19:31 <NishanthMenon> weird question - context: The SoC that I am working with is heterogenous, with the debug system accessible both from SoC address space and over jtag. Since I am running linux on the big ARM, i can directly access the debug address space from SoC view - so i can in theory use it to debug the cortex-M for example. We do have support for openOCD to get via jtag , but i am a bit lost how to describe access from SoC address view side.

19:31 <NishanthMenon> or even if openocd is the right tool even

19:35 <Hawk777> Hm, I’m not an expert, but maybe this could be done as a new adapter driver that implements a high-level interface, like HLA and some of the STLINKs do, I think? Not quite sure.

19:45 <NishanthMenon> hmm..

20:26 <borneoa_> NishanthMenon: the driver xlnx-pcie-xvc access an ARM DAP through a PCIe memory mapper address space. But in your case you don't have the DAP and you want to access the coresight devices directly from the memory map. I cannot find better reference for you than xlnx-pcie-xvc, and I cannot imagine yet any blocking point if you want to use this driver as a reference for making your own driver.

20:26 <NishanthMenon> thanks borneoa_

20:29 <borneoa_> NishanthMenon: in my experience, the coresight devices are mapped at different address if you access from DAP or from system memory. Maybe not your case. But if the address is different, it means you cannot rely on the debug ROM tables and you will have to hardcore addresses in the config file, or better you should have a translation from ROM addresses to system addresses; most of the times a simple offset is enough

20:30 <NishanthMenon> yes - it is a bitmask actually

20:42 <borneoa_> NishanthMenon: it could be specified in the driver config through a dedicated command

20:43 <NishanthMenon> thanks borneoa_ I am also looking at rshim driver as well..

20:45 <borneoa_> NishanthMenon: I see this kind of driver as something generic and useful. ARM also defines something similar but with some proprietary firmware on one CPU in order to debug the other CPU. Using openocd is a nice alternative.

20:45 <NishanthMenon> yes indeed..

20:45 <borneoa_> NishanthMenon: you are right! Also rshim is a valid reference