Closed Bug 1487150 Opened 6 years ago Closed 6 years ago

Akamai edge chooses TLS 1.3 draft-23 (symantec.com, ibm.com, and others affected)

Categories

(NSS :: Libraries, defect)

defect
Not set
normal

Tracking

(geckoview62 unaffected, firefox-esr52 unaffected, firefox-esr60 unaffected, firefox61 unaffected, firefox62 unaffected, firefox63+ fixed)

RESOLVED WORKSFORME
Tracking Status
geckoview62 --- unaffected
firefox-esr52 --- unaffected
firefox-esr60 --- unaffected
firefox61 --- unaffected
firefox62 --- unaffected
firefox63 + fixed

People

(Reporter: cpeterson, Unassigned)

References

Details

(Keywords: regression, Whiteboard: [sitewait])

[Tracking Requested - why for this release]:

Kai, I bisected this regression to NSS bug 1470914.

When loading https://www.symantec.com/ in today's Firefox 63 Nightly, I get the follow network error page:

  Secure Connection Failed
  An error occurred during a connection to www.symantec.com. SSL received a malformed Server Hello handshake message. Error code: SSL_ERROR_RX_MALFORMED_SERVER_HELLO 

I assume this network error is related to distrusting Symantec certs, but I would then expect to see the cert error page with the yellow warning border like https://www.paypal.com:

  Warning: Potential Security Risk Ahead.
  Error code: MOZILLA_PKIX_ERROR_ADDITIONAL_POLICY_CONSTRAINT_FAILED
Flags: needinfo?(kaie)
Works on nightly 2018-08-27, broken on 2018-08-29. It's in the NSS_3_39_BETA2 uplift, https://hg.mozilla.org/mozilla-central/rev/4a4b97e9087c
Looks like this belongs in NSS, as it happened between 5bc69334e84f and 0776cf816a6d in NSS. Most of those patches are yours, mt, can you take a look?
Assignee: nobody → nobody
Component: Security: PSM → Libraries
Flags: needinfo?(martin.thomson)
Product: Core → NSS
Target Milestone: --- → 3.39
Version: Trunk → trunk
This is a bug in the TLS 1.3 implementation on the server.  Gotta say, I didn't expect this one.

We send them a supported_versions extension (in my testing: 002b 0009 08 0304 0303 0302 0301, indicating TLS 1.0 to 1.3 inclusive).  They send one back with 0x7f17.  That's TLS 1.3 draft-23.

I suspect that what is happening is that they see TLS 1.3 and map that to the final version number internally, losing the distinction between draft and final.  So they assume that they understand the advertised version.  Then they write out the draft version.

(Separately, I realize that I have some work to do on downgrade, which was the first thought that both ekr and I had here.)
Flags: needinfo?(martin.thomson)
Blocks: 1483129
Let's try outreach with Symantec.
Flags: needinfo?(miket)
Whiteboard: [needscontact]
(In reply to Mike Taylor [:miketaylr] (62 Regression Engineering Owner) from comment #4)
> Let's try outreach with Symantec.

I've sent an email to our partner list -- will update here.
Flags: needinfo?(miket)
Whiteboard: [needscontact] → [sitewait]
Oops, working in parallel.  I reached out to someone at Symantec who tells me that they traced the issue to Akamai who have escalated the issue.
Summary: symantec.com Server Hello handshake Error code: SSL_ERROR_RX_MALFORMED_SERVER_HELLO → Akamai edge chooses TLS 1.3 draft-23 (symantec.com, ibm.com, and others affected)
Update after talking to Akamai: they have confirmed that OpenSSL appears to have a bug.  They are working on it.
Flags: needinfo?(kaie)
Akamai is disabling the TLS 1.3 beta this week, until we update to the official 1.1.1 OpenSSL release.
Status: NEW → RESOLVED
Closed: 6 years ago
Resolution: --- → WORKSFORME
Nightly 63.0 and as of last night 64.0 is working just fine and handles https://www.tls13.net/ perfectly.
You need to log in before you can comment on or make changes to this bug.