Re: [PATCH, RFC] usbmon: correct computing of the ISO packets withmmap

[NÃmeth MÃrton]

NÃmeth MÃrton wrote:
> Hi Pete,
> Pete Zaitcev wrote:
>> On Sun, 14 Nov 2010 21:24:46 +0100
>> NÃmeth MÃrton <nm127@xxxxxxxxxxx> wrote:
>>
>>> Sure, I'm happy as long as I can capture the whole data content of the ISO
>>> packets.
>> Great. So, what do you think about the attached? It differs from your
>> code in one important aspect: output buffer sizes are not adjusted.
>> Your patch attempts that, but uses actual_length, which is not
>> defined during submission events. So I skipped that.
> 
> I think there are four cases for ISO communication URBs:
> 
> ev_type == 'S' && usb_urb_dir_in(urb)  ---> we don't need the data
> ev_type == 'C' && usb_urb_dir_in(urb)  ---> data is available, we'll need it
> ev_type == 'S' && usb_urb_dir_out(urb)  ---> data is available, we'll need it
> ev_type == 'C' && usb_urb_dir_out(urb)  ---> we don't need the data
> 
> I cannot say for sure which field contains the length in case of ISO out submit
> URBs.
> 
> I'll test the patch later in my environment.

I tested this patch with ISO in traffic and it works for me.

>> diff --git a/drivers/usb/mon/mon_bin.c b/drivers/usb/mon/mon_bin.c
>> index 44cb37b..15c5e46 100644
>> --- a/drivers/usb/mon/mon_bin.c
>> +++ b/drivers/usb/mon/mon_bin.c
>> @@ -437,6 +437,28 @@ static unsigned int mon_bin_get_data(const struct mon_reader_bin *rp,
>>  	return length;
>>  }
>>  
>> +/*
>> + * This is the look-ahead pass in case of 'Ci', when actual_length cannot
>> + * be used to determine the length of the whole contiguous buffer.
>> + */
>> +static unsigned int mon_bin_collate_isodesc(const struct mon_reader_bin *rp,
>> +    struct urb *urb, unsigned int ndesc)
>> +{
>> +	struct usb_iso_packet_descriptor *fp;
>> +	unsigned int length;
>> +
>> +	length = 0;
>> +	fp = urb->iso_frame_desc;
>> +	while (ndesc-- != 0) {
>> +		if (fp->status == 0 && fp->actual_length != 0) {
>> +			if (fp->offset + fp->actual_length > length)
>> +				length = fp->offset + fp->actual_length;
> 
> I don't know whether the compiler will optimize the two times computing the
> expression (fp->offset + fp->actual_length). Maybe I would use a local variable
> to compute the sum before the "if" and then use just the local variable.
> 
>> +		}
>> +		fp++;
>> +	}
>> +	return length;
>> +}
>> +
>>  static void mon_bin_get_isodesc(const struct mon_reader_bin *rp,
>>      unsigned int offset, struct urb *urb, char ev_type, unsigned int ndesc)
>>  {
>> @@ -479,6 +501,10 @@ static void mon_bin_event(struct mon_reader_bin *rp, struct urb *urb,
>>  	/*
>>  	 * Find the maximum allowable length, then allocate space.
>>  	 */
>> +	urb_length = (ev_type == 'S') ?
>> +	    urb->transfer_buffer_length : urb->actual_length;
>> +	length = urb_length;
>> +
>>  	if (usb_endpoint_xfer_isoc(epd)) {
>>  		if (urb->number_of_packets < 0) {
>>  			ndesc = 0;
>> @@ -487,14 +513,16 @@ static void mon_bin_event(struct mon_reader_bin *rp, struct urb *urb,
>>  		} else {
>>  			ndesc = urb->number_of_packets;
>>  		}
>> +		if (ev_type == 'C' && usb_urb_dir_in(urb))
>> +			length = mon_bin_collate_isodesc(rp, urb, ndesc);
>>  	} else {
>>  		ndesc = 0;
>>  	}
>>  	lendesc = ndesc*sizeof(struct mon_bin_isodesc);
>>  
>> -	urb_length = (ev_type == 'S') ?
>> -	    urb->transfer_buffer_length : urb->actual_length;
>> -	length = urb_length;
>> +	/* not an issue unless there's a subtle bug in a HCD somewhere */
>> +	if (length >= urb->transfer_buffer_length)
>> +		length = urb->transfer_buffer_length;
>>  
>>  	if (length >= rp->b_size/5)
>>  		length = rp->b_size/5;
>>
>>
> 
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/