diff mbox series

X.509: Add missing IMPLICIT annotations to AKID ASN.1 module

Message ID be8ab09429d55c6cfc52ee0e43bf021ffb384152.1695720715.git.lukas@wunner.de (mailing list archive)
State New
Headers show
Series X.509: Add missing IMPLICIT annotations to AKID ASN.1 module | expand

Commit Message

Lukas Wunner Sept. 26, 2023, 9:46 a.m. UTC
The ASN.1 module in RFC 5280 appendix A.1 uses EXPLICIT TAGS whereas the
one in appendix A.2 uses IMPLICIT TAGS.

The kernel's simplified asn1_compiler.c always uses EXPLICIT TAGS, hence
definitions from appendix A.2 need to be annotated as IMPLICIT for the
compiler to generate RFC-compliant code.

In particular, GeneralName is defined in appendix A.2:

GeneralName ::= CHOICE {
        otherName                       [0] OtherName,
        ...
        dNSName                         [2] IA5String,
        x400Address                     [3] ORAddress,
        directoryName                   [4] Name,
        ...
        }

Because appendix A.2 uses IMPLICIT TAGS, the IA5String tag (0x16) of a
dNSName is not rendered.  Instead, the string directly succeeds the
[2] tag (0x82).

Likewise, the SEQUENCE tag (0x30) of an OtherName is not rendered.
Instead, only the constituents of the SEQUENCE are rendered:  An OID tag
(0x06), a [0] tag (0xa0) and an ANY tag.  That's three consecutive tags
instead of a single encompassing tag.

The situation is different for x400Address and directoryName choices:
They reference ORAddress and Name, which are defined in appendix A.1,
therefore use EXPLICIT TAGS.

The AKID ASN.1 module is missing several IMPLICIT annotations, hence
isn't RFC-compliant.  In the unlikely event that an AKID contains other
elements beside a directoryName, users may see parse errors.

Add the missing annotations but do not tag this commit for stable as I
am not aware of any issue reports.  Fixes are only eligible for stable
if they're "obviously correct" and with ASN.1 there's no such thing.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
---
Found this while bringing up PCI device authentication, which involves
validating the Subject Alternative Name in certificates.

I double-checked all ASN.1 modules in the tree and this seems to be
the only one affected by the issue.

 crypto/asymmetric_keys/x509_akid.asn1 | 24 +++++++++++++++++-------
 1 file changed, 17 insertions(+), 7 deletions(-)

Comments

Herbert Xu Oct. 5, 2023, 10:25 a.m. UTC | #1
On Tue, Sep 26, 2023 at 11:46:41AM +0200, Lukas Wunner wrote:
> The ASN.1 module in RFC 5280 appendix A.1 uses EXPLICIT TAGS whereas the
> one in appendix A.2 uses IMPLICIT TAGS.
> 
> The kernel's simplified asn1_compiler.c always uses EXPLICIT TAGS, hence
> definitions from appendix A.2 need to be annotated as IMPLICIT for the
> compiler to generate RFC-compliant code.
> 
> In particular, GeneralName is defined in appendix A.2:
> 
> GeneralName ::= CHOICE {
>         otherName                       [0] OtherName,
>         ...
>         dNSName                         [2] IA5String,
>         x400Address                     [3] ORAddress,
>         directoryName                   [4] Name,
>         ...
>         }
> 
> Because appendix A.2 uses IMPLICIT TAGS, the IA5String tag (0x16) of a
> dNSName is not rendered.  Instead, the string directly succeeds the
> [2] tag (0x82).
> 
> Likewise, the SEQUENCE tag (0x30) of an OtherName is not rendered.
> Instead, only the constituents of the SEQUENCE are rendered:  An OID tag
> (0x06), a [0] tag (0xa0) and an ANY tag.  That's three consecutive tags
> instead of a single encompassing tag.
> 
> The situation is different for x400Address and directoryName choices:
> They reference ORAddress and Name, which are defined in appendix A.1,
> therefore use EXPLICIT TAGS.
> 
> The AKID ASN.1 module is missing several IMPLICIT annotations, hence
> isn't RFC-compliant.  In the unlikely event that an AKID contains other
> elements beside a directoryName, users may see parse errors.
> 
> Add the missing annotations but do not tag this commit for stable as I
> am not aware of any issue reports.  Fixes are only eligible for stable
> if they're "obviously correct" and with ASN.1 there's no such thing.
> 
> Signed-off-by: Lukas Wunner <lukas@wunner.de>
> ---
> Found this while bringing up PCI device authentication, which involves
> validating the Subject Alternative Name in certificates.
> 
> I double-checked all ASN.1 modules in the tree and this seems to be
> the only one affected by the issue.
> 
>  crypto/asymmetric_keys/x509_akid.asn1 | 24 +++++++++++++++++-------
>  1 file changed, 17 insertions(+), 7 deletions(-)

Patch applied.  Thanks.
diff mbox series

Patch

diff --git a/crypto/asymmetric_keys/x509_akid.asn1 b/crypto/asymmetric_keys/x509_akid.asn1
index 1a33231..c7818ff 100644
--- a/crypto/asymmetric_keys/x509_akid.asn1
+++ b/crypto/asymmetric_keys/x509_akid.asn1
@@ -14,15 +14,15 @@  CertificateSerialNumber ::= INTEGER ({ x509_akid_note_serial })
 GeneralNames ::= SEQUENCE OF GeneralName
 
 GeneralName ::= CHOICE {
-	otherName			[0] ANY,
-	rfc822Name			[1] IA5String,
-	dNSName				[2] IA5String,
+	otherName			[0] IMPLICIT OtherName,
+	rfc822Name			[1] IMPLICIT IA5String,
+	dNSName				[2] IMPLICIT IA5String,
 	x400Address			[3] ANY,
 	directoryName			[4] Name ({ x509_akid_note_name }),
-	ediPartyName			[5] ANY,
-	uniformResourceIdentifier	[6] IA5String,
-	iPAddress			[7] OCTET STRING,
-	registeredID			[8] OBJECT IDENTIFIER
+	ediPartyName			[5] IMPLICIT EDIPartyName,
+	uniformResourceIdentifier	[6] IMPLICIT IA5String,
+	iPAddress			[7] IMPLICIT OCTET STRING,
+	registeredID			[8] IMPLICIT OBJECT IDENTIFIER
 	}
 
 Name ::= SEQUENCE OF RelativeDistinguishedName
@@ -33,3 +33,13 @@  AttributeValueAssertion ::= SEQUENCE {
 	attributeType		OBJECT IDENTIFIER ({ x509_note_OID }),
 	attributeValue		ANY ({ x509_extract_name_segment })
 	}
+
+OtherName ::= SEQUENCE {
+	type-id			OBJECT IDENTIFIER,
+	value			[0] ANY
+	}
+
+EDIPartyName ::= SEQUENCE {
+	nameAssigner		[0] ANY OPTIONAL,
+	partyName		[1] ANY
+	}