1134437 - Delay move to PFS cipher suites

Status: RESOLVED FIXED

(Core :: WebRTC: Networking, defect)

Description

[msander]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/40.0.2214.111 Safari/537.36

Steps to reproduce:

The move to support only PFS cipher suites in Fx 37 will cause interoperability problems for 0penTok mobile clients using older versions of webrtc.org code.

In addition to supporting browsers like FireFox and Chrome, we also have a number of partners who use WebRTC using native applications built on top of native mobile SDKs. These applications routinely rely on interoperability with browsers as part of their use-case.

The versions of mobile SDKs out in the wild are built on top of the webrtc.org code-base that is unable to do ECDHE suites and in turn cause interoperaibility issues when FF is not able to act as a server for DHE suites.

Both FF37 and Chrome interoperability has been resolved by a) https://bugzilla.mozilla.org/show_bug.cgi?id=1052610 and b) https://code.google.com/p/chromium/issues/detail?id=406458 respectively.

We are in the process of updating our mobile SDKs to support using PFS cipher suites but it usually takes more than 6 weeks for the average mobile application update cycle. As it stands today, interoperability with FireFox 37 will be broken once FF37 moves to production - https://wiki.mozilla.org/RapidRelease/Calendar#Future_branch_dates. 

We were wondering if support for non-PFS ciphers can be enabled in FF37 and PFS support pushed to FF38. This would help increase the percentage of end-users using mobile applications to receive the update in time to prevent interoperability issues with FireFox when FF38 lands into production.

Comment 1

[Adam Roach [:abr]]

Tagging EKR and Brad, since they've given this some thought.

Comment 2

[Eric Rescorla (:ekr)]

My interest is in not delaying the transition to PFS-only indefinitely.

If we land https://bugzilla.mozilla.org/show_bug.cgi?id=102794 will there
be any objection to turning this on?

Comment 3

[Brad Lassey [:blassey] (use needinfo?)]

I am OK with pushing this off to 38, but not any further than that (it is an ESR release).

Comment 4

[Maire Reavy [:mreavy]]

Martin -- Would you have time to write a patch for this bug (to delay enforcing cipher suites in Fx37 only)?   I'm copying Richard who I'm hoping has the time to review such a patch when it's ready.  

FYI: Fx37 uplifts to Beta next week, and I believe Beta-build1 "goes to build" on Tuesday/Wednesday.  We don't have to land this in time for build1 of Beta (this isn't a fire drill), but it would be nice.

Comment 5

[msander]

Delaying this until Fx38 will be really helpful, thank you.

Comment 6

[Martin Thomson [:mt:]]

I'm going to re-enable both TLS 1.2 and the static RSA cipher suites in Aurora only.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f388153986c8
https://reviewboard.mozilla.org/r/4053

Comment 7

[Martin Thomson [:mt:]]

Attachment: MozReview Request: bz://1134437/mt

/r/4055 - Bug 1134437 - Enable TLS 1.2 and static RSA ciphersuites, r=abr

Pull down this commit:

hg pull review -r 725160562d1deaf20750dca18ab27cccda32390f

Comment 8

[Adam Roach [:abr]]

https://reviewboard.mozilla.org/r/4055/#review3255

To the extent that I understand the problem, this appears to address it. I'm not an expert on the suites that we're re-enabling, but trust that Martin got that part right.

Comment 9

[Martin Thomson [:mt:]]

Comment on attachment 8566878 [details]
MozReview Request: bz://1134437/mt

Approval Request Comment
[Feature/regressing bug #]: bug 1052610
[User impact if declined]: some WebRTC applications will not interoperate with Firefox, we've been asked to give them a little more time to update
[Describe test coverage new/current, TreeHerder]: TBPL and unit tests are OK with this; the area has good automated coverage
[Risks and why]: only that this is no longer considered "good" crypto
[String/UUID change made/needed]: none

Note: aurora/37 only please, we don't want this in 38

Comment 10

[Lawrence Mandel [:lmandel] (use needinfo)]

Comment on attachment 8566878 [details]
MozReview Request: bz://1134437/mt

OK. Let's defer the security change to 38 and re-enable TLS 1.2 and static RSA ciphersuites in 37. Aurora+

Comment 11

[Martin Thomson [:mt:]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/366436d0a508

Comment 12

[Maire Reavy [:mreavy]]

Hi Michael -- The patch for this has now landed on Aurora (Fx37) and should be in the Aurora Nightly build tomorrow (Feb 21).  Can you help us test this and verify that this patch works as expected for you?  Thanks.

Comment 13

[msander]

Yes, I'll let you know.

Comment 14

[msander]

This is working as expected, thank you!

Comment 16

[Martin Thomson [:mt:]]

Attachment: MozReview Request: Bug 1134437 - Enable TLS 1.2 and static RSA ciphersuites, r=abr