LAUNCHXL-CC2640R2：CC2640R2 reconnect fail with MAC, Android Host

Sean Cheng84

Other Parts Discussed in Thread: BLE-STACK, CC2640

Hi I have problem as below

• Scenario: HID device with CC2640R2. CC2640R2 is always failed to reconnect with Mac and Android Host. But there is no issue with Windows 10 Host. For OK case, after reconnect, CC2640R2 will send HID Report to Host.

• Test Environment:
SDK: simplelink_cc2640r2_sdk_2_40_00_32
HW: CC2640R2 launchpad

• Test steps:
1. CC2640R2 connect, just works pairing & bonding with Host.
2. On Host, turn off Bluetooth.
3. On Host, turn on Bluetooth.
4. CC2640R2 try to reconnect, but always fail.

Host	Bluetooth	OS	HID Function test (1st connect)	Reconnect test
Android phone	V4.2	android 7.0/ android 8.0	PASS	FAIL
Windows notebook	V5.0	windows10 Professional ver1809	PASS	PASS
Windows notebook	V5.0	windows10 Professional ver1803	PASS	PASS
Macbook	V4.0	MacOS High Sierra	PASS	FAIL

• Device log: CC2640R2 got GAPBOND_PAIRING_STATE BONED SUCCESS. But for GATT_Notification, the return value is blepending (0x16)

• Sniff log: Attach two logs
mac os sniffer : NG case.
Windows sniffer : OK case.
Below left snapshot is with MAC (NG case), right is with Windows10 (OK case). I find for the left, the encryption process is always fail.

mac os sniffer.rar Windows sniffer.rar

over 4 years ago

0 J Lindh over 4 years ago

TI__Guru 57865 points

Hi Sean,
Thank you for the very detailed post! I see that the peripheral device sends LL_LENGTH_REQ in the NG case, with an unknown response. Does that occur on the Android as well? Also, I am quite surprised that the Mac doesn't send a LL_FEATURE_REQ, as the Win10 does. Again, how does the Android behave in this case?

Also, can you confirm that pairing works in all three cases (Mac/Android/Win10) although it's the reconnect (bonding) that is troublesome for Mac/android? If that is the case, any difference in the pairing process between the centrals?

0 Sean Cheng84 over 4 years ago in reply to J Lindh

Prodigy 130 points

attach is Andorid sniffer
open sniffer file, you can use link for Frontline Protocol Analysis System

http://fte.com/support/CPAS-download.aspx?demo=BPA%20600&iid=21

i have no idea how to fix this bug could you help analyzing and give some suggestion

2625.Android.rar

0 J Lindh over 4 years ago in reply to Sean Cheng84

TI__Guru 57865 points

Hi Sean,
I am actually running on a macOS which Frontline haven't added support for yet. Any chance you could share a print screen similar to the other ones?

0 Sean Cheng84 over 4 years ago in reply to J Lindh

Prodigy 130 points

Hi J Linh:

Can you use virtual machine for Windows ?

i try to export file have lost Details data for .cfa file

as below photo

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

Can you confirm that pairing works in all three cases (Mac/Android/Win10) although it's the reconnect (bonding) that is troublesome for Mac/android? If that is the case, any difference in the pairing process between the centrals?

Are you using the BLE-Stack or BLE5-Stack?

Just to be sure: Is the CC2640R2 device the master or the slave in the connection? (It looks like from the sniffer log that the master device sends an LL_START_ENC_REQ, this packet should only be sent by the slave so it's strange behavior.)

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Hi Marie H:
I using the BLE-Stack.
CC2640R2 is slave in the connection.

I found some difference as below

1. Mac os have response for Read by Group Type,but Windows os not.
2. Mac os have Read Request for Report Map,but windows os not.
3. Windows os have Read Request for Boot Mouse Input Report , but Mac os not
4. Windows os have Read Request for Boot Keyboard Output Report, but Mac os not

MAC OS:

Windows OS:

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

Can you confirm that pairing works in all three cases (Mac/Android/Win10) although it's the reconnect (bonding) that is troublesome for Mac/android?

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Yes,pairing works in all three cases,
it's get success at GAPBOND_PAIRING_STATE_COMPLETE in all three cases

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

Are you getting any errors on the master side when this happens?

I would like to try to reproduce the issue here. Can you tell me exactly what android devices you were using when this happened?

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

No, i didn't get any errors on the master side.
android device is ZenFone 4 (ZE554KL)/ ZenFone 3 Deluxe (ZS550KL)

By the way,i send example code in private message on Saturday.
Do you reproduce the issue?

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

I think the easiest will be for me to reproduce using a macbook. Can you give me step-by step instructions? I.e. what kind of app you use on the mac side etc.

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

I don't have use app on the mac side.
Because CC2640 is HID Report for mouse.
HID Report control X&Y point for mouse at ble connected.

step as below

1.turn on bluetooth,connect CC2640_MA

2.connected success,and then mouse cursor can auto move by HID Report.

3.turn off bluetooth,and then turn on bluetooth wait reconnected.

4.mac os show connected,but mouse cursor does't move.(CC2640R2 device log got GAPBOND_PAIRING_STATE BONED SUCCESS. But for GATT_Notification, the return value is blepending (0x16) on this time.)

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

Thanks for the instructions!

I looked at the source code you sent me, looking at the projectspec file this project uses simplelink_cc2640r2_sdk_1_50_00_58 (not simplelink_cc2640r2_sdk_2_40_00_32). Did you make changes in the project options that are not reflected in the files you sent me?

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Hi Marie H:

Sorry ,i use IAR for IDE.

IAR use relative path

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

I followed the steps you posted, but on my side I can't see any issue.

1.turn on bluetooth,connect CC2640_MA
2.connected success,and then mouse cursor can auto move by HID Report.
3.turn off bluetooth,and then turn on bluetooth wait reconnected.
4.mac os show connected,but mouse cursor does't move.(CC2640R2 device log got GAPBOND_PAIRING_STATE BONED SUCCESS. But for GATT_Notification, the return value is blepending (0x16) on this time.)

number 4 for me is: Mac OS show reconnected and the cursor starts moving around in circles. I let it run for around 5 minutes.

0 Marie H over 4 years ago in reply to Marie H

TI__Guru 73541 points

I tested with MacBook C02WC2PQHW2R (High Sierra).

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Hi Marie:

What IDE are you using?

CCS/IAR?

My Version is 8.20.2 for IAR

This hex file can help me test if there is no issue?

2744.test.7z

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

I am using IAR 8.20.2.

Did you make any code changes in this hex file?

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Hi Marie H:
No,so this hex file test can't see any issue?
Can you give hex flile& sniffer for me confirm if it is any issue?

Thanks

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

i Sean,

I have attached the hex file.

Looking at the sniffer log you posted, both the master and the slave are sending unknown LL control packets. I find this very mysterious, which is why I would like to reproduce the issue here.

ble_hid_adv_remote_cc2640r2rc_app_FlashROM_StackLibrary.hex

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

Are you testing on a LaunchPad?

0 Jerry Kuo over 4 years ago in reply to Marie H

TI__Expert 3525 points

Hi Marie,

yes, with CC2640R2 launchpad.

Customer (Sean) tested "hid_emu_kbd" project (from SIMPLELINK-CC2640R2-SDK-BLE-EXAMPLE-PACK_1.50.00.62) with Macbook. It works fine (no such reconnect problem). So, he ports "hid_emu_kbd" to simplelink_cc2640r2_sdk_2_40_00_32. But the reconnect problem happen. Is the reconnect problem caused by the newer SDK ?

Please provide any advise about why for GATT_Notification, the return value is blepending (0x16) ?

Thank you.

Attachment is the hex file of hid_emu_kbd ported to SDK_2_40.

ble_hid_emu_kbd_cc2640r2lp_app_FlashROM_StackLibrary.zip

0 Marie H over 4 years ago in reply to Jerry Kuo

TI__Guru 73541 points

Hi Jerry,

I don't know any reason why this should happen.

I am getting confused since the sniffer logs you have posted do not match what you're describing:
- you say the CC2640R2 is getting blePending error messages and the macbook is not receiving any more HID information
- the sniffer logs show the CC2640R2 is continuously sending notifications, there are no retransmissions.

Can you go over the math of
a) How much data the sensor is generating
b) How often the CC2640R2 is trying to send a GATT notification
c) How long the connection interval is
to see if it's possible to send as many packets as is being queued in the availeble time.

You should also look into whether the MAX_NUM_PDU and MAX_PDU_SIZE defines are creating a bottle-neck.

0 Sean Cheng84 over 4 years ago in reply to Marie H

Prodigy 130 points

Hi Marie:
a) How much data the sensor is generating
It's mouse data for simulation.

b) How often the CC2640R2 is trying to send a GATT notification
It's 10ms to send Hid report .

c) How long the connection interval is
connection interval is 10ms.

MAX_NUM_PUD use Default.

0 Marie H over 4 years ago in reply to Sean Cheng84

TI__Guru 73541 points

Hi Sean,

What I mean is that how many bytes are generated compared to how many bytes can be sent? Since you are generating one packet per 10 ms and sending one packet per 10 ms you should look at the number of bytes being generated vs. MAX_PDU_SIZE.

Bluetooth®︎

Bluetooth forum

LAUNCHXL-CC2640R2：CC2640R2 reconnect fail with MAC, Android Host