debugfs question ...

[Greg KH]

On Tue, Apr 30, 2019 at 12:45:08PM +0100, rdq at metamail.co wrote:
> Greetings,
> 
> My I2C sensor driver has a debugfs entry for development purposes.
> Everything works fine with the exception of the read operation. When 'cat'
> is used, the read operation is called repeatedly and indefinitely. If the
> read() is changed to return 0 then, as expected, nothing is displayed. 
> 
> The pattern for the implementation is (AFAICT) right out of  the book (shown
> below). 
> 
> What am I missing? Any thoughts much appreciated.
> 
> TAIA.
> 
> RDQ
> 
> static ssize_t sc031gs_reg_read_file(struct file *file, char __user
> *user_buf,
> 				   size_t count, loff_t *ppos)
> {
> 	char *buf = 0;
> 	ssize_t total = 0;
> 	struct sc031gs_dev *sensor = file->private_data;
> 	if (!sensor)
> 		return -EINVAL;
> 	if (*ppos < 0 || !count)
> 		return -EINVAL;
> 	buf = kmalloc(count, GFP_KERNEL);
> 	if (!buf)
> 		return -ENOMEM;
> 	total = snprintf(buf,count,"Hello world\n");
> 	if (total >= 0) {
> 		if (copy_to_user(user_buf, buf, total)) {
> 			kfree(buf);
> 			return -EFAULT;
> 		}
> 		*ppos += total;
> 	}
> 	kfree(buf);
> 	return total;
> }

You are returning a "short" read, and then disallowing ppos to be set to
a non-zero value?  That's a recipie for disaster :(

Also, you allow userspace to allocate as much memory as it asks for?
Not good :(

Why not just use the simple_read_from_buffer() call?  That handles all
of the "housekeeping" for you, and your function can be _much_ simpler.

good luck!

greg k-h