Hi Paul,
On Tue, 25 Mar 2008 15:32:36 +0900, Paul Mundt wrote:
> This adds support for the SMBus adapter found in the various FPGAs on
> the Renesas Highlander platforms. Particularly the R0P7780LC0011RL and
> R0P7785LC0011RL FPGAs.
>
> Functionality is fairly restricted, in that only byte and block data
> transfers are supported. Normal/fast mode and IRQ/polling are also
> supported. Primarily used for various RTCs and thermal sensors.
>
> Signed-off-by: Paul Mundt <[EMAIL PROTECTED]>
>
Review:
> ---
>
> drivers/i2c/busses/Kconfig | 12
> drivers/i2c/busses/Makefile | 1
> drivers/i2c/busses/i2c-highlander.c | 500
> ++++++++++++++++++++++++++++++++++++
> 3 files changed, 513 insertions(+)
>
> diff --git a/drivers/i2c/busses/Kconfig b/drivers/i2c/busses/Kconfig
> index 476b0bb..3376d53 100644
> --- a/drivers/i2c/busses/Kconfig
> +++ b/drivers/i2c/busses/Kconfig
> @@ -146,6 +146,18 @@ config I2C_GPIO
> This is a very simple bitbanging I2C driver utilizing the
> arch-neutral GPIO API to control the SCL and SDA lines.
>
> +config I2C_HIGHLANDER
> + tristate "Highlander FPGA SMBus interface"
> + depends on SH_HIGHLANDER
> + help
> + If you say yes to this option, support will be included for
> + the SMBus interface located in the FPGA on various Highlander
> + boards, particularly the R0P7780LC0011RL and R0P7785LC0011RL
> + FPGAs. This is wholly unrelated to the SoC I2C.
> +
> + This driver can also be built as a module. If so, the module
> + will be called i2c-highlander.
> +
> config I2C_HYDRA
> tristate "CHRP Apple Hydra Mac I/O I2C interface"
> depends on PCI && PPC_CHRP && EXPERIMENTAL
> diff --git a/drivers/i2c/busses/Makefile b/drivers/i2c/busses/Makefile
> index ea7068f..d179bce 100644
> --- a/drivers/i2c/busses/Makefile
> +++ b/drivers/i2c/busses/Makefile
> @@ -14,6 +14,7 @@ obj-$(CONFIG_I2C_BLACKFIN_TWI) += i2c-bfin-twi.o
> obj-$(CONFIG_I2C_DAVINCI) += i2c-davinci.o
> obj-$(CONFIG_I2C_ELEKTOR) += i2c-elektor.o
> obj-$(CONFIG_I2C_GPIO) += i2c-gpio.o
> +obj-$(CONFIG_I2C_HIGHLANDER) += i2c-highlander.o
> obj-$(CONFIG_I2C_HYDRA) += i2c-hydra.o
> obj-$(CONFIG_I2C_I801) += i2c-i801.o
> obj-$(CONFIG_I2C_I810) += i2c-i810.o
> diff --git a/drivers/i2c/busses/i2c-highlander.c
> b/drivers/i2c/busses/i2c-highlander.c
> new file mode 100644
> index 0000000..f6b92d2
> --- /dev/null
> +++ b/drivers/i2c/busses/i2c-highlander.c
> @@ -0,0 +1,500 @@
> +/*
> + * Renesas Solutions Highlander FPGA I2C/SMBus support.
> + *
> + * Supported devices: R0P7780LC0011RL, R0P7785LC0011RL
> + *
> + * Copyright (C) 2008 Paul Mundt
> + * Copyright (C) 2008 Renesas Solutions Corp.
> + * Copyright (C) 2008 Atom Create Engineering Co., Ltd.
> + *
> + * This file is subject to the terms and conditions of the GNU General
> + * Public License version 2. See the file "COPYING" in the main directory
> + * of this archive for more details.
> + */
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/interrupt.h>
> +#include <linux/i2c.h>
> +#include <linux/platform_device.h>
> +#include <linux/completion.h>
> +#include <linux/io.h>
> +#include <linux/delay.h>
> +
> +#define SMCR 0x00
> +#define SMCR_START (1 << 0)
> +#define SMCR_IRIC (1 << 1)
> +#define SMCR_BBSY (1 << 2)
> +#define SMCR_ACKE (1 << 3)
> +#define SMCR_RST (1 << 4)
> +#define SMCR_IEIC (1 << 6)
> +
> +#define SMSMADR 0x02
> +
> +#define SMMR 0x04
> +#define SMMR_MODE0 (1 << 0)
> +#define SMMR_MODE1 (1 << 1)
> +#define SMMR_CAP (1 << 3)
> +#define SMMR_TMMD (1 << 4)
> +#define SMMR_SP (1 << 7)
> +
> +#define SMSADR 0x06
> +#define SMTRDR 0x46
> +
> +struct highlander_i2c_dev {
> + struct device *dev;
> + void __iomem *base;
> + struct i2c_adapter adapter;
> + struct completion cmd_complete;
> + int irq;
> + u8 *buf;
> + size_t buf_len;
Your driver is a bit inconsistent with these two last struct members.
You only access them in highlander_i2c_smbus_xfer(), the rest of the
time you pass them to functions as extra parameters (while you are
_also_ passing a pointer to the struct highlander_i2c_dev.) Please
choose a strategy and stick to it: either use these struct members to
carry the values around, or drop these members and use local variables
instead. Mixing both is inefficient and confusing.
> +};
> +
> +static int iic_force_poll, iic_force_normal, iic_timeout = 1000;
> +
> +static inline void highlander_i2c_irq_enable(struct highlander_i2c_dev *dev)
> +{
> + iowrite16(ioread16(dev->base + SMCR) | SMCR_IEIC, dev->base + SMCR);
> +}
> +
> +static inline void highlander_i2c_irq_disable(struct highlander_i2c_dev *dev)
> +{
> + iowrite16(ioread16(dev->base + SMCR) & ~SMCR_IEIC, dev->base + SMCR);
> +}
> +
> +static inline void highlander_i2c_start(struct highlander_i2c_dev *dev)
> +{
> + iowrite16(ioread16(dev->base + SMCR) | SMCR_START, dev->base + SMCR);
> +}
> +
> +static inline void highlander_i2c_done(struct highlander_i2c_dev *dev)
> +{
> + iowrite16(ioread16(dev->base + SMCR) | SMCR_IRIC, dev->base + SMCR);
> +}
> +
> +static inline void highlander_i2c_setup(struct highlander_i2c_dev *dev)
There's no point in forcing this function to be inlined, it's definitely
not performance critical.
> +{
> + u16 smmr;
> +
> + smmr = ioread16(dev->base + SMMR);
> + smmr |= SMMR_TMMD;
> +
> + if (iic_force_normal)
> + smmr &= ~SMMR_SP;
> + else
> + smmr |= SMMR_SP;
> +
> + iowrite16(smmr, dev->base + SMMR);
> +}
> +
A comment about why the following 2 functions are needed and/or what
they are doing would be nice to have. Also, the order of the parameters
is rather error-prone, most C functions copying things from one buffer
to another have the destination as first parameter, for example memcpy.
These functions seem to only support even values for len (except for
1), this should be mentioned.
> +static void smbus_write_data(u8 *src, u16 *dst, int len)
> +{
> + int i, j;
> +
> + if (len == 1)
> + *dst = *src << 8;
> + else {
> + j = 0;
> + for (i = 0; i < len; i += 2) {
> + *(dst + j) = *(src + i) << 8 | *(src + i + 1);
> + j++;
> + }
> + }
> +}
> +
> +static void smbus_read_data(u16 *src, u8 *dst, int len)
> +{
> + int i, j;
> +
> + if (len == 1)
> + *dst = *src >> 8;
> + else {
> + j = 0;
> + for (i = 0; i < len; i += 2) {
> + *(dst + i) = *(src + j) >> 8;
> + *(dst + i + 1) = *(src + j) & 0x00ff;
> + j++;
> + }
> + }
> +}
If I read the code above correctly, you are merely converting 16-bit
words from cpu-endian to long-endian and back, so using be16_to_cpu and
cpu_to_be16 would perform better. If the Highlander is big-endian, you
should be able to let the compiler optimize out most of the code.
> +
> +static void highlander_i2c_command(struct highlander_i2c_dev *dev, u8
> command, int len)
> +{
> + u16 cmd[32];
> + int i, j;
> +
> + j = 0;
> + if (len == 1)
> + cmd[j++] = (command << 8);
> + else
> + for (i = 0; i < len; i += 2)
> + cmd[j++] = (command << 8) | command;
> +
> + for (i = 0; i < j; i++) {
> + iowrite16(cmd[i], dev->base + SMSADR + (i * sizeof(u16)));
> + dev_dbg(dev->dev, "command data[%x] 0x%04x\n", i, cmd[i]);
> + }
> +}
> +
> +static int highlander_i2c_wait_for_bbsy(struct highlander_i2c_dev *dev)
> +{
> + unsigned long timeout;
> +
> + timeout = jiffies + msecs_to_jiffies(iic_timeout);
> + while (ioread16(dev->base + SMCR) & SMCR_BBSY) {
> + if (time_after(jiffies, timeout)) {
> + dev_warn(dev->dev, "timeout waiting for bus ready\n");
> + return -ETIMEDOUT;
> + }
> +
> + msleep(1);
> + }
> +
> + return 0;
> +}
> +
> +static int highlander_i2c_reset(struct highlander_i2c_dev *dev)
> +{
> + iowrite16(ioread16(dev->base + SMCR) | SMCR_RST, dev->base + SMCR);
> + return highlander_i2c_wait_for_bbsy(dev);
> +}
> +
> +static int highlander_i2c_wait_for_ack(struct highlander_i2c_dev *dev)
> +{
> + u16 tmp = ioread16(dev->base + SMCR);
> +
> + if ((tmp & (SMCR_IRIC | SMCR_ACKE)) == SMCR_ACKE) {
> + dev_warn(dev->dev, "ack abnormality\n");
> + return highlander_i2c_reset(dev);
> + }
> +
> + return 0;
> +}
> +
> +static irqreturn_t highlander_i2c_irq(int irq, void *dev_id)
> +{
> + struct highlander_i2c_dev *dev = dev_id;
> +
> + highlander_i2c_done(dev);
> + complete(&dev->cmd_complete);
> +
> + return IRQ_HANDLED;
> +}
> +
> +static void highlander_i2c_poll(struct highlander_i2c_dev *dev)
> +{
> + unsigned long timeout;
> + u16 smcr;
> +
> + timeout = jiffies + msecs_to_jiffies(iic_timeout);
> + for (;;) {
> + smcr = ioread16(dev->base + SMCR);
> +
> + /*
> + * Don't bother checking ACKE here, this and the reset
> + * are handled in highlander_i2c_wait_xfer_done() when
> + * waiting for the ACK.
> + */
> +
> + if (smcr & SMCR_IRIC)
> + return;
> + if (time_after(jiffies, timeout))
> + break;
> +
> + cpu_relax();
> + cond_resched();
> + }
> +
> + dev_err(dev->dev, "polling timed out\n");
> +}
> +
> +static inline int highlander_i2c_wait_xfer_done(struct highlander_i2c_dev
> *dev)
> +{
> + if (dev->irq)
> + wait_for_completion_interruptible_timeout(&dev->cmd_complete,
> + msecs_to_jiffies(iic_timeout));
> + else
> + /* busy looping, the IRQ of champions */
> + highlander_i2c_poll(dev);
> +
> + return highlander_i2c_wait_for_ack(dev);
> +}
> +
> +static int highlander_i2c_read(struct highlander_i2c_dev *dev,
> + u8 *buf, unsigned int len)
> +{
> + int i, cnt;
> + u16 data[16];
> +
> + if (highlander_i2c_wait_for_bbsy(dev)) {
> + dev_warn(dev->dev, "timed out\n");
highlander_i2c_wait_for_bbsy already prints a warning message.
> + return -EAGAIN;
> + }
> +
> + highlander_i2c_start(dev);
> +
> + if (highlander_i2c_wait_xfer_done(dev)) {
> + dev_err(dev->dev, "Arbitration loss\n");
> + return -EIO;
I think you should return -EAGAIN here as well... On arbitration loss
the caller will certainly want to retry later, same as if the bus was
busy?
> + }
> +
> + cnt = (len + 1) >> 1;
> + for (i = 0; i < cnt; i++) {
> + data[i] = ioread16(dev->base + SMTRDR + (i * sizeof(u16)));
> + dev_dbg(dev->dev, "read data[%x] 0x%04x\n", i, data[i]);
> + }
> +
> + smbus_read_data(data, buf, len);
> +
> + /*
> + * The R0P7780LC0011RL FPGA on the SH7780-based Highlanders
> + * needs a significant delay in the read path. SH7785 Highlanders
> + * don't have this issue, so restrict it entirely to those.
> + */
> + if (mach_is_r7780rp() || mach_is_r7780mp())
> + mdelay(1000);
A one second busy-wait is nasty. Can't you use msleep here instead?
Having to wait for 1 second after each read makes this driver pretty
unusable on these machines anyway, doesn't it? :(
> +
> + return 0;
> +}
> +
> +static int highlander_i2c_write(struct highlander_i2c_dev *dev,
> + u8 *buf, unsigned int len)
> +{
> + int i, cnt;
> + u16 data[16];
> +
> + smbus_write_data(buf, data, len);
> +
> + cnt = (len + 1) >> 1;
> + for (i = 0; i < cnt; i++) {
> + iowrite16(data[i], dev->base + SMTRDR + (i * sizeof(u16)));
> + dev_dbg(dev->dev, "write data[%x] 0x%04x\n", i, data[i]);
> + }
> +
> + if (highlander_i2c_wait_for_bbsy(dev)) {
> + dev_warn(dev->dev, "timed out\n");
highlander_i2c_wait_for_bbsy already prints a warning message.
> + return -EAGAIN;
> + }
> +
> + highlander_i2c_start(dev);
> +
> + return highlander_i2c_wait_xfer_done(dev);
> +}
> +
> +static int highlander_i2c_smbus_xfer(struct i2c_adapter *adap, u16 addr,
> + unsigned short flags, char read_write,
> + u8 command, int size,
> + union i2c_smbus_data *data)
> +{
> + struct highlander_i2c_dev *dev = i2c_get_adapdata(adap);
> + int read = read_write & I2C_SMBUS_READ;
> + u16 tmp;
> +
> + init_completion(&dev->cmd_complete);
> +
> + dev_dbg(dev->dev, "addr %04x, command %02x, read_write %d, size %d\n",
> + addr, command, read_write, size);
> +
> + /* Defaults */
> + dev->buf = NULL;
> + dev->buf_len = 0;
These initializations shouldn't be needed.
> +
> + /*
> + * Set up the buffer and transfer size
> + */
> + switch (size) {
> + case I2C_SMBUS_BYTE_DATA:
> + if (data)
> + dev->buf = &data->byte;
It is not valid for data to be null for I2C_SMBUS_BYTE_DATA, so this
test isn't needed.
> + dev->buf_len = 1;
> + break;
> + case I2C_SMBUS_BLOCK_DATA:
> + case I2C_SMBUS_I2C_BLOCK_DATA:
> + dev->buf = &data->block[1];
> + dev->buf_len = data->block[0];
> + break;
That's definitely not correct. Given that you don't differentiate
between I2C_SMBUS_BLOCK_DATA and I2C_SMBUS_I2C_BLOCK_DATA anywhere in
the code, I take it that the device implements either I2C block
transactions or SMBus block transactions but not both. Which one is it?
> + default:
> + dev_err(dev->dev, "bogus command %d\n", size);
> + return -EINVAL;
> + }
> +
> + /*
> + * Encode the mode setting
> + */
> + tmp = ioread16(dev->base + SMMR);
> + tmp &= ~(SMMR_MODE0 | SMMR_MODE1);
> +
> + switch (dev->buf_len) {
> + case 1:
> + /* default */
> + break;
> + case 8:
> + tmp |= SMMR_MODE0;
> + break;
> + case 16:
> + tmp |= SMMR_MODE1;
> + break;
> + case 32:
> + tmp |= (SMMR_MODE0 | SMMR_MODE1);
> + break;
> + default:
> + dev_err(dev->dev, "bogus xfer size %d\n", dev->buf_len);
> + return -EINVAL;
> + }
Wait, do you mean that this device only supports block sizes of 8, 16
and 32. Oh my. At least please change "bogus" to "unsupported" in the
message above - if something is bogus here, that's the hardware.
> +
> + iowrite16(tmp, dev->base + SMMR);
> +
> + /* Ensure we're in a sane state */
> + highlander_i2c_done(dev);
> +
> + /* Set slave address */
> + iowrite16(((addr & 0x7f) << 1) | read, dev->base + SMSMADR);
Masking is not needed, addr is a 7-bit value to start with.
> +
> + highlander_i2c_command(dev, command, dev->buf_len);
> +
> + if (read)
> + return highlander_i2c_read(dev, dev->buf, dev->buf_len);
> + else
> + return highlander_i2c_write(dev, dev->buf, dev->buf_len);
> +}
> +
> +static u32 highlander_i2c_func(struct i2c_adapter *adapter)
> +{
> + return I2C_FUNC_SMBUS_BYTE_DATA | I2C_FUNC_SMBUS_BLOCK_DATA;
Doesn't match the cases handled in highlander_i2c_smbus_xfer.
> +}
> +
> +static const struct i2c_algorithm highlander_i2c_algo = {
> + .smbus_xfer = highlander_i2c_smbus_xfer,
> + .functionality = highlander_i2c_func,
> +};
> +
> +static int highlander_i2c_probe(struct platform_device *pdev)
> +{
> + struct highlander_i2c_dev *dev;
> + struct i2c_adapter *adap;
> + struct resource *res;
> + int ret;
> +
> + res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + if (unlikely(!res)) {
> + dev_err(&pdev->dev, "no mem resource\n");
> + return -ENODEV;
> + }
> +
> + dev = kzalloc(sizeof(struct highlander_i2c_dev), GFP_KERNEL);
> + if (unlikely(!dev))
> + return -ENOMEM;
> +
> + dev->base = ioremap_nocache(res->start, res->end - res ->start + 1);
No spaces around ->.
> + if (unlikely(!dev->base)) {
> + ret = -ENXIO;
> + goto err;
> + }
> +
> + dev->dev = &pdev->dev;
> + platform_set_drvdata(pdev, dev);
> +
> + dev->irq = platform_get_irq(pdev, 0);
> + if (dev->irq && !iic_force_poll) {
> + ret = request_irq(dev->irq, highlander_i2c_irq, IRQF_DISABLED,
> + pdev->name, dev);
> + if (unlikely(ret))
> + goto err_unmap;
> +
> + highlander_i2c_irq_enable(dev);
> + } else {
> + dev_notice(&pdev->dev, "no IRQ, using polling mode\n");
> + highlander_i2c_irq_disable(dev);
> + }
> +
> + highlander_i2c_setup(dev);
> +
> + adap = &dev->adapter;
> + i2c_set_adapdata(adap, dev);
> + adap->owner = THIS_MODULE;
> + adap->class = I2C_CLASS_HWMON;
> + strncpy(adap->name, "HL FPGA I2C adapter", sizeof(adap->name));
strncpy is broken, please use strlcpy instead.
> + adap->algo = &highlander_i2c_algo;
> + adap->dev.parent = &pdev->dev;
> +
> + adap->nr = pdev->id;
> + ret = i2c_add_numbered_adapter(adap);
> + if (unlikely(ret)) {
> + dev_err(&pdev->dev, "failure adding adapter\n");
> + goto err_free_irq;
> + }
> +
> + /*
> + * Reset the adapter
> + */
> + ret = highlander_i2c_reset(dev);
> + if (unlikely(ret)) {
> + dev_err(&pdev->dev, "controller didn't come up\n");
> + goto err_free_irq;
The error path is broken here, you return with an i2c adapter
registered but all its resources freed.
In fact, you should reset the adapter before registering it, otherwise
there's a race where someone could use the adapter before it's
operational.
> + }
> +
> + return 0;
> +
> +err_free_irq:
> + free_irq(dev->irq, dev);
> +err_unmap:
> + iounmap(dev->base);
> +err:
> + kfree(dev);
> +
> + platform_set_drvdata(pdev, NULL);
> +
> + return ret;
> +}
> +
> +static int highlander_i2c_remove(struct platform_device *pdev)
> +{
> + struct highlander_i2c_dev *dev = platform_get_drvdata(pdev);
> +
> + i2c_del_adapter(&dev->adapter);
> +
> + if (dev->irq)
> + free_irq(dev->irq, dev);
Here you test for dev->irq before calling free_irq(), but in the error
path above you didn't?
> +
> + iounmap(dev->base);
> + kfree(dev);
> +
> + platform_set_drvdata(pdev, NULL);
> +
> + return 0;
> +}
> +
> +static struct platform_driver highlander_i2c_driver = {
> + .driver = {
> + .name = "i2c-highlander",
> + .owner = THIS_MODULE,
> + },
> +
> + .probe = highlander_i2c_probe,
> + .remove = highlander_i2c_remove,
I guess that the remove function could be made __devexit or even __exit?
> +};
> +
> +static int __init highlander_i2c_init(void)
> +{
> + return platform_driver_register(&highlander_i2c_driver);
> +}
> +
> +static void __exit highlander_i2c_exit(void)
> +{
> + platform_driver_unregister(&highlander_i2c_driver);
> +}
> +
> +module_init(highlander_i2c_init);
> +module_exit(highlander_i2c_exit);
> +
> +MODULE_AUTHOR("Paul Mundt");
> +MODULE_DESCRIPTION("Renesas Highlander FPGA I2C/SMBus adapter");
> +MODULE_LICENSE("GPL v2");
> +
> +module_param(iic_force_poll, bool, 0);
> +module_param(iic_force_normal, bool, 0);
> +module_param(iic_timeout, int, 0);
> +
> +MODULE_PARM_DESC(iic_force_poll, "Force polling mode");
> +MODULE_PARM_DESC(iic_force_normal, "Force normal mode (100 kHz), default is
> fast mode (400 kHz)");
> +MODULE_PARM_DESC(iic_timeout, "Force timeout value in msecs (default 1000
> ms)");
This is just setting the timeout value, not "forcing" it.
--
Jean Delvare
_______________________________________________
i2c mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/i2c
