1. 26 Oct, 2016 1 commit
  2. 25 Oct, 2016 1 commit
    • Jakub Pawlowski's avatar
      Fix random crashes in HID related code · e318faa7
      Jakub Pawlowski authored
      Operation on characteristics/descriptors shouldn't access GATT database
      when it's executed. This could happen while service rediscovery is in
      progress.
      
      Bug: 32240759
      Test: connect to HID device
      Change-Id: Ie2b6e6b451456204b1cea1e500df9a0ff949a9ef
      (cherry picked from commit d8f09d07)
      e318faa7
  3. 23 Oct, 2016 1 commit
  4. 20 Oct, 2016 1 commit
  5. 05 Oct, 2016 1 commit
  6. 03 Oct, 2016 1 commit
  7. 29 Sep, 2016 1 commit
  8. 28 Sep, 2016 2 commits
    • gitbuildkicker's avatar
    • Andre Eisenbach's avatar
      Fix 100 LE device connection limitation · 013c32bf
      Andre Eisenbach authored
      Currently after 100 devices are added to the device security database,
      any subsequent LE connection to a newly discovered LE device address
      will fail as entries in the security record database are not reused.
      
      This patch removes a device record if the device itself is removed and
      also ensures that the oldest device security record is deleted if the
      limit is reached to ensure a new record can be allocated.
      
      Bug: 31625900
      Change-Id: I22f6c82c64a9a9bfb2a16d79182903e5aa011355
      013c32bf
  9. 26 Sep, 2016 1 commit
  10. 25 Sep, 2016 1 commit
  11. 23 Sep, 2016 1 commit
    • chaoyu.x.wu's avatar
      Fix the BLE connection failure with some mice · 58c27b48
      chaoyu.x.wu authored
      If enable the "connection parameters update" during GATT discovery,
      it may cause some device connection failure due to connection timeout.
      Since the the "connection parameter update" is enabled again when GATT
      discovery completed, remove the enable of "connection parameter update"
      after SMP complete to fix the issue.
      
      This issue was already fixed in commit 2d41fe1c, but then it was
      re-introduced in commit 444a8da8, which also removed comment explaining
      the issue, added back in this patch.
      
      Bug: 29060797
      Change-Id: I42b0b36056821c30d887484e22bfcbd04ea7ca03
      (cherry picked from commit 1ec17e06)
      58c27b48
  12. 18 Sep, 2016 1 commit
  13. 16 Sep, 2016 7 commits
    • Pavlin Radoslavov's avatar
      Add BT_HCI_UNKNOWN_MESSAGE_TYPE log event · 9ef3ce8f
      Pavlin Radoslavov authored
      If the received HCI type is unknown, then log an event and abort.
      The most likely reason for that to happen is if the UART stream
      is corrupted. We cannot recover from that, and there is not much
      else we can do.
      
      Also, fixed a bug in an HCI-related unit test that was exposed
      by the above change.
      
      Bug: 31432127
      Change-Id: Ia888c485f177af4962268bf8f593b27fd7a4b080
      (cherry picked from commit 27ec0c78)
      (cherry picked from commit e1c8ed17)
      9ef3ce8f
    • Pavlin Radoslavov's avatar
      Add BT_HCI_UNKNOWN_MESSAGE_TYPE log event · e1c8ed17
      Pavlin Radoslavov authored
      If the received HCI type is unknown, then log an event and abort.
      The most likely reason for that to happen is if the UART stream
      is corrupted. We cannot recover from that, and there is not much
      else we can do.
      
      Also, fixed a bug in an HCI-related unit test that was exposed
      by the above change.
      
      Bug: 31432127
      Change-Id: Ia888c485f177af4962268bf8f593b27fd7a4b080
      (cherry picked from commit 27ec0c78)
      e1c8ed17
    • Jakub Pawlowski's avatar
      Fix bad address type assumption · 1e688824
      Jakub Pawlowski authored
      LE related code should not use BTM_ACL_IS_CONNECTED, because it assumes
      classic transport.
      
      In this particular case, connection created event, which was result of
      advertising, caused cancellation of "LE Create Connection".
      Attempt to send "LE Create Connection Cancel" was checking if connection
      is already established, but was failing to do so becuase
      BTM_ACL_IS_CONNECTED was used instead of btm_bda_to_acl with proper
      transport. This caused just created connection to be dropped.
      
      Bug: 31442085
      Change-Id: Ia345a3d00e8567160f4abb2a31c4130ee69d67a3
      (cherry picked from commit f5566938)
      1e688824
    • Andre Eisenbach's avatar
    • Jakub Pawlowski's avatar
      Add a missing case for BTA_DM_SEARCH_CANCEL_CMPL_EVT · 5f98c241
      Jakub Pawlowski authored
      Cancelling discovery was not handled properly in
      btif_dm_search_services_evt and was causing crashes due to assert in
      default case.
      
      Bug: 31442085
      Change-Id: If9a2c203730c9b06df5ff8d8e251f0c35addcb82
      (cherry picked from commit 7df43240)
      5f98c241
    • Jakub Pawlowski's avatar
      Fix NPE in l2cble_notify_le_connection · 5e61112a
      Jakub Pawlowski authored
      Accessing p_lcb must be guarded with NULL check.
      
      Bug: 31442085
      Change-Id: I0f6e3f8e0247c5e8946c4d098f8a3ba442f5fb28
      (cherry picked from commit b4458c93)
      5e61112a
    • Andre Eisenbach's avatar
      Add security record for devices found through LE batch scan · b4a5ec95
      Andre Eisenbach authored
      Currently, starting an LE batch scan are not added to the stack internal
      device database, which means that the address type and device type
      needed for a connection are not available.
      
      This patch ensures that devices found during an LE scan are added to the
      device database to ensure connection attempts will succeed.
      
      Bug: 31309376
      Change-Id: Ic52d7eaa03d17dcbbd848af9140038bc84ea3bd0
      (cherry picked from commit e0295e2c)
      b4a5ec95
  14. 07 Sep, 2016 1 commit
  15. 06 Sep, 2016 1 commit
  16. 04 Sep, 2016 1 commit
  17. 01 Sep, 2016 1 commit
    • Jakub Pawlowski's avatar
      Fix HID crashing on reconnect · 0afaaa41
      Jakub Pawlowski authored
      This patch fixes crashes on reconnection to HID device.
      All those issues are already fixed in AOSP.
      
      Bug: 31221398
      Change-Id: I43b64d0b11b1f548cb373332777ea92d8a075466
      0afaaa41
  18. 29 Aug, 2016 8 commits
  19. 23 Aug, 2016 4 commits
    • Jakub Pawlowski's avatar
      Make gatt_attr use opportunistic connection · de0eabbc
      Jakub Pawlowski authored
      am: 970d9d18
      
      Change-Id: Iaff1375e619409afe9570b6fb79867070a339902
      de0eabbc
    • Jakub Pawlowski's avatar
      GATT: don't update link status for apps that were never connected · a7dd7079
      Jakub Pawlowski authored
      am: aab757d3
      
      Change-Id: Id9397aa59aae3570821c9838d428541cfb6a878a
      a7dd7079
    • Jakub Pawlowski's avatar
      Make gatt_attr use opportunistic connection · 970d9d18
      Jakub Pawlowski authored
      This patch creates new connection mode - opportunistic connection. When
      such connection is made, no call to gatt_update_app_hold_link_status
      will be made when it's started.
      
      This means that connecting and disconnecting in this mode won't trigger
      disconnect timer. When other, app creates regular connection and then
      disconnects, the physical connection might get disconnected.
      
      Opportunistic connection is used only for code setting CCC right now.
      
      Bug: 30186455
      Change-Id: Ia5be7682b6c3dcb4993060f818dee603aef3e19e
      970d9d18
    • Jakub Pawlowski's avatar
      GATT: don't update link status for apps that were never connected · aab757d3
      Jakub Pawlowski authored
      Right now, in GATT_Deregister we make a call to
      gatt_update_app_use_link_flag, which might result in a call to
      GATT_SetIddleTimeout on links that this client was never connected to.
      
      In order to fix that, we must check if gatt_update_app_hold_link_status
      made any changes to p_tcb->app_hold_link. If not, then skip updating
      the timer.
      
      Bug: 30186455
      Change-Id: I6f18ec7fe4e80e048adc9e360cae703b7d64d62a
      aab757d3
  20. 21 Aug, 2016 1 commit
  21. 19 Aug, 2016 1 commit
  22. 18 Aug, 2016 2 commits
    • Jakub Pawlowski's avatar
      GATT: don't update link status for apps that were never connected · d921c78f
      Jakub Pawlowski authored
      Right now, in GATT_Deregister we make a call to
      gatt_update_app_use_link_flag, which might result in a call to
      GATT_SetIddleTimeout on links that this client was never connected to.
      
      In order to fix that, we must check if gatt_update_app_hold_link_status
      made any changes to p_tcb->app_hold_link. If not, then skip updating
      the timer.
      
      Change-Id: I6f18ec7fe4e80e048adc9e360cae703b7d64d62a
      d921c78f
    • Jakub Pawlowski's avatar
      Make gatt_attr use opportunistic connection · 8ed030ff
      Jakub Pawlowski authored
      This patch creates new connection mode - opportunistic connection. When
      such connection is made, no call to gatt_update_app_hold_link_status
      will be made when it's started.
      
      This means that connecting and disconnecting in this mode won't trigger
      disconnect timer. When other, app creates regular connection and then
      disconnects, the physical connection might get disconnected.
      
      Opportunistic connection is used only for code setting CCC right now.
      
      Bug: 30186455
      Change-Id: Ia5be7682b6c3dcb4993060f818dee603aef3e19e
      8ed030ff