linux终端设备uart驱动分析

一:前言

接着前面的终端控制台分析,接下来分析serial的驱动.在linux中,serial也对应着终端,通常被称为串口终端.在shell上,我们看到的/dev/ttyS*就是串口终端所对应的设备节点.

在分析具体的serial驱动之前.有必要先分析uart驱动架构.uart是Universal Asynchronous Receiver and Transmitter的缩写.翻译成中文即为”通用异步收发器”.它是串口设备驱动的封装层.

二:uart驱动架构概貌

 

uart设备是继tty_driver的又一层封装.实际上uart_driver就是对应tty_driver.在它的操作函数中,将操作转入uart_port.

在写操作的时候,先将数据放入一个叫做circ_buf的环形缓存区.然后uart_port从缓存区中取数据,将其写入到串口设备中.

当uart_port从serial设备接收到数据时,会将设备放入对应line discipline的缓存区中.

这样.用户在编写串口驱动的时候,只先要注册一个uart_driver.它的主要作用是定义设备节点号.然后将对设备的各项操作封装在uart_port.驱动工程师没必要关心上层的流程,只需按硬件规范将uart_port中的接口函数完成就可以了.

 

三:uart驱动中重要的数据结构及其关联

我们可以自己考虑下,基于上面的架构代码应该要怎么写.首先考虑以下几点:

1: 一个uart_driver通常会注册一段设备号.即在用户空间会看到uart_driver对应有多个设备节点.例如:

/dev/ttyS0  /dev/ttyS1

每个设备节点是对应一个具体硬件的,从上面的架构来看,每个设备文件应该对应一个uart_port.

也就是说:uart_device怎么同多个uart_port关系起来?怎么去区分操作的是哪一个设备文件?

 

2:每个uart_port对应一个circ_buf,所以uart_port必须要和这个缓存区关系起来

 

回忆tty驱动架构中.tty_driver有一个叫成员指向一个数组,即tty->ttys.每个设备文件对应设数组中的一项.而这个数组所代码的数据结构为tty_struct. 相应的tty_struct会将tty_driver和ldisc关联起来.

那在uart驱动中,是否也可用相同的方式来处理呢?

将uart驱动常用的数据结构表示如下:

 

结合上面提出的疑问.可以很清楚的看懂这些结构的设计.

 

四:uart_driver的注册操作

Uart_driver注册对应的函数为: uart_register_driver()代码如下:

int uart_register_driver(struct uart_driver *drv)

{

     struct tty_driver *normal = NULL;

     int i, retval;

 

     BUG_ON(drv->state);

 

     /*

      * Maybe we should be using a slab cache for this, especially if

      * we have a large number of ports to handle.

      */

     drv->state = kzalloc(sizeof(struct uart_state) * drv->nr, GFP_KERNEL);

     retval = -ENOMEM;

     if (!drv->state)

         goto out;

 

     normal  = alloc_tty_driver(drv->nr);

     if (!normal)

         goto out;

 

     drv->tty_driver = normal;

 

     normal->owner      = drv->owner;

     normal->driver_name    = drv->driver_name;

     normal->name       = drv->dev_name;

     normal->major      = drv->major;

     normal->minor_start    = drv->minor;

     normal->type       = TTY_DRIVER_TYPE_SERIAL;

     normal->subtype        = SERIAL_TYPE_NORMAL;

     normal->init_termios   = tty_std_termios;

     normal->init_termios.c_cflag = B9600 | CS8 | CREAD | HUPCL | CLOCAL;

     normal->init_termios.c_ispeed = normal->init_termios.c_ospeed = 9600;

     normal->flags      = TTY_DRIVER_REAL_RAW | TTY_DRIVER_DYNAMIC_DEV;

     normal->driver_state    = drv;

     tty_set_operations(normal, &uart_ops);

 

     /*

      * Initialise the UART state(s).

      */

     for (i = 0; i < drv->nr; i++) {

         struct uart_state *state = drv->state + i;

 

         state->close_delay     = 500;    /* .5 seconds */

         state->closing_wait    = 30000;  /* 30 seconds */

 

         mutex_init(&state->mutex);

     }

 

     retval = tty_register_driver(normal);

 out:

     if (retval < 0) {

         put_tty_driver(normal);

         kfree(drv->state);

     }

     return retval;

}

从上面代码可以看出.uart_driver中很多数据结构其实就是tty_driver中的.将数据转换为tty_driver之后,注册tty_driver.然后初始化uart_driver->state的存储空间.

这样,就会注册uart_driver->nr个设备节点.主设备号为uart_driver-> major. 开始的次设备号为uart_driver-> minor.

值得注意的是.在这里将tty_driver的操作集统一设为了uart_ops.其次,在tty_driver-> driver_state保存了这个uart_driver.这样做是为了在用户空间对设备文件的操作时,很容易转到对应的uart_driver.

另外:tty_driver的flags成员值为: TTY_DRIVER_REAL_RAW | TTY_DRIVER_DYNAMIC_DEV.里面包含有TTY_DRIVER_DYNAMIC_DEV标志.结合之前对tty的分析.如果包含有这个标志,是不会在初始化的时候去注册device.也就是说在/dev/下没有动态生成结点(如果是/dev下静态创建了这个结点就另当别论了^_^).

 

五: uart_add_one_port()操作

在前面提到.在对uart设备文件过程中.会将操作转换到对应的port上,这个port跟uart_driver是怎么关联起来的呢?这就是uart_add_ont_port()的主要工作了.

顾名思义,这个函数是在uart_driver增加一个port.代码如下:

int uart_add_one_port(struct uart_driver *drv, struct uart_port *port)

{

     struct uart_state *state;

     int ret = 0;

     struct device *tty_dev;

 

     BUG_ON(in_interrupt());

 

     if (port->line >= drv->nr)

         return -EINVAL;

 

     state = drv->state + port->line;

 

     mutex_lock(&port_mutex);

     mutex_lock(&state->mutex);

     if (state->port) {

         ret = -EINVAL;

         goto out;

     }

 

     state->port = port;

     state->pm_state = -1;

 

     port->cons = drv->cons;

     port->info = state->info;

 

     /*

      * If this port is a console, then the spinlock is already

      * initialised.

      */

     if (!(uart_console(port) && (port->cons->flags & CON_ENABLED))) {

         spin_lock_init(&port->lock);

         lockdep_set_class(&port->lock, &port_lock_key);

     }

 

     uart_configure_port(drv, state, port);

 

     /*

      * Register the port whether it's detected or not.  This allows

      * setserial to be used to alter this ports parameters.

      */

     tty_dev = tty_register_device(drv->tty_driver, port->line, port->dev);

     if (likely(!IS_ERR(tty_dev))) {

         device_can_wakeup(tty_dev) = 1;

         device_set_wakeup_enable(tty_dev, 0);

     } else

         printk(KERN_ERR "Cannot register tty device on line %d\n",

                port->line);

 

     /*

      * Ensure UPF_DEAD is not set.

      */

     port->flags &= ~UPF_DEAD;

 

 out:

     mutex_unlock(&state->mutex);

     mutex_unlock(&port_mutex);

 

     return ret;

}

首先这个函数不能在中断环境中使用. Uart_port->line就是对uart设备文件序号.它对应的也就是uart_driver->state数组中的uart_port->line项.

它主要初始化对应uart_driver->state项.接着调用uart_configure_port()进行port的自动配置.然后注册tty_device.如果用户空间运行了udev或者已经配置好了hotplug.就会在/dev下自动生成设备文件了.

 

六:设备节点的open操作

在用户空间执行open操作的时候,就会执行uart_ops->open. Uart_ops的定义如下:

static const struct tty_operations uart_ops = {

     .open         = uart_open,

     .close        = uart_close,

     .write        = uart_write,

     .put_char = uart_put_char,

     .flush_chars  = uart_flush_chars,

     .write_room   = uart_write_room,

     .chars_in_buffer= uart_chars_in_buffer,

     .flush_buffer = uart_flush_buffer,

     .ioctl        = uart_ioctl,

     .throttle = uart_throttle,

     .unthrottle   = uart_unthrottle,

     .send_xchar   = uart_send_xchar,

     .set_termios  = uart_set_termios,

     .stop         = uart_stop,

     .start        = uart_start,

     .hangup       = uart_hangup,

     .break_ctl    = uart_break_ctl,

     .wait_until_sent= uart_wait_until_sent,

#ifdef CONFIG_PROC_FS

     .read_proc    = uart_read_proc,

#endif

     .tiocmget = uart_tiocmget,

     .tiocmset = uart_tiocmset,

};

对应open的操作接口为uart_open.代码如下:

static int uart_open(struct tty_struct *tty, struct file *filp)

{

     struct uart_driver *drv = (struct uart_driver *)tty->driver->driver_state;

     struct uart_state *state;

     int retval, line = tty->index;

 

     BUG_ON(!kernel_locked());

     pr_debug("uart_open(%d) called\n", line);

 

     /*

      * tty->driver->num won't change, so we won't fail here with

      * tty->driver_data set to something non-NULL (and therefore

      * we won't get caught by uart_close()).

      */

     retval = -ENODEV;

     if (line >= tty->driver->num)

         goto fail;

 

     /*

      * We take the semaphore inside uart_get to guarantee that we won't

      * be re-entered while allocating the info structure, or while we

      * request any IRQs that the driver may need.  This also has the nice

      * side-effect that it delays the action of uart_hangup, so we can

      * guarantee that info->tty will always contain something reasonable.

      */

     state = uart_get(drv, line);

     if (IS_ERR(state)) {

         retval = PTR_ERR(state);

         goto fail;

     }

 

     /*

      * Once we set tty->driver_data here, we are guaranteed that

      * uart_close() will decrement the driver module use count.

      * Any failures from here onwards should not touch the count.

      */

     tty->driver_data = state;

     tty->low_latency = (state->port->flags & UPF_LOW_LATENCY) ? 1 : 0;

     tty->alt_speed = 0;

     state->info->tty = tty;

 

     /*

      * If the port is in the middle of closing, bail out now.

      */

     if (tty_hung_up_p(filp)) {

         retval = -EAGAIN;

         state->count--;

         mutex_unlock(&state->mutex);

         goto fail;

     }

 

     /*

      * Make sure the device is in D0 state.

      */

     if (state->count == 1)

         uart_change_pm(state, 0);

 

     /*

      * Start up the serial port.

      */

     retval = uart_startup(state, 0);

 

     /*

      * If we succeeded, wait until the port is ready.

      */

     if (retval == 0)

         retval = uart_block_til_ready(filp, state);

     mutex_unlock(&state->mutex);

 

     /*

      * If this is the first open to succeed, adjust things to suit.

      */

     if (retval == 0 && !(state->info->flags & UIF_NORMAL_ACTIVE)) {

         state->info->flags |= UIF_NORMAL_ACTIVE;

 

         uart_update_termios(state);

     }

 

 fail:

     return retval;

}

在这里函数里,继续完成操作的设备文件所对应state初始化.现在用户空间open这个设备了.即要对这个文件进行操作了.那uart_port也要开始工作了.即调用uart_startup()使其进入工作状态.当然,也需要初始化uart_port所对应的环形缓冲区circ_buf.即state->info-> xmit.

特别要注意,在这里将tty->driver_data = state;这是因为以后的操作只有port相关了,不需要去了解uart_driver的相关信息.

跟踪看一下里面调用的两个重要的子函数. uart_get()和uart_startup().先分析uart_get().代码如下:

static struct uart_state *uart_get(struct uart_driver *drv, int line)

{

     struct uart_state *state;

     int ret = 0;

 

     state = drv->state + line;

     if (mutex_lock_interruptible(&state->mutex)) {

         ret = -ERESTARTSYS;

         goto err;

     }

 

     state->count++;

     if (!state->port || state->port->flags & UPF_DEAD) {

         ret = -ENXIO;

         goto err_unlock;

     }

 

     if (!state->info) {

         state->info = kzalloc(sizeof(struct uart_info), GFP_KERNEL);

         if (state->info) {

              init_waitqueue_head(&state->info->open_wait);

              init_waitqueue_head(&state->info->delta_msr_wait);

 

              /*

               * Link the info into the other structures.

               */

              state->port->info = state->info;

 

              tasklet_init(&state->info->tlet, uart_tasklet_action,

                        (unsigned long)state);

         } else {

              ret = -ENOMEM;

              goto err_unlock;

         }

     }

     return state;

 

 err_unlock:

     state->count--;

     mutex_unlock(&state->mutex);

 err:

     return ERR_PTR(ret);

}

从代码中可以看出.这里注要是操作是初始化state->info.注意port->info就是state->info的一个副本.即port直接通过port->info可以找到它要操作的缓存区.

 

uart_startup()代码如下:

static int uart_startup(struct uart_state *state, int init_hw)

{

     struct uart_info *info = state->info;

     struct uart_port *port = state->port;

     unsigned long page;

     int retval = 0;

 

     if (info->flags & UIF_INITIALIZED)

         return 0;

 

     /*

      * Set the TTY IO error marker - we will only clear this

      * once we have successfully opened the port.  Also set

      * up the tty->alt_speed kludge

      */

     set_bit(TTY_IO_ERROR, &info->tty->flags);

 

     if (port->type == PORT_UNKNOWN)

         return 0;

 

     /*

      * Initialise and allocate the transmit and temporary

      * buffer.

      */

     if (!info->xmit.buf) {

         page = get_zeroed_page(GFP_KERNEL);

         if (!page)

              return -ENOMEM;

 

         info->xmit.buf = (unsigned char *) page;

         uart_circ_clear(&info->xmit);

     }

 

     retval = port->ops->startup(port);

     if (retval == 0) {

         if (init_hw) {

              /*

               * Initialise the hardware port settings.

               */

              uart_change_speed(state, NULL);

 

              /*

               * Setup the RTS and DTR signals once the

               * port is open and ready to respond.

               */

              if (info->tty->termios->c_cflag & CBAUD)

                   uart_set_mctrl(port, TIOCM_RTS | TIOCM_DTR);

         }

 

         if (info->flags & UIF_CTS_FLOW) {

              spin_lock_irq(&port->lock);

              if (!(port->ops->get_mctrl(port) & TIOCM_CTS))

                   info->tty->hw_stopped = 1;

              spin_unlock_irq(&port->lock);

         }

 

         info->flags |= UIF_INITIALIZED;

 

         clear_bit(TTY_IO_ERROR, &info->tty->flags);

     }

 

     if (retval && capable(CAP_SYS_ADMIN))

         retval = 0;

 

     return retval;

}

在这里,注要完成对环形缓冲,即info->xmit的初始化.然后调用port->ops->startup( )将这个port带入到工作状态.其它的是一个可调参数的设置,就不详细讲解了.

 

七:设备节点的write操作

Write操作对应的操作接口为uart_write( ).代码如下:

static int

uart_write(struct tty_struct *tty, const unsigned char *buf, int count)

{

     struct uart_state *state = tty->driver_data;

     struct uart_port *port;

     struct circ_buf *circ;

     unsigned long flags;

     int c, ret = 0;

 

     /*

      * This means you called this function _after_ the port was

      * closed.  No cookie for you.

      */

     if (!state || !state->info) {

         WARN_ON(1);

         return -EL3HLT;

     }

 

     port = state->port;

     circ = &state->info->xmit;

 

     if (!circ->buf)

         return 0;

 

     spin_lock_irqsave(&port->lock, flags);

     while (1) {

         c = CIRC_SPACE_TO_END(circ->head, circ->tail, UART_XMIT_SIZE);

         if (count < c)

              c = count;

         if (c <= 0)

              break;

         memcpy(circ->buf + circ->head, buf, c);

         circ->head = (circ->head + c) & (UART_XMIT_SIZE - 1);

         buf += c;

         count -= c;

         ret += c;

     }

     spin_unlock_irqrestore(&port->lock, flags);

 

     uart_start(tty);

     return ret;

}

 

Uart_start()代码如下:

static void uart_start(struct tty_struct *tty)

{

     struct uart_state *state = tty->driver_data;

     struct uart_port *port = state->port;

     unsigned long flags;

 

     spin_lock_irqsave(&port->lock, flags);

     __uart_start(tty);

     spin_unlock_irqrestore(&port->lock, flags);

}

static void __uart_start(struct tty_struct *tty)

{

     struct uart_state *state = tty->driver_data;

     struct uart_port *port = state->port;

 

     if (!uart_circ_empty(&state->info->xmit) && state->info->xmit.buf &&

         !tty->stopped && !tty->hw_stopped)

         port->ops->start_tx(port);

}

 

显然,对于write操作而言,它就是将数据copy到环形缓存区.然后调用port->ops->start_tx()将数据写到硬件寄存器.

 

八:Read操作

Uart的read操作同Tty的read操作相同,即都是调用ldsic->read()读取read_buf中的内容.有对这部份内容不太清楚的,参阅<< linux设备模型之tty驱动架构>>.

 

九:小结

本小节是分析serial驱动的基础.在理解了tty驱动架构之后,再来理解uart驱动架构应该不是很难.随着我们在linux设备驱动分析的深入,越来越深刻的体会到,linux的设备驱动架构很多都是相通的.只要深刻理解了一种驱动架构.举一反三.也就很容易分析出其它架构的驱动了.