i2c驱动架构（史上最全） davinc dm368 i2c驱动分析 .

预备知识

在阅读本文最好先熟悉一种i2c设备的驱动程序，并且浏览一下i2c-core.c以及芯片提供商的提供的i2c总线驱动（i2c-davinci.c）。标题党请见谅！

其实i2c接口非常的简单，即使用51单片的gpio来模拟i2c，编写一个e2prom或者其他i2c接口的驱动程序，也不是什么难事，几百行代码就能搞定。

但是Linux的i2c驱动体系结构却有相当的复杂度，不管是叫linux i2c驱动还是单片机i2c驱动，其根本还是操作soc芯片内部的i2c模块（也叫i2c adapter）（读写i2c相关的寄存器）来产生start、stop还有ack信号而已。

linux设备驱动到底复杂在什么地方？

假设soc芯片dm368有两个i2c adapter（368内部真正只有一个i2c模块）：i2c_adapter1，i2c_adapter1；然后外部有三个i2c接口的设备i2c_device1，i2c_device2，i2c_device3。

现在要求在裸机下写出他们的驱动函数。那么肯定要写出6个不同的驱动函数：

i2c_adapter1_ReadWrite_i2c_device1();
i2c_adapter1_ReadWrite_i2c_device2()
i2c_adapter1_ReadWrite_i2c_device3()
i2c_adapter2_ReadWrite_i2c_device1()
i2c_adapter2_ReadWrite_i2c_device2()
i2c_adapter2_ReadWrite_i2c_device3()

i2c_adapter1_ReadWrite_i2c_device1();
i2c_adapter1_ReadWrite_i2c_device2()
i2c_adapter1_ReadWrite_i2c_device3()
i2c_adapter2_ReadWrite_i2c_device1()
i2c_adapter2_ReadWrite_i2c_device2()
i2c_adapter2_ReadWrite_i2c_device3()

设想一共有m个i2c adapter和n个外设i2c device，那么将需要m*n个驱动。并且这m*n个驱动程序必要会有很大部分重复的代码，而且不利于驱动程序的移植。

如果采用adapter和device分离的思想来写这样的驱动会是怎样呢？

图1

这样分离之后，只需要m+n个驱动，而且Adapter和Device的几乎没有耦合性，增加一个Adapter或者device并不会影响其余的驱动。

这就是分离思想带来的好处。除此之外，linux虽然是C写的，但是大量使用了面向对象的变成方法（可以理解为分层的思想），仅仅分离细想和分层思想的引入，就大大增加了linux设备驱动的复杂度。

linux驱动中 i2c驱动架构

图2

上图完整的描述了linux i2c驱动架构，虽然I2C硬件体系结构比较简单，但是i2c体系结构在linux中的实现却相当复杂。那么我们如何编写特定i2c接口器件（比如，ov2715，需要i2c来配置寄存器）的驱动程序？就是说上述架构中的那些部分需要我们完成，而哪些是linux内核已经完善的或者是芯片提供商（TI davinci平台已经做好的）已经提供的？

架构层次分类

第一层：提供i2c adapter的硬件驱动，探测、初始化i2c adapter（如申请i2c的io地址和中断号），驱动soc控制的i2c adapter在硬件上产生信号（start、stop、ack）以及处理i2c中断。覆盖图中的硬件实现层

第二层：提供i2c adapter的algorithm，用具体适配器的xxx_xferf()函数来填充i2c_algorithm的master_xfer函数指针，并把赋值后的i2c_algorithm再赋值给i2c_adapter的algo指针。覆盖图中的访问抽象层、i2c核心层

第三层：实现i2c设备驱动中的i2c_driver接口，用具体的i2c device设备的attach_adapter()、detach_adapter()方法赋值给i2c_driver的成员函数指针。实现设备device与总线（或者叫adapter）的挂接。覆盖图中的driver驱动层

第四层：实现i2c设备所对应的具体device的驱动，i2c_driver只是实现设备与总线的挂接，而挂接在总线上的设备则是千差万别的，eeprom和ov2715显然不是同一类的device，所以要实现具体设备device的write()、read()、ioctl()等方法，赋值给file_operations，然后注册字符设备（多数是字符设备）。覆盖图中的driver驱动层

第一层和第二层又叫i2c总线驱动(bus)，第三第四属于i2c设备驱动(device driver)。在linux驱动架构中，几乎不需要驱动开发人员再添加bus，因为linux内核几乎集成所有总线bus，如usb、pci、i2c等等。并且总线bus中的【与特定硬件相关的代码】已由芯片提供商编写完成，例如TI davinci平台i2c总线bus与硬件相关的代码在内核目录/drivers/i2c/buses下的i2c-davinci.c源文件中；而三星的s3c-2440平台i2c总线bus为/drivers/i2c/buses/i2c-s3c2410.c

第三第四层又叫设备驱动层与特定device相干的就需要驱动工程师来实现了。

明确了方向后，再来具体分析。

具体分析

i2c_adapter与i2c_client的关系与i2c硬件体系中设配器与设备的关系一致，即i2c_client依附于i2c_adapter，由于一个适配器上可以连接多个i2c设备device，所以相应的，i2c_adapter也可以被多个i2c_client依附，在i2c_adapter中包含i2c_client的链表。同一类的i2c设备device对应一个驱动driver。driver与device的关系是一对多的关系。

现在，我们就来看一下这几个重要的结构体，分别是i2c_driver i2c_client i2c_adapter，也可以先忽略他们，待会回过头来看会更容易理解

1、i2c_driver

struct i2c_driver {
    int id;
    unsigned int class;

    int (*attach_adapter)(struct i2c_adapter *);
    int (*detach_adapter)(struct i2c_adapter *);

    int (*detach_client)(struct i2c_client *);

    int (*command)(struct i2c_client *client,unsigned int cmd, void *arg);
    struct device_driver driver;
    struct list_head list;
};

struct i2c_driver {
	int id;
	unsigned int class;

	int (*attach_adapter)(struct i2c_adapter *);
	int (*detach_adapter)(struct i2c_adapter *);

	int (*detach_client)(struct i2c_client *);

	int (*command)(struct i2c_client *client,unsigned int cmd, void *arg);
	struct device_driver driver;
	struct list_head list;
};

2、i2c_client

struct i2c_client {
    unsigned int flags;     /* div., see below      */
    unsigned short addr;        /* chip address - NOTE: 7bit    */
                    /* addresses are stored in the  */
                    /* _LOWER_ 7 bits       */
    struct i2c_adapter *adapter;    /* the adapter we sit on    */
    struct i2c_driver *driver;  /* and our access routines  */
    int usage_count;        /* How many accesses currently  */
                    /* to the client        */
    struct device dev;      /* the device structure     */
    struct list_head list;
    char name[I2C_NAME_SIZE];
    struct completion released;
};

struct i2c_client {
	unsigned int flags;		/* div., see below		*/
	unsigned short addr;		/* chip address - NOTE: 7bit 	*/
					/* addresses are stored in the	*/
					/* _LOWER_ 7 bits		*/
	struct i2c_adapter *adapter;	/* the adapter we sit on	*/
	struct i2c_driver *driver;	/* and our access routines	*/
	int usage_count;		/* How many accesses currently  */
					/* to the client		*/
	struct device dev;		/* the device structure		*/
	struct list_head list;
	char name[I2C_NAME_SIZE];
	struct completion released;
};

3、i2c_adapter

struct i2c_adapter {
    struct module *owner;
    unsigned int id;
    unsigned int class;
    struct i2c_algorithm *algo;/* the algorithm to access the bus   */
    void *algo_data;

    /* --- administration stuff. */
    int (*client_register)(struct i2c_client *);
    int (*client_unregister)(struct i2c_client *);

    /* data fields that are valid for all devices   */
    struct mutex bus_lock;
    struct mutex clist_lock;

    int timeout;
    int retries;
    struct device dev;      /* the adapter device */
    struct class_device class_dev;  /* the class device */

    int nr;
    struct list_head clients;
    struct list_head list;
    char name[I2C_NAME_SIZE];
    struct completion dev_released;
    struct completion class_dev_released;
};

struct i2c_adapter {
	struct module *owner;
	unsigned int id;
	unsigned int class;
	struct i2c_algorithm *algo;/* the algorithm to access the bus	*/
	void *algo_data;

	/* --- administration stuff. */
	int (*client_register)(struct i2c_client *);
	int (*client_unregister)(struct i2c_client *);

	/* data fields that are valid for all devices	*/
	struct mutex bus_lock;
	struct mutex clist_lock;

	int timeout;
	int retries;
	struct device dev;		/* the adapter device */
	struct class_device class_dev;	/* the class device */

	int nr;
	struct list_head clients;
	struct list_head list;
	char name[I2C_NAME_SIZE];
	struct completion dev_released;
	struct completion class_dev_released;
};

4、i2c_algorithm

struct i2c_algorithm {
    int (*master_xfer)(struct i2c_adapter *adap,struct i2c_msg *msgs,
                       int num);
    int (*slave_send)(struct i2c_adapter *,char*,int);
    int (*slave_recv)(struct i2c_adapter *,char*,int);
    u32 (*functionality) (struct i2c_adapter *);
};

struct i2c_algorithm {
	int (*master_xfer)(struct i2c_adapter *adap,struct i2c_msg *msgs, 
	                   int num);
	int (*slave_send)(struct i2c_adapter *,char*,int);
	int (*slave_recv)(struct i2c_adapter *,char*,int);
	u32 (*functionality) (struct i2c_adapter *);
};

【i2c_adapter与i2c_algorithm】

i2c_adapter对应与物理上的一个适配器，而i2c_algorithm对应一套通信方法，一个i2c适配器需要i2c_algorithm中提供的（i2c_algorithm中的又是更下层与硬件相关的代码提供）通信函数来控制适配器上产生特定的访问周期。缺少i2c_algorithm的i2c_adapter什么也做不了，因此i2c_adapter中包含其使用i2c_algorithm的指针。

i2c_algorithm中的关键函数master_xfer()用于产生i2c访问周期需要的start stop ack信号，以i2c_msg（即i2c消息）为单位发送和接收通信数据。i2c_msg也非常关键，调用驱动中的发送接收函数需要填充该结构体

/*
 * I2C Message - used for pure i2c transaction, also from /dev interface
 */
struct i2c_msg {
    __u16 addr; /* slave address            */
    __u16 flags;
    __u16 len;      /* msg length               */
    __u8 *buf;      /* pointer to msg data          */
};

/*
 * I2C Message - used for pure i2c transaction, also from /dev interface
 */
struct i2c_msg {
	__u16 addr;	/* slave address			*/
 	__u16 flags;		
 	__u16 len;		/* msg length				*/
 	__u8 *buf;		/* pointer to msg data			*/
};

【i2c_driver和i2c_client】

i2c_driver对应一套驱动方法，其主要函数是attach_adapter()和detach_client()，i2c_client对应真实的i2c物理设备device，每个i2c设备都需要一个i2c_client来描述，i2c_driver与i2c_client的关系是一对多。一个i2c_driver上可以支持多个同等类型的i2c_client.

【i2c_adapter和i2c_client】

i2c_adapter和i2c_client的关系与i2c硬件体系中适配器和设备的关系一致，即i2c_client依附于i2c_adapter,由于一个适配器上可以连接多个i2c设备，所以i2c_adapter中包含依附于它的i2c_client的链表。

从图1图2中都可以看出，linux内核对i2c架构抽象了一个叫核心层core的中间件，它分离了设备驱动device driver和硬件控制的实现细节（如操作i2c的寄存器），core层不但为上面的设备驱动提供封装后的内核注册函数，而且还为小面的硬件时间提供注册接口（也就是i2c总线注册接口），可以说core层起到了承上启下的作用。

我们先看一下i2c-core为外部提供的核心函数（选取部分），i2c-core对应的源文件为i2c-core.c，位于内核目录/driver/i2c/i2c-core.c

EXPORT_SYMBOL(i2c_add_adapter);
EXPORT_SYMBOL(i2c_del_adapter);
EXPORT_SYMBOL(i2c_del_driver);
EXPORT_SYMBOL(i2c_attach_client);
EXPORT_SYMBOL(i2c_detach_client);

EXPORT_SYMBOL(i2c_transfer);

EXPORT_SYMBOL(i2c_add_adapter);
EXPORT_SYMBOL(i2c_del_adapter);
EXPORT_SYMBOL(i2c_del_driver);
EXPORT_SYMBOL(i2c_attach_client);
EXPORT_SYMBOL(i2c_detach_client);

EXPORT_SYMBOL(i2c_transfer);

如果看过i2c设备驱动程序的人一定对上面几个函数比较熟悉。

i2c_transfer()函数，i2c_transfer()函数本身并不具备驱动适配器物理硬件完成消息交互的能力，它只是寻找到i2c_adapter对应的i2c_algorithm，并使用i2c_algorithm的master_xfer()函数真正的驱动硬件流程，代码清单如下，不重要的已删除。

int i2c_transfer(struct i2c_adapter * adap, struct i2c_msg *msgs, int num)
{
    int ret;
    if (adap->algo->master_xfer) {//如果master_xfer函数存在，则调用，否则返回错误
        ret = adap->algo->master_xfer(adap,msgs,num);//这个函数在硬件相关的代码中给algorithm赋值
        return ret;
    } else {
        return -ENOSYS;
    }
}

int i2c_transfer(struct i2c_adapter * adap, struct i2c_msg *msgs, int num)
{
	int ret;
	if (adap->algo->master_xfer) {//如果master_xfer函数存在，则调用，否则返回错误
		ret = adap->algo->master_xfer(adap,msgs,num);//这个函数在硬件相关的代码中给algorithm赋值
		return ret;
	} else {
		return -ENOSYS;
	}
}

当一个具体的client被侦测到并被关联的时候，设备和sysfs文件将被注册。相反的，在client被取消关联的时候，sysfs文件和设备也被注销，驱动开发人员需开发i2c设备驱动时，需要调用下列函数。程序清单如下

int i2c_attach_client(struct i2c_client *client)
{
    ...
    device_register(&client->dev);
    device_create_file(&client->dev, &dev_attr_client_name);
    ...
    return 0;
}

int i2c_attach_client(struct i2c_client *client)
{
	...
	device_register(&client->dev);
	device_create_file(&client->dev, &dev_attr_client_name);
	...
	return 0;
}

int i2c_detach_client(struct i2c_client *client)
{
    ...
    device_remove_file(&client->dev, &dev_attr_client_name);
    device_unregister(&client->dev);
    ...
    return res;
}

int i2c_detach_client(struct i2c_client *client)
{
	...
	device_remove_file(&client->dev, &dev_attr_client_name);
	device_unregister(&client->dev);
	...
	return res;
}

i2c_add_adapter()函数和i2c_del_adapter()在i2c-davinci.c中有调用，稍后分析

/* -----
 * i2c_add_adapter is called from within the algorithm layer,
 * when a new hw adapter registers. A new device is register to be
 * available for clients.
 */
int i2c_add_adapter(struct i2c_adapter *adap)
{
    ...
    device_register(&adap->dev);
    device_create_file(&adap->dev, &dev_attr_name);
    ...
    /* inform drivers of new adapters */
    list_for_each(item,&drivers) {
        driver = list_entry(item, struct i2c_driver, list);
        if (driver->attach_adapter)
            /* We ignore the return code; if it fails, too bad */
            driver->attach_adapter(adap);
    }
    ...
}

/* -----
 * i2c_add_adapter is called from within the algorithm layer,
 * when a new hw adapter registers. A new device is register to be
 * available for clients.
 */
int i2c_add_adapter(struct i2c_adapter *adap)
{
	...
	device_register(&adap->dev);
	device_create_file(&adap->dev, &dev_attr_name);
	...
	/* inform drivers of new adapters */
	list_for_each(item,&drivers) {
		driver = list_entry(item, struct i2c_driver, list);
		if (driver->attach_adapter)
			/* We ignore the return code; if it fails, too bad */
			driver->attach_adapter(adap);
	}
	...
}

int i2c_del_adapter(struct i2c_adapter *adap)
{
    ...
    list_for_each(item,&drivers) {
        driver = list_entry(item, struct i2c_driver, list);
        if (driver->detach_adapter)
            if ((res = driver->detach_adapter(adap))) {
            }
    }
    ...
    list_for_each_safe(item, _n, &adap->clients) {
        client = list_entry(item, struct i2c_client, list);

        if ((res=client->driver->detach_client(client))) {

        }
    }
    ...
    device_remove_file(&adap->dev, &dev_attr_name);
    device_unregister(&adap->dev);

}

int i2c_del_adapter(struct i2c_adapter *adap)
{
	...
	list_for_each(item,&drivers) {
		driver = list_entry(item, struct i2c_driver, list);
		if (driver->detach_adapter)
			if ((res = driver->detach_adapter(adap))) {
			}
	}
	...
	list_for_each_safe(item, _n, &adap->clients) {
		client = list_entry(item, struct i2c_client, list);

		if ((res=client->driver->detach_client(client))) {

		}
	}
	...
	device_remove_file(&adap->dev, &dev_attr_name);
	device_unregister(&adap->dev);

}

i2c-davinci.c是实现与硬件相关功能的代码集合，这部分是与平台相关的，也叫做i2c总线驱动，这部分代码是这样添加到系统中的

static struct platform_driver davinci_i2c_driver = {
    .probe      = davinci_i2c_probe,
    .remove     = davinci_i2c_remove,
    .driver     = {
        .name   = "i2c_davinci",
        .owner  = THIS_MODULE,
    },
};

/* I2C may be needed to bring up other drivers */
static int __init davinci_i2c_init_driver(void)
{
    return platform_driver_register(&davinci_i2c_driver);
}
subsys_initcall(davinci_i2c_init_driver);

static void __exit davinci_i2c_exit_driver(void)
{
    platform_driver_unregister(&davinci_i2c_driver);
}
module_exit(davinci_i2c_exit_driver);

static struct platform_driver davinci_i2c_driver = {
	.probe		= davinci_i2c_probe,
	.remove		= davinci_i2c_remove,
	.driver		= {
		.name	= "i2c_davinci",
		.owner	= THIS_MODULE,
	},
};

/* I2C may be needed to bring up other drivers */
static int __init davinci_i2c_init_driver(void)
{
	return platform_driver_register(&davinci_i2c_driver);
}
subsys_initcall(davinci_i2c_init_driver);

static void __exit davinci_i2c_exit_driver(void)
{
	platform_driver_unregister(&davinci_i2c_driver);
}
module_exit(davinci_i2c_exit_driver);

并且，i2c适配器控制硬件发送接收数据的函数在这里赋值给i2c-algorithm，i2c_davinci_xfer稍加修改就可以在裸机中控制i2c适配器

static struct i2c_algorithm i2c_davinci_algo = {
    .master_xfer    = i2c_davinci_xfer,
    .functionality  = i2c_davinci_func,
};

static struct i2c_algorithm i2c_davinci_algo = {
	.master_xfer	= i2c_davinci_xfer,
	.functionality	= i2c_davinci_func,
};

然后在davinci_i2c_probe函数中，将i2c_davinci_algo添加到添加到algorithm系统中

adap->algo = &i2c_davinci_algo;

adap->algo = &i2c_davinci_algo;

梳理图

有时候代码比任何文字描述都来得直接，但是过多的代码展示反而让人觉得枯燥。这个时候，需要一幅图来梳理一下上面的内容，请看图3。

图3

好了，上面这些代码的展示是告诉我们，linux内核和芯片提供商为我们的的驱动程序提供了 i2c驱动的框架，以及框架底层与硬件相关的代码的实现。剩下的就是针对挂载在i2c两线上的i2c设备了device，如at24c02，例如ov2715，而编写的具体设备驱动了，这里的设备就是硬件接口外挂载的设备，而非硬件接口本身（soc硬件接口本身的驱动可以理解为总线驱动）。

在理解了i2c驱动架构后，我们接下来再作两方面的分析工作：一是具体的i2c设备ov2715驱动源码分析，二是davinci平台的i2c总线驱动源码。

ov2715设备i2c驱动源码分析

ov2715为200万的CMOS Sensor，芯片的寄存器控制通过i2c接口完成，i2c设备地址为0x6c，寄存器地址为16位两个字节，寄存器值为8位一个字节，可以理解为一般的字符设备。
该驱动程序并非只能用于ov2715，因此源码中存在支持多个设备地址的机制。
该字符设备的用到的结构体有两个，如下

typedef struct {

  int devAddr;

  struct i2c_client client;   //!< Data structure containing general access routines.
  struct i2c_driver driver;   //!< Data structure containing information specific to each client.

  char name[20];
  int nameSize;
  int users;

} I2C_Obj;

typedef struct {

  int devAddr;

  struct i2c_client client;   //!< Data structure containing general access routines.
  struct i2c_driver driver;   //!< Data structure containing information specific to each client.
  
  char name[20];
  int nameSize;
  int users;
  
} I2C_Obj;

#define I2C_DEV_MAX_ADDR  (0xFF)
#define I2C_TRANSFER_BUF_SIZE_MAX   (256)
typedef struct {

  struct cdev cdev;             /* Char device structure    */
  int     major;
  struct semaphore semLock;

  I2C_Obj *pObj[I2C_DEV_MAX_ADDR];

  uint8_t reg[I2C_TRANSFER_BUF_SIZE_MAX];
  uint16_t reg16[I2C_TRANSFER_BUF_SIZE_MAX];
  uint8_t buffer[I2C_TRANSFER_BUF_SIZE_MAX*4];

} I2C_Dev;

#define I2C_DEV_MAX_ADDR  (0xFF)
#define I2C_TRANSFER_BUF_SIZE_MAX   (256)
typedef struct {

  struct cdev cdev;             /* Char device structure    */
  int     major;
  struct semaphore semLock;
    
  I2C_Obj *pObj[I2C_DEV_MAX_ADDR];

  uint8_t reg[I2C_TRANSFER_BUF_SIZE_MAX];
  uint16_t reg16[I2C_TRANSFER_BUF_SIZE_MAX];
  uint8_t buffer[I2C_TRANSFER_BUF_SIZE_MAX*4];
  
} I2C_Dev;

一个I2C_Obj描述一个设备，devAddr保存该设备的地址，I2C_Obj内嵌到结构体I2C_Dev，I2C_Dev管理该驱动所支持的所有设备，尽管支持多个设备，但i2c适配器只有一个，因此需要一个信号量semLock来保护该共享资源，同时只能向一个设备读写数据。成员变量cdev是我们所熟知的，每个字符设备驱动中几乎总会有一个结构体包含它，major用于保存该驱动的主设备编号，reg数组为寄存器地址为8位的寄存器地址缓冲区，reg16为寄存器地址为16的寄存器地址缓冲区。同时可以读写多个寄存器地址的值。buffer为读写的寄存器值

使用I2C_Dev构建一个全局变量gI2C_dev，在驱动的多个地方均需要它。
下面先从字符设备的基本框架入手，然后深入该驱动的细节部分。
首先是该字符设备的初始化和退出函数

int I2C_devInit(void)
{
  int     result, i;
  dev_t   dev = 0;

  result = alloc_chrdev_region(&dev, 0, 1, I2C_DRV_NAME);//分配字符设备空间

  for(i=0; i<I2C_DEV_MAX_ADDR; i++)
  {
    gI2C_dev.pObj[i]=NULL;
  }

  gI2C_dev.major = MAJOR(dev);//保存设备主编号
  sema_init(&gI2C_dev.semLock, 1);//信号量初始化
  cdev_init(&gI2C_dev.cdev, &gI2C_devFileOps);//使用gI2C_devFileOps初始化该字符设备，gI2C_devFileOps见下文
  gI2C_dev.cdev.owner = THIS_MODULE;//常规赋值
 gI2C_dev.cdev.ops = &gI2C_devFileOps;//常规赋值 result = cdev_add(&gI2C_dev.cdev, dev, 1);//添加设备到字符设备中 return result;}void I2C_devExit(void){ dev_t devno = MKDEV(gI2C_dev.major, 0); cdev_del(&gI2C_dev.cdev);//从字符设备中删除该设备 unregister_chrdev_region(devno, 1);//回收空间}
gI2c_devFileOps全局变量，驱动初始化会用到该结构体变量
struct file_operations gI2C_devFileOps = {
  .owner = THIS_MODULE,
  .open = I2C_devOpen,
  .release = I2C_devRelease,
  .ioctl = I2C_devIoctl,
};

int I2C_devInit(void)
{
  int     result, i;
  dev_t   dev = 0;

  result = alloc_chrdev_region(&dev, 0, 1, I2C_DRV_NAME);//分配字符设备空间
  
  for(i=0; i<I2C_DEV_MAX_ADDR; i++)
  {
    gI2C_dev.pObj[i]=NULL;
  }

  gI2C_dev.major = MAJOR(dev);//保存设备主编号
  sema_init(&gI2C_dev.semLock, 1);//信号量初始化
  cdev_init(&gI2C_dev.cdev, &gI2C_devFileOps);//使用gI2C_devFileOps初始化该字符设备，gI2C_devFileOps见下文
  gI2C_dev.cdev.owner = THIS_MODULE;//常规赋值
 gI2C_dev.cdev.ops = &gI2C_devFileOps;//常规赋值 result = cdev_add(&gI2C_dev.cdev, dev, 1);//添加设备到字符设备中 return result;}void I2C_devExit(void){ dev_t devno = MKDEV(gI2C_dev.major, 0); cdev_del(&gI2C_dev.cdev);//从字符设备中删除该设备 unregister_chrdev_region(devno, 1);//回收空间}
gI2c_devFileOps全局变量，驱动初始化会用到该结构体变量
struct file_operations gI2C_devFileOps = {
  .owner = THIS_MODULE,
  .open = I2C_devOpen,
  .release = I2C_devRelease,
  .ioctl = I2C_devIoctl,
};

该驱动只实现了三个函数,open,release和ioctl，对于i2c设备来说，这已经足够了。
在I2C_devOpen和I2C_devOpen中并没有做实际的工作，重要的工作均在I2C_devIoctl这个ioctl中完成。I2C_devIoctl代码展示（将影响结构条理的代码去掉，稍后在做详细分析）

int I2C_devIoctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg)
{
  I2C_Obj *pObj;
  int status=0;
  I2C_TransferPrm transferPrm;

  pObj = (I2C_Obj *)filp->private_data;

  if(!I2C_IOCTL_CMD_IS_VALID(cmd))
    return -1;
  cmd = I2C_IOCTL_CMD_GET(cmd);//cmd命令转换，防止混淆，具体原因参见上一篇文章：ioctl中的cmd

  down_interruptible(&gI2C_dev.semLock);      //信号量down

  switch(cmd)
  {
    case I2C_CMD_SET_DEV_ADDR://命令1，设置设备地址
      filp->private_data = I2C_create(arg);

    case I2C_CMD_WRITE:  //命令2，写寄存器值

      status = copy_from_user(&transferPrm, (void *)arg, sizeof(transferPrm));
      ...

      break;
    case I2C_CMD_READ:  //命令3，读寄存器值

      status = copy_from_user(&transferPrm, (void *)arg, sizeof(transferPrm));
      ...

      break;
    default:
      status = -1;
      break;
  }

  up(&gI2C_dev.semLock);      //信号量up

  return status;
}

int I2C_devIoctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg)
{
  I2C_Obj *pObj;
  int status=0;
  I2C_TransferPrm transferPrm;
  
  pObj = (I2C_Obj *)filp->private_data;

  if(!I2C_IOCTL_CMD_IS_VALID(cmd))
    return -1;
  cmd = I2C_IOCTL_CMD_GET(cmd);//cmd命令转换，防止混淆，具体原因参见上一篇文章：ioctl中的cmd

  down_interruptible(&gI2C_dev.semLock);      //信号量down
  
  switch(cmd)
  {
    case I2C_CMD_SET_DEV_ADDR://命令1，设置设备地址
      filp->private_data = I2C_create(arg);

    case I2C_CMD_WRITE:  //命令2，写寄存器值
      
      status = copy_from_user(&transferPrm, (void *)arg, sizeof(transferPrm));
      ...
            
      break;
    case I2C_CMD_READ:  //命令3，读寄存器值
    
      status = copy_from_user(&transferPrm, (void *)arg, sizeof(transferPrm));
      ...
      
      break;
    default:
      status = -1;
      break;    
  }

  up(&gI2C_dev.semLock);      //信号量up

  return status;
}

以上三个命令中最重要最复杂的是第一个I2C_CMD_SET_DEV_ADDR，设置设备地址，之所以重要和复杂，因为在I2C_create()函数中，将通过i2c-core提供的函数把该驱动程序和底层的i2c_adapter联系起来。下面是I2C_create()函数源码

void *I2C_create(int devAddr) {

    int ret;
    struct i2c_driver *driver;
    struct i2c_client *client = client;
    I2C_Obj *pObj;

    devAddr >>= 1;

    if(devAddr>I2C_DEV_MAX_ADDR)  //变量合法性判断
      return NULL;

    if(gI2C_dev.pObj[devAddr]!=NULL) {  //变量合法性判断，如果该地址的设备已经创建，则调过，防止上层错误调用
      // already allocated, increment user count, and return the allocated handle
      gI2C_dev.pObj[devAddr]->users++;
      return gI2C_dev.pObj[devAddr];
    }

    pObj = (void*)kmalloc( sizeof(I2C_Obj), GFP_KERNEL); //为pObj分配空间
    gI2C_dev.pObj[devAddr] = pObj;  //将分配的空间地址保存在全局变量里
    memset(pObj, 0, sizeof(I2C_Obj));

    pObj->client.adapter = NULL;
    pObj->users++;    //用户基数，初始化为0，当前设为1
    pObj->devAddr = devAddr;  //保存设备地址

    gI2C_curAddr = pObj->devAddr;  //gI2C_curAddr为全局的整型变量，用于保存当前的设备地址
    driver = &pObj->driver;  //将成员变量driver单独抽取出来，因为线面要使用driver来初始化驱动

    pObj->nameSize=0;//i2c设备名称，注意，这里不是在/dev下面的设备节点名
    pObj->name[pObj->nameSize++] = 'I';
    pObj->name[pObj->nameSize++] = '2';
    pObj->name[pObj->nameSize++] = 'C';
    pObj->name[pObj->nameSize++] = '_';
    pObj->name[pObj->nameSize++] = 'A' + ((pObj->devAddr >> 0) & 0xF);
    pObj->name[pObj->nameSize++] = 'B' + ((pObj->devAddr >> 4) & 0xF);
    pObj->name[pObj->nameSize++] = 0;

    driver->driver.name = pObj->name; //保存刚才设置的name
    driver->id = I2C_DRIVERID_MISC;
    driver->attach_adapter = I2C_attachAdapter;   //这个很重要，将驱动连接到i2c适配器上，在后面分析
    driver->detach_client = I2C_detachClient;    //这个很重，在后面分析

    if((ret = i2c_add_driver(driver)))  //使用i2c-core（i2c_register_driver函数）的接口，注册该驱动，i2c_add_driver实质调用了driver_register()
    {
        printk( KERN_ERR "I2C: ERROR: Driver registration failed (address=%x), module not inserted.\n", pObj->devAddr);
    }

    if(ret<0) {

      gI2C_dev.pObj[pObj->devAddr] = NULL;
      kfree(pObj);
      return NULL;
    }
    return pObj;
}

void *I2C_create(int devAddr) {

    int ret;
    struct i2c_driver *driver;
    struct i2c_client *client = client;
    I2C_Obj *pObj;

    devAddr >>= 1;
    
    if(devAddr>I2C_DEV_MAX_ADDR)  //变量合法性判断
      return NULL;
   
    if(gI2C_dev.pObj[devAddr]!=NULL) {	//变量合法性判断，如果该地址的设备已经创建，则调过，防止上层错误调用
      // already allocated, increment user count, and return the allocated handle
      gI2C_dev.pObj[devAddr]->users++;
      return gI2C_dev.pObj[devAddr];
    }
    
    pObj = (void*)kmalloc( sizeof(I2C_Obj), GFP_KERNEL); //为pObj分配空间
    gI2C_dev.pObj[devAddr] = pObj;  //将分配的空间地址保存在全局变量里
    memset(pObj, 0, sizeof(I2C_Obj));
  
    pObj->client.adapter = NULL;
    pObj->users++;    //用户基数，初始化为0，当前设为1
    pObj->devAddr = devAddr;  //保存设备地址
    
    gI2C_curAddr = pObj->devAddr;  //gI2C_curAddr为全局的整型变量，用于保存当前的设备地址
    driver = &pObj->driver;  //将成员变量driver单独抽取出来，因为线面要使用driver来初始化驱动

    pObj->nameSize=0;//i2c设备名称，注意，这里不是在/dev下面的设备节点名
    pObj->name[pObj->nameSize++] = 'I';
    pObj->name[pObj->nameSize++] = '2';
    pObj->name[pObj->nameSize++] = 'C';
    pObj->name[pObj->nameSize++] = '_';   
    pObj->name[pObj->nameSize++] = 'A' + ((pObj->devAddr >> 0) & 0xF);
    pObj->name[pObj->nameSize++] = 'B' + ((pObj->devAddr >> 4) & 0xF);
    pObj->name[pObj->nameSize++] = 0;

    driver->driver.name = pObj->name; //保存刚才设置的name
    driver->id = I2C_DRIVERID_MISC;
    driver->attach_adapter = I2C_attachAdapter;   //这个很重要，将驱动连接到i2c适配器上，在后面分析
    driver->detach_client = I2C_detachClient;	//这个很重，在后面分析

    if((ret = i2c_add_driver(driver)))	//使用i2c-core（i2c_register_driver函数）的接口，注册该驱动，i2c_add_driver实质调用了driver_register()
    {
        printk( KERN_ERR "I2C: ERROR: Driver registration failed (address=%x), module not inserted.\n", pObj->devAddr);
    }

    if(ret<0) {

      gI2C_dev.pObj[pObj->devAddr] = NULL;
      kfree(pObj);    
      return NULL;
    }
    return pObj;
}

其他两个命令是I2C_CMD_WRITE和I2C_CMD_READ，这个比较简单，只需设置寄存器地址的大小以及寄存器值的大小，然后通过i2c-core 提供的i2c_transfer()函数发送即可。例如I2C_wirte()

int I2C_write(I2C_Obj *pObj, uint8_t *reg, uint8_t *buffer, uint8_t count, uint8_t dataSize)
{
  uint8_t i;
  int err;
  struct i2c_client *client;
    struct i2c_msg msg[1];
    unsigned char data[8];

  if(pObj==NULL)
    return -ENODEV;

  client = &pObj->client;//得到client信息
  if(!client->adapter)
    return -ENODEV;

  if(dataSize<=0||dataSize>4)
    return -1;

  for(i=0; i<count; i++) {

    msg->addr = client->addr;//设置要写的i2c设备地址
    msg->flags= 0;//一直为0
    msg->buf  = data;//date为准备i2c通信的缓冲区，这个缓冲区除了不包含设备地址外，要包括要目标寄存器地址，和要写入该寄存器的值

    data[0] = reg[i];//寄存器地址赋值

    if(dataSize==1) {//寄存器值长度为1
       data[1]  = buffer[i];//寄存器值赋值
       msg->len = 2;     //设置data长度为2
    }   else if(dataSize==2) {//寄存器值长度为2
       data[1] = buffer[2*i+1];
       data[2] = buffer[2*i];
       msg->len = 3;
    }
    err = i2c_transfer(client->adapter, msg, 1);//调用i2c-core中的i2c_transfer发送i2c数据
    if( err < 0 )
      return err;
    }

  return 0;
}

int I2C_write(I2C_Obj *pObj, uint8_t *reg, uint8_t *buffer, uint8_t count, uint8_t dataSize)
{
  uint8_t i;
  int err;
  struct i2c_client *client;
	struct i2c_msg msg[1];
	unsigned char data[8];

  if(pObj==NULL)
    return -ENODEV;

  client = &pObj->client;//得到client信息
  if(!client->adapter)
    return -ENODEV;  
  
  if(dataSize<=0||dataSize>4)
    return -1;
    
  for(i=0; i<count; i++) {
  
    msg->addr = client->addr;//设置要写的i2c设备地址
    msg->flags= 0;//一直为0
    msg->buf  = data;//date为准备i2c通信的缓冲区，这个缓冲区除了不包含设备地址外，要包括要目标寄存器地址，和要写入该寄存器的值
		
    data[0] = reg[i];//寄存器地址赋值
		
    if(dataSize==1) {//寄存器值长度为1
       data[1]  = buffer[i];//寄存器值赋值
       msg->len = 2;  	//设置data长度为2	
    }	else if(dataSize==2) {//寄存器值长度为2
       data[1] = buffer[2*i+1];
       data[2] = buffer[2*i];
       msg->len = 3;
    } 
    err = i2c_transfer(client->adapter, msg, 1);//调用i2c-core中的i2c_transfer发送i2c数据
    if( err < 0 )
      return err;
    }
  
  return 0;
}

现在，我们重点分析上一段代码void *I2C_create(int devAddr)函数中的i2c_driver结构体部分的代码，下面的代码是从上面I2C_create抽取出来的

driver->driver.name = pObj->name;
driver->id = I2C_DRIVERID_MISC;
driver->attach_adapter = I2C_attachAdapter;
driver->detach_client = I2C_detachClient;

    driver->driver.name = pObj->name;
    driver->id = I2C_DRIVERID_MISC;
    driver->attach_adapter = I2C_attachAdapter;
    driver->detach_client = I2C_detachClient;

在i2c_driver结构体中针对attach_adapter有这样的说明：

/* Notifies the driver that a new bus has appeared. This routine
 * can be used by the driver to test if the bus meets its conditions
 * & seek for the presence of the chip(s) it supports. If found, it
 * registers the client(s) that are on the bus to the i2c admin. via
 * i2c_attach_client.
 */

	/* Notifies the driver that a new bus has appeared. This routine
	 * can be used by the driver to test if the bus meets its conditions
	 * & seek for the presence of the chip(s) it supports. If found, it 
	 * registers the client(s) that are on the bus to the i2c admin. via
	 * i2c_attach_client.
	 */

意思是通知驱动，i2c适配器已经就绪了，这时可以讲device的driver连接到总线bus上。所以I2C_attachAdapter的作用就是检测client，然后将client连接上来。attach_adapter和detach_client由内核驱动自动调用，我们只需在调用的时候实现必要的功能即可，如下代码展示

int I2C_attachAdapter(struct i2c_adapter *adapter)
{
    return I2C_detectClient(adapter, gI2C_curAddr);
}

int I2C_detectClient(struct i2c_adapter *adapter, int address)
{
    I2C_Obj *pObj;
    struct i2c_client *client;
    int err = 0;

    if(address > I2C_DEV_MAX_ADDR) {
      printk( KERN_ERR "I2C: ERROR: Invalid device address %x\n", address);
      return -1;
    }

    pObj = gI2C_dev.pObj[address];
    if(pObj==NULL) {
      printk( KERN_ERR "I2C: ERROR: Object not found for address %x\n", address);
      return -1;
    }

    client = &pObj->client;

    if(client->adapter)
      return -EBUSY;  /* our client is already attached */

    memset(client, 0x00, sizeof(struct i2c_client));
    client->addr = pObj->devAddr;
    client->adapter = adapter;
    client->driver = &pObj->driver;

    if((err = i2c_attach_client(client)))
    {
        printk( KERN_ERR "I2C: ERROR: Couldn't attach %s (address=%x)\n", pObj->name, pObj->devAddr);
        client->adapter = NULL;
        return err;
    }
    return 0;
}

int I2C_attachAdapter(struct i2c_adapter *adapter)
{
    return I2C_detectClient(adapter, gI2C_curAddr);
}

int I2C_detectClient(struct i2c_adapter *adapter, int address)
{
    I2C_Obj *pObj;
    struct i2c_client *client;
    int err = 0;
    
    if(address > I2C_DEV_MAX_ADDR) {
      printk( KERN_ERR "I2C: ERROR: Invalid device address %x\n", address);        
      return -1;
    }
      
    pObj = gI2C_dev.pObj[address];
    if(pObj==NULL) {
      printk( KERN_ERR "I2C: ERROR: Object not found for address %x\n", address);    
      return -1;
    }

    client = &pObj->client;

    if(client->adapter)
      return -EBUSY;  /* our client is already attached */

    memset(client, 0x00, sizeof(struct i2c_client));
    client->addr = pObj->devAddr;
    client->adapter = adapter;
    client->driver = &pObj->driver;

    if((err = i2c_attach_client(client)))
    {
        printk( KERN_ERR "I2C: ERROR: Couldn't attach %s (address=%x)\n", pObj->name, pObj->devAddr);
        client->adapter = NULL;
        return err;
    }
    return 0;
}

最终I2C_detectClient()函数调用了i2c-core中的i2c_attach_client，从名字上就能看出什么意思，连接client设备。
当内核驱动准备删除该驱动时会自动调用i2c_driver的成员函数：detech_client，因此我们需要实现删除client设备的函数然后赋值给改函数指针，detech_client的说明如下：

/* tells the driver that a client is about to be deleted & gives it
 * the chance to remove its private data. Also, if the client struct
 * has been dynamically allocated by the driver in the function above,
 * it must be freed here.
 */

	/* tells the driver that a client is about to be deleted & gives it 
	 * the chance to remove its private data. Also, if the client struct
	 * has been dynamically allocated by the driver in the function above,
	 * it must be freed here.
	 */

下面是detech_client调用的函数代码清单，该函数最终调用了i2c-core提供的i2c_detach_client，用于取消client设备的连接

int I2C_detachClient(struct i2c_client *client)
{
    int err;

    if(!client->adapter)
        return -ENODEV; /* our client isn't attached */

    if((err = i2c_detach_client(client))) {
        printk( KERN_ERR "Client deregistration failed (address=%x), client not detached.\n", client->addr);
        return err;
    }

    client->adapter = NULL;

    return 0;
}

int I2C_detachClient(struct i2c_client *client)
{
    int err;

    if(!client->adapter)
        return -ENODEV; /* our client isn't attached */

    if((err = i2c_detach_client(client))) {
        printk( KERN_ERR "Client deregistration failed (address=%x), client not detached.\n", client->addr);
        return err;
    }

    client->adapter = NULL;

    return 0;
}

ov2715设备的i2c驱动源码的分析就到这里，至于平台相关的i2c总线驱动分析就放到下一篇文章里分析，因为这部分多数情况下并不需要我们亲自去实现。但是对于理解i2c驱动架构来说，还是有很大帮助的。