Message-ID: <YH2k5xnD/+CKnMBQ@hovoldconsulting.com>
Date:   Mon, 19 Apr 2021 17:42:31 +0200
From:   Johan Hovold <johan@...nel.org>
To:     Pho Tran <photranvan0712@...il.com>
Cc:     gregkh@...uxfoundation.org, linux-usb@...r.kernel.org,
        linux-kernel@...r.kernel.org, Hung.Nguyen@...abs.com,
        Tung.Pham@...abs.com, Pho Tran <pho.tran@...abs.com>
Subject: Re: [PATCH v9] USB: serial: cp210x: Add support for GPIOs on CP2108

On Thu, Apr 08, 2021 at 05:36:07PM +0700, Pho Tran wrote:
> From: Pho Tran <pho.tran@...abs.com>
> 
> Similar to other CP210x devices, GPIO interfaces (gpiochip) should be
> supported for CP2108.
> 
> CP2108 has 4 serial interfaces but only 1 set of GPIO pins are shared
> to all of those interfaces. So, just need to initialize GPIOs of CP2108
> with only one interface (I use interface 0). It means just only 1 gpiochip
> device file will be created for CP2108.
> 
> CP2108 has 16 GPIOs, So data types of several variables need to be is u16
> instead of u8(in struct cp210x_serial_private). This doesn't affect other
> CP210x devices.
> 
> Because CP2108 has 16 GPIO pins, the parameter passed by cp210x functions
> will be different from other CP210x devices. So need to check part number
> of the device to use correct data format  before sending commands to
> devices.
> 
> Like CP2104, CP2108 have GPIO pins with configurable options. Therefore,
> should be mask all pins which are not in GPIO mode in cp2108_gpio_init()
> function.
> 
> Signed-off-by: Pho Tran <pho.tran@...abs.com>
> ---
> 
> 04/08/2021: Patch v8 Fixed build warning reported by kernel test robot
> with ARCH=i386
> 04/05/2021: Patch v7 Modified commit message follow Greg's comment.
> 04/05/2021: Patch v6 Fixed build warning reported by kernel test robot
> with ARCH=x86_64
> 03/15/2021: Patch v5 Modified code according to comment of Johan:
> 	1. Unified the handling of CP2108 and other types and
> 	take care about endianness.
> 	2. Used suitable types data for variable.
> 	3. Fixed cp2108_gpio_init and add more detail on
> 	commit message and comment.
> 	4. Dropped some of the ones that don't add any value.

You left out a few changes here. You changed how the altfunctions were
detected and fixed the gpio_init error handling. Please include all
relevant changes in your changelogs.

> 03/12/2021: Patch v4 used git send-mail instead of send patch by manual
> follow the instructions of Johan Hovold <johan@...nel.org>.
> 03/05/2021: Patch v3 modified format and contents of changelog follow feedback
> from Johan Hovold <johan@...nel.org>.
> 03/04/2021: Patch v2 modified format patch as comment from
> Johan Hovold <johan@...nel.org>:
> 	1. Break commit message lines at 80 cols
> 	2. Use kernel u8 and u16 instead of the c99 ones.
> 03/01/2021: Initialed submission of patch "Make the CP210x driver work with
> GPIOs of CP2108.".
> 
>  drivers/usb/serial/cp210x.c | 254 +++++++++++++++++++++++++++++++-----
>  1 file changed, 220 insertions(+), 34 deletions(-)
> 
> diff --git a/drivers/usb/serial/cp210x.c b/drivers/usb/serial/cp210x.c
> index 7bec1e730b20..3812aac2b015 100644
> --- a/drivers/usb/serial/cp210x.c
> +++ b/drivers/usb/serial/cp210x.c
> @@ -245,9 +245,9 @@ struct cp210x_serial_private {
>  #ifdef CONFIG_GPIOLIB
>  	struct gpio_chip	gc;
>  	bool			gpio_registered;
> -	u8			gpio_pushpull;
> -	u8			gpio_altfunc;
> -	u8			gpio_input;
> +	u16			gpio_pushpull;
> +	u16			gpio_altfunc;
> +	u16			gpio_input;
>  #endif
>  	u8			partnum;
>  	speed_t			min_speed;
> @@ -399,6 +399,18 @@ static struct usb_serial_driver * const serial_drivers[] = {
>  #define CP210X_PARTNUM_CP2102N_QFN20	0x22
>  #define CP210X_PARTNUM_UNKNOWN	0xFF
>  
> +/*
> + * CP2108 Define bit locations for EnhancedFxn_IFCx
> + * Refer to https://www.silabs.com/documents/public/application-notes/an978-cp210x-usb-to-uart-api-specification.pdf
> + * for more information.

Thanks for the reference, but no need to refer to this one more than
once; the quad-port-config comment below should be enough.

> + */
> +#define EF_IFC_GPIO_TXLED		0x01
> +#define EF_IFC_GPIO_RXLED		0x02
> +#define EF_IFC_GPIO_RS485		0x04
> +#define EF_IFC_GPIO_RS485_LOGIC 0x08
> +#define EF_IFC_GPIO_CLOCK		0x10
> +#define EF_IFC_DYNAMIC_SUSPEND	0x40
> +

These should go after the quad-port-config structure with the other port
config defines.

And shouldn't they have a CP2108_ prefix?

>  /* CP210X_GET_COMM_STATUS returns these 0x13 bytes */
>  struct cp210x_comm_status {
>  	__le32   ulErrors;
> @@ -500,6 +512,45 @@ struct cp210x_single_port_config {
>  	u8	device_cfg;
>  } __packed;
>  
> +/*
> + * Quad Port Config definitions
> + * Refer to https://www.silabs.com/documents/public/application-notes/an978-cp210x-usb-to-uart-api-specification.pdf
> + * for more information.
> + * CP210X_VENDOR_SPECIFIC, CP210X_GET_PORTCONFIG call reads these 0x49 bytes
> + * on a CP2108 chip.
> + * CP2108 Quad Port State structure(used in Quad Port Config structure)

You can drop this last sentence; clear from the code.

> + */
> +struct cp210x_quad_port_state {
> +	__le16 gpio_mode_PB0;
> +	__le16 gpio_mode_PB1;
> +	__le16 gpio_mode_PB2;
> +	__le16 gpio_mode_PB3;
> +	__le16 gpio_mode_PB4;
> +
> +

Double new line.

> +	__le16 gpio_lowpower_PB0;
> +	__le16 gpio_lowpower_PB1;
> +	__le16 gpio_lowpower_PB2;
> +	__le16 gpio_lowpower_PB3;
> +	__le16 gpio_lowpower_PB4;
> +
> +	__le16 gpio_latch_PB0;
> +	__le16 gpio_latch_PB1;
> +	__le16 gpio_latch_PB2;
> +	__le16 gpio_latch_PB3;
> +	__le16 gpio_latch_PB4;

Use lowercase pb for consistency?

> +};
> +
> +// Cp2108 Quad Port Config structure

Please avoid // comments. CP2108 (uppercase P)?

> +struct cp210x_quad_port_config {
> +	struct cp210x_quad_port_state reset_state;
> +	struct cp210x_quad_port_state suspend_state;
> +	u8 ipdelay_IFC[4];
> +	u8 enhancedfxn_IFC[4];

Lowercase ifc?

> +	u8 enhancedfxn_device;
> +	u8 extclkfreq[4];
> +} __packed;
> +
>  /* GPIO modes */
>  #define CP210X_SCI_GPIO_MODE_OFFSET	9
>  #define CP210X_SCI_GPIO_MODE_MASK	GENMASK(11, 9)
> @@ -510,6 +561,9 @@ struct cp210x_single_port_config {
>  #define CP210X_GPIO_MODE_OFFSET		8
>  #define CP210X_GPIO_MODE_MASK		GENMASK(11, 8)
>  
> +#define CP2108_GPIO_MODE_OFFSET		0
> +#define CP2108_GPIO_MODE_MASK		GENMASK(15, 0)
> +

Perhaps you can skip these (see below).

>  /* CP2105 port configuration values */
>  #define CP2105_GPIO0_TXLED_MODE		BIT(0)
>  #define CP2105_GPIO1_RXLED_MODE		BIT(1)
> @@ -526,12 +580,31 @@ struct cp210x_single_port_config {
>  #define CP210X_2NCONFIG_GPIO_RSTLATCH_IDX	587
>  #define CP210X_2NCONFIG_GPIO_CONTROL_IDX	600
>  
> -/* CP210X_VENDOR_SPECIFIC, CP210X_WRITE_LATCH call writes these 0x2 bytes. */
> -struct cp210x_gpio_write {
> +/*
> + * CP210X_VENDOR_SPECIFIC, CP210X_WRITE_LATCH call writes these
> + * 0x04 bytes on CP2108.
> + */
> +struct cp210x_16gpios_write {
> +	__le16	mask;
> +	__le16	state;
> +};
> +
> +/*
> + * CP210X_VENDOR_SPECIFIC, CP210X_WRITE_LATCH call writes these
> + * 0x02 bytes on CP2102N, Cp2103, Cp2104 and CP2105.
> + */
> +struct cp210x_8gpios_write {
>  	u8	mask;
>  	u8	state;
>  };
>  
> +//Struct cp210x_gpio_write include devices have both of 8 gpios and 16 gpios.
> +struct cp210x_gpio_write {
> +	struct cp210x_8gpios_write cp210x_8gpios;
> +	struct cp210x_16gpios_write cp210x_16gpios;
> +};
> +
> +

This doesn't make any sense. First I though you used a union, but also
that would seem a bit unnecessary.

Just keep the current struct as is and add a second one for 16 bits
(e.g. struct cp210x_gpio_write16).

>  /*
>   * Helper to get interface number when we only have struct usb_serial.
>   */
> @@ -1298,21 +1371,46 @@ static int cp210x_gpio_get(struct gpio_chip *gc, unsigned int gpio)
>  	struct cp210x_serial_private *priv = usb_get_serial_data(serial);
>  	u8 req_type = REQTYPE_DEVICE_TO_HOST;
>  	int result;
> -	u8 buf;
> -
> -	if (priv->partnum == CP210X_PARTNUM_CP2105)
> -		req_type = REQTYPE_INTERFACE_TO_HOST;
> +	u16 buf;
> +	__le16 wbuf;

You don't need both, but it may be cleaner to use u8 buf[2] for the
transfer buffer.

>  	result = usb_autopm_get_interface(serial->interface);
>  	if (result)
>  		return result;
> -
> -	result = cp210x_read_vendor_block(serial, req_type,
> -					  CP210X_READ_LATCH, &buf, sizeof(buf));
> -	usb_autopm_put_interface(serial->interface);
> +/*
> + * This function will be read latch value of gpio and storage to buf(16bit)
> + * where bit 0 is GPIO0, bit 1 is GPIO1, etc. Up to GPIOn where n is
> + * total number of GPIO pins the interface supports.
> + * Interfaces on CP2102N supports 7 GPIOs
> + * Interfaces on CP2103 amd CP2104 supports 4 GPIOs
> + * Enhanced interfaces on CP2105 support 3 GPIOs
> + * Standard interfaces on CP2105 support 4 GPIOs
> + * Interfaces on CP2108 supports 16 GPIOs
> + */

Missing indentation.

But not sure this adds much to what can be understood from the code.

> +	switch (priv->partnum) {
> +	/*
> +	 * Request type to Read_Latch of CP2105 and Cp2108
> +	 * is 0xc1 <REQTYPE_INTERFACE_TO_HOST>
> +	 */

That is clear from the code (but please try to be consistent with
capitalisation of CP210x).

> +	case CP210X_PARTNUM_CP2108:
> +		req_type = REQTYPE_INTERFACE_TO_HOST;
> +		result = cp210x_read_vendor_block(serial, req_type,
> +						CP210X_READ_LATCH, &wbuf, sizeof(__le16));
> +		break;
> +	case CP210X_PARTNUM_CP2105:
> +		req_type = REQTYPE_INTERFACE_TO_HOST;
> +		result = cp210x_read_vendor_block(serial, req_type,
> +						CP210X_READ_LATCH, &wbuf, sizeof(u8));
> +		break;
> +	default:
> +		result = cp210x_read_vendor_block(serial, req_type,
> +						CP210X_READ_LATCH, &wbuf, sizeof(u8));
> +		break;
> +	}

Just do the read after the switch into a common u8 buf[2] and use the
full length only for CP2108.

>  	if (result < 0)
>  		return result;

You leak the PM counter reference here.

> -
> +	buf = le16_to_cpu(wbuf);

And then this can be le16_to_cpup((__le16 *)buf).

> +	usb_autopm_put_interface(serial->interface);
>  	return !!(buf & BIT(gpio));
>  }
>  
> @@ -1321,37 +1419,49 @@ static void cp210x_gpio_set(struct gpio_chip *gc, unsigned int gpio, int value)
>  	struct usb_serial *serial = gpiochip_get_data(gc);
>  	struct cp210x_serial_private *priv = usb_get_serial_data(serial);
>  	struct cp210x_gpio_write buf;
> +	u16 wIndex;
>  	int result;
>  
> -	if (value == 1)
> -		buf.state = BIT(gpio);
> -	else
> -		buf.state = 0;
> -
> -	buf.mask = BIT(gpio);
> +	if (value == 1) {
> +		buf.cp210x_8gpios.state = BIT(gpio);
> +		buf.cp210x_16gpios.state = cpu_to_le16(BIT(gpio));
> +	} else {
> +		buf.cp210x_8gpios.state = 0;
> +		buf.cp210x_16gpios.state = 0;
> +	}
> +	buf.cp210x_8gpios.mask = BIT(gpio);
> +	buf.cp210x_16gpios.mask = cpu_to_le16(BIT(gpio));

This is just just really messy. Please keep the original struct as is,
and possibly add another one for cp2108 with a suitable suffix (e.g.
"16"). It should be fine to keep both on the stack.

Here you could use two shared u16 for state and mask. Do the endianess
conversion when setting up the cp2108 struct.

>  	result = usb_autopm_get_interface(serial->interface);
>  	if (result)
>  		goto out;
>  
> -	if (priv->partnum == CP210X_PARTNUM_CP2105) {
> +	switch (priv->partnum) {
> +	case CP210X_PARTNUM_CP2108:
>  		result = cp210x_write_vendor_block(serial,
> -						   REQTYPE_HOST_TO_INTERFACE,
> -						   CP210X_WRITE_LATCH, &buf,
> -						   sizeof(buf));
> -	} else {
> -		u16 wIndex = buf.state << 8 | buf.mask;
> -
> +							REQTYPE_HOST_TO_INTERFACE,
> +							CP210X_WRITE_LATCH, &buf.cp210x_16gpios,
> +							sizeof(buf.cp210x_16gpios));
> +		break;
> +	case CP210X_PARTNUM_CP2105:
> +		result = cp210x_write_vendor_block(serial,
> +							REQTYPE_HOST_TO_INTERFACE,
> +							CP210X_WRITE_LATCH, &buf.cp210x_8gpios,
> +							sizeof(buf.cp210x_8gpios));
> +		break;
> +	default:
> +		wIndex = buf.cp210x_8gpios.state << 8 | buf.cp210x_8gpios.mask;
>  		result = usb_control_msg(serial->dev,
> -					 usb_sndctrlpipe(serial->dev, 0),
> -					 CP210X_VENDOR_SPECIFIC,
> -					 REQTYPE_HOST_TO_DEVICE,
> -					 CP210X_WRITE_LATCH,
> -					 wIndex,
> -					 NULL, 0, USB_CTRL_SET_TIMEOUT);
> +							usb_sndctrlpipe(serial->dev, 0),
> +							CP210X_VENDOR_SPECIFIC,
> +							REQTYPE_HOST_TO_DEVICE,
> +							CP210X_WRITE_LATCH,
> +							wIndex,
> +							NULL, 0, USB_CTRL_SET_TIMEOUT);
> +		break;

The indentation is just way off here. No need to do open-parenthesis
alignment; two tabs is enough. And try to stay below 80 columns.

>  	}
> -
>  	usb_autopm_put_interface(serial->interface);
> +

Random whitespace changes.

>  out:
>  	if (result < 0) {
>  		dev_err(&serial->interface->dev, "failed to set GPIO value: %d\n",
> @@ -1420,6 +1530,73 @@ static int cp210x_gpio_set_config(struct gpio_chip *gc, unsigned int gpio,
>  	return -ENOTSUPP;
>  }
>  
> +static int cp2108_gpio_init(struct usb_serial *serial)
> +{
> +	struct cp210x_serial_private *priv = usb_get_serial_data(serial);
> +	struct cp210x_quad_port_config config;
> +	u16 gpio_latch;
> +	u16 temp;
> +	int result;
> +	u8 i;
> +
> +	result = cp210x_read_vendor_block(serial, REQTYPE_DEVICE_TO_HOST,
> +					  CP210X_GET_PORTCONFIG, &config,
> +					  sizeof(config));
> +	if (result < 0)
> +		return result;
> +	priv->gc.ngpio = 16;
> +	temp = le16_to_cpu(config.reset_state.gpio_mode_PB1);
> +	priv->gpio_pushpull = (temp & CP2108_GPIO_MODE_MASK) >> CP2108_GPIO_MODE_OFFSET;
> +	temp = le16_to_cpu(config.reset_state.gpio_latch_PB1);
> +	gpio_latch = (temp & CP2108_GPIO_MODE_MASK) >> CP2108_GPIO_MODE_OFFSET;

Perhaps you can just skip the mask and shift (and temp) above since all
sixteen bits of PB1 are used for the GPIOs.

> +	/*
> +	 * Mark all pins which are not in GPIO mode
> +	 * Refer to table 9.1: GPIO Mode alternate Functions on CP2108 datasheet:
> +	 * https://www.silabs.com/documents/public/data-sheets/cp2108-datasheet.pdf
> +	 * Alternate Functions of GPIO0 to GPIO3 is determine by enhancedfxn_IFC[0]
> +	 * and the same for other pins, enhancedfxn_IFC[1]: GPIO4 to GPIO7,
> +	 * enhancedfxn_IFC[2]: GPIO8 to GPIO11, enhancedfxn_IFC[3]: GPIO12 to GPIO15.
> +	 */
> +	for (i = 0; i < 4; i++) {
> +		switch (config.enhancedfxn_IFC[i]) {

This doesn't look correct. Isn't enhancedfxn_IFC a bitmask so that you
need to check each bit?

> +		case EF_IFC_GPIO_TXLED:
> +			priv->gpio_altfunc |= BIT(i * 4);
> +			break;
> +		case EF_IFC_GPIO_RXLED:
> +			priv->gpio_altfunc |= BIT((i * 4) + 1);
> +			break;
> +		case EF_IFC_GPIO_RS485_LOGIC:

This isn't a alternate function bit AFAICT. It just determines the logic
level when in RS485 mode. It would be good to include that comment from
an978 with the defines.

> +		case EF_IFC_GPIO_RS485:
> +			priv->gpio_altfunc |= BIT((i * 4) + 2);
> +			break;
> +		case EF_IFC_GPIO_CLOCK:
> +			priv->gpio_altfunc |= BIT((i * 4) + 3);
> +			break;
> +		case EF_IFC_DYNAMIC_SUSPEND:
> +			priv->gpio_altfunc |= BIT(i * 4);
> +			priv->gpio_altfunc |= BIT((i * 4) + 1);
> +			priv->gpio_altfunc |= BIT((i * 4) + 2);
> +			priv->gpio_altfunc |= BIT((i * 4) + 3);
> +			break;
> +		}

How does EF_IFC_DYNAMIC_SUSPEND work? Why does that prevent using all
four pins as GPIO?

> +	}
> +	/*
> +	 * Like CP2102N, CP2108 has also no strict input and output pin
> +	 * modes.
> +	 * Do the same input mode emulation as CP2102N.
> +	 */
> +	for (i = 0; i < priv->gc.ngpio; ++i) {
> +		/*
> +		 * Set direction to "input" iff pin is open-drain and reset
> +		 * value is 1.
> +		 */
> +		if (!(priv->gpio_pushpull & BIT(i)) && (gpio_latch & BIT(i)))
> +			priv->gpio_input |= BIT(i);
> +	}
> +
> +	return 0;
> +}
> +
>  /*
>   * This function is for configuring GPIO using shared pins, where other signals
>   * are made unavailable by configuring the use of GPIO. This is believed to be
> @@ -1649,6 +1826,15 @@ static int cp210x_gpio_init(struct usb_serial *serial)
>  	case CP210X_PARTNUM_CP2102N_QFN20:
>  		result = cp2102n_gpioconf_init(serial);
>  		break;
> +	case CP210X_PARTNUM_CP2108:
> +		/*
> +		 * The GPIOs are not tied to any specific port so onlu register

typo: only

> +		 * once for interface 0.
> +		 */
> +		if (cp210x_interface_num(serial) != 0)
> +			return 0;
> +		result = cp2108_gpio_init(serial);
> +		break;
>  	default:
>  		return 0;
>  	}

Johan

Powered by blists - more mailing lists