On Fri, Sep 12, 2025 at 02:14:49PM -0700, Fan Wu wrote: > On Fri, Sep 12, 2025 at 6:14 AM Lukas Wunner <lukas@xxxxxxxxx> wrote: > > On Thu, Sep 11, 2025 at 10:53:56PM +0000, wufan@xxxxxxxxxx wrote: > > > +++ b/crypto/asymmetric_keys/x509_cert_parser.c > > > @@ -623,7 +625,7 @@ int x509_process_extension(void *context, size_t hdrlen, > > > if (v[0] != (ASN1_CONS_BIT | ASN1_SEQ)) > > > return -EBADMSG; > > > if (vlen < 2) > > > return -EBADMSG; > > > if (v[1] != vlen - 2) > > > return -EBADMSG; > > > - if (vlen >= 4 && v[1] != 0 && v[2] == ASN1_BOOL && v[3] == 1) > > > + if (vlen >= 5 && v[1] != 0 && v[2] == ASN1_BOOL && v[3] == 1 && v[4] != 0) > > > ctx->cert->pub->key_eflags |= 1 << KEY_EFLAG_CA; > > > return 0; > > > } > > > > Your patch is correct, however the conditions ... > > > > vlen >= 5 && v[1] != 0 && v[2] == ASN1_BOOL && v[3] == 1 > > > > ... all check well-formedness of the BasicConstraints object, > > so it seems if any of those checks fails, -EBADMSG should be returned. > > > > The check "if (vlen < 2)" could be changed to "if (vlen < 5)" because > > 5 bytes seems to be the minimum size of a well-formed BasicConstraints > > object. Then the "vlen >= 5" and "v[1] != 0" checks can be dropped. > > Actually, we need to be careful here. OpenSSL produces > BasicConstraints with CA:FALSE as just an empty SEQUENCE: > > 06 03 55 1d 13 | 01 01 ff | 04 02 | 30 00 > [----OID------] [critical] [OCTET] [empty SEQ] I see, thanks for the explanation. This behavior of OpenSSL doesn't seem spec-compliant, or is it? RFC 5280 sec 4.2.1.9 says the pathLenConstraint is optional, but the cA boolean is not optional. Is there a rule that booleans need not be rendered if they are false? BTW, I note that X.690 sec 11.1 says that for DER encoding, all bits of a "true" boolean must be set, hence the 0xff value. But I'm fine with your more permissive approach which checks for a non-zero value, hence also allows BER encoding per X.690 sec 8.2.2. Thanks! Lukas
