Some devices with an internal multi-tap JTAG router require a vendor
specific bypass instruction to bypass the master TAP when addressing
slave taps internal to the same device. On these devices the standard
bypass instruction bypasses the whole device.
Change-Id: I4506f0e67c9e4dfe39b7fa18c63d67900313e594
Signed-off-by: Henrik Nordström <henrik.nordstrom@addiva.se>
Reviewed-on: https://review.openocd.org/c/openocd/+/8041
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
swd_connect_multidrop() sent DORMANT_TO_SWD and called
swd_multidrop_select_inner(). DORMANT_TO_SWD sequence ends
with a LINE_RESET sequence.
swd_multidrop_select_inner() sent LINE_RESET sequence again.
It was useless in this case.
swd_connect_multidrop() emited JTAG_TO_DORMANT and DORMANT_TO_SWD
sequences before connecting each DAP in SWD multidrop bus.
It is sufficient to emit JTAG_TO_DORMANT and DORMANT_TO_SWD
just once and emit the shorter LINE_RESET instead for subsequent DAPs.
Introduce a global variable swd_multidrop_in_swd_state
and use it to control what sequence is emitted.
In case of reconnect after an error, always use the full switch
JTAG_TO_DORMANT and DORMANT_TO_SWD.
Change-Id: Iba21620f6a9680793208bf398960ed0eb59df3b1
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7218
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
SWD multidrop requires some initialization once before connecting
all daps. Provide an optional pre-connect dap operation.
Change-Id: I778215c512c56423a425dda80ab19a739f22f285
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7542
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The original code supported ADIv5 only, just one SELECT register
with some reserved bits - the pseudo value DP_SELECT_INVALID was
just fine to indicate the DP SELECT register is in an unknown state.
Added ADIv6 support required DP SELECT and SELECT1 registers
without reserved bits. Therefore DP_SELECT_INVALID value became
reachable as a (fortunately not really used) ADIv6 AP ADDR.
JTAG DPBANKSEL setting support introduced with ADIv6 does not
honor DP_SELECT_INVALID correctly: required select value
gets compared to DP_SELECT_INVALID value and the most common zero
bank does not trigger DP SELECT write.
DP banked registers need just to set DP SELECT. ADIv6 AP register
addressing scheme may use both DP SELECT and SELECT1. This further
complicates using a single invalid value.
Moreover the difference how the SWD line reset influences
DPBANKSEL field between ADIv5 and ADIv6 deserves better handling
than setting select cache to zero and then to DP_SELECT_INVALID
in a very specific code positions.
Introduce bool flags indicating the validity of each SELECT
register and one SWD specific for DPBANKSEL field.
Use the latter to prevent selecting DP BANK before taking
the connection out of reset by reading DPIDR.
Treat DP SELECT and SELECT1 individually in ADIv6 64-bit mode.
Update comments to reflect the difference between ADIv5 and ADIv6
in SWD line reset.
Change-Id: Ibbb0b06cb592be072571218b666566a13d8dff0e
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-on: https://review.openocd.org/c/openocd/+/7541
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
In loops that typically take longer time to complete, check if there is
a pending shutdown request. If so, terminate the loop.
This allows to respond to a signal requesting a shutdown during some
loops which do not return control to main OpenOCD loop.
Change-Id: Iace0b58eddde1237832d0f9333a7c7b930565674
Signed-off-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8032
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The field backup_working_area is always used as a boolean value.
Use bool type for backup_working_area.
Change-Id: I55c68d717dbbe9e5caf60fd1db368527c6d1b995
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8036
Tested-by: jenkins
Reviewed-by: zapb <dev@zapb.de>
This functionality can be useful for;
1-ESP flashing code to load flasher stub on target and
write/read/erase flash.
2-ESP GCOV command uses some of these functions to run
onboard routines to dump coverage info.
This is high level api for the Espressif xtensa and riscv targets
Signed-off-by: Erhan Kurubas <erhan.kurubas@espressif.com>
Change-Id: I5e618b960bb6566ee618d4ba261f51af97a7cb0e
Reviewed-on: https://review.openocd.org/c/openocd/+/7759
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
When user requested a change on cp0 status register,
it may contain changes on EXL/ERL bits, and changes on
these bits could lead to differnt behaviours on writing
to other cp0 registers.
Change-Id: Ic83039988c29c06ee134226b52de943c46d19da2
Signed-off-by: Walter Ji <walter.ji@oss.cipunited.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7914
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Instead of returning an 'error string', throw an error. This makes it
much easier to handle errors in Tcl scripts or in tools that use Tcl RPC.
Change-Id: I75c48750cfad7430fa5e6bc88fe04ebd59d34cea
Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8006
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The mips opcode macro for the SRL opcode was using the wrong constant
value:
SRA -- Shift right arithmetic
Encoding: 0000 00-- ---t tttt dddd dhhh hh00 0011
SRL -- Shift right logical
Encoding: 0000 00-- ---t tttt dddd dhhh hh00 0010
This corrects the opcode constant for SRL and adds the SRA opcode for
completeness.
There is only one user of MIPS32_OP_SRL in src/flash/nor/cfi.c:
Since the mask constant (0x00000080 for the DQ7 mask) shifted in this
case would never have the sign bit set, it worked fine even though it
was accidentally using the SRA opcode instead of SRL.
Change-Id: I0a80746e2075c7df1ce35b9db00d9d0b997a3feb
Signed-off-by: Tobias Diedrich <ranma+openocd@tdiedrich.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/3613
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Oleksij Rempel <linux@rempel-privat.de>
We were previously inadvertently clearing the top 32 bits of SCTLR_EL1
during read_memory/write_memory as a result of using 32-bit operations
to access the register and because the fields used to temporarily
store the register were 32-bit. Fix it.
Change-Id: I657d7f949e1f7ab6bf90609e3f91cae09cade31a
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7939
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Associativity=0 means that only one CMO is needed for all ways.
It could also mean that the cache is 1-way, but this is less likely.
Currently Associativity=0 causes us to hang in the while loop in
decode_cache_reg. Fix it by skipping the loop in this case. We can let
way_shift be set to the arbitrary value of 0 because in the case where
Associativity=0 we only ever shift 0 by it before ORing it into the
CMO operand.
Change-Id: I7c1de68d33f6b3ed627cbb1e2401d43185e4c1e3
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7916
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
In the unusual (and even incorrect) case of running the command
target create ...
before defining an adapter and the associated transport, the
command causes a segmentation fault. E.g.:
openocd -c 'target create cpu cortex-m -endian little'
Check that get_current_transport() returns a valid pointer before
referencing it.
Change-Id: I9796a7e92196ef3df5c7152b27c34102045dc9e7
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7962
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
Enhance `mips32_read_config_regs` to better detect MIPS32 ISA,
DSP, and FPU features, allowing user to get more detailed
target information.
Most of these information will be used in MIPS m/iAptiv support.
Change-Id: I23571a626ec64fa019acac91bdbfcb434373bfc1
Signed-off-by: Walter Ji <walter.ji@oss.cipunited.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7911
Tested-by: jenkins
Reviewed-by: Oleksij Rempel <linux@rempel-privat.de>
This patch forces a read of register from the target
in the get_reg command in case the register cache
does not hold a valid value at that moment.
Note that the command "reg" already handles it
correctly, no fix is needed there.
Change-Id: I75fad25188e94ee4e06162ab6d600ea24dbf590a
Signed-off-by: Marek Vrbka <marek.vrbka@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7958
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Reviewed-by: zapb <dev@zapb.de>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Previously, when listing the watchpoints, OpenOCD printed
numbers 0, 1 and 2 representing READ, WRITE and ACCESS type
watchpoints.
This patch changes it to 'r', 'w' and 'a'. This increases the
clarity as what type the watchpoint actually is.
Change-Id: I9eac72dfd0bb2a9596a5b0c080a3f584556ed599
Signed-off-by: Marek Vrbka <marek.vrbka@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7909
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
This patch adds the "all" option to the rwp command.
It removes all watchpoints, much like rbp all removes
all breakpoints.
Change-Id: Id58dd103085e558f17afa4a287888cf085566ca9
Signed-off-by: Marek Vrbka <marek.vrbka@codasip.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7907
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Add support for the ARM MCRR/MRRC instructions which require the use of
two registers to transfer a 64-bit co-processor registers. We are going
to use this in a subsequent patch in order to properly dump 64-bit page
table descriptors that exist on ARMv7A with VMSA extensions.
We make use of r0 and r1 to transfer 64-bit quantities to/from DCC.
Change-Id: Ic4975026c1ae4f2853795575ac7701d541248736
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Michael Chalfant <michael.chalfant@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/5228
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Currently, instruction cache is being invalidated in
arc_{un,}set_breakpoint() regardless of whether the breakpoint's type is
HW or SW. For SW breakpoints, this has no net effect as the caches are
flushed as a by-product of overwriting instructions in main memory and
is thus merely unnecessary; but for HW breakpoints this invalidation is
not preceded by a flush and might lead to loss of data. This patch
removes the invalidate() call altogether to correct this undesired
behavior for HW breakpoints.
With this patch applied, all supported HW breakpoint tests from the gdb
testsuite are now passing with the arc-openocd backend.
Change-Id: I3d252b97f01f1a1e2bf0eb8fb257bdab0c544bc2
Signed-off-by: Artemiy Volkov <artemiy@synopsys.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7767
Tested-by: jenkins
Reviewed-by: Evgeniy Didin <didin@synopsys.com>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The comment above armv8_dpm_read_current_registers() doesn't match
the implementation, as the function reads all the registers from
ARMV8_PC and above.
The registers currently read are not relevant to answer to the
usual GDB initial request through the 'g' packet. Plus the lack of
differentiation per core state (AArch32 vs AArch64) causes the
read of not existing registers in AArch32 triggering errors, as
tentatively fixed by https://review.openocd.org/5517/
Fix the code to read the registers initially required by GDB.
Modify the comment to report the register list in AArch32 and in
AArch64.
Keep the extra checks inside the read loop, even if they are
mostly irrelevant; this could prevent errors if someone needs to
extend the number of registers to read.
The current implementation of the register's description in
OpenOCD does not allow to discriminate among AArch32 and AArch64
registers. Add a TODO comment to highlight it.
Change-Id: Icd47d93c19a9e1694a7b51bbc5ca7e21a578df41
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/7887
Tested-by: jenkins
