您好,我正在为我的字符驱动程序编写一个 write() 方法,我想知道为什么当我将数据从用户复制到我的内核缓冲区时,我的缓冲区包含随机乱码。下面是我正在使用的方法。
ssize_t dev_write(struct file *filp, const char __user *buff, size_t count, loff_t *offp){// This function looks good
struct my_char_structure *my_dev = filp->private_data;
char *offset; // This points to my buffer
size_t np, left_to_print = count; // These are the bytes left to write
//my_dev->set.write is a double pointer to a sentence where
//set.write = array[sentence number][character in sentence]
//set.write if set.write points to non-allocated memory, the sentence number is NULL
//My device only holds 100 sentences in all
if(my_dev->set.write == NULL){// Write is a double pointer that points
printk(KERN_ALERT "Device has no more room");
return -ENOMEM; // Look up
}
//you can ignore the commented out stuff but I want to check to see if I'm referencing
//a sentence that has been filled out. other wise the sentence is null and may be
//written to. I will add this to my code later for traversing sentences.
/*if(*(my_dev->set.write) != NULL){
my_dev->set.write ++;
dev_write(filp,buff,count,offp);
exit 0;
}*/
if (down_interruptible(&my_dev->sem))
return -ERESTARTSYS;
// Here I'm referencing a memory segment that acts as a pointer to a sentence which
//I write to. *(my_dev->set.write) is a pointer to a char buff which I will place chars
// which is essentially my string
*(my_dev->set.write) = kmalloc(count,GFP_KERNEL);
offset = *(my_dev->set.write); // my offset points to that buffer as well
//A sentence can only be the size of count, which is passed by the user
//thats why I allocate count bytes for memory.
if((np = copy_from_user(offset,buff,left_to_print)) < 0)
goto erro_out;
left_to_print -= (count-np);
offset += (count-np);
offp += (count-np);
//For debbuging purposes I have printk(). My offset points to my buffer, however
//prints jibberish. In user space I insert \0 at the end of the string.
printk(KERN_ALERT " 4 :: write && left is ... %ld. The string %s", left_to_print, offset); // %s prints jibberish
return left_to_print; // change to count
erro_out:// if there was an error I free the sentence so that it may be written to.
kfree(*(my_dev->set.write));
printk(KERN_ALERT "Print error %ld",np);
return -EFAULT; // look this up
}
如果我的方法有点啰嗦,下面显示了我的方法摘要。这对于第一次写入设备应该是好的。
ssize_t dev_write(struct file *filp, const char __user *buff, size_t count, loff_t *offp){// This function looks good
struct my_char_structure *my_dev = filp->private_data;
static char *offset;
if (down_interruptible(&my_dev->sem))
return -ERESTARTSYS;
*(my_dev->set.write) = kmalloc(count,GFP_KERNEL);
offset = *(my_dev->set.write);
if(copy_from_user(offset,buff,left_to_print) < 0)
goto erro_out;
// This is my question, why does %s print jibberish
printk(KERN_ALERT " 4 :: write && left is ... %ld. The string %s", left_to_print, offset);
return 0; // change to count
erro_out:
kfree(*(my_dev->set.write));
printk(KERN_ALERT "Print error %ld",np);
return -EFAULT; // look this up
}
我使用 echo "my sentence">/dev/my_dev 来写入我的设备。我不在乎回声是否有效。我只希望 printk() 显示“我的句子”,这意味着 copy_from 正在工作。我还尝试使用 echo "my sentence\0">/dev/my_dev 来遵守在字符串末尾粘贴空字符的规则。
感谢您的帮助。
最佳答案
copy_from_user
的第一个参数需要一个目标地址。通过像这样传递偏移量:
if((np = copy_from_user(offset,buff,left_to_print)) < 0)
你给了它错误的写入地址。你应该传递你的缓冲区,比如:
if((np = copy_from_user(_my_buffer_ + offset,buff,left_to_print)) < 0)
请确保您分配了足够的空间,以便您可以将 offset
添加到地址,并注意我假设您的缓冲区是 sizeof
计算结果为的数据类型1.
关于c - 如何从内核缓冲区打印字符串?我使用 copy_from_user() 对吗?,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/9813783/