On this chip, the ndmreset bit in the RISC-V debug module doesn't
trigger a system reset like it should. To work around this, add a custom
"reset-assert" handler in its config file that resets the system by
writing to memory-mapped registers.
I've tested this workaround on a Sipeed Longan Nano dev board with a
GD32VF103CBT6 chip. It works correctly for both "reset run" and "reset
halt" (halting at pc=0 for the latter).
I originally submitted[1] this workaround to the riscv-openocd fork of
OpenOCD. That fork's maintainers accepted it, but have not upstreamed it
like they have several other of my changes.
[1] https://github.com/riscv/riscv-openocd/pull/538
Change-Id: I7482990755b300fcbe4963c9a599d599bc02684d
Signed-off-by: Thomas Hebb <tommyhebb@gmail.com>
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6957
Tested-by: jenkins
Reviewed-by: zapb <dev@zapb.de>
Add support for the TI K3 family AM243 SoC. This SoC is built on the
same base of AM642, so reuse the configuration with the exception of
Cortex-A53 which is not available on this device.
For further details, see https://www.ti.com/lit/pdf/spruim2
Change-Id: I971ba878b0f503e5120f6853634776eb61d05080
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7946
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The AM2x family of K3 SoCs typically do not contain a Cortex-A53 or
A72 processor. So, make the cpu "up" functions available when armv8
processor count > 0.
Change-Id: I985b194fe7cc63e4134ad84ccd921cc456eb412f
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7944
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Convert the memory access ap port num as a variable to allow support
for the AM2x family of K3 SoCs.
Change-Id: Ibd96c94055721f60d95179dab21d014c15b0f562
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7943
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Convert the Cortex-R5 ap port num as a variable to allow support for
the AM2x family of K3 SoCs.
Change-Id: I7dc8b459dca8b5f21395230b5cb782b14538bd48
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7942
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Convert the sysctrl ap port num as a variable to allow support for the
AM2x family of K3 SoCs.
Change-Id: I1b5b55e48240e6654779dd636fdf07bca055e192
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7941
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Firstly, create the TPIU nrf52.tpiu if using the nrf52 target. This is
standard, using AP 0 and TPIU base address 0xE0040000.
Secondly, add a pre_enable handler for this TPIU which configures the
TRACEMUX field of the TRACECONFIG register. This register is reset
every time the MCU resets, so the pre_enable handler creates a
reset-end handler to ensure the register remains set.
Change-Id: I408b20fc03dc2060c21bad0c21ed713eee55a113
Signed-off-by: Frank Plowman <post@frankplowman.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7901
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
While we can read and write from memory from the view of various
processors, all K3 debug systems have a AXI Access port that allows
us to directly access memory from debug interface. This port is
especially useful in the following scenarios:
1. Debug cache related behavior on processors as this provides a
direct bypass path.
2. Processor has crashed or inaccessible for some reason (low power
state etc.)
3. Scenarios prior to the processor getting active.
4. Debug MMU or address translation issues (example: TI's Region
Address Table {RAT} translation table used to physically map
SoC address space into R5/M4F processor address space)
The AXI-AP port is the same for all processors in TI's K3 family.
To prevent a circular-loop scenario for axi-ap accessing debug memory
with dmem (direct memory access debug), enable this only when dmem is
disabled.
Change-Id: Ie4ca9222f034ffc2fa669fb5124a5f8e37b65e3b
Reported-by: Dubravko Srsan <dubravko.srsan@dolotron.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7899
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
The Texas Instruments' K3 devices are a mix of AMP and SMP systems.
The operating systems used on these processors can vary dramatically
as well. Introduce a RTOS array variable, which is keyed off the cpu
to identify which RTOS is used on that CPU. This can be "auto" or
"hwthread" in case of SMP debug etc.
For example:
AM625 with an general purpose M4F running Zephyr and 4 A53s running SMP
Linux could be invoked by:
openocd -c 'set V8_SMP_DEBUG 1' -c 'set RTOS(am625.cpu.gp_mcu) Zephyr' \
-c "set RTOS(am625.cpu.a53.0) hwthread" -f board/ti_am625evm.cfg
Change-Id: Ib5e59fa2583b3115e5799658afcdd0ee91935e82
Reported-by: Dubravko Srsan <dubravko.srsan@dolotron.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7898
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Describe the SMP Armv8 cores in SMP configuration with coreid
explicitly called out. This allows for gdb session to call the smp
behavior clearly.
Change-Id: Ie43be22db64737bbb66181f09d3c83567044f3ac
Signed-off-by: Dubravko Srsan <dubravko.srsan@dolotron.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7897
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
When _v8_smp_targets is used with V8_SMP_DEBUG=1, describe the targets
as SMP targets. However, the variable expansion is not in the context of
a proc, and a typo in referring to global $_v8_smp_targets causes this
to fail. Just refer to $_v8_smp_targets directly.
Change-Id: Iffe5fd2703bed6a9c840284285e70b8a8ce84e17
Signed-off-by: Dubravko Srsan <dubravko.srsan@dolotron.com>
Signed-off-by: Nishanth Menon <nm@ti.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7896
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Add support for the TI K3 family AM62P SoC. This SoC is built on the
same base of AM62A7, so reuse the configuration with the exception of
the JTAG ID and the actual name used for the R5 core (moved from main
domain to wakeup domain).
For further details, see https://www.ti.com/lit/pdf/spruj83
Signed-off-by: Nishanth Menon <nm@ti.com>
Change-Id: I3a80be9e71204ed7697e51ac1ad488ef405744ef
Reviewed-on: https://review.openocd.org/c/openocd/+/7892
Reviewed-by: Bryan Brattlof <hello@bryanbrattlof.com>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
$_soc is set per platform, no point in duplicating _CHIPNAME to
explicitly set the information provided by $_CHIPNAME itself.
So move it out after the check for CHIP_NAME
Signed-off-by: Nishanth Menon <nm@ti.com>
Change-Id: I60d30d6a9a2ce352f66c5bc03075e4ba638e3062
Reviewed-on: https://review.openocd.org/c/openocd/+/7889
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
