1361099 - provide a way to map all dns to constant (probly localhost)

Status: RESOLVED FIXED

(Core :: Networking: DNS, defect)

Description

[Patrick McManus [:mcmanus]]

// When non empty all non-localhost DNS queries (including IP addresses)
// resolve to this value. The value can be a name or an IP address.
// domains mapped to localhost with localDomains stay localhost.
pref("network.dns.forceResolve", "");

Testing is the primary use case here - replay captive data on a 'fake server' by directing all traffic to it at the DNS level. Chrome has something similar.

Comment 1

[Patrick McManus [:mcmanus]]

does this meet the immediate need?

Comment 2

[Patrick McManus [:mcmanus]]

Attachment: Bug 1361099 - provide a way to map all dns to constant

// When non empty all non-localhost DNS queries (including IP addresses)
// resolve to this value. The value can be a name or an IP address.
// domains mapped to localhost with localDomains stay localhost.
pref("network.dns.forceResolve", "");

Testing is the primary use case here - replay captive data on a 'fake
server' by directing all traffic to it at the DNS level. Chrome has
something similar.

Review commit: https://reviewboard.mozilla.org/r/135180/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/135180/

Comment 3

[Benjamin Smedberg]

Yes, I believe this meets the immediate need.

Comment 4

[Valentin Gosu [:valentin] (he/him)]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138174

::: netwerk/dns/nsDNSService2.cpp:718
(Diff revision 1)
>          return NS_OK;
>      }
>  
> +    if (mForceResolveOn) {
> +        MutexAutoLock lock(mLock);
> +        if (!IsASCII(aInput) ||

It's not clear why the input needs to be non-ascii, based on the patch description.

Comment 5

[Valentin Gosu [:valentin] (he/him)]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138178

Also, the win10 try is all orange. It might be because the test failed, and the pref didn't get cleared?
do_register_cleanup would be the proper way to fix it.

::: netwerk/test/unit/test_dns_service.js:12
(Diff revision 1)
>    onLookupComplete: function(inRequest, inRecord, inStatus) {
>      var answer = inRecord.getNextAddrAsString();
>      do_check_true(answer == "127.0.0.1" || answer == "::1");
>  
> +    nextTest();
>      do_test_finished();

I think the test might become racy. Leave in just one do_test_pending() in run_test(), and move do_test_finished() to testsDone

Comment 6

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/135180/diff/1-2/

Comment 7

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138178

the non win-10 problem is because there is a "wrapper" for that test I forgot about that tries to run it in content (the main one ones in chrome).. which is ok for the feature, but not for the test which sets prefs.

after fixing that in rev2 win10 is still a disaster, but the other platforms look ok. I don't believe the test is racy and causing impacts beyond xpcshell - those test suites are clearly new sessions.

so either the patch is busted or it is based on something busted. hmm.

Comment 8

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/135180/diff/2-3/

Comment 9

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/135180/diff/3-4/

Comment 10

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138174

> It's not clear why the input needs to be non-ascii, based on the patch description.

rev3

Comment 11

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138178

the diff between r3 and r4 is a rebase - no other changes. fyi

Comment 12

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138178

it seems quite likely that win10-taskcluster is just not working.. see https://treeherder.mozilla.org/#/jobs?repo=try&revision=b0fac48a0a88b8b0aa716fbf431fec44a6d7c640 which is unrelated to this bug but looks very similar.

I'll rerun with a more traditional try syntax

Comment 13

[Patrick McManus [:mcmanus]]

Comment on attachment 8863417 [details]
Bug 1361099 - provide a way to map all dns to constant

https://reviewboard.mozilla.org/r/135180/#review138178

> I think the test might become racy. Leave in just one do_test_pending() in run_test(), and move do_test_finished() to testsDone

I don't see any evidence it is racy.. pending/finsished work as semaphores and test2 is declared pending before test1 is marked finished (its part of the work of test1)

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/c76bc8652af1
provide a way to map all dns to constant r=valentin

Comment 15

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/c76bc8652af1