On Thu, 11 Jun 2015 23:06:54 -0500
Rob Herring <robherring2@xxxxxxxxx> wrote:
> Adding Neil Brown who has been doing a UART slave binding[1].
Thanks for the heads-up.
This is quite a different approach to configuring a device that is attached to
a UART, but then I think it is trying to handle different issues than my
approach and the two can easily co-exist.
My need was for a driver to be activated when the tty device was opened.
So the uart needs to examine the child in devicetree, find the "compatible"
driver, probe it, and allow it to intercept "open" requests.
This need is to be able to configure a particular driver that is attached to a
particular line discipline.
So (if I understand correctly), the /dev/ttyXX would be opened, an line
discipline would be selected, and ioctl to that line discipline would select
the driver and then that drive goes and looks in devicetree, find the
configuration, and makes use of it.
It may be a pipe dream, but it might be nice if these could be combined. When
the uart looks in device tree, if it finds a child which specifies a
line-discipline, it could activate that line discipline directly which could
then extract other information and activate the relevant drive, which could
then gather the relevant configuration information.
I do think my current code would get confused if it found this device tree node
under a uart.
It currently thinks any device tree node with a "compatible" attribute implies
a device that will have a driver. So it won't register the tty device until
that driver appears. That would be a problem for this NFC driver.
It probably makes sense to change my approach so that the tty still gets
registered immediately, but when the driver appears it gets connected in to the
tty somehow. I'll have to give that some thought.
Thanks,
NeilBrown
>
> On Thu, Jun 11, 2015 at 7:00 AM, Vincent Cuissard <cuissard@xxxxxxxxxxx> wrote:
> > Add support of Marvell NFC chip controlled over UART
> >
> > Signed-off-by: Vincent Cuissard <cuissard@xxxxxxxxxxx>
> > ---
> > .../devicetree/bindings/net/nfc/nfcmrvl.txt | 29 +++
> > drivers/nfc/nfcmrvl/Kconfig | 11 +
> > drivers/nfc/nfcmrvl/Makefile | 3 +
> > drivers/nfc/nfcmrvl/nfcmrvl.h | 7 +
> > drivers/nfc/nfcmrvl/uart.c | 225 +++++++++++++++++++++
> > include/linux/platform_data/nfcmrvl.h | 9 +
> > net/nfc/nci/uart.c | 1 -
> > 7 files changed, 284 insertions(+), 1 deletion(-)
> > create mode 100644 Documentation/devicetree/bindings/net/nfc/nfcmrvl.txt
> > create mode 100644 drivers/nfc/nfcmrvl/uart.c
> >
> > diff --git a/Documentation/devicetree/bindings/net/nfc/nfcmrvl.txt b/Documentation/devicetree/bindings/net/nfc/nfcmrvl.txt
> > new file mode 100644
> > index 0000000..7c4a0cc
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/net/nfc/nfcmrvl.txt
> > @@ -0,0 +1,29 @@
> > +* Marvell International Ltd. NCI NFC Controller
> > +
> > +Required properties:
> > +- compatible: Should be "mrvl,nfc-uart".
>
> This should describe the chip, not the connection.
>
> marvell is the preferred vendor prefix. mrvl is deprecated.
>
>
> > +Optional SoC specific properties:
> > +- pinctrl-names: Contains only one value - "default".
> > +- pintctrl-0: Specifies the pin control groups used for this controller.
>
> What pins are you controlling exactly? The UART signals on the host?
>
> > +- reset-n-io: Output GPIO pin used to reset the chip (active low).
>
> Should be suffixed with "-gpio". Is this reset for the chip or only NFC?
>
> > +- hci-muxed: Specifies that the chip is muxing NCI over HCI frames.
> > +
> > +Optional UART-based chip specific properties:
> > +- flow-control: Specifies that the chip is using RTS/CTS.
>
> Is this a setting in the NFC chip or just how the UART is configured.
> If the latter, we already have standard UART property to enable
> flow-control on a UART.
>
> > +- break-control: Specifies that the chip needs specific break management.
>
> Can you be more specific?
>
> > +
> > +Example (for ARM-based BeagleBoard Black with 88W8887 on UART5):
> > +
> > +&uart5 {
> > + status = "okay";
> > +
> > + nfcmrvluart: nfcmrvluart@5 {
>
> Unit addresses should have a corresponding reg property. You don't
> need a reg property here, so drop the unit address. Names should allow
> be generic, so just "nfc" is probably sufficient here. Or maybe
> "uart-slave".
>
> How are you doing the binding for the rest of the 8887 chip which has
> also has BT and WiFi. We need to figure out how to do the whole chip
> which is a bit more complicated, but quite common. We could split up
> each function, but then I worry that there may be some
> interdependencies between functions on the chip.
>
> > + compatible = "mrvl,nfc-uart";
> > +
> > + reset-n-io = <&gpio3 16 0>;
> > +
> > + hci-muxed;
> > + flow-control;
> > + }
> > +};
>
> Rob
>
> [1] https://lwn.net/Articles/643878/
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
