[Tinyos-devel] 2.0.2 CC2420

Thu Jul 5 10:33:10 PDT 2007

So we have two votes for including the software address recognition from
Mike and Ryan.  There's not really an optional way of including it, except
by have a completely separate CC2420ReceiveP in an app directory that
overrides the default.  Having two separate CC2420ReceiveP's would make
configuration management more difficult.  

Instead of making two, how about we use software address recognition as the
default as suggested and nudge someone to characterize both implementations
at some point in a test bed to quantify the difference in performance?  We
know the address check is not optimized, but if performance really doesn't
suffer that much, then the benefits outweigh the loss.

As per Ryan's comments, CC2420Config.setAddressRecognition(bool on) and
CC2420Config.isAddressRecognitionEnabled() have been added, along with an
equivalent preprocessing #define: CC2420_NO_ADDRESS_RECOGNITION.  The
isAddressRecognitionEnabled() is not checked when performing SACK's, since
we should always perform an address check when SACK'ing.  

Another unit test should be added for this functionality, but I don't have
time to create it right now.  I'll regression test the rest of the stack to
make sure nothing broke and update CC2420BaseStation to use the new
preprocessing definition configurations.  After regression testing is
complete, I'll commit these changes.

Thanks for all your input guys,
-David

-----Original Message-----
From: J. Ryan Stinnett [mailto:jryans at rice.edu] 
Sent: Wednesday, July 04, 2007 1:26 PM
To: David Moss
Cc: 'Chieh-Jan (Mike) Liang'; tinyos-devel at millennium.berkeley.edu
Subject: Re: RE: [Tinyos-devel] 2.0.2 CC2420

David Moss wrote:
> I would be very hesitant to allow the baseline CC2420ReceiveP to perform
its
> former software address check to support this edge case because
> acknowledgement performance may suffer as a whole, which we've already
seen
> and quantified to some degree.  
> 
> There are a few timing issues going on with acks in the background to take
> into consideration:  After the Tx mote sends its packet, the SPI bus must
be
> acquired by the Rx mote.  This resource acquiring may add significant
delay
> depending on what else is going on with the SPI bus (flash, other task
> handling).  By the time the Rx mote acquires its SPI resource and is able
to
> download the packet, the Tx mote's ack wait period is winding down.  The
Rx
> mote must SACK as fast as possible to ensure the ack is received by the Tx
> mote within the ack wait window.  The fastest way to SACK is to download
the
> first byte (length) and the second byte (fcf) to see if A) we have a valid
> packet and B) if the packet requested an ack.  If we wanted to test for
the
> address, we'd have to download several more bytes before being able to see
> the destination address.  True, it's only a few more bytes, but it's
> certainly not optimized.
> 
> Another way around this case is to do what the BaseStationCC2420 does,
which
> is override a module from the CC2420 stack with a version that does what
you
> want.  I've made it fairly easy to do:  there's an FCF_LENGTH parameter in
> CC2420ReceiveP which is currently set to 2 (because the FCF is 2 bytes).
By
> setting it to 7, Receive will make sure the packet's length is at least 7
> bytes, read those 7 bytes and SACK, then continue reading the rest of the
> packet.  Because we're reading in 7 bytes from the header before SACK'ing
> instead of 2, the destination address will be available and we can check
it.
> (attached, this file isn't tested).

I'm not sure I understand how reading these 7 bytes works around the 
issue of reading extra bytes in the paragraph above this one.  I can see 
how having to download the entire packet before SACK would degrade 
performance, however if it can be done by getting only a few more bytes 
to check the address, I think that's great and should be available *as 
an option* but not be the default, since as you say it is not optimal.

- Ryan

> 
> If other users agree that this really isn't an edge case and we should see
> this software address check version as the default, then by all means it
> should be in there.  However, SACK's do not perform as well as hardware
> ack's, and it's been my focus to get them optimized as much as possible.
> 
> -David
> 
> 
> 
> 
> 
> -----Original Message-----
> From: Chieh-Jan (Mike) Liang [mailto:cliang4 at cs.jhu.edu] 
> Sent: Wednesday, July 04, 2007 11:20 AM
> To: David Moss
> Cc: tinyos-devel at millennium.berkeley.edu
> Subject: Re: [Tinyos-devel] 2.0.2 CC2420
> 
>> ....  If you want the base
>> station to disable address recognition to sniff all packets, and also
send
>> out SACK's when it receives a packet with its address, that's another
> issue.
>> -David
>>
> 
> I think this could be useful. In addition, adding address check should 
> not affect the two applications you mentioned: base station and sniffer. 
> Any thoughts?
> 
> Thanks
> 
> Mike
> 
>> -----Original Message-----
>> From: J. Ryan Stinnett [mailto:jryans at rice.edu] 
>> Sent: Tuesday, July 03, 2007 10:45 PM
>> To: David Moss
>> Cc: 'Chieh-Jan (Mike) Liang'; tinyos-devel at millennium.berkeley.edu
>> Subject: Re: RE: [Tinyos-devel] 2.0.2 CC2420
>>
>> Does this have any effect on snooping with CC2420?  Just to confirm, the 
>> address recognition feature needs to be disabled as in BaseStationCC2420 
>> before snooping will work, is that correct?
>>
>> Thanks,
>> Ryan
>>
>> David Moss wrote:
>>> Yup - there's a good reason for this.  The destination address is
checked
>>> automatically by hardware while receiving the packet transmission.  So
by
>>> the time the uC gets the signal that a new packet has arrived, the
packet
>>> has already gone through the address/pan check.  Therefore, we don't
need
>> a
>>> second address check in software before issuing a SACK.
>>>
>>> -David
>>>
>>>
>>> -----Original Message-----
>>> From: David Moss [mailto:dmm at rincon.com] 
>>> Sent: Tuesday, July 03, 2007 9:45 PM
>>> To: 'Chieh-Jan (Mike) Liang'
>>> Cc: 'tinyos-devel at Millennium.Berkeley.EDU'
>>> Subject: RE: [Tinyos-devel] 2.0.2 CC2420
>>>
>>> Yup - there's a good reason for this.  The destination address is
checked
>>> automatically by hardware while receiving the packet transmission.  So
by
>>> the time the uC gets the signal that a new packet has arrived, the
packet
>>> has already gone through the address/pan check.  Therefore, we don't
need
>> a
>>> second address check in software before issuing a SACK.
>>>
>>> -David
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: Chieh-Jan (Mike) Liang [mailto:cliang4 at cs.jhu.edu] 
>>> Sent: Tuesday, July 03, 2007 7:41 PM
>>> To: David Moss
>>> Cc: tinyos-devel at Millennium.Berkeley.EDU
>>> Subject: Re: [Tinyos-devel] 2.0.2 CC2420
>>>
>>> Something about SACK in the new CC2420 stack is not clear to me. In the 
>>> old implementation, before SACK is strobed, the mote ID is compared with

>>> the packet destination address. However, this check is not performed any

>>> more. Is there any reason behind this?
>>>
>>> Thank you
>>>
>>> Mike
>>>
>>> David Moss wrote:
>>>> Just updated the CC2420 stack in preparation for 2.0.2.  The file 
>>>> architecture has been re-arranged a little bit, making it easier to see

>>>> how the stack is put together.  As a result, the .platform files that 
>>>> reference the cc2420 were updated. 
>>>>
>>>>  
>>>>
>>>> Test results are included. Let me know if you have any issues.
>>>>
>>>> -David
>>>>
>>>>  
>>>>
>>>>  
>>>>
>>>> Recent Updates (from the readme):
>>>>
>>>> * New chip SPI bus arbitration working with Receive and Transmit.
>>>>
>>>>  
>>>>
>>>> * Applied TUnit automated unit testing to CC2420 development
>>>>
>>>>   > Caught lots of bugs, especially through regression testing
>>>>
>>>>   > Source code in tinyos-2.x-contribs/tunit/
>>>>
>>>>  
>>>>
>>>> * Applied TEP115 behavior to CC2420 SplitControl in Csma and Lpl
>>>>
>>>>  
>>>>
>>>> * Updated ActiveMessageAddressC to provide the ActiveMessageAddress 
>>>> interface
>>>>
>>>>   > Updated CC2420ConfigP to handle 
>>>> ActiveMessageAddress.addressChanged() and
>>>>
>>>>     sync automatically upon address change events.
>>>>
>>>>  
>>>>
>>>> * Updated CC2420Config interface to enable/disable sw/hw
> acknowledgements
>>>>  
>>>>
>>>> * Updated CC2420ConfigP to share register editing through single
>> functions
>>>>  
>>>>
>>>> * Acknowledge after packet length and FCF check out valid. 
>>>>
>>>>   > The destination address is confirmed in hardware, so we don't need
>>>>
>>>>     to download the entire header before acking.
>>>>
>>>>  
>>>>
>>>> * Moved the getHeader() and getMetadata() commands to an internal
>>> interface
>>>>   called CC2420PacketBody, provided by CC2420PacketC
>>>>
>>>>  
>>>>
>>>> * Separated core functionality into different sub-packages/directories
>>>>
>>>>   > Updated micaz, telosb, intelmote2 .platform files
>>>>
>>>>   > Logically organizes code
>>>>
>>>>  
>>>>
>>>> * Updated some LPL architecture
>>>>
>>>>   > Removed continuous modulation because it didn't work 100% and I 
>>>> don't have
>>>>
>>>>     time to make it work. 
>>>>
>>>>   > Decreased backoffs and decreased on-time for detects, saving
energy.
>>>>
>>>>  
>>>>
>>>> * Updated to the new AMPacket interface; made the radio set the
outbound
>>>>
>>>>   packet's destpan after send().
>>>>
>>>>  
>>>>
>>>>  
>>>>
>>>>  
>>>>
>>>>
>>>>
------------------------------------------------------------------------
>>>>
>>>> _______________________________________________
>>>> Tinyos-devel mailing list
>>>> Tinyos-devel at Millennium.Berkeley.EDU
>>>>
>>
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-devel
>>>