RE: [RFC PATCH v3 1/2] usb: typec: USB Type-C Port Manager (tcpm)

[Jun Li]

Hi Guenter,

> -----Original Message-----
> From: linux-usb-owner@xxxxxxxxxxxxxxx [mailto:linux-usb-
> owner@xxxxxxxxxxxxxxx] On Behalf Of Guenter Roeck
> Sent: Wednesday, August 24, 2016 5:11 AM
> To: Felipe Balbi <felipe.balbi@xxxxxxxxxxxxxxx>
> Cc: Chandra Sekhar Anagani <chandra.sekhar.anagani@xxxxxxxxx>; Bruce
> Ashfield <bruce.ashfield@xxxxxxxxxxxxx>; Bin Gao <bin.gao@xxxxxxxxx>;
> Pranav Tipnis <pranav.tipnis@xxxxxxxxx>; Heikki Krogerus
> <heikki.krogerus@xxxxxxxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx; linux-
> usb@xxxxxxxxxxxxxxx; Guenter Roeck <groeck@xxxxxxxxxxxx>
> Subject: [RFC PATCH v3 1/2] usb: typec: USB Type-C Port Manager (tcpm)
> 
> This driver implements the USB Type-C Power Delivery state machine
> for both source and sink ports. Alternate mode support is not
> fully implemented.
> 
> The driver attaches to the USB Type-C class code implemented in
> the following patches.
> 
> 	usb: typec: add driver for Intel Whiskey Cove PMIC USB Type-C PHY
> 	usb: USB Type-C connector class
> 
> This driver only implements the state machine. Lower level drivers are
> responsible for
> - Reporting VBUS status and activating VBUS
> - Setting CC lines and providing CC line status
> - Setting line polarity
> - Activating and deactivating VCONN
> - Setting the current limit
> - Activating and deactivating PD message transfers
> - Sending and receiving PD messages
> 
> The driver provides both a functional API as well as callbacks for
> lower level drivers.
> 
> Signed-off-by: Guenter Roeck <groeck@xxxxxxxxxxxx>
> ---
> v3:
> - Improve TCPM state machine resiliency if there are spurious CC line
> changes
>   while the state machine is in a transient change (waiting for a timeout)
> - Update current limit after CC voltage level changes on a port which is
> not
>   PD capable.
> 
> v2:
> - Only update polarity if setting it was successful
>   If setting the CC line polarity in the driver was not successful,
>   don't update the internal polarity state.
> - All PD messages are little endian; convert to and from CPU endianness.
> - Avoid comparisons against NULL.
> - Use u8/u16/u32 instead of uint8_t/uint16_t/uint32_t consistently.
> - Callbacks into tcpm need to be lockless to avoid timing problems
>   in low level drivers.
> - Simplify callbacks; tcpm can request the current state of cc/vbus
>   when it is ready to use it.
> 
>  drivers/usb/typec/Kconfig  |    7 +
>  drivers/usb/typec/Makefile |    1 +
>  drivers/usb/typec/tcpm.c   | 3163
> ++++++++++++++++++++++++++++++++++++++++++++
>  drivers/usb/typec/tcpm.h   |  137 ++
>  include/linux/usb/pd.h     |  282 ++++
>  include/linux/usb/pd_bdo.h |   31 +
>  include/linux/usb/pd_vdo.h |  412 ++++++
>  7 files changed, 4033 insertions(+)
>  create mode 100644 drivers/usb/typec/tcpm.c
>  create mode 100644 drivers/usb/typec/tcpm.h
>  create mode 100644 include/linux/usb/pd.h
>  create mode 100644 include/linux/usb/pd_bdo.h
>  create mode 100644 include/linux/usb/pd_vdo.h
> 

...

> +
> +static void run_state_machine(struct tcpm_port *port)
> +{
> +	int ret;
> +
> +	port->enter_state = port->state;
> +	switch (port->state) {
> +	/* SRC states */
> +	case SRC_UNATTACHED:
> +		tcpm_swap_complete(port, -ENOTCONN);
> +		tcpm_src_detach(port);
> +		tcpm_set_cc(port, TYPEC_CC_RP_DEF);
> +		if (port->typec_caps.type == TYPEC_PORT_DRP)
> +			tcpm_set_state(port, SNK_UNATTACHED, PD_T_DRP_SNK);

With this and below, after disconnect, the DRP port state machine will be
in infinite loop of state transition between SRC_UNATTACHED <--> SNK_UNATTACHED,
correct?

Li Jun
...

> +	/* SNK states */
> +	case SNK_UNATTACHED:
> +		tcpm_swap_complete(port, -ENOTCONN);
> +		tcpm_snk_detach(port);
> +		tcpm_set_cc(port, TYPEC_CC_RD);
> +		if (port->typec_caps.type == TYPEC_PORT_DRP)
> +			tcpm_set_state(port, SRC_UNATTACHED, PD_T_DRP_SRC);
> +		break;
>