On Mon, Jan 30, 2017 at 11:44 PM, Sebastian Reichel [off-list ref] wrote:

Hi,

On Tue, Jan 31, 2017 at 04:02:41AM +0100, Sebastian Reichel wrote:

quoted

On Wed, Jan 25, 2017 at 02:23:38PM -0600, Rob Herring wrote:

quoted

Here's V3 of the serdev bus support with all the review feedback so far
incorporated. Greg applied the first 4 patches already. I also found a
problem in the tty_port_wakeup changes and have fixed that.

I've added serdev support to the BT hci_ll.c driver as it turns out it works
on newer TI chips too, but just needed firmware loading. That avoids the
impossible mess of cleaning up the TI-ST driver, and it's in a much closer
state to being ready for upstream. I've sent that out separately.

Changelog is in individual patches. Previous versions are here[1][2]. This
series and the mentioned drivers can be found here[3].

Rob

[1] http://www.mail-archive.com/linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org/msg1304151.html
[2] http://www.mail-archive.com/linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org/msg1311650.html
[3] git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux.git serial-bus-v4

I rebased my nokia bluetooth driver on top of [3]. It's not yet
working, but everything added by this patchset (and
"bluetooth: hci_uart: add serdev driver support library") seems
to work as expected, so for the whole series:

Tested-By: Sebastian Reichel <redacted>

Thanks.

FWIW, I got bluetooth on top of your patchset working on N950. N900
is not yet tested and probably does not yet work (it's more picky
about flow control and usage of the wakeup gpios). You can find a
branch with the additional patches @ [4]. I had to add serdev
functions to wait for TX buffer depletion and checking CTS. Apart
from that flow control only worked correctly when also taking care
of RTS (like hci_uart_set_flow_control does).

I'd originally copied hci_uart_set_flow_control, but then realized
that it didn't really do what I wanted. What I have now is turn on or
off h/w control (i.e. off is no RTS/CTS signals connected). What
hci_uart_set_flow_control does is enable h/w flow control
(enable=false) or force assert RTS (enable=true). I guess it didn't
help that I used the same _set_flow_control naming with different
behavior. I did expect we'd also need something for RTS control, but
figured we'd add it as needed and wanted to come up with something a
bit clearer as to the function (i.e. serdev_device_pause_rx or
serdev_device_assert_rts).

I noticed, that you did not yet write any DT bindings for your
bluetooth slaves. Did you already decide for a location?

Marvell BT is already in bindings/net/.

Rob

[4] git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-n900.git serial-bus-v4-nokia-bt

-- Sebastian