Merge lp:~mterry/unity8/cancel-pam-harder into lp:unity8

[Michael Terry]

Fix a possible crash in our PAM threading code. (LP: #1425362)

So it was possible for a PAM subthread (which does the actual authenticating) to get mismatched state with the main thread. When this happened (i.e. when the subthread tried to use a pam_handle object that the main thread had already closed), a crash occurs. Like LP: #1425362.

So I've done a few cleanups here:
- Use a cancellable QConcurrent method (::mapped instead of ::run). Being able to cancel the thread is a stronger hammer than just hoping we answer all questions from the thread and it will close itself.
- We now block when the subthread sends a prompt request to the main thread. This makes handling those signals more predictable when we're trying to close the thread.
- We will process events while waiting for the subthread to close, so that we can actually process the above signals in the main thread.

I've added a dumb little test that tries to authenticate/cancel/authenticate 100 times in quick succession. Without this branch, we'd crash reliably anywhere from attempt 2 to 7. With this branch, we're fine all 100.

Fix a possible crash in our PAM threading code. (LP: #1425362)

So it was possible for a PAM subthread (which does the actual authenticating) to get mismatched state with the main thread. When this happened (i.e. when the subthread tried to use a pam_handle object that the main thread had already closed), a crash occurs. Like LP: #1425362.

So I've done a few cleanups here:
- Use a cancellable QConcurrent method (::mapped instead of ::run). Being able to cancel the thread is a stronger hammer than just hoping we answer all questions from the thread and it will close itself.
- We now block when the subthread sends a prompt request to the main thread. This makes handling those signals more predictable when we're trying to close the thread.
- We will process events while waiting for the subthread to close, so that we can actually process the above signals in the main thread.

I've added a dumb little test that tries to authenticate/cancel/authenticate 100 times in quick succession. Without this branch, we'd crash reliably anywhere from attempt 2 to 7. With this branch, we're fine all 100.

Users didn't notice this bug, because the timing is tough to reproduce naturally (I would have thought, but then bug 1425362 noticed in a VM... Maybe the slowness of a VM helps surface this).

== Checklist ==

 * Are there any related MPs required for this MP to build/function as expected? Please list.
 No

 * Did you perform an exploratory manual test run of your code change and any related functionality?
 Yes... I added a test anyway, that stresses this code

 * Did you make sure that your branch does not contain spurious tags?
 Yes

 * If you changed the packaging (debian), did you subscribe the ubuntu-unity team to this MP?
 NA

 * If you changed the UI, has there been a design review?
 NA

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1636
http://jenkins.qa.ubuntu.com/job/unity8-ci/5366/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1565/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/530/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/531/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/531/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1563/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5366/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1636
http://jenkins.qa.ubuntu.com/job/unity8-ci/5367/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1566/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/531/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/532/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/532/console
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1564/console

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5367/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1637
http://jenkins.qa.ubuntu.com/job/unity8-ci/5371/
Executed test runs:
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1582
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/535/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/536
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/536
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1403
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1580
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1580/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18433

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5371/rebuild

review: Needs Fixing (continuous-integration)

[Michael Terry (mterry)]

Marco Trevisan (the original bug filer) notes that this branch fixes his crash.

[Albert Astals Cid (aacid)]

Tests never seem to finish now

 6/22 Test #6: testGreeterPam ...................***Timeout 1500.00 sec

review: Needs Fixing

[Albert Astals Cid (aacid)]

The test passes for me locally but fails in CI, i wonder if either it needs some extra dependency we don't have on CI or if it needs X/xvfb

[Michał Sawicz (saviq)]

Interestingly, this is what fails for me under sbuild:

22/23 Test #23: wizardsystemtest .................***Failed 15.22 sec
********* Start testing of SystemTest *********
Config: Using QtTest library 5.4.0, Qt 5.4.0 (x86_64-little_endian-lp64 shared (dynamic) release build; by GCC 4.9.2)
PASS : SystemTest::initTestCase()
PASS : SystemTest::testEnable()
QInotifyFileSystemWatcherEngine::addPaths: inotify_add_watch failed: No such file or directory
PASS : SystemTest::testDisable()
QInotifyFileSystemWatcherEngine::addPaths: inotify_add_watch failed: No such file or directory
FAIL! : SystemTest::testNoticeChanges() Compared values are not the same
   Actual (((spy.count()))): 3
   Expected (4) : 4
   Loc: [/«BUILDDIR»/unity8-8.02+15.04.20150302/tests/plugins/Wizard/tst_system.cpp(120)]
PASS : SystemTest::cleanupTestCase()
Totals: 4 passed, 1 failed, 0 skipped, 0 blacklisted
********* Finished testing of SystemTest *********

review: Needs Fixing

[Michael Terry (mterry)]

Huh. During build it passes in jenkins (since make test is run during build). It's just during the qmluitest run that it fails.

[Michael Terry (mterry)]

<Saviq> mterry, huh, the failure I reported actually is there in trunk for me...

So that leaves just the jenkins failure. I'm trying to see how the qmluitests environment is different from the build one.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1638
http://jenkins.qa.ubuntu.com/job/unity8-ci/5393/
Executed test runs:
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1668
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/557/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/558
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/558/console
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1480
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1666
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1666/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18599

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5393/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1639
http://jenkins.qa.ubuntu.com/job/unity8-ci/5409/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1708/console
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/573
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/574
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/574
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1514/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1706
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1706/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18721

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5409/rebuild

review: Needs Fixing (continuous-integration)

[Michael Terry (mterry)]

OK, addressed the build failure in jenkins. The remaining qmluitest failures are issues in trunk.

[Francis Ginther (fginther)]

The above mako testing failed due to an unbootable image making it into the pipeline. The ci job has been restarted.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1639
http://jenkins.qa.ubuntu.com/job/unity8-ci/5414/
Executed test runs:
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1717
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/578
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/579
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/579
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1523
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1715
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1715/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18735

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5414/rebuild

review: Needs Fixing (continuous-integration)

[Albert Astals Cid (aacid)]

Small fix:
 * You can change the two futureWatcher.future().something() to futureWatcher.something()

About QCoreApplication::processEvents it's a bit on the nasty side, is it possible to put the code in a slot attached to the futurewatcher finished signal?

Also does respond(QString()) need to happen repeteadly inside the loop or just once?

review: Needs Information

[Michael Terry (mterry)]

> You can change the two futureWatcher.future().something()
> to futureWatcher.something()

Done.

> About QCoreApplication::processEvents it's a bit on the nasty
> side, is it possible to put the code in a slot attached to
> the futurewatcher finished signal?

This would mean holding on to a map of QFuture objects to pam_handle objects so we'd know which ones need to have pam_end called on them when the thread is done. And keeping a set of waiting-QFuture-responses for each thread.

I just figured threading code and PAM code are each complicated enough that I didn't want to try to be fancy. So I favored the direct route.

But I agree that a manual processEvents isn't super sexy. I can change it to the above if you'd like.

> Also does respond(QString()) need to happen
> repeteadly inside the loop or just once?

PAM can offer multiple prompts in a row. The child-thread will block on each response one at a time in a small loop. QFuture.cancel() will not interrupt a waiting child-thread.

But if you can promise that it *will* interrupt the thread the second it stops waiting, before it can get to waiting on a second prompt, then we only need respond() once. And we could even get rid of the loop and just call processEvents once.

I was not so confident in how cancel() worked. But I haven't dealt with Qt threads often.

[Albert Astals Cid (aacid)]

As far as i know, no, cancel won't terminate the thread, actually i think would not cancel authenticateWithPam in the middle of an execution, it's just that the future has not started executing and is cancelled before doing so.

As discussed on irc, please "add code to verify that all respond() have been properly dispatched"

review: Needs Fixing

[Albert Astals Cid (aacid)]

add code to the test :D

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1640
http://jenkins.qa.ubuntu.com/job/unity8-ci/5415/
Executed test runs:
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1723
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/579
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/580
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/580
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1529
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1721
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1721/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18754

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5415/rebuild

review: Needs Fixing (continuous-integration)

[Albert Astals Cid (aacid)]

 * Did you perform an exploratory manual test run of the code change and any related functionality?
Yes

 * Did CI run pass?
No, but the added one passes

 * Did you make sure that the branch does not contain spurious tags?
Yes

review: Approve

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1641
http://jenkins.qa.ubuntu.com/job/unity8-ci/5418/
Executed test runs:
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/1729
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/582
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/583
    FAILURE: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/583/console
    UNSTABLE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/1535
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1727
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/1727/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/18764

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5418/rebuild

review: Needs Fixing (continuous-integration)

[Michał Sawicz (saviq)]

This prevents me from unlocking unity8 on my desktop. The password entry field goes wiggle-wiggle-wiggle every 5s or so without any interaction.

The unity8.log prints a seemingly related:
> This backend doesn't support multiple users

review: Needs Fixing

[Michael Terry (mterry)]

OK, try this again. I was able to reproduce your problem on the desktop, Saviq. Fixed it.

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1644
http://jenkins.qa.ubuntu.com/job/unity8-ci/5604/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/2408/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/767
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/769
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/769
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/2098/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/2406
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/2406/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/19862

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5604/rebuild

review: Needs Fixing (continuous-integration)

[PS Jenkins bot (ps-jenkins)]

FAILED: Continuous integration, rev:1644
http://jenkins.qa.ubuntu.com/job/unity8-ci/5605/
Executed test runs:
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-vivid-touch/2416/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity-phablet-qmluitests-vivid/768
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-amd64-ci/770
    SUCCESS: http://jenkins.qa.ubuntu.com/job/unity8-vivid-i386-ci/770
    FAILURE: http://jenkins.qa.ubuntu.com/job/generic-deb-autopilot-runner-vivid-mako/2105/console
    SUCCESS: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/2414
        deb: http://jenkins.qa.ubuntu.com/job/generic-mediumtests-builder-vivid-armhf/2414/artifact/work/output/*zip*/output.zip
    SUCCESS: http://s-jenkins.ubuntu-ci:8080/job/touch-flash-device/19875

Click here to trigger a rebuild:
http://s-jenkins.ubuntu-ci:8080/job/unity8-ci/5605/rebuild

review: Needs Fixing (continuous-integration)

[Albert Astals Cid (aacid)]

review: Abstain

[Albert Astals Cid (aacid)]

I retested the saviq was having and i can't reproduce anymore.

review: Approve

=== modified file 'plugins/LightDM/liblightdm/GreeterPrivate.cpp'
--- plugins/LightDM/liblightdm/GreeterPrivate.cpp	2015-01-20 11:50:19 +0000
+++ plugins/LightDM/liblightdm/GreeterPrivate.cpp	2015-04-23 00:56:41 +0000
@@ -56,7 +56,8 @@
                 this, SLOT(handleMessage(pam_handle *, QString, QLightDM::Greeter::MessageType)));
         // This next connect is how we pass ResponseFutures between threads
         connect(this, SIGNAL(showPrompt(pam_handle *, QString, QLightDM::Greeter::PromptType, QLightDM::GreeterImpl::ResponseFuture)),
-                this, SLOT(handlePrompt(pam_handle *, QString, QLightDM::Greeter::PromptType, QLightDM::GreeterImpl::ResponseFuture)));
+                this, SLOT(handlePrompt(pam_handle *, QString, QLightDM::Greeter::PromptType, QLightDM::GreeterImpl::ResponseFuture)),
+                Qt::BlockingQueuedConnection);
     }
 
     ~GreeterImpl()
@@ -68,6 +69,13 @@
     {
         // Clear out any existing PAM interactions first
         cancelPam();
+        if (pamHandle != nullptr) {
+            // While we were cancelling pam above, we processed Qt events.
+            // Which may have allowed someone to call start() on us again.
+            // In which case, we'll bail on our current start() call.
+            // This isn't racy because it's all in the same thread.
+            return;
+        }
 
         AppData *appData = new AppData();
         appData->impl = this;
@@ -79,7 +87,7 @@
 
         if (pam_start("lightdm", username.toUtf8(), &conversation, &pamHandle) == PAM_SUCCESS) {
             appData->handle = pamHandle;
-            futureWatcher.setFuture(QtConcurrent::run(authenticateWithPam, pamHandle));
+            futureWatcher.setFuture(QtConcurrent::mapped(QList<pam_handle*>() << pamHandle, authenticateWithPam));
         } else {
             delete appData;
             greeterPrivate->authenticated = false;
@@ -88,7 +96,7 @@
         }
     }
 
-    static int authenticateWithPam(pam_handle* pamHandle)
+    static int authenticateWithPam(pam_handle* const& pamHandle)
     {
         int pamStatus = pam_authenticate(pamHandle, 0);
         if (pamStatus == PAM_SUCCESS) {
@@ -242,11 +250,22 @@
 private:
     void cancelPam()
     {
-        // Unfortunately we can't simply cancel our QFuture because QtConcurrent::run doesn't support cancel
         if (pamHandle != nullptr) {
+            QFuture<int> pamFuture = futureWatcher.future();
             pam_handle *handle = pamHandle;
             pamHandle = nullptr; // to disable normal finishPam() handling
-            while (respond(QString())); // clear our local queue of QFutures
+            pamFuture.cancel();
+
+            // Note the empty loop, we just want to clear the futures queue.
+            // Any further prompts from the pam thread will be immediately
+            // responded to/dismissed in handlePrompt().
+            while (respond(QString()));
+
+            // Now let signal/slot handling happen so the thread can finish
+            while (!pamFuture.isFinished()) {
+                QCoreApplication::processEvents();
+            }
+
             pam_end(handle, PAM_CONV_ERR);
         }
     }
@@ -266,6 +285,11 @@
 {
 }
 
+GreeterPrivate::~GreeterPrivate()
+{
+    delete m_impl;
+}
+
 void GreeterPrivate::handleAuthenticate()
 {
     m_impl->start(authenticationUser);

=== modified file 'plugins/LightDM/liblightdm/GreeterPrivate.h'
--- plugins/LightDM/liblightdm/GreeterPrivate.h	2015-01-20 11:50:19 +0000
+++ plugins/LightDM/liblightdm/GreeterPrivate.h	2015-04-23 00:56:41 +0000
@@ -30,7 +30,7 @@
 {
 public:
     explicit GreeterPrivate(Greeter* parent=0);
-    virtual ~GreeterPrivate() = default;
+    virtual ~GreeterPrivate();
 
     // These variables may not be used by all subclasses, that's no problem
     bool authenticated;

=== modified file 'tests/plugins/LightDM/CMakeLists.txt'
--- tests/plugins/LightDM/CMakeLists.txt	2015-03-02 14:31:08 +0000
+++ tests/plugins/LightDM/CMakeLists.txt	2015-04-23 00:56:41 +0000
@@ -1,8 +1,38 @@
+add_definitions(
+    -DCURRENT_SOURCE_DIR="${CMAKE_CURRENT_SOURCE_DIR}"
+    )
+include_directories(
+    ${CMAKE_CURRENT_BINARY_DIR}
+    )
+
 add_executable(GreeterDBusTestExec
     dbus.cpp
     ${CMAKE_SOURCE_DIR}/plugins/LightDM/Greeter.cpp
     )
 qt5_use_modules(GreeterDBusTestExec Core DBus Quick Test)
+target_link_libraries(GreeterDBusTestExec
+    -L${CMAKE_BINARY_DIR}/tests/mocks/LightDM/liblightdm
+    -llightdm-qt5-2
+    )
+target_include_directories(GreeterDBusTestExec PUBLIC
+    ${CMAKE_SOURCE_DIR}/plugins/LightDM
+    ${CMAKE_SOURCE_DIR}/tests/mocks/LightDM
+    )
+add_binary_qml_test(GreeterDBus "${CMAKE_BINARY_DIR}/tests/mocks/LightDM/liblightdm" MockLightDM "QML2_IMPORT_PATH=${CMAKE_BINARY_DIR}/tests/mocks")
+
+add_executable(GreeterPamTestExec
+    pam.cpp
+    ${CMAKE_SOURCE_DIR}/plugins/LightDM/liblightdm/GreeterPrivate.cpp
+    )
+qt5_use_modules(GreeterPamTestExec Concurrent Core Test)
+target_link_libraries(GreeterPamTestExec
+    integratedLightDM
+    )
+target_include_directories(GreeterPamTestExec PUBLIC
+    ${CMAKE_SOURCE_DIR}/plugins/LightDM/liblightdm
+    )
+add_qmltest_target(testGreeterPam "${CMAKE_CURRENT_BINARY_DIR}/GreeterPamTestExec" TRUE FALSE)
+add_dependencies(testGreeterPam GreeterPamTestExec)
 
 add_executable(GreeterUsersModelTestExec
     usersmodel.cpp
@@ -10,24 +40,13 @@
     ${CMAKE_SOURCE_DIR}/plugins/Utils/unitysortfilterproxymodelqml.cpp
     )
 qt5_use_modules(GreeterUsersModelTestExec Core Test)
-
-include_directories(
-    ${CMAKE_CURRENT_BINARY_DIR}
+target_link_libraries(GreeterUsersModelTestExec
+    -L${CMAKE_BINARY_DIR}/tests/mocks/LightDM/liblightdm
+    -llightdm-qt5-2
+    )
+target_include_directories(GreeterUsersModelTestExec PUBLIC
     ${CMAKE_SOURCE_DIR}/plugins/LightDM
     ${CMAKE_SOURCE_DIR}/plugins/Utils
     ${CMAKE_SOURCE_DIR}/tests/mocks/LightDM
     )
-
-target_link_libraries(GreeterDBusTestExec
-    MockLightDM
-    )
-
-target_link_libraries(GreeterUsersModelTestExec
-    MockLightDM
-    )
-
-add_definitions(-DCURRENT_SOURCE_DIR="${CMAKE_CURRENT_SOURCE_DIR}")
-
-add_binary_qml_test(GreeterDBus "${CMAKE_BINARY_DIR}/tests/mocks/LightDM/liblightdm" MockLightDM "QML2_IMPORT_PATH=${CMAKE_BINARY_DIR}/tests/mocks")
-
 add_binary_qml_test(GreeterUsersModel "${CMAKE_BINARY_DIR}/tests/mocks/LightDM/liblightdm" MockLightDM "LIBLIGHTDM_MOCK_MODE=full")

=== added file 'tests/plugins/LightDM/pam.cpp'
--- tests/plugins/LightDM/pam.cpp	1970-01-01 00:00:00 +0000
+++ tests/plugins/LightDM/pam.cpp	2015-04-23 00:56:41 +0000
@@ -0,0 +1,52 @@
+/*
+ * Copyright (C) 2015 Canonical, Ltd.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; version 3.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "GreeterPrivate.h"
+
+#include <QtTest>
+
+class GreeterPamTest : public QObject
+{
+    Q_OBJECT
+
+private Q_SLOTS:
+
+    void init()
+    {
+        m_greeterpriv = new QLightDM::GreeterPrivate();
+    }
+
+    void cleanup()
+    {
+        delete m_greeterpriv;
+        QTRY_COMPARE(QThreadPool::globalInstance()->activeThreadCount(), 0);
+    }
+
+    void testRapidFireAuthentication()
+    {
+        m_greeterpriv->authenticationUser = qgetenv("USER");
+        for (int i = 0; i < 100; i++) {
+            m_greeterpriv->handleAuthenticate();
+        }
+    }
+
+private:
+    QLightDM::GreeterPrivate *m_greeterpriv;
+};
+
+QTEST_GUILESS_MAIN(GreeterPamTest)
+
+#include "pam.moc"