diff mbox

[v5,2/2] i2c: sunxi: add P2WI (Push/Pull 2 Wire Interface) controller support

Message ID 1402408036-5235-3-git-send-email-boris.brezillon@free-electrons.com (mailing list archive)
State New, archived
Headers show

Commit Message

Boris BREZILLON June 10, 2014, 1:47 p.m. UTC
The P2WI controller looks like an SMBus controller which only supports byte
data transfers. But, it differs from standard SMBus protocol on several
aspects:
- it supports only one slave device, and thus drop the address field
- it adds a parity bit every 8bits of data
- only one read access is required to read a byte (instead of a write
  followed by a read access in standard SMBus protocol)
- there's no Ack bit after each byte transfer

This means this bus cannot be used to interface with standard SMBus
devices (the only known device to support this interface is the AXP221
PMIC).

Signed-off-by: Boris BREZILLON <boris.brezillon@free-electrons.com>
---
 drivers/i2c/busses/Kconfig          |  12 ++
 drivers/i2c/busses/Makefile         |   1 +
 drivers/i2c/busses/i2c-sun6i-p2wi.c | 356 ++++++++++++++++++++++++++++++++++++
 3 files changed, 369 insertions(+)
 create mode 100644 drivers/i2c/busses/i2c-sun6i-p2wi.c

Comments

Arnd Bergmann June 10, 2014, 1:54 p.m. UTC | #1
On Tuesday 10 June 2014 15:47:16 Boris BREZILLON wrote:
> 
> +config I2C_SUN6I_P2WI
> +       tristate "Allwinner sun6i internal P2WI controller"
> +       depends on ARCH_SUNXI
> +       help
> +         If you say yes to this option, support will be included for the
> +         P2WI (Push/Pull 2 Wire Interface) controller embedded in some sunxi
> +         SOCs.
> +         The P2WI looks like an SMBus controller (which supports only byte
> +         accesses), except that it only supports one slave device.
> +         This interface is used to connect to specific PMIC devices (like the
> +         AXP221).
> +

Sorry for the stupid question, but why is this an i2c driver if the
hardware protocol is completely different?

I understand that a lot of devices can be driven using either spi or
i2c, and we have two sets of {directories,maintainers,bus_types,...}
for them. Your description sounds like this is a separate option
that isn't any closer to i2c than it is to spi.

Would it perhaps be better to expose it only as a regmap rather than
an i2c host?

	Arnd
Maxime Ripard June 10, 2014, 2:36 p.m. UTC | #2
Hi Arnd,

On Tue, Jun 10, 2014 at 03:54:56PM +0200, Arnd Bergmann wrote:
> On Tuesday 10 June 2014 15:47:16 Boris BREZILLON wrote:
> > 
> > +config I2C_SUN6I_P2WI
> > +       tristate "Allwinner sun6i internal P2WI controller"
> > +       depends on ARCH_SUNXI
> > +       help
> > +         If you say yes to this option, support will be included for the
> > +         P2WI (Push/Pull 2 Wire Interface) controller embedded in some sunxi
> > +         SOCs.
> > +         The P2WI looks like an SMBus controller (which supports only byte
> > +         accesses), except that it only supports one slave device.
> > +         This interface is used to connect to specific PMIC devices (like the
> > +         AXP221).
> > +
> 
> Sorry for the stupid question, but why is this an i2c driver if the
> hardware protocol is completely different?

It's not completely different. It deviates, but still looks very
similar to i2c, and to be precise, SMBus.

You'll have the full discussion that led to do this in i2c here:
http://www.spinics.net/lists/linux-i2c/msg15066.html

Also, one significant thing to take into account is that the
communication with a device starts as I2C, only to switch to this
protocol after some initialization sequence.

> I understand that a lot of devices can be driven using either spi or
> i2c, and we have two sets of {directories,maintainers,bus_types,...}
> for them. Your description sounds like this is a separate option
> that isn't any closer to i2c than it is to spi.

That's not true. It's *much* closer from I2C than it is from SPI.

> Would it perhaps be better to expose it only as a regmap rather than
> an i2c host?

That could be a solution, but is it a common practice to define a bus
adapter driver in a regmap driver?

Maxime
Arnd Bergmann June 10, 2014, 2:47 p.m. UTC | #3
On Tuesday 10 June 2014 16:36:04 Maxime Ripard wrote:
> On Tue, Jun 10, 2014 at 03:54:56PM +0200, Arnd Bergmann wrote:
> > On Tuesday 10 June 2014 15:47:16 Boris BREZILLON wrote:
> > > 
> > > +config I2C_SUN6I_P2WI
> > > +       tristate "Allwinner sun6i internal P2WI controller"
> > > +       depends on ARCH_SUNXI
> > > +       help
> > > +         If you say yes to this option, support will be included for the
> > > +         P2WI (Push/Pull 2 Wire Interface) controller embedded in some sunxi
> > > +         SOCs.
> > > +         The P2WI looks like an SMBus controller (which supports only byte
> > > +         accesses), except that it only supports one slave device.
> > > +         This interface is used to connect to specific PMIC devices (like the
> > > +         AXP221).
> > > +
> > 
> > Sorry for the stupid question, but why is this an i2c driver if the
> > hardware protocol is completely different?
> 
> It's not completely different. It deviates, but still looks very
> similar to i2c, and to be precise, SMBus.
> 
> You'll have the full discussion that led to do this in i2c here:
> http://www.spinics.net/lists/linux-i2c/msg15066.html
> 
> Also, one significant thing to take into account is that the
> communication with a device starts as I2C, only to switch to this
> protocol after some initialization sequence.

Ok, sounds good.

> > I understand that a lot of devices can be driven using either spi or
> > i2c, and we have two sets of {directories,maintainers,bus_types,...}
> > for them. Your description sounds like this is a separate option
> > that isn't any closer to i2c than it is to spi.
> 
> That's not true. It's *much* closer from I2C than it is from SPI.

Ok.

> > Would it perhaps be better to expose it only as a regmap rather than
> > an i2c host?
> 
> That could be a solution, but is it a common practice to define a bus
> adapter driver in a regmap driver?

No, not yet.

Maybe Boris can just put an explanation into the changeset description
of the driver so other people are able to find it more easily.

	Arnd
Wolfram Sang June 10, 2014, 7:45 p.m. UTC | #4
Hi,

compiling this driver gives me:

drivers/i2c/busses/i2c-sun6i-p2wi.c: In function 'p2wi_probe':
drivers/i2c/busses/i2c-sun6i-p2wi.c:272:2: error: implicit declaration of function 'devm_reset_control_get' [-Werror=implicit-function-declaration]
drivers/i2c/busses/i2c-sun6i-p2wi.c:272:13: warning: assignment makes pointer from integer without a cast [enabled by default]

> diff --git a/drivers/i2c/busses/i2c-sun6i-p2wi.c b/drivers/i2c/busses/i2c-sun6i-p2wi.c
> new file mode 100644
> index 0000000..4f7b93f
> --- /dev/null
> +++ b/drivers/i2c/busses/i2c-sun6i-p2wi.c
> @@ -0,0 +1,356 @@
> +/*
> + * P2WI (Push-Pull Two Wire Interface) bus driver.
> + *
> + * Author: Boris BREZILLON <boris.brezillon@free-electrons.com>
> + *
> + * This file is licensed under the terms of the GNU General Public License
> + * version 2.  This program is licensed "as is" without any warranty of any
> + * kind, whether express or implied.
> + *
> + * The P2WI controller looks like an SMBus controller which only supports byte
> + * data transfers. But, it differs from standard SMBus protocol on several
> + * aspects:
> + * - it supports only one slave device, and thus drop the address field
> + * - it adds a parity bit every 8bits of data
> + * - only one read access is required to read a byte (instead of a write
> + *   followed by a read access in standard SMBus protocol)
> + * - there's no Ack bit after each byte transfer
> + *
> + * This means this bus cannot be used to interface with standard SMBus
> + * devices (the only known device to support this interface is the AXP221
> + * PMIC).
> + *
> + */
> +#include <linux/kernel.h>
> +#include <linux/slab.h>
> +#include <linux/module.h>
> +#include <linux/spinlock.h>

Unneeded. Please check other includes, too.

> +#include <linux/i2c.h>
> +#include <linux/interrupt.h>
> +#include <linux/platform_device.h>
> +#include <linux/io.h>
> +#include <linux/of.h>
> +#include <linux/of_device.h>
> +#include <linux/of_irq.h>
> +#include <linux/clk.h>
> +#include <linux/err.h>
> +#include <linux/delay.h>
> +#include <linux/clk.h>
> +#include <linux/reset.h>
> +

...

> +	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +	p2wi->regs = devm_ioremap_resource(dev, r);
> +	if (IS_ERR(p2wi->regs)) {
> +		ret = PTR_ERR(p2wi->regs);
> +		return ret;
> +	}

return PTR_ERR(...)?

> +
> +	snprintf(p2wi->adapter.name, sizeof(p2wi->adapter.name), pdev->name);
> +	irq = platform_get_irq(pdev, 0);
> +	if (irq < 0) {
> +		dev_err(dev, "failed to retrieve irq: %d\n", ret);
> +		return irq;
> +	}
> +
> +	p2wi->clk = devm_clk_get(dev, NULL);
> +	if (IS_ERR(p2wi->clk)) {
> +		ret = PTR_ERR(p2wi->clk);
> +		dev_err(dev, "failed to retrieve clk: %d\n",
> +			ret);

dangling ret.

> +		return ret;
> +	}
> +

...

> +MODULE_AUTHOR("Boris BREZILLON <boris.brezillon@free-electrons.com>");
> +MODULE_DESCRIPTION("Allwinner P2WI driver");
> +MODULE_LICENSE("GPL");

"GPL v2"

Thanks,

   Wolfram
Boris BREZILLON June 11, 2014, 7:42 a.m. UTC | #5
Hello Wolfram,

On 10/06/2014 21:45, Wolfram Sang wrote:
> Hi,
>
> compiling this driver gives me:
>
> drivers/i2c/busses/i2c-sun6i-p2wi.c: In function 'p2wi_probe':
> drivers/i2c/busses/i2c-sun6i-p2wi.c:272:2: error: implicit declaration of function 'devm_reset_control_get' [-Werror=implicit-function-declaration]
> drivers/i2c/busses/i2c-sun6i-p2wi.c:272:13: warning: assignment makes pointer from integer without a cast [enabled by default]

Yes, there's a missing "depends on RESET_CONTROLLER" in the Kconfig
entry, which I didn't notice because this option was selected by the
sun6i platform.

>> diff --git a/drivers/i2c/busses/i2c-sun6i-p2wi.c b/drivers/i2c/busses/i2c-sun6i-p2wi.c
>> new file mode 100644
>> index 0000000..4f7b93f
>> --- /dev/null
>> +++ b/drivers/i2c/busses/i2c-sun6i-p2wi.c
>> @@ -0,0 +1,356 @@
>> +/*
>> + * P2WI (Push-Pull Two Wire Interface) bus driver.
>> + *
>> + * Author: Boris BREZILLON <boris.brezillon@free-electrons.com>
>> + *
>> + * This file is licensed under the terms of the GNU General Public License
>> + * version 2.  This program is licensed "as is" without any warranty of any
>> + * kind, whether express or implied.
>> + *
>> + * The P2WI controller looks like an SMBus controller which only supports byte
>> + * data transfers. But, it differs from standard SMBus protocol on several
>> + * aspects:
>> + * - it supports only one slave device, and thus drop the address field
>> + * - it adds a parity bit every 8bits of data
>> + * - only one read access is required to read a byte (instead of a write
>> + *   followed by a read access in standard SMBus protocol)
>> + * - there's no Ack bit after each byte transfer
>> + *
>> + * This means this bus cannot be used to interface with standard SMBus
>> + * devices (the only known device to support this interface is the AXP221
>> + * PMIC).
>> + *
>> + */
>> +#include <linux/kernel.h>
>> +#include <linux/slab.h>
>> +#include <linux/module.h>
>> +#include <linux/spinlock.h>
> Unneeded. Please check other includes, too.

Okay, I'll clean it up.
>
>> +#include <linux/i2c.h>
>> +#include <linux/interrupt.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/io.h>
>> +#include <linux/of.h>
>> +#include <linux/of_device.h>
>> +#include <linux/of_irq.h>
>> +#include <linux/clk.h>
>> +#include <linux/err.h>
>> +#include <linux/delay.h>
>> +#include <linux/clk.h>
>> +#include <linux/reset.h>
>> +
> ...
>
>> +	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +	p2wi->regs = devm_ioremap_resource(dev, r);
>> +	if (IS_ERR(p2wi->regs)) {
>> +		ret = PTR_ERR(p2wi->regs);
>> +		return ret;
>> +	}
> return PTR_ERR(...)?

Oops, some vestige of my previous dev_err call.

>
>> +
>> +	snprintf(p2wi->adapter.name, sizeof(p2wi->adapter.name), pdev->name);
>> +	irq = platform_get_irq(pdev, 0);
>> +	if (irq < 0) {
>> +		dev_err(dev, "failed to retrieve irq: %d\n", ret);
>> +		return irq;
>> +	}
>> +
>> +	p2wi->clk = devm_clk_get(dev, NULL);
>> +	if (IS_ERR(p2wi->clk)) {
>> +		ret = PTR_ERR(p2wi->clk);
>> +		dev_err(dev, "failed to retrieve clk: %d\n",
>> +			ret);
> dangling ret.

I'll fix it.

Thanks,

Boris
Boris BREZILLON June 11, 2014, 7:52 a.m. UTC | #6
Hello Arnd,

On 10/06/2014 16:47, Arnd Bergmann wrote:
> On Tuesday 10 June 2014 16:36:04 Maxime Ripard wrote:
>> On Tue, Jun 10, 2014 at 03:54:56PM +0200, Arnd Bergmann wrote:
>>> On Tuesday 10 June 2014 15:47:16 Boris BREZILLON wrote:
>>>> +config I2C_SUN6I_P2WI
>>>> +       tristate "Allwinner sun6i internal P2WI controller"
>>>> +       depends on ARCH_SUNXI
>>>> +       help
>>>> +         If you say yes to this option, support will be included for the
>>>> +         P2WI (Push/Pull 2 Wire Interface) controller embedded in some sunxi
>>>> +         SOCs.
>>>> +         The P2WI looks like an SMBus controller (which supports only byte
>>>> +         accesses), except that it only supports one slave device.
>>>> +         This interface is used to connect to specific PMIC devices (like the
>>>> +         AXP221).
>>>> +
>>> Sorry for the stupid question, but why is this an i2c driver if the
>>> hardware protocol is completely different?
>> It's not completely different. It deviates, but still looks very
>> similar to i2c, and to be precise, SMBus.
>>
>> You'll have the full discussion that led to do this in i2c here:
>> http://www.spinics.net/lists/linux-i2c/msg15066.html
>>
>> Also, one significant thing to take into account is that the
>> communication with a device starts as I2C, only to switch to this
>> protocol after some initialization sequence.
> Ok, sounds good.
>
>>> I understand that a lot of devices can be driven using either spi or
>>> i2c, and we have two sets of {directories,maintainers,bus_types,...}
>>> for them. Your description sounds like this is a separate option
>>> that isn't any closer to i2c than it is to spi.
>> That's not true. It's *much* closer from I2C than it is from SPI.
> Ok.
>
>>> Would it perhaps be better to expose it only as a regmap rather than
>>> an i2c host?
>> That could be a solution, but is it a common practice to define a bus
>> adapter driver in a regmap driver?
> No, not yet.
>
> Maybe Boris can just put an explanation into the changeset description
> of the driver so other people are able to find it more easily.

You mean in my commit message ?
I thought it was already explaining the subtle differences between P2WI
and the SMBus protocols.

What would you like me to add to this explanation ?
Something about the I2C to P2WI initialization part ?

Best Regards,

Boris
Arnd Bergmann June 11, 2014, 8:07 a.m. UTC | #7
On Wednesday 11 June 2014 09:52:52 Boris BREZILLON wrote:
> You mean in my commit message ?
> I thought it was already explaining the subtle differences between P2WI
> and the SMBus protocols.
> 
> What would you like me to add to this explanation ?
> Something about the I2C to P2WI initialization part ?
> 

What I'd like to see is a mention of previous discussion that
concluded that it should be an i2c driver rather than a new
type of driver, and why.

	Arnd
diff mbox

Patch

diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
index c94db1c..09bce1c 100644
--- a/drivers/i2c/busses/Kconfig
+++ b/drivers/i2c/busses/Kconfig
@@ -771,6 +771,18 @@  config I2C_STU300
 	  This driver can also be built as a module. If so, the module
 	  will be called i2c-stu300.
 
+config I2C_SUN6I_P2WI
+	tristate "Allwinner sun6i internal P2WI controller"
+	depends on ARCH_SUNXI
+	help
+	  If you say yes to this option, support will be included for the
+	  P2WI (Push/Pull 2 Wire Interface) controller embedded in some sunxi
+	  SOCs.
+	  The P2WI looks like an SMBus controller (which supports only byte
+	  accesses), except that it only supports one slave device.
+	  This interface is used to connect to specific PMIC devices (like the
+	  AXP221).
+
 config I2C_TEGRA
 	tristate "NVIDIA Tegra internal I2C controller"
 	depends on ARCH_TEGRA
diff --git a/drivers/i2c/busses/Makefile b/drivers/i2c/busses/Makefile
index 18d18ff..58feeb9 100644
--- a/drivers/i2c/busses/Makefile
+++ b/drivers/i2c/busses/Makefile
@@ -75,6 +75,7 @@  obj-$(CONFIG_I2C_SIMTEC)	+= i2c-simtec.o
 obj-$(CONFIG_I2C_SIRF)		+= i2c-sirf.o
 obj-$(CONFIG_I2C_ST)		+= i2c-st.o
 obj-$(CONFIG_I2C_STU300)	+= i2c-stu300.o
+obj-$(CONFIG_I2C_SUN6I_P2WI)	+= i2c-sun6i-p2wi.o
 obj-$(CONFIG_I2C_TEGRA)		+= i2c-tegra.o
 obj-$(CONFIG_I2C_VERSATILE)	+= i2c-versatile.o
 obj-$(CONFIG_I2C_WMT)		+= i2c-wmt.o
diff --git a/drivers/i2c/busses/i2c-sun6i-p2wi.c b/drivers/i2c/busses/i2c-sun6i-p2wi.c
new file mode 100644
index 0000000..4f7b93f
--- /dev/null
+++ b/drivers/i2c/busses/i2c-sun6i-p2wi.c
@@ -0,0 +1,356 @@ 
+/*
+ * P2WI (Push-Pull Two Wire Interface) bus driver.
+ *
+ * Author: Boris BREZILLON <boris.brezillon@free-electrons.com>
+ *
+ * This file is licensed under the terms of the GNU General Public License
+ * version 2.  This program is licensed "as is" without any warranty of any
+ * kind, whether express or implied.
+ *
+ * The P2WI controller looks like an SMBus controller which only supports byte
+ * data transfers. But, it differs from standard SMBus protocol on several
+ * aspects:
+ * - it supports only one slave device, and thus drop the address field
+ * - it adds a parity bit every 8bits of data
+ * - only one read access is required to read a byte (instead of a write
+ *   followed by a read access in standard SMBus protocol)
+ * - there's no Ack bit after each byte transfer
+ *
+ * This means this bus cannot be used to interface with standard SMBus
+ * devices (the only known device to support this interface is the AXP221
+ * PMIC).
+ *
+ */
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/spinlock.h>
+#include <linux/i2c.h>
+#include <linux/interrupt.h>
+#include <linux/platform_device.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/of_irq.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/delay.h>
+#include <linux/clk.h>
+#include <linux/reset.h>
+
+
+/* P2WI registers */
+#define P2WI_CTRL		0x0
+#define P2WI_CCR		0x4
+#define P2WI_INTE		0x8
+#define P2WI_INTS		0xc
+#define P2WI_DADDR0		0x10
+#define P2WI_DADDR1		0x14
+#define P2WI_DLEN		0x18
+#define P2WI_DATA0		0x1c
+#define P2WI_DATA1		0x20
+#define P2WI_LCR		0x24
+#define P2WI_PMCR		0x28
+
+/* CTRL fields */
+#define P2WI_CTRL_START_TRANS		BIT(7)
+#define P2WI_CTRL_ABORT_TRANS		BIT(6)
+#define P2WI_CTRL_GLOBAL_INT_ENB	BIT(1)
+#define P2WI_CTRL_SOFT_RST		BIT(0)
+
+/* CLK CTRL fields */
+#define P2WI_CCR_SDA_OUT_DELAY(v)	(((v) & 0x7) << 8)
+#define P2WI_CCR_MAX_CLK_DIV		0xff
+#define P2WI_CCR_CLK_DIV(v)		((v) & P2WI_CCR_MAX_CLK_DIV)
+
+/* STATUS fields */
+#define P2WI_INTS_TRANS_ERR_ID(v)	(((v) >> 8) & 0xff)
+#define P2WI_INTS_LOAD_BSY		BIT(2)
+#define P2WI_INTS_TRANS_ERR		BIT(1)
+#define P2WI_INTS_TRANS_OVER		BIT(0)
+
+/* DATA LENGTH fields*/
+#define P2WI_DLEN_READ			BIT(4)
+#define P2WI_DLEN_DATA_LENGTH(v)	((v - 1) & 0x7)
+
+/* LINE CTRL fields*/
+#define P2WI_LCR_SCL_STATE		BIT(5)
+#define P2WI_LCR_SDA_STATE		BIT(4)
+#define P2WI_LCR_SCL_CTL		BIT(3)
+#define P2WI_LCR_SCL_CTL_EN		BIT(2)
+#define P2WI_LCR_SDA_CTL		BIT(1)
+#define P2WI_LCR_SDA_CTL_EN		BIT(0)
+
+/* PMU MODE CTRL fields */
+#define P2WI_PMCR_PMU_INIT_SEND		BIT(31)
+#define P2WI_PMCR_PMU_INIT_DATA(v)	(((v) & 0xff) << 16)
+#define P2WI_PMCR_PMU_MODE_REG(v)	(((v) & 0xff) << 8)
+#define P2WI_PMCR_PMU_DEV_ADDR(v)	((v) & 0xff)
+
+#define P2WI_MAX_FREQ			6000000
+
+struct p2wi {
+	struct i2c_adapter adapter;
+	struct completion complete;
+	unsigned int status;
+	void __iomem *regs;
+	struct clk *clk;
+	struct reset_control *rstc;
+	int slave_addr;
+};
+
+static irqreturn_t p2wi_interrupt(int irq, void *dev_id)
+{
+	struct p2wi *p2wi = dev_id;
+	unsigned long status;
+
+	status = readl(p2wi->regs + P2WI_INTS);
+	p2wi->status = status;
+
+	/* Clear interrupts */
+	status &= (P2WI_INTS_LOAD_BSY | P2WI_INTS_TRANS_ERR |
+		   P2WI_INTS_TRANS_OVER);
+	writel(status, p2wi->regs + P2WI_INTS);
+
+	complete(&p2wi->complete);
+
+	return IRQ_HANDLED;
+}
+
+static u32 p2wi_functionality(struct i2c_adapter *adap)
+{
+	return I2C_FUNC_SMBUS_BYTE_DATA;
+}
+
+static int p2wi_smbus_xfer(struct i2c_adapter *adap, u16 addr,
+			   unsigned short flags, char read_write,
+			   u8 command, int size, union i2c_smbus_data *data)
+{
+	struct p2wi *p2wi = i2c_get_adapdata(adap);
+	unsigned long dlen = P2WI_DLEN_DATA_LENGTH(1);
+
+	if (p2wi->slave_addr >= 0 && addr != p2wi->slave_addr) {
+		dev_err(&adap->dev, "invalid P2WI address\n");
+		return -EINVAL;
+	}
+
+	if (!data)
+		return -EINVAL;
+
+	writel(command, p2wi->regs + P2WI_DADDR0);
+
+	if (read_write == I2C_SMBUS_READ)
+		dlen |= P2WI_DLEN_READ;
+	else
+		writel(data->byte, p2wi->regs + P2WI_DATA0);
+
+	writel(dlen, p2wi->regs + P2WI_DLEN);
+
+	if (readl(p2wi->regs + P2WI_CTRL) & P2WI_CTRL_START_TRANS) {
+		dev_err(&adap->dev, "P2WI bus busy\n");
+		return -EBUSY;
+	}
+
+	reinit_completion(&p2wi->complete);
+
+	writel(P2WI_INTS_LOAD_BSY | P2WI_INTS_TRANS_ERR | P2WI_INTS_TRANS_OVER,
+	       p2wi->regs + P2WI_INTE);
+
+	writel(P2WI_CTRL_START_TRANS | P2WI_CTRL_GLOBAL_INT_ENB,
+	       p2wi->regs + P2WI_CTRL);
+
+	wait_for_completion(&p2wi->complete);
+
+	if (p2wi->status & P2WI_INTS_LOAD_BSY) {
+		dev_err(&adap->dev, "P2WI bus busy\n");
+		return -EBUSY;
+	}
+
+	if (p2wi->status & P2WI_INTS_TRANS_ERR) {
+		dev_err(&adap->dev, "P2WI bus xfer error\n");
+		return -ENXIO;
+	}
+
+	if (read_write == I2C_SMBUS_READ)
+		data->byte = readl(p2wi->regs + P2WI_DATA0);
+
+	return 0;
+}
+
+static const struct i2c_algorithm p2wi_algo = {
+	.smbus_xfer = p2wi_smbus_xfer,
+	.functionality = p2wi_functionality,
+};
+
+static const struct of_device_id p2wi_of_match_table[] = {
+	{ .compatible = "allwinner,sun6i-a31-p2wi" },
+	{}
+};
+MODULE_DEVICE_TABLE(of, p2wi_of_match_table);
+
+static int p2wi_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *np = dev->of_node;
+	struct device_node *childnp;
+	unsigned long parent_clk_freq;
+	u32 clk_freq = 100000;
+	struct resource *r;
+	struct p2wi *p2wi;
+	u32 slave_addr;
+	int clk_div;
+	int irq;
+	int ret;
+
+	of_property_read_u32(np, "clock-frequency", &clk_freq);
+	if (clk_freq > P2WI_MAX_FREQ) {
+		dev_err(dev,
+			"required clock-frequency (%u Hz) is too high (max = 6MHz)",
+			clk_freq);
+		return -EINVAL;
+	}
+
+	if (of_get_child_count(np) > 1) {
+		dev_err(dev, "P2WI only supports one slave device\n");
+		return -EINVAL;
+	}
+
+	p2wi = devm_kzalloc(dev, sizeof(struct p2wi), GFP_KERNEL);
+	if (!p2wi)
+		return -ENOMEM;
+
+	p2wi->slave_addr = -1;
+
+	/*
+	 * Authorize a p2wi node without any children to be able to use an
+	 * i2c-dev from userpace.
+	 * In this case the slave_addr is set to -1 and won't be checked when
+	 * launching a P2WI transfer.
+	 */
+	childnp = of_get_next_available_child(np, NULL);
+	if (childnp) {
+		ret = of_property_read_u32(childnp, "reg", &slave_addr);
+		if (ret) {
+			dev_err(dev, "invalid slave address on node %s\n",
+				childnp->full_name);
+			return -EINVAL;
+		}
+
+		p2wi->slave_addr = slave_addr;
+	}
+
+	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	p2wi->regs = devm_ioremap_resource(dev, r);
+	if (IS_ERR(p2wi->regs)) {
+		ret = PTR_ERR(p2wi->regs);
+		return ret;
+	}
+
+	snprintf(p2wi->adapter.name, sizeof(p2wi->adapter.name), pdev->name);
+	irq = platform_get_irq(pdev, 0);
+	if (irq < 0) {
+		dev_err(dev, "failed to retrieve irq: %d\n", ret);
+		return irq;
+	}
+
+	p2wi->clk = devm_clk_get(dev, NULL);
+	if (IS_ERR(p2wi->clk)) {
+		ret = PTR_ERR(p2wi->clk);
+		dev_err(dev, "failed to retrieve clk: %d\n",
+			ret);
+		return ret;
+	}
+
+	ret = clk_prepare_enable(p2wi->clk);
+	if (ret) {
+		dev_err(dev, "failed to enable clk: %d\n", ret);
+		return ret;
+	}
+
+	parent_clk_freq = clk_get_rate(p2wi->clk);
+
+	p2wi->rstc = devm_reset_control_get(dev, NULL);
+	if (IS_ERR(p2wi->rstc)) {
+		ret = PTR_ERR(p2wi->rstc);
+		dev_err(dev, "failed to retrieve reset controller: %d\n", ret);
+		goto err_clk_disable;
+	}
+
+	ret = reset_control_deassert(p2wi->rstc);
+	if (ret) {
+		dev_err(dev, "failed to deassert reset line: %d\n",
+			ret);
+		goto err_clk_disable;
+	}
+
+	init_completion(&p2wi->complete);
+	p2wi->adapter.dev.parent = dev;
+	p2wi->adapter.algo = &p2wi_algo;
+	p2wi->adapter.owner = THIS_MODULE;
+	p2wi->adapter.dev.of_node = pdev->dev.of_node;
+	platform_set_drvdata(pdev, p2wi);
+	i2c_set_adapdata(&p2wi->adapter, p2wi);
+
+	ret = devm_request_irq(dev, irq, p2wi_interrupt, 0, pdev->name, p2wi);
+	if (ret) {
+		dev_err(dev, "can't register interrupt handler irq%d: %d\n",
+			irq, ret);
+		goto err_reset_assert;
+	}
+
+	writel(P2WI_CTRL_SOFT_RST, p2wi->regs + P2WI_CTRL);
+
+	clk_div = parent_clk_freq / clk_freq;
+	if (!clk_div) {
+		dev_warn(dev,
+			 "clock-frequency is too high, setting it to %lu Hz\n",
+			 parent_clk_freq);
+		clk_div = 1;
+	} else if (clk_div > P2WI_CCR_MAX_CLK_DIV) {
+		dev_warn(dev,
+			 "clock-frequency is too low, setting it to %lu Hz\n",
+			 parent_clk_freq / P2WI_CCR_MAX_CLK_DIV);
+		clk_div = P2WI_CCR_MAX_CLK_DIV;
+	}
+
+	writel(P2WI_CCR_SDA_OUT_DELAY(1) | P2WI_CCR_CLK_DIV(clk_div),
+	       p2wi->regs + P2WI_CCR);
+
+	ret = i2c_add_adapter(&p2wi->adapter);
+	if (!ret)
+		return 0;
+
+err_reset_assert:
+	reset_control_assert(p2wi->rstc);
+
+err_clk_disable:
+	clk_disable_unprepare(p2wi->clk);
+
+	return ret;
+}
+
+static int p2wi_remove(struct platform_device *dev)
+{
+	struct p2wi *p2wi = platform_get_drvdata(dev);
+
+	reset_control_assert(p2wi->rstc);
+	clk_disable_unprepare(p2wi->clk);
+	i2c_del_adapter(&p2wi->adapter);
+
+	return 0;
+}
+
+static struct platform_driver p2wi_driver = {
+	.probe	= p2wi_probe,
+	.remove	= p2wi_remove,
+	.driver	= {
+		.owner = THIS_MODULE,
+		.name = "i2c-sunxi-p2wi",
+		.of_match_table = p2wi_of_match_table,
+	},
+};
+module_platform_driver(p2wi_driver);
+
+MODULE_AUTHOR("Boris BREZILLON <boris.brezillon@free-electrons.com>");
+MODULE_DESCRIPTION("Allwinner P2WI driver");
+MODULE_LICENSE("GPL");