rtt从设计上、框架上、各种模块支持性上都应该国内领先了，必须表扬下。但是，我感觉并不好用，更有莫名的不稳定，比如随便想一个：
1 文件写write SD卡，会有一个不大不打小的概率会失败，磁盘太重要了，这问题我往哪里找啊！
2 i2c 的读写动作，网上清一色的绕过rt_device_write，都是直接调用的i2c虚拟驱动层的函数，这么个重要且小的问题愣是不解决！搞的我自己扩展了内核，又在设备层之上做了个api读写接口，如下：
#include "i2c_if.h"

int i2c_if_read(   rt_device_t i2c_dev, 
                    uint16_t slave_addr, 
                    uint16_t addr, 
                    uint16_t addr_len,      // number of bytes, 1 or 2
                    void *buf, int len)
{
    int ret;
    rt_off_t pos_flag, pos = slave_addr >> 1;
    if (addr_len == 2) addr = (addr >> 8) | (addr << 8);
    struct rt_i2c_bus_device *bus = (struct rt_i2c_bus_device *)i2c_dev->user_data;
    
    RT_ASSERT(bus);
    
    rt_mutex_take(&bus->lock, RT_WAITING_FOREVER);
    
    pos_flag = pos | (RT_I2C_NO_STOP << 16);
    if ((ret=rt_device_write(i2c_dev, pos_flag, &addr, addr_len)) == addr_len) {
        pos_flag = pos | (RT_I2C_FORCE_RESTART << 16);
        ret = rt_device_read(i2c_dev, pos_flag, buf, len);
    }

rt_mutex_release(&bus->lock);
    
    return ret;
}

int i2c_if_write(  rt_device_t i2c_dev, 
                    uint16_t slave_addr, 
                    uint16_t addr, 
                    uint16_t addr_len,      // number of bytes, 1 or 2
                    const void *buf, int len)
{
    int ret;
    rt_off_t pos_flag, pos = slave_addr >> 1;
    if (addr_len == 2) addr = (addr >> 8) | (addr << 8);
    struct rt_i2c_bus_device *bus = (struct rt_i2c_bus_device *)i2c_dev->user_data;
    
    RT_ASSERT(bus);
    
    rt_mutex_take(&bus->lock, RT_WAITING_FOREVER);
    
    pos_flag = pos | (RT_I2C_NO_STOP << 16);
    if ((ret=rt_device_write(i2c_dev, pos_flag, &addr, addr_len)) == addr_len) {
        pos_flag = pos | (RT_I2C_NO_START << 16);
        ret = rt_device_write(i2c_dev, pos_flag, buf, len);
    }

rt_mutex_release(&bus->lock);
    
    return ret;
}

基本的file I2c uart net socket io，一定要好用且稳定，必须是傻瓜式的。否则，带给用户的全是担心和不信任.