Provide a customizable hook for handling target-specific GDB queries
Valgrind-clean, no new Clang analyzer warnings
Signed-off-by: Ian Thompson <ianst@cadence.com>
Change-Id: I684a259ed29f3651cbce668101cff421e522f79e
Reviewed-on: https://review.openocd.org/c/openocd/+/7082
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The command 'init' causes the execution of few lower level
commands, e.g. 'target init', and switches from command mode
COMMAND_CONFIG to COMMAND_EXEC, with an intermediate switch back
to mode COMMAND_CONFIG.
A timed target polling can occur during the execution of 'init'
and the target's status can trigger the execution of some events.
E.g. if a target has been left halted by a previous execution of
OpenOCD, the first poll will find the target halted, calling the
corresponding 'halted' event.
The event handler can use commands that can only be executed in
mode COMMAND_EXEC. If the poll happens while OpenOCD is in mode
COMMAND_CONFIG, the triggered handler will fail.
Prevent the target polling to operate during the execution of the
'init' command.
Change-Id: Ia435a5d2039be9b247e2336616dab53ed5d983ac
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7007
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
The same flag 'jtag_poll' is currently used as local data for the
command 'poll' and to temporarily mask the target polling.
This can cause unexpected behavior if the command 'poll' is
executed while polling is temporarily masked.
Add a new flag 'jtag_poll_en' to hold the temporarily mask
condition and keep 'jtag_poll' for the 'poll' command only.
While there, change the initial assignment of 'jtag_poll' using
the proper boolean value.
Change-Id: I18dcf7c65b07aefadf046caaa2fcd2d74fa6fbae
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7009
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
GDB remote serial protocol specifies breakpoint/watchpoint packet
responses can be an empty string to indicate the specified breakpoint
type is not supported. Add support for this response alongside existing
"OK", "E NN" replies.
Signed-off-by: Ian Thompson <ianst@cadence.com>
Change-Id: Iaf6280e4c936eb95a92bc80cc74d451ebb328dc3
Reviewed-on: https://review.openocd.org/c/openocd/+/7102
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
After a certain RTOS has been configured there is no mechanism
to go back to no RTOS support. It may be useful for debugging
purposes. With the provided modification, the "none" option
of RTOS is provided as a valid option.
It has been tested in two different board (Cortex M4 and Cortex M33).
Documentation has also been updated.
Signed-off-by: Asier Llano <allano@hubbell.com>
Change-Id: I602210bff31ccadd41c41e9454c52b5fffa1671e
Reviewed-on: https://review.openocd.org/c/openocd/+/7092
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tcl commands 'arm reg', 'arm mcr/mrc' do not work on M-profile based devices.
Isolate them from 'arm core_state' and 'arm disassemble' and do not chain
them from armv7m_command_handlers.
Change-Id: I2c6befdf82575e95cf05ed158ab5e6faa1a182c3
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7101
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Tcl command 'arm core_state' was exposed even on Cortex-M devices.
However it returned message "Unsupported Command" without error status
on such device.
Set the only possible arm->core_state ARM_STATE_THUMB in armv7m init.
Block setting core_state to arm on Cortex-M.
Change-Id: I9525553ac8863a6cf77bbacbcd57e354b6cfe1ca
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7100
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
If a Cortex-M MCU become unresponsive during a debug session and
re-examination fails to find MEM-AP, debug_ap pointer is set to NULL.
Eventual call of cortex_m_poll() dereferences debug_ap.
Check debug_ap validity at the begin of cortex_m_poll().
Change-Id: I9519f48760c91a48a9e5e8c34634d247098cb14a
Fixes: 35a503b08d (arm_adi_v5: add ap refcount and add get/put around ap use)
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7108
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
AARCH64_COMMON_MAGIC was same as CORTEX_A_COMMON_MAGIC, probably copy pasta.
Define unique AARCH64_COMMON_MAGIC
Change-Id: Ie30e0028453a1fce5624ecad9bf73d5ac3791281
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6997
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Unify common_magic type to unsigned int
Move common_magic to be the first member of the struct
Add unsigned specifier to xxx_COMMON_MAGIC #defines
Change-Id: If961d33232698529514ba3720e04418baf6dc6fe
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/6996
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The flash/nor subsystem uses bus_width and chip_width for CFI
external flash only. Drop setting these values for internal flash.
Change-Id: I64e79ab38b6e39e845ff96fbf4f60145e3b9690a
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7098
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The working area allocation routines use 4 byte word alignment.
In the corner case the size of the working area is not aligned,
target_alloc_working_area() of size = target_get_working_area_avail()
will fail because the size gets aligned up and does not fit to the area
which size is aligned down.
Align down the result of target_get_working_area_avail() to cope with that
corner case.
While on it use fancy ALIGN_... macros instead of bitwise and operator.
Change-Id: Ia2a1e861c401c2c78fe6323379a3776fb4f47b06
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7096
Tested-by: jenkins
Reviewed-by: Erhan Kurubas <erhan.kurubas@espressif.com>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Previously, if the target was halted before the FreeRTOS scheduler was
started but after at least one thread was created, then the current thread
would be set to whichever thread had the highest priority. In addition to
being misleading, because that thread is not actually running, it can
cause issues with gdb. For instance, breaking somewhere before the first
thread is created will show the current execution as the current thread,
but stepping over a line that creates a thread will cause the current
thread to switch to the newly created thread and the current execution to
disappear. The sudden disappearance of the current execution thread seems
to confuse some versions of gdb.
With this change, the value of xSchedulerRunning is checked to determine
whether the scheduler has been started. If it hasn't, then a fake
'current execution' thread is always created and made the current thread.
Signed-off-by: Frank Dischner <frank.dischner@gmail.com>
Change-Id: Ide0fe7d9ffb9fac95cee4c805735f434c7c4934d
Reviewed-on: https://review.openocd.org/c/openocd/+/6935
Tested-by: jenkins
Reviewed-by: Asier Llano <asierllano@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
When there are no rtos threads or none are active, a fake thread with
ID 1 is created for the current execution, but the current thread ID was
never set to this new fake ID. This would lead to an incorrect attempt to
read stacked registers for this fake thread. Explicitly setting the
current thread ID to the fake ID ensures that the registers are read from
the core instead of calling freertos_get_thread_reg_list.
Signed-off-by: Frank Dischner <frank.dischner@gmail.com>
Change-Id: I694509a0e01df089429b20ff1b879fc0592b532d
Reviewed-on: https://review.openocd.org/c/openocd/+/6934
Tested-by: jenkins
Reviewed-by: Asier Llano <asierllano@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
We found that the NPCX has an issue with the byte lanes so that non byte
aligned writes aren't working. To overcome this, for byte accesses we
copy the byte to be written to all of the byte lanes.
doc: Document command nu_npcx_quirks
Signed-off-by: benjbender <benjbender@gmail.com>
[Andreas Fritiofson: Squashed commits]
Signed-off-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Change-Id: I9ef63bf692f4e68f57459e1ec33f3abcbf533cd2
Reviewed-on: https://review.openocd.org/c/openocd/+/6630
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
This patch adds support for DAP interface to Cadence vdebug driver.
It implements a new transport layer for dapdirect_swd.
Change-Id: I64b02a9e1ce91e552e07fca692879655496f88b6
Signed-off-by: Jacek Wuwer <jacekmw8@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/6965
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
