diff mbox series

[v3,2/3] i2c: add support for HiSilicon I2C controller

Message ID 1616411413-7177-3-git-send-email-yangyicong@hisilicon.com
State New
Headers show
Series Add support for HiSilicon I2C controller | expand

Commit Message

Yicong Yang March 22, 2021, 11:10 a.m. UTC
Add HiSilicon I2C controller driver for the Kunpeng SoC. It provides
the access to the i2c busses, which connects to the eeprom, rtc, etc.

The driver works with IRQ mode, and supports basic I2C features and 10bit
address. The DMA is not supported.

Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 drivers/i2c/busses/Kconfig    |  10 +
 drivers/i2c/busses/Makefile   |   1 +
 drivers/i2c/busses/i2c-hisi.c | 525 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 536 insertions(+)
 create mode 100644 drivers/i2c/busses/i2c-hisi.c

Comments

Yicong Yang March 24, 2021, 9:30 a.m. UTC | #1
Hi Dmitry,

Thanks for reviewing this.

On 2021/3/22 23:21, Dmitry Osipenko wrote:
> Hello Yicong,

> 

> 22.03.2021 14:10, Yicong Yang пишет:

>> Add HiSilicon I2C controller driver for the Kunpeng SoC. It provides

>> the access to the i2c busses, which connects to the eeprom, rtc, etc.

>>

>> The driver works with IRQ mode, and supports basic I2C features and 10bit

>> address. The DMA is not supported.

>>

>> Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>

>> ---

>>  drivers/i2c/busses/Kconfig    |  10 +

>>  drivers/i2c/busses/Makefile   |   1 +

>>  drivers/i2c/busses/i2c-hisi.c | 525 ++++++++++++++++++++++++++++++++++++++++++

>>  3 files changed, 536 insertions(+)

>>  create mode 100644 drivers/i2c/busses/i2c-hisi.c

> 

> ...

>> +

>> +#define NSEC_TO_CYCLES(ns, clk_rate_khz) (DIV_ROUND_UP_ULL((clk_rate_khz) * (ns), NSEC_PER_MSEC))

> 

> This is a very long line, you should split it into two.


will split.

> 

> Parens around DIV_ROUND_UP_ULL aren't needed.

> 


will drop the parens.

> ...

>> +static void hisi_i2c_enable_int(struct hisi_i2c_controller *ctlr, u32 mask)

>> +{

>> +	writel(mask, ctlr->iobase + HISI_I2C_INT_MASK);

> 

> Why you don't use relaxed versions of readl/writel? Do you really need

> to insert memory barriers?

> 


this will not be used during the transfer process, so a relaxed version of readl/writel
will not have performance enhancement.

the barriers are necessary as i want to make the operations in order to avoid potential
problems caused by reordering.

>> +}

>> +

>> +static void hisi_i2c_disable_int(struct hisi_i2c_controller *ctlr, u32 mask)

>> +{

>> +	writel((~mask) & HISI_I2C_INT_ALL, ctlr->iobase + HISI_I2C_INT_MASK);

>> +}

>> +

>> +static void hisi_i2c_clear_int(struct hisi_i2c_controller *ctlr, u32 mask)

>> +{

>> +	writel(mask, ctlr->iobase + HISI_I2C_INT_CLR);

>> +}

>> +

>> +static void hisi_i2c_handle_errors(struct hisi_i2c_controller *ctlr)

>> +{

>> +	u32 int_err = ctlr->xfer_err, reg;

>> +

>> +	if (int_err & HISI_I2C_INT_FIFO_ERR) {

>> +		reg = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);

>> +

>> +		if (reg & HISI_I2C_FIFO_STATE_RX_RERR)

>> +			dev_err(ctlr->dev, "rx fifo error read.\n");

> 

> The dot "." in the end of error messages is unnecessary.

> 


i'd like to keep this, as i think this is rather driver specific and not
violating any rules.

> ...

>> +/*

>> + * Initialize the transfer information and start the I2C bus transfer.

>> + * We only configure the transfer and do some pre/post works here, and

>> + * wait for the transfer done. The major transfer process is performed

>> + * in the IRQ handler.

>> + */

>> +static int hisi_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,

>> +				int num)

>> +{

>> +	struct hisi_i2c_controller *ctlr = i2c_get_adapdata(adap);

>> +	DECLARE_COMPLETION_ONSTACK(done);

>> +	int ret = num;

>> +

>> +	hisi_i2c_reset_xfer(ctlr);

>> +	ctlr->completion = &done;

>> +	ctlr->msg_num = num;

>> +	ctlr->msgs = msgs;

>> +

>> +	hisi_i2c_start_xfer(ctlr);

>> +

>> +	if (!wait_for_completion_timeout(ctlr->completion, adap->timeout)) {

>> +		hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);

> 

> This doesn't save you from racing with the interrupt handler. It looks

> like you need to enable/disable IRQ around the completion, similarly to

> what NVIDIA Tegra I2C driver does.

> 


thanks for suggestion.

the hardware between tegra and this one is a little different as we don't provide
a way to reinit the controller. so {synchronize,disable}_irq() after mask
the interrupt here will avoid the race.

Thanks,
Yicong

> .

>
Yicong Yang March 24, 2021, 10:07 a.m. UTC | #2
Hi Andy,

Thanks for reviewing this.

On 2021/3/23 0:59, Andy Shevchenko wrote:
> On Mon, Mar 22, 2021 at 07:10:12PM +0800, Yicong Yang wrote:

>> Add HiSilicon I2C controller driver for the Kunpeng SoC. It provides

>> the access to the i2c busses, which connects to the eeprom, rtc, etc.

>>

>> The driver works with IRQ mode, and supports basic I2C features and 10bit

>> address. The DMA is not supported.

> 

> ...

> 

>> +#include <linux/acpi.h>

> 

> Hadn't noticed how you are using this header.


will remove this.

> 

>> +#include <linux/bits.h>

>> +#include <linux/bitfield.h>

>> +#include <linux/completion.h>

>> +#include <linux/i2c.h>

>> +#include <linux/interrupt.h>

>> +#include <linux/io.h>

>> +#include <linux/module.h>

> 

> Missed mod_devicetable.h.

> Probably missed property.h, but not sure.>


i think this has been included implicitly as the module compilation worked well.


>> +#include <linux/platform_device.h>

> 

> ...

> 

>> +#define HISI_I2C_INT_ALL	0x1f

> 

> GENMASK() ?


will change to that.

> 

> ...

> 

>> +#define HISI_I2C_INT_ERR	(HISI_I2C_INT_TRANS_ERR | \

>> +				 HISI_I2C_INT_FIFO_ERR)

> 

> Either one line, or it will look better like

> #define HISI_I2C_INT_ERR					\

> 	(HISI_I2C_INT_TRANS_ERR | HISI_I2C_INT_FIFO_ERR)

> 


sure. that looks better.

> ...

> 

>> +#define HISI_I2C_STD_SPEED_MODE		0x0

>> +#define HISI_I2C_FAST_SPEED_MODE	0x1

>> +#define HISI_I2C_HIGH_SPEED_MODE	0x2

> 

> Why not plain decimal numbers?

> 


it's something will be written to the register, so i make them
hexadecimal to better corresponding to the register fields.

> ...

> 

>> +struct hisi_i2c_controller {

>> +	struct device *dev;

> 

>> +	struct i2c_adapter adapter;

> 

> If you put this as a first member, the container_of() become a no-op for this.

> But I dunno if it's used against this structure.

> 


the container_of() is not used in this drivers. but it's fine to swap these
two fields. thanks for the information.

>> +	void __iomem *iobase;

>> +	int irq;

>> +

>> +	/* Intermediates for recording the transfer process */

>> +	struct completion *completion;

>> +	struct i2c_msg *msgs;

>> +	int msg_num;

>> +	int msg_tx_idx;

>> +	int buf_tx_idx;

>> +	int msg_rx_idx;

>> +	int buf_rx_idx;

>> +	u16 tar_addr;

>> +	u32 xfer_err;

>> +

>> +	/* I2C bus configuration */

>> +	u32 scl_fall_time;

>> +	u32 scl_rise_time;

>> +	u32 sda_hold_time;

>> +	u64 clk_rate_khz;

>> +	u32 bus_freq_hz;

>> +	u32 spk_len;

>> +};

> 

> ...

> 

>> +	struct i2c_timings t;

>> +

>> +	i2c_parse_fw_timings(ctlr->dev, &t, true);

>> +	ctlr->bus_freq_hz = t.bus_freq_hz;

> 

>> +	ctlr->scl_fall_time = t.scl_fall_ns;

>> +	ctlr->scl_rise_time = t.scl_rise_ns;

>> +	ctlr->sda_hold_time = t.sda_hold_ns;

> 

> Why not simply to have the timings structure embedded into hisi_i2c_controller

> one?

> 


not all the fields of the timing structures are needed to be stored, only three
of them are necessary.

> ...

> 

>> +	ctlr->dev = dev;

> 

> Would it make sense to assign aster getting IRQ resource...

> 

>> +	ctlr->iobase = devm_platform_ioremap_resource(pdev, 0);

>> +	if (IS_ERR(ctlr->iobase))

>> +		return PTR_ERR(ctlr->iobase);

>> +

>> +	ctlr->irq = platform_get_irq(pdev, 0);

>> +	if (ctlr->irq < 0)

>> +		return ctlr->irq;

> 

> ...somewhere here?


i cannot see any differences and some other drivers also assign the 'dev' before getting IRQ
resources.

are there any considerations on this?

> 

>> +	hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);

> 

> ...

> 

>> +	ctlr->clk_rate_khz = DIV_ROUND_UP_ULL(ctlr->clk_rate_khz, 1000);

> 

> HZ_PER_KHZ ?


the macro is not public. seems it's redundant to have this macro which
will only be used once.

Thanks,
Yicong

>
Yicong Yang March 24, 2021, 10:21 a.m. UTC | #3
On 2021/3/23 1:04, Andy Shevchenko wrote:
> On Mon, Mar 22, 2021 at 07:10:12PM +0800, Yicong Yang wrote:

>> Add HiSilicon I2C controller driver for the Kunpeng SoC. It provides

>> the access to the i2c busses, which connects to the eeprom, rtc, etc.

>>

>> The driver works with IRQ mode, and supports basic I2C features and 10bit

>> address. The DMA is not supported.

> 

> ...

> 

>> +static const char *hisi_i2c_speed_string(u32 bus_freq_hz)

>> +{

>> +	switch (bus_freq_hz) {

>> +	case I2C_MAX_STANDARD_MODE_FREQ:

>> +		return "100K";

>> +	case I2C_MAX_FAST_MODE_FREQ:

>> +		return "400K";

>> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:

>> +		return "3.4M";

>> +	default:

>> +		return "unknown";

>> +	}

>> +}

> 

> Just realized that if you print the name of the mode (and maybe frequency

> value) then it can be moved to generic I²C code and other will benefit out of

> this (DesignWare is the first in my mind).


sure, that's good. but the i2c core doesn't make use of the speed mode
information so maybe print of this information is rather driver depended.

>
Andy Shevchenko March 24, 2021, 11:15 a.m. UTC | #4
On Wed, Mar 24, 2021 at 06:07:17PM +0800, Yicong Yang wrote:
> On 2021/3/23 0:59, Andy Shevchenko wrote:

> > On Mon, Mar 22, 2021 at 07:10:12PM +0800, Yicong Yang wrote:


...

> > Missed mod_devicetable.h.

> > Probably missed property.h, but not sure.>

> 

> i think this has been included implicitly as the module compilation worked well.


The rule of thumb is to include headers you are direct user of. The implicit
inclusion is allowed if and only if there is a guarantee by explicit inclusion
that an implicit one will be included no matter what. I don't remember we have
such guarantee for mod_devicetable.h.

On top of that, it's performance wise to explicitly include, otherwise it's an
additional burden for compiler and thus on electrical plant and hence not
environment friendly. And all this for peanuts.

...

> >> +#define HISI_I2C_STD_SPEED_MODE		0x0

> >> +#define HISI_I2C_FAST_SPEED_MODE	0x1

> >> +#define HISI_I2C_HIGH_SPEED_MODE	0x2

> > 

> > Why not plain decimal numbers?

> 

> it's something will be written to the register, so i make them

> hexadecimal to better corresponding to the register fields.


Are they bits? No. Why hex numbers? Please, give a better justification.

...

> >> +	ctlr->bus_freq_hz = t.bus_freq_hz;

> > 

> >> +	ctlr->scl_fall_time = t.scl_fall_ns;

> >> +	ctlr->scl_rise_time = t.scl_rise_ns;

> >> +	ctlr->sda_hold_time = t.sda_hold_ns;

> > 

> > Why not simply to have the timings structure embedded into hisi_i2c_controller

> > one?

> > 

> 

> not all the fields of the timing structures are needed to be stored, only three

> of them are necessary.


Four as far as I see. But for the sake of standardization I would keep a whole
structure. It will be easier to grep and find users of it (looking into data
structures only).

...

> >> +	ctlr->dev = dev;

> > 

> > Would it make sense to assign aster getting IRQ resource...

> > 

> >> +	ctlr->iobase = devm_platform_ioremap_resource(pdev, 0);

> >> +	if (IS_ERR(ctlr->iobase))

> >> +		return PTR_ERR(ctlr->iobase);

> >> +

> >> +	ctlr->irq = platform_get_irq(pdev, 0);

> >> +	if (ctlr->irq < 0)

> >> +		return ctlr->irq;

> > 

> > ...somewhere here?

> 

> i cannot see any differences and some other drivers also assign the 'dev' before getting IRQ

> resources.

> 

> are there any considerations on this?


Of course. The rule of thumb: be lazy. Why do you assign earlier if
a) there are no users;
b) in between it may bail out with error
?

Give a better justification why it should be there.

...

> >> +	ctlr->clk_rate_khz = DIV_ROUND_UP_ULL(ctlr->clk_rate_khz, 1000);

> > 

> > HZ_PER_KHZ ?

> 

> the macro is not public. seems it's redundant to have this macro which

> will only be used once.


It will be easier to see how many users of it we have. It's not needed to make
it public right now, but keeping something like

#define HZ_PER_KHZ  1000L

in this file will be helpful.

Give a better justification why it should not be there.

-- 
With Best Regards,
Andy Shevchenko
Andy Shevchenko March 24, 2021, 11:16 a.m. UTC | #5
On Wed, Mar 24, 2021 at 06:21:36PM +0800, Yicong Yang wrote:
> On 2021/3/23 1:04, Andy Shevchenko wrote:

> > On Mon, Mar 22, 2021 at 07:10:12PM +0800, Yicong Yang wrote:


...

> >> +static const char *hisi_i2c_speed_string(u32 bus_freq_hz)

> >> +{

> >> +	switch (bus_freq_hz) {

> >> +	case I2C_MAX_STANDARD_MODE_FREQ:

> >> +		return "100K";

> >> +	case I2C_MAX_FAST_MODE_FREQ:

> >> +		return "400K";

> >> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:

> >> +		return "3.4M";

> >> +	default:

> >> +		return "unknown";

> >> +	}

> >> +}

> > 

> > Just realized that if you print the name of the mode (and maybe frequency

> > value) then it can be moved to generic I²C code and other will benefit out of

> > this (DesignWare is the first in my mind).

> 

> sure, that's good. but the i2c core doesn't make use of the speed mode

> information so maybe print of this information is rather driver depended.


Yes, but it's useful. And since we will have at least two users of this it is a
good justification to use I²C core to keep and provide this API.

-- 
With Best Regards,
Andy Shevchenko
Dmitry Osipenko March 24, 2021, 12:26 p.m. UTC | #6
24.03.2021 12:30, Yicong Yang пишет:
...
>>> +static void hisi_i2c_enable_int(struct hisi_i2c_controller *ctlr, u32 mask)

>>> +{

>>> +	writel(mask, ctlr->iobase + HISI_I2C_INT_MASK);

>>

>> Why you don't use relaxed versions of readl/writel? Do you really need

>> to insert memory barriers?

>>

> 

> this will not be used during the transfer process, so a relaxed version of readl/writel

> will not have performance enhancement.

> 

> the barriers are necessary as i want to make the operations in order to avoid potential

> problems caused by reordering.


The iomap is strongly ordered, hence register accesses are always
ordered. The barrier ensures that CPU memory accesses are finished
before h/w registers are touched. Looks like you don't need to worry
about the memory barrier in the case of this driver.

>>> +}

>>> +

>>> +static void hisi_i2c_disable_int(struct hisi_i2c_controller *ctlr, u32 mask)

>>> +{

>>> +	writel((~mask) & HISI_I2C_INT_ALL, ctlr->iobase + HISI_I2C_INT_MASK);

>>> +}

>>> +

>>> +static void hisi_i2c_clear_int(struct hisi_i2c_controller *ctlr, u32 mask)

>>> +{

>>> +	writel(mask, ctlr->iobase + HISI_I2C_INT_CLR);

>>> +}

>>> +

>>> +static void hisi_i2c_handle_errors(struct hisi_i2c_controller *ctlr)

>>> +{

>>> +	u32 int_err = ctlr->xfer_err, reg;

>>> +

>>> +	if (int_err & HISI_I2C_INT_FIFO_ERR) {

>>> +		reg = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);

>>> +

>>> +		if (reg & HISI_I2C_FIFO_STATE_RX_RERR)

>>> +			dev_err(ctlr->dev, "rx fifo error read.\n");

>>

>> The dot "." in the end of error messages is unnecessary.

>>

> 

> i'd like to keep this, as i think this is rather driver specific and not

> violating any rules.


The common kernel style is *not* to have the dot + some other messages
in this driver already don't have it. Should be better if you could
remove it.

>>> +/*

>>> + * Initialize the transfer information and start the I2C bus transfer.

>>> + * We only configure the transfer and do some pre/post works here, and

>>> + * wait for the transfer done. The major transfer process is performed

>>> + * in the IRQ handler.

>>> + */

>>> +static int hisi_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,

>>> +				int num)

>>> +{

>>> +	struct hisi_i2c_controller *ctlr = i2c_get_adapdata(adap);

>>> +	DECLARE_COMPLETION_ONSTACK(done);

>>> +	int ret = num;

>>> +

>>> +	hisi_i2c_reset_xfer(ctlr);

>>> +	ctlr->completion = &done;

>>> +	ctlr->msg_num = num;

>>> +	ctlr->msgs = msgs;

>>> +

>>> +	hisi_i2c_start_xfer(ctlr);

>>> +

>>> +	if (!wait_for_completion_timeout(ctlr->completion, adap->timeout)) {

>>> +		hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);

>>

>> This doesn't save you from racing with the interrupt handler. It looks

>> like you need to enable/disable IRQ around the completion, similarly to

>> what NVIDIA Tegra I2C driver does.

>>

> 

> thanks for suggestion.

> 

> the hardware between tegra and this one is a little different as we don't provide

> a way to reinit the controller. so {synchronize,disable}_irq() after mask

> the interrupt here will avoid the race.


The disable/enable will be ideal, but synchronize should be good enough
as well.
diff mbox series

Patch

diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
index 05ebf75..f6b8b71 100644
--- a/drivers/i2c/busses/Kconfig
+++ b/drivers/i2c/busses/Kconfig
@@ -645,6 +645,16 @@  config I2C_HIGHLANDER
 	  This driver can also be built as a module.  If so, the module
 	  will be called i2c-highlander.
 
+config I2C_HISI
+	tristate "HiSilicon I2C controller"
+	depends on (ARM64 && ACPI) || COMPILE_TEST
+	help
+	  Say Y here if you want to have Hisilicon I2C controller support
+	  available on the Kunpeng Server.
+
+	  This driver can also be built as a module. If so, the module
+	  will be called i2c-hisi.
+
 config I2C_IBM_IIC
 	tristate "IBM PPC 4xx on-chip I2C interface"
 	depends on 4xx
diff --git a/drivers/i2c/busses/Makefile b/drivers/i2c/busses/Makefile
index 615f35e..e1c9292 100644
--- a/drivers/i2c/busses/Makefile
+++ b/drivers/i2c/busses/Makefile
@@ -63,6 +63,7 @@  obj-$(CONFIG_I2C_EMEV2)		+= i2c-emev2.o
 obj-$(CONFIG_I2C_EXYNOS5)	+= i2c-exynos5.o
 obj-$(CONFIG_I2C_GPIO)		+= i2c-gpio.o
 obj-$(CONFIG_I2C_HIGHLANDER)	+= i2c-highlander.o
+obj-$(CONFIG_I2C_HISI)		+= i2c-hisi.o
 obj-$(CONFIG_I2C_HIX5HD2)	+= i2c-hix5hd2.o
 obj-$(CONFIG_I2C_IBM_IIC)	+= i2c-ibm_iic.o
 obj-$(CONFIG_I2C_IMG)		+= i2c-img-scb.o
diff --git a/drivers/i2c/busses/i2c-hisi.c b/drivers/i2c/busses/i2c-hisi.c
new file mode 100644
index 0000000..8656f71
--- /dev/null
+++ b/drivers/i2c/busses/i2c-hisi.c
@@ -0,0 +1,525 @@ 
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * HiSilicon I2C Controller Driver for Kunpeng SoC
+ *
+ * Copyright (c) 2021 HiSilicon Limited.
+ */
+
+#include <linux/acpi.h>
+#include <linux/bits.h>
+#include <linux/bitfield.h>
+#include <linux/completion.h>
+#include <linux/i2c.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#define HISI_I2C_FRAME_CTRL		0x0000
+#define   HISI_I2C_FRAME_CTRL_SPEED_MODE	GENMASK(1, 0)
+#define   HISI_I2C_FRAME_CTRL_ADDR_TEN	BIT(2)
+#define HISI_I2C_SLV_ADDR		0x0004
+#define   HISI_I2C_SLV_ADDR_VAL		GENMASK(9, 0)
+#define   HISI_I2C_SLV_ADDR_GC_S_MODE	BIT(10)
+#define   HISI_I2C_SLV_ADDR_GC_S_EN	BIT(11)
+#define HISI_I2C_CMD_TXDATA		0x0008
+#define   HISI_I2C_CMD_TXDATA_DATA	GENMASK(7, 0)
+#define   HISI_I2C_CMD_TXDATA_RW	BIT(8)
+#define   HISI_I2C_CMD_TXDATA_P_EN	BIT(9)
+#define   HISI_I2C_CMD_TXDATA_SR_EN	BIT(10)
+#define HISI_I2C_RXDATA			0x000c
+#define   HISI_I2C_RXDATA_DATA		GENMASK(7, 0)
+#define HISI_I2C_SS_SCL_HCNT		0x0010
+#define HISI_I2C_SS_SCL_LCNT		0x0014
+#define HISI_I2C_FS_SCL_HCNT		0x0018
+#define HISI_I2C_FS_SCL_LCNT		0x001c
+#define HISI_I2C_HS_SCL_HCNT		0x0020
+#define HISI_I2C_HS_SCL_LCNT		0x0024
+#define HISI_I2C_FIFO_CTRL		0x0028
+#define   HISI_I2C_FIFO_RX_CLR		BIT(0)
+#define   HISI_I2C_FIFO_TX_CLR		BIT(1)
+#define   HISI_I2C_FIFO_RX_AF_THRESH	GENMASK(7, 2)
+#define   HISI_I2C_FIFO_TX_AE_THRESH	GENMASK(13, 8)
+#define HISI_I2C_FIFO_STATE		0x002c
+#define   HISI_I2C_FIFO_STATE_RX_RERR	BIT(0)
+#define   HISI_I2C_FIFO_STATE_RX_WERR	BIT(1)
+#define   HISI_I2C_FIFO_STATE_RX_EMPTY	BIT(3)
+#define   HISI_I2C_FIFO_STATE_TX_RERR	BIT(6)
+#define   HISI_I2C_FIFO_STATE_TX_WERR	BIT(7)
+#define   HISI_I2C_FIFO_STATE_TX_FULL	BIT(11)
+#define HISI_I2C_SDA_HOLD		0x0030
+#define   HISI_I2C_SDA_HOLD_TX		GENMASK(15, 0)
+#define   HISI_I2C_SDA_HOLD_RX		GENMASK(23, 16)
+#define HISI_I2C_FS_SPK_LEN		0x0038
+#define   HISI_I2C_FS_SPK_LEN_CNT	GENMASK(7, 0)
+#define HISI_I2C_HS_SPK_LEN		0x003c
+#define   HISI_I2C_HS_SPK_LEN_CNT	GENMASK(7, 0)
+#define HISI_I2C_INT_MSTAT		0x0044
+#define HISI_I2C_INT_CLR		0x0048
+#define HISI_I2C_INT_MASK		0x004C
+#define HISI_I2C_TRANS_STATE		0x0050
+#define HISI_I2C_TRANS_ERR		0x0054
+#define HISI_I2C_VERSION		0x0058
+
+#define HISI_I2C_INT_ALL	0x1f
+#define HISI_I2C_INT_TRANS_CPLT	BIT(0)
+#define HISI_I2C_INT_TRANS_ERR	BIT(1)
+#define HISI_I2C_INT_FIFO_ERR	BIT(2)
+#define HISI_I2C_INT_RX_FULL	BIT(3)
+#define HISI_I2C_INT_TX_EMPTY	BIT(4)
+#define HISI_I2C_INT_ERR	(HISI_I2C_INT_TRANS_ERR | \
+				 HISI_I2C_INT_FIFO_ERR)
+
+#define HISI_I2C_STD_SPEED_MODE		0x0
+#define HISI_I2C_FAST_SPEED_MODE	0x1
+#define HISI_I2C_HIGH_SPEED_MODE	0x2
+
+#define HISI_I2C_TX_FIFO_DEPTH		64
+#define HISI_I2C_RX_FIFO_DEPTH		64
+#define HISI_I2C_TX_F_AE_THRESH		1
+#define HISI_I2C_RX_F_AF_THRESH		60
+
+#define NSEC_TO_CYCLES(ns, clk_rate_khz) (DIV_ROUND_UP_ULL((clk_rate_khz) * (ns), NSEC_PER_MSEC))
+
+struct hisi_i2c_controller {
+	struct device *dev;
+	struct i2c_adapter adapter;
+	void __iomem *iobase;
+	int irq;
+
+	/* Intermediates for recording the transfer process */
+	struct completion *completion;
+	struct i2c_msg *msgs;
+	int msg_num;
+	int msg_tx_idx;
+	int buf_tx_idx;
+	int msg_rx_idx;
+	int buf_rx_idx;
+	u16 tar_addr;
+	u32 xfer_err;
+
+	/* I2C bus configuration */
+	u32 scl_fall_time;
+	u32 scl_rise_time;
+	u32 sda_hold_time;
+	u64 clk_rate_khz;
+	u32 bus_freq_hz;
+	u32 spk_len;
+};
+
+static const char *hisi_i2c_speed_string(u32 bus_freq_hz)
+{
+	switch (bus_freq_hz) {
+	case I2C_MAX_STANDARD_MODE_FREQ:
+		return "100K";
+	case I2C_MAX_FAST_MODE_FREQ:
+		return "400K";
+	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
+		return "3.4M";
+	default:
+		return "unknown";
+	}
+}
+
+static void hisi_i2c_enable_int(struct hisi_i2c_controller *ctlr, u32 mask)
+{
+	writel(mask, ctlr->iobase + HISI_I2C_INT_MASK);
+}
+
+static void hisi_i2c_disable_int(struct hisi_i2c_controller *ctlr, u32 mask)
+{
+	writel((~mask) & HISI_I2C_INT_ALL, ctlr->iobase + HISI_I2C_INT_MASK);
+}
+
+static void hisi_i2c_clear_int(struct hisi_i2c_controller *ctlr, u32 mask)
+{
+	writel(mask, ctlr->iobase + HISI_I2C_INT_CLR);
+}
+
+static void hisi_i2c_handle_errors(struct hisi_i2c_controller *ctlr)
+{
+	u32 int_err = ctlr->xfer_err, reg;
+
+	if (int_err & HISI_I2C_INT_FIFO_ERR) {
+		reg = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);
+
+		if (reg & HISI_I2C_FIFO_STATE_RX_RERR)
+			dev_err(ctlr->dev, "rx fifo error read.\n");
+
+		if (reg & HISI_I2C_FIFO_STATE_RX_WERR)
+			dev_err(ctlr->dev, "rx fifo error write.\n");
+
+		if (reg & HISI_I2C_FIFO_STATE_TX_RERR)
+			dev_err(ctlr->dev, "tx fifo error read.\n");
+
+		if (reg & HISI_I2C_FIFO_STATE_TX_WERR)
+			dev_err(ctlr->dev, "tx fifo error write.\n");
+	}
+}
+
+static int hisi_i2c_start_xfer(struct hisi_i2c_controller *ctlr)
+{
+	struct i2c_msg *msg = ctlr->msgs;
+	u32 reg;
+
+	reg = readl(ctlr->iobase + HISI_I2C_FRAME_CTRL);
+	reg &= ~HISI_I2C_FRAME_CTRL_ADDR_TEN;
+	if (msg->flags & I2C_M_TEN)
+		reg |= HISI_I2C_FRAME_CTRL_ADDR_TEN;
+	writel(reg, ctlr->iobase + HISI_I2C_FRAME_CTRL);
+
+	reg = readl(ctlr->iobase + HISI_I2C_SLV_ADDR);
+	reg &= ~HISI_I2C_SLV_ADDR_VAL;
+	reg |= FIELD_PREP(HISI_I2C_SLV_ADDR_VAL, msg->addr);
+	writel(reg, ctlr->iobase + HISI_I2C_SLV_ADDR);
+
+	reg = readl(ctlr->iobase + HISI_I2C_FIFO_CTRL);
+	reg |= HISI_I2C_FIFO_RX_CLR | HISI_I2C_FIFO_TX_CLR;
+	writel(reg, ctlr->iobase + HISI_I2C_FIFO_CTRL);
+	reg &= ~(HISI_I2C_FIFO_RX_CLR | HISI_I2C_FIFO_TX_CLR);
+	writel(reg, ctlr->iobase + HISI_I2C_FIFO_CTRL);
+
+	hisi_i2c_clear_int(ctlr, HISI_I2C_INT_ALL);
+	hisi_i2c_enable_int(ctlr, HISI_I2C_INT_ALL);
+
+	return 0;
+}
+
+static void hisi_i2c_reset_xfer(struct hisi_i2c_controller *ctlr)
+{
+	ctlr->msg_num = 0;
+	ctlr->xfer_err = 0;
+	ctlr->msg_tx_idx = 0;
+	ctlr->msg_rx_idx = 0;
+	ctlr->buf_tx_idx = 0;
+	ctlr->buf_rx_idx = 0;
+}
+
+/*
+ * Initialize the transfer information and start the I2C bus transfer.
+ * We only configure the transfer and do some pre/post works here, and
+ * wait for the transfer done. The major transfer process is performed
+ * in the IRQ handler.
+ */
+static int hisi_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
+				int num)
+{
+	struct hisi_i2c_controller *ctlr = i2c_get_adapdata(adap);
+	DECLARE_COMPLETION_ONSTACK(done);
+	int ret = num;
+
+	hisi_i2c_reset_xfer(ctlr);
+	ctlr->completion = &done;
+	ctlr->msg_num = num;
+	ctlr->msgs = msgs;
+
+	hisi_i2c_start_xfer(ctlr);
+
+	if (!wait_for_completion_timeout(ctlr->completion, adap->timeout)) {
+		hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);
+		i2c_recover_bus(&ctlr->adapter);
+		dev_err(ctlr->dev, "bus transfer timeout\n");
+		ret = -EIO;
+	}
+
+	if (ctlr->xfer_err) {
+		hisi_i2c_handle_errors(ctlr);
+		ret = -EIO;
+	}
+
+	hisi_i2c_reset_xfer(ctlr);
+	ctlr->completion = NULL;
+
+	return ret;
+}
+
+static u32 hisi_i2c_functionality(struct i2c_adapter *adap)
+{
+	return I2C_FUNC_I2C | I2C_FUNC_10BIT_ADDR | I2C_FUNC_SMBUS_EMUL;
+}
+
+static const struct i2c_algorithm hisi_i2c_algo = {
+	.master_xfer	= hisi_i2c_master_xfer,
+	.functionality	= hisi_i2c_functionality,
+};
+
+static int hisi_i2c_read_rx_fifo(struct hisi_i2c_controller *ctlr)
+{
+	struct i2c_msg *cur_msg;
+	u32 fifo_state;
+
+	while (ctlr->msg_rx_idx < ctlr->msg_num) {
+		cur_msg = ctlr->msgs + ctlr->msg_rx_idx;
+
+		if (!(cur_msg->flags & I2C_M_RD)) {
+			ctlr->msg_rx_idx++;
+			continue;
+		}
+
+		fifo_state = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);
+		while (!(fifo_state & HISI_I2C_FIFO_STATE_RX_EMPTY) &&
+		       ctlr->buf_rx_idx < cur_msg->len) {
+			cur_msg->buf[ctlr->buf_rx_idx++] = readl(ctlr->iobase + HISI_I2C_RXDATA);
+			fifo_state = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);
+		}
+
+		if (ctlr->buf_rx_idx == cur_msg->len) {
+			ctlr->buf_rx_idx = 0;
+			ctlr->msg_rx_idx++;
+		}
+
+		if (fifo_state & HISI_I2C_FIFO_STATE_RX_EMPTY)
+			break;
+	}
+
+	return 0;
+}
+
+static void hisi_i2c_xfer_msg(struct hisi_i2c_controller *ctlr)
+{
+	int max_write = HISI_I2C_TX_FIFO_DEPTH;
+	bool need_restart = false, last_msg;
+	struct i2c_msg *cur_msg;
+	u32 cmd, fifo_state;
+
+	while (ctlr->msg_tx_idx < ctlr->msg_num) {
+		cur_msg = ctlr->msgs + ctlr->msg_tx_idx;
+		last_msg = (ctlr->msg_tx_idx == ctlr->msg_num - 1);
+
+		/* Signal the SR bit when we start transferring a new message */
+		if (ctlr->msg_tx_idx && !ctlr->buf_tx_idx)
+			need_restart = true;
+
+		fifo_state = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);
+		while (!(fifo_state & HISI_I2C_FIFO_STATE_TX_FULL) &&
+		       ctlr->buf_tx_idx < cur_msg->len &&
+		       max_write) {
+			cmd = 0;
+
+			if (need_restart) {
+				cmd |= HISI_I2C_CMD_TXDATA_SR_EN;
+				need_restart = false;
+			}
+
+			/* Signal the STOP bit at the last frame of the last message */
+			if (ctlr->buf_tx_idx == cur_msg->len - 1 &&
+			    last_msg)
+				cmd |= HISI_I2C_CMD_TXDATA_P_EN;
+
+			if (cur_msg->flags & I2C_M_RD)
+				cmd |= HISI_I2C_CMD_TXDATA_RW;
+			else
+				cmd |= FIELD_PREP(HISI_I2C_CMD_TXDATA_DATA,
+						  cur_msg->buf[ctlr->buf_tx_idx]);
+
+			writel(cmd, ctlr->iobase + HISI_I2C_CMD_TXDATA);
+			ctlr->buf_tx_idx++;
+			max_write--;
+
+			fifo_state = readl(ctlr->iobase + HISI_I2C_FIFO_STATE);
+		}
+
+		/* Update the transfer index after per message transfer is done. */
+		if (ctlr->buf_tx_idx == cur_msg->len) {
+			ctlr->buf_tx_idx = 0;
+			ctlr->msg_tx_idx++;
+		}
+
+		if ((fifo_state & HISI_I2C_FIFO_STATE_TX_FULL) ||
+		    max_write == 0)
+			break;
+	}
+}
+
+static irqreturn_t hisi_i2c_irq(int irq, void *context)
+{
+	struct hisi_i2c_controller *ctlr = context;
+	u32 int_stat;
+
+	int_stat = readl(ctlr->iobase + HISI_I2C_INT_MSTAT);
+	hisi_i2c_clear_int(ctlr, int_stat);
+	if (!(int_stat & HISI_I2C_INT_ALL))
+		return IRQ_NONE;
+
+	if (int_stat & HISI_I2C_INT_TX_EMPTY)
+		hisi_i2c_xfer_msg(ctlr);
+
+	if (int_stat & HISI_I2C_INT_ERR) {
+		ctlr->xfer_err = int_stat;
+		goto out;
+	}
+
+	/* Drain the rx fifo before finish the transfer */
+	if (int_stat & (HISI_I2C_INT_TRANS_CPLT | HISI_I2C_INT_RX_FULL))
+		hisi_i2c_read_rx_fifo(ctlr);
+
+out:
+	if (int_stat & HISI_I2C_INT_TRANS_CPLT || ctlr->xfer_err) {
+		hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);
+		hisi_i2c_clear_int(ctlr, HISI_I2C_INT_ALL);
+		complete(ctlr->completion);
+	}
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * Helper function for calculating and configuring the HIGH and LOW
+ * periods of SCL clock. The caller will pass the ratio of the
+ * counts (divide / divisor) according to the target speed mode,
+ * and the target registers.
+ */
+static void hisi_i2c_set_scl(struct hisi_i2c_controller *ctlr,
+			     u32 divide, u32 divisor,
+			     u32 reg_hcnt, u32 reg_lcnt)
+{
+	u32 total_cnt, t_scl_hcnt, t_scl_lcnt, scl_fall_cnt, scl_rise_cnt;
+	u32 scl_hcnt, scl_lcnt;
+
+	/* Total SCL clock cycles per speed period */
+	total_cnt = DIV_ROUND_UP_ULL(ctlr->clk_rate_khz * 1000, ctlr->bus_freq_hz);
+	/* Total HIGH level SCL clock cycles including edges */
+	t_scl_hcnt = DIV_ROUND_UP_ULL(total_cnt * divide, divisor);
+	/* Total LOW level SCL clock cycles including edges */
+	t_scl_lcnt = total_cnt - t_scl_hcnt;
+	/* Fall edge SCL clock cycles */
+	scl_fall_cnt = NSEC_TO_CYCLES(ctlr->scl_fall_time, ctlr->clk_rate_khz);
+	/* Rise edge SCL clock cycles */
+	scl_rise_cnt = NSEC_TO_CYCLES(ctlr->scl_rise_time, ctlr->clk_rate_khz);
+
+	/* Calculated HIGH and LOW periods of SCL clock */
+	scl_hcnt = t_scl_hcnt - ctlr->spk_len - 7 - scl_fall_cnt;
+	scl_lcnt = t_scl_lcnt - 1 - scl_rise_cnt;
+
+	writel(scl_hcnt, ctlr->iobase + reg_hcnt);
+	writel(scl_lcnt, ctlr->iobase + reg_lcnt);
+}
+
+static void hisi_i2c_configure_bus(struct hisi_i2c_controller *ctlr)
+{
+	u32 reg, sda_hold_cnt, speed_mode;
+	struct i2c_timings t;
+
+	i2c_parse_fw_timings(ctlr->dev, &t, true);
+	ctlr->bus_freq_hz = t.bus_freq_hz;
+	ctlr->spk_len = NSEC_TO_CYCLES(t.digital_filter_width_ns, ctlr->clk_rate_khz);
+	ctlr->scl_fall_time = t.scl_fall_ns;
+	ctlr->scl_rise_time = t.scl_rise_ns;
+	ctlr->sda_hold_time = t.sda_hold_ns;
+
+	switch (ctlr->bus_freq_hz) {
+	case I2C_MAX_FAST_MODE_FREQ:
+		speed_mode = HISI_I2C_FAST_SPEED_MODE;
+		hisi_i2c_set_scl(ctlr, 26, 76, HISI_I2C_FS_SCL_HCNT, HISI_I2C_FS_SCL_LCNT);
+		break;
+	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
+		speed_mode = HISI_I2C_HIGH_SPEED_MODE;
+		hisi_i2c_set_scl(ctlr, 6, 22, HISI_I2C_HS_SCL_HCNT, HISI_I2C_HS_SCL_LCNT);
+		break;
+	case I2C_MAX_STANDARD_MODE_FREQ:
+	default:
+		speed_mode = HISI_I2C_STD_SPEED_MODE;
+
+		/* For default condition force the bus speed to standard mode. */
+		ctlr->bus_freq_hz = I2C_MAX_STANDARD_MODE_FREQ;
+		hisi_i2c_set_scl(ctlr, 40, 87, HISI_I2C_SS_SCL_HCNT, HISI_I2C_SS_SCL_LCNT);
+		break;
+	}
+
+	reg = readl(ctlr->iobase + HISI_I2C_FRAME_CTRL);
+	reg &= ~HISI_I2C_FRAME_CTRL_SPEED_MODE;
+	reg |= FIELD_PREP(HISI_I2C_FRAME_CTRL_SPEED_MODE, speed_mode);
+	writel(reg, ctlr->iobase + HISI_I2C_FRAME_CTRL);
+
+	sda_hold_cnt = NSEC_TO_CYCLES(ctlr->sda_hold_time, ctlr->clk_rate_khz);
+
+	reg = FIELD_PREP(HISI_I2C_SDA_HOLD_TX, sda_hold_cnt);
+	writel(reg, ctlr->iobase + HISI_I2C_SDA_HOLD);
+
+	writel(ctlr->spk_len, ctlr->iobase + HISI_I2C_FS_SPK_LEN);
+
+	reg = FIELD_PREP(HISI_I2C_FIFO_RX_AF_THRESH, HISI_I2C_RX_F_AF_THRESH);
+	reg |= FIELD_PREP(HISI_I2C_FIFO_TX_AE_THRESH, HISI_I2C_TX_F_AE_THRESH);
+	writel(reg, ctlr->iobase + HISI_I2C_FIFO_CTRL);
+}
+
+static int hisi_i2c_probe(struct platform_device *pdev)
+{
+	struct hisi_i2c_controller *ctlr;
+	struct device *dev = &pdev->dev;
+	struct i2c_adapter *adapter;
+	u32 hw_version;
+	int ret;
+
+	ctlr = devm_kzalloc(dev, sizeof(*ctlr), GFP_KERNEL);
+	if (!ctlr)
+		return -ENOMEM;
+
+	ctlr->dev = dev;
+	ctlr->iobase = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(ctlr->iobase))
+		return PTR_ERR(ctlr->iobase);
+
+	ctlr->irq = platform_get_irq(pdev, 0);
+	if (ctlr->irq < 0)
+		return ctlr->irq;
+
+	hisi_i2c_disable_int(ctlr, HISI_I2C_INT_ALL);
+
+	ret = devm_request_irq(dev, ctlr->irq, hisi_i2c_irq,
+			       0, "hisi-i2c", ctlr);
+	if (ret) {
+		dev_err(dev, "failed to request irq handler, ret = %d.\n", ret);
+		return ret;
+	}
+
+	ret = device_property_read_u64(dev, "clk_rate", &ctlr->clk_rate_khz);
+	if (ret) {
+		dev_err(dev, "failed to get clock frequency, ret = %d.\n", ret);
+		return ret;
+	}
+	ctlr->clk_rate_khz = DIV_ROUND_UP_ULL(ctlr->clk_rate_khz, 1000);
+
+	hisi_i2c_configure_bus(ctlr);
+
+	adapter = &ctlr->adapter;
+	snprintf(adapter->name, sizeof(adapter->name),
+		 "HiSilicon I2C Controller %s", dev_name(dev));
+	adapter->owner = THIS_MODULE;
+	adapter->algo = &hisi_i2c_algo;
+	adapter->dev.parent = dev;
+	i2c_set_adapdata(adapter, ctlr);
+
+	ret = devm_i2c_add_adapter(dev, adapter);
+	if (ret) {
+		dev_err(dev, "failed to add i2c adapter, ret = %d.\n", ret);
+		return ret;
+	}
+
+	hw_version = readl(ctlr->iobase + HISI_I2C_VERSION);
+	dev_info(ctlr->dev, "speed mode %s. hw version 0x%x.\n",
+		 hisi_i2c_speed_string(ctlr->bus_freq_hz), hw_version);
+
+	return 0;
+}
+
+static const struct acpi_device_id hisi_i2c_acpi_ids[] = {
+	{ "HISI03D1", 0 },
+	{ }
+};
+MODULE_DEVICE_TABLE(acpi, hisi_i2c_acpi_ids);
+MODULE_ALIAS("platform:hisi-i2c");
+
+static struct platform_driver hisi_i2c_driver = {
+	.probe		= hisi_i2c_probe,
+	.driver		= {
+		.name	= "hisi-i2c",
+		.acpi_match_table = hisi_i2c_acpi_ids,
+	},
+};
+
+module_platform_driver(hisi_i2c_driver);
+
+MODULE_AUTHOR("Yicong Yang <yangyicong@hisilicon.com>");
+MODULE_DESCRIPTION("HiSilicon I2C Controller Driver");
+MODULE_LICENSE("GPL");