Martin Blumenstingl [off-list ref] writes:

Kevin,

On Sun, Jan 15, 2017 at 11:32 PM, Martin Blumenstingl
[off-list ref] wrote:

quoted

This adds the missing kernel bits for Bluetooth support on the
Tronsmart Vega S95 (GXBB based) boards as well as for the GXL
P230/P231 and GXM Q200/Q201 reference boards.

The Bluetooth functionality on these boards is provided by the
SDIO wifi/Bluetooth combo-chip (Broadcom bcm43xx based). The
Bluetooth module on that combo-chip has to be taken out of reset,
which is taken care of the GPIO in the sdio_pwrseq.

Once the module is taken out of reset it can be set up from userspace
using the "hciattach" tool from bluez, which talks to the Bluetooth
module which is connected to one of the serial ports (in our case
uart_A). To get the Bluetooth module initialized within the timeout
defined by "hciattach" (and to achieve usable speeds for Bluetooth
transfers) the communication uses a speed of 2000000 baud, which was
not supported by meson_uart before.

NOTE: The .dts-changes from this series depends on my previous series
"add support for uart_AO_B" - see [0]


[0] http://lists.infradead.org/pipermail/linux-amlogic/2017-January/001982.html


Martin Blumenstingl (4):
  tty: serial: meson: allow baud-rates higher than 115200
  ARM64: dts: meson-gx: add the serial CTS and RTS pin groups

Neil just informed me that Rob Herring is working on defining the UART
Bluetooth device properly using devicetree (preparation for this is
named "[PATCH v2 0/9] Serial slave device bus", see [0] for the
mailing-list conversation and [1] for a WiP git repo).

Ah, good news.  I knew Rob had been talking about that for awhile, so
I'm glad to see it coming into existence.  Thanks for the pointers.

This means that there will be a better solution than the one proposed
in the meson-gx-p23x-q20x and meson-gxbb-vega-s95 patches (patches #1
and #2 are not affected by this),> namely:
- I'm currently (ab)using sdio-pwrseq node to power on the UART
Bluetooth module, this can solved by specifying a bluetooth {  } node
inside the &uart_A node in the future and providing the corresponding
GPIOs there

Good.  I didn't like the (ab)use of sdio_pwrseq GPIO resets for the
bluetooth either.

- all the userspace commands will not be necessary once
drivers/bluetooth/hci_bcm.c is changed to the serio framework and gets
devicetree support

Cool.  Maybe you can setup a WIP branch based on Rob's work for broader
testing on Amlogic boards until this gets merged?

In the mean time, I've applied patch 2/4 and hopefully the drivers/tty
patch will go through the serial tree.

Kevin