Hi!

> commit e7b931bee739e8a77ae216e613d3b99342b6dec0 upstream.
> 
> The driver would happily overwrite its write buffer with user data in
> 256 byte increments due to a removed buffer-space sanity check.

> +++ b/drivers/usb/serial/iuu_phoenix.c
> @@ -697,14 +697,16 @@ static int iuu_uart_write(struct tty_str
>  	struct iuu_private *priv = usb_get_serial_port_data(port);
>  	unsigned long flags;
>  
> -	if (count > 256)
> -		return -ENOMEM;
> -
>  	spin_lock_irqsave(&priv->lock, flags);
>  
> +	count = min(count, 256 - priv->writelen);
> +	if (count == 0)
> +		goto out;
> +
>  	/* fill the buffer */
>  	memcpy(priv->writebuf + priv->writelen, buf, count);
>  	priv->writelen += count;
> +out:
>  	spin_unlock_irqrestore(&priv->lock, flags);
>  
>  	return count;

Ok, so... goto and label is unneccessary, memcpy will do the right
thing with count == 0.

But what is worse, this changes return value in the error case;
returning 0 instead of -ENOMEM. I don't believe 0 is appropriate
return code here.

(It should block on the write buffer if blocking or return -EAGAIN if
nonblocking, right?)

Best regards,
								Pavel
-- 
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany