[PATCH] 0051 Lightweight CAs: lookup correct issuer for OCSP responses
Hi all, The attached patch makes sure that the right authority is used to create OCSP responses. Note that OCSP requests may ask about certs from more than one issuer - even though this is crazy the heuristic used is to simply use issuer of the first CertID in the request. Note that OCSP response validation of certificates issued by sub-CAs currently fails due to a separate issue[1]. [1] https://fedorahosted.org/pki/ticket/1632
Well, it would help to attach the patch :) On Thu, Oct 01, 2015 at 10:43:51PM +1000, Fraser Tweedale wrote:
Hi all,
The attached patch makes sure that the right authority is used to create OCSP responses. Note that OCSP requests may ask about certs from more than one issuer - even though this is crazy the heuristic used is to simply use issuer of the first CertID in the request.
Note that OCSP response validation of certificates issued by sub-CAs currently fails due to a separate issue[1].
[1] https://fedorahosted.org/pki/ticket/1632
_______________________________________________ Pki-devel mailing list Pki-devel@redhat.com https://www.redhat.com/mailman/listinfo/pki-devel
Couple of comments .. 1. First off, there is a typo in the comments on the method. I think you mean .. 3. Either we WERE the issuing CA, or we .. rather than "were not" 2. We can go with the heuristic of taking the first CA, but I do not think we should leak information about other certs if the CA is incorrect. The way the code is now, we will still return data on whether a particular cert serial number is valid -- even if that cert was not issued on that CA. A simple solution is to simply pass code to processRequest() to ignore the request if the issuer is not correct and not return a response for that request. Ade On Thu, 2015-10-01 at 22:51 +1000, Fraser Tweedale wrote:
Well, it would help to attach the patch :)
On Thu, Oct 01, 2015 at 10:43:51PM +1000, Fraser Tweedale wrote:
Hi all,
The attached patch makes sure that the right authority is used to create OCSP responses. Note that OCSP requests may ask about certs from more than one issuer - even though this is crazy the heuristic used is to simply use issuer of the first CertID in the request.
Note that OCSP response validation of certificates issued by sub -CAs currently fails due to a separate issue[1].
[1] https://fedorahosted.org/pki/ticket/1632
_______________________________________________ Pki-devel mailing list Pki-devel@redhat.com https://www.redhat.com/mailman/listinfo/pki-devel
Pki-devel mailing list Pki-devel@redhat.com https://www.redhat.com/mailman/listinfo/pki-devel
On Mon, Feb 22, 2016 at 12:02:49PM -0500, Ade Lee wrote:
Couple of comments ..
1. First off, there is a typo in the comments on the method. I think you mean ..
3. Either we WERE the issuing CA, or we .. rather than "were not"
2. We can go with the heuristic of taking the first CA, but I do not think we should leak information about other certs if the CA is incorrect. The way the code is now, we will still return data on whether a particular cert serial number is valid -- even if that cert was not issued on that CA.
A simple solution is to simply pass code to processRequest() to ignore the request if the issuer is not correct and not return a response for that request.
RFC 6960 says: The response MUST include a SingleResponse for each certificate in the request. So the best we can do is return 'unknown' status in this case. I've attached updated patch 0051-2 - the only change is the comment fixup - and two new patches: 0074 refactors digest lookup and adds support for SHA-2 algos, and 0075 changes the OCSP behaviour to return 'unknown' cert status for certs that from a different issuer. Cheers, Fraser
After clarification and discussions with cfu, ACK. On Wed, 2016-03-02 at 15:13 +1000, Fraser Tweedale wrote:
On Mon, Feb 22, 2016 at 12:02:49PM -0500, Ade Lee wrote:
Couple of comments ..
1. First off, there is a typo in the comments on the method. I think you mean ..
3. Either we WERE the issuing CA, or we .. rather than "were not"
2. We can go with the heuristic of taking the first CA, but I do not think we should leak information about other certs if the CA is incorrect. The way the code is now, we will still return data on whether a particular cert serial number is valid -- even if that cert was not issued on that CA.
A simple solution is to simply pass code to processRequest() to ignore the request if the issuer is not correct and not return a response for that request.
RFC 6960 says:
The response MUST include a SingleResponse for each certificate in the request.
So the best we can do is return 'unknown' status in this case.
I've attached updated patch 0051-2 - the only change is the comment fixup - and two new patches: 0074 refactors digest lookup and adds support for SHA-2 algos, and 0075 changes the OCSP behaviour to return 'unknown' cert status for certs that from a different issuer.
Cheers, Fraser
On Thu, Mar 03, 2016 at 03:35:39PM -0500, Ade Lee wrote:
After clarification and discussions with cfu, ACK.
Thanks! Pushed to master: - afe1d7205ae32c272e15dbf42475da4a79b5c9bc Lightweight CAs: lookup correct issuer for OCSP responses - 04214b3d3405750cbbda228554c0d9f087a59170 Move OCSP digest name lookup to CertID class - c0c1834465438844ff542514127b80b568c1afd8 Do not leak status of certs issued by other CAs I also filed a ticket for this feature: https://fedorahosted.org/pki/ticket/2230 Cheers, Fraser
participants (2)
-
Ade Lee -
Fraser Tweedale