Steve’s Software Trek

I created a BACnet Development Kit, compiled the open source BACnet Protocol Stack hosted on SourceForge.net firmware using gcc-avr, and attempted to load the firmware onto the kit using an Atmel AVR ISP mkII USB programmer.  I connected the AVR ISP mkII to my Ubuntu Hardy Linux PC and attached the 2×3 ISP cable to the development kit board, and attempted to program using avrdude:

$ avrdude -c avrispmkII -p m644p -P usb -e -U flash:w:bacnet.hex

avrdude: usb_open(): cannot read serial number "error sending..."
avrdude: usb_open(): cannot read product name "error sending..."
avrdude: usbdev_open(): error setting configuration...
avrdude: usbdev_open(): did not find any USB device "usb"

Wow! That wasn’t nice.  A search on the Internet turned up some answers about how to fix the problem, and I modified my /etc/udev/avarice.rules file to look like this:

SUBSYSTEM!="usb_device", ACTION!="add", GOTO="avarice_end"

# Atmel Corp. JTAG ICE mkII
SYSFS{idVendor}=="03eb", SYSFS{idProduct}=="2103", MODE="660", GROUP="dialout"
# Atmel Corp. AVRISP mkII
SYSFS{idVendor}=="03eb", SYSFS{idProduct}=="2104", MODE="660", GROUP="dialout"
# Atmel Corp. Dragon
SYSFS{idVendor}=="03eb", SYSFS{idProduct}=="2107", MODE="660", GROUP="dialout"

LABEL="avarice_end"

I disconnected and reconnected my AVR ISP mkII, and tried again.  I still had the same results.  I checked that I was a member of the dialout group:

$ groups
skarg adm dialout cdrom floppy audio dip video plugdev users fuse lpadmin admin

I looked at the symbolic link in the /etc/udev/rules.d/ directory and also consumed the README file.  My file had the following name:

/etc/udev/rules.d/z60_avarice.rules

So I changed the filename:

$ sudo mv /etc/udev/rules.d/z60_avarice.rules /etc/udev/rules.d/60-avarice.rules

I unplugged the AVR ISP mkII, and plugged it in again.  I tried to run avrdude again:

avrdude -c avrispmkII   \
     -p m644p -P usb -e        \
     -U flash:w:bacnet.hex

avrdude: AVR device initialized and ready to accept instructions

Reading | ################################################## | 100% 0.01s

avrdude: Device signature = 0x1e960a
avrdude: erasing chip
avrdude: reading input file "bacnet.hex"
avrdude: input file bacnet.hex auto detected as Intel Hex
avrdude: writing flash (57406 bytes):

Writing | ################################################## | 100% 6.31s

avrdude: 57406 bytes of flash written
avrdude: verifying flash memory against bacnet.hex:
avrdude: load data flash data from input file bacnet.hex:
avrdude: input file bacnet.hex auto detected as Intel Hex
avrdude: input file bacnet.hex contains 57406 bytes
avrdude: reading on-chip flash data:

Reading | ################################################## | 100% 5.41s

avrdude: verifying ...
avrdude: 57406 bytes of flash verified

avrdude: safemode: Fuses OK

avrdude done.  Thank you.

The kit programmed successfully, and began operation.

I wrote an article about Analyzing BACnet with Wireshark for ASHRAE Journal special supplement BACnet Today.  ASHRAE is very particular about endorsing specific company products, and removed some references I had to information about how BACnet MS/TP can be supported on Wireshark by adding an external interface which sends Ethernet SNAP protocol packets.  The external interface was originally a Cimetrics U+4 device which connects BACnet MS/TP via USB to a computer, and uses a special driver to appear as a network interface to Wireshark.

In April, 2008, I was informed that the Cimetrics U+4 packets showed up in Wireshark, but were not decoded correctly.  I asked for a sample of the capture, and reverse engineered the header.  I then created the BACnet MS/TP decoding in Wireshark in order to display the MS/TP data, and passed the BACnet Network Layer and Application Layer information along to the existing dissectors.

The beauty of the BACnet protocol is that it is layered.  The NPDU (network layer) and APDU (application layer) encodings are the same regardless of the datalink layer used.  The NDPU layer is decoded by packet-bacnet.c, which passes the APDU layer to packet-bacapp.c.  The BACnet MS/TP datalink layer is decoded by packet-mstp.c which I created, and passes the NPDU, if it exists, to packet-bacnet.c, for decoding.  The ARCNET, Ethernet, and BACnet/IP (packet-bvlc.c) all pass a relevant packet to packet-bacnet.c for BACnet decoding.  I created packet-cimetrics.c for the Cimetrics header which passes the BACnet MS/TP datalink layer to packet-mstp.c.

The BACnet MS/TP patches were accepted into the Wireshark repository on May 13, 2008, checked in as SVN 25291.  The SVN 25881 was a fix to the Length decoding.  The BACnet MS/TP decoding eventually made it into the Wireshark 1.1 development release, but is not present in the Wireshark 1.0 stable release.

Wireshark now sees packets with the Cimetrics U+4 signature encapsulated in the SNAP protocol as MS/TP packets, and it is able to open a file that has been saved in PCAP format using the MS/TP WTAP signature.  But I didn’t have a Cimetrics U+4 device.  So I used a standard RS485 converter, my open source BACnet stack, and wrote a simple application to packetize the MS/TP frames and saves them to PCAP format.  I could now capture MS/TP packets and view them!

In August, 2008, I received an email from John in New Zealand.  His intention was “to use an existing simple VB application that captures MS/TP traffic from a low cost USB to RS-485 converter, to send the MS/TP packets using Winsock/UDP to Wireshark for final dissection and display.”  I figured that we could do this by making use of the existing SNAP protocol dissector that I created for the Cimetrics packets.

I reused the code that I had written to save in PCAP format, and wrote a simple application to packetize the MS/TP frames and resend them on Ethernet as the Cimetrics U+4 signature encapsulated in the SNAP protocol.  I got it to work under Linux, but would probably need to use WinPcap to get it to work under Windows.

Sending those SNAP protocol packets on Ethernet allows the display of BACnet MS/TP packets in realtime alongside BACnet/IP and BACnet Ethernet.  This makes it very easy to relate traffic going to and from workstations and routers with what is happening on the BACnet MS/TP network.

There are some sample captures (see cimetrics_mstp.pcap and mstp_wtap.cap) that demostrate some of the abilities of Wireshark.

I use a USB to RS-485 converter from SerialGear USB-COMi-SI-M, and am told that the B&B Electronics USOPTL4 works very well.  I am also told that the USB to RS-485 adapters that use the FTDI chipset are great for this as they allow the use of 76800 baud which is usually the stumbling block for native MS/TP interfacing with a standard PC serial port.

I shot some low light photography at the youth group Halloween dance using a Casio Exilim point-n-shoot, and some of the photos had digital noise, or grain (as we called it in the days of film).

I searched the internet for some assistance in removing the digital noise using the GNU Image Manipulation Program (GIMP), and stumbled upon a well written post about how to remove digital noise.  The technique makes use of the GREYCstoration plugin for the GIMP.

I use the GIMP on Kubuntu Linux, and wondered if there was already a package created for GREYCstoration.

$ apt-cache search greyc
gimp-plugin-registry - A repository of optional extensions for The GIMP

I found it!  So I installed it:

$ sudo apt-get install gimp-plugin-registry

After restarting the GIMP, here are the results:

Low light photo (scaled)

Low light photo (GREYCstoration, scaled)

I created a ring buffer library for embedded systems in C some years ago and have been using it when I need to add a simple queue to my software.  I knew it worked correctly since I created unit tests to validate the code.  However, I noticed that it wasn’t interrupt safe since one of the consumer variables is shared by the producer (the count).  This library requires disabling interrupts when consuming.  Not what I wanted.

There are a number of difficulties when designing ring buffers, mostly dealing with efficiencies of code or data, or the struggle between consumers and producers.  My generic ringbuf library had used a single fill count, but the count had to be updated by the interrupt service routine (producer) and by the routine that retrieved the data (consumer).  I tried to use read and write offset indices, but that required that I always keep an element open since the you can’t tell if the ring is empty or full when the head and tail are the same.

Absolute indices appeared to be interrupt safe, although they require that the buffer length be a power of two.  That seemed to be flexible enough for most embedded work.  I modified and tested my ring buffer library using that method, but my ringbuf, which allowed for a fixed length element,  seemed to be overkill for what I was using it for - gathering bytes from a serial interrupt service routine.  Its API also had a slight flaw such that if the consumer retrieved the data (via Pop rather than just viewing the first element), the producer was free to put new data into the buffer, overwriting the data just retrieved.

I had started my ring buffer library as a simple character (one byte) FIFO - first in, first out.  So, I revisited that code, and modified it to use absolute indices.  A simple byte ring buffer, first in, first out.  Now I had interrupt safe code that would work in my serial interrupt service routine.

Why go to the trouble to make a library module to handle a ring buffer when slapping the following quick and dirty ring buffer code (from Stefan) into the serial module would have worked as well?

   #define N 128
   volatile unsigned int head, tail;
   volatile char buffer[N];
   unsigned int inuse() { return head - tail; }
   void put(char c) { if (inuse() != N) { buffer[head++%N] = c; } }
   void get(char* c) { if (inuse() != 0) { *c = buffer[tail++%N]; } }

Because I needed to know that it worked correctly, and I like reusable library modules.  Especially high quality reusable modules.  Especially high quality reusable modules that include unit tests to validate functionality.  Unit tests that also serve as guides for implementers.

The new module files are ringbuf.c and ringbuf.h for the ring buffer with fixed sized elements, and fifo.c and fifo.h for the ring buffer with byte sized elements. All my versions of the ring buffer and the unit testing framework can be found in the ringbuf.zip file.

I drove to the BACnet meetings in nearby Atlanta and Bill photographed my license plate.  We need lofty goals, don’t we?

I wanted an easy way to convert music to MP3 from various formats under Konqueror in KDE on Kubuntu Linux.  I found an interesting post about creating a KDE Service Menu which got me started in the right direction.  However, I didn’t like the way ffmpeg converted to 64kbps MP3 quality and the bad sounding results regardless of the bitrate settings that I chose.  I tried lame, and was quite pleased with the results.  I also didn’t like the way the service menu used the entire filename without stripping the dotted extension, so I copied another service menu perl example.  My “Convert to MP3″ service menu script goes into the following location:

/usr/share/apps/konqueror/servicemenus/

File convertmp3.desktop:

[Desktop Entry]
ServiceTypes=audio/*
Actions=convertToMP3

[Desktop Action convertToMP3]
Name=Convert to MP3
Exec=lame -V2 %f "`echo %f | perl -pe 's/\.[^.]+$//'`.mp3"

Enjoy!

I kept getting “Invalid Credentials” in Thunderbird when connected to my Gmail IMAP account.  I found the solution to be:

https://www.google.com/accounts/UnlockCaptcha?

Birds are fast.  I guess that is why they have been able to survive all these years.  But I want to take a photograph of them!  So, I have learned to be fast.  Not just fast in taking the shot, but also fast regarding the shutter speed.

I usually use a longer lense (300mm) to get in close, and didn’t realize how much shaking is going on when I am photographing something.  Since a lot of birds eat in the morning or evening, and the setting is my back yard which is wooded, the birds are not in full sun light.  The other challenge in a wooded environment is auto focus.  Too many things for the auto focus to focus on, and how does it know I am looking at a bird?

1) Focus Problem  2) Lens Shake  3) Missed the Bird

So here is what I do:

1. Increase the ISO.  I set my Canon Digital Rebel XTi to ISO 1600.
2. Use shutter priority - TV - mode.  Set the shutter speed to 1/125s or faster.
3. Use the rapid fire - AI Servo - mode for taking lots of photos very fast.
4. Use manual focus (sometimes).

The BACnet committee met in Salt Lake City at the end of June.  We had a large number of Addenda that had finished public review, and had lots of comments for which to respond.  Salt Lake City had some interesting sites and some street items that I had not seen before.  The “Look” on the pavement, as well as the flags used at some crosswalks.

I attended Mass at Cathedral of the Madeleine on Saturday night.  It was a beautiful church and a thought provoking sermon.

I ate some wonderful food in Salt Lake City, including a fabulous carrot cake at Squatter’s Pub Brewery, and some delightful salmon at Cucina Toscana.

Our fun BACnet committee and working groups met in Germantown at the end of April for a week. The weather was perfect, and the Montgomery College Germantown Campus was beautiful and lively.

We had our traditional Tuesday night BACnet dinner at Dogfish Head Brewery.

Most of the working groups met.