See https://github.com/riscv-collab/riscv-openocd/issues/1139
riscv013_invalidate_cached_progbuf() was failing to zeroize the final
buffer array element. Use memset() instead of a manual loop to zeroize
it in order to address this and simplify the code.
(1) Error code and 'skip_reason' string were replaced with memory access
status. It allows to specify whether OpenOCD should exit the access
early.
(2) Slightly refactored 'read_memory' and 'write_memory' functions.
Checkpatch-ignore: MACRO_ARG_PRECEDENCE, MULTISTATEMENT_MACRO_USE_DO_WHILE
Checkpatch-ignore: TRAILING_SEMICOLON
Signed-off-by: Farid Khaydari <f.khaydari@syntacore.com>
Fixes#1043
There were multiple issuese with DMI logging:
1. Address was assumed to be the same (#1043).
2. Reported IDLE count was not affected by a reset of the delays.
3. VLA were used.
These issues are addressed in the commit.
Change-Id: I82f45505e8a62dfdd7dcb418784975fe10180109
Signed-off-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
There was a conflict:
1. commit 2cd8ebf44d ("breakpoints: use 64-bit type for watchpoint mask
and value")
2. commit 0bf3373e80 ("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>
This is the fast path for when there is a mismatch in the leading whole
bytes, which means we should return true to indicate not equal like all
the other cases here and in the surrounding functions. Otherwise we'll
incorrectly report _buf1 == _buf2 if and only if there are mismatches in
the leading whole bytes.
This was introduced during the refactor and optimisation referenced
below.
The only in-tree caller of this is jtag_check_value_inner, which will
just fail to catch some errors. However, downstream in riscv-openocd it
gets used in the riscv target to determine whether an IR scan is needed
to select the debug module, and with an IRLEN >= 8 this breaks resetting
if the encoding for the DMI isn't all-ones in its leading whole bytes
(to match BYPASS), since it will believe they are the same and not do an
IR scan, failing (with "At least one TAP shouldn't be in BYPASS mode")
in the subsequent DR scan due to the TAP still being recorded as having
bypass set (and really having an instruction of either BYPASS or
IDCODE).
Fixes: e4ee891759 ("improve buf_cmp and buf_cmp_mask helpers")
Change-Id: Ic4f7ed094429abc4c06a775eb847a8b3ddf2e2d6
Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8489
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Tested-by: jenkins
Two different sizes uint8_t and uint32_t was used for this value
without a good reason.
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Change-Id: I4bb60cc5397ffd0d37e7034e3930e62793140c8d
Reviewed-on: https://review.openocd.org/c/openocd/+/8439
Reviewed-by: Andreas Bolsch <hyphen0break@gmail.com>
Tested-by: jenkins
Could be handy for dummy transfer size detection.
Signed-off-by: Tomas Vanek <vanekt@fbl.cz>
Change-Id: Ibb485218f6c2ff9066910bb58be0fc614b77add3
Reviewed-on: https://review.openocd.org/c/openocd/+/8438
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Reviewed-by: Andreas Bolsch <hyphen0break@gmail.com>
Use the TRAX interface DEBUGPC if available.
Otherwise use default stop-and-go profiling.
ESP32-S3, before this patch:
Internal: 8 samples/second
FT2232H: 12 samples/second
After this patch:
Internal: 18ksamples/second
FT2232H: 100ksamples/second
Change-Id: I681f0bccf4263c1e24f38be511e3b3aec8bf4d60
Signed-off-by: Richard Allen <rsaxvc@rsaxvc.net>
Reviewed-on: https://review.openocd.org/c/openocd/+/8431
Reviewed-by: Erhan Kurubas <erhan.kurubas@espressif.com>
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Ian Thompson <ianst@cadence.com>
Reviewed-by: Yurii Shutkin <yurii.shutkin@gmail.com>
Simplify the code using the target endianness independent API.
Change-Id: I39f720d0db9cf24eb41d7f359e4321bbc2045658
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8474
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Flash banks created in kinetis_create_missing_banks did not populate
bank->minimal_write_gap. The default value of 0 was interpreted as
FLASH_WRITE_CONTINUOUS. This created unnecessary large padding if your
binary had a gap in the populated flash. It also caused flash errors
when loading with GDB because the erroneously padded pages were not
erased first. Tested using an S32k148 using s32k.cfg.
Change-Id: I9b7af698e29ac2c4f5fc8ecd82fa7f4b1a0d43f1
Signed-off-by: daniellizewski <daniellizewski@geotab.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8463
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
When multiple jlink programmers are connected and no specific serial
or USB location is specified, print out the detected serial numbers.
Signed-off-by: Marcus Nilsson <brainbomb@gmail.com>
Change-Id: I280da2b85363f7054c5f466637120427cadcf7d1
Reviewed-on: https://review.openocd.org/c/openocd/+/8356
Reviewed-by: Mark Zhuang <mark.zhuang@spacemit.com>
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Looks like 7f2d3e2925 introduced a regression by incorrectly assigning
threads. The title of the commit message says that the intention was to
"derive threadid from SMP index", this is not what happens, however.
Instead threadid is assigned based on an index of all examined targets
in an SMP group.
This introduces two logical errors.
*Error 1*
Here is the code that assigns threads to harts:
```
foreach_smp_target(head, target->smp_targets) {
struct target *curr = head->target;
if (!target_was_examined(curr))
continue;
threadid_t tid = threads_found + 1;
hwthread_fill_thread(rtos, curr, threads_found, tid);
```
Now, imagine a situation when we have two targets: `target.A` and
`target.B`. Let's assume that `target.A` is NOT examined (it could be
under reset, for example). Then, according to the algorithm when
assigning thread identifiers `target.B` will be assigned tid of 1. The
respected inferior on GDB side will be called `Thread 1`.
Now, imagine that `target.A` activates and succefully examined - OpenOCD
will re-assign thread identifiers. And now on GDB side `Thread 1` will
represent the state of `target.A`. Which is incorrect.
*Error 2*
The reverse mapping between `threadid` and targets does not take the
state of targets into account.
```
static struct target *
hwthread_find_thread(struct target *target, threadid_t thread_id)
...
threadid_t tid = 1;
foreach_smp_target(head, target->smp_targets) {
if (thread_id == tid)
head->target;
++tid;
}
```
So the constructed mapping is incorrect. Since in example above
`Thread 1` will get mapped to `target.A`.
*Solution:*
It seems that threadids should be assigned based on position of the
thread in an smp group disregarding the target state.
Change-Id: Ib93b7ed3bb03696afdf56a105b333e22b9ec69b5
Signed-off-by: Parshintsev Anatoly <anatoly.parshintsev@syntacore.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8471
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Mark Zhuang <mark.zhuang@spacemit.com>
Tested-by: jenkins
Reviewed-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
hide_csrs should not emit warnings on an attempt to hide non-exitents CSR.
hide_csrs funcitonality is intended to be used for scenarios when we don`t
want certain groups of registers to be available in GDB. Typically this is
needed to simplify integration with various IDE. In such scenarious it may
be impractical/unfeseable to figure out which register is present on a
target. So reporting a situation when a user wants to hide a non-existent
register creates way too much noise. This commit reduces severity of
relevant debug message to LOG_TARGET_DEBUG
Change-Id: Icbb982c4bcce7586fe35b6b004d0874d6014d5a7
This reverts commit e56dc61697.
The reverted commit claims to be the same as
b201a5db23, but it's not -- it changes the
warning in `riscv_reg_impl_expose_csrs()` instead of the one in
`riscv_reg_impl_hide_csrs()`.
If no mask is given, the value in the option register is replaced
completely. If a mask is set, only those bits that are set in the mask
are transferred into the option register; the others remain unchanged.
Change-Id: If488a10f92d7dcc0e0f192aef5e67c255fd529c3
Signed-off-by: Ondřej Hošek <ondra.hosek@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8466
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Use mbstowcs() to get required length of wide character string and
include space for terminating null wide character.
Change-Id: I668de6f0acc9b3ec5aca033d870dd9ef354f9077
Signed-off-by: Marcus Nilsson <brainbomb@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8232
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
Integer values are interpreted by TCL as decimal, binary, octal
or hexadecimal if prepended with '0d', '0b', '0o' or '0x'
respectively.
The case of '0' prefix has been interpreted as octal till TCL 8.6
but is interpreted as part of a decimal number by JimTCL and from
TCL 9.
Align str_to_buf() to latest TCL syntax by:
- addding support for '0d', '0b' and '0o' prefix;
- dropping support for '0' prefix.
Change-Id: I708ef72146d75b7bf429df329a0269cf48700a44
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8465
Tested-by: jenkins
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Before the change, if the user wrote to any `tdata*` register, OpenOCD
would sometimes start to disable all the triggers (by writing zeroes to
`tdata1`) and re-enable them again (by witing all trigger registers to the
values read before for each `tselect` value), e.g. on `step`
(see `disable/enable_triggers()`).
There are a couple of issues with such approach:
1. RISC-V Debug Specification does not require custom register types
to support re-enabling by such sequence of writes (e.g. some custom
trigger type may require writing a custom CSR to enable it).
2. OpenOCD may still overwrite these triggers when a user asks to set a
new WP.
This commit introduces `riscv reserve_trigger ...` command to explicitly
mark the triggers OpenOCD should not touch.
Such approach allows to separate management of custom triggers and
offload it onto the user (e.g. disable/enable such triggers by setting up
an event handler on `step`-related events).
Change-Id: I3339000445185ab221368442a070f412bf44bfab
Signed-off-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
Currently, during register file examination:
1. A read of an XPR is attempted via 64-bit abstract access.
2. If such a read fails (e.g. connection unstable) XLEN is assumed to be
32.
3. Then `misa` is read. Since `misa` is a CSR and it may be only
readable via program buffer, `s0` should be readable beforehand (at
least some assumption about `xlen` should be made).
4. Before the commit, the `misa.mxl` field was not checked against
`xlen`, therefore erroneous info may have been reported to the user.
Moreover, the `examine()` would pass indicating no error at all.
5. After the commit, `misa.mxl` is checked against `xlen` value.
Change-Id: I3fe5bd6742e564e6de782aad9ed10e65c0728923
Signed-off-by: Evgeniy Naydanov <evgeniy.naydanov@syntacore.com>
* removed `progbuf_size` field from `riscv_info`; added getter
* moved `impebreak` field from `riscv_info` to `riscv013_info`
as implementation dependent field; added getter
Signed-off-by: Farid Khaydari <f.khaydari@syntacore.com>
Instead of making people type this in all the time, just default to
"SEGGER RTT" so more things work out of the box.
Change-Id: I147142cf0a755e635d3f66e047be2eb5049cf511
Signed-off-by: Karl Palsson <karl.palsson@marel.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8354
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Open output file in binary mode to disable EOL
conversion on Windows (and sometimes cygwin depending
on installation settings and path).
Change-Id: I38276dd1af011ce5781b0264b7cbb08c32a1a2ad
Signed-off-by: Richard Allen <rsaxvc@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8278
Reviewed-by: Karl Palsson <karlp@tweak.au>
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
This is useful for setting a reset catch on a CPU that is being
brought out of reset.
Change-Id: Id8fe9bc3f75fd170f207f470a9f3b0faba7f24c1
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8422
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Reduce the amount of boilerplate by moving cti_regs into its only
user, making it a local variable and removing the now-redundant
p_val pointer.
Change-Id: I778cc1e960532fae1ac1a952c6ff19c54e578a5f
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8421
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Some targets start up with the sticky overrun bit set. On such targets
we need to clear it in order to avoid subsequent incorrect reads.
Change-Id: I3e939a9e092de6fcea9494d3179a3386aa1701d2
Signed-off-by: Peter Collingbourne <pcc@google.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8420
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
Tested-by: jenkins
Reviewed-by: Tomas Vanek <vanekt@fbl.cz>
While at it, fix some small coding style issues.
Change-Id: Ifb8e78b55d29a06d69a3ce71d12d0040777aef13
Signed-off-by: Marc Schink <dev@zapb.de>
Reviewed-on: https://review.openocd.org/c/openocd/+/8423
Tested-by: jenkins
Reviewed-by: Antonio Borneo <borneo.antonio@gmail.com>
The function str_to_buf() can be simplified by writing directly
the intermediate results in the output buffer.
Such simplification improves the readability and also makes
scan-build happy, as it does not trigger anymore the warning:
src/helper/binarybuffer.c:328:8: warning: Use of memory
allocated with size zero [unix.Malloc]
if ((b256_buf[(buf_len / 8)] & mask) != 0x0) {
Change-Id: I1cef9a1ec5ff0e5841ba582610f273e89e7a81da
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8396
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Tested-by: jenkins
The name 'buf_len' is misleading, as it usually refers to the byte
length of a buffer. Here we use it for the length in bits.
Rename it as 'buf_bitsize'.
While there, fix checkpatch error by changing the index type to
'unsigned int'.
Change-Id: I78855ed79a346d996d9c0100d94d14c64a36b228
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8395
Tested-by: jenkins
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
With 'radix' always zero and '_detected_radix' always NULL, drop
the two parameters and simplify str_to_buf().
While there:
- drop some redundant assert(),
- drop the re-check for the base prefix,
- simplify str_strip_number_prefix_if_present() and rename it, as
the prefix MUST be present,
- fix a minor typo,
- update the doxygen description of str_to_buf().
Change-Id: I1abdc8ec0587b23881953d3094101c04d5bb1c58
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8394
Tested-by: jenkins
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Commit 53b94fad58 ("binarybuffer: Fix str_to_buf() parsing
function") introduces the helper command_parse_str_to_buf() to
parse as number a string on TCL command-line.
The parameter 'radix' can specify the base (decimal, octal,
hexadecimal, or auto-detected).
TCL is supposed to use decimal numbers by default, while octal and
hexadecimal numbers must be prefixed respectively with '0' and
'0x' (or '0X').
This would require the helper to always run auto-detection of the
base, thus always set the 'radix' parameter to zero. This makes
the parameter useless.
Keeping the 'radix' parameter can open the door to future abuse of
TCL syntax, E.g. a command can require an octal value without the
mandatory TCL '0' prefix; the octal value cannot be the result of
TCL expression.
To prevent any future abuse of the 'radix' parameter, drop it.
Change-Id: I88855bd83b4e08e8fdcf86a2fa5ef3269dd4ad57
Signed-off-by: Antonio Borneo <borneo.antonio@gmail.com>
Reviewed-on: https://review.openocd.org/c/openocd/+/8393
Tested-by: jenkins
Reviewed-by: Jan Matyas <jan.matyas@codasip.com>
Tommy Murphy
Evgeniy Naydanov
Farid Khaydari
Evgeniy Naydanov
Jessica Clarke
Tomas Vanek
Richard Allen
Antonio Borneo
Parshintsev Anatoly
daniellizewski
Marcus Nilsson
Ondřej Hošek
Tim Newsome
Anatoly Parshintsev
Karl Palsson
Richard Allen
Marc Schink
Peter Collingbourne