由于工作的原因，对SPI的理解最为深刻，也和SPI最有感情了，之前工作都是基于OSEK操作系统上进行实现，也在US/OS3上实现过SPI驱动的实现和测试，但是都是基于基本的寄存器操作，没有一个系统软件架构的思想，感觉linux SPI驱动很强大，水很深，废话少说，SPI总线上有两类设备：一类是主机端，通常作为SOC系统的一个子模块出现，比如很多嵌入式MPU中都常常包含SPI模块。一类是从机被控端，例如一些SPI接口的Flash、传感器等等。主机端是SPI总线的控制者，通过使用SPI协议主动发起SPI总线上的会话。而受控端则被动接受SPI主控端的指令，并作出响应的响应，本文读者前提是必须熟练掌握linux Platform总线驱动模型 和基本字符设备驱动的实现。
SPI总线由MISO（串行数据输入）、MOSI（串行数据输出）、SCK（串行移位时钟）、CS（使能信号）4个信号线组成。SPI常用四种数据传输模式，主要差别在于：输出串行同步时钟极性（CPOL）和相位（CPHA）可以进行配置。如果CPOL= 0，串行同步时钟的空闲状态为低电平；如果CPOL= 1，串行同步时钟的空闲状态为高电平。如果CPHA= 0，在串行同步时钟的前沿（上升或下降）数据被采样；如果CPHA = 1，在串行同步时钟的后沿（上升或下降）数据被采样。同I2C子系统类似，SPI子系统分为3个部分，分别是SPI核心层、主控制器驱动和协议驱动，通俗一点就是SPI核心层主要完成1.定义并注册SPI总线spi_bus_type和控制器类spi_master_class;2.提供spi_driver,spi_device和spi_master的分配，创建，注册和注销；3.实现SPI通信方法的上层代码。主控制器驱动对应I2C的适配器驱动，SPI用spi_master来描述相应的控制器，通常用spi_bitbang来控制实际的数据传输，功能非常类似与i2c_algorithm。协议驱动类比I2C设备驱动，可以理解成客户端驱动，下面详细分析和实现基于linux3.14.78内核版本 SPI在S3C2440/6410上驱动移植.
Step1，实现SPI控制器的设备接口（相对应有两种方式，一种实现方式是S3c2440纯粹通过配置来实现，另一种实现方式针对m25p10，独立编写单独的spi驱动模块）
首先针对S3C2440，SPI控制器的设备接口在drivers/spi/spidev.c中实现，下图为相应的软件流程图，spidev.c中的spidev_init()作为模块初始化函数，在系统启动或者模块加载是被调用，主要完成以下三种操作：
1、调用register_chrdev()为SPI控制器注册主设备号为153，次设备号范围为0~255，文件操作集合为spidev_fops的字符设备；
2、调用class_create（）注册一个名为“spidev”的设备类；
3、调用spi_register_driver（）向系统添加SPI控制器的设备驱动spidev_spi；
spi_register_driver（）将驱动spidev_spi添加到SPI核心层注册的spi_bus_type总线上，注意，该总线属于spi总线，Spi总线对应的总线类型为spi_bus_type，在内核的drivers/spi/spi.c中定义，对应的匹配规则是（高版本中的匹配规则会稍有变化，引入了id_table，可以匹配多个spi设备名称）详见以下代码。由于spi总线的匹配方式是检查spi_device.modalias与spi_driver.driver.name是否相同，而spidev_spi的driver.name是“spidev”,so 只有spi_bus_type总线上的modalias为“spidev”的设备，才可以与spidev_spi驱动匹配。由于S3C2440拥有两个SPI控制器，对应的平台信息是s3c_device_spi0和s3c_device_spi1,可以将其添加到机器配置文件的My2440_devices数组中，驱动中probe方法中用到的总线编号、片选总数等信息来源于平台数据，这些数据需要用户添加，添加方式有两种，平台设备的平台数据类型是S3C2410_SPI_info，在板级初始化文件中可以为两个SPI平台设备分别定义和添加这些平台数据，此外为了支持SPI控制器设备接口功能，还需要在机器配置文件为SPI控制器设备添加并注册spi_board_info对象，详见下面代码。
设备与驱动匹配后，通过调用spidev_spi的probe方法来绑定工作见代码。SPI控制器设备操作集合是spidev_fops，主要包括spidev_write,spidev_read,spidev_ioctl,spidev_open和spidev_release。下面的代码注释中详尽分析了spidev_read（）函数的实现过程，spidev_write（）的分析和实现方法与read类似，spidev_open（）会遍历device_list链表，找出其中设备号与打开设备文件的inode.i_rdev相等的spidev_data对象，并将该对象记录在文件的私有数据filp->private_data中，以供read和write函数获取，此外要实现全双工的传输需要借助控制器设备的ioctl方法，对应实现函数是spidev_ioctl()，全双工传输也是通过spidev_sync()完成的，与半双工传输唯一不同的是在消息的传输段中，tx_buffer和rx_buffer同时被设置，它还提供获取和修改时钟模式、子宽、最大时钟频率属性的命令。
1 static struct s3c2410_spi_info s3c2410_spi0_platdata = { 2 .pin_cs = S3C2410_GPG2, 3 .num_cs = 2, 4 .bus_num = 0, 5 .gpio_setup = s3c24xx_spi_gpiocfg_bus0_gpe11_12_13, 6 }; 7 static struct spi_board_info s3c2410_spi0_board[] = 8 { 9 [0] = { 10 .modalias = "spidev", 11 .bus_num = 0, 12 .chip_select = 0, 13 .max_speed_hz = 500 * 1000, 14 }, 15 [1] = { 16 .modalias = "at25", 17 .platform_data = &at25_eeprom_data, 18 .bus_num = 0, 19 .chip_select = 1, 20 .max_speed_hz = 500 * 1000, 21 } 22 }; 23 static struct s3c2410_spi_info s3c2410_spi1_platdata = { 24 .pin_cs = S3C2410_GPG3, 25 .num_cs = 1, 26 .bus_num = 1, 27 .gpio_setup = s3c24xx_spi_gpiocfg_bus1_gpg5_6_7, 28 }; 29 static struct spi_board_info s3c2410_spi1_board[] = 30 { 31 [0] = { 32 .modalias = "spidev", 33 .bus_num = 1, 34 .chip_select = 0, 35 .max_speed_hz = 500 * 1000, 36 } 37 }; 38 static void __init My2440_machine_init(void) 39 { 40 s3c24xx_fb_set_platdata(&My2440_fb_info); 41 s3c_i2c0_set_platdata(NULL); 42 i2c_register_board_info(0, i2c_devices, ARRAY_SIZE(i2c_devices)); 43 s3c_device_spi0.dev.platform_data= &s3c2410_spi0_platdata; 44 spi_register_board_info(s3c2410_spi0_board, ARRAY_SIZE(s3c2410_spi0_board)); 45 s3c_device_spi1.dev.platform_data= &s3c2410_spi1_platdata; 46 spi_register_board_info(s3c2410_spi1_board, ARRAY_SIZE(s3c2410_spi1_board)); 47 s3c_device_nand.dev.platform_data = &My2440_nand_info; 48 s3c_device_sdi.dev.platform_data = &My2440_mmc_cfg; 49 platform_add_devices(My2440_devices, ARRAY_SIZE(My2440_devices)); 50 }
1 struct bus_type spi_bus_type = { 2 .name = "spi", 3 .dev_attrs = spi_dev_attrs, 4 .match = spi_match_device, 5 .uevent = spi_uevent, 6 .suspend = spi_suspend, 7 .resume = spi_resume, 8 };
static int spi_match_device(struct device *dev, struct device_driver *drv) { const struct spi_device *spi = to_spi_device(dev); return strcmp(spi->modalias, drv->name) == 0; }
static int spidev_probe(struct spi_device *spi) { struct spidev_data *spidev; int status; unsigned long minor; /* Allocate driver data, 分配并初始化一个spidev_data对象，用来描述SPI控制器设备驱动操作的控制器设备，包含它的设备号，传输缓存和内嵌spi_device*/ spidev = kzalloc(sizeof(*spidev), GFP_KERNEL); if (!spidev) return -ENOMEM; /* Initialize the driver data ，将与spidev_spi驱动匹配的spi_device赋值给spidev_data.spi，通过find_first_zero_bit()从位图minors中分配到次设备号，
与主设备号SPIDEV_MAJOR构成设备号并赋值给spidev_data.devt,然后调用device_create（）在/dev下创建名称格式为“spidev%d%d”的字符设备文件，第一个%d表示控制器编号，
即spi_device.master.bus_num，第二个对应片选号，即spi_device.chip_select*/ spidev->spi = spi; spin_lock_init(&spidev->spi_lock); mutex_init(&spidev->buf_lock); INIT_LIST_HEAD(&spidev->device_entry); /* If we can allocate a minor number, hook up this device. * Reusing minors is fine so long as udev or mdev is working. */ mutex_lock(&device_list_lock); minor = find_first_zero_bit(minors, N_SPI_MINORS); if (minor < N_SPI_MINORS) { struct device *dev; spidev->devt = MKDEV(SPIDEV_MAJOR, minor); dev = device_create(spidev_class, &spi->dev, spidev->devt, spidev, "spidev%d.%d", spi->master->bus_num, spi->chip_select); status = PTR_ERR_OR_ZERO(dev); } else { dev_dbg(&spi->dev, "no minor number available!\n"); status = -ENODEV; } if (status == 0) { set_bit(minor, minors);
/*将spidev_data对象添加到一个名为device_list的设备列表中，最后将spidev_data对象设置为对应spi_device的驱动数据，将其值赋值给dpi_device.dev.driver_data*/ list_add(&spidev->device_entry, &device_list); } mutex_unlock(&device_list_lock); if (status == 0) spi_set_drvdata(spi, spidev); else kfree(spidev); return status; }
/*spidev_read(),spidev_write()实现的读写操作都是半双工的传输，要实现全双工传输需要调用spidev_ioctl（）操作，在spidev_read（）中，通过调用spidev_sync_read()
同步读取指定长度数据到spidev_data_buffer中，然后调用copy_to_user()将读取的数据复制到用户空间缓冲*/
static ssize_t spidev_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos) { struct spidev_data *spidev; ssize_t status = 0; /* chipselect only toggles at start or end of operation */ if (count > bufsiz) return -EMSGSIZE; spidev = filp->private_data; mutex_lock(&spidev->buf_lock); status = spidev_sync_read(spidev, count); if (status > 0) { unsigned long missing; missing = copy_to_user(buf, spidev->buffer, status); if (missing == status) status = -EFAULT; else status = status - missing; } mutex_unlock(&spidev->buf_lock); return status; }
/*spidev_sync_read（）中定义一个spi_transfer和spi_message对象，使用spidev_data.buffer初始化spi_transfer.rx_buf，
然后将spi_transfer对象添加到spi_message中，调用spidev_sync()传输信息；*/
spidev_sync_read(struct spidev_data *spidev, size_t len)
{ struct spi_transfer t = { .rx_buf = spidev->buffer, .len = len, }; struct spi_message m; spi_message_init(&m); spi_message_add_tail(&t, &m); return spidev_sync(spidev, &m); }
/*spidev_sync()实现同步传输的机制：首先为传入的消息对象的complete和context成员赋值，context被设置为完成接口done，complete被设置成spidev_complete（），然后
调用spi_async()启动异步传输，spi_async()返回后，通过调用wait_for_completion（）阻塞调用进程，直到异步传输完成，spi_message的complete方法，即spidev_complete（）
被调用来唤醒被阻塞的进程，异步传输接口函数spi_async()属于SPI核心层，在include/linux/spi/spi.h中定义并实现，它只调用了spi_device所属控制器的transfer方法*/
static ssize_t spidev_sync(struct spidev_data *spidev, struct spi_message *message) { DECLARE_COMPLETION_ONSTACK(done); int status; message->complete = spidev_complete; message->context = &done; spin_lock_irq(&spidev->spi_lock); if (spidev->spi == NULL) status = -ESHUTDOWN; else status = spi_async(spidev->spi, message); spin_unlock_irq(&spidev->spi_lock); if (status == 0) { wait_for_completion(&done); status = message->status; if (status == 0) status = message->actual_length; } return status; }
针对FS_S5PC100上的M25P10芯片，
//spi_device对应的含义是挂接在spi总线上的一个设备，所以描述它的时候应该明确它自身的设备特性、传输要求、及挂接在哪个总线上 static struct spi_board_info s3c_spi_devs[] __initdata = { { .modalias = "m25p10", .mode = SPI_MODE_0, //CPOL=0, CPHA=0 此处选择具体数据传输模式 .max_speed_hz = 10000000, //最大的spi时钟频率 /* Connected to SPI-0 as 1st Slave */ .bus_num = 0, //设备连接在spi控制器0上 .chip_select = 0, //片选线号，在S5PC100的控制器驱动中没有使用它作为片选的依据，而是选择了下文controller_data里的方法 .controller_data = &smdk_spi0_csi[0], }, }; static struct s3c64xx_spi_csinfo smdk_spi0_csi[] = { [0] = { .set_level = smdk_m25p10_cs_set_level, .fb_delay = 0x3, }, }; static void smdk_m25p10_cs_set_level(int high) //spi控制器会用这个方法设置cs { u32 val; val = readl(S5PC1XX_GPBDAT); if (high) val |= (1<<3); else val &= ~(1<<3); writel(val, S5PC1XX_GPBDAT); } spi_register_board_info(s3c_spi_devs, ARRAY_SIZE(s3c_spi_devs));//注册spi_board_info,这个代码会把spi_board_info注册要链表board_list上。 //spi_master的注册会在spi_register_board_info之后，spi_master注册的过程中会调用scan_boardinfo扫描board_list，找到挂接在它上面的spi设备，然后创建并注册spi_device。 static void scan_boardinfo(struct spi_master *master) { struct boardinfo *bi; mutex_lock(&board_lock); list_for_each_entry(bi, &board_list, list) { struct spi_board_info *chip = bi->board_info; unsigned n; for (n = bi->n_board_info; n > 0; n--, chip++) { if (chip->bus_num != master->bus_num) continue; /* NOTE: this relies on spi_new_device to * issue diagnostics when given bogus inputs */ (void) spi_new_device(master, chip); //创建并注册了spi_device } } mutex_unlock(&board_lock); } //spi_driver.c linux内核中的/driver/mtd/devices/m25p80.c驱动为参考 static struct spi_driver m25p80_driver = { //spi_driver的构建 .driver = { .name = "m25p80", .bus = &spi_bus_type, .owner = THIS_MODULE, }, .probe = m25p_probe, .remove = __devexit_p(m25p_remove), */ }; spi_register_driver(&m25p80_driver);//spi driver的注册 //在有匹配的spi device时，会调用m25p_probe,根据传入的spi_device参数，可以找到对应的spi_master。接下来就可以利用spi子系统为我们完成数据交互了 static int __devinit m25p_probe(struct spi_device *spi)
Step2，实现SPI platform 总线匹配
S3C2440和S5PC100的SPI控制器驱动的实现同样采用了Platform驱动模型，针对S3C2440在drivers/spi/spi_s3c24xx.c中，模块初始化函数s3c24xx_spi_init()完成了platform驱动的注册，名称是“s3c2410-spi”,与该驱动匹配的平台设备在arch/arm/plat-samsung/devs.c中定义，并且与驱动同名，针对S5PC100，详见下面代码,匹配完成后调用相应的probe：
//Platform_device，SPI控制器对应platform_device的定义方式，同样以S5PC100中的SPI控制器为例，参看arch/arm/plat-s5pc1xx/dev-spi.c文件 struct platform_device s3c_device_spi0 = { .name = "s3c64xx-spi", //名称，要和Platform_driver匹配 .id = 0, //第0个控制器，S5PC100中有3个控制器 .num_resources = ARRAY_SIZE(s5pc1xx_spi0_resource), //占用资源的种类 .resource = s5pc1xx_spi0_resource, //指向资源结构数组的指针 .dev = { .dma_mask = &spi_dmamask, //dma寻址范围 .coherent_dma_mask = DMA_BIT_MASK(32), //可以通过关闭cache等措施保证一致性的dma寻址范围 .platform_data = &s5pc1xx_spi0_pdata, //特殊的平台数据，参看后文 }, }; static struct s3c64xx_spi_cntrlr_info s5pc1xx_spi0_pdata = { .cfg_gpio = s5pc1xx_spi_cfg_gpio, //用于控制器管脚的IO配置 .fifo_lvl_mask = 0x7f, .rx_lvl_offset = 13, }; static int s5pc1xx_spi_cfg_gpio(struct platform_device *pdev) { switch (pdev->id) { case 0: s3c_gpio_cfgpin(S5PC1XX_GPB(0), S5PC1XX_GPB0_SPI_MISO0); s3c_gpio_cfgpin(S5PC1XX_GPB(1), S5PC1XX_GPB1_SPI_CLK0); s3c_gpio_cfgpin(S5PC1XX_GPB(2), S5PC1XX_GPB2_SPI_MOSI0); s3c_gpio_setpull(S5PC1XX_GPB(0), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPB(1), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPB(2), S3C_GPIO_PULL_UP); break; case 1: s3c_gpio_cfgpin(S5PC1XX_GPB(4), S5PC1XX_GPB4_SPI_MISO1); s3c_gpio_cfgpin(S5PC1XX_GPB(5), S5PC1XX_GPB5_SPI_CLK1); s3c_gpio_cfgpin(S5PC1XX_GPB(6), S5PC1XX_GPB6_SPI_MOSI1); s3c_gpio_setpull(S5PC1XX_GPB(4), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPB(5), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPB(6), S3C_GPIO_PULL_UP); break; case 2: s3c_gpio_cfgpin(S5PC1XX_GPG3(0), S5PC1XX_GPG3_0_SPI_CLK2); s3c_gpio_cfgpin(S5PC1XX_GPG3(2), S5PC1XX_GPG3_2_SPI_MISO2); s3c_gpio_cfgpin(S5PC1XX_GPG3(3), S5PC1XX_GPG3_3_SPI_MOSI2); s3c_gpio_setpull(S5PC1XX_GPG3(0), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPG3(2), S3C_GPIO_PULL_UP); s3c_gpio_setpull(S5PC1XX_GPG3(3), S3C_GPIO_PULL_UP); break; default: dev_err(&pdev->dev, "Invalid SPI Controller number!"); return -EINVAL; } //platform_driver，参看drivers/spi/spi_s3c64xx.c文件 static struct platform_driver s3c64xx_spi_driver = { .driver = { .name = "s3c64xx-spi", //名称，和platform_device对应 .owner = THIS_MODULE, }, .remove = s3c64xx_spi_remove, .suspend = s3c64xx_spi_suspend, .resume = s3c64xx_spi_resume, }; platform_driver_probe(&s3c64xx_spi_driver, s3c64xx_spi_probe)；//注册s3c64xx_spi_driver //和平台中注册的platform_device匹配后，调用s3c64xx_spi_probe。然后根据传入的platform_device参数，构建一个用于描述SPI控制器的结构体spi_master，并注册。spi_register_master(master)。后续注册的spi_device需要选定自己的spi_master，并利用spi_master提供的传输功能传输spi数据。和I2C类似，SPI也有一个描述控制器的对象叫spi_master,其主要成员是主机控制器的序号（系统中可能存在多个SPI主机控制器）、片选数量、SPI模式和时钟设置用到的函数、数据传输用到的函数等; struct spi_master { struct device dev; s16 bus_num; //表示是SPI主机控制器的编号。由平台代码决定 u16 num_chipselect; //控制器支持的片选数量，即能支持多少个spi设备 int (*setup)(struct spi_device *spi); //针对设备设置SPI的工作时钟及数据传输模式等。在spi_add_device函数中调用。 int (*transfer)(struct spi_device *spi, struct spi_message *mesg); //实现数据的双向传输，可能会睡眠 void (*cleanup)(struct spi_device *spi); //注销时调用 };
static int s3c24xx_spi_probe(struct platform_device *pdev) { struct s3c2410_spi_info *pdata; struct s3c24xx_spi *hw; struct spi_master *master; struct resource *res; int err = 0; master = spi_alloc_master(&pdev->dev, sizeof(struct s3c24xx_spi));//分配SPI控制器结构及驱动私有数据 if (master == NULL) { dev_err(&pdev->dev, "No memory for spi_master\n"); return -ENOMEM; } hw = spi_master_get_devdata(master); memset(hw, 0, sizeof(struct s3c24xx_spi)); hw->master = master; hw->pdata = pdata = dev_get_platdata(&pdev->dev); hw->dev = &pdev->dev; if (pdata == NULL) { dev_err(&pdev->dev, "No platform data supplied\n"); err = -ENOENT; goto err_no_pdata; } platform_set_drvdata(pdev, hw);//将SPI控制器私有数据作为平台设备驱动数据，便于通过相应接口获得 init_completion(&hw->done);//初始化完成接口 /* initialise fiq handler */ s3c24xx_spi_initfiq(hw); /* setup the master state. */ /* the spi->mode bits understood by this driver: */ master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH; master->num_chipselect = hw->pdata->num_cs;//设置控制器片选总数和总线编号 master->bus_num = pdata->bus_num; /* setup the state for the bitbang driver 初始化bitbang驱动的相关成员*/ hw->bitbang.master = hw->master; hw->bitbang.setup_transfer = s3c24xx_spi_setupxfer; hw->bitbang.chipselect = s3c24xx_spi_chipsel; hw->bitbang.txrx_bufs = s3c24xx_spi_txrx; hw->master->setup = s3c24xx_spi_setup; hw->master->cleanup = s3c24xx_spi_cleanup; dev_dbg(hw->dev, "bitbang at %p\n", &hw->bitbang); /* find and map our resources 这是平台驱动中最常规的工作，找到并映射资源*/ res = platform_get_resource(pdev, IORESOURCE_MEM, 0); hw->regs = devm_ioremap_resource(&pdev->dev, res); if (IS_ERR(hw->regs)) { err = PTR_ERR(hw->regs); goto err_no_pdata; } hw->irq = platform_get_irq(pdev, 0); if (hw->irq < 0) { dev_err(&pdev->dev, "No IRQ specified\n"); err = -ENOENT; goto err_no_pdata; } err = devm_request_irq(&pdev->dev, hw->irq, s3c24xx_spi_irq, 0, pdev->name, hw); if (err) { dev_err(&pdev->dev, "Cannot claim IRQ\n"); goto err_no_pdata; } hw->clk = devm_clk_get(&pdev->dev, "spi");//获取SPI的时钟资源 if (IS_ERR(hw->clk)) { dev_err(&pdev->dev, "No clock for device\n"); err = PTR_ERR(hw->clk); goto err_no_pdata; } /* setup any gpio we can 设置片选方法并配置片选引脚 */ if (!pdata->set_cs) { if (pdata->pin_cs < 0) { dev_err(&pdev->dev, "No chipselect pin\n"); err = -EINVAL; goto err_register; } err = devm_gpio_request(&pdev->dev, pdata->pin_cs, dev_name(&pdev->dev)); if (err) { dev_err(&pdev->dev, "Failed to get gpio for cs\n"); goto err_register; } hw->set_cs = s3c24xx_spi_gpiocs; gpio_direction_output(pdata->pin_cs, 1); } else hw->set_cs = pdata->set_cs; s3c24xx_spi_initialsetup(hw);//使能SPI时钟，初始化2440 SPI控制器寄存器及片选引脚等 /* register our spi controller 注册SPI控制器，进而完成对控制器对象的分配初始化和注册，注册控制器会扫描
board_list链表，从中取出spi_board_info创建spi_device设备，因此在执行probe方法时，只要board_list链表
上有控制器对应的spi_board_info，就能创建出控制器的设备对象，并与控制器设备驱动匹配，进而创建出用于访问该控制器的设备文件
*/ err = spi_bitbang_start(&hw->bitbang); if (err) { dev_err(&pdev->dev, "Failed to register SPI master\n"); goto err_register; } return 0; err_register: clk_disable(hw->clk); err_no_pdata: spi_master_put(hw->master); return err; }
Step3，分析实现SPI总线通信方法
SPI控制器通常由spi_bitbang来完成实际数据的数据传输，spi_bitbang的定义如下：
struct spi_bitbang { spinlock_t lock; /*操作工作队列时使用的自旋锁*/ u8 busy; u8 use_dma; u8 flags; /* extra spi->mode support */ struct spi_master *master; /* setup_transfer() changes clock and/or wordsize to match settings * for this transfer; zeroes restore defaults from spi_device.为特定的传输设置时钟、字宽等属性的方法 */ int (*setup_transfer)(struct spi_device *spi, struct spi_transfer *t); void (*chipselect)(struct spi_device *spi, int is_on); #define BITBANG_CS_ACTIVE 1 /* normally nCS, active low */ #define BITBANG_CS_INACTIVE 0 /* txrx_bufs()为实际的传输方法 may handle dma mapping for transfers that don't * already have one (transfer.{tx,rx}_dma is zero), or use PIO */ int (*txrx_bufs)(struct spi_device *spi, struct spi_transfer *t); /* txrx_word[SPI_MODE_*]()按字传输的方法 just looks like a shift register */ u32 (*txrx_word[4])(struct spi_device *spi, unsigned nsecs, u32 word, u8 bits); };
两者的SPI总线通信控制方法不同，在上面spi_bitbang结构可以看出，消息传输启动利用的是spi_bibang_txrx_bufs(),在2440 SPI驱动中是通过s3c24xx_spi_txrx()来实现，他将传输段中请求数据的缓冲区和长度赋值给s3c24xx_spi对象的相关成员，然后主动发送一个字节数据，启动传输过程，之后使用进程阻塞在完成接口s3c24xx_spi.done上，其他数据的传输将在中断处理函数中完成，中断检查到发送完成后，通知完成接口，唤醒之前阻塞在其上的进程（这里特别说明下6410SPI通信方法，与2440不同，采用DMA传输，相应的实现函数为s3c64xx_spi_transfer(),利用工作队列进行延迟调度），m25p80SPI总线通信控制详尽代码如下，说明一点，spi_message的消息可以同步传输也可以异步传输，正常产品中基本上都是异步传输，传输结束时有spi_message.complete()方法完成，下面的例子以同步实现简单实现：
#include <linux/platform_device.h> #include <linux/spi/spi.h> #include <linux/init.h> #include <linux/module.h> #include <linux/device.h> #include <linux/interrupt.h> #include <linux/mutex.h> #include <linux/slab.h> // kzalloc #include <linux/delay.h> #define FLASH_PAGE_SIZE 256 /* Flash Operating Commands */ #define CMD_READ_ID 0x9f #define CMD_WRITE_ENABLE 0x06 #define CMD_BULK_ERASE 0xc7 #define CMD_READ_BYTES 0x03 #define CMD_PAGE_PROGRAM 0x02 #define CMD_RDSR 0x05 /* Status Register bits. */ #define SR_WIP 1 /* Write in progress */ #define SR_WEL 2 /* Write enable latch */ /* ID Numbers */ #define MANUFACTURER_ID 0x20 #define DEVICE_ID 0x1120 /* Define max times to check status register before we give up. */ #define MAX_READY_WAIT_COUNT 100000 #define CMD_SZ 4 struct m25p10a { struct spi_device *spi; struct mutex lock; char erase_opcode; char cmd[ CMD_SZ ]; }; /* * Internal Helper functions */ /* * Read the status register, returning its value in the location * Return the status register value. * Returns negative if error occurred. */ static int read_sr(struct m25p10a *flash) { ssize_t retval; u8 code = CMD_RDSR; u8 val; retval = spi_write_then_read(flash->spi, &code, 1, &val, 1); if (retval < 0) { dev_err(&flash->spi->dev, "error %d reading SR\n", (int) retval); return retval; } return val; } /* * Service routine to read status register until ready, or timeout occurs. * Returns non-zero if error. */ static int wait_till_ready(struct m25p10a *flash) { int count; int sr; /* one chip guarantees max 5 msec wait here after page writes, * but potentially three seconds (!) after page erase. */ for (count = 0; count < MAX_READY_WAIT_COUNT; count++) { if ((sr = read_sr(flash)) < 0) break; else if (!(sr & SR_WIP)) return 0; /* REVISIT sometimes sleeping would be best */ } printk( "in (%s): count = %d\n", count ); return 1; } /* * Set write enable latch with Write Enable command. * Returns negative if error occurred. */ static inline int write_enable( struct m25p10a *flash ) { flash->cmd[0] = CMD_WRITE_ENABLE; return spi_write( flash->spi, flash->cmd, 1 ); } /* * Erase the whole flash memory * * Returns 0 if successful, non-zero otherwise. */ static int erase_chip( struct m25p10a *flash ) { /* Wait until finished previous write command. */ if (wait_till_ready(flash)) return -1; /* Send write enable, then erase commands. */ write_enable( flash ); flash->cmd[0] = CMD_BULK_ERASE; return spi_write( flash->spi, flash->cmd, 1 ); } /* * Read an address range from the flash chip. The address range * may be any size provided it is within the physical boundaries. */ static int m25p10a_read( struct m25p10a *flash, loff_t from, size_t len, char *buf ) { int r_count = 0, i; flash->cmd[0] = CMD_READ_BYTES; flash->cmd[1] = from >> 16; flash->cmd[2] = from >> 8; flash->cmd[3] = from; #if 1 struct spi_transfer st[2]; struct spi_message msg; spi_message_init( &msg ); memset( st, 0, sizeof(st) ); flash->cmd[0] = CMD_READ_BYTES; flash->cmd[1] = from >> 16; flash->cmd[2] = from >> 8; flash->cmd[3] = from; st[ 0 ].tx_buf = flash->cmd; st[ 0 ].len = CMD_SZ; spi_message_add_tail( &st[0], &msg ); st[ 1 ].rx_buf = buf; st[ 1 ].len = len; spi_message_add_tail( &st[1], &msg ); mutex_lock( &flash->lock ); /* Wait until finished previous write command. */ if (wait_till_ready(flash)) { mutex_unlock( &flash->lock ); return -1; } spi_sync( flash->spi, &msg ); r_count = msg.actual_length - CMD_SZ; printk( "in (%s): read %d bytes\n", __func__, r_count ); for( i = 0; i < r_count; i++ ) { printk( "0x%02x\n", buf[ i ] ); } mutex_unlock( &flash->lock ); #endif return 0; } /* * Write an address range to the flash chip. Data must be written in * FLASH_PAGE_SIZE chunks. The address range may be any size provided * it is within the physical boundaries. */ static int m25p10a_write( struct m25p10a *flash, loff_t to, size_t len, const char *buf ) { int w_count = 0, i, page_offset; struct spi_transfer st[2]; struct spi_message msg; #if 1 if (wait_till_ready(flash)) { //读状态，等待ready mutex_unlock( &flash->lock ); return -1; } #endif write_enable( flash ); //写使能 spi_message_init( &msg ); memset( st, 0, sizeof(st) ); flash->cmd[0] = CMD_PAGE_PROGRAM; flash->cmd[1] = to >> 16; flash->cmd[2] = to >> 8; flash->cmd[3] = to; st[ 0 ].tx_buf = flash->cmd; st[ 0 ].len = CMD_SZ; spi_message_add_tail( &st[0], &msg ); st[ 1 ].tx_buf = buf; st[ 1 ].len = len; spi_message_add_tail( &st[1], &msg ); mutex_lock( &flash->lock ); /* get offset address inside a page */ page_offset = to % FLASH_PAGE_SIZE; /* do all the bytes fit onto one page? */ if( page_offset + len <= FLASH_PAGE_SIZE ) { // yes st[ 1 ].len = len; printk("%d, cmd = %d\n", st[ 1 ].len, *(char *)st[0].tx_buf); //while(1) { spi_sync( flash->spi, &msg ); } w_count = msg.actual_length - CMD_SZ; } else { // no } printk( "in (%s): write %d bytes to flash in total\n", __func__, w_count ); mutex_unlock( &flash->lock ); return 0; } static int check_id( struct m25p10a *flash ) { char buf[10] = {0}; flash->cmd[0] = CMD_READ_ID; spi_write_then_read( flash->spi, flash->cmd, 1, buf, 3 ); printk( "Manufacture ID: 0x%x\n", buf[0] ); printk( "Device ID: 0x%x\n", buf[1] | buf[2] << 8 ); return buf[2] << 16 | buf[1] << 8 | buf[0]; } static int m25p10a_probe(struct spi_device *spi) { int ret = 0; struct m25p10a *flash; char buf[ 256 ]; printk( "%s was called\n", __func__ ); flash = kzalloc( sizeof(struct m25p10a), GFP_KERNEL ); if( !flash ) { return -ENOMEM; } flash->spi = spi; mutex_init( &flash->lock ); /* save flash as driver's private data */ spi_set_drvdata( spi, flash ); check_id( flash ); //读取ID #if 1 ret = erase_chip( flash ); //擦除 if( ret < 0 ) { printk( "erase the entirely chip failed\n" ); } printk( "erase the whole chip done\n" ); memset( buf, 0x7, 256 ); m25p10a_write( flash, 0, 20, buf); //0地址写入20个7 memset( buf, 0, 256 ); m25p10a_read( flash, 0, 25, buf ); //0地址读出25个数 #endif return 0; } static int m25p10a_remove(struct spi_device *spi) { return 0; } static struct spi_driver m25p10a_driver = { .probe = m25p10a_probe, .remove = m25p10a_remove, .driver = { .name = "m25p10a", }, }; static int __init m25p10a_init(void) { return spi_register_driver(&m25p10a_driver); } static void __exit m25p10a_exit(void) { spi_unregister_driver(&m25p10a_driver); } module_init(m25p10a_init); module_exit(m25p10a_exit); MODULE_DESCRIPTION("m25p10a driver for FS_S5PC100"); MODULE_LICENSE("GPL");
参考文献 1.《深入Linux内核架构》 2.《Linux设备驱动开发详解》