Fixes a regression from 753a92952: if cb was negative, read_bytes
overflowed, because the type was changed from int to size_t.
Also fixes incorrect timeout calculation: timeout_ms was 200, but
each iteration delayed for 10ms, and reduced timeout_ms by 1. This
made the effective timeout to be 2000ms.
The previous approach was to allocate an array on the stack, and
have the array extend past the stack size. This worked by would
result in SP being moved near the end of the stack. If an interrupt
triggered at that time, interrupt prologue would try to save the
context to the stack, tripping the stack overflow watchpoint.
Replacing this with the approach which doesn't move the SP and simply
writes to decreasing addresses from SP, until stack overflow check
triggers.
riscv32-esp-elf toolchain (used for ESP32-C3) can also be used for
ESP32-S2 RISC-V ULP coprocessor.
This removes the riscv-none-embed-gcc toolchain which was originally
used for the ULP, and updates the docs and CMake files to use
riscv32-esp-elf.
Some flags are cleaned up and workarounds removed from CMake toolchain
file.
This adds support for the retargetable locking implementation in
newlib 3. This feature will be enabled in the future toolchain builds.
With the present version of the toolchain, this code doesn't get used.
When _RETARGETABLE_LOCKING gets enabled, newlib locking implementation
will be modified as follows:
- Legacy ESP-specific _lock_xxx functions are preserved. This is done
because ROM copies of newlib in ESP32 and ESP32-S2 rely on these
functions through the function pointer table. Also there is some
code in IDF which still uses these locking functions.
- New __retarget_lock_xxx functions are introduced. Newlib expects
these functions to be provided by the system. These functions work
pretty much the same way as the ESP-specific _lock_xxx functions,
except one major difference: _lock_acquire receives the lock pointer
by value, and as such doesn't support lazy initialization.
- Static locks used by newlib are now explicitly initialized at
startup. Since it is unlikely that these static locks are used at
the same time, all compatible locks are set to point to the same
mutex. This saves a bit of RAM. Note that there are still many locks
not initialized statically, in particular those inside FILE
structures.
This follows a similar approach as UART core dump handling in
idf_monitor. Panic handler message is detected in the output,
collected into a file, and the file is passed to the decoding script.
In this case, the decoding script acts as a tiny GDB server, so we
can ask GDB to perform the backtrace.
Currently IDF_VERSION_* variables are not available to the component
CMakeLists.txt files at the requirements expansion stage. This makes
it harder to write component CMakeLists files compatible with
different IDF versions.
Include version.cmake from the requirements expansion script, add a
build system test.
Tools: Windows Installer add pre-installation screen with System checks for Python and Windows Defender
Closes IDF-1819
See merge request espressif/esp-idf!11152
Wi-Fi enables and disables ADC when exiting and entering sleep mode.
Coordinate ADC power state with other modules, using adc_power_acquire
and adc_power_release.
Since idf_monitor decodes anything that looks like a code address and
starts with 0x, bootloader logs often get annotated with function
names such as WindowOverflow and other random and scary looking things
unrelated to the issue the user is facing. Print the addresses without
0x to avoid confusion by decoded function names. Print hexadecimal
size with 'h' suffix to distinguish it from the decimal value that
follows.
- line was truncated because 64 characters were not sufficient
- length passed to snprintf should be full buffer length, not -1
- make the width of lock name field fixed
- fix alignment of lock type column
Multiple ssl-related features have been backported from Python 3.x
to Python 2.7.9. This adds a fallback so that idf_tools.py can work
on older versions.
`uart_ll_get_txfifo_len` returns the number of bytes available in the
TX FIFO; The condition we need is "FIFO empty", not "FIFO has free
space". `uart_ll_is_tx_idle` does that, and also ensures that the last
character popped from the TX FIFO has been fully transmitted.
This commit fixes an issue with gdbstub, where it would list threads
with TIDs 1 to N in qfThreadInfo/qsThreadInfo responses, and then
would tell GDB that the current TID is 0 in the qC response. This
caused an assertion failure in GDB, because it couldn't find the
thread structure corresponding to TID 0:
src/gdb/gdb/thread.c:93: internal-error: thread_info* inferior_thread(): Assertion `tp' failed.
The issue was caused by the logic of qfThreadInfo/qsThreadInfo.
If the "paniced" task index was 1, the code would report it in the
response to qfThreadInfo, and then mistakenly skip task with index 0
in qsThreadInfo, due to the use of pre-increment instead of a
post-increment.
With that issue fixed, GDB assertion doesn't happen anymore. However
the code contained a deeper problem, which manifested itself in the
fact that GDB would incorrectly show task index 0 as the current task,
after the above fix.
Previous version of the code assumed that when GDB requests the thread
list, it uses the first thread returned by the target as the "default"
thread, and subsequently shows the user that the program is stopped
in that thread. This assumption was incorrect. In fact, after
connecting to a remote target, GDB obtains information about the
"default" or "current" thread from two sources:
1. the 'thread' special register indicated in the status response
($T00thread;00000001#ee)
2. if the target has only sent the plain stop response ($T00#ee), GDB
would ask for the current thread using a qC packet.
With that in mind, it is not necessary to report the paniced task as
the first task in qfThreadInfo response. We can simply returns the
tasks in their natural order, and then indicate the current task in
the qS packet response.
However even that change does not fully resolve the issues with task
list. The previous version of this code also incorrectly interpreted
the meaning of GDB TIDs -1 and 0. When GDB sends an "Hg0" command
early in the connection process, it doesn't expect the server to set
task 0 as the current task, as the code assumed. Rather, it tells the
server to "set any (arbitrary) task as the current one", and the most
logical thing to do for the server that is already in "stopped" state
is to keep the current task selection.
Since TID 0 has a special meaning in GDB remote protocol, gdbstub code
is now modified to map task indices (which start from 0) to GDB TIDs.
GDB TIDs are arbitrary, and for simplicity we keep the same order and
start counting them from 1.
The summary of all the above changes is:
1. Use "task index + 1" as the TID reported to GDB
2. Report the tasks in natural order; don't complicate the code to
make the paniced task first in the list.
3. Centralize modification of 'current_task_index' and 'regfile'
in the new 'set_active_task' function, to improve encapsulation.
The remaining 4 kB had been reserved for storing RF calibration and
BT stack state since 4e092be6. However, these features never got
implemented. If we ever need to place RF related data into RTC slow
memory, we can do this by creating a variable with RTC_NOINIT_ATTR
instead.
Closes https://github.com/espressif/esp-idf/issues/3993
Ivan Grokhotkov