KHJKKarl Hans Janke Kollaborativhttp://www.khjk.org/Sven Moritz Hallbergsm@khjk.org2013-08-22T01:46:00ZThe Arduino as a simple JTAG adapterhttp://www.khjk.org/log/2013/aug/jtagbang.htmlpescopesco@khjk.org2013-08-22T01:46:00Z
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/img/2013/cpld.medium.jpg"><img src="/img/2013/cpld.klein.jpg" alt="cpld.klein.jpg" /></a
></div
></div
><p />A good while ago I won one of the free PCBs
<a href="http://dangerousprototypes.com/store/">regularly given away</a
>
by DIY hardware shop
<a href="http://dangerousprototypes.com/">Dangerous Prototypes</a
>.
My board of choice was a
<a href="http://dangerousprototypes.com/docs/XC9500XL_CPLD_breakout_board">CPLD breakout board</a
>,
for the Xilinx XC9572XL.
<a href="http://en.wikipedia.org/wiki/CPLD/">CPLDs</a
>
are the smaller brother of FPGAs:
<q>programmable logic</q
> chips that can be made to act
as any integrated circuit
within the device's limits.
The XC9572XL is programmed via a
standard <a href="http://en.wikipedia.org/wiki/JTAG/">JTAG</a
>
interface.
I did not have anything that speaks JTAG
so went looking if my Arduino
can be turned into an appropriate programmer.
The <a href="https://github.com/sowbug/JTAGWhisperer">solution that I found</a
>,
however, did not work;
so I built my own.
<p /><q>Normally</q
>,
to program a CPLD, or FPGA,
one buys an expensive interface cable
and uses it with the software development suite
supplied by the particular chip's vendor.
Of course there are plenty of DIY alternatives;
in fact, Dangerous Prototypes
<a href="http://dangerousprototypes.com/docs/Bus_Pirate">sell one</a
>
<a href="http://dangerousprototypes.com/docs/Bus_Blaster">or two</a
>.
One of my goals with this project was, however,
to spend next to no money on it.
I got the circuit board for free,
the parts cost around 3EUR,
and I had already done a similar job
with my
<a href="http://www.khjk.org/log/2009/mar/arduino-avrprog.html">Arduino Atmel programmer</a
>.
<p />So after soldering the board
I flashed the abovementioned
JTAG code onto the Arduino.
This was my second time SMD-soldering
so I was not expecting the board to work on first try.
But even after checking every connection with a multimeter,
<span style="font-family:monospace">JTAGWhisperer</span
> would
do apparently nothing
after receiving the first chunk of data.
I eventually gave up searching for the cause.
<p />Instead I decided to write a very simple
Arduino program that allows
direct interaction
with the JTAG interface from
a serial terminal.
It is called <span style="font-family:monospace">jtagbang</span
> because
it is essentially bit-banging on the
JTAG pins.
By pure coincidence,
it also requires frequent use
of the exclamation mark (<q>bang</q
>)
when talking to it.
<p />I didn't know anything about JTAG until three days ago.
Now I know that it is awesome.
The point of JTAG is to connect to
any number of chips in some circuit design,
taking up next to no space on the board,
requiring only very simple support from the chip,
and allowing the user to inspect and manipulate
virtually every pin and connection
at any time
without touching anything.
I call it <em>fucking magic</em
>.
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/img/2013/extest.medium.jpg"><img src="/img/2013/extest.klein.jpg" alt="extest.klein.jpg" /></a
></div
><div class="floatcaption">These LEDs are lit because I told the chip
I needed those outputs on for testing purposes.
</div
></div
><p />Unfortunately I cannot explain the magic in the space of this post,
however, here is a link to the
<a href="http://standards.ieee.org/findstds/standard/1149.1-2013.html">IEEE specification</a
>.
While IEEE doesn't want you to read their standards,
someone has helpfully put
<a href="http://www.slideshare.net/SonLeVu/jtag-specification-ieee-std-1149-12001">the 2001 version on slideshare</a
>…
Reading that spec is still not much fun,
but I made a drawing of the
<a href="https://twitter.com/thequux/status/370004863851061248">important part</a
>.
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/img/2013/tapstates.medium.jpg"><img src="/img/2013/tapstates.klein.jpg" alt="tapstates.klein.jpg" /></a
></div
></div
><p />So, long story short:
Upload the attached sketch to an Arduino,
take a peek at the top of the file maybe,
and connect to it with a terminal emulator (read <span style="font-family:monospace">minicom</span
>)
or the Arduino IDE's serial monitor (set to line-ending <q>Newline</q
>).
Enter a capital <span style="font-family:monospace">X</span
> and it will interrogate the
JTAG interface
to find all the connected devices (chips).
It lists their built-in identification codes which take the form
of 32 bits in four groups:
<p /><pre><code>59604093 [0101 1001011000000100 00001001001 1]
</code
></pre
><p />The groups are, from most to least significant bit:
4-bit product version (<em>5</em
>),
16-bit product code (<em>9604</em
> <a href="http://bsdl.info/list.htm?search=1001011000000100000010010011">is the
XC9572XL</a
>),
11-bit manufacturer code (<em>00001001001</em
> is Xilinx),
and one bit that is always 1 for thaumaturgic reasons.
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/img/2013/jtagenum.medium.png"><img src="/img/2013/jtagenum.klein.png" alt="jtagenum.klein.png" /></a
></div
><div class="floatcaption">I should find a PC mainboard to try this with.
</div
></div
><p />Next, I need to get the CPLD programmed.
Xilinx uses (X)SVF files for this,
a file format that describes what to do
on a JTAG interface in a more high-level fashion
than my bit-banging.
I need a <q>player</q
> for this format that translates
standard SVF commands into
<q>bang language</q
>
and vice-versa.
The good thing is that I can now do this
in a high-level programming language of my choice
entirely on the host
instead of cramming it into the Arduino.
<p />The adventure
after that
will be
learning VHDL
and designing an actual integrated circuit.
<p /><b>Attachment:</b
>
<a href="/log/2013/aug/jtagbang.ino">jtagbang.ino</a
> (v0.1)
<p />I am releasing the code under the terms of the
quite permissive <a href="http://opensource.org/licenses/ISC">ISC license</a
>.
JSON with blobs, still context-freehttp://www.khjk.org/log/2012/jun/datalang.htmlpescopesco@khjk.org2012-06-11T00:00:00Z
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/log/2012/jun/datalang.medium.jpg"><img src="/log/2012/jun/datalang.klein.jpg" alt="datalang.klein.jpg" /></a
></div
><div class="floatcaption">Pen and paper design.
</div
></div
><p />Inspired by their talk
<a href="http://media.ccc.de/browse/congress/2011/28c3-4763-en-the_science_of_insecurity.html"><q>The Science of Insecurity</q
></a
>
I took Meredith Patterson and Sergey Bratus by their word and
tried to solve my next network communication problem
without crossing the line beyond deterministic context-free
languages.
<p />The upshot of said talk was that most if not all security problems
stem from the fact that some software component could not
foresee the consequences of its input.
From a language-theoretic point of view,
the problem boils down
to <em>recognizing</em
> the set (<q>language</q
>) of acceptable inputs.
There are different classes of languages
whose recognizers require increasingly complex mechanisms.
Things are basically pleasant with
<em>regular</em
> languages and one step up,
aforementioned <em>deterministic context-free</em
> ones.
Up to this point we can algorithmically decide
whether two specifications describe the same language;
whether two peers in the network are cleanly interoperable.
<p />When I was looking for a good
<a href="http://en.wikipedia.org/wiki/Comparison_of_data_serialization_formats">data serialization format</a
>,
in addition to my original requirements,
I went looking for one that had a deterministic context-free grammar.
Incidentally, one of the things I wanted to be able to do
was efficiently transfer relatively large blocks of arbitrary data.
Unfortunately, what immediately catapults you into the
land of (mildly) context-<em>sensitive</em
> languages are length fields.
<p />JSON (as implied by the title) would have been my favorite choice,
but the best way to put binary blobs in it is
by encoding them as Base64-encoded strings.
<p /><pre><code>{ "message": "Hi, Bob!"
, "attachment": "ZGFmdXFpc3RoaXM="
}
</code
></pre
><p />For one thing, this means overhead in encoding time, decoding time and data volume.
Also it is unsatisfying because one property of JSON is
self-descriptiveness:
recognizing a JSON value reveals its type.
Base64 blobs would be hidden in strings and force the recipient
to know exactly where to expect them.
<p />A short note about overhead and efficiency concerns.
It is generally rightfully considered foolish to optimize prematurely.
From <em>most</em
> standpoints, computers are fast, bandwidth is cheap
and you are probably wasting ten times as much elsewhere
as avoiding Base64 would ever save.
Nevertheless, optimizing for efficiency is not useless
and in the right place, a constant factor can make all the difference.
Base64 <em>will</em
> turn your 3GB download into a 4GB one.
More importantly,
I am treating this endeavor as much as an academic as a practical one,
asking <q>could we</q
> as much as <q>do we want to</q
>.
So below is the answer I came up with.
<p />The idea is to break binary data into chunks of uniform size.
I chose 4096 bytes rather arbitrarily.
Allow one final chunk of variable length and encode that one in Base64.
So every 4kB, there is one character (<span style="font-family:monospace">#</span
>)
which means <q>another 4k coming</q
>.
There need not be any such <q>raw chunks</q
>;
they are always followed by exactly one
(possibly empty) Base64 string enclosed in <span style="font-family:monospace">%</span
>.
Examples:
<p /><pre><code>#.....#.....%ZGFmdXFp%
#.....#.....%%
%ZGFmdXFp%
</code
></pre
><p />This syntax is added to JSON, along with a few other extensions.
<p /><b> Design goals </b
>
<p /><ul><li>Stay deterministic context-free.
</li
><li>Avoid escaping or re-encoding every byte in binary blobs.
</li
><li>Simple grammar.
</li
><li>Self-describing structure.
</li
><li>Stay reasonably human-readable and human-writable.
</li
><li>Minimize attack surface for bugs.
</li
></ul
><p /><ul><li>Plus: Allow exact representation of binary floating point numbers.
</li
><li>Plus: Allow strings to use any character encoding.
</li
><li>Plus: Do not use newlines as syntax, allow arbitrary values to be serialized
into single lines.
</li
><li>Plus: Provide for easy streaming of values.
</li
></ul
><p /><b> Non-goals </b
>
<p /><ul><li>Optimal size.
</li
><li>Optimal speed.
</li
></ul
><p />Notwithstanding the goal to support efficiency with large blobs,
this format is not meant to squeeze every last bit out of everything.
That conflicts with self-descriptiveness and is what
<a href="https://developers.google.com/protocol-buffers/">Protocol Buffers</a
>
are for.
<p /><b> Characteristics </b
>
<p /><ul><li>Proper superset of UTF-8-encoded JSON.
</li
><li>Types:
<ul><li>Null
</li
><li>Boolean
</li
><li>Number
</li
><li>Byte-Array
</li
><li>String
</li
><li>List (= JSON <q>arrays</q
>)
</li
><li>Record (= JSON <q>objects</q
>)
</li
></ul
></li
><li>Defined in <a href="http://www.faqs.org/rfcs/rfc5234.html">ABNF</a
>.
<ul><li>~150 lines.
</li
><li>Transcribes trivially to a PEG.
</li
></ul
></li
><li>All syntax is ASCII.
<ul><li>No other external encodings allowed.
</li
></ul
></li
><li>Strings are tagged with their encoding.
<ul><li>No tag means UTF-8.
</li
><li>JSON-style unicode escapes <span style="font-family:monospace">\u....</span
> supported for compatibility.
<ul><li>UTF-16 surrogate pairs recognized by grammar.
</li
><li>Parsers SHOULD properly recode these for UTF-8 strings.
</li
></ul
></li
></ul
></li
><li>Arbitrary bytes in strings via hexadecimal escapes <span style="font-family:monospace">\x..</span
>.
</li
><li>Numbers are arbitrary-precision rationals.
</li
><li>Hexadecimal notation for numbers supported.
<ul><li>Including hexadecimal fraction and exponent notation.
</li
></ul
></li
><li>Top-level <q>document</q
> consists of one value of any type.
<ul><li>JSON only allows arrays and objects.
</li
></ul
></li
><li>Defined syntax for top-level value streams.
<ul><li>Values terminated by newlines.
</li
><li>Allows parser for rule <span style="font-family:monospace">stream-element</span
> to
simply be called repeatedly on input stream.
</li
></ul
></li
></ul
><p /><b> Example </b
>
<p /><pre><code>{ "null": null
, "boolean": true
, "integer": 1234
, "rational": 1234.56
, "exponent": 1234.56e2
, "hexadecimal": 0x123AB.CDxE
, "bytes": %ZGFmdXFp%
, "string": "Hello"
, "encoding": "Mot\xF6rhead"_latin1
, "list": [23,"skidoo"]
, "record": {}
}
</code
></pre
><p /><b> Show me the code! </b
>
<p />Glad you asked!
The child currently carries the rather stupid working title <q>datalang</q
>
and resides in a <a href="http://code.khjk.org/datalang/">repository</a
> here:
<p /><pre><code>darcs get http://code.khjk.org/datalang/
</code
></pre
><p />Included is the <a href="http://code.khjk.org/datalang/grammar.abnf">ABNF grammar</a
>
as well as a <a href="http://code.khjk.org/datalang/parser.c">demo parser</a
>
implemented in C using <a href="http://github.com/UpstandingHackers/hammer"><span style="font-family:monospace">hammer</span
></a
>.
Oh right, <span style="font-family:monospace">hammer</span
>.
Given that this post has already turned into a novel,
I am going to save that for later.
<p /><b>PS:</b
>
If anyone thinks of a better name than <em>datalang</em
>,
your suggestion is very welcome at my easily-guessed email address.
Fingerprints are so 90shttp://www.khjk.org/log/2012/apr/smp.htmlpescopesco@khjk.org2012-04-24T22:22:00Z
<p /><div class="float" style="float:right"><div class="floatcontent"><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2012/apr/smp/p01.klein.jpg" alt="p01.klein.jpg" /></div
></div
></div
></div
><p />A few years back I prepared a presentation on the so-called
<em>Socialist Millionaires' Protocol</em
> (SMP) for a university seminar.
SMP is a solution to the problem of key authentication
devised for <a href="http://www.cypherpunks.ca/otr/">OTR</a
> (Off-the-Record),
the system for instant-messaging encryption.
<p />Today I held a short version of the presentation
for non-mathematicians at the <a href="http://www.hamburg.ccc.de/">CCC Hamburg</a
>.
For the benefit of the Internet,
the awesomely hand-made <a href="/log/2012/apr/smp/slides/">slides</a
>
are in English.
There is also a handy <a href="/log/2012/apr/smp/handout.pdf">hand-out</a
>
with a protocol diagram.
<p />The <a href="/log/2012/apr/smp/smp.pdf">written presentation</a
> for the course
is <strong>completely in German</strong
> and math-rich.
I did try hard to make it a clear read for the so-inclined.
Have fun! :)
An introduction to Bitcoinhttp://www.khjk.org/log/2011/jul/bitcoin.htmlpescopesco@khjk.org2011-07-05T10:30:00Z
<p />I held a little intro talk about <a href="http://www.bitcoin.org/">Bitcoin</a
> last night
at a local Linux meetup kinda thing. It was a light technical description of
what the system is and how it works.
<p />Here are the <a href="/log/2011/jul/btc.pdf">slides</a
> and
their LaTeX <a href="/log/2011/jul/btc.zip">sources</a
>. That is all.
Blind signature basicshttp://www.khjk.org/log/2011/apr/blindsigs.htmlpescopesco@khjk.org2011-04-29T21:00:00Z
<p /><div class="float" style="float:left"><div class="floatcontent"><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2011/apr/signature.klein.jpg" alt="signature.klein.jpg" /></div
></div
></div
></div
><p />I'm starting work on my diploma thesis this month.
The exact topic isn't set in stone yet,
but it will be something crypto.
If everything goes dreamy-awesome,
I'll find something nice to write about lattice-based
blind signatures or somesuch.
Background:
<p /><ul><li>Blind signatures are used for e-cash.
I'm fascinated with e-cash.
</li
><li>Lattice-based methods are candidates for quantum-resistant systems.
Appearently there's lots of current research going on.
</li
><li>I've done some lattice theory at Charles University in Prague.
Never thought I'd see it again, so wouldn't that be something.
</li
><li>There's zero knowledge in there somewhere.
Another fascination.
</li
></ul
><p />So, time to sum up the basics.
As far as my history serves,
David Chaum invented blind signatures in the 80s for electronic voting
but nobody wanted to buy that,
so he also invented electronic cash.
Then he got really paranoid and didn't sell it either.
Real quick summary. ;)
<p />Anyway…
<p />The principle is to mix whatever you want signed
(electronic voting ballot, 100 EUR banknote)
with a random <q>blinding factor</q
>
and divide that out only after Trent
(your government, bank) has signed.
Thus Trent cannot recognize and connect the note to you
when it comes back to him later.
<p />The classic algorithm is based on RSA and is painted up fast.
Unfortunately, my awesome markup language still has no fancy math support
so you have to live with ASCII art:
<p /><pre><code>m = message to be signed
e = public "encryption" (i.e. verification) exponent
n = public modulus
d = secret "decryption" (i.e. signing) exponent
k = blinding factor (just a random number)
x^(de) = x^(ed) = x (mod n) -- RSA property
Alice prepares: mk^e -- blinded message
Trent signs: (mk^e)^d = m^d k
Alice unblinds: m^d k / k = m^d -- signed message
Bob can check: (m^d)^e = m
</code
></pre
><p />One might think that signing something completely blindly might be a bad idea.
After all, a bank needs to know the value of the note it is signing.
To ensure <em>any</em
> desired property of the signed document,
Trent can require a <q>cut-and-choose</q
> step.
In this case Alice must give him <em>n</em
> different but equivalent messages.
He chooses one of them and asks Alice to unblind all the others.
Trent signs the remaining blinded one if
the others satisfy the desired property.
Alice's chance to cheat of <em>1:n</em
> can be made unattractive
by attaching a suitable penalty.
Making a stupid little time waster with Javascripthttp://www.khjk.org/log/2011/mar/memory.htmlpescopesco@khjk.org2011-03-14T22:13:00Z
<p /><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2011/mar/memory.klein.png" alt="memory.klein.png" /></div
><div class="floatcaption">Awesome!
</div
></div
><p />I used the weekend to code
<a href="http://kompilierfreizeit.de/memory">a little memory game</a
>
in Javascript.
Apart from being pointless and annoying
(especially if your visual memory sucks as much as mine),
I think it turned out lovely!
<p />I was hoping to gain some experience with user-facing Javascript
from it, as
previous excursions into the
<a href="http://javascript.crockford.com/javascript.html">misunderstood programming language</a
>
have been minimal in <a href="http://www.khjk.org/log/2010/aug/pixls.html">one way</a
>
or <a href="http://www.khjk.org/log/2010/nov/jsbigint.html">another</a
>.
So, this time I got to use some actual objects.
Structured data, woo!
<p />I poked around the design space a little to see what was up with
Crockford's take on <a href="http://javascript.crockford.com/prototypal.html">prototypal inheritance</a
>
vs. <a href="http://stackoverflow.com/questions/383402/is-javascript-s-new-keyword-considered-harmful">others' assertions</a
>
that his dislike of the <span style="font-family:monospace">new</span
>-operator was ill-founded.
You can look at <a href="http://kompilierfreizeit.de/memory/memorygame.js">the code</a
>
to see what I settled on in this instance,
but maybe I'll leave that discussion for another post.
Suffice it to say, it's a bit of a mish-mash
but I'm sure it will crystallize nicely.
<p />Oh yeah, and this is going onto my <strong>side project</strong
>
(I desperately needed one!) site
<a href="http://kompilierfreizeit.de/">kompilierfreizeit.de</a
>.
I'm going to collect some nice other time wasters there,
but won't say it out too loud, yet.
That would just make me feel obligated…
<p />Have fun!
Tron (1982) awesomenesseshttp://www.khjk.org/log/2011/jan/tron.htmlpescopesco@khjk.org2011-01-27T00:30:00Z
<p /><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2011/jan/tron-pacman.klein.jpg" alt="tron-pacman.klein.jpg" /></div
></div
><p />I just (re)watched the original Tron. Highlights:
<p /><ul><li>Adam, Tron's user, has the phrase
<a href="http://en.wikipedia.org/wiki/Klaatu_barada_nikto"><q>gort klaatu barada niktu</q
></a
>
posted up in large letters on his cubicle wall.
This is not a reference to <q>Army of Darkness</q
> (1992, duh) but to
<q>The Day the Earth Stood Still</q
>.
I should watch that.
</li
><li>The screen in the control room of bad guy's ship shows pac-man in one scene
(see image above).
</li
><li>The image of Moses in Southpark episode <q>Jewbilee</q
> (s03e09) is modeled after
the Master Control Program (see image below).
</li
><li>Spotted <a href="http://en.wikipedia.org/wiki/Ken_Perlin">Kenneth Perlin</a
>
in the credits.
</li
></ul
><p /><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2011/jan/mcp-moses.klein.jpg" alt="mcp-moses.klein.jpg" /></div
></div
>carpalx keyboard layouts for X11http://www.khjk.org/log/2011/jan/carpalx.htmlpescopesco@khjk.org2011-01-23T17:14:00Z
<p /><a href="http://mkweb.bcgsc.ca/carpalx/"><span style="font-family:monospace">carpalx</span
></a
> is a program for the algorithmic
evaluation and optimization of keyboard layouts.
The site is a wealth of information and adds a nice scientific angle
to the endless discussion of alternatives to the QWERTY layout.
First of all
it provides evaluations and comparisons of several popular layouts,
including <a href="http://en.wikipedia.org/wiki/Dvorak_Simplified_Keyboard">Dvorak</a
>
and <a href="http://colemak.com/">Colemak</a
>.
In addition
it describes a number of layouts generated by the program,
optimized to certain parameters and side conditions.
Within the given effort metric,
they improve significantly over all alternatives.
<p /><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2011/jan/qgmlwb.klein.png" alt="qgmlwb.klein.png" /></div
><div class="floatcaption"><span style="font-family:monospace">carpalx</span
>' fully optimized layout with Colemak lettermask, <q>QGMLWB</q
>.
</div
></div
><p />I've created an XKB definition file that allows the five featured layouts
to be used with X11.
In fact, it also includes the <q>worst layout</q
> TNWMLC
which <em>maximizes</em
> the effort metric. Don't use that, though.
<p />To install, copy <a href="/log/2011/jan/carpalx.xkb"><span style="font-family:monospace">carpalx.xkb</span
></a
> to your XKB
symbols directory (path my differ slightly)
<p /><pre><code>$ cp carpalx.xkb /usr/share/X11/xkb/symbols/carpalx
</code
></pre
><p />and add the following lines to <span style="font-family:monospace">/usr/share/X11/xkb/symbols.dir</span
>:
<p /><pre><code>-dp----- a------- carpalx(qgmlwb)
--p----- a------- carpalx(qgmlwy)
--p----- a------- carpalx(qfmlwy)
--p----- a------- carpalx(qwkrfy)
--p----- a------- carpalx(qwyrfm)
--p----- a------- carpalx(tnwmlc)
</code
></pre
><p />Now it should be possible to load the layouts with, e.g.
<p /><pre><code>$ setxkbmap carpalx # defaults to QGMLWB
$ setxkbmap -layout carpalx -variant qwkrfy # to select other variants
</code
></pre
><p />Technically, the following files should also be updated:
<p /><pre><code>/usr/share/X11/xkb/rules/xorg.lst
/usr/share/X11/xkb/rules/xorg.xml
</code
></pre
><p />For the time being I couldn't be bothered, but
I think that's necessary if the layouts are to show up in Gnome and KDE control
panels.
<p /><b>PS:</b
>
Dear search engines, X11 means X Window System,
means this lets you use carpalx layouts on Linux (desktops).
Or FreeBSD for that matter.
Getting FreeBSD on the Net via UMTShttp://www.khjk.org/log/2010/dec/umts-stick.htmlpescopesco@khjk.org2010-12-28T01:35:00Z
<p />So, now I know how to make these USB UMTS stick thingies work…
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/log/2010/dec/umts-stick.medium.jpg"><img src="/log/2010/dec/umts-stick.klein.jpg" alt="umts-stick.klein.jpg" /></a
></div
></div
><p />I've been meaning to get one for a while,
so I got a cheap SIM-locked stick with a prepaid plan
for this year's post-christmas gadget shopping.
I'm recording how I got it to work as much for my own sake
as anyone else's…
<p />For FreeBSD, you need the <span style="font-family:monospace">u3g</span
> driver.
My model is a HUAWEI E1750,
which is one of several supported by the driver.
Here's what <span style="font-family:monospace">dmesg</span
> shows:
<p /><pre><code>ugen3.2: <vendor 0x12d1> at usbus3
ugen3.2: <vendor 0x12d1> at usbus3 (disconnected)
ugen3.2: <HUAWEI Technology> at usbus3
u3g0: <HUAWEI Technology HUAWEI Mobile, class 0/0,
rev 2.00/0.00, addr 2> on usbus3
u3g0: Found 4 ports.
</code
></pre
><p />Note the <q>4 ports</q
> message.
These things act like an old modem,
presenting a virtual serial port on which they accept
<q>AT</q
> commands.
Just like old times.
Actually, they often present several ports serving
different functions.
<p /><pre><code>$ ls /dev/cuaU?.?
/dev/cuaU0.0 /dev/cuaU0.1 /dev/cuaU0.2 /dev/cuaU0.3
</code
></pre
><p />You have to find the one to use by experimentation.
I learned that <span style="font-family:monospace">screen</span
> can actually be used as a terminal emulator
for this (instead of the clunky old <span style="font-family:monospace">minicom</span
>):
<p /><pre><code>$ screen /dev/cuaU0.0
</code
></pre
><p />Type <span style="font-family:monospace">AT</span
> and see if you get an <span style="font-family:monospace">OK</span
> reply.
<p />Most sticks also present mass storage devices.
There is usually an emulated <q>CD</q
> drive that contains
driver software for Windows and MacOS.
Some models appearently require this <q>disk</q
> to be <q>ejected</q
>
before even switching on the modem part.
I'm not sure if mine does;
the <span style="font-family:monospace">u3g</span
> manpage states that it should happen
transparently (look for <q>u3gstub</q
>).
<p />This model also has a microSD slot that can be used
like any regular usb storage device.
<p /><pre><code>cd0 at umass-sim0 bus 0 scbus0 target 0 lun 0
cd0: <HUAWEI Mass Storage 2.31> Removable CD-ROM SCSI-2 device
[...]
da0 at umass-sim1 bus 1 scbus1 target 0 lun 0
da0: <HUAWEI SD Storage 2.31> Removable Direct Access SCSI-2 device
</code
></pre
><p />Since the SIM card is usually protected by a PIN,
one needs to enter this before it will perform its function.
These and other UMTS-specific functions are simply accessed by
special AT commands.
Many have a <q>query</q
> variant ending in a question mark that will
report the current status and a <q>set</q
> function with an equals sign.
<p /><pre><code>> AT+CPIN?
+CPIN: SIM PIN
OK
> AT+CPIN=1234
OK
> AT+CPIN?
+CPIN: READY
OK
</code
></pre
><p />If you don't care about feedback,
the PIN entry can be done by writing directly to the device file.
<p /><pre><code>$ echo "AT+CPIN=1234" > /dev/cuaU0.0
</code
></pre
><p />My device now signals its ready status by blinking its LED in
a different color and rhythm.
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/log/2010/dec/umts-fiddling.medium.jpg"><img src="/log/2010/dec/umts-fiddling.klein.jpg" alt="umts-fiddling.klein.jpg" /></a
></div
><div class="floatcaption">Lots of fiddling later…
</div
></div
><p />Now for the PPP connection (just like old times!),
in <span style="font-family:monospace">/etc/ppp/ppp.conf</span
>:
<p /><pre><code>o2:
set log Phase Chat LCP IPCP CCP tun command +connect
set device /dev/cuaU0.0
set speed 460800
set dial "ABORT BUSY ABORT NO\\sCARRIER ABORT ERROR TIMEOUT 5 \
\"\" \
AT OK-AT-OK \
AT+CFUN=1 OK-AT-OK \
AT+CSQ OK \
AT+CGDCONT=1,\\\"IP\\\",\\\"internet\\\" OK \
AT+CGACT? OK \
AT+CGATT? OK \
ATD*99***1# CONNECT"
set timeout 180 # 3 minute idle timer (the default)
set ifaddr 10.0.0.1/0 10.0.0.2/0 255.255.255.0 0.0.0.0
set vj slotcomp off
set crtscts on
add default HISADDR # set default route
enable dns # set dns servers in resolv.conf
</code
></pre
><p />There is a bunch of stuff in there I just copied from examples
on the web. Let's see:
<p /><pre><code>set log Phase Chat LCP IPCP CCP tun command +connect
</code
></pre
><p />This makes <span style="font-family:monospace">ppp</span
> write all kinds of stuff to syslog (<span style="font-family:monospace">/var/log/ppp.log</span
>).
Very useful if something goes wrong, but completely optional.
The keywords after <q>set log</q
> select different parts of the whole
exchange and can be left out individually if desired.
<p /><pre><code>set speed 460800
</code
></pre
><p />No idea if this is necessary.
All kinds of <q>baud rates</q
> seem to just work.
<p /><pre><code>set dial "ABORT BUSY ABORT NO\\sCARRIER ABORT ERROR ...
</code
></pre
><p />The <q>dial</q
> setting is the conversation script <span style="font-family:monospace">ppp</span
> follows
in order to set off the right sequence of AT commands that
open the connection.
This is actually executed by chat(8).
See that manpage for the syntax.
The first line sets abort conditions.
If the modem responds with <q>BUSY</q
>, <q>NO CARRIER</q
> or <q>ERROR</q
>
to any command, <span style="font-family:monospace">ppp</span
> will consider the connection attempt failed.
<p /><pre><code>AT OK-AT-OK
</code
></pre
><p />This one uses a <q>subexpect/subresponse pair</q
> (-AT-OK) and
just makes sure that the modem responds (two tries).
<p /><pre><code>AT+CFUN=1 OK-AT-OK
</code
></pre
><p />This command turns the device <q>on</q
> if it
doesn't do it automatically.
Not sure what <q>on</q
> means exactly,
maybe that it actually goes on the air looking for its network.
My device does this automatically
after entering the PIN.
(I think that's what the changed LED pattern actually means.)
Anyway, the command just becomes a no-op
when it is issued again.
<p /><pre><code>AT+CSQ OK
</code
></pre
><p />This asks for the signal strength.
Not really necessary,
but maybe it fails if there is absolutely no signal.
<p /><pre><code>AT+CGDCONT=1,\\\"IP\\\",\\\"internet\\\" OK
</code
></pre
><p />This is somewhat important.
It sets the <q>APN</q
>.
I'm actually not sure what that is,
but I'm guessing some kind of logical access point
that the mobile network operator can define.
It's just called <q>internet</q
> in my case,
you have to set the right one for your network.
The Web has lists of these.
Google for <q>umts apn setting</q
>.
<p />Actually, the command sets a bundle of parameters known
as a <q>PDP context</q
>.
There can be many of these,
identified by numbers,
and the modem can be switched around among them.
<p /><pre><code>AT+CGACT? OK
</code
></pre
><p />Asks the modem which of the above-mentioned contexts
is active.
Not sure why this is necessary or useful,
copied it from an example.
<p /><pre><code>AT+CGATT? OK
</code
></pre
><p />Asks the modem whether it is attached to the GPRS service.
Presumably this can fail.
<p /><pre><code>ATD*99***1# CONNECT
</code
></pre
><p />Finally, this is the <q>dial</q
> command that opens the PPP
connection.
The number 99 seems to be standard.
The part <q>***1</q
> is optional and specifies the PDP context to
use.
<p />BTW, I'm guessing the ATD command can also be used to dial
the magic numbers used for activating prepaid time on the card.
Will have to try this later.
<p />Enough AT commands, back to <span style="font-family:monospace">ppp</span
> settings:
<p /><pre><code>set ifaddr 10.0.0.1/0 10.0.0.2/0 255.255.255.0 0.0.0.0
</code
></pre
><p />This is another important one.
I didn't look it up in the docs,
but one thing it accomplishes is explicitly setting the
point-to-point address on the network interface.
Appearently the necessity of this depends on the setup
of the peer, i.e. some supply their own address, some don't.
Mine didn't and I always got this in the log:
<p /><pre><code>tun0: Warning: iface add: ioctl(SIOCAIFADDR, 10.150.178.217
-> 0.0.0.0): Destination address required
tun0: Error: ipcp_InterfaceUp: unable to set ip address
</code
></pre
><p />The last two interesting ones are these:
<p /><pre><code>add default HISADDR # set default route
enable dns # set dns servers in resolv.conf
</code
></pre
><p />And that's basically it. Now
<p /><pre><code>$ ppp -ddial o2
</code
></pre
><p />will establish the connection,
disappear into the background,
and keep it open until killed.
There are a few other modes besides <span style="font-family:monospace">ddial</span
>,
consult the manpage.
<p /><strong>Interesting links:</strong
>
<p /><ul><li><a href="http://wiki.bsdforen.de/howto/umts_mit_huawei">A HOWTO in German language</a
>
at <a href="http://www.bsdforen.de/"><span style="font-family:monospace">bsdforen.de</span
></a
>
</li
><li><a href="http://docs.google.com/viewer?a=v&q=cache:RF06UMENUgkJ:www.wireless-products.dk/PDF-filer/RF-Solution-PDF/GPRS%2520AT%2520Command%2520List.pdf+at+cgatt"><q>AT commands for GPRS</q
></a
>,
some technical documentation from Wavecom at Google docs.
</li
><li><a href="http://www.xs4all.nl/~ernstagn/GPRS-HOWTO/GPRS-HOWTO-5.html"><q>Controlling your <span style="font-family:monospace">modem</span
></q
></a
>,
chapter 5 of <q>The Linux GPRS HOWTO</q
>
</li
><li><a href="http://docs.google.com/viewer?a=v&q=cache:5prtAfMEu7UJ:www.ripe.net/ripe/meetings/ripe-53/presentations/gprs.pdf+umts+apn"><q>UMTS/GPRS system/overview from an IP addressing perspective</q
></a
> (talk slides on Google docs)
</li
></ul
>das keyboard.http://www.khjk.org/log/2010/dec/das-keyboard.htmlpescopesco@khjk.org2010-12-26T14:46:00Z
<p />Yesterday's server fiddling was performed
on one of those things you can only get as presents.
<p /><div class="float" style="float:none"><div class="floatcontent"><a href="/log/2010/dec/das-keyboard.medium.jpg"><img src="/log/2010/dec/das-keyboard.klein.jpg" alt="das-keyboard.klein.jpg" /></a
></div
><div class="floatcaption"><q>das keyboard</q
>, model ultimate S, US layout.
</div
></div
><p />The <a href="http://www.daskeyboard.com/">official marketing</a
>
makes it look awesome.
I knew I wanted what they promised.
All the reviews on the web were distinctly positive.
But then there aren't many for a niche product like this.
The only way to get it is by mail order.
No stores carry it.
None of my friends have one.
No way to try before purchase.
And of course it's expensive.
<p />So I put it on my Amazon wishlist,
mentioning something about fetishism in the comment area.
<p />Here's the feature run-down:
It's a full-size USB/PS2 keyboard without labels;
<a href="http://en.wikipedia.org/wiki/Buckling_spring">buckling spring switches</a
>
like the infamous <a href="http://www.aibpc.com/?p=439">Model M</a
>;
n-key rollover;
color: black.
A keyboard for typists.
<p />My first impression was, erm,
I'll let the twitter quote stand for itself:
<p /><div class="float" style="float:none"><div class="floatcontent"><img src="/log/2010/dec/orgasmic.klein.png" alt="orgasmic.klein.png" /></div
></div
><p />The fine click,
perfectly synchronized by the laws of physics to the
actual moment of contact.
Words can not describe it.
<p />So yeah, I'm very happy so far.
As for the build, it seems sturdy,
has a good weight and proper rubber pads.
Let's hope it lasts a good few years!