1529222 - Disable support for MD2 in S/MIME multipart; nsCMSMessage should use HASH_GetHashOidTagByHashType

Status: RESOLVED FIXED

(MailNews Core :: Security: S/MIME, enhancement)

Description

[Kai Engert (:KaiE:)]

The code landed in bug 1507218 added function nsCMSMessage::GetIntHashToOidHash which Bob thinks should be provided by a general function in NSS.

We added HASH_GetHashOidTagByHashType in bug 1526336.

Once we made that change, we should remove the dependency on symbols like nsICryptoHash::MD2 from TB, which bug 1526302 wants to remove.

This bug blocks the landing of bug 1526302.

Comment 1

[Kai Engert (:KaiE:)]

Dana, J.C., TB will require NSS function HASH_GetHashOidTagByHashType to be exported.

It's a very small mapping function. Can we export it into Firefox/Gecko, or do you prefer to avoid it, and have only TB export it?

Comment 2

[Dana Keeler (she/her) (use needinfo) [:keeler] (on leave)]

Exporting it from gecko seems fine, although where are we in terms of having thunderbird being able to export its own set of symbols?

Comment 3

[Kai Engert (:KaiE:)]

(In reply to Dana Keeler (she/her) (use needinfo) (:keeler for reviews) from comment #2)

although where are we in terms of having thunderbird being able to export its own set of symbols?

We are able to do that, we can use a file with "extra" exports. We just have to define a configuration variable in the TB build, and Gecko's nss.symbols file will include it, see bug 1529228 comment 1.

Comment 4

[Kai Engert (:KaiE:)]

I'm sneaking in the export of the API to bug 1529227.

Comment 5

[Kai Engert (:KaiE:)]

Attachment: 1529222-v1.patch

Minor cleanup that I promised to do in bug 1526302

Comment 6

[Kai Engert (:KaiE:)]

Comment on attachment 9047661 [details] [diff] [review]
1529222-v1.patch

needs revision

Comment 7

[Kai Engert (:KaiE:)]

The existing code still allowed the processing of incoming signed S/MIME messages that use the MD2 has algorithm for the message digest.

MD2 is insecure, we should stop supporting it.

(I'll file a separate bug to discuss if and when we should potentially disable MD5.)

Comment 8

[Kai Engert (:KaiE:)]

Attachment: 1529222-v2.patch

Comment 9

[Kai Engert (:KaiE:)]

FYI, I'm mixing the cleanup with the removal of MD2 support, because Firefox/PSM wants to disable support for MD2 as a hashing algorithm.

Comment 10

[Kai Engert (:KaiE:)]

The use of MD2 has been discouraged since 2004:
https://tools.ietf.org/html/rfc3850#page-3

Comment 11

[Kai Engert (:KaiE:)]

The attached patch disables parsing of the MD2 header in multipart S/MIME messages. NSS might support still allow MD2 signed messages in other places where signatures are verified. This will need separate checking.

Comment 12

[Magnus Melin [:mkmelin]]

Comment on attachment 9047665 [details] [diff] [review]
1529222-v2.patch

Review of attachment 9047665 [details] [diff] [review]:
-----------------------------------------------------------------

LGTM, r=mkmelin

Comment 13

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/e5731ae4eb09
Disable support for MD2 in S/MIME multipart; nsCMSMessage should use HASH_GetHashOidTagByHashType. r=mkmelin