Difference between revisions of "JTAG Finder"

From eLinux.org
Jump to: navigation, search
m (Fix ugly link.)
 
(2 intermediate revisions by 2 users not shown)
Line 2: Line 2:
  
 
Due to his talk held at the [http://ph-neutral.org ph-neutral], this article is undergoing a major rework.
 
Due to his talk held at the [http://ph-neutral.org ph-neutral], this article is undergoing a major rework.
 +
 +
  
 
== (How) Does this work? ==
 
== (How) Does this work? ==
* The slides from my talk at [http://ph-neutral.org ph-neutral] 0x7d6 are here: [http://hunz.org/jtag.pdf pdf] [http://hunz.org/jtag.odp odp].  
+
* The slides from my talk at [http://ph-neutral.org ph-neutral] 0x7d6 are [[Media:Jtag.pdf|here]] [http://hunz.org/jtag.odp odp].  
(Hey! hunz.org does not respond. What about mirrors?) They explain how it works.
+
(hunz.org seems dead so pdf version now uploaded to elinux, enjoy)
  
 
* It did work with
 
* It did work with

Latest revision as of 11:13, 11 November 2011

User:Hunz is searching a maintainer for this project! Since he currently doesn't have the time/will to realize the further plans.

Due to his talk held at the ph-neutral, this article is undergoing a major rework.


(How) Does this work?

(hunz.org seems dead so pdf version now uploaded to elinux, enjoy)

  • It did work with
  • It didn't work with
    • a grandstream SIP phone yet (maybe some (n)TRST) problem or JTAG disabled
    • some kind of D-Link VPN gateway? (didn't ask for the exact name)

What does it look like?

scroll down to understand the meaning of the output

Further plans / What's missing?

Better methods

IDCODE

One should be able to reduce the number of necessary tries to (n * (n-1)) + (n-2) by reading the IDCODE from the Data Register:
Only TCK and TMS should be necessary for this, so one can look for TDO afterwards.
This method is described here: http://www.xilinx.com/xlnx/xil_ans_display.jsp?iLanguageID=1&iCountryID=1&getPagePath=11857
Possible problems:

  • This doesn't work if a devices uses Bypass as default instruction, does it?
  • One doesn't know the IDCODE in advance, thus detecting this pattern it is harder.

shift DR

shifting DR instead of IR might be faster? This is the case when the DR is smaller than IR.
The IDCODE must be 32 bits long, thus the DR register is 32 bits long if the IDCODE is loaded.
The IR length is usually about 36bits?


Better software

  • retry on timeout
  • live display for the status of each pin (at least while not scanning)
  • IO entry field (to disable single pins or assign const values to them)
  • (n)TRST support is missing (low active and with high active pulldown transistor)
  • detect hi impedance pins by looking for pins that change their state with pullups (100k) on/off
  • warn that hi impedance pins might be disconnected from the jtag!
  • try the Hi + hi impedance pins for TMS, TCK, TDI first and
  • the Low ones for nTRST first
  • read IDCODE?
  • assist the user in creating the relocations file (pullups on for all pins, user shall put cable between gnd and pin 1..30, one pin at a time)


Better hardware

  • Arduino / Freeduino based hardware?
    • Enable easy / fast deployment of the jtag finder
    • USB connected or not (standalone) operation mode
  • 1.8V support( needed for xbox 360 for example)
    • AVR alone won't do here (except for tiny2313 but it doesn't have that much pins)
    • thus I'm thinking about a cpld
    • could use a whole gang of tiny2313's perhaps
  • The current hardware is just a proof of concept. Let's build something fiddle-free.
  • I'd like to design a circuit board for this - if you want one as well please add yourself to this list:
    • Hunz
    • Obsidian
  • Let's collect some suggestions for the hardware here as well. My current suggestion is:
    • USB powered, USB-to-Serial interface (115200 baud)
    • socket for an ATMega16L or 32L? running at 7.3728 MHz (for the 115200 baud)
      • DIL or SMD? (TODO: compare the costs)
      • Obsidian: Wouldn't DIL make it easier for people to build their own boards?
    • bootloader for firmware updates
    • floppy connector compatible 2x17 pin header for the cable to the testpins
    • extra STK500 compatible pinheaders (maybe unpopulated) without the 330 Ohm resistors for other applications
    • costs: about 15-20 Euros
  • PLEASE add your comments/suggestions!
    • are 30 IOs enough or do you want more?
      • Certainly sounds enough for me for the first design.
    • the ATmega16/32 got 1/2kbyte of SRAM - it's enough for the JTAG scanner but maybe you'd like to do some other things with the board as well?
    • do you want more processing power? again it's enough for the JTAG scanner but one might want to do other things which need more processing power
    • or would you prefer to keep it simple, stupid?

Better firmware (AVR)

  • assembler for the scanning function -> speedup
  • TX buffer, interrupt based TX while doing sth. other
  • full duplex for uart? scan new pair while TX for old one running?
  • TX only interesting pins?
  • access the current TCK/TMS/TDI lines via RAM (ld/st) instead of IO (in/out) -> no case/if needed - just add the port offset
  • better delay
  • shift TMS/TDI, read TDO support - to use it as openwince-jtag cable
  • autonomous scanning?
  • 7.x MHz support

The Hardware

Explanations

not much to explain - had an ATmega32 lying around...


The Software

Explanations

  • there's a file named relocations.avr
    • it maps floppy connector / cable pin number to AVR port/pin
    • AVR port A = 0, B = 1, C = 2, D = 3
    • each AVR port got 8 pins
    • first line in the relocations file is pin 1, 2nd is pin 2, etc.
    • if there's a 13 in line 5 it means: cable pin 5 = avr port 13/8 (=1 = port B), AVR pin 13 mod 8 = pin 5
  • use pullups
    • should be useful for nearly any device to avoid thrashing input lines
    • disabling it leads to false positives
  • delay
    • currently (1.8MHz), 0 is safe (about 100kHz)
    • higher values will slow down the scanning process
  • shift pattern
    • you can leave this alone
  • # shifts
    • often you can reduce this to speed up the scanning process
    • number of times the 8bit-pattern is shifted into current TDI pin
    • so the number of bits shifted is #shifts * 8
    • detection fails if the value is too small - retry with a bigger value if necessary
    • 15 should be ok for most devices

Output

  • 4/3/2: TDO=6:11/20 means:
    • for TCK pin 4,
    • TMS pin 3,
    • TDI pin 2,
    • pin 6 ist a TDO suspect - the pattern was shifted in 20 times in at pin 2 (TDI) and found 11 times at pin 6
  • more advanced: output while scanning a blackberry 7290
    • scroll upwards and use the gallery

If it doesn't work

  • have you tried a higher number of shifts?
  • have you tried increasing the delay?
  • have you run the scan a 2nd and a 3rd time?
  • nTRST might be pulled low - the current software doesn't yet support this
    • enhance the software ;) (ask me if you need help - scroll down for contact)

JTAG might be disabled

  • by removing some 0-Ohm resistors on the circuit board (look for a connection to the testpins at missing resistors)
  • by adding additional resistors to set pins to a constant value
    • if you can't pullup/down a pin with 330 Ohm it's either Vcc/GND or disabled using this method
    • look for a connected resistor and try removing it
  • by a burned fuse directly in the chip
    • bad luck for mortals like us ;-(
  • by software
    • maybe keeping the cpu in reset state might help?

Comment from user of many JTAG tools

I think this tool can be quite dangerous when used on some JTAG ports, if you look at [1], for example MPC BDM port (also usable as a JTAG connection) is bad one. If the port has only TCK/TDO/TMS/TDI/nTRST/nSRST pins, then nothing wrong *should* happen, but if there are other signals, be very careful.

Also if you just try random IR/DR scans on a JTAG-enabled chip you may destroy it. For example some CPLDs have special IR command that erases internal flash memory, but this command requires clocking and some termination. When run improperly it can destroy the flash.

Links and related projects

Contact

  • Jabber: hunz at jabber.berlin.ccc.de
  • silcnet channel: c3a