WL18xx Bluetooth HCI init failures with TI BTS profiles

Dave Milici

Other Parts Discussed in Thread: WL1831

Hi all--

We have an AM335x BeagleBone-like board with WL1831 module onboard, running Linux kernel 3.8.13 with BlueZ stack enabled.

We modified the UART port assignments for interfacing with the WL18xx HCI UART lines, and tried various pinmux options enabling/disabling RTS/CTS. We also have GPIO assigned for manually toggling BT_EN.

So far the WL18xx chip only gets enumerated via 'hciattach' when using the generic profile without flow control. However this has not been sufficient for testing the BT radio for nearby devices with 'hcitool scan'.

When we use the TI-specific profile with 'hciattach', it switches HCI comm-link to faster baud rate with flow control enabled, but initialization times out. There is a brief burst of activity on UART TxD/RxD lines observed with RTS/CTS lines going low.

When RTS/CTS is enabled for UART3:

root@beaglebone:~# ./bt_en.sh 
root@beaglebone:~# hciattach /dev/ttyO3 texas 115200 flow
Found a Texas Instruments' chip!
Firmware file : /lib/firmware/TIInit_11.8.32.bts
Loaded BTS script version 1
texas: changing baud rate to 3000000, flow control to 1
Initialization timed out.

When RTS/CTS is disabled for UART3:

root@beaglebone:~# ./bt_en.sh
root@beaglebone:~# hciattach /dev/ttyO3 any 115200 noflow
[  245.020416] Bluetooth: HCI UART driver ver 2.2
[  245.025277] Bluetooth: HCI H4 protocol initialized
[  245.030373] Bluetooth: HCI BCSP protocol initialized
[  245.035638] Bluetooth: HCILL protocol initialized
[  245.040625] Bluetooth: HCIATH3K protocol initialized
[  245.045883] Bluetooth: HCI Three-wire UART (H5) protocol initialized
Device setup complete  
root@beaglebone:~# [  245.475620] Bluetooth: BNEP (Ethernet Emulation) ver 1.3
[  245.481356] Bluetooth: BNEP filters: protocol multicast
[  245.493115] Bluetooth: BNEP socket layer initialized
[  245.588394] Bluetooth: RFCOMM TTY layer initialized
[  245.594295] Bluetooth: RFCOMM socket layer initialized
[  245.599742] Bluetooth: RFCOMM ver 1.11

root@beaglebone:~# hciconfig -a
hci0:   Type: BR/EDR  Bus: UART
        BD Address: 98:7B:F3:D8:E9:C2  ACL MTU: 1021:6  SCO MTU: 180:4
        UP RUNNING PSCAN
        RX bytes:924 acl:0 sco:0 events:40 errors:0
        TX bytes:1010 acl:0 sco:0 commands:40 errors:0
        Features: 0xff 0xfe 0x2d 0xfe 0xdb 0xff 0x7b 0x87
        Packet type: DM1 DM3 DM5 DH1 DH3 DH5 HV1 HV2 HV3
        Link policy: RSWITCH HOLD SNIFF
        Link mode: SLAVE ACCEPT
        Name: 'beaglebone-0'
        Class: 0x420100
        Service Classes: Networking, Telephony
        Device Class: Computer, Uncategorized
        HCI Version: 4.0 (0x6)  Revision: 0x0
        LMP Version: 4.0 (0x6)  Subversion: 0xac20
        Manufacturer: Texas Instruments Inc. (13)

However this generic profile is unable to discover any nearby BT devices via 'hcitool scan'.

Note in trying to get the UART configured, we also applied the RTS/CTS serial driver patch posted on TI wiki at:

http://processors.wiki.ti.com/index.php/WL18xx_Platform_Integration_Guide

It seems evident that RTS/CTS handshaking is required for loading TI BTS profiles with 'hciattach'.

Is there a way to get this basic BT functionality working without adding more SW layers (like WLink8 driver patch set or Bluetopia stack)?

--DaveM

over 7 years ago

0 Dave Milici over 7 years ago

Prodigy 70 points

Hi all--

In going through the process for applying TI kernel 3.8 drivers patches for WLink8 support, I found one of the TI wikis referenced another BTS file. 

http://git.ti.com/ti-bt/service-packs/blobs/raw/7783edec65c91e0112961bdb7acad6db6a050092/initscripts/TIInit_11.8.32.bts

This has the same version number v11.8.32 as the one we had been using for 'hciattach', but this one is about 2KB larger and it actually works.

Prior to finding that, I had found older BTS files on another wiki for Wl12xx, some of which had been posted as 115K baud versions. Why not keep these BTS files sorted out in one location?

--DaveM

0 Iain Hunter over 7 years ago in reply to Dave Milici

TI__Genius 16130 points

Hi Dave,
The bts files are kept at git.ti.com/.../initscripts
The one you identified is about a year old, the latest git.ti.com/.../TIInit_11.8.32.bts is service pack 3.9.
The first thing these init scripts do is increase the baudrate to 3M.
Iain

0 Dave Milici over 7 years ago in reply to Iain Hunter

Prodigy 70 points

Hi Iain--

Thanks for the update. I notice the one you're pointing to is about 1KB larger than the last one I found.

So that's at least 3 different files labeled v11.8.32. It would be nice if some docs explained the versioning.

Would the Bluetopia stack be looking to download the same BTS file as BlueZ 'hciattach'?

--DaveM

0 Iain Hunter over 7 years ago in reply to Dave Milici

TI__Genius 16130 points

Hi Dave,
The numbering corresponds to the device - more precisely the ROM code version embedded in the bluetooth core. So a WL183x will request initialisation script 11_8_32.bts. So you should take the latest, which is service pack 3.9 from the git I pointed to. Both Bluetopia and hciattach load this file.

It is mandatory to have RTS/CTS enabled for Bluetooth operation. Without flow control you will not be able to successfully load the initialisation script. Part of the script is a delay while PLL in WL183x locks - flow control is used to hold off the host from continuing to communicate during this locking period.
Iain

0 Dave Milici over 7 years ago in reply to Iain Hunter

Prodigy 70 points

Hi Iain--

OK this latest BTS file also works after downloading via 'hciattach'.

Thanks very much for your followup explanation.

--DaveM

Wi-Fi

Wi-Fi forum

WL18xx Bluetooth HCI init failures with TI BTS profiles