net: dsa: microchip: add KSZ9477 I2C driver

drivers/net/dsa/microchip/ksz9477_i2c.c new file +0 -0

new file mode 100644
index 0000000..29511e9

drivers/net/dsa/microchip/ksz9477_i2c.c new file +12 -0

+static int ksz_i2c_read24(struct ksz_device *dev, u32 reg, u32 *val)
+{
+	int ret;
+
+	*val = 0;
+	ret = ksz_i2c_read(dev, reg, (u8 *)val, 3);
+	if (!ret) {
+		*val = be32_to_cpu(*val);
+		/* convert to 24bit */
+		*val >>= 8;
+	}
+

Is there any specific reason that ksz_i2c_read24 & ksz_i2c_write24 in ksz9477_i2c.c
but, other functions are in ksz_i2c.h?

Quoted 4 lines

Is there any specific reason that ksz_i2c_read24 & ksz_i2c_write24 in ksz9477_i2c.c
but, other functions are in ksz_i2c.h?

There are specific registers in KSZ9477 such that using 24-bit is more efficient than reading 32-bit first and writing 32-bit. The other older switches mostly use 8-bit, while one uses 16-bit.

From: Tristram Ha <[email protected]>

Add KSZ9477 I2C driver support. The code ksz9477.c and ksz_common.c are used together to generate the I2C driver.

Signed-off-by Tristram Ha <[email protected]>

Can this use regmap instead ?

Quoted 2 lines

Can this use regmap instead ?

To tell the truth I do not know how.

Some customers, like Sergio, seem to be using KSZ9897 with I2C in their projects and so cannot wait further.

Quoted 7 lines · On 12/19/2018 11:30 PM, [email protected] wrote:

Quoted 2 lines

Can this use regmap instead ?

To tell the truth I do not know how.

Some customers, like Sergio, seem to be using KSZ9897 with I2C in their projects and so cannot wait further.

Well, I just did 'git grep regmap_i2c drivers' and 'git grep regmap_spi drivers/' and found eg.
drivers/iio/pressure/zpa2326_i2c.c drivers/iio/pressure/zpa2326_spi.c

There's plenty of drivers using regmap in drivers/ to demonstrate how to use it. And there's always include/linux/regmap.h .

In this specific (DSA) case, register either regmap_spi or regmap_i2c in the ksz_switch_register(), pop the regmap pointer into ksz_device and then use it in ksz_read*()/ksz_write*(). You can then, in a subsequent patch, massage the regmap accesses directly into the code instead and drop the ksz_read*()/ksz_write*() functions altogether, since those would be just regmap function wrappers.

And if you're still lost, I can start on the conversion myself.

Quoted 24 lines · Marek Vasut

Quoted 7 lines · On 12/19/2018 11:30 PM, [email protected] wrote:

Quoted 2 lines

Can this use regmap instead ?

To tell the truth I do not know how.

Some customers, like Sergio, seem to be using KSZ9897 with I2C in their projects and so cannot wait further.

Well, I just did 'git grep regmap_i2c drivers' and 'git grep regmap_spi drivers/' and found eg.
drivers/iio/pressure/zpa2326_i2c.c drivers/iio/pressure/zpa2326_spi.c

There's plenty of drivers using regmap in drivers/ to demonstrate how to use it. And there's always include/linux/regmap.h .

In this specific (DSA) case, register either regmap_spi or regmap_i2c in the ksz_switch_register(), pop the regmap pointer into ksz_device and then use it in ksz_read*()/ksz_write*(). You can then, in a subsequent patch, massage the regmap accesses directly into the code instead and drop the ksz_read*()/ksz_write*() functions altogether, since those would be just regmap function wrappers.

Quoted 1 line

And if you're still lost, I can start on the conversion myself.

Actually, wait a few hours, I'll try to prepare a regmap patchset.

Quoted 17 lines

Quoted 16 lines

Well, I just did 'git grep regmap_i2c drivers' and 'git grep regmap_spi drivers/' and found eg.
drivers/iio/pressure/zpa2326_i2c.c drivers/iio/pressure/zpa2326_spi.c

There's plenty of drivers using regmap in drivers/ to demonstrate how to use it. And there's always include/linux/regmap.h .

In this specific (DSA) case, register either regmap_spi or regmap_i2c in the ksz_switch_register(), pop the regmap pointer into ksz_device and then use it in ksz_read*()/ksz_write*(). You can then, in a subsequent patch, massage the regmap accesses directly into the code instead and drop the ksz_read*()/ksz_write*() functions altogether, since those would be just regmap function wrappers.

And if you're still lost, I can start on the conversion myself.

Quoted 1 line

Actually, wait a few hours, I'll try to prepare a regmap patchset.

Try this [1], it should be a start. The last two patches might need some adjustments, so please test and let me know how that works (or better yet, provide fixes). Adding i2c regmap should then be trivial :)

[1] https://patchwork.ozlabs.org/cover/1016432/

Quoted 23 lines

Quoted 15 lines

Quoted 15 lines

Well, I just did 'git grep regmap_i2c drivers' and 'git grep regmap_spi drivers/' and found eg.
drivers/iio/pressure/zpa2326_i2c.c drivers/iio/pressure/zpa2326_spi.c

There's plenty of drivers using regmap in drivers/ to demonstrate how to use it. And there's always include/linux/regmap.h .

In this specific (DSA) case, register either regmap_spi or regmap_i2c in the ksz_switch_register(), pop the regmap pointer into ksz_device and then use it in ksz_read*()/ksz_write*(). You can then, in a subsequent patch, massage the regmap accesses directly into the code instead and drop the ksz_read*()/ksz_write*() functions altogether, since those would be just regmap function wrappers.

Quoted 3 lines

Quoted 1 line

And if you're still lost, I can start on the conversion myself.

Actually, wait a few hours, I'll try to prepare a regmap patchset.

Try this [1], it should be a start. The last two patches might need some adjustments, so please test and let me know how that works (or better yet, provide fixes). Adding i2c regmap should then be trivial :)

Quoted 1 line

[1] https://patchwork.ozlabs.org/cover/1016432/

I did look at the code in regmap_i2c.c and regmap_spi.c.

So the assumption is there is a regmap_read function that automatically maps to regmap_i2c_read or regmap_spi_read depending on the device.

But the SPI access does not use the register directly. It has to convert it into a special format recognized by the chip. The size used by SPI is 4, while I2C access uses the standard way and the size is 2.

From: Tristram Ha <[email protected]>

Add KSZ9477 I2C driver support. The code ksz9477.c and ksz_common.c are used together to generate the I2C driver.

Signed-off-by Tristram Ha <[email protected]>

drivers/net/dsa/microchip/ksz_i2c.h new file +57 -0

new file mode 100644
index 0000000..b9af0a8
@@ -0,0 +1,69 @@
+/* SPDX-License-Identifier: GPL-2.0
+ * Microchip KSZ series I2C access common header
+ *
+ * Copyright (C) 2018 Microchip Technology Inc.
+ *	Tristram Ha <[email protected]>
+ */
+
+#ifndef __KSZ_I2C_H
+#define __KSZ_I2C_H
+
+/* Chip dependent I2C access */
+static int ksz_i2c_read(struct ksz_device *dev, u32 reg, u8 *data,
+			unsigned int len);
+static int ksz_i2c_write(struct ksz_device *dev, u32 reg, void *data,
+			 unsigned int len);
+
+static int ksz_i2c_read8(struct ksz_device *dev, u32 reg, u8 *val)
+{
+	return ksz_i2c_read(dev, reg, val, 1);
+}
+
+static int ksz_i2c_read16(struct ksz_device *dev, u32 reg, u16 *val)
+{
+	int ret = ksz_i2c_read(dev, reg, (u8 *)val, 2);
+
+	if (!ret)
+		*val = be16_to_cpu(*val);
+
+	return ret;
+}
+
+static int ksz_i2c_read32(struct ksz_device *dev, u32 reg, u32 *val)
+{
+	int ret = ksz_i2c_read(dev, reg, (u8 *)val, 4);
+
+	if (!ret)
+		*val = be32_to_cpu(*val);
+
+	return ret;
+}
+
+static int ksz_i2c_write8(struct ksz_device *dev, u32 reg, u8 value)
+{
+	return ksz_i2c_write(dev, reg, &value, 1);
+}
+
+static int ksz_i2c_write16(struct ksz_device *dev, u32 reg, u16 value)
+{
+	value = cpu_to_be16(value);
+	return ksz_i2c_write(dev, reg, &value, 2);
+}
+
+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}

drivers/net/dsa/microchip/ksz_i2c.h new file +2 -0

+
+#endif

Signature · 1 line

1.9.1

Quoted 10 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +10 -0

+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_read(dev, reg, data, len);
+}

drivers/net/dsa/microchip/ksz_i2c.h line 67

Quoted 5 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +5 -0

+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_write(dev, reg, data, len);
+}

This header file makes no sense. Please move the functions into .c

Hi!

Quoted 10 lines

Quoted 10 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +10 -0

+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_read(dev, reg, data, len);
+}

Quoted 8 lines

drivers/net/dsa/microchip/ksz_i2c.h line 67

Quoted 5 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +5 -0

+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_write(dev, reg, data, len);
+}

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

Pavel

Signature · 2 lines

DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

Quoted 23 lines · Wed, Dec 19, 2018 at 05:08:57PM CET, [email protected] wrote:

Hi!

Quoted 10 lines

Quoted 10 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +10 -0

+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_read(dev, reg, data, len);
+}

Quoted 8 lines

drivers/net/dsa/microchip/ksz_i2c.h line 67

Quoted 5 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +5 -0

+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_write(dev, reg, data, len);
+}

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

Quoted 2 lines

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Quoted 27 lines · Jiri Pirko

Quoted 25 lines · Wed, Dec 19, 2018 at 05:08:57PM CET, [email protected] wrote:

Hi!

Quoted 10 lines

Quoted 10 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +10 -0

+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_read(dev, reg, data, len);
+}

Quoted 8 lines

drivers/net/dsa/microchip/ksz_i2c.h line 67

Quoted 5 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +5 -0

+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_write(dev, reg, data, len);
+}

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

Quoted 4 lines

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Ok, sorry for the noise.

Quoted 7 lines

Quoted 7 lines

Quoted 6 lines

Quoted 1 line

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

Quoted 6 lines

Quoted 4 lines

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Ok, sorry for the noise.

If you were to use regmap, this whole boilerplate would go away ...

Quoted 12 lines

Quoted 12 lines

Quoted 11 lines

Quoted 6 lines

Quoted 1 line

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Quoted 3 lines

Quoted 1 line

Ok, sorry for the noise.

If you were to use regmap, this whole boilerplate would go away ...

Sometimes I am confused about this review process.

The new code is same as previous code submitted. The old code was accepted, but then there are objections to the new code.

So if I change the new code, should I go back to correct the old code?

The ksz9477_i2c.c should be the same as ksz9477_i2c.c while ksz_i2c.h matches ksz_spi.h. The idea is ksz_i2c.h can be used by another switch driver like ksz####_i2c.c.

So if I change the new code, should I go back to correct the old code?

Quoted 2 lines · On Wed, Dec 19, 2018 at 10:50:16PM +0000, [email protected] wrote:

So if I change the new code, should I go back to correct the old code?

Other operating systems would worry about breaking code which is working fine, but re-writing ancient code is The Linux Way.

Quoted 6 lines · Dan Carpenter

Quoted 2 lines · On Wed, Dec 19, 2018 at 10:50:16PM +0000, [email protected] wrote:

So if I change the new code, should I go back to correct the old code?

Other operating systems would worry about breaking code which is working fine, but re-writing ancient code is The Linux Way.

So ... when will we finally rewrite Linux in rust ? :-)

Quoted 8 lines · Marek Vasut

Quoted 6 lines · Dan Carpenter

Quoted 2 lines · On Wed, Dec 19, 2018 at 10:50:16PM +0000, [email protected] wrote:

So if I change the new code, should I go back to correct the old code?

Other operating systems would worry about breaking code which is working fine, but re-writing ancient code is The Linux Way.

Quoted 1 line

So ... when will we finally rewrite Linux in rust ? :-)

Probably we'll move more modules to userspace in time.

Wasn't there a joke proof of concept a while back that let people write kernel modules in perl?

Quoted 12 lines

Quoted 12 lines

Quoted 11 lines

Quoted 6 lines

Quoted 1 line

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Quoted 3 lines

Quoted 1 line

Ok, sorry for the noise.

If you were to use regmap, this whole boilerplate would go away ...

Sometimes I am confused about this review process.

The new code is same as previous code submitted. The old code was accepted, but then there are objections to the new code.

Quoted 18 lines · On Wed 2018-12-19 22:50:16, [email protected] wrote:

Quoted 7 lines

Quoted 7 lines

Quoted 7 lines

Quoted 6 lines

Quoted 1 line

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

Quoted 8 lines

Quoted 6 lines

Quoted 4 lines

1) It makes no sense to have header files for things like this. The functions are only used within the single .c file.

2) You cannot inline them, as they are used as ops.

Ok, sorry for the noise.

If you were to use regmap, this whole boilerplate would go away ...

Sometimes I am confused about this review process.

Quoted 2 lines

The new code is same as previous code submitted. The old code was accepted, but then there are objections to the new code.

Sometimes you just need to reject review comment...

It took extremely long to merge this driver; that does not help, either.

Hi!

Quoted 10 lines

Quoted 10 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +10 -0

+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+	value = cpu_to_be32(value);
+	return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_read(dev, reg, data, len);
+}

Quoted 8 lines

drivers/net/dsa/microchip/ksz_i2c.h line 67

Quoted 5 lines

drivers/net/dsa/microchip/ksz_i2c.h new file +5 -0

+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+	return ksz_i2c_write(dev, reg, data, len);
+}

This header file makes no sense. Please move the functions into .c

No, that would make code bigger & slower.

It makes sense to me. But I'd add "inline" keyword to make the goal explicit.

From: Tristram Ha <[email protected]>

Add KSZ9477 I2C driver support. The code ksz9477.c and ksz_common.c are used together to generate the I2C driver.

Signed-off-by Tristram Ha <[email protected]>

---
v1
- Return error code from i2c_transfer
- Change GPL license
- Change author

 drivers/net/dsa/microchip/Kconfig       |   6 +
 drivers/net/dsa/microchip/Makefile      |   1 +
 drivers/net/dsa/microchip/ksz9477_i2c.c | 196 ++++++++++++++++++++++++++++++++
 drivers/net/dsa/microchip/ksz_i2c.h     |  69 +++++++++++
 4 files changed, 272 insertions(+)
 create mode 100644 drivers/net/dsa/microchip/ksz9477_i2c.c
 create mode 100644 drivers/net/dsa/microchip/ksz_i2c.h

drivers/net/dsa/microchip/Kconfig +6 -0

index bea29fd..fd94441 100644
@@ -9,6 +9,12 @@ menuconfig NET_DSA_MICROCHIP_KSZ9477
        help
          This driver adds support for Microchip KSZ9477 switch chips.
 
+config NET_DSA_MICROCHIP_KSZ9477_I2C
+       tristate "KSZ9477 series I2C connected switch driver"
+       depends on NET_DSA_MICROCHIP_KSZ9477 && I2C
+       help
+         Select to enable support for registering switches configured through I2C.
+
 config NET_DSA_MICROCHIP_KSZ9477_SPI
        tristate "KSZ9477 series SPI connected switch driver"
        depends on NET_DSA_MICROCHIP_KSZ9477 && SPI

drivers/net/dsa/microchip/Makefile +1 -0

index 3142c18..dbcc5db 100644
@@ -1,3 +1,4 @@
 obj-$(CONFIG_NET_DSA_MICROCHIP_KSZ_COMMON)     += ksz_common.o
 obj-$(CONFIG_NET_DSA_MICROCHIP_KSZ9477)                += ksz9477.o
+obj-$(CONFIG_NET_DSA_MICROCHIP_KSZ9477_I2C)    += ksz9477_i2c.o
 obj-$(CONFIG_NET_DSA_MICROCHIP_KSZ9477_SPI)    += ksz9477_spi.o

drivers/net/dsa/microchip/ksz9477_i2c.c new file +196 -0

new file mode 100644
index 0000000..29511e9
@@ -0,0 +1,196 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Microchip KSZ9477 series register access through I2C
+ *
+ * Copyright (C) 2018 Microchip Technology Inc.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/i2c.h>
+
+#include "ksz_priv.h"
+#include "ksz_i2c.h"
+
+/* Enough to read all switch port registers. */
+#define I2C_TX_BUF_LEN                 0x100
+
+static int ksz9477_i2c_read_reg(struct i2c_client *i2c, u32 reg, u8 *val,
+                               unsigned int len)
+{
+       struct i2c_msg msg[2];
+       int ret;
+
+       val[0] = (u8)(reg >> 8);
+       val[1] = (u8)reg;
+
+       msg[0].addr = i2c->addr;
+       msg[0].flags = 0;
+       msg[0].len = 2;
+       msg[0].buf = val;
+
+       msg[1].addr = i2c->addr;
+       msg[1].flags = I2C_M_RD;
+       msg[1].len = len;
+       msg[1].buf = &val[2];
+
+       ret = i2c_transfer(i2c->adapter, msg, 2);
+       if (ret == 2)
+               ret = 0;
+       return ret;
+}
+
+static int ksz9477_i2c_write_reg(struct i2c_client *i2c, u32 reg, u8 *val,
+                                unsigned int len)
+{
+       struct i2c_msg msg;
+       int ret;
+
+       val[0] = (u8)(reg >> 8);
+       val[1] = (u8)reg;
+
+       msg.addr = i2c->addr;
+       msg.flags = 0;
+       msg.len = 2 + len;
+       msg.buf = val;
+
+       ret = i2c_transfer(i2c->adapter, &msg, 1);
+       if (ret == 1)
+               ret = 0;
+       return ret;
+}
+
+static int ksz_i2c_read(struct ksz_device *dev, u32 reg, u8 *data,
+                       unsigned int len)
+{
+       struct i2c_client *i2c = dev->priv;
+       int ret;
+
+       ret = ksz9477_i2c_read_reg(i2c, reg, dev->txbuf, len);
+       if (!ret)
+               memcpy(data, &dev->txbuf[2], len);
+       return ret;
+}
+
+static int ksz_i2c_write(struct ksz_device *dev, u32 reg, void *data,
+                        unsigned int len)
+{
+       struct i2c_client *i2c = dev->priv;
+
+       if (len > I2C_TX_BUF_LEN)
+               len = I2C_TX_BUF_LEN;
+       memcpy(&dev->txbuf[2], data, len);
+       return ksz9477_i2c_write_reg(i2c, reg, dev->txbuf, len);
+}
+
+static int ksz_i2c_read24(struct ksz_device *dev, u32 reg, u32 *val)
+{
+       int ret;
+
+       *val = 0;
+       ret = ksz_i2c_read(dev, reg, (u8 *)val, 3);
+       if (!ret) {
+               *val = be32_to_cpu(*val);
+               /* convert to 24bit */
+               *val >>= 8;
+       }
+
+       return ret;
+}
+
+static int ksz_i2c_write24(struct ksz_device *dev, u32 reg, u32 value)
+{
+       /* make it to big endian 24bit from MSB */
+       value <<= 8;
+       value = cpu_to_be32(value);
+       return ksz_i2c_write(dev, reg, &value, 3);
+}
+
+static const struct ksz_io_ops ksz9477_i2c_ops = {
+       .read8 = ksz_i2c_read8,
+       .read16 = ksz_i2c_read16,
+       .read24 = ksz_i2c_read24,
+       .read32 = ksz_i2c_read32,
+       .write8 = ksz_i2c_write8,
+       .write16 = ksz_i2c_write16,
+       .write24 = ksz_i2c_write24,
+       .write32 = ksz_i2c_write32,
+       .get = ksz_i2c_get,
+       .set = ksz_i2c_set,
+};
+
+static int ksz9477_i2c_probe(struct i2c_client *i2c,
+                            const struct i2c_device_id *i2c_id)
+{
+       struct ksz_device *dev;
+       int ret;
+
+       dev = ksz_switch_alloc(&i2c->dev, &ksz9477_i2c_ops, i2c);
+       if (!dev)
+               return -ENOMEM;
+
+       if (i2c->dev.platform_data)
+               dev->pdata = i2c->dev.platform_data;
+
+       dev->txbuf = devm_kzalloc(dev->dev, 2 + I2C_TX_BUF_LEN, GFP_KERNEL);
+
+       ret = ksz9477_switch_register(dev);
+
+       /* Main DSA driver may not be started yet. */
+       if (ret)
+               return ret;
+
+       i2c_set_clientdata(i2c, dev);
+
+       return 0;
+}
+
+static int ksz9477_i2c_remove(struct i2c_client *i2c)
+{
+       struct ksz_device *dev = i2c_get_clientdata(i2c);
+
+       if (dev)
+               ksz_switch_remove(dev);
+
+       return 0;
+}
+
+static void ksz9477_i2c_shutdown(struct i2c_client *i2c)
+{
+       struct ksz_device *dev = i2c_get_clientdata(i2c);
+
+       if (dev && dev->dev_ops->shutdown)
+               dev->dev_ops->shutdown(dev);
+}
+
+static const struct i2c_device_id ksz9477_i2c_id[] = {
+       { "ksz9477-switch", 0 },
+       {},
+};
+
+MODULE_DEVICE_TABLE(i2c, ksz9477_i2c_id);
+
+static const struct of_device_id ksz9477_dt_ids[] = {
+       { .compatible = "microchip,ksz9477" },
+       { .compatible = "microchip,ksz9897" },
+       {},
+};
+MODULE_DEVICE_TABLE(of, ksz9477_dt_ids);
+
+static struct i2c_driver ksz9477_i2c_driver = {
+       .driver = {
+               .name   = "ksz9477-switch",
+               .owner  = THIS_MODULE,
+               .of_match_table = of_match_ptr(ksz9477_dt_ids),
+       },
+       .probe  = ksz9477_i2c_probe,
+       .remove = ksz9477_i2c_remove,
+       .shutdown = ksz9477_i2c_shutdown,
+       .id_table = ksz9477_i2c_id,
+};
+
+module_i2c_driver(ksz9477_i2c_driver);
+
+MODULE_AUTHOR("Tristram Ha <[email protected]>");
+MODULE_DESCRIPTION("Microchip KSZ9477 Series Switch I2C access Driver");
+MODULE_LICENSE("GPL v2");

drivers/net/dsa/microchip/ksz_i2c.h new file +69 -1

new file mode 100644
index 0000000..b9af0a8
@@ -0,0 +1,69 @@
+/* SPDX-License-Identifier: GPL-2.0
+ * Microchip KSZ series I2C access common header
+ *
+ * Copyright (C) 2018 Microchip Technology Inc.
+ *     Tristram Ha <[email protected]>
+ */
+
+#ifndef __KSZ_I2C_H
+#define __KSZ_I2C_H
+
+/* Chip dependent I2C access */
+static int ksz_i2c_read(struct ksz_device *dev, u32 reg, u8 *data,
+                       unsigned int len);
+static int ksz_i2c_write(struct ksz_device *dev, u32 reg, void *data,
+                        unsigned int len);
+
+static int ksz_i2c_read8(struct ksz_device *dev, u32 reg, u8 *val)
+{
+       return ksz_i2c_read(dev, reg, val, 1);
+}
+
+static int ksz_i2c_read16(struct ksz_device *dev, u32 reg, u16 *val)
+{
+       int ret = ksz_i2c_read(dev, reg, (u8 *)val, 2);
+
+       if (!ret)
+               *val = be16_to_cpu(*val);
+
+       return ret;
+}
+
+static int ksz_i2c_read32(struct ksz_device *dev, u32 reg, u32 *val)
+{
+       int ret = ksz_i2c_read(dev, reg, (u8 *)val, 4);
+
+       if (!ret)
+               *val = be32_to_cpu(*val);
+
+       return ret;
+}
+
+static int ksz_i2c_write8(struct ksz_device *dev, u32 reg, u8 value)
+{
+       return ksz_i2c_write(dev, reg, &value, 1);
+}
+
+static int ksz_i2c_write16(struct ksz_device *dev, u32 reg, u16 value)
+{
+       value = cpu_to_be16(value);
+       return ksz_i2c_write(dev, reg, &value, 2);
+}
+
+static int ksz_i2c_write32(struct ksz_device *dev, u32 reg, u32 value)
+{
+       value = cpu_to_be32(value);
+       return ksz_i2c_write(dev, reg, &value, 4);
+}
+
+static int ksz_i2c_get(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+       return ksz_i2c_read(dev, reg, data, len);
+}
+
+static int ksz_i2c_set(struct ksz_device *dev, u32 reg, void *data, size_t len)
+{
+       return ksz_i2c_write(dev, reg, data, len);
+}
+
+#endif
--

1.9.1