How to read memory to get some information before target create？

[lz-bro]

For multi-cluster RISC-V core debugging, all clusters belong to the same tap. Actually, not all clusters may be used because there is partial-good situation. So how to read memory to get partial-good information before target create.

[TommyMurphyTM1234]

For multi-cluster RISC-V core debugging, all clusters belong to the same tap.

What exactly do you mean by "cluster" here?

Actually, not all clusters may be used because there is partial-good situation.

What does "partial-good situation" mean?

So how to read memory to get partial-good information before target create.

As far as I know the OpenOCD memory access commands require that the target has been created before they can be used.

• https://openocd.org/doc/html/General-Commands.html#:~:text=15.3%20Memory%20access%20commands

Maybe you can use raw/low level JTAG commands to do what you need?

• https://openocd.org/doc-release/html/JTAG-Commands.html

[aap-sc]

@lz-bro

general note: this tracker is probably not the best place to ask these questions, since it looks like a general OpenOCD question, not RISC-V specific one. Probably if you ask in OpenOCD mailing list you can get a more detailed answers.

back to your question: if you mean that some targets require additional initialization before the can be accessed/examined, then OpenOCD supports "deferred examine". See "-defer-examine" argument to target create in OpenOCD documentation. Maybe this could work for you?

Additionally we already had questions related to "how to read memory before init". Here is an example of a discussion: #1038 . If this is something that you want, you could take a look there (though, there are no detailed instructions there, only general guidance).

[lz-bro]

Thanks, my question was answered.

[lz-bro]

@TommyMurphyTM1234 @aap-sc I can not use irscan/drscan commands to read memory before target create, because irscan/drscan commands must be used after init and target create command must be used before init.

[aap-sc]

@lz-bro you can. Just create taps and don't create targets

[lz-bro]

@lz-bro you can. Just create taps and don't create targets

@aap-sc If I want to use irscan/drscan commands to read memory get some information, then target create based on the information， how can I do this?

[aap-sc]

@lz-bro it depends on what you need to do. You have several options.

1. Write a script that just launches OpenOCD two times. The first time you do low-level initialization and write some information in a file (jimtcl allows you to do that). Then run OpenOCD for the second time passing all the necessary parameters to your configs. This flow works quite well from my experience.
2. Alternatively , you can still create the targets but request a deferred examine (see -defer-examine flag). Then you don't need to run OpenOCD several times and you can do everything you need in one session.

[lz-bro]

@aap-sc Thanks for reply. For the second option, it is possible that some targets are not needed in the end base on the information.
for example：
If I use irscan/drscan commands to read memory and get 0b0011, I would target create 2 harts.
if I get 0b1111, I would target create 4 harts.
Can this be handled in one session?

[aap-sc]

@aap-sc Thanks for reply. For the second option, it is possible that some targets are not needed in the end base on the information. for example： If I use irscan/drscan commands to read memory and get 0b0011, I would target create 2 harts. if I get 0b1111, I would target create 4 harts. Can this be handled in one session?

No, I don't think so. AFAIK OpenOCD does not allow you to dynamically modify your target list. In your case I would recommend you to go with option 1.

[aap-sc]

@aap-sc Thanks for reply. For the second option, it is possible that some targets are not needed in the end base on the information. for example： If I use irscan/drscan commands to read memory and get 0b0011, I would target create 2 harts. if I get 0b1111, I would target create 4 harts. Can this be handled in one session?

No, I don't think so. AFAIK OpenOCD does not allow you to dynamically modify your target list. In your case I would recommend you to go with option 1.

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

[lz-bro]

@aap-sc Thanks for reply. For the second option, it is possible that some targets are not needed in the end base on the information. for example： If I use irscan/drscan commands to read memory and get 0b0011, I would target create 2 harts. if I get 0b1111, I would target create 4 harts. Can this be handled in one session?

No, I don't think so. AFAIK OpenOCD does not allow you to dynamically modify your target list. In your case I would recommend you to go with option 1.

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

How can I invoke openocd from the configuration file itself，Use exec command Error.

[aap-sc]

How can I invoke openocd from the configuration file itself，Use exec command Error.

Could you provide more context? I don't understand the question.

[lz-bro]

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

Is it means that I can invoke openocd from the .tcl, but I did not succeed, could you help me? like,

set command "$OPENOCD_PATH/openocd -f $CONFIG_FILE "

set output [exec $command]

set _CHIPNAME riscv
jtag newtap $_CHIPNAME cpu -irlen 5 -expected-id $TAPID

# handle output to determine how many targets to create
...

[aap-sc]

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

Do note, however, that .tcl itself supports this kind of scripting. So in theory you could invoke openocd from the configuration file itself. So in theory it may be possible to write this in a manner that is transparent to the user

Is it means that I can invoke openocd from the .tcl, but I did not succeed, could you help me? like,

set command "$OPENOCD_PATH/openocd -f $CONFIG_FILE "

set output [exec $command]

set _CHIPNAME riscv
jtag newtap $_CHIPNAME cpu -irlen 5 -expected-id $TAPID

# handle output to determine how many targets to create
...

here is an example that works for me (file test.tcl):

echo "hello, this is an OpenOCD!"

set CmdResult [exec build/Release/install_openocd/openocd/bin/openocd -c "echo \"     hello from second OpenOCD\"" -c shutdown]

echo "Below we have the help from additional openocd run:"
echo "$CmdResult"

echo "bye!"

shutdown

I run it as:

build/Release/install_openocd/openocd/bin/openocd -f test.tcl

this produces the following output :

Open On-Chip Debugger 0.12.0+dev-aap_sc_bleh_blue_experimental-g84705d15-dirty (2024-12-09-20:50)
Licensed under GNU GPL v2
For bug reports, read
	http://openocd.org/doc/doxygen/bugs.html
hello, this is an OpenOCD!
Below we have the help from additional openocd run:
Open On-Chip Debugger 0.12.0+dev-aap_sc_bleh_blue_experimental-g84705d15-dirty (2024-12-09-20:50)
Licensed under GNU GPL v2
For bug reports, read
	http://openocd.org/doc/doxygen/bugs.html
     hello from second OpenOCD
shutdown command invoked
bye!
shutdown command invoked

If you need something more complex -you can try to build on top of this example... You can also try to provide more context on the nature of the failure (I still don't quite get what do you mean by "did not succeed").