[PATCH v6 0/5] Add support for HiSilicon I2C controller

[PATCH v6 0/5] Add support for HiSilicon I2C controller

[Yicong Yang]

Add driver and MAINTAINERS for HiSilicon I2C controller on Kunpeng SoC. Also
provide the devm_*() variants for adding the I2C adapters. Add a public
api to provide I2C frequency mode strings and convert designware driver
to use it.

Change since v5:
- address the comment from Dmitry and add his Reviewed-by
- address the comment from Jarkko and add his Reviewed-by and Tested-by
- add Jarkko's Acked-by for designware patch
Link: https://lore.kernel.org/linux-i2c/1617113966-40498-1-git-send-email-yangyicong@hisilicon.com/

Change since v4:
- and Andy's Reviewed-by
- attach Andy's patch of switch designware driver to use i2c_freq_mode_string()
Link: https://lore.kernel.org/linux-i2c/1617109549-4013-1-git-send-email-yangyicong@hisilicon.com/
Link: https://lore.kernel.org/linux-i2c/20210330134633.29889-1-andriy.shevchenko@linux.intel.com/

Change since v3:
- split the bus mode string api to I2C as suggested by Andy
- simplify the devm variants and change the export format
- address the comments of the HiSilicon I2C driver from Andy and Dmitry, thanks!
Link: https://lore.kernel.org/linux-i2c/1616411413-7177-1-git-send-email-yangyicong@hisilicon.com/

Change since v2:
- handle -EPROBE_DEFER case when get irq number by platform_get_irq()
Link: https://lore.kernel.org/linux-i2c/1615296137-14558-1-git-send-email-yangyicong@hisilicon.com/

Change since v1:
- fix compile test error on 32bit arch, reported by intel lkp robot:
  64 bit division without using kernel wrapper in probe function.
Link:https://lore.kernel.org/linux-i2c/1615016946-55670-1-git-send-email-yangyicong@hisilicon.com/

Andy Shevchenko (1):
  i2c: designware: Switch over to i2c_freq_mode_string()

Yicong Yang (4):
  i2c: core: add managed function for adding i2c adapters
  i2c: core: add api to provide frequency mode strings
  i2c: add support for HiSilicon I2C controller
  MAINTAINERS: Add maintainer for HiSilicon I2C driver

 MAINTAINERS                                |   7 +
 drivers/i2c/busses/Kconfig                 |  10 +
 drivers/i2c/busses/Makefile                |   1 +
 drivers/i2c/busses/i2c-designware-master.c |  20 +-
 drivers/i2c/busses/i2c-hisi.c              | 505 +++++++++++++++++++++++++++++
 drivers/i2c/i2c-core-base.c                |  26 ++
 include/linux/i2c.h                        |  21 ++
 7 files changed, 574 insertions(+), 16 deletions(-)
 create mode 100644 drivers/i2c/busses/i2c-hisi.c

-- 
2.8.1

[PATCH v6 1/5] i2c: core: add managed function for adding i2c adapters

[Yicong Yang]

Some I2C controller drivers will only unregister the I2C
adapter in their .remove() callback, which can be done
by simply using a managed variant to add the I2C adapter.

So add the managed functions for adding the I2C adapter.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 drivers/i2c/i2c-core-base.c | 26 ++++++++++++++++++++++++++
 include/linux/i2c.h         |  1 +
 2 files changed, 27 insertions(+)

diff --git a/drivers/i2c/i2c-core-base.c b/drivers/i2c/i2c-core-base.c
index 63ebf72..de9402c 100644
--- a/drivers/i2c/i2c-core-base.c
+++ b/drivers/i2c/i2c-core-base.c
@@ -1703,6 +1703,32 @@ void i2c_del_adapter(struct i2c_adapter *adap)
 }
 EXPORT_SYMBOL(i2c_del_adapter);
 
+static void devm_i2c_del_adapter(void *adapter)
+{
+	i2c_del_adapter(adapter);
+}
+
+/**
+ * devm_i2c_add_adapter - device-managed variant of i2c_add_adapter()
+ * @dev: managing device for adding this I2C adapter
+ * @adapter: the adapter to add
+ * Context: can sleep
+ *
+ * Add adapter with dynamic bus number, same with i2c_add_adapter()
+ * but the adapter will be auto deleted on driver detach.
+ */
+int devm_i2c_add_adapter(struct device *dev, struct i2c_adapter *adapter)
+{
+	int ret;
+
+	ret = i2c_add_adapter(adapter);
+	if (ret)
+		return ret;
+
+	return devm_add_action_or_reset(dev, devm_i2c_del_adapter, adapter);
+}
+EXPORT_SYMBOL_GPL(devm_i2c_add_adapter);
+
 static void i2c_parse_timing(struct device *dev, char *prop_name, u32 *cur_val_p,
 			    u32 def_val, bool use_def)
 {
diff --git a/include/linux/i2c.h b/include/linux/i2c.h
index 5662265..10bd0b0 100644
--- a/include/linux/i2c.h
+++ b/include/linux/i2c.h
@@ -844,6 +844,7 @@ static inline void i2c_mark_adapter_resumed(struct i2c_adapter *adap)
  */
 #if IS_ENABLED(CONFIG_I2C)
 int i2c_add_adapter(struct i2c_adapter *adap);
+int devm_i2c_add_adapter(struct device *dev, struct i2c_adapter *adapter);
 void i2c_del_adapter(struct i2c_adapter *adap);
 int i2c_add_numbered_adapter(struct i2c_adapter *adap);
 
-- 
2.8.1

[PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Yicong Yang]

Some I2C drivers like Designware and HiSilicon will print the
bus frequency mode information, so add a public one that everyone
can make use of.

Tested-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 include/linux/i2c.h | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/include/linux/i2c.h b/include/linux/i2c.h
index 10bd0b0..7268180 100644
--- a/include/linux/i2c.h
+++ b/include/linux/i2c.h
@@ -47,6 +47,26 @@ typedef int (*i2c_slave_cb_t)(struct i2c_client *client,
 #define I2C_MAX_HIGH_SPEED_MODE_FREQ	3400000
 #define I2C_MAX_ULTRA_FAST_MODE_FREQ	5000000
 
+static inline const char *i2c_freq_mode_string(u32 bus_freq_hz)
+{
+	switch (bus_freq_hz) {
+	case I2C_MAX_STANDARD_MODE_FREQ:
+		return "Standard Mode (100 kHz)";
+	case I2C_MAX_FAST_MODE_FREQ:
+		return "Fast Mode (400 kHz)";
+	case I2C_MAX_FAST_MODE_PLUS_FREQ:
+		return "Fast Mode Plus (1.0 MHz)";
+	case I2C_MAX_TURBO_MODE_FREQ:
+		return "Turbo Mode (1.4 MHz)";
+	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
+		return "High Speed Mode (3.4 MHz)";
+	case I2C_MAX_ULTRA_FAST_MODE_FREQ:
+		return "Ultra Fast Mode (5.0 MHz)";
+	default:
+		return "Unknown Mode";
+	}
+}
+
 struct module;
 struct property_entry;
 
-- 
2.8.1

[PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Yicong Yang]

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.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 drivers/i2c/busses/Kconfig    |  10 +
 drivers/i2c/busses/Makefile   |   1 +
 drivers/i2c/busses/i2c-hisi.c | 505 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 516 insertions(+)
 create mode 100644 drivers/i2c/busses/i2c-hisi.c

diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
index 05ebf75..eddf7bf 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 || 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..4284935
--- /dev/null
+++ b/drivers/i2c/busses/i2c-hisi.c
@@ -0,0 +1,505 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * HiSilicon I2C Controller Driver for Kunpeng SoC
+ *
+ * Copyright (c) 2021 HiSilicon Technologies Co., Ltd.
+ */
+
+#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/mod_devicetable.h>
+#include <linux/platform_device.h>
+#include <linux/property.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	GENMASK(4, 0)
+#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		0
+#define HISI_I2C_FAST_SPEED_MODE	1
+#define HISI_I2C_HIGH_SPEED_MODE	2
+
+#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 HZ_PER_KHZ	1000
+
+#define NSEC_TO_CYCLES(ns, clk_rate_khz) \
+	DIV_ROUND_UP_ULL((clk_rate_khz) * (ns), NSEC_PER_MSEC)
+
+struct hisi_i2c_controller {
+	struct i2c_adapter adapter;
+	void __iomem *iobase;
+	struct device *dev;
+	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 */
+	struct i2c_timings t;
+	u64 clk_rate_khz;
+	u32 spk_len;
+};
+
+static void hisi_i2c_enable_int(struct hisi_i2c_controller *ctlr, u32 mask)
+{
+	writel_relaxed(mask, ctlr->iobase + HISI_I2C_INT_MASK);
+}
+
+static void hisi_i2c_disable_int(struct hisi_i2c_controller *ctlr, u32 mask)
+{
+	writel_relaxed((~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_relaxed(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);
+		synchronize_irq(ctlr->irq);
+		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 * HZ_PER_KHZ, ctlr->t.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->t.scl_fall_ns, ctlr->clk_rate_khz);
+	/* Rise edge SCL clock cycles */
+	scl_rise_cnt = NSEC_TO_CYCLES(ctlr->t.scl_rise_ns, 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;
+
+	i2c_parse_fw_timings(ctlr->dev, &ctlr->t, true);
+	ctlr->spk_len = NSEC_TO_CYCLES(ctlr->t.digital_filter_width_ns, ctlr->clk_rate_khz);
+
+	switch (ctlr->t.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->t.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->t.sda_hold_ns, 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->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;
+
+	ctlr->dev = dev;
+
+	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, HZ_PER_KHZ);
+
+	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 is %s. hw version 0x%x\n",
+		 i2c_freq_mode_string(ctlr->t.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);
+
+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");
-- 
2.8.1

[PATCH v6 4/5] MAINTAINERS: Add maintainer for HiSilicon I2C driver

[Yicong Yang]

Add maintainer for HiSilicon I2C driver.

Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 MAINTAINERS | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index 8d23b0e..da2754a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -8040,6 +8040,13 @@ F:	drivers/crypto/hisilicon/hpre/hpre.h
 F:	drivers/crypto/hisilicon/hpre/hpre_crypto.c
 F:	drivers/crypto/hisilicon/hpre/hpre_main.c
 
+HISILICON I2C CONTROLLER DRIVER
+M:	Yicong Yang <yangyicong@hisilicon.com>
+L:	linux-i2c@vger.kernel.org
+S:	Maintained
+W:	https://www.hisilicon.com
+F:	drivers/i2c/busses/i2c-hisi.c
+
 HISILICON LPC BUS DRIVER
 M:	john.garry@huawei.com
 S:	Maintained
-- 
2.8.1

[PATCH v6 5/5] i2c: designware: Switch over to i2c_freq_mode_string()

[Yicong Yang]

From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Use generic i2c_freq_mode_string() helper to print chosen bus speed.

Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
 drivers/i2c/busses/i2c-designware-master.c | 20 ++++----------------
 1 file changed, 4 insertions(+), 16 deletions(-)

diff --git a/drivers/i2c/busses/i2c-designware-master.c b/drivers/i2c/busses/i2c-designware-master.c
index dd27b9d..b64c4c8 100644
--- a/drivers/i2c/busses/i2c-designware-master.c
+++ b/drivers/i2c/busses/i2c-designware-master.c
@@ -35,10 +35,10 @@ static void i2c_dw_configure_fifo_master(struct dw_i2c_dev *dev)
 
 static int i2c_dw_set_timings_master(struct dw_i2c_dev *dev)
 {
-	const char *mode_str, *fp_str = "";
 	u32 comp_param1;
 	u32 sda_falling_time, scl_falling_time;
 	struct i2c_timings *t = &dev->timings;
+	const char *fp_str = "";
 	u32 ic_clk;
 	int ret;
 
@@ -153,22 +153,10 @@ static int i2c_dw_set_timings_master(struct dw_i2c_dev *dev)
 
 	ret = i2c_dw_set_sda_hold(dev);
 	if (ret)
-		goto out;
-
-	switch (dev->master_cfg & DW_IC_CON_SPEED_MASK) {
-	case DW_IC_CON_SPEED_STD:
-		mode_str = "Standard Mode";
-		break;
-	case DW_IC_CON_SPEED_HIGH:
-		mode_str = "High Speed Mode";
-		break;
-	default:
-		mode_str = "Fast Mode";
-	}
-	dev_dbg(dev->dev, "Bus speed: %s%s\n", mode_str, fp_str);
+		return ret;
 
-out:
-	return ret;
+	dev_dbg(dev->dev, "Bus speed: %s\n", i2c_freq_mode_string(t->bus_freq_hz));
+	return 0;
 }
 
 /**
-- 
2.8.1

Re: [PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 1062 bytes --]


> diff --git a/include/linux/i2c.h b/include/linux/i2c.h
> index 10bd0b0..7268180 100644
> --- a/include/linux/i2c.h
> +++ b/include/linux/i2c.h
> @@ -47,6 +47,26 @@ typedef int (*i2c_slave_cb_t)(struct i2c_client *client,
>  #define I2C_MAX_HIGH_SPEED_MODE_FREQ	3400000
>  #define I2C_MAX_ULTRA_FAST_MODE_FREQ	5000000
>  
> +static inline const char *i2c_freq_mode_string(u32 bus_freq_hz)
> +{
> +	switch (bus_freq_hz) {
> +	case I2C_MAX_STANDARD_MODE_FREQ:
> +		return "Standard Mode (100 kHz)";
> +	case I2C_MAX_FAST_MODE_FREQ:
> +		return "Fast Mode (400 kHz)";
> +	case I2C_MAX_FAST_MODE_PLUS_FREQ:
> +		return "Fast Mode Plus (1.0 MHz)";
> +	case I2C_MAX_TURBO_MODE_FREQ:
> +		return "Turbo Mode (1.4 MHz)";
> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
> +		return "High Speed Mode (3.4 MHz)";
> +	case I2C_MAX_ULTRA_FAST_MODE_FREQ:
> +		return "Ultra Fast Mode (5.0 MHz)";
> +	default:
> +		return "Unknown Mode";
> +	}
> +}

Any reason ehy this is an inline function? My gut feeling says it would
be better added to the core?


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 302 bytes --]


Only super minor stuff. Thanks to all the contributors and reviewers!

> +#define HZ_PER_KHZ	1000

KHZ_PER_HZ?

> +	ret = devm_i2c_add_adapter(dev, adapter);
> +	if (ret) {
> +		dev_err(dev, "failed to add i2c adapter, ret = %d\n", ret);

No need to print that. The core prints messages on failures.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 1/5] i2c: core: add managed function for adding i2c adapters

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 741 bytes --]

On Wed, Mar 31, 2021 at 09:36:26PM +0800, Yicong Yang wrote:
> Some I2C controller drivers will only unregister the I2C
> adapter in their .remove() callback, which can be done
> by simply using a managed variant to add the I2C adapter.
> 
> So add the managed functions for adding the I2C adapter.

Yes, I think we can do this. We just need to pay attention that people
make sure interrupts are disabled and/or freed when the adapter gets
removed. But this is also true for the current situation. I.e. removing
the adapter manually in remove() and then relying on devm_ to free the
interrupt is a bit dangerous.

> +	return devm_add_action_or_reset(dev, devm_i2c_del_adapter, adapter);

Cool, I didn't know this function.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 0/5] Add support for HiSilicon I2C controller

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 460 bytes --]

On Wed, Mar 31, 2021 at 09:36:25PM +0800, Yicong Yang wrote:
> Add driver and MAINTAINERS for HiSilicon I2C controller on Kunpeng SoC. Also
> provide the devm_*() variants for adding the I2C adapters. Add a public
> api to provide I2C frequency mode strings and convert designware driver
> to use it.

Thanks, this series is in a very good condition. One question about
patch 2, the rest are minor nits (see the patches themselves). It will
make it for 5.13.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Yicong Yang]

On 2021/4/7 4:02, Wolfram Sang wrote:
> 
> Only super minor stuff. Thanks to all the contributors and reviewers!
> 
>> +#define HZ_PER_KHZ	1000
> 
> KHZ_PER_HZ?

that doesn't match what we want. we want the count of HZs per one KHZ.

> 
>> +	ret = devm_i2c_add_adapter(dev, adapter);
>> +	if (ret) {
>> +		dev_err(dev, "failed to add i2c adapter, ret = %d\n", ret);
> 
> No need to print that. The core prints messages on failures.
> 

i check the code and find a little difference. the print in the core doesn't
have the device prefix so it may not tell us which device's registeration fails.
but considering we have device name embedded in the adapter name for this
driver, it's ok to remove this message.

Re: [PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Yicong Yang]

On 2021/4/7 3:54, Wolfram Sang wrote:
> 
>> diff --git a/include/linux/i2c.h b/include/linux/i2c.h
>> index 10bd0b0..7268180 100644
>> --- a/include/linux/i2c.h
>> +++ b/include/linux/i2c.h
>> @@ -47,6 +47,26 @@ typedef int (*i2c_slave_cb_t)(struct i2c_client *client,
>>  #define I2C_MAX_HIGH_SPEED_MODE_FREQ	3400000
>>  #define I2C_MAX_ULTRA_FAST_MODE_FREQ	5000000
>>  
>> +static inline const char *i2c_freq_mode_string(u32 bus_freq_hz)
>> +{
>> +	switch (bus_freq_hz) {
>> +	case I2C_MAX_STANDARD_MODE_FREQ:
>> +		return "Standard Mode (100 kHz)";
>> +	case I2C_MAX_FAST_MODE_FREQ:
>> +		return "Fast Mode (400 kHz)";
>> +	case I2C_MAX_FAST_MODE_PLUS_FREQ:
>> +		return "Fast Mode Plus (1.0 MHz)";
>> +	case I2C_MAX_TURBO_MODE_FREQ:
>> +		return "Turbo Mode (1.4 MHz)";
>> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
>> +		return "High Speed Mode (3.4 MHz)";
>> +	case I2C_MAX_ULTRA_FAST_MODE_FREQ:
>> +		return "Ultra Fast Mode (5.0 MHz)";
>> +	default:
>> +		return "Unknown Mode";
>> +	}
>> +}
> 
> Any reason ehy this is an inline function? My gut feeling says it would
> be better added to the core?
> 

it's not a complicated function so i didn't think it'll make much difference,
so i just put it in the header along with the coresponding macro definitions.
do you want me to move it to the core?

Thanks

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Jarkko Nikula]

Hi

On 3/31/21 4:36 PM, Yicong Yang wrote:
> +	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, HZ_PER_KHZ);
> +

I'd use a temp variable here for reading the "clk_rate" property in 
Hertz and calculating the derived kHz value from it. As a bonus allow to 
use u32 for clk_rate_khz instead of u64. u32 will still provide plenty 
of headroom :-)

Reason for temp variable is for me it's confusing to see statement like 
"rate_khz = rate_khz / 1000".

Jarkko

Re: [PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Andy Shevchenko]

On Wed, Apr 07, 2021 at 04:29:29PM +0800, Yicong Yang wrote:
> On 2021/4/7 3:54, Wolfram Sang wrote:
> > 
> >> diff --git a/include/linux/i2c.h b/include/linux/i2c.h
> >> index 10bd0b0..7268180 100644
> >> --- a/include/linux/i2c.h
> >> +++ b/include/linux/i2c.h
> >> @@ -47,6 +47,26 @@ typedef int (*i2c_slave_cb_t)(struct i2c_client *client,
> >>  #define I2C_MAX_HIGH_SPEED_MODE_FREQ	3400000
> >>  #define I2C_MAX_ULTRA_FAST_MODE_FREQ	5000000
> >>  
> >> +static inline const char *i2c_freq_mode_string(u32 bus_freq_hz)
> >> +{
> >> +	switch (bus_freq_hz) {
> >> +	case I2C_MAX_STANDARD_MODE_FREQ:
> >> +		return "Standard Mode (100 kHz)";
> >> +	case I2C_MAX_FAST_MODE_FREQ:
> >> +		return "Fast Mode (400 kHz)";
> >> +	case I2C_MAX_FAST_MODE_PLUS_FREQ:
> >> +		return "Fast Mode Plus (1.0 MHz)";
> >> +	case I2C_MAX_TURBO_MODE_FREQ:
> >> +		return "Turbo Mode (1.4 MHz)";
> >> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
> >> +		return "High Speed Mode (3.4 MHz)";
> >> +	case I2C_MAX_ULTRA_FAST_MODE_FREQ:
> >> +		return "Ultra Fast Mode (5.0 MHz)";
> >> +	default:
> >> +		return "Unknown Mode";
> >> +	}
> >> +}
> > 
> > Any reason ehy this is an inline function? My gut feeling says it would
> > be better added to the core?
> > 
> 
> it's not a complicated function so i didn't think it'll make much difference,
> so i just put it in the header along with the coresponding macro definitions.
> do you want me to move it to the core?

I guess exporting will save few dozens of bytes if the function is used more
than once. (All strings will be duplicated or multiplied in that case)

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Yicong Yang]

On 2021/4/7 16:32, Jarkko Nikula wrote:
> Hi
> 
> On 3/31/21 4:36 PM, Yicong Yang wrote:
>> +    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, HZ_PER_KHZ);
>> +
> 
> I'd use a temp variable here for reading the "clk_rate" property in Hertz and calculating the derived kHz value from it. As a bonus allow to use u32 for clk_rate_khz instead of u64. u32 will still provide plenty of headroom :-)
> 
> Reason for temp variable is for me it's confusing to see statement like "rate_khz = rate_khz / 1000".
> 

I can get this addressed in the updated version. Thanks for the suggestion!

Re: [PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Yicong Yang]

On 2021/4/7 18:08, Andy Shevchenko wrote:
> On Wed, Apr 07, 2021 at 04:29:29PM +0800, Yicong Yang wrote:
>> On 2021/4/7 3:54, Wolfram Sang wrote:
>>>
>>>> diff --git a/include/linux/i2c.h b/include/linux/i2c.h
>>>> index 10bd0b0..7268180 100644
>>>> --- a/include/linux/i2c.h
>>>> +++ b/include/linux/i2c.h
>>>> @@ -47,6 +47,26 @@ typedef int (*i2c_slave_cb_t)(struct i2c_client *client,
>>>>  #define I2C_MAX_HIGH_SPEED_MODE_FREQ	3400000
>>>>  #define I2C_MAX_ULTRA_FAST_MODE_FREQ	5000000
>>>>  
>>>> +static inline const char *i2c_freq_mode_string(u32 bus_freq_hz)
>>>> +{
>>>> +	switch (bus_freq_hz) {
>>>> +	case I2C_MAX_STANDARD_MODE_FREQ:
>>>> +		return "Standard Mode (100 kHz)";
>>>> +	case I2C_MAX_FAST_MODE_FREQ:
>>>> +		return "Fast Mode (400 kHz)";
>>>> +	case I2C_MAX_FAST_MODE_PLUS_FREQ:
>>>> +		return "Fast Mode Plus (1.0 MHz)";
>>>> +	case I2C_MAX_TURBO_MODE_FREQ:
>>>> +		return "Turbo Mode (1.4 MHz)";
>>>> +	case I2C_MAX_HIGH_SPEED_MODE_FREQ:
>>>> +		return "High Speed Mode (3.4 MHz)";
>>>> +	case I2C_MAX_ULTRA_FAST_MODE_FREQ:
>>>> +		return "Ultra Fast Mode (5.0 MHz)";
>>>> +	default:
>>>> +		return "Unknown Mode";
>>>> +	}
>>>> +}
>>>
>>> Any reason ehy this is an inline function? My gut feeling says it would
>>> be better added to the core?
>>>
>>
>> it's not a complicated function so i didn't think it'll make much difference,
>> so i just put it in the header along with the coresponding macro definitions.
>> do you want me to move it to the core?
> 
> I guess exporting will save few dozens of bytes if the function is used more
> than once. (All strings will be duplicated or multiplied in that case)
> 

yes, that's one concern. since we don't need this to perform fast, an inline
one maybe unnecessary.

Re: [PATCH v6 2/5] i2c: core: add api to provide frequency mode strings

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 464 bytes --]


> > I guess exporting will save few dozens of bytes if the function is used more
> > than once. (All strings will be duplicated or multiplied in that case)
> > 
> 
> yes, that's one concern. since we don't need this to perform fast, an inline
> one maybe unnecessary.

Exactly. I also don't see an advantage of the function being inline. But
potential disadvantage, even if just small memory overhead. So, I'd
still rather see it as a core function.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 386 bytes --]


> Reason for temp variable is for me it's confusing to see statement like
> "rate_khz = rate_khz / 1000".

Yes. And with this clearer calculation, we can maybe skip the HZ_PER_KHZ
define completely and just use plain '1000' as a factor/divider because
it then becomes obvious. I still find the define more confusing than
helpful TBH. But I'll leave the final decision to Yicong Yang.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Yicong Yang]

On 2021/4/8 7:04, Wolfram Sang wrote:
> 
>> Reason for temp variable is for me it's confusing to see statement like
>> "rate_khz = rate_khz / 1000".
> 
> Yes. And with this clearer calculation, we can maybe skip the HZ_PER_KHZ
> define completely and just use plain '1000' as a factor/divider because
> it then becomes obvious. I still find the define more confusing than
> helpful TBH. But I'll leave the final decision to Yicong Yang.
> 

HZ_PER_KHZ macro are defined separately in other places of the kernel.
Andy suggested to have this defined and used so that one day we can factor
this macro out to the public. :)

Re: [PATCH v6 3/5] i2c: add support for HiSilicon I2C controller

[Andy Shevchenko]

On Thu, Apr 8, 2021 at 10:17 AM Yicong Yang <yangyicong@hisilicon.com> wrote:
> On 2021/4/8 7:04, Wolfram Sang wrote:
> >
> >> Reason for temp variable is for me it's confusing to see statement like
> >> "rate_khz = rate_khz / 1000".
> >
> > Yes. And with this clearer calculation, we can maybe skip the HZ_PER_KHZ
> > define completely and just use plain '1000' as a factor/divider because
> > it then becomes obvious. I still find the define more confusing than
> > helpful TBH. But I'll leave the final decision to Yicong Yang.
> >
>
> HZ_PER_KHZ macro are defined separately in other places of the kernel.
> Andy suggested to have this defined and used so that one day we can factor
> this macro out to the public. :)

Right, and I'm still for the idea to have a macro defined.

-- 
With Best Regards,
Andy Shevchenko