org> writes:

> The code looks fine, but the explanation needs some tweaks:
>
> On Mon, Apr 04, 2022 at 11:25:54AM +0100, Luís Henriques wrote:
>> fscrypt keys have used the $FSTYP as prefix.  However this format is being
>> deprecated -- newer kernels already allow the usage of the generic
>> 'fscrypt:' prefix for ext4 and f2fs.  This patch allows the usage of this
>> new prefix for testing filesystems that have never supported the old
>> format, but keeping the $FSTYP prefix for filesystems that support it, so
>> that old kernels can be tested.
>
> This explanation is inconsistent with the code, which uses FSTYP for only ext4
> and f2fs, and fscrypt for everything else including ubifs.
>
> A better explanation would be something like "Only use $FSTYP on filesystems
> that never supported the 'fscrypt' prefix, i.e. ext4 and f2fs."
>
>> +# Keys are named $FSTYP:KEYDESC where KEYDESC is the 16-character key descriptor
>> +# hex string.  Newer kernels (ext4 4.8 and later, f2fs 4.6 and later) also allow
>> +# the common key prefix "fscrypt:" in addition to their filesystem-specific key
>> +# prefix ("ext4:", "f2fs:").  It would be nice to use the common key prefix, but
>> +# for now use the filesystem- specific prefix for these 2 filesystems to make it
>> +# possible to test older kernels, and the "fscrypt" prefix for anything else.
>> +_get_fs_keyprefix()
>
> The first part of this comment sort of implies that FSTYP is the default and
> "fscrypt" is the exception, but it should be the other way around.
>
> How about:
>
> # When fscrypt keys are added using the legacy mechanism (process-subscribed
> # keyrings rather than filesystem keyrings), they are normally named
> # "fscrypt:KEYDESC" where KEYDESC is the 16-character key descriptor hex string.
> # However, ext4 and f2fs didn't add support for the "fscrypt" prefix until
> # kernel v4.8 and v4.6, respectively.  Before that, they used "ext4" and "f2fs",
> # respectively.  To allow testing ext4 and f2fs encryption on kernels older than
> # this, we use these filesystem-specific prefixes for ext4 and f2fs.

Doh!  Yes, of course I need to adjust the documentation.  Sorry, I'll send
v3 shortly.  Thanks!

Cheers,
diff mbox series

Patch

diff --git a/common/encrypt b/common/encrypt
index f90c4ef05a3f..6dae7708d52b 100644
--- a/common/encrypt
+++ b/common/encrypt
@@ -250,6 +250,24 @@  _num_to_hex()
 	fi
 }
 
+# Keys are named $FSTYP:KEYDESC where KEYDESC is the 16-character key descriptor
+# hex string.  Newer kernels (ext4 4.8 and later, f2fs 4.6 and later) also allow
+# the common key prefix "fscrypt:" in addition to their filesystem-specific key
+# prefix ("ext4:", "f2fs:").  It would be nice to use the common key prefix, but
+# for now use the filesystem- specific prefix for these 2 filesystems to make it
+# possible to test older kernels, and the "fscrypt" prefix for anything else.
+_get_fs_keyprefix()
+{
+	case $FSTYP in
+	ext4|f2fs)
+		echo $FSTYP
+		;;
+	*)
+		echo fscrypt
+		;;
+	esac
+}
+
 # Add the specified raw encryption key to the session keyring, using the
 # specified key descriptor.
 _add_session_encryption_key()
@@ -268,18 +286,11 @@  _add_session_encryption_key()
 	#	};
 	#
 	# The kernel ignores 'mode' but requires that 'size' be 64.
-	#
-	# Keys are named $FSTYP:KEYDESC where KEYDESC is the 16-character key
-	# descriptor hex string.  Newer kernels (ext4 4.8 and later, f2fs 4.6
-	# and later) also allow the common key prefix "fscrypt:" in addition to
-	# their filesystem-specific key prefix ("ext4:", "f2fs:").  It would be
-	# nice to use the common key prefix, but for now use the filesystem-
-	# specific prefix to make it possible to test older kernels...
-	#
 	local mode=$(_num_to_hex 0 4)
 	local size=$(_num_to_hex 64 4)
+	local prefix=$(_get_fs_keyprefix)
 	echo -n -e "${mode}${raw}${size}" |
-		$KEYCTL_PROG padd logon $FSTYP:$keydesc @s >>$seqres.full
+		$KEYCTL_PROG padd logon $prefix:$keydesc @s >>$seqres.full
 }
 
 #
@@ -302,7 +313,8 @@  _generate_session_encryption_key()
 _unlink_session_encryption_key()
 {
 	local keydesc=$1
-	local keyid=$($KEYCTL_PROG search @s logon $FSTYP:$keydesc)
+	local prefix=$(_get_fs_keyprefix)
+	local keyid=$($KEYCTL_PROG search @s logon $prefix:$keydesc)
 	$KEYCTL_PROG unlink $keyid >>$seqres.full
 }
 
@@ -310,7 +322,8 @@  _unlink_session_encryption_key()
 _revoke_session_encryption_key()
 {
 	local keydesc=$1
-	local keyid=$($KEYCTL_PROG search @s logon $FSTYP:$keydesc)
+	local prefix=$(_get_fs_keyprefix)
+	local keyid=$($KEYCTL_PROG search @s logon $prefix:$keydesc)
 	$KEYCTL_PROG revoke $keyid >>$seqres.full
 }