Re: [PATCH 2/2] staging: comedi: drivers: usbduxfast.c: fix for DMA buffers on stack

[walter harms <wharms@bfs.de>]

Am 22.02.2013 20:06, schrieb Greg KH:
> On Fri, Feb 22, 2013 at 07:58:35PM +0100, walter harms wrote:
>>
>>
>> Am 22.02.2013 19:07, schrieb Kumar Amit Mehta:
>>> fix for instances of DMA buffer on stack(being passed to usb_control_msg) for
>>> the USB-DUXfast Board driver.
>>>
>>> Signed-off-by: Kumar Amit Mehta <gmate.amit@gmail.com>
>>> ---
>>>  drivers/staging/comedi/drivers/usbduxfast.c |   30 ++++++++++++++++-----------
>>>  1 file changed, 18 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/drivers/staging/comedi/drivers/usbduxfast.c b/drivers/staging/comedi/drivers/usbduxfast.c
>>> index 4bf5dd0..1ba0e3d 100644
>>> --- a/drivers/staging/comedi/drivers/usbduxfast.c
>>> +++ b/drivers/staging/comedi/drivers/usbduxfast.c
>>> @@ -436,10 +436,14 @@ static void usbduxfastsub_ai_Irq(struct urb *urb)
>>>  static int usbduxfastsub_start(struct usbduxfastsub_s *udfs)
>>>  {
>>>  	int ret;
>>> -	unsigned char local_transfer_buffer[16];
>>> +	unsigned char *local_transfer_buffer;
>>> +
>>> +	local_transfer_buffer = kmalloc(1, GFP_KERNEL);
>>> +	if (!local_transfer_buffer)
>>> +		return -ENOMEM;
>>>  
>>>  	/* 7f92 to zero */
>>> -	local_transfer_buffer[0] = 0;
>>> +	*local_transfer_buffer = 0;
>>>  	/* bRequest, "Firmware" */
>>>  	ret = usb_control_msg(udfs->usbdev, usb_sndctrlpipe(udfs->usbdev, 0),
>>>  			      USBDUXFASTSUB_FIRMWARE,
>>> @@ -450,22 +454,25 @@ static int usbduxfastsub_start(struct usbduxfastsub_s *udfs)
>>>  			      local_transfer_buffer,
>>>  			      1,      /* Length */
>>>  			      EZTIMEOUT);    /* Timeout */
>>> -	if (ret < 0) {
>>> +	if (ret < 0)
>>>  		dev_err(&udfs->interface->dev,
>>>  			"control msg failed (start)\n");
>>> -		return ret;
>>> -	}
>>>  
>>> -	return 0;
>>> +	kfree(local_transfer_buffer);
>>> +	return ret;
>>>  }
>>>  
>>>  static int usbduxfastsub_stop(struct usbduxfastsub_s *udfs)
>>>  {
>>>  	int ret;
>>> -	unsigned char local_transfer_buffer[16];
>>> +	unsigned char *local_transfer_buffer;
>>> +
>>> +	local_transfer_buffer = kmalloc(1, GFP_KERNEL);
>>> +	if (!local_transfer_buffer)
>>> +		return -ENOMEM;
>>>  
>>>  	/* 7f92 to one */
>>> -	local_transfer_buffer[0] = 1;
>>> +	*local_transfer_buffer = 1;
>>>  	/* bRequest, "Firmware" */
>>>  	ret = usb_control_msg(udfs->usbdev, usb_sndctrlpipe(udfs->usbdev, 0),
>>>  			      USBDUXFASTSUB_FIRMWARE,
>>> @@ -474,13 +481,12 @@ static int usbduxfastsub_stop(struct usbduxfastsub_s *udfs)
>>>  			      0x0000,	/* Index */
>>>  			      local_transfer_buffer, 1,	/* Length */
>>>  			      EZTIMEOUT);	/* Timeout */
>>> -	if (ret < 0) {
>>> +	if (ret < 0)
>>>  		dev_err(&udfs->interface->dev,
>>>  			"control msg failed (stop)\n");
>>> -		return ret;
>>> -	}
>>>  
>>> -	return 0;
>>> +	kfree(local_transfer_buffer);
>>> +	return ret;
>>>  }
>>>  
>>>  static int usbduxfastsub_upload(struct usbduxfastsub_s *udfs,
>>
>> mmh, it seems a bit overheat to alloc 1 byte.
>> Could one of the driver maintainers please comment
>> is that realy need ?
> 
> Yes it is needed.
> 
>> or is it possible to pass one byte
>> in a register ? (aka char/int) without allocating ?
> 
> Nope, the USB host controllers must be able to DMA to this memory
> location, so you have to allocate it dynamically, sorry.
> 
> thanks,

thx for clarification.

@Kumar Amit Mehta:
Would you mind to add this as comment ? Allocating one byte does not
look clever so maybe will come up with the idea of changing that.

re,
 wh