Difference between revisions of "JTAG"

Revision as of 15:46, 30 March 2009

[SSinyagin] I'm trying to get use of JTAG for U-Boot debugging and probably FPGA debugging in the future. This page is not yet finished, and the work is in progress.

Building a JTAG connector for apf27Dev board

First you need a decent JTAG connector. The description below produces a 20-pin male JTAG connectior with standard ARM pinout.

List of material and part numbers at Conrad:

• 2x20-pin header, 2.54mm pitch (Conrad: 741973)
• ATA/IDE cable (Conrad: 971742)
• 2x10-pin IDC low profile header, 2.54mm pitch (Conrad: 743534)
• 20-way flat cable, 1.27mm pitch (Conrad: 609463)

Assembling it all together:

1. Solder the 40-pin header onto the J19 connector on the apf27dev board
2. Cut off a ~10cm piece from the ATA cable with the 40pin plug at the end.
3. Cut off ~10cm from 20-way flat cable
4. Carefully attach the 20-pin header to the flat cable. The thing is easy to break and it requires some forcing. Better buy a spare piece in advance.
5. Use a knife and a cutting pad and dissect the cable endings, ~3cm long. Dissect the whole width of the 20-way cable and only pins 40 to 31 on the 40-way ATA cable.
6. Strip the cable endings
7. Solder the two cables together, as specified below:

JTAG Adapter

I'm using the Amontec JTAGkey adapter. It has a male 20-pin plug and a 20-way female-to-female cable which fits directly into the cable as described above. A cheaper solution would be to use the Amontec JTAGkey-Tiny and install a 20-pin female plug on on the connector cable for apf27dev.

Configuring OpenOCD

yet to be done. Basic tests show that cable works as expected, but board-specific settings need some more time to figure out.

telnet_port 4444
gdb_port 3333
# GDB can also flash my flash!
gdb_memory_map enable
gdb_flash_program enable


interface ft2232
ft2232_device_desc "Amontec JTAGkey"
ft2232_layout jtagkey
ft2232_vid_pid 0x0403 0xcff8
jtag_khz 1000


reset_config trst_only combined
jtag_ntrst_delay 200

set  _CHIPNAME imx27
set  _ENDIAN little

# The bs tap
set _BSTAPID 0x1b900f0f
jtag newtap $_CHIPNAME bs \
  -irlen 4 -ircapture 0x1 -irmask 0xf -expected-id $_BSTAPID

# The CPU tap
set _CPUTAPID 0x07926121
jtag newtap $_CHIPNAME cpu \
  -irlen 4 -ircapture 0x1 -irmask 0xf -expected-id $_CPUTAPID

set _TARGETNAME [format "%s.cpu" $_CHIPNAME]
target create $_TARGETNAME arm926ejs \
    -endian $_ENDIAN -chain-position $_TARGETNAME -variant arm926ejs

$_TARGETNAME configure \
  -work-area-virt 0xffff4c00 -work-area-phys 0xffff4c00 \
  -work-area-size  0x8000 -work-area-backup 1


arm7_9 dcc_downloads enable


jtag newtap xc3s200a.fpga fpga \
	-irlen 6 \
	-irmask 0x3f \
	-ircapture 0x9 \
	-expected-id 0x2218093

Working with BDI2000

Here is a configuration file I use. It does not work yet but some settings are already right. More informations will be added later, when it works.

; bdiGDB configuration for ARMadeus APF27 board
; ---------------------------------------------
; Jonathan ILIAS-PILLET
; These settings do not work yet
[INIT]
; to be done

[TARGET]
CPUTYPE     ARM926E             ; processor core
CLOCK       6                   ; JTAG clock 3 = 4 MHz, 6 = 200KHz (setting used for testing)
WAKEUP	    500	                ; millisecond to wait after a reset to let target start
SCANPRED    1 6                 ; JTAG chain starts with FGPA (spartan3), it has a 6 bits Instruction Register
TRST        OPENDRAIN           ; pullup provided by iMX27 (§7.4 JTAG Controller Pin List)
RESET	    HARD 200            ;
ENDIAN      LITTLE              ; memory model is little endian
;VECTOR      CATCH 0x1f          ; not used now
BREAKMODE   HARD	        ; hardware breakpoints
;BREAKMODE   SOFT 0xDFFFDFFF     ;SOFT or HARD, ARM / Thumb break code
BDIMODE     AGENT

[HOST]
DEBUGPORT   2001                ; TCP port to connect GDB to
;FORMAT      ELF                 ; format of image files
;LOAD        MANUAL              ; load code manually after reset
PROMPT      APF27> 

[FLASH]
; to be done

[REGS]
;FILE	$reg926e.def   ; disabled while testing

Open issues

• TRST does not stop the CPU
• FPGA chip is not visible when OpenOCD detcts the TAPs.

most probably pull-up resistors are needed.