Unbricked my LSP at last.
I had a unit that would not boot-up.
Used openocd to dump memory from 0xfffc0000 for 0x40000 bytes -that being the 256k Flash-
>dump_image flash.bin 0xfffc0000 0x40000
An hex 'compare' with 'u-boot.buffalo.updated' showed a 'FF' 'hole' in sector (27dec) starting at 0xfffdb000.
Actual 'hole' was 0x1bbba=0xfffdbbba for 0x446 bytes all were were 'ff's, which I assumed to be my boot-up problem.
It would be very interesting to establish how this came about.
Having messed about with drath's modified version of openocd -and, many thanks for his input on IRC-, I managed to repair my 'hole' by re-writing that part of the flash.
1> erased the sector number 27 (that's starting at 0xfffdb000 for 0x1000 bytes)
> flash probe 0
flash 'cfi' found at 0xfffc0000
> flash erase 0 27 27
erased sectors 27 through 27 on flash bank 0 in 0s 391000us
2> Copied the 0x1000 chunk of binary from 'u-boot.buffalo.updated' in 'LS-GL_FW_103-jtymod5-fixed' from offset 0x1b000 - 0x1bfff (that covered my strange block of 'ff's). Obviously you need to be using the same version of u-boot that you have installed on the unit.
3> Tranlated the binary into a series of 'write byte' commands that could be used with openocd's script command
where Var = DB000 to Var = DBFFF and Byte would be the byte for that location
mwb 0xFFFC5555 0xAA
mwb 0xFFFC2AAA 0x55
mwb 0xFFFC5555 0xA0
mwb 0xFFF<var> <byte>
for 4096 repetitions
mwb 0xFFFC0000 0xF0
4> Ran the script with the script command from openocd and after about 15 min it had flashed that sector with the good binary (flashing the full 64 sectors won't be very fast!).
Now my new 'dump_image' and 'u-boot.buffalo.updated' compared exactly, with the exception of size obviously.
At first the unit didn't seem to have changed, but once I unplugged the Jtag and started the LSP all was fine and I've again got a working unit
Thanks to everyone here who contibuted to this Jtag project.