04:47 <NishanthMenon> borneoa_: thanks for the hint https://github.com/nmenon/nm-dmem-openocd/blob/e01d0fe0e65ea73a3069efc92f627b20e0622af2/src/jtag/drivers/dmem.c worked like a charm.. https://github.com/nmenon/nm-dmem-openocd/blob/e01d0fe0e65ea73a3069efc92f627b20e0622af2/tcl/board/ti_am625evm-self.cfg + https://github.com/nmenon/nm-dmem-openocd/blob/e01d0fe0e65ea73a3069efc92f627b20e0622af2/tcl/target/ti_k3.cfg

04:48 <NishanthMenon> we did however uncover a hardwired firewall rule in the architecture so for R5s, we might need to introduce an emulation mode.. but yeah, i can happily debug M4F sitting on A53, no jtag pods needed

04:49 <NishanthMenon> using /dev/mem was'nt probably the smartest of choices though.. ;)

05:24 <borneoa_> NishanthMenon: OpenOCD sits in Linux userland. If you don't want to write a dedicated Linux kernel driver to access the memory, then /dev/mem is "the" way to go.

05:25 <NishanthMenon> borneoa_: yup.. https://github.com/Mellanox/rshim-user-space/blob/master/src/rshim.c -> I need to look closer as to what the folks had done..

05:25 <NishanthMenon> in kernel driver would have been nice.. but then... it would be just a single OS :D

05:28 <borneoa_> NishanthMenon: and it would be a silly duplicate of dev/mem. It would not be accepted upstream....

05:29 <NishanthMenon> more than that it'd conflict with usage of coresight drivers :(

05:43 <borneoa_> NishanthMenon: if you want to propose your new OpenOCD driver in Gerrit, I would be pleased to review it.

05:43 <NishanthMenon> borneoa_: aah thanks. I need to write up the emulation mode as well.. which I am attempting right now

05:44 <NishanthMenon> will add for review ofcourse.

05:44 <NishanthMenon> if there is strong pushback on /dev/mem, will consider switching over to userspace driver or other alternatives

08:04 <borneoa_> NishanthMenon: we have already some adapter driver using /dev/mem for fast going bit banging. Should be ok using it in this driver too.

08:06 <NishanthMenon> borneoa_: thanks for the hint. Will look closer to see if I can avoid duplicating code.

08:11 <borneoa_> NishanthMenon: I don't expect there is anything that can be shared to avoid duplications. But using the same way to open and mmap could help in the future if we add some security feature. As far as I know, today there is no per address access control on /dev/map

08:11 <NishanthMenon> Gotcha