977870 - insanity::pkix: consume the rest of input when a CertID doesn't match in an OCSP response

Status: RESOLVED FIXED

(Core :: Security: PSM, defect)

Description

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

When not using NSS's OCSP cache, the test cases "ocsp-stapling-good-other.example.com" and "ocsp-stapling-empty.example.com" in test_ocsp_stapling.js cause VerifyEncodedOCSPResponse to return SEC_ERROR_BAD_DER, which it shouldn't. It should (probably) instead return SEC_ERROR_OCSP_UNKNOWN_CERT or SEC_ERROR_OCSP_MALFORMED_RESPONSE, respectively.

Comment 1

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

Attachment: patch

It turns out, when using Nested or NestedOf, if the whole input isn't consumed, SEC_ERROR_BAD_DER is returned. In CertID, if the response doesn't match the certificate we're interested in, der::Success is returned, but input hasn't been consumed. So, SEC_ERROR_BAD_DER ultimately gets returned. This patch patch makes it so the input gets consumed in these cases.

Comment 2

[Brian Smith (:briansmith, :bsmith, use NEEDINFO?)]

Comment on attachment 8391473 [details] [diff] [review]
patch

Review of attachment 8391473 [details] [diff] [review]:
-----------------------------------------------------------------

It seems like we need a test for the case where there are two SingleResponses in the OCSP response and the first one doesn't match. I will add this as part of bug 916629.

Comment 3

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

I'm going to handle the other case where we return SEC_ERROR_BAD_DER in bug 984567.

Comment 4

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

Attachment: patch v1.1

This is just a subset of the previous patch - carrying over r+.

Comment 5

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/c2bc849c4169

Comment 6

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/c2bc849c4169