Linux设备驱动之I2C架构分析 ------------------------------------------ 本文系本站原创,欢迎转载! 转载请注明出处:http://ericxiao.cublog.cn/ ------------------------------------------ 一:前言 I2c是philips提出的外设总线.I2C只有两条线,一条串行数据线:SDA,一条是时钟线SCL.正因为这样,它方便了工程人员的布线.另外,I2C是一种多主机控制总线.它和USB总线不同,USB是基于master-slave机制,任何设备的通信必须由主机发起才可以.而I2C是基于multi master机制.一同总线上可允许多个master.关于I2C协议的知识,这里不再赘述.可自行下载spec阅读即可. 二:I2C架构概述 在linux中,I2C驱动架构如下所示:
如上图所示,每一条I2C对应一个adapter.在kernel中,每一个adapter提供了一个描述的结构(struct i2c_adapter),也定义了adapter支持的操作(struct i2c_adapter).再通过i2c core层将i2c设备与i2c adapter关联起来. 三:adapter注册 在kernel中提供了两个adapter注册接口,分别为i2c_add_adapter()和i2c_add_numbered_adapter().由于在系统中可能存在多个adapter,因为将每一条I2C总线对应一个编号,下文中称为I2C总线号.这个总线号的PCI中的总线号不同.它和硬件无关,只是软件上便于区分而已. 对于i2c_add_adapter()而言,它使用的是动态总线号,即由系统给其分析一个总线号,而i2c_add_numbered_adapter()则是自己指定总线号,如果这个总线号非法或者是被占用,就会注册失败. 分别来看一下这两个函数的代码: int i2c_add_adapter(struct i2c_adapter *adapter) { int id, res = 0; retry: if (idr_pre_get(&i2c_adapter_idr, GFP_KERNEL) == 0) return -ENOMEM; mutex_lock(&core_lock); /* "above" here means "above or equal to", sigh */ res = idr_get_new_above(&i2c_adapter_idr, adapter, __i2c_first_dynamic_bus_num, &id); mutex_unlock(&core_lock); if (res < 0) { if (res == -EAGAIN) goto retry; return res; } adapter->nr = id; return i2c_register_adapter(adapter); } 在这里涉及到一个idr结构.idr结构本来是为了配合page cache中的radix tree而设计的.在这里我们只需要知道,它是一种高效的搜索树,且这个树预先存放了一些内存.避免在内存不够的时候出现问题.所在,在往idr中插入结构的时候,首先要调用idr_pre_get()为它预留足够的空闲内存,然后再调用idr_get_new_above()将结构插入idr中,该函数以参数的形式返回一个id.以后凭这个id就可以在idr中找到相对应的结构了.对这个数据结构操作不太理解的可以查阅本站<< linux文件系统之文件的读写>>中有关radix tree的分析. 注意一下idr_get_new_above(&i2c_adapter_idr, adapter,__i2c_first_dynamic_bus_num, &id)的参数的含义,它是将adapter结构插入到i2c_adapter_idr中,存放位置的id必须要大于或者等于__i2c_first_dynamic_bus_num, 然后将对应的id号存放在adapter->nr中.调用i2c_register_adapter(adapter)对这个adapter进行进一步注册. 看一下另外一人注册函数: i2c_add_numbered_adapter( ),如下所示: int i2c_add_numbered_adapter(struct i2c_adapter *adap) { int id; int status; if (adap->nr & ~MAX_ID_MASK) return -EINVAL; retry: if (idr_pre_get(&i2c_adapter_idr, GFP_KERNEL) == 0) return -ENOMEM; mutex_lock(&core_lock); /* "above" here means "above or equal to", sigh; * we need the "equal to" result to force the result */ status = idr_get_new_above(&i2c_adapter_idr, adap, adap->nr, &id); if (status == 0 && id != adap->nr) { status = -EBUSY; idr_remove(&i2c_adapter_idr, id); } mutex_unlock(&core_lock); if (status == -EAGAIN) goto retry; if (status == 0) status = i2c_register_adapter(adap); return status; } 对比一下就知道差别了,在这里它已经指定好了adapter->nr了.如果分配的id不和指定的相等,便返回错误. 过一步跟踪i2c_register_adapter().代码如下: static int i2c_register_adapter(struct i2c_adapter *adap) { int res = 0, dummy; mutex_init(&adap->bus_lock); mutex_init(&adap->clist_lock); INIT_LIST_HEAD(&adap->clients); mutex_lock(&core_lock); /* Add the adapter to the driver core. * If the parent pointer is not set up, * we add this adapter to the host bus. */ if (adap->dev.parent == NULL) { adap->dev.parent = &platform_bus; pr_debug("I2C adapter driver [%s] forgot to specify " "physical device\n", adap->name); } sprintf(adap->dev.bus_id, "i2c-%d", adap->nr); adap->dev.release = &i2c_adapter_dev_release; adap->dev.class = &i2c_adapter_class; res = device_register(&adap->dev); if (res) goto out_list; dev_dbg(&adap->dev, "adapter [%s] registered\n", adap->name); /* create pre-declared device nodes for new-style drivers */ if (adap->nr < __i2c_first_dynamic_bus_num) i2c_scan_static_board_info(adap); /* let legacy drivers scan this bus for matching devices */ dummy = bus_for_each_drv(&i2c_bus_type, NULL, adap, i2c_do_add_adapter); out_unlock: mutex_unlock(&core_lock); return res; out_list: idr_remove(&i2c_adapter_idr, adap->nr); goto out_unlock; } 首先对adapter和adapter中内嵌的struct device结构进行必须的初始化.之后将adapter内嵌的struct device注册. 在这里注意一下adapter->dev的初始化.它的类别为i2c_adapter_class,如果没有父结点,则将其父结点设为platform_bus.adapter->dev的名字为i2c + 总线号. 测试一下: [eric@mochow i2c]$ cd /sys/class/i2c-adapter/ [eric@mochow i2c-adapter]$ ls i2c-0 可以看到,在我的PC上,有一个I2C adapter,看下详细信息: [eric@mochow i2c-adapter]$ tree . `-- i2c-0 |-- device -> ../../../devices/pci0000:00/0000:00:1f.3/i2c-0 |-- name |-- subsystem -> ../../../class/i2c-adapter `-- uevent 3 directories, 2 files 可以看到,该adapter是一个PCI设备. 继续往下看: 之后,在注释中看到,有两种类型的driver,一种是new-style drivers,另外一种是legacy drivers New-style drivers是在2.6近版的kernel加入的.它们最主要的区别是在adapter和i2c driver的匹配上. 3.1: new-style 形式的adapter注册 对于第一种,也就是new-style drivers,将相关代码再次列出如下: if (adap->nr < __i2c_first_dynamic_bus_num) i2c_scan_static_board_info(adap); 如果adap->nr 小于__i2c_first_dynamic_bus_num的话,就会进入到i2c_scan_static_board_info(). 结合我们之前分析的adapter的两种注册分式: i2c_add_adapter()所分得的总线号肯会不会小于__i2c_first_dynamic_bus_num.只有i2c_add_numbered_adapter()才有可能满足: (adap->nr < __i2c_first_dynamic_bus_num) 而且必须要调用i2c_register_board_info()将板子上的I2C设备信息预先注册时才会更改__i2c_first_dynamic_bus_num的值.在x86上只没有使用i2c_register_board_info()的.因此,x86平台上的分析可以忽略掉new-style driver的方式.不过,还是详细分析这种情况下. 首先看一下i2c_register_board_info(),如下: int __init i2c_register_board_info(int busnum, struct i2c_board_info const *info, unsigned len) { int status; mutex_lock(&__i2c_board_lock); /* dynamic bus numbers will be assigned after the last static on
e */ if (busnum >= __i2c_first_dynamic_bus_num) __i2c_first_dynamic_bus_num = busnum + 1; for (status = 0; len; len--, info++) { struct i2c_devinfo *devinfo; devinfo = kzalloc(sizeof(*devinfo), GFP_KERNEL); if (!devinfo) { pr_debug("i2c-core: can't register boardinfo!\n"); status = -ENOMEM; break; } devinfo->busnum = busnum; devinfo->board_info = *info; list_add_tail(&devinfo->list, &__i2c_board_list); } mutex_unlock(&__i2c_board_lock); return status; } 跟踪i2c_scan_static_board_info():代码如下: static void i2c_scan_static_board_info(struct i2c_adapter *adapter) { struct i2c_devinfo *devinfo; mutex_lock(&__i2c_board_lock); list_for_each_entry(devinfo, &__i2c_board_list, list) { if (devinfo->busnum == adapter->nr && !i2c_new_device(adapter, &devinfo->board_info)) printk(KERN_ERR "i2c-core: can't create i2c%d-%04x\n", i2c_adapter_id(adapter), devinfo->board_info.addr); } mutex_unlock(&__i2c_board_lock); } 该函数遍历挂在__i2c_board_list链表上面的i2c设备的信息,也就是我们在启动的时候指出的i2c设备的信息. 如果指定设备是位于adapter所在的I2C总线上,那么,就调用i2c_new_device().代码如下: struct i2c_client * i2c_new_device(struct i2c_adapter *adap, struct i2c_board_info const *info) { struct i2c_client *client; int status; client = kzalloc(sizeof *client, GFP_KERNEL); if (!client) return NULL; client->adapter = adap; client->dev.platform_data = info->platform_data; device_init_wakeup(&client->dev, info->flags & I2C_CLIENT_WAKE); client->flags = info->flags & ~I2C_CLIENT_WAKE; client->addr = info->addr; client->irq = info->irq; strlcpy(client->name, info->type, sizeof(client->name)); /* a new style driver may be bound to this device when we * return from this function, or any later moment (e.g. maybe * hotplugging will load the driver module). and the device * refcount model is the standard driver model one. */ status = i2c_attach_client(client); if (status < 0) { kfree(client); client = NULL; } return client; } 我们又遇到了一个新的结构:struct i2c_client,不要被这个结构吓倒了,其实它就是一个嵌入struct device的I2C设备的封装.它和我们之前遇到的struct usb_device结构的作用是一样的. 首先,在clinet里保存该设备的相关消息.特别的, client->adapter指向了它所在的adapter. 特别的,clinet->name为info->name.也是指定好了的. 一切初始化完成之后,便会调用i2c_attach_client( ).看这个函数的字面意思,是将clinet关联起来.到底怎么样关联呢?继续往下看: int i2c_attach_client(struct i2c_client *client) { struct i2c_adapter *adapter = client->adapter; int res = 0; //初始化client内嵌的dev结构 //父结点为所在的adapter,所在bus为i2c_bus_type client->dev.parent = &client->adapter->dev; client->dev.bus = &i2c_bus_type; //如果client已经指定了driver,将driver和内嵌的dev关联起来 if (client->driver) client->dev.driver = &client->driver->driver; //指定了driver, 但不是newstyle的 if (client->driver && !is_newstyle_driver(client->driver)) { client->dev.release = i2c_client_release; client->dev.uevent_suppress = 1; } else client->dev.release = i2c_client_dev_release; //clinet->dev的名称 snprintf(&client->dev.bus_id[0], sizeof(client->dev.bus_id), "%d-%04x", i2c_adapter_id(adapter), client->addr); //将内嵌的dev注册 res = device_register(&client->dev); if (res) goto out_err; //将clinet链到adapter->clients中 mutex_lock(&adapter->clist_lock); list_add_tail(&client->list, &adapter->clients); mutex_unlock(&adapter->clist_lock); dev_dbg(&adapter->dev, "client [%s] registered with bus id %s\n", client->name, client->dev.bus_id); //如果adapter->cleinet_reqister存在,就调用它 if (adapter->client_register) { if (adapter->client_register(client)) { dev_dbg(&adapter->dev, "client_register " "failed for client [%s] at 0x%02x\n", client->name, client->addr); } } return 0; out_err: dev_err(&adapter->dev, "Failed to attach i2c client %s at 0x%02x " "(%d)\n", client->name, client->addr, res); return res; } 参考上面添加的注释,应该很容易理解这段代码了,就不加详细分析了.这个函数的名字不是i2c_attach_client()么?怎么没看到它的关系过程呢? 这是因为:在代码中设置了client->dev所在的bus为i2c_bus_type .以为只需要有bus为i2c_bus_type的driver注册,就会产生probe了.这个过程呆后面分析i2c driver的时候再来详细分析. Legacy形式的adapter注册代码片段如下: dummy = bus_for_each_drv(&i2c_bus_type, NULL, adap, i2c_do_add_adapter); 这段代码遍历挂在i2c_bus_type上的驱动,然后对每一个驱动和adapter调用i2c_do_add_adapter(). 代码如下: static int i2c_do_add_adapter(struct device_driver *d, void *data) { struct i2c_driver *driver = to_i2c_driver(d); struct i2c_adapter *adap = data; if (driver->attach_adapter) { /* We ignore the return code; if it fails, too bad */ driver->attach_adapter(adap); } return 0; } 该函数很简单,就是调用driver的attach_adapter()接口. 到此为止,adapter的注册已经分析完了. 四:i2c driver注册 在分析i2c driver的时候,有必要先分析一下i2c架构的初始化 代码如下: static int __init i2c_init(void) { int retval; retval = bus_register(&i2c_bus_type); if (retval) return retval; retval = class_register(&i2c_adapter_class); if (retval) goto bus_err; retval = i2c_add_driver(&dummy_driver); if (retval) goto class_err; return 0; class_err: class_unregister(&i2c_adapter_class); bus_err: bus_unregister(&i2c_bus_type); return retval; } subsys_initcall(i2c_init); 很明显,i2c_init()会在系统初始化的时候被调用. 在i2c_init中,先注册了i2c_bus_type的bus,i2c_adapter_class的class.然后再调用i2c_add_driver()注册了一个i2c driver. I2c_bus_type结构如下: static struct bus_type i2c_bus_type = { .name = "i2c", .dev_attrs = i2c_dev_attrs, .match = i2c_device_match, .uevent = i2c_device_uevent, .probe = i2c_device_probe, .remove = i2c_device_remove, .shutdown = i2c_device_shutdown, .suspend = i2c_device_suspend, .resume = i2c_device_resume, }; 这个结构先放在这里吧,以后还会用到里面的信息的. 从上面的初始化函数里也看到了,注册i2c driver的接口为i2c_add_driver().代码如下: static inline int i2c_add_driver(struct i2c_driver *driver) { return i2c_register_driver(THIS_MODULE, driver); } 继续跟踪: int i2c_register_driver(struct module *owner, struct i2c_driver *driver) { int res; /* new style driver methods can't mix with legacy ones */ //如果是一个newstyle的driver.但又定义了attach_adapter/detach_adapter.非法 if (is_newstyle_driver(driver)) { if (driver->attach_adapter || driver->detach_adapter || driver->detach_client) { printk(KERN_WARNING "i2c-core: driver [%s] is confused\n", driver->driver.name); return -EINVAL; } } /* add the driver to the list of i2c drivers in the driver core */ //关联到i2c_bus_types driver->driver.owner = owner; driver->driver.bus = &i2c_bus_type; /* for new style drivers, when registration returns the driver core * will have called probe() for all matching-but-unbound devices. */ //注册内嵌的driver res = driver_register(&driver->driver); if (res) return res; mutex_lock(&core_lock); pr_debug("i2c-core: driver [%s] registered\n", driver->driver.name); /* legacy drivers scan i2c busses directly */ //遍历所有的adapter,对其都调用driver->attach_adapter if (driver->attach_adapter) { struct i2c_adapter *adapter; down(&i2c_adapter_class.sem); list_for_each_entry(adapter, &i2c_adapter_class.devices, dev.node) { driver->attach_adapter(adapter); } up(&i2c_adapter_class.sem); } mutex_unlock(&core_lock); return 0; } 这里也有两种形式的区分,对于第一种,只需要将内嵌的driver注册就可以了,对于legacy的情况,对每一个adapter都调用driver->attach_adapter(). 现在,我们可以将adapter和i2c driver关联起来考虑一下了: 1:如果是news style形式的,在注册adapter的时候,将它上面的i2c 设备转换成了struct client.struct client->dev->bus又指定了和i2c driver同一个bus.因为,它们可以发生probe. 2:如果是legacy形式,就直接找到对应的对象,调用driver->attach_adapter(). I2c_bus_type的操作主要存在于new-style形式的驱动中.接下来分析一下对应的probe过程: 5.1:match过程分析 Match对应的操作函数为i2c_device_match().代码如下 static int i2c_device_match(struct device *dev, struct device_driver *drv) { struct i2c_client *client = to_i2c_client(dev); struct i2c_driver *driver = to_i2c_driver(drv); /* make legacy i2c drivers bypass driver model probing entirely; * such drivers scan each i2c adapter/bus themselves. */ if (!is_newstyle_driver(driver)) return 0; /* match on an id table if there is one */ if (driver->id_table) return i2c_match_id(driver->id_table, client) != NULL; return 0; } 如果该驱动不是一个new-style形式的.或者driver没有定义匹配的id_table.都会匹配失败. 继续跟踪进i2c_match_id(): static const struct i2c_device_id *i2c_match_id(const struct i2c_device_id *id, const struct i2c_client *client) { while (id->name[0]) { if (strcmp(client->name, id->name) == 0) return id; id++; } return NULL; } 由此可见.如果client的名字和driver->id_table[]中的名称匹配即为成功. 5.2:probe过程分析 static int i2c_device_probe(struct device *dev) { struct i2c_client *client = to_i2c_client(dev); struct i2c_driver *driver = to_i2c_driver(dev->driver); const struct i2c_device_id *id; int status; if (!driver->probe) return -ENODEV; client->driver = driver; dev_dbg(dev, "probe\n"); if (driver->id_table) client); else > status = driver->probe(client, id); if (status) client->driver = NULL; return status; } 这个函数也很简单,就是将probe流程回溯到i2c driver的probe() 六:其它的扩展 分析完adapter和i2c driver的注册之后,好像整个架构也差不多了,其它,扩展的东西还有很多. 我们举一个legacy形式的例子,这个例子是在kernel中随便搜索出来的: 在linux-2.6.26.3/drivers/hwmon/ad7418.c中,初始化函数为: static int __init ad7418_init(void) { return i2c_add_driver(&ad7418_driver); } i2c_driver ad7418_driver结构如下: static struct i2c_driver ad7418_driver = { .driver = { .name = "ad7418", }, .attach_adapter = ad7418_attach_adapter, .detach_client = ad7418_detach_client, }; 该结构中没有probe()函数,可以断定是一个legacy形式的驱动.这类驱动注册的时候,会调用driver的attach_adapter函数.在这里也就是ad7418_attach_adapter. 这个函数代码如下: static int ad7418_attach_adapter(struct i2c_adapter *adapter) { if (!(adapter->class & I2C_CLASS_HWMON)) return 0; } 在这里我们又遇到了一个i2c-core中的函数,i2c_probe().在分析这个函数之前,先来看下addr_data是什么? #define I2C_CLIENT_MODULE_PARM(var,desc) \ static unsigned short var[I2C_CLIENT_MAX_OPTS] = I2C_CLIENT_DEFAULTS; \ static unsigned int var##_num; \ module_param_array(var, short, &var##_num, 0); \ MODULE_PARM_DESC(var,desc) #define I2C_CLIENT_MODULE_PARM_FORCE(name) \ I2C_CLIENT_MODULE_PARM(force_##name, \ "List of adapter,address pairs which are " \ "unquestionably assumed to contain a `" \ # name "' chip") #define I2C_CLIENT_INSMOD_COMMON \ I2C_CLIENT_MODULE_PARM(probe, "List of adapter,address pairs to scan " \ "additionally"); \ I2C_CLIENT_MODULE_PARM(ignore, "List of adapter,address pairs not to " \ "scan"); \ static const struct i2c_client_address_data addr_data = { \ .normal_i2c = normal_i2c, \ .probe = probe, \ .ignore = ignore, \ .forces = forces, \ } #define I2C_CLIENT_FORCE_TEXT \ "List of adapter,address pairs to boldly assume to be present" 由此可知道,addr_data中的三个成员都是模块参数.在加载模块的时候可以用参数的方式对其赋值.三个模块参数为别为probe,ignore,force.另外需要指出的是normal_i2c不能以模块参数的方式对其赋值,只能在驱动内部静态指定. 从模块参数的模述看来, probe是指"List of adapter,address pairs to scan additionally" Ignore是指"List of adapter,address pairs not to scan " Force是指"List of adapter,address pairs to boldly assume to be present" 事实上,它们里面的数据都是成对出现的.前面一部份表示所在的总线号,ANY_I2C_BUS表示任一总线.后一部份表示设备的地址. 现在可以来跟踪i2c_probe()的代码了.如下: int i2c_probe(struct i2c_adapter *adapter, const struct i2c_client_address_data *address_data, int (*found_proc) (struct i2c_adapter *, int, int)) { int i, err; int adap_id = i2c_adapter_id(adapter); /* Force entries are done first, and are not affected by ignore entries */ //先扫描force里面的信息,注意它是一个二级指针.ignore里的信息对它是无效的 if (address_data->forces) { const unsigned short * const *forces = address_data->forces; int kind; for (kind = 0; forces[kind]; kind++) { for (i = 0; forces[kind][i] != I2C_CLIENT_END; i += 2) { if (forces[kind][i] == adap_id || forces[kind][i] == ANY_I2C_BUS) { dev_dbg(&adapter->dev, "found force " "parameter for adapter %d, " "addr 0x%02x, kind %d\n", adap_id, forces[kind][i + 1], kind); err = i2c_probe_address(adapter, forces[kind][i + 1], kind, found_proc); if (err) return err; } } } } /* Stop here if we can't use SMBUS_QUICK */ //如果adapter不支持quick.不能够遍历这个adapter上面的设备 if (!i2c_check_functionality(adapter, I2C_FUNC_SMBUS_QUICK)) { if (address_data->probe[0] == I2C_CLIENT_END && address_data->normal_i2c[0] == I2C_CLIENT_END) return 0; dev_warn(&adapter->dev, "SMBus Quick command not supported, " "can't probe for chips\n"); return -1; } /* Probe entries are done second, and are not affected by ignore entries either */ //遍历probe上面的信息.ignore上的信息也对它是没有影响的 for (i = 0; address_data->probe[i] != I2C_CLIENT_END; i += 2) { if (address_data->probe[i] == adap_id || address_data->probe[i] == ANY_I2C_BUS) { dev_dbg(&adapter->dev, "found probe parameter for " "adapter %d, addr 0x%02x\n", adap_id, address_data->probe[i + 1]); err = i2c_probe_address(adapter, address_data->probe[i + 1], -1, found_proc); if (err) return err; } } /* Normal entries are done last, unless shadowed by an ignore entry */ //最后遍历normal_i2c上面的信息.它上面的信息不能在ignore中. for (i = 0; address_data->normal_i2c[i] != I2C_CLIENT_END; i += 1) { int j, ignore; ignore = 0; for (j = 0; address_data->ignore[j] != I2C_CLIENT_END; j += 2) { if ((address_data->ignore[j] == adap_id || address_data->ignore[j] == ANY_I2C_BUS) && address_data->ignore[j + 1] == address_data->normal_i2c[i]) { dev_dbg(&adapter->dev, "found ignore " "parameter for adapter %d, " "addr 0x%02x\n", adap_id, address_data->ignore[j + 1]); ignore = 1; break; } } if (ignore) continue; dev_dbg(&adapter->dev, "found normal entry for adapter %d, " "addr 0x%02x\n", adap_id, address_data->normal_i2c[i]); err = i2c_probe_address(adapter, address_data->normal_i2c[i], -1, found_proc); if (err) return err; } return 0; } 这段代码很简单,结合代码上面添加的注释应该很好理解.如果匹配成功,则会调用i2c_probe_address ().这个函数代码如下: static int i2c_probe_address(struct i2c_adapter *adapter, int addr, int kind, int (*found_proc) (struct i2c_adapter *, int, int)) { int err; /* Make sure the address is valid */ //地址小于0x03或者大于0x77都是不合法的 if (addr < 0x03 || addr > 0x77) { dev_warn(&adapter->dev, "Invalid probe address 0x%02x\n", addr); return -EINVAL; } /* Skip if already in use */ //adapter上已经有这个设备了 if (i2c_check_addr(adapter, addr)) return 0; /* Make sure there is something at this address, unless forced */ //如果kind小于0.检查adapter上是否有这个设备 if (kind < 0) { if (i2c_smbus_xfer(adapter, addr, 0, 0, 0, I2C_SMBUS_QUICK, NULL) < 0) return 0; /* prevent 24RF08 corruption */ if ((addr & ~0x0f) == 0x50) i2c_smbus_xfer(adapter, addr, 0, 0, 0, I2C_SMBUS_QUICK, NULL); } /* Finally call the custom detection function */ //调用回调函数 err = found_proc(adapter, addr, kind); /* -ENODEV can be returned if there is a chip at the given address but it isn't supported by this chip driver. We catch it here as this isn't an error. */ if (err == -ENODEV) err = 0; if (err) dev_warn(&adapter->dev, "Client creation failed at 0x%x (%d)\n", addr, err); return err; } 首先,对传入的参数进行一系列的合法性检查.另外,如果该adapter上已经有了这个地址的设备了.也会返回失败.所有adapter下面的设备都是以adapter->dev为父结点的.因此只需要遍历adapter->dev下面的子设备就可以得到当前地址是不是被占用了. 如果kind < 0.还得要adapter检查该总线是否有这个地址的设备.方法是向这个地址发送一个Read的Quick请求.如果该地址有应答,则说明这个地址上有这个设备.另外还有一种情况是在24RF08设备的特例. 如果adapter上确实有这个设备,就会调用驱动调用时的回调函数. 在上面涉及到了IIC的传输方式,有疑问的可以参考intel ICH5手册的有关smbus部份. 跟踪i2c_smbus_xfer().代码如下: s32 i2c_smbus_xfer(struct i2c_adapter * adapter, u16 addr, unsigned short flags, char read_write, u8 command, int size, union i2c_smbus_data * data) { s32 res; flags &= I2C_M_TEN | I2C_CLIENT_PEC; if (adapter->algo->smbus_xfer) { mutex_lock(&adapter->bus_lock); res = adapter->algo->smbus_xfer(adapter,addr,flags,read_write, command,size,data); mutex_unlock(&adapter->bus_lock); } else res = i2c_smbus_xfer_emulated(adapter,addr,flags,read_write, command,size,data); return res; } 如果adapter有smbus_xfer()函数,则直接调用它发送,否则,也就是在adapter不支持smbus协议的情况下,调用i2c_smbus_xfer_emulated()继续处理. 跟进i2c_smbus_xfer_emulated().代码如下: static s32 i2c_smbus_xfer_emulated(struct i2c_adapter * adapter, u16 addr, unsigned short flags, char read_write, u8 command, int size, union i2c_smbus_data * data) { /* So we need to generate a series of msgs. In the case of writing, we need to use only one message; when reading, we need two. We initialize most things with sane defaults, to keep the code below somewhat simpler. */ //写操作只会进行一次交互,而读操作,有时会有两次操作. //因为有时候读操作要先写command,再从总线上读数据 //在这里为了代码的简洁.使用了两个缓存区,将两种情况统一起来. unsigned char msgbuf0[I2C_SMBUS_BLOCK_MAX+3]; unsigned char msgbuf1[I2C_SMBUS_BLOCK_MAX+2]; //一般来说,读操作要交互两次.例外的情况我们在下面会接着分析 int num = read_write == I2C_SMBUS_READ?2:1; //与设备交互的数据,一般在msg[0]存放写入设备的信息,在msb[1]里存放接收到的 //信息.不过也有例外的 //msg[2]的初始化,默认发送缓存区占一个字节,无接收缓存 struct i2c_msg msg[2] = { { addr, flags, 1, msgbuf0 }, { addr, flags | I2C_M_RD, 0, msgbuf1 } }; int i; u8 partial_pec = 0; //将要发送的信息copy到发送缓存区的第一字节 msgbuf0[0] = command; switch(size) { //quick类型的,其它并不传输有效数据,只是将地址写到总线上,等待应答即可 //所以将发送缓存区长度置为0 .再根据读/写操作,调整msg[0]的标志位 //这类传输只需要一次总线交互 case I2C_SMBUS_QUICK: msg[0].len = 0; /* Special case: The read/write field is used as data */ msg[0].flags = flags | (read_write==I2C_SMBUS_READ)?I2C_M_RD:0; num = 1; break; case I2C_SMBUS_BYTE: //BYTE类型指一次写和读只有一个字节.这种情况下,读和写都只会交互一次 //这种类型的读有例外,它读取出来的数据不是放在msg[1]中的,而是存放在msg[0] if (read_write == I2C_SMBUS_READ) { /* Special case: only a read! */ msg[0].flags = I2C_M_RD | flags; num = 1; } break; case I2C_SMBUS_BYTE_DATA: //Byte_Data是指命令+数据的传输形式.在这种情况下,写只需要一次交互,读却要两次 //第一次将command写到总线上,第二次要转换方向.要将设备地址和read标志写入总线. //应回答之后再进行read操作 //写操作占两字节,分别是command+data.读操作的有效数据只有一个字节 //交互次数用初始化值就可以了 if (read_write == I2C_SMBUS_READ) msg[1].len = 1; else { msg[0].len = 2; msgbuf0[1] = data->byte; } break; case I2C_SMBUS_WORD_DATA: //Word_Data是指命令+双字节的形式.这种情况跟Byte_Data的情况类似 //两者相比只是交互的数据大小不同 if (read_write == I2C_SMBUS_READ) msg[1].len = 2; else { msg[0].len=3; msgbuf0[1] = data->word & 0xff; msgbuf0[2] = data->word >> 8; } break; case I2C_SMBUS_PROC_CALL: //Proc_Call的方式与write 的Word_Data相似,只不过写完Word_Data之后,要等待它的应答 //应该它需要交互两次,一次写一次读 num = 2; /* Special case */ read_write = I2C_SMBUS_READ; msg[0].len = 3; msg[1].len = 2; msgbuf0[1] = data->word & 0xff; msgbuf0[2] = data->word >> 8; break; case I2C_SMBUS_BLOCK_DATA: //Block_Data:指command+N段数据的情况. //如果是读操作,它首先要写command到总线,然后再读N段数据.要写的command已经 //放在msg[0]了.现在只需要将msg[1]的标志置I2C_M_RECV_LEN位,msg[1]有效长度为1字节.因为 //adapter驱动会处理好的.现在现在还不知道要传多少段数据. //对于写的情况:msg[1]照例不需要.将要写的数据全部都放到msb[0]中.相应的也要更新 //msg[0]中的缓存区长度 if (read_write == I2C_SMBUS_READ) { msg[1].flags |= I2C_M_RECV_LEN; msg[1].len = 1; /* block length will be added by the underlying bus driver */ } else { //data->block[0]表示后面有多少段数据.总长度要加2是因为command+count+N段数据 msg[0].len = data->block[0] + 2; if (msg[0].len > I2C_SMBUS_BLOCK_MAX + 2) { dev_err(&adapter->dev, "smbus_access called with " "invalid block write size (%d)\n", data->block[0]); return -1; } for (i = 1; i < msg[0].len; i++) msgbuf0[i] = data->block[i-1]; } break; case I2C_SMBUS_BLOCK_PROC_CALL: //Proc_Call:表示写完Block_Data之后,要等它的应答消息它和Block_Data相比,只是多了一部份应答而已 num = 2; /* Another special case */ read_write = I2C_SMBUS_READ; if (data->block[0] > I2C_SMBUS_BLOCK_MAX) { dev_err(&adapter->dev, "%s called with invalid " "block proc call size (%d)\n", __func__, data->block[0]); return -1; } msg[0].len = data->block[0] + 2; for (i = 1; i < msg[0].len; i++) msgbuf0[i] = data->block[i-1]; msg[1].flags |= I2C_M_RECV_LEN; msg[1].len = 1; /* block length will be added by the underlying bus driver */ break; case I2C_SMBUS_I2C_BLOCK_DATA: //I2c Block_Data与Block_Data相似,只不过read的时候,数据长度是预先定义好了的.另外 //与Block_Data相比,中间不需要传输Count字段.(Count表示数据段数目) if (read_write == I2C_SMBUS_READ) { msg[1].len = data->block[0]; } else { msg[0].len = data->block[0] + 1; if (msg[0].len > I2C_SMBUS_BLOCK_MAX + 1) { dev_err(&adapter->dev, "i2c_smbus_xfer_emulated called with " "invalid block write size (%d)\n", data->block[0]); return -1; } for (i = 1; i <= data->block[0]; i++) msgbuf0[i] = data->block[i]; } break; default: dev_err(&adapter->dev, "smbus_access called with invalid size (%d)\n", size); return -1; } //如果启用了PEC.Quick和I2c Block_Data是不支持PEC的 i = ((flags & I2C_CLIENT_PEC) && size != I2C_SMBUS_QUICK && size != I2C_SMBUS_I2C_BLOCK_DATA); if (i) { /* Compute PEC if first message is a write */ //如果第一个操作是写操作 if (!(msg[0].flags & I2C_M_RD)) { //如果只是写操作 if (num == 1) /* Write only */ //如果只有写操作,写缓存区要扩充一个字节,用来存放计算出来的PEC i2c_smbus_add_pec(&msg[0]); else /* Write followed by read */ //如果后面还有读操作,先计算前面写部份的PEC(注意这种情况下不需要 //扩充写缓存区,因为不需要发送PEC.只会接收到PEC) partial_pec = i2c_smbus_msg_pec(0, &msg[0]); } /* Ask for PEC if last message is a read */ //如果最后一次是读消息.还要接收到来自slave的PEC.所以接收缓存区要扩充一个字节 if (msg[num-1].flags & I2C_M_RD) msg[num-1].len++; } if (i2c_transfer(adapter, msg, num) < 0) return -1; /* Check PEC if last message is a read */ //操作完了之后,如果最后一个操作是PEC的读操作.检验后面的PEC是否正确 if (i && (msg[num-1].flags & I2C_M_RD)) { if (i2c_smbus_check_pec(partial_pec, &msg[num-1]) < 0) return -1; } //操作完了,现在可以将数据放到data部份返回了. if (read_write == I2C_SMBUS_READ) switch(size) { case I2C_SMBUS_BYTE: data->byte = msgbuf0[0]; break; case I2C_SMBUS_BYTE_DATA: data->byte = msgbuf1[0]; break; case I2C_SMBUS_WORD_DATA: case I2C_SMBUS_PROC_CALL: data->word = msgbuf1[0] | (msgbuf1[1] << 8); break; case I2C_SMBUS_I2C_BLOCK_DATA: for (i = 0; i < data->block[0]; i++) data->block[i+1] = msgbuf1[i]; break; case I2C_SMBUS_BLOCK_DATA: case I2C_SMBUS_BLOCK_PROC_CALL: for (i = 0; i < msgbuf1[0] + 1; i++) data->block[i] = msgbuf1[i]; break; } return 0; } 在这个函数添上了很详细的注释,配和intel的datasheet,应该很容易看懂.在上面的交互过程中,调用了子函数i2c_transfer().它的代码如下所示: int i2c_transfer(struct i2c_adapter * adap, struct i2c_msg *msgs, int num) { int ret; if (adap->algo->master_xfer) { #ifdef DEBUG for (ret = 0; ret < num; ret++) { dev_dbg(&adap->dev, "master_xfer[%d] %c, addr=0x%02x, " "len=%d%s\n", ret, (msgs[ret].flags & I2C_M_RD) 'R' : 'W', msgs[ret].addr, msgs[ret].len, (msgs[ret].flags & I2C_M_RECV_LEN) ? "+" : ""); } #endif if (in_atomic() || irqs_disabled()) { ret = mutex_trylock(&adap->bus_lock); if (!ret) /* I2C activity is ongoing. */ return -EAGAIN; } else { mutex_lock_nested(&adap->bus_lock, adap->level); } ret = adap->algo->master_xfer(adap,msgs,num); mutex_unlock(&adap->bus_lock); return ret; } else { dev_dbg(&adap->dev, "I2C level transfers not supported\n"); return -ENOSYS; } } 因为在这里的同步用的是mutex.首先判断判断是否充许睡眠,如果不允许,尝试获锁.如果获锁失败,则返回,这样的操作是避免进入睡眠,我们在后面也可以看到,实际的传输工作交给了adap->algo->master_xfer()完成. 在这里,我们终于把i2c_probe_address()的执行分析完了,经过这个分析,我们也知道了数据是怎么样传输的.我们接着i2c_probe()往下看.如果i2c_probe_address()成功.说明总线上确实有这样的设备.那么就会调用驱动中的回调函数.在ad7148的驱动中,如下所示: return i2c_probe(adapter, &addr_data, ad7418_detect); 也就是说,要调用的回调函数是ad7418_detect().这个函数中我们只分析和i2c框架相关的部份.代码片段如下所示: static int ad7418_detect(struct i2c_adapter *adapter, int address, int kind) { struct i2c_client *client; …… …… client->addr = address; client->adapter = adapter; client->driver = &ad7418_driver; i2c_set_clientdata(client, data); …… …… if ((err = i2c_attach_client(client))) goto exit_free; …… …… } 结合上面关于new-style形式的驱动分析.发现这里走的是同一个套路,即初始化了client.然后调用i2c_attach_client().后面的流程就跟上面分析的一样了.只不过,不相同的是,这里clinet已经指定了驱动为ad7418_driver.应该在注册clinet->dev之后,就不会走bus->match和bus->probe的流程了. 七:i2c dev节点操作 现在来分析上面架构图中的i2c-dev.c中的部份.这个部份为用户空间提供了操作adapter的接口.这部份代码其实对应就晃一个模块.它的初始化函数为: module_init(i2c_dev_init); i2c_dev_init()代码如下: static int __init i2c_dev_init(void) { int res; printk(KERN_INFO "i2c /dev entries driver\n"); res = register_chrdev(I2C_MAJOR, "i2c", &i2cdev_fops); if (res) goto out; i2c_dev_class = class_create(THIS_MODULE, "i2c-dev"); if (IS_ERR(i2c_dev_class)) goto out_unreg_chrdev; res = i2c_add_driver(&i2cdev_driver); if (res) goto out_unreg_class; return 0; out_unreg_class: class_destroy(i2c_dev_class); out_unreg_chrdev: unregister_chrdev(I2C_MAJOR, "i2c"); out: printk(KERN_ERR "%s: Driver Initialisation failed\n", __FILE__); return res; } res = i2c_add_driver(&i2cdev_driver); if (res) goto out_unreg_class; i2cdev_driver定义如下: static struct i2c_driver i2cdev_driver = { .driver = { .name = "dev_driver", }, .id = I2C_DRIVERID_I2CDEV, .attach_adapter = i2cdev_attach_adapter, .detach_adapter = i2cdev_detach_adapter, .detach_client = i2cdev_detach_client, }; 也就是说,当它注册或者有新的adapter注册后,就会它的attach_adapter()函数.该函数代码如下: static int i2cdev_attach_adapter(struct i2c_adapter *adap) { struct i2c_dev *i2c_dev; int res; i2c_dev = get_free_i2c_dev(adap); if (IS_ERR(i2c_dev)) return PTR_ERR(i2c_dev); /* register this i2c device with the driver core */ i2c_dev->dev = device_create(i2c_dev_class, &adap->dev, MKDEV(I2C_MAJOR, adap->nr), "i2c-%d", adap->nr); if (IS_ERR(i2c_dev->dev)) { res = PTR_ERR(i2c_dev->dev); goto error; } res = device_create_file(i2c_dev->dev, &dev_attr_name); if (res) goto error_destroy; pr_debug("i2c-dev: adapter [%s] registered as minor %d\n", adap->name, adap->nr); return 0; error_destroy: device_destroy(i2c_dev_class, MKDEV(I2C_MAJOR, adap->nr)); error: return_i2c_dev(i2c_dev); return res; } 这个函数也很简单,首先调用get_free_i2c_dev()分配并初始化了一个struct i2c_dev结构,使i2c_dev->adap指向操作的adapter.之后,该i2c_dev会被链入链表i2c_dev_list中.再分别以I2C_MAJOR, adap->nr为主次设备号创建了一个device.如果此时系统配置了udev或者是hotplug,那么就么在/dev下自动创建相关的设备节点了. 刚才我们说过,所有主设备号为I2C_MAJOR的设备节点的操作函数是i2cdev_fops.它的定义如下所示: static const struct file_operations i2cdev_fops = { .owner = THIS_MODULE, .llseek = no_llseek, .read = i2cdev_read, .write = i2cdev_write, .ioctl = i2cdev_ioctl, .open = i2cdev_open, .release = i2cdev_release, }; 7.1:i2c dev的open操作 Open操作对应的函数为i2cdev_open().代码如下: static int i2cdev_open(struct inode *inode, struct file *file) { unsigned int minor = iminor(inode); struct i2c_client *client; struct i2c_adapter *adap; struct i2c_dev *i2c_dev; //以次设备号从i2c_dev_list链表中取得i2c_dev i2c_dev = i2c_dev_get_by_minor(minor); if (!i2c_dev) return -ENODEV; //以apapter的总线号从i2c_adapter_idr中找到adapter adap = i2c_get_adapter(i2c_dev->adap->nr); if (!adap) return -ENODEV; /* This creates an anonymous i2c_client, which may later be * pointed to some address using I2C_SLAVE or I2C_SLAVE_FORCE. * * This client is ** NEVER REGISTERED ** with the driver model * or I2C core code!! It just holds private copies of addressing * information and maybe a PEC flag. */ //分配并初始化一个i2c_client结构 client = kzalloc(sizeof(*client), GFP_KERNEL); if (!client) { i2c_put_adapter(adap); return -ENOMEM; } snprintf(client->name, I2C_NAME_SIZE, "i2c-dev %d", adap->nr); client->driver = &i2cdev_driver; //clinet->adapter指向操作的adapter client->adapter = adap; //关联到file file->private_data = client; return 0; } 注意这里分配并初始化了一个struct i2c_client结构.但是没有注册这个clinet.此外,这个函数中还有一个比较奇怪的操作.不是在前面已经将i2c_dev->adap指向要操作的adapter么?为什么还要以adapter->nr为关键字从i2c_adapter_idr去找这个操作的adapter呢?注意了,调用i2c_get_adapter()从总线号nr找到操作的adapter的时候,还会增加module的引用计数.这样可以防止模块意外被释放掉.也许有人会有这样的疑问,那 i2c_dev->adap->nr操作,如果i2c_dev->adap被释放掉的话,不是一样会引起系统崩溃么?这里因为,在i2cdev_attach_adapter()间接的增加了一次adapter的一次引用计数.如下: tatic int i2cdev_attach_adapter(struct i2c_adapter *adap) { ．．．．．． i2c_dev->dev = device_create(i2c_dev_class, &adap->dev, MKDEV(I2C_MAJOR, adap->nr), "i2c-%d", adap->nr); ．．．．．． } 看到了么，i2c_dev内嵌的device是以adap->dev为父结点,在device_create()中会增次adap->dev的一次引用计数. 好了,open()操作到此就完成了. 7.2:read操作 Read操作对应的操作函数如下示: static ssize_t i2cdev_read (struct file *file, char __user *buf, size_t count, loff_t *offset) { char *tmp; int ret; struct i2c_client *client = (struct i2c_client *)file->private_data; if (count > 8192) count = 8192; tmp = kmalloc(count,GFP_KERNEL); if (tmp==NULL) return -ENOMEM; pr_debug("i2c-dev: i2c-%d reading %zd bytes.\n", iminor(file->f_path.dentry->d_inode), count); ret = i2c_master_recv(client,tmp,count); if (ret >= 0) ret = copy_to_user(buf,tmp,count)?-EFAULT:ret; kfree(tmp); return ret; } 首先从file结构中取得struct i2c_clinet.然后在kernel同分配相同长度的缓存区,随之调用i2c_master_recv()从设备中读取数据.再将读取出来的数据copy到用户空间中. I2c_master_recv()代码如下: int i2c_master_recv(struct i2c_client *client, char *buf ,int count) { struct i2c_adapter *adap=client->adapter; struct i2c_msg msg; int ret; msg.addr = client->addr; msg.flags = client->flags & I2C_M_TEN; msg.flags |= I2C_M_RD; msg.len = count; msg.buf = buf; ret = i2c_transfer(adap, &msg, 1); /* If everything went ok (i.e. 1 msg transmitted), return #bytes transmitted, else error code. */ return (ret == 1) ? count : ret; } 看完前面的代码之后,这个函数应该很简单了,就是为读操作初始化了一个i2c_msg.然后调用i2c_tanster().代码中的client->flags & I2C_M_TEN表示adapter是否采用10位寻址的方式.在这里就不再详细分析了. 另外,有人可能看出了一个问题.这里clinet->addr是从哪来的呢?对,在read之前应该还要有一步操作来设置clinet->addr的值.这个过程是ioctl的操作.ioctl可以设置PEC标志,重试次数,超时时间,和发送接收数据等,我们在这里只看一下clinet->addr的设置.代码片段如下示: static int i2cdev_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg) { ．．．．．． ．．．．．． switch ( cmd ) { case I2C_SLAVE: case I2C_SLAVE_FORCE: /* NOTE: devices set up to work with "new style" drivers * can't use I2C_SLAVE, even when the device node is not * bound to a driver. Only I2C_SLAVE_FORCE will work. * * Setting the PEC flag here won't affect kernel drivers, * which will be using the i2c_client node registered with * the driver model core. Likewise, when that client has * the PEC flag already set, the i2c-dev driver won't see * (or use) this setting. */ if ((arg > 0x3ff) || (((client->flags & I2C_M_TEN) == 0) && arg > 0x7f)) return -EINVAL; if (cmd == I2C_SLAVE && i2cdev_check_addr(client->adapter, arg)) return -EBUSY; /* REVISIT: address could become busy later */ client->addr = arg; return 0; ．．．．．． ．．．．．． } 由此可见,调用I2C_SLAVE或者I2C_SLAVE_FORCE的Ioctl就会设置clinet->addr.另外,注释中也说得很清楚了.如果是I2C_SLAVE的话,还会调用其所长i2cdev_check_addr().进行地址检查,如果adapter已经关联到这个地址的设备,就会检查失败. 7.2:write操作 Write操作如下所示: static ssize_t i2cdev_write (struct file *file, const char __user *buf, size_t count, loff_t *offset) { int ret; char *tmp; struct i2c_client *client = (struct i2c_client *)file->private_data; if (count > 8192) count = 8192; tmp = kmalloc(count,GFP_KERNEL); if (tmp==NULL) return -ENOMEM; if (copy_from_user(tmp,buf,count)) { kfree(tmp); return -EFAULT; } pr_debug("i2c-dev: i2c-%d writing %zd bytes.\n", iminor(file->f_path.dentry->d_inode), count); ret = i2c_master_send(client,tmp,count); kfree(tmp); return ret; } 该操作比较简单,就是将用户空间的数据发送到i2c 设备. 八:小结 在本节中,分析了i2c的框架设计.这个框架大体上沿用了Linux的设备驱动框架,不过之中又做了很多变通.在之后的分析中,会分别举一个adapter和i2c device的例子来详细描述一下有关i2c driver的设计
中财网