1018355 - IPv6 tethering failure

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Wifi, defect)

Description

[pgravel]

Attachment: ipv6tether_ffos.log

USB and wifi tethering cannot be turned on when on an IPv6 connection.

If you look at the attached logs, the failure occurs due to 
 D TetherController: setDnsForwarders(0 = '2620:0:ccc:0:0:0:0:2')
 E TetherController: Failed to parse DNS server '2620:0:ccc:0:0:0:0:2'

This is because netd's TetherController only supports IPv4 DNSes and in gecko's NetworkUtils::setUSBTethering() the data call's IPv6 DNSes are directly passed in to netd. 

As you can see in the second log attaches, Android seems to use the hard-coded DNSes 8.8.8.8 and 8.8.4.4 when toggling on tethering, regardless of the base data call's connection type. (ref: https://android.googlesource.com/platform/frameworks/base/+/master/services/java/com/android/server/connectivity/Tethering.java#1246).

Comment 1

[pgravel]

Attachment: ipv6tether_android.log

Comment 2

[Preeti Raghunath(:Preeti)]

Please review for WiFi

Comment 3

[Ken Chang[:kenkai]]

Chuck, this bug should have been fixed by your patches before, right?

Comment 4

[Ken Chang[:kenkai]]

After discussing with Vincent, he can take this bug.

Comment 5

[Chuck Lee [:chucklee, Triggered by NI, Review, and Feedback requests]]

(In reply to Ken Chang[:ken] from comment #3)
> Chuck, this bug should have been fixed by your patches before, right?

AFAIK, we only deal with wifi but not tethering on IPv6.
The problem here is we set IPv6 info from data call to netd but it doesn't support it.

Hardcode DNS Server to IPv4 seems wired to me.
First, while using IPv4 DNS Server on IPv6 network, I think we also need a ipv4-in-ipv6 tunnel(or some special work in gonk) for the IPv4 DNS query and other IPv4 traffic on IPv6 network. Or we are just making the log looks right but the tethering still fails.
Also, it might break some private services from service provider, if they use some domain name that only can be resolved by the DNS Server from data call.

I suggest we test if force IPv4 DNS Server do make tethering first.

Comment 6

[Vincent Chang[:vchang][changyihsin]]

It should be ipv4 address(8.8.8.8/8.8.4.4) if you don't modify the default dns settings for tethering.wifi.dns1/tethering.wifi.dns2. 
http://dxr.mozilla.org/mozilla-central/source/dom/wifi/WifiWorker.js#61

Or do you modify them in https://github.com/mozilla-b2g/gaia/blob/v1.4/build/config/common-settings.json ?

Comment 7

[pgravel]

I haven't touched Gaia. Here's where gecko grabs the DNS from the data call in both the wifi- and usb-tethering cases.
(wifi) http://lxr.mozilla.org/mozilla-central/source/dom/system/gonk/NetworkUtils.cpp#1552
(usb) http://lxr.mozilla.org/mozilla-central/source/dom/system/gonk/NetworkUtils.cpp#1578

Comment 8

[Preeti Raghunath(:Preeti)]

Vincent

ni to answer comment 7

Comment 9

[Vincent Chang[:vchang][changyihsin]]

Attachment: Patch v1.0

Prevent setting the ipv6 address to dns forwarder.

Comment 10

[Vincent Chang[:vchang][changyihsin]]

I don't have IPV6 environment right here. Can you help to verify if the patch works for you ?

Comment 11

[Chuck Lee [:chucklee, Triggered by NI, Review, and Feedback requests]]

Comment on attachment 8434679 [details] [diff] [review]
Patch v1.0

Review of attachment 8434679 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good, with a little suggestion. Thanks.

::: dom/system/gonk/NetworkUtils.cpp
@@ +1556,5 @@
>    getIFProperties(GET_CHAR(mExternalIfname), interfaceProperties);
>  
>    if (strcmp(interfaceProperties.dns1, "")) {
> +    int type = getIpType(interfaceProperties.dns1);
> +    if (type != AF_INET6) {

I suggest use |type == AF_INET| in case we got expected network types.

Comment 12

[Preeti Raghunath(:Preeti)]

Vincent

Anything pending for landing here?

Comment 13

[Vincent Chang[:vchang][changyihsin]]

Per comment 10, I am waiting for the test result from QC.

Comment 14

[Vincent Chang[:vchang][changyihsin]]

Commit the patch first. 
https://hg.mozilla.org/integration/b2g-inbound/rev/340f8450649e

Comment 15

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/340f8450649e

Comment 16

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-b2g30_v1_4/rev/c29f4a3d409c

Comment 17

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/3a5de06a61d8

Comment 18

[pgravel]

verified behavior is now the same as android. thanks!

Comment 19

[Brogan Zumwalt [Inactive]]

New test case needs to be written to address bug.

Comment 20

[KTucker [:KTucker][Inactive 3/4/2016]]

New test case has been written in moztrap:

https://moztrap.mozilla.org/manage/case/14237/

Comment 21

[Kai-Zhen Li [:seinlin][:kli]]

Vincent, can we also get this fixed on v2.0? Thanks!

Comment 22

[Victor Chang]

I think you're looking for a different "vchang" :-)

Comment 23

[Kai-Zhen Li [:seinlin][:kli]]

NI the correct person.

Comment 24

[Vincent Chang[:vchang][changyihsin]]

I think it's there already. 
http://hg.mozilla.org/releases/mozilla-b2g32_v2_0/file/692c93509dc9/dom/system/gonk/NetworkUtils.cpp#l1560

Comment 25

[Kai-Zhen Li [:seinlin][:kli]]

(In reply to Vincent Chang[:vchang] from comment #24)
> I think it's there already. 
> http://hg.mozilla.org/releases/mozilla-b2g32_v2_0/file/692c93509dc9/dom/
> system/gonk/NetworkUtils.cpp#l1560

Thanks!