Re: [PATCH 2/3] drivers:input:ps2 Added sunxi A20 ps2 driver, changed makefile and Kconfig

From: Maxime Ripard
Date: Fri Dec 05 2014 - 05:05:22 EST

Next message: Arnd Bergmann: "Re: [PATCH] optimize ktime_divns for constant divisors"
Previous message: Andreas Larsson: "[PATCH] sparc32,leon: Align ccall_info to prevent unaligned traps on crosscall"
In reply to: Dmitry Torokhov: "Re: [PATCH 2/3] drivers:input:ps2 Added sunxi A20 ps2 driver, changed makefile and Kconfig"
Next in thread: Arnd Bergmann: "Re: [PATCH 2/3] drivers:input:ps2 Added sunxi A20 ps2 driver, changed makefile and Kconfig"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi,

Thanks for working on this.

First, make sure to read Documentation/SubmittingPatches. The
guidelines in there might seem tedious, but it's there to enforce that
we keep the kernel homogenous.

Most notably, you should make sure that:
- you run scripts/checkpatch.pl on your patches
- you have a commit log
- your commit title stays below 80 chars
- you have a Signed-off-by in your commit log

On Thu, Dec 04, 2014 at 04:23:44AM +0530, vishnupatekar wrote:
> ---
> drivers/input/serio/Kconfig | 9 ++
> drivers/input/serio/Makefile | 1 +
> drivers/input/serio/sunxi-ps2.c | 305 +++++++++++++++++++++++++++++++++++++++
> 3 files changed, 315 insertions(+)
> create mode 100644 drivers/input/serio/sunxi-ps2.c
>
> diff --git a/drivers/input/serio/Kconfig b/drivers/input/serio/Kconfig
> index bc2d474..1a86e41 100644
> --- a/drivers/input/serio/Kconfig
> +++ b/drivers/input/serio/Kconfig
> @@ -281,4 +281,13 @@ config HYPERV_KEYBOARD
> To compile this driver as a module, choose M here: the module will
> be called hyperv_keyboard.
>
> +config SERIO_SUNXI_PS2
> + tristate "Allwinner Sun7i-A20 PS/2 controller"
> + default m
> + help
> + Say Y here if you have Sun7i-A20 Allwinner PS/2 ports.
> +
> + To compile this driver as a module, choose M here: the
> + module will be called sunxi-ps2.
> +
> endif
> diff --git a/drivers/input/serio/Makefile b/drivers/input/serio/Makefile
> index 815d874..0fa0f78 100644
> --- a/drivers/input/serio/Makefile
> +++ b/drivers/input/serio/Makefile
> @@ -29,3 +29,4 @@ obj-$(CONFIG_SERIO_ARC_PS2) += arc_ps2.o
> obj-$(CONFIG_SERIO_APBPS2) += apbps2.o
> obj-$(CONFIG_SERIO_OLPC_APSP) += olpc_apsp.o
> obj-$(CONFIG_HYPERV_KEYBOARD) += hyperv-keyboard.o
> +obj-$(CONFIG_SERIO_SUNXI_PS2) += sunxi-ps2.o
> diff --git a/drivers/input/serio/sunxi-ps2.c b/drivers/input/serio/sunxi-ps2.c
> new file mode 100644
> index 0000000..ccd7b29
> --- /dev/null
> +++ b/drivers/input/serio/sunxi-ps2.c
> @@ -0,0 +1,305 @@
> +/*
> + * sunxi-ps2.c Support for Allwinner A20 PS2 host controller
> + *
> + * Author: Aaron.maoye <leafy.myeh@xxxxxxxxxxxxxx>
> + * Vishnu Patekar <vishnupatekar0510@xxxxxxxxx>
> + * Based on sunxi-ps2.c 3.0 kernel

Mentionning that it's Allwinner's 3.0 kernel would be good, it doesn't
make much sense otherwise.

> +#include <linux/module.h>
> +#include <linux/serio.h>
> +#include <linux/interrupt.h>
> +#include <linux/errno.h>
> +#include <linux/slab.h>
> +#include <linux/list.h>
> +#include <linux/io.h>
> +#include <linux/of_address.h>
> +#include <linux/of_device.h>
> +#include <linux/of_irq.h>
> +#include <linux/of_platform.h>
> +#include <linux/clk.h>
> +#include <linux/delay.h>
> +
> +#define DRIVER_NAME "sunxi-ps2"
> +
> +#define RESSIZE(res) (((res)->end - (res)->start)+1)
> +
> +#define SW_PS2_GCTRL (0x00)
> +#define SW_PS2_DATA (0x04)
> +#define SW_PS2_LCTRL (0x08)
> +#define SW_PS2_LSTAT (0x0c)
> +#define SW_PS2_FCTRL (0x10)
> +#define SW_PS2_FSTAT (0x14)
> +#define SW_PS2_CLKDR (0x18)

The SW_ prefix is a left over from Allwinner I suppose?

Using a prefix like sun4i_ps2/sun7i_ps2 depending on the outcome of
the discussion you had with Hans would seem better, across all your
driver.

> +
> +/* SW_PS2_GCTRL */
> +#define SWPS2_BUSEN (1 << 0)
> +#define SWPS2_MASTER (1 << 1)
> +#define SWPS2_RESET (1 << 2)
> +#define SWPS2_INTEN (1 << 3)

We usually have the convention of having the register name mentionned
in the bits offsets, and the bits declared right after their register.

Something like:

#define SUN4I_PS2_GCTRL_REG 0x00
#define SUN4I_PS2_GCTRL_BUS_EN BIT(0)

#define SUN4I_PS2_LCTRL_REG 0x08
#define SUN4I_PS2_LCTRL_PAR_ERR_EN BIT(1)

etc....

> +#define SWPS2_INTFLAG (1 << 3)

This is defined to the same value than _INTEN, is this intentionnal?

> +
> +/* SW_PS2_LCTRL */
> +#define SWPS2_LCTL_NOACK (0x0 << 18)
> +#define SWPS2_LCTL_TXDTOEN (0x1 << 8)
> +#define SWPS2_LCTL_STOPERREN (0x1 << 3)
> +#define SWPS2_LCTL_ACKERREN (0x1 << 2)
> +#define SWPS2_LCTL_PARERREN (0x1 << 1)
> +#define SWPS2_LCTL_RXDTOEN (0x1 << 0)
> +
> +/* SW_PS2_FSTAT */
> +#define SWPS2_FSTA_RXRDY (1 << 0)
> +#define SWPS2_FSTA_RXOF (1 << 1)
> +#define SWPS2_FSTA_RXUF (1 << 2)
> +#define SWPS2_FSTA_TXRDY (1 << 8)
> +#define SWPS2_FSTA_TXOF (1 << 9)
> +#define SWPS2_FSTA_TXUF (1 << 10)
> +
> +#define SW_PS2_SAMPLE_CLK (1000000)
> +#define SW_PS2_SCLK (125000)
> +
> +struct sunxips2data {
> + int irq;
> + spinlock_t ps2_lock;
> + void __iomem *base_address; /* virt address of control registers*/
> + struct serio *serio; /* serio*/
> + struct device *dev;
> + struct clk *pclk;
> +};
> +
> +/*********************/
> +/* Interrupt handler */
> +/*********************/
> +static irqreturn_t sunxips2_interrupt(int irq, void *dev_id)
> +{
> + struct sunxips2data *drvdata = dev_id;
> + u32 intr_status;
> + u32 fifo_status;
> + unsigned char byte;
> + u32 rval;
> + u32 error = 0;
> +
> + spin_lock(&drvdata->ps2_lock);
> +
> + /* Get the PS/2 interrupts and clear them */
> + intr_status = readl(drvdata->base_address + SW_PS2_LSTAT);
> + fifo_status = readl(drvdata->base_address + SW_PS2_FSTAT);
> +
> + /*Check Line Status Register*/
> + if (intr_status & 0x10f) {
> + if (intr_status & 0x08)
> + dev_info(drvdata->dev, "PS/2 Stop Bit Error!");
> + if (intr_status & 0x04)
> + dev_info(drvdata->dev, "PS/2 Acknowledge Error!\n");
> + if (intr_status & 0x02)
> + dev_info(drvdata->dev, "PS/2 Parity Error!\n");
> + if (intr_status & 0x100)
> + dev_info(drvdata->dev, "PS/2 Transmit Data Timeout!\n");
> + if (intr_status & 0x01)
> + dev_info(drvdata->dev, "PS/2 Receive Data Timeout!\n");
> +
> + writel(readl(drvdata->base_address + SW_PS2_GCTRL)|0x4, drvdata->base_address + SW_PS2_GCTRL);/*reset PS/2 controller*/
> + writel(0x10f, drvdata->base_address + SW_PS2_LSTAT);
> +
> + error = 1;
> + }
> +
> + /*Check FIFO Status Register*/
> + if (fifo_status & 0x0606) {
> + if (fifo_status & 0x400)
> + dev_info(drvdata->dev, "PS/2 Tx FIFO Underflow!\n");
> + if (fifo_status & 0x200)
> + dev_info(drvdata->dev, "PS/2 Tx FIFO Overflow!\n");
> + if (fifo_status & 0x04)
> + dev_info(drvdata->dev, "PS/2 Rx FIFO Underflow!\n");
> + if (fifo_status & 0x02)
> + dev_info(drvdata->dev, "PS/2 Rx FIFO Overflow!\n");
> +
> + writel(readl(drvdata->base_address + SW_PS2_GCTRL)|0x4, drvdata->base_address + SW_PS2_GCTRL); /*reset PS/2 controller*/
> + writel(0x707, drvdata->base_address + SW_PS2_FSTAT);
> + error = 1;
> + }
> +
> + rval = (fifo_status >> 16) & 0x3;
> + while (!error && rval--) {
> + byte = readl(drvdata->base_address + SW_PS2_DATA) & 0xff;
> + dev_info(drvdata->dev, "PS/2 Receive %02x\n", byte);
> + serio_interrupt(drvdata->serio, byte, 0);
> + }
> +
> + writel(intr_status, drvdata->base_address + SW_PS2_LSTAT);
> + writel(fifo_status, drvdata->base_address + SW_PS2_FSTAT);
> +
> + spin_unlock(&drvdata->ps2_lock);
> +
> + return IRQ_HANDLED;
> +}
> +
> +static int sunxips2_open(struct serio *pserio)
> +{
> + struct sunxips2data *drvdata = pserio->port_data;
> + u32 src_clk = 0;
> + u32 clk_scdf;
> + u32 clk_pcdf;
> + u32 rval;
> +
> + /*Set Line Control And Enable Interrupt*/
> + rval = SWPS2_LCTL_TXDTOEN|SWPS2_LCTL_STOPERREN|SWPS2_LCTL_ACKERREN|SWPS2_LCTL_PARERREN|SWPS2_LCTL_RXDTOEN;
> + writel(rval, drvdata->base_address + SW_PS2_LCTRL);
> +
> + /*Reset FIFO*/
> + writel(0x3<<16 | 0x607, drvdata->base_address + SW_PS2_FCTRL);
> +
> + src_clk = clk_get_rate(drvdata->pclk);
> +
> + if (!src_clk) {
> + dev_info(drvdata->dev, "w_ps2c_set_sclk error, source clock is 0.");
> + return -1;
> + }
> +
> + /*Set Clock Divider Register*/
> + clk_scdf = ((src_clk + (SW_PS2_SAMPLE_CLK>>1)) / SW_PS2_SAMPLE_CLK - 1);
> + clk_pcdf = ((SW_PS2_SAMPLE_CLK + (SW_PS2_SCLK>>1)) / SW_PS2_SCLK - 1);
> + rval = (clk_scdf<<8) | clk_pcdf;/* | (PS2_DEBUG_SEL<<16);*/

What is this supposed to do? Calculating and rounding the dividers?

You should look into DIV_ROUND_* if that's so

> + writel(rval, drvdata->base_address + SW_PS2_CLKDR);
> +
> + /*Set Global Control Register*/
> + rval = SWPS2_RESET|SWPS2_INTEN|SWPS2_MASTER|SWPS2_BUSEN;
> + writel(rval, drvdata->base_address + SW_PS2_GCTRL);
> +
> + udelay(100);

Why is that udelay needed?

> +
> + return 0;
> +}
> +
> +static void sunxips2_close(struct serio *pserio)
> +{
> + struct sunxips2data *drvdata = pserio->port_data;
> +
> + spin_lock(&drvdata->ps2_lock);

spin_lock_irqsave would be better I guess.

> + /* Disable the PS2 interrupts */
> + writel(0, drvdata->base_address + SW_PS2_GCTRL);
> + spin_unlock(&drvdata->ps2_lock);
> +}
> +
> +static int sunxips2_write(struct serio *pserio, unsigned char val)
> +{
> + struct sunxips2data *drvdata = (struct sunxips2data *)pserio->port_data;
> + u32 timeout = 10000;
> +
> + do {
> + if (readl(drvdata->base_address + SW_PS2_FSTAT) & SWPS2_FSTA_TXRDY) {
> + writel(val, drvdata->base_address + SW_PS2_DATA);
> + return 0;
> + }
> + } while (timeout--);

Please use time_before() here

> + return -1;

And return a meaningful error.

> +static int sunxips2_probe(struct platform_device *ofdev)

The ofdev name is measleading. This is a platform_device structure, it
could be probed by other mechanisms than OF, so please use a different
name here. Usually pdev is used.

> +{
> + struct resource *res; /* IO mem resources */
> + struct sunxips2data *drvdata;
> + struct serio *serio;
> + struct device *dev = &ofdev->dev;
> + unsigned int irq;
> + int error;
> +
> + drvdata = devm_kzalloc(dev, sizeof(struct sunxips2data), GFP_KERNEL);
> + serio = devm_kzalloc(dev, sizeof(struct serio), GFP_KERNEL);
> + if (!drvdata || !serio)
> + error = -ENOMEM;
> +
> + /* Request clock */
> + drvdata->pclk = clk_get(dev, NULL);

You can use devm_clk_get.

> + if (IS_ERR(drvdata->pclk))
> + dev_dbg(dev, "couldn't get clock %li\n",
> + PTR_ERR(drvdata->pclk));

This is an error, you should treat it as such.

> + if (!IS_ERR(drvdata->pclk)) {
> + error = clk_prepare_enable(drvdata->pclk);
> + if (error < 0) {
> + dev_err(dev, "failed to enable clock %d\n", error);
> + return error;
> + }
> + }
> +
> + /* IO */
> + res = platform_get_resource(ofdev, IORESOURCE_MEM, 0);
> + drvdata->base_address = devm_ioremap_resource(dev, res);
> + if (IS_ERR(drvdata->base_address)) {
> + dev_err(dev, "failed to map registers\n");
> + error = PTR_ERR(drvdata->base_address);
> + }

ditto.

> + serio->id.type = SERIO_8042;
> + serio->write = sunxips2_write;
> + serio->open = sunxips2_open;
> + serio->close = sunxips2_close;
> + serio->port_data = drvdata;
> + serio->dev.parent = dev;
> + strlcpy(serio->name, dev_name(dev), sizeof(serio->name));
> + strlcpy(serio->phys, dev_name(dev), sizeof(serio->phys));
> +
> + platform_set_drvdata(ofdev, drvdata);
> + serio_register_port(serio);
> +
> + /* Get IRQ for the device */
> + irq = irq_of_parse_and_map(ofdev->dev.of_node, 0);
> + if (!irq) {
> + dev_err(dev, "no IRQ found\n");
> + return -ENODEV;
> + }

You can use platform_get_irq, for consistency with how you retrieve
resources.

You're also no unregistering the driver from the serio framework.

> + drvdata->irq = irq;
> + drvdata->serio = serio;
> + drvdata->dev = dev;
> + error = devm_request_any_context_irq(drvdata->dev, drvdata->irq, &sunxips2_interrupt, 0,
> + DRIVER_NAME, drvdata);

request_irq is enough here.

> + if (error) {
> + dev_err(drvdata->dev,
> + "Couldn't allocate interrupt %d : error: %d\n", drvdata->irq, error);
> + return error;
> + }
> + return 0; /* success */
> +}
> +
> +static int sunxips2_remove(struct platform_device *of_dev)
> +{
> + struct sunxips2data *drvdata = platform_get_drvdata(of_dev);
> +
> + if (!IS_ERR(drvdata->pclk)) {
> + clk_disable_unprepare(drvdata->pclk);
> + clk_put(drvdata->pclk);
> + }
> + serio_unregister_port(drvdata->serio);
> + mdelay(2);

Why a mdelay?

> + return 0;
> +}
> +
> +/* Match table for of_platform binding */
> +static const struct of_device_id sunxips2_of_match[] = {
> + { .compatible = "allwinner,sun7i-a20-ps2", },
> + { },
> +};
> +
> +MODULE_DEVICE_TABLE(of, sunxips2_of_match);
> +
> +/*platform driver structure*/
> +static struct platform_driver sunxips2_of_driver = {
> + .probe = sunxips2_probe,
> + .remove = sunxips2_remove,
> + .driver = {
> + .name = DRIVER_NAME,
> + .owner = THIS_MODULE,

You can drop the owner field, it's already set by
module_platform_driver.

> + .of_match_table = sunxips2_of_match,
> + },
> +};
> +module_platform_driver(sunxips2_of_driver);
> +
> +MODULE_AUTHOR("Aaron.maoye<leafy.myeh@xxxxxxxxxxxxxx, "
> + "Vishnu Patekar <vishnupatekar0510@xxxxxxxxx>");

Usually, you define two MODULE_AUTHOR.

Maxime

--
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

Attachment: signature.asc
Description: Digital signature

Next message: Arnd Bergmann: "Re: [PATCH] optimize ktime_divns for constant divisors"
Previous message: Andreas Larsson: "[PATCH] sparc32,leon: Align ccall_info to prevent unaligned traps on crosscall"
In reply to: Dmitry Torokhov: "Re: [PATCH 2/3] drivers:input:ps2 Added sunxi A20 ps2 driver, changed makefile and Kconfig"
Next in thread: Arnd Bergmann: "Re: [PATCH 2/3] drivers:input:ps2 Added sunxi A20 ps2 driver, changed makefile and Kconfig"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]