[openssl-users] [External] Re: Correct the check of RSA_FLAG_SIGN_VER
John.Eichenberger at Honeywell.com
Wed Apr 4 16:06:36 UTC 2018
Your answer #1 below presumes that RSA_new_method() is called AFTER RSA_set_method(). Is that a valid presumption? How is that documented as a requirement?
When the flag is set in a call to RSA_set_method() after a call to RSA_new(), the flag gets ignored with the current implementation.
Intermec by Honeywell
Principal Engineer: Sustaining Engineering
From: openssl-users [mailto:openssl-users-bounces at openssl.org] On Behalf Of Richard Levitte
Sent: Tuesday, April 03, 2018 7:43 PM
To: openssl-users at openssl.org
Subject: [External] Re: [openssl-users] Correct the check of RSA_FLAG_SIGN_VER
In message <BN6PR07MB3553B5B834436F685450D434F6A50 at BN6PR07MB3553.namprd07.prod.outlook.com> on Tue, 3 Apr 2018 22:55:06 +0000, "Eichenberger, John" <John.Eichenberger at Honeywell.com> said:
John.Eichenberger> Honeywell Internal
John.Eichenberger> I think I have an answer for why this commit needed to be reverted:
John.Eichenberger> Author: Dr. Stephen Henson <steve at openssl.org> John.Eichenberger> Author date: 2 years ago (12/20/2015 10:18:43 AM) John.Eichenberger> Commit date: 2 years ago (12/20/2015 11:27:03 AM) John.Eichenberger> Commit hash: 6656ba7152dfe4bba865e327dd362ea08544aa80
John.Eichenberger> Children: 1c7de36f62
John.Eichenberger> Parent(s): 17592f323a John.Eichenberger> John.Eichenberger> Don't check RSA_FLAG_SIGN_VER.
John.Eichenberger> Reviewed-by: Richard Levitte <levitte at openssl.org> John.Eichenberger> John.Eichenberger> The change made in that commit was to simply remove John.Eichenberger> the attempt to check for the RSA_FLAG_SIGN_VER flag.
John.Eichenberger> But that's not what is wrong with this code that John.Eichenberger> required changing. The change should be to add John.Eichenberger> "meth" prior to flags:
Well, not quite, actually. We can easily study the code prior to this change by looking at the 1.0.2 source.
1. in RSA_new_method(), which is used to create new instances of the
RSA structure, there's this line:
ret->flags = ret->meth->flags & ~RSA_FLAG_NON_FIPS_ALLOW;
So that makes the check of rsa->flags valid, no need to go via
2. In rsa.h (crypto/rsa/rsa.h in 1.0.2), you'll find this comment in
the middle of the definition of rsa_meth_st
* New sign and verify functions: some libraries don't allow arbitrary
* data to be signed/verified: this allows them to be used. Note: for
* this to work the RSA_public_decrypt() and RSA_private_encrypt() should
* *NOT* be used RSA_sign(), RSA_verify() should be used instead. Note:
* for backwards compatibility this functionality is only enabled if the
* RSA_FLAG_SIGN_VER option is set in 'flags'.
Do note the note on backward compatibility... you see, there were
versions of OpenSSL where the fields 'rsa_sign' and 'rsa_verify'
didn't exist (they appeared in OpenSSL 0.9.5), so for the sake of
allowing older applications to work with the newer OpenSSL without
recompilation, we required all new RSA method implementations to
use that flag to have the 'rsa_sign' and 'rsa_verify' functions
used. Without that flag, those functions were assumed not to
exist, that the RSA method structure was pre-0.9.5. However, this
was somewhere in 2000.
Fast forward to 2015, when we were starting to make certain types
opaque, and someone noticed this flag still hanging around, and we
figured that 0.9.5 is long gone, and 1.0.1 was a year away from its
end of life, and we figured that the reason to have this flag at
all was a matter of years long past, it was time to simply drop its
use. It had grown to become irrelevant a long time ago.
Richard Levitte levitte at openssl.org
OpenSSL Project http://www.openssl.org/~levitte/
openssl-users mailing list
To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-users
More information about the openssl-users