[krbdev.mit.edu #7596] [Comment] PKINIT should allow missing DH param Q

[Greg Hudson via RT]

http://kerborg-prod-app-1.mit.edu/rt/Ticket/Display.html?id=7596
This is a comment.  It is not sent to the Requestor(s):

Ticket 8398 (commit e5c77a11341a79e6af1e5aef7c587a5b75a9e378, "Add PKINIT
support for OpenSSL 1.1.0", going into release 1.15) attempted to preserve
this workaround using a custom ASN.1 type with an optional q parameter. 
However, the same commit added a helper dup_dh_params() with error-checking on
the results of BN_dup(), replacing the old code which had no error-checking. 
This error-checking would erroneously trigger if q is unset in the DH
parameters, causing server_process_dh() and
therefore pkinit_server_return_padata() to return ENOMEM.

Since no one has reported the regression to my knowledge, I would guess that
Microsoft fixed its bug and the workaround is no longer needed.  I have asked
Microsoft dochelp for clarification of which versions omit the q value.

If we need to restore the workaround, dup_dh_params() can check for (oldq !=
NULL && q == NULL) instead of q == NULL.  Also, we could do the parsing more
simply in OpenSSL 1.1 by trying d2i_DHxparams() and falling back to
d2i_DHparams().  The latter function will read domain parameters using the
PKCS#3 DHParameter type, which has no q value.  (OpenSSL 1.0 does not have
DHxparams support, so we're stuck with custom encoding and decoding to handle
the RFC 3279 type.)