Antonio Borneo [Mon, 30 Dec 2024 11:14:24 +0000 (12:14 +0100)]
target: cortex_m: add support of ARMv8.1-M register 'vpr'
The register 'vpr' is present when MVFR1.MVE is not zero.
For the moment, reuse the existing flag 'fp_feature'. To be
reviewed for the case of MVE supported without floating point.
The documentation of GDB [1] reports that the register 'vpr'
should be represented as 3 fields.
Richard Pasek [Wed, 11 Dec 2024 05:43:57 +0000 (00:43 -0500)]
Add Linux SPI device SWD adapter support
To alleviate the need to bitbang SWD, I've written a SWD SPI
implementation. This code is inspired by the work of luppy@appkaki.com
as shown at github.com/lupyuen/openocd-spi but with the desire to be
more generic. This implementation makes use of the more common 4 wire
SPI port using full duplex transfers to be able to capture the SWD ACK
bits when a SWD TX operation is in progress.
TEST:
Connects successfully with the following combinations:
Hosts:
Raspberry Pi 4B
Unnamed Qualcomm SoC with QUPv3 based SPI port
Targets:
Raspberry Pi 2040
Nordic nRF52840
NXP RT500
Change-Id: Ic2f38a1806085d527e6f999a3d15aea6f32d1019 Signed-off-by: Richard Pasek <rpasek@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8645 Reviewed-by: Tomas Vanek <vanekt@fbl.cz> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Reviewed-by: zapb <dev@zapb.de> Tested-by: jenkins
Antonio Borneo [Tue, 31 Dec 2024 13:47:02 +0000 (14:47 +0100)]
helper: list: rename macro clashing with sys/queue.h
The macro named LIST_HEAD() clashed with a macro of same name in
the GNU libc file sys/queue.h.
This causes a warning in MacOS build due to some other system file
including sys/queue.h.
Rename LIST_HEAD() as OOCD_LIST_HEAD().
Checkpatch-ignore: MACRO_ARG_REUSE
Change-Id: Ic653edec77425a58251d64f56c9f5f6c645ba0cd Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com> Reported-by: Andrew Shelley <ashelley@btinternet.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8683 Tested-by: jenkins Reviewed-by: Tomas Vanek <vanekt@fbl.cz> Reviewed-by: Andy <andrewjohnshelley@gmail.com>
R. Diez [Thu, 28 Nov 2024 20:47:34 +0000 (21:47 +0100)]
amend angie build definitions to fix make dist
"make dist" was broken because GNU Make was using a built-in rule
to try to build angie from angie.c . This is a limitation in Automake
when you add a whole subdir with the same name to EXTRA_DIST.
The Automake doc actually discourages adding whole subdirs.
Change-Id: I85ea4ecbd529b060c70f83bcfda7522e1730480d Signed-off-by: R. Diez <rdiezmail-openocd@yahoo.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8600 Tested-by: jenkins Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tomas Vanek [Tue, 10 Dec 2024 07:53:49 +0000 (08:53 +0100)]
drivers/cmsis_dap_usb_bulk: allow waiting for bulk write
No driver directly working with the USB hardware needs additional
time to complete the write op, they always return transfer complete
status immediately after submitting the transfer.
Although there is implemented correct waiting path in cmsis_dap_usb_write()
it was marked by error logs to catch any suspicious behaviour during
debugging of asynchronous libusb transfers.
However there are drivers which need waiting to finish write op:
at least usbipd-win, IP tunnelled USB driver, was reported
to flood the log with the related errors.
Change LOG_ERROR to LOG_DEBUG_IO in the code waiting to finish write op.
Tomas Vanek [Tue, 10 Dec 2024 07:26:48 +0000 (08:26 +0100)]
drivers/cmsis_dap: use blocking flag instead of wait timeout
CMSIS-DAP bulk backend read op used two timeouts: transfer timeout
used in libusb_fill_bulk_transfer() and wait timeout used optionally
in libusb_handle_events_timeout_completed().
The real usage is limited to two cases only:
1) blocking read: the same timeout is used for both transfer
and wait
2) non-blocking read: transfer timeout is used in
libusb_fill_bulk_transfer(),
libusb_handle_events_timeout_completed() is called with zero timeout.
Use blocking flag as read op parameter to distinguish between
these two cases.
See also [1]
Link: [1] 8596: jtag: cmsis_dap: include helper/time_support.h | https://review.openocd.org/c/openocd/+/8596 Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Change-Id: Ia755f17dc72bb9ce8e02065fee6a064f8eec6661
Reviewed-on: https://review.openocd.org/c/openocd/+/8639 Tested-by: jenkins Reviewed-by: Paul Fertser <fercerpav@gmail.com>
Tomas Vanek [Fri, 25 Oct 2024 15:17:22 +0000 (17:17 +0200)]
tcl/interface: Find proper alias for RP1 on Raspberry Pi 5
Previously, Linux assigned gpiochip numbers sequentially depending on
when the chip driver was probed. As RP1 is on the end of a PCIe link, it
is probed later than the on-board chips (including expanders connected
over SPI/I2C). This meant that RP1's gpiochip assignment was at an
offset that could potentially change.
A downstream kernel patch now assigns fixed offsets for RP1 and the
onboard gpiochips. Query the device tree to get proper GPIO_CHIP index.
Change-Id: I759978d4b3021c815a7d9febb41961cd1d3d185c Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/8650 Reviewed-by: Jonathan Bell <jonathan@raspberrypi.com> Tested-by: jenkins
Marc Schink [Fri, 29 Nov 2024 20:29:18 +0000 (20:29 +0000)]
tcl/target: Add support for Nordic nRF54L series
The RISC-V coprocessor is currently not supported. It is attached to the
DAP via AP#2 but the AP implementation is unknown.
The nRFL54L series uses resistive RAM (RRAM) as non-volatile memory
which can be programmed directly. Since it does not fit in the current
flash memory infrastructure of OpenOCD there is no NVM support so far.
Change-Id: I9934af4fd3bb8b7272954fc4b17638c7dabbbee0 Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8609 Tested-by: jenkins Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Marc Schink [Sat, 30 Nov 2024 14:32:09 +0000 (15:32 +0100)]
target/xtensa: Remove 'ERROR: ' prefix in error log
Remove the prefix since it is redundant. While at it, also
get rid of the useless exclamation mark.
Change-Id: I16fd6a88b533fac19b4c622cf9740fd32ba7892c Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8611 Reviewed-by: Richard Allen <rsaxvc@gmail.com> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins
R. Diez [Fri, 29 Nov 2024 20:09:50 +0000 (21:09 +0100)]
configure.ac: switch from $host to $host_os
Suggested during review https://review.openocd.org/c/openocd/+/8533
Only the OS part was being checked anyway.
The aim is to facilitate merging all $host_os checks in the future.
Change-Id: Idce1d5872cf19ef423429fa0c3b2ff7ee3945332 Signed-off-by: R. Diez <rdiezmail-openocd@yahoo.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8607 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins
R. Diez [Fri, 29 Nov 2024 20:24:23 +0000 (21:24 +0100)]
configure.ac: enable the Dummy adapter by default
The Dummy adapter is useful when developing generic JimTcl code.
Besides, the distributed BUGS file states that you should
try to reproduce any crashes with the Dummy adapter, so
it does not make sense that it is not enabled by default.
Change-Id: I145de06de4d2c0011619b1b941200b63e200db23 Signed-off-by: R. Diez <rdiezmail-openocd@yahoo.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8608 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins Reviewed-by: Paul Fertser <fercerpav@gmail.com>
Antonio Borneo [Fri, 22 Nov 2024 17:06:40 +0000 (18:06 +0100)]
target: aarch64: fix out-of-bound access to array
The command 'arm core_state' uses the enum in 'arm->core_state' as
an index in the table of strings to print the core state.
With [1] the enum has been extended with the new state for AArch64
but not the corresponding table of strings.
This causes an access after the limit of arm_state_strings[].
Rewrite the table using c99 array designators to better show the
link between the enum list and the table.
Add the function arm_core_state_string() to check for out-of-bound
values allover the file.
Change-Id: I06473c2c8088b38ee07118bcc9e49bc8eafbc6e2 Fixes: [1] 9cbfc9feb35c ("arm_dpm: Add new state ARM_STATE_AARCH64") Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8594 Tested-by: jenkins
Marc Schink [Wed, 23 Oct 2024 13:42:16 +0000 (15:42 +0200)]
doc/manual: Add section about logging
The log messages are very inconsistent across the code base. Add a
guideline for log messages to help improve consistency. The guideline is
based on the most commonly used style in the current code base.
Change-Id: I076d68abe588dd04b59580379e97b82d537def23 Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8576 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
rtt server: fix for dropped data when target has no space
rtt_write_channel may write less data than requested,
default device buffer size for channel 0 is 16 bytes,
so currently anything larger than this is dropped.
This fix implements per connection buffer and uses the
connection->input_pending flag to retry writes.
Pete Moore [Mon, 16 Dec 2024 13:02:14 +0000 (14:02 +0100)]
jimtcl: Fix command not found
During the ./configure build stage, error './configure.gnu: line 1: -e:
command not found' can occur.
Problem: the -e flag with echo is not portable. While some shells support
it (e.g., Bash), others (e.g., POSIX /bin/sh on some systems) do not.
Solution: replacing echo -e with printf, since printf is POSIX-compliant
and works consistently across different shells.
Change-Id: I9efbba662599a61bbe1fc56a33dc1ee7ad58826c Signed-off-by: Pete Moore <pmoore@mozilla.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8653 Tested-by: jenkins Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Reviewed-by: Andreas Fritiofson <andreas.fritiofson@gmail.com>
Pete Moore [Mon, 16 Dec 2024 15:55:06 +0000 (16:55 +0100)]
flash/nor/sfdp: Fix broken DEBUG log line on macOS
https://review.openocd.org/c/openocd/+/8439 changed variable `words`
from uint8_t to unsigned int in sfdp.c but failed to update the
LOG_DEBUG line to reflect the new type. On macOS this caused:
src/flash/nor/sfdp.c:107:28: error: format specifies type 'unsigned
char' but the argument has type 'unsigned int' [-Werror,-Wformat]
The formatting of the debug line has been updated to reflect the updated
type.
Change-Id: Ifc7ddb1279ab2603901c969d9c09af847f3a3caf Signed-off-by: Pete Moore <pmoore@mozilla.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8660 Tested-by: jenkins Reviewed-by: Tomas Vanek <vanekt@fbl.cz> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Henrik Mau [Wed, 13 Nov 2024 15:43:23 +0000 (15:43 +0000)]
target/xtensa: add maskisr command support for NX
Add maskisr command support to Xtensa NX targets allowing masking
of interrupts during single stepping.
Change-Id: I3835479de8015f1a2842afd1aeab24829e385031 Signed-off-by: Henrik Mau <henrik.mau@analog.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8575 Reviewed-by: Ian Thompson <ianst@cadence.com> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins
Antonio Borneo [Sun, 10 Nov 2024 13:30:25 +0000 (14:30 +0100)]
driver: stlink: get adapter speed through adapter_get_speed_khz()
The stlink driver, both in dapdirect and in HLA modes, pretends to
store locally the value of the adapter speed in order to use it
later-on during adapter initialization.
It doesn't work in dapdirect mode since the code to store locally
the value will not be executed until the adapter is already fully
initialized.
This cause an issue in dapdirect mode:
- due to the local value, still kept at -1, the adapter will be
initialized to the lowest clock speed (5 KHz on stlink v2 in SWD
mode);
- after the adapter initialization the framework will again set
the speed with the value requested by the user.
Some target, like nRF51822, only accepts JTAG/SWD speed in a
defined range of frequencies. The initial speed of 5 KHz used by
dapdirect can be out of range, making the target debug port not
working.
The adapter framework already stores the value of speed and makes
it available through adapter_get_speed_khz().
Drop struct hl_interface_param::initial_interface_speed.
Let the code to use adapter_get_speed_khz().
Change-Id: Ie11bf0234574f2a9180d3d3a16efb78e08dfcd86 Reported-by: Andrzej Sierżęga <asier70@gmail.com> Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8574 Reviewed-by: Andrzej Sierżęga <asier70@gmail.com> Tested-by: jenkins
Antonio Borneo [Mon, 10 Apr 2023 10:02:55 +0000 (12:02 +0200)]
checkpatch: exclude gerrit's Change-Id line from commit description
Checkpatch rejects patches that have empty commit description and
logs them with:
WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one
But if the patch has a gerrit's Change-Id line placed before the
line Signed-off-by, then checkpatch considers the Change-Id line
as a valid commit description text.
Use the Change-Id tag as a marker of the end of the commit message,
thus not counting its line as part of the commit description.
This patch is not relevant for the Linux kernel development process
as gerrit is not involved and the Change-Id tag is rejected. But
other projects, like OpenOCD, base the development on gerrit and
reuse kernel's checkpatch with flag '--ignore GERRIT_CHANGE_ID'.
This patch has been refused [1] in Linux upstream because it has
not been considered relevant for that project.
Let's take it as another add-on in OpenOCD checkpatch.
Jan Matyas [Thu, 10 Oct 2024 18:35:19 +0000 (20:35 +0200)]
autoconf: Add support for code coverage
Add support for code coverage collection. This helps
developers to check if their test scenarios really exercised
all the OpenOCD functionality that they intended to test.
- Option --enable-gcov has been added to configure.ac
which enables the coverage collection using Gcov. (Disabled
by default.)
- The steps to collect and inspect the coverage have been
described in HACKING file.
Change-Id: I259e401937a255e7ad7f155359a0b7787e4d0752 Signed-off-by: Jan Matyas <jan.matyas@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8521 Tested-by: jenkins Reviewed-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
target/breakpoints: fix types in `watchpoint_add_internal()`
There was a conflict:
1. commit 2cd8ebf44d1a ("breakpoints: use 64-bit type for watchpoint mask
and value")
2. commit 0bf3373e808a ("target/breakpoints: Use 'unsigned int' for
length")
The second commit was created erlier, but merged later so the types of
`mask` and `value` became `uint32_t` in `watchpoint_add_internal()`.
This created a bug:
`WATCHPOINT_IGNORE_DATA_VALUE_MASK` is defined as `(~(uint64_t)0)`.
Truncation to uint32_t makes it so the comparisons with the constant
don't work.
Change-Id: I19c414c351f52aff72a60330d83c29db7bbca375 Signed-off-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8500 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Reviewed-by: Jan Matyas <jan.matyas@codasip.com> Tested-by: jenkins Reviewed-by: zapb <dev@zapb.de> Reviewed-by: Karl Palsson <karlp@tweak.au>
Marek Kraus [Wed, 23 Oct 2024 14:27:33 +0000 (16:27 +0200)]
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
Marc Schink [Fri, 30 Aug 2024 08:47:55 +0000 (10:47 +0200)]
tcl/interface: Restructure parport config files
Provide cable specific configuration files like for the FTDI interface.
Depcrecate the old configuration files but keep them until the next
release for compatibility reasons.
Change-Id: I436bd60779a107120c9e1b1f0b8a69a39a240ad4 Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8514 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins
Antonio Borneo [Sat, 12 Oct 2024 09:48:05 +0000 (11:48 +0200)]
stlink: deprecate HLA support
The STLink API that supports dap-direct is available from STLink
firmware v2j24, published in early 2015.
We can reasonably expect that any old STLink still in use today
has got at least one firmware update during the last 10 years.
Most of the board files in upstream OpenOCD still use the STLink
in HLA mode. This limits the test coverage of the dap-direct code,
which was introduced in OpenOCD v0.11.0.
- Rename interface/stlink.cfg as interface/stlink-hla.cfg to still
provide support for HLA, adding a deprecated message.
- Rename interface/stlink-dap.cfg as interface/stlink.cfg to make
dap-direct the default trasport.
- Add a redirect file interface/stlink-dap.cfg for users that have
out-of-tree custom board files.
- Update all the board files to the new setup.
- Remove STLink HLA mentions from the documentation, while adding
a reference to interface/stlink-hla.cfg
Checkpatch-ignore: LONG_LINE
Change-Id: I99366bb03cd3b83f8f408514e657f30e59813063 Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8523 Tested-by: jenkins Reviewed-by: Andrzej Sierżęga <asier70@gmail.com>
Antonio Borneo [Mon, 15 Jul 2024 09:56:36 +0000 (11:56 +0200)]
target: cortex_m: fix polling for target kept under reset
In multi-target SoC not all the targets are running simultaneously
and some target could be powered off or kept under reset.
Commit 4892e32294c6 ("target/cortex_m: allow poll quickly get out
of TARGET_RESET state") does not considers the case of a target
that is kept in reset and expects the target to change state from
TARGET_RESET immediately.
This causes OpenOCD to log continuously:
Info : [stm32mp15x.cm4] external reset detected
Info : [stm32mp15x.cm4] external reset detected
Info : [stm32mp15x.cm4] external reset detected
Read again dhcsr to detect the 'stable' reset status and quit,
waiting for next poll to re-check the target's status.
Change-Id: Ic66029b988404a1599bb99bc66d4a8845b8b02c6 Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com> Fixes: 4892e32294c6 ("target/cortex_m: allow poll quickly get out of TARGET_RESET state")
Reviewed-on: https://review.openocd.org/c/openocd/+/8399 Reviewed-by: Tomas Vanek <vanekt@fbl.cz> Tested-by: jenkins
R. Diez [Sat, 1 Jun 2024 19:20:31 +0000 (21:20 +0200)]
configure: show adapter Xilinx XVC/PCIe in the configuration summary
Adapter Xilinx XVC/PCIe was not appearing in the configuration summary
because of the wrong variable name: build_xlnx_pcie_xvc
instead of enable_xlnx_pcie_xvc.
Also build this adapter automatically on Linux.
Change-Id: I69ea92f550052b9ce55ce32597ac446a15a87388 Signed-off-by: R. Diez <rdiezmail-openocd@yahoo.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8312 Tested-by: jenkins Reviewed-by: R. Diez <rdiez-2006@rd10.de> Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Jan Matyas [Thu, 3 Oct 2024 07:59:24 +0000 (09:59 +0200)]
gitignore: Start ignoring ".vscode"
To help the developers who use Visual Studio Code IDE,
ignore the ".vscode" folder in Git. This folder contains
local configuration of the VSCode workspace.
Change-Id: I1d54d8ce2bd0680f2fa1fb773bb33c786bdcc608 Signed-off-by: Jan Matyas <jan.matyas@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8518 Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Tested-by: jenkins
Antonio Borneo [Fri, 27 Sep 2024 20:36:52 +0000 (22:36 +0200)]
openocd: fix build with jimtcl 0.83
In jimtcl 0.82, the include file jim.h included in turn stdio.h
This made redundant to include the former in openocd source files.
Since jimtcl 0.83, jim.h drops the include of stdio.h, causing
openocd build to fail.
Include stdio.h in the files that need it.
Change-Id: Ic81c9b273d7520f4d2d8c32bc3e0a6bcfffb67ed Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8512 Reviewed-by: Jonathan McDowell <noodles-openocd@earth.li> Tested-by: jenkins
Antonio Borneo [Sun, 22 Sep 2024 16:32:21 +0000 (18:32 +0200)]
openocd: build: allow more socket connections on Windows
Cross compiling OpenOCD for Windows forces the maximum number of
open files (including sockets) to 64. See in include file
psdk_inc/_fd_types.h:
#ifndef FD_SETSIZE
#define FD_SETSIZE 64
#endif
This limit is far lower than the default value 1024 used by Linux.
In pull request #644 [1] in risc-v fork it's reported that:
- each socket server of OpenOCD (GDB, telnet, ...) uses one FD;
- each active connection to a socket server uses another FD;
- multi-core devices with 32 or more cores, each having a GDB
connection, already saturates the 64 available FD at the 26th
GDB connection.
The patch [2] proposed and merged in risc-v fork adds the compile
flag
-DFD_SETSIZE=128
to all the host types. While this looks fine for Windows, it
reduces the default value for Linux and other OS.
Add the compile flag FD_SETSIZE only to cross compile for Windows.
Antonio Borneo [Sun, 8 Sep 2024 19:20:48 +0000 (21:20 +0200)]
target: convert 'unsigned' to 'unsigned int'
Conversion done with
checkpatch --fix-inplace -types UNSPECIFIED_INT
Ignore the cast as they could be better addressed.
Fix only minor additional checkpatch issue (spacing and line
length).
Use Checkpatch-ignore below for the function pointers in the file
'armv7a_cache_l2x.h' that do not assign the identifier names to
the function arguments.
Most of these struct are unused and should be fixed or dropped.
Walter Ji [Thu, 22 Feb 2024 09:06:53 +0000 (17:06 +0800)]
target/mips32: add dsp access support for gdb
Change order of dsp register name array and removed hi0 and lo0
to comply with gdb definition of dsp in mips-dsp.xml, the regs
name array is now mapping corresponding dsp accumulator names
onto `mips32_regs` and `core_regs` instead of mapping to instr
arrays in dsp functions.
feature now requires a place to store cached dsp registers.
Add dsp registers to reg_list for gdb to access them.
Add dsp module enable detection to avoid DSP Disabled exception
while reading dsp accumulators.
Add dsp register reading procedure in `mips32_pracc_read_regs`
and writing procedure in `mips32_pracc_write_regs`.
Change-Id: Iacc335da030ab85989922c81aac7925b3dc17459 Signed-off-by: Walter Ji <walter.ji@oss.cipunited.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8476 Tested-by: jenkins Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com> Reviewed-by: Oleksij Rempel <linux@rempel-privat.de>
projects / users / borneoa / openocd-next.git / log