1132540 - coastcapitalsavings.com displays ssl_error_no_cypher_overlap error

Status: VERIFIED FIXED

(Web Compatibility :: Site Reports, defect)

Description

[u279076]

There is a recent regression in Firefox when loading coastcapitalsavings.com. The page used to load fine but in a recent Nightly it now just displays an SSL error. I tried going back to the January 15, 2015 Nightly and the page loads fine. 

I will see if I can track down the regression window.

Comment 1

[:Cykesiopka]

This is a TLS intolerance issue: https://www.ssllabs.com/ssltest/analyze.html?d=www.coastcapitalsavings.com&s=208.69.252.179

In particular, see this line:
> TLS version intolerance 	TLS 1.1  TLS 1.2  TLS 1.3  TLS 1.98  TLS 2.98   PROBLEMATIC

FWIW, this site also appears to be RC4 only.

Comment 2

[u279076]

Last good revision: 2cb22c058add
First bad revision: 3094601af679
Pushlog: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=2cb22c058add&tochange=3094601af679

Perhaps this could be caused by bug 1126413?

Comment 3

[u279076]

(In reply to Cykesiopka from comment #1)
> This is a TLS intolerance issue:
> https://www.ssllabs.com/ssltest/analyze.html?d=www.coastcapitalsavings.
> com&s=208.69.252.179
> 
> In particular, see this line:
> > TLS version intolerance 	TLS 1.1  TLS 1.2  TLS 1.3  TLS 1.98  TLS 2.98   PROBLEMATIC
> 
> FWIW, this site also appears to be RC4 only.

Does that mean this is an issue that Coast Capital Savings needs to fix on their end?

Comment 4

[:Cykesiopka]

(In reply to Anthony Hughes, QA Mentor (:ashughes) from comment #3)
> (In reply to Cykesiopka from comment #1)
> > This is a TLS intolerance issue:
> > https://www.ssllabs.com/ssltest/analyze.html?d=www.coastcapitalsavings.
> > com&s=208.69.252.179
> > 
> > In particular, see this line:
> > > TLS version intolerance 	TLS 1.1  TLS 1.2  TLS 1.3  TLS 1.98  TLS 2.98   PROBLEMATIC
> > 
> > FWIW, this site also appears to be RC4 only.
> 
> Does that mean this is an issue that Coast Capital Savings needs to fix on
> their end?

Well, I withdraw my assertion that this is because of TLS intolerance (it may or may not be), but yes, the issues need to be fixed on the server end eventually.

Comment 5

[u279076]

Okay, I will reach out to their webmaster to inform them of this issue.

Comment 6

[Masatoshi Kimura [:emk]]

Bug 1126413 shouldn't change the behavior. Maybe bug 1124039? Anyway, the site should fix the issue as  Cykesiopka said.

Comment 7

[Kyle Repinski]

This is an problem with my bank's website as well; for the moment I'll be adding them to security.tls.insecure_fallback_hosts in addition to letting them know about the issue.

Comment 8

[u279076]

[Tracking Requested - why for this release]: Nominating for tracking since this could become a larger support issue as Firefox 38 gets closer to release, assuming multiple banks are affected and don't fix it in time.

Comment 9

[Lukas Blakk [:lsblakk] use ?needinfo]

We definitely want to keep an eye on this and it will also impact ESR.  Matt - is this something sec QA  can work on, pre-check popular banking websites so we have an idea of what to expect and can get ahead of some Tech Evangelism?

Comment 10

[Matt Wobensmith [:mwobensmith][:matt:]]

Bug 1124039 looks suspect, as the site here attempts to connect with TLS 1.0 and the RC4 cypher. Haven't read all of the related bugs there, so I can't say for sure.

As with any TLS/SSL change, this requires compatibility testing. However, I think we first should find out if bug 1124039 indeed is the issue. If so, I'll focus my testing and results in that bug and we can make this one a dependency. 

Masatoshi, can you confirm? Thank you.

Comment 11

[Masatoshi Kimura [:emk]]

I can't connect www.coastcapitalsavings.com without enabling RC4 even if the fallback limit is 1.
I can connect www.coastcapitalsavings.com with RC4 enabled even if the fallback limit is 3.
Bug 1124039 will disable RC4 unless the site is whitelisted. So definitely RC4 is the issue, not a TLS intolerant issue. (I don't know why SSL Labs claims the TLS intolerance, but SSL Labs is not always accurate.)

Comment 12

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

(In reply to Masatoshi Kimura [:emk] from comment #11)
> (I don't know why SSL Labs
> claims the TLS intolerance, but SSL Labs is not always accurate.)

SSL Labs uses 0x0303 for both the record-level version number and client_version for its ClientHello, like MSIE and Java. NSS uses 0x0301 for the record layer and 0x0303 for client_version. According to one study, the SSL Labs way has ~5% more intolerance than the NSS way.

Also, some other aspects of the SSL Labs ClientHello may be different.

Comment 13

[Masatoshi Kimura [:emk]]

The site uses TLS_ECDHE_RSA_WITH_AES_256_CBC_SHA now.

Comment 14

[u279076]

Verified fixed with Firefox Nightly 39.0a1 20150302030204.

Comment 15

[u279076]

Cleaning up the flags on this bug since it was a server-side issue and not a Firefox bug.