nRF52840: hci_usb timeout during initialization by bluez

[Eystorhaf]

Hey there,

encountered a problem with hci_usb sample on nrF52840,

Hardware:
USB-Devices:
nordic nRF52840 DK & laird DVK-BL654-1.0

USB-Host:
PC / x86 Ubuntu LTS 18.04 / Bluez 4.48 / Kernel 4.15.0-36-generic

Firmware:
zephyr-master @8b20f9d

While initializing the hci-device bluez gets a timeout:
[Thu Oct 18 02:36:18 2018] Bluetooth: hci0: command 0x1002 tx timeout

Here the output of:

btmon: https://pastebin.com/ypNTmTgC
journalct -f | grep bluetoothd* :https://pastebin.com/PT4V6Dpf
dmesg -T: https://pastebin.com/VQQ0tf1x
zephyr uart debug console: https://pastebin.com/v14JUCux

Wondering whether this is a usb or hci stack problem?

Best,

[jhedberg]

I don't know why the controller would not respond to Read Local Supported Commands. It's something that's used even as part of a combined build HCI initialization.

[jhedberg]

The only special thing I can think of regarding HCI_Read_Local_Supported_Commands is that its return parameters (as part of the Command Complete event) are probably some of the longest of any HCI command. Perhaps that could be an issue here (possibly with the way the HCI transport is implemented).

[uhrheber]

I'm getting the same error, both on pca10056 and pca10059.
Mint 19, kernel 4.15.0-36-generic, bluez 5.48.
The same thing happens on an older Mint 18.

[cvinayak]

@jhedberg @uhrheber I just now built the hci_usb sample, and it works out-of-the-box as a btusb device on my archlinux (on my virtualbox) when flashed on my nRF52840 DK. sorry, mine worked as I was on a branch that still used the usb_dc_nrf5.c and not the now usb_dc_nrfx.c implementation.

[pawelzadrozniak]

The device does not enumerate on Windows. Looks like HCI ACL packet is interpreted incorrectly:
https://pastebin.com/0BUYatVF
Error in hci.c (added length logging):
if (buf->len < len) { BT_ERR("Invalid HCI ACL packet length, %d < %d -", buf->len, len); return -EINVAL; }

17218 equals 0x4342 - see the attached USB log:

[cvinayak]

@pawelzadrozniak I am suspecting a regression in 68cb7b3

[pawelzadrozniak]

Looks like Windows 7 does not support this type of device.
The device enumerates in Linux, however I am getting bus fault BT thread:
https://pastebin.com/ffugF07U

Ubuntu 18.04, bluez 5.48.

[finikorg]

Looks like Windows 7 does not support this type of device.
The device enumerates in Linux, however I am getting bus fault BT thread:
https://pastebin.com/ffugF07U

I suggest to create separate issue about Windows support, that never worked for me also (I haven't tested much), probably we need to add some special tweaks...

[henrikbrixandersen]

I can confirm this issue. I tried reverting 68cb7b3 and related commits - that works.

[uhrheber]

henrikbrixandersen, I'm not familiar with git (yet). How do I do that?

[pawelzadrozniak]

Timeout issue fixed in #10934 .

[henrikbrixandersen]

I can confirm that the timeout issue was fixed in #10934. Can this issue be closed?

[cvinayak]

Fixed by #10934

Thank you @henrikbrixandersen