[openssl-users] proposed changes to ruby-openssl create_extension

[Michael Richardson]

Like Bob Moskowitz who has been posting about IDevID, I have also been
creating certificates with custom/private extensions in aid of creating
IDevIDs.

I'm one of the authors of both:
    https://datatracker.ietf.org/doc/draft-ietf-anima-bootstrapping-keyinfra/
and https://datatracker.ietf.org/doc/draft-ietf-anima-voucher/

I want to create certificates with custom/private extensions programatically,
and I found it very difficult to do using the current ruby-openssl modules.

I was sure that it must be possible from the C-API, and found that this
following change to ruby's interface worked well for me:
    https://github.com/ruby/openssl/pull/141


    } I haven't found a seperate openssl-ruby list as yet, so please  {
    } excuse the BCC, and as this is not a security issue, I haven't  {
    } used that address.  Please redirect me.                         {


The critical change being:
-    ext = X509V3_EXT_nconf_nid(conf, ctx, nid, RSTRING_PTR(valstr));
+    ext = X509V3_EXT_nconf(conf, ctx, RSTRING_PTR(oid), RSTRING_PTR(value));

Because EXT_nconf does all the nid lookups, and processes the generics
itself, no point in the ruby-ssl code limiting itself to OIDs predefined
in objects.h by it's use of nid lookups directly.


    ** I'm asking the Openssl team if I'm using the API reasonably **
    ** Clearly, I have some possible garbage that has leaked in    **

My code looks like:

    @idevid.add_extension(ef.create_extension(
                          "subjectAltName",
                          sprintf("otherName:1.3.6.1.4.1.46930.1;UTF8:%s",
                                   self.sanitized_eui64),
                          false))

which let me put a private extension having my private hardware number into
the SAN.  That works just great, I think.  I have in fact extracted the
result in metdtls code in my constrained device a few months ago.

After my changes above, I could now also do: (46930 being my PEN)

@idevid.add_extension(ef.create_extension(
                      "1.3.6.1.4.1.46930.2",
                      "ASN1:UTF8String:http://www.sandelman.ca",
                      false))

which would insert a MASAURL (using a PEN until we get an id-pe OID
allocated) as described in: https://tools.ietf.org/html/draft-ietf-anima-bootstrapping-keyinfra-07#section-2.2

Of concern is that when I look at the resulting certificate:

dooku-[fountain/spec/certs](2.3.0) mcr 10006 %openssl x509 -noout -text -in
12-00-00-66-4D-02.crt
Certificate:
...
            X509v3 Subject Alternative Name:
                othername:<unsupported>
            1.3.6.1.4.1.46930.2:
                ..http://www.sandelman.ca

Looking at a hexdump I see "0x0c" and "0x17" prior to the http, but maybe
it's a length or something.... I wondered if there was garbage or a UTF-8 BOM
or something inserted..  so, I pointed asn1parse at the result, and I see:

400:d=5  hl=2 l=   9 prim: OBJECT            :1.3.6.1.4.1.46930.2
411:d=5  hl=2 l=  25 prim: OCTET STRING      [HEX DUMP]:0C17687474703A2F2F7777772E73616E64656C6D616E2E6361

So the 0xc and 0x17 are really there.

Clearly, it's not being take as an UTF8String (because I would expect to see
UTF8STRING as the type if it were), yet the ASN1:UTF8String is in fact being
processed somehow.

If you want to see the whole certificate result, as sample/test data to my
Registrar:
  https://github.com/mcr/fountain/blob/master/spec/certs/12-00-00-66-4D-02.crt

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr at sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 487 bytes
Desc: not available
URL: <http://mta.openssl.org/pipermail/openssl-users/attachments/20170828/69d0c641/attachment.sig>