GitLab

mNetworkFactory and mUntrustedNetworkFactory are being dumped twice. The
second time, potentially causing null exceptions.
Removed the second calls (were left over from bad rebase).

BUG=28530520

Change-Id: I3312ac83a7ea2d4027036ea85702caa74985753d

We sometimes fetch the channel list often (every 20 seconds),
so logging the message on every attempt to fetch the list is
spammy.

Demote the message to verbose level, to reduce spam.

BUG=28425940
TEST=manual

Manual test:
- install new code
- reboot device
- after reboot, sleep 60
$ adb logcat -d \
  | egrep 'getting valid|ScannerImpl: Starting wifi scan for freqs'
  -> match for "ScannerImpl...", but not "getting valid"

Change-Id: I56a8f941ced0f291f0c37e9279e5dda635351cb2

Added an exponential back off retry mechanism for networks
found by PNO scan but rejected by QNS due to their low RSSI
values.

Bug: 28402107
Change-Id: I8a69afb3d22fbd2f479073e6087893c24e5a5d39

When new scan results are processed, networks already stored as the
available networks may have updated configs passed in, but they may also
have null configs.  The null configs should not be used to update the
stored config.  Added a check to determine if the passed in config is
not null before the update.

In addition, debugging output also reported the value of
HasEverConnected as false for networks with a null config.  This was
updated to report null_config instead.

Added tests covering config updates.
Added tests for debugging output.

BUG: 28451079
Change-Id: Iff9888ab87c61619b2f865516eca22d87eb4f4b8

Keep track of the network BSSID of the last connection attempt.
This BSSID is checked in case the phone is still in the process of
connecting to that network. mWifiInfo is checked for the currently
connected network. This is for the case that firmware performed
roaming automatically.

While there, addressed a checkstyle issue.

Bug: 28462151
Change-Id: Ia9316048e0549036de1c20bf1f3099cfcda9a75e
Test: WiFi unit tests

* changes:
  WifiLastResortWatchdog trigger logic
  WifiLastResortWatchdog failure counting

Added logging of various metrics to the WifiLastResortWatchdog. These
metrics count the number of times the Watchdog triggers, and stats
counting the number of networks present at failure time for different
failure types.

BUG=27856474

Change-Id: If43836b1c33791fefb8000196b231c312161feef

Implemented trigger condition logic that is checked after a failure is
counted. The condition takes into account: failure counts, wifi
connected state, past connections to saved networks and whether the
watchdog has fired since the wifi environment changed.

Please note: This change only implements the logic for deciding when to
trigger, not the trigger itself.

BUG=27856709

Change-Id: I91ad6127a72ca24c72a5520348c4fff83f16b559

Move the execution of this method to the thread of
WifiStateMachine to address a potential WifiConnectivityManager
concurrent access issue.

While there, remove a dead method.

Bug: 28369823
Change-Id: I49f424637ee6897495daedbf4b1cc06477e78798

With WifiScanner also running on the WifiStateMachine thread,
WifiConnectivityManager is now exectued on a single thread context.

While there, fixed a few checkstyle issues.

Bug: 28369823
Change-Id: Ibf2b65be42d243deb61a68380a9011f4acc663b4

Implemented watchdog methods and calls to them from WifiStateMachine to
count association rejection, authentication failure & dhcp failures.

This will allow the watchdog to determine when all available networks
are failing, and Wifi Framework has no other options, however this Cl
does not implement that logic.

BUG=27855690
BUG=27856523

Change-Id: I416bfd0e0340823de5ca04f2d040bff7a751b677

This reverts commit ced2db17.

Reverting due to test failure.
BUG=28385150

Change-Id: I86ffd7f017727ee94508102cd7a2962dcf803ffc

BUG: 28242396
Change-Id: I53c0e91d2b9ee2731c0f74ef6376f4fb4e67f2e9

Sending ANQP request when wifi is connecting or connected
may fail the connection.
This CL fixes it by adding a parameter to WifiConfigManager
::updateSavedNetworkWithNewScanDetail(). This boolean value
can indicate that we are connecting or connected.
Link debouncing is also considered as connecting in this case.

BUG=26494190
TEST=runtest frameworks-wifi
Change-Id: Ic4dfd0c0ba0392da68b6705d18fb82d92f44e508

Also make logging in WifiScanningService not static and add some tests
to verify some logging behavior.

Fixes: 28274869
Change-Id: I879200778dcadb4e67d5a85d42442baf3ceef6c9

Create a separate method to fetch the saved network from scan detail.
This can be used for caching dtim values in WifiStateMachine.

BUG: 26494190
TEST: Compiles & unit-test passes

Change-Id: Ib5819fd72afc3dcf691735deccb237874eac3f8d

WifiLogger::getOutput() sometimes takes more than a few
seconds because it fetchs the entire logcat buffer then takes
the most recent 127 lines.
Since it is a blocking call, it may increase the chance that
WifiService being killed by WatchDog.

This change uses logcat -t option, only fetching the
most recent 127 lines. It significantly reduces the time of the
blocking call.

BUG=28298654
TEST=unit tests, manual test
Manual test:
- boot device
$ adb logcat -d | wc -l
  -> repeat until >127
$ adb bugreport > /tmp/bugreport
- check that ‘system log:’ section contains 127 lines.

Change-Id: Ifcfb9e53a7c8b731e76cab1f25cec1bc0c83f33e

b/27905553  This fixes CTS failures in testProtectedBroadcasts in
android.permission.cts.TelephonyManagerPermissionTest.

Currently the "state" variable can be null, so the equality condition 
has been rearranged to avoid the NPE.

Change-Id: I0326df8680c6e37bc2151a38f8f74305b3f8de79

* changes:
  WifiConnectivityManager: Unit tests for rate limit
  WifiConnectivityManager: Connection rate limit

By default, VERBOSE level logging is disabled for native code.
However, when debugging Wifi issues, we'll sometimes want to
include VERBOSE level log messages from the HAL implementation.

Update WifiStateMachine's verbose-debugging logic, to support
this case.

Note that VERBOSE level messages will not be compiled in
on user builds. Nor will system server be permitted to change
the logging level on user builds. For these reasons, the
log level is only modified on non-user builds.

BUG=27857554
TEST=manual

Manual test (bullhead + forthcoming HAL patch)
- On userdebug build with 'Enable Wi-Fi Verbose Logging' set:
  $ adb reboot && adb wait-for-device && adb logcat -d | grep 'V WifiHAL'
  -> matching lines seen
  $ adb shell getprop | grep WifiHAL
  [log.tag.WifiHAL]: [V]
- On userdebug build with 'Enable Wi-Fi Verbose Logging' unset:
  $ adb reboot && adb wait-for-device && adb logcat -d | grep 'V WifiHAL'
  -> NO matching lines
  $ adb shell getprop | grep WifiHAL
  [log.tag.WifiHAL]: [D]
- On user build with 'Enable Wi-Fi Verbose Logging' set:
  $ adb logcat -d | grep 'V WifiHAL'
  -> NO matching lines
  $ adb shell getprop | grep WifiHAL
  -> NO matching lines

Change-Id: I43af6852cd445f9a7daee04981f253a6c0ebc499

Adding unit tests to verify the connection attempt rate limiting
functionality. Had to change the way system time is being fetched in
WifiConnectivityManager to achieve this.

BUG: 28242396
Change-Id: I2bfae3a9ea60cba1c3c1405dd5eb58208f12c2db
TEST: Compiles & unit-test passes

Implement a rate limit for connection attempts to prevent us flapping
between networks.

While there,
Cleanup all the static time values to be more readable.

BUG: 28242396
Change-Id: I3e2476674e7239e000b93b95751acec3ea7bc8ce

Remove some redundant code in enableNetwork and move some the network
status updation from state machine to ConfigManager.

BUG: 28317135
Change-Id: I375af87bd971352d3066a150e046571f00de44fd
TEST: Compiles & unit-test passes

There is a lot of instability/unknowns for using ePno/Gscans for
connectivity purposes. Unless we've all of them hashed out, turn them
off.

BUG: 28300874
BUG: 28278197
Change-Id: Ifd650032237ac510687f0d374d31f7c9f13d18bd
TEST: Compiles & unit-test passes (Will send for ACTS testing).

Also made the constants final.

BUG:28317300
Change-Id: I32ad6d1e4dbd3f5cc1cdbde8803b66e48b9928c6