tcl/target/bl702: implement full software reset

In previous implementation, it was known that it does not perform
full reset, and that some peripherals, such as GLB core,
which handles among other stuff GPIOs, was not reset.

It was presumed, that full reset by software is not possible,
although, by accident, even when comment says that
CTRL_PWRON_RESET is set to 1, it is not
(value written into 0x40000018 supposed to be 0x7, not 0x6).

CTRL_PWRON_RESET indeed triggers full "power-on like" reset,
so this method is implemented in this commit.
There are some workarounds to make reset seamless, without any
error messages, which are described in comments of TCL script.

Only down-side of this reset is, that chip is halted after reset
bit later in BootROM than previous implementation,
but it's still good.

Change-Id: Ife2cdcc6a2d96a2e24039bfec149705baf046318
Signed-off-by: Marek Kraus <gamelaster@outlook.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8529
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
This commit is contained in:
Marek Kraus 2024-10-23 16:27:33 +02:00 committed by Antonio Borneo
parent bcebc84882
commit b68d23da3c
1 changed files with 14 additions and 1 deletions

View File

@ -34,6 +34,10 @@ $_TARGETNAME configure -work-area-phys 0x22020000 -work-area-size 0x10000 -work-
# Internal RC ticks on 32 MHz, so this speed should be safe to use.
adapter speed 4000
# Debug Module's ndmreset resets only Trust Zone Controller, so we need to do SW reset instead.
# CTRL_PWRON_RESET triggers full "power-on like" reset.
# This means that pinmux configuration to access JTAG is reset as well, and configured back early
# in BootROM.
$_TARGETNAME configure -event reset-assert-pre {
halt
@ -55,6 +59,15 @@ $_TARGETNAME configure -event reset-assert-pre {
# Do reset
# In GLB_SWRST_CFG2, clear CTRL_SYS_RESET, CTRL_CPU_RESET and CTRL_PWRON_RESET
mmw 0x40000018 0x0 0x00000007
# Since this full software reset resets GPIO pinmux as well, we will lose access
# to JTAG right away after writing to register. This chip doesn't support abstract
# memory access, so when this is done by progbuf or sysbus, OpenOCD will fail to read
# if write was successful or not, and will print error about that. Since receiving of
# this error is expected, we will turn off log printing for a moment,
set lvl [lindex [debug_level] 1]
debug_level -1
# In GLB_SWRST_CFG2, set CTRL_SYS_RESET, CTRL_CPU_RESET and CTRL_PWRON_RESET to 1
mmw 0x40000018 0x6 0x0
catch {mmw 0x40000018 0x7 0x0}
debug_level $lvl
}