13:55 <pdgendt> hi, I have an ATMEGA128 sitting in a daisy chain, and this controller has a JTAG bug: The JTAG instruction IDCODE is not working correctly. Data to succeeding devices are

13:55 <pdgendt> replaced by all-ones during Update-DR.

13:56 <pdgendt> How can I configure openocd to skip the IDCODE test or to do it manually?

14:00 <PaulFertser> pdgendt: hm, does OpenOCD actually stop if idcode doesn't match?

14:00 <PaulFertser> pdgendt: I think it reports a warning and carries on.

14:01 <pdgendt> ok, and how would I proceed to try and flash the device?

14:02 <pdgendt> if I try "reset halt" I get: JTAG scan chain interrogation failed: all zeroes

14:03 <PaulFertser> pdgendt: what are the other devices in the chain?

14:04 <pdgendt> an ATMEGA128 and an ATMEGA1284p

14:04 <pdgendt> I use the sysfsgpio driver

14:04 <PaulFertser> pdgendt: I think you probably do not need "reset halt" to try flashing either.

14:05 <PaulFertser> pdgendt: OpenOCD doesn't really support debug of AVRs, only flashing, so there's no real reset or target state handling.

14:06 <pdgendt> ok, flashing is the only thing I need, so that's fine

14:06 <PaulFertser> Target state is always returned as halted anyway.

14:06 <pdgendt> but if I try to read for example it says the Device ID and manufacturer IDs are all zero

14:06 <PaulFertser> pdgendt: to read using what command?

14:07 <pdgendt> flash read_bank

14:07 <PaulFertser> pdgendt: not sure that's really supported. Can you try flashing right away?

14:08 <pdgendt> well, I did add some code which I found on gerrit (not merged) were support for eeprom and fuses was added, so it could be that I have a different code base atm

14:08 <pdgendt> but it works if I have a single device on the jtag chain

14:08 <PaulFertser> pdgendt: reading doesn't seem to be supported at all anyway.

14:09 <pdgendt> I have added write_bank/read_bank/verify_bank for flash/eeprom/fuses

14:09 <PaulFertser> pdgendt: when OpenOCD talks to a device in the chain it loads BYPASS instruction in all other devices.

14:10 <PaulFertser> pdgendt: is it possible you specify wrong IR length in your target configs?

14:11 <pdgendt> I doubt that, again it works with a single atmega1284p but on another board we have an atmega128+atmega1284p

14:12 <PaulFertser> pdgendt: wrong IR length might work for a single device on the chain.

14:14 <pdgendt> jtag newtap $_CHIPNAME cpu -irlen 4 -ircapture 0x1 -irmask 0xf -expected-id $_CPUTAPID

14:15 <PaulFertser> Does this match both datasheets?

14:18 <pdgendt> yes, just verified in both datasheets

14:18 <PaulFertser> And does it fail the same flashing both the first device on the chain and the second?

14:18 <pdgendt> yes

14:20 <PaulFertser> You can run openocd with -d3 to see more of what it does internally

14:21 <pdgendt> PaulFertser: I'll try to poke it some more, thank you for your time already

14:22 <PaulFertser> pdgendt: I do not expect IDCODE bug to have any effect there but you'll need to dig deeper to see what's actually happening.

14:22 <PaulFertser> pdgendt: also, you didn't provide a single log yet, so there's not much food for thought.

14:29 <pdgendt> PaulFertser: I uploaded a log here: https://pastebin.com/wQqHb9ye not sure if this is helpful yet

14:41 <pdgendt> PaulFertser: it could be that my reset pin is inversed on the board, I will verify this first

14:41 <PaulFertser> pdgendt: yes, looks like no communication takes place at all

15:45 <pdgendt> PaulFertser: I did have some pins misconfigured, now I do get some response from the controllers. One question though, the chip ID has some version bits, is there a way to ignore this part of the ID?

15:47 <PaulFertser> pdgendt: what part of openocd complains?

15:48 <pdgendt> Warn : JTAG tap: atmega128.cpu UNEXPECTED: 0x5970203f (mfg: 0x01f (Atmel), part: 0x9702, ver: 0x5)

15:48 <pdgendt> Error: JTAG tap: atmega128.cpu expected 1 of 1: 0x0970203f (mfg: 0x01f (Atmel), part: 0x9702, ver: 0x0)

15:52 <PaulFertser> pdgendt: but that doesn't prevent other actions

15:53 <pdgendt> Ok, great! I also found the -ignore-version, sorry for the noise