266651 - case sensitive comparison done on wildcard hostnames in certs

Status: RESOLVED FIXED

(NSS :: Libraries, defect, P2)

Description

[John David Galt]

User-Agent:       Mozilla/5.0 (Windows; U; Win98; rv:1.7.3) Gecko/20040913 Firefox/0.10
Build Identifier: Mozilla/5.0 (Windows; U; Win98; rv:1.7.3) Gecko/20040913 Firefox/0.10

This happened on a private site, so I'll have to "genericize" the domain name,
but:  When opening the page https://secure.myexampledomain.com/, I get a popup
saying that the site has presented an invalid certificate because the
certificate actually belongs to Secure.MyExampleDomain.com (the same domain name
except for case).

Reproducible: Always
Steps to Reproduce:

Actual Results:  
I hit "OK" to accept the certificate and was able to continue.

Expected Results:  
It should have accepted the certificate as valid without comment, since domain
names are inherently case blind.

Comment 1

[Ali Ebrahim]

Moving to NSS.

Comment 2

[Wan-Teh Chang]

Nelson, could you look at this?  Thanks.

Comment 3

[Nelson Bolyard (seldom reads bugmail)]

Here is a guess at an alternative explanation.  
Submittor, please tell us if this explanation is correct.

The cert probably has a domain name in the Common Name attribute of the 
SubjectName, and also one or more domain names as subject alternative names.
The domain name given in the Common Name attribute is probably NOT also 
listed among the subject alternative names.  The name being tested in the
URL is the name in the subject's CN, and not one of the alternative names.

If this explanation matches the facts, then this bug is invalid.  The 
software is working according to the RFC 2818, which states:

   If a subjectAltName extension of type dNSName is present, that MUST
   be used as the identity. Otherwise, the (most specific) Common Name
   field in the Subject field of the certificate MUST be used. 

Comment 4

[John David Galt]

I'm not in a position to know that.  I don't own the target site.

Comment 5

[Nelson Bolyard (seldom reads bugmail)]

You don't need to own the target site.  
You only need to get the cert from the target site and examine it.  

We need evidence to support the hypothesis.  
Otherwise, all we have is conjecture.  
In the absence of evidence of the real problem, we'll eventually
have to mark this bug invalid.  

Comment 6

[Ken Key]

Attachment: portreg.c fix for case insensitivity when CN contains a "shexp".

It is not stated in the description of this bug, but if the private certificate

the user had contained a wildcard, then what the user describes is exactly what

this patch fixes.

Comment 7

[Ken Key]

It is not stated, but if the certificate they were testing with contained a
wildcard, a bug in security/nss/lib/util/portreg.c will cause it to be
compared in a case-sensitive manner and fail.

This is happening in the real-world today at:

    https://xpennstatewrestlingclub.merchantquest.net/

who's certificate has a CN containing *.MerchantQuest.Net and firefox is
complaining about the mismatch.

To recreate, just create a server wildcard cert using mixed case and try
to access it from firefox.  Firefox complains.  Try a mixed case fqdn CN
and firefox accepts it.

This is against head-of-tree as of Nov. 11th, 2004.

Comment 8

[Ken Key]

Attachment: Real Diff against Head-Of-Tree

Sorry, old diff was against a sticky tag.  Real diff against Head-Of-Tree

Comment 9

[Ken Key]

By the way, we have notified https://xpennstatewrestlingclub.merchantquest.net/ of
the issue.  They may request a new cert that is all lowercase to work
around this problem, so my real-world example my not work by the time someone
else gets around to testing against it.  However, it is easily replicated on 
a private web server by using a wild-card and mixed case in the CN of a
cert.

Comment 10

[Nelson Bolyard (seldom reads bugmail)]

Ken, thanks for the patch!  Would you like me to add your name and email 
address to the list of contributors to that file?  If so, please indicate
the email address you'd like me to use.  

Comment 11

[Nelson Bolyard (seldom reads bugmail)]

Comment on attachment 165582 [details] [diff] [review]
Real Diff against Head-Of-Tree

Will check in when I get attribution feedback from patch contributor.

Comment 12

[Ken Key]

(In reply to comment #11)
> (From update of attachment 165582 [details] [diff] [review])
> Will check in when I get attribution feedback from patch contributor.
> 

Glad to finally be able to be in a position to give something back to
the project.  The little patch hardly seems worthy of an attribution,
but there is a coolness factor I can't resist.

  Ken Key <[e-mail]>

works for me.

Thanks,
K^2

Comment 13

[Nelson Bolyard (seldom reads bugmail)]

Bug is not platform dependent.
Fix tested.  About to checkin.

Comment 14

[Nelson Bolyard (seldom reads bugmail)]

Checking in portreg.c;
/cvsroot/mozilla/security/nss/lib/util/portreg.c,v  <--  portreg.c
new revision: 1.4; previous revision: 1.3

Thanks, Ken!