After startup is done and app_main is reached, all calls to
heap_caps_add_region_with_caps (adding a new heap to the list of registered
heaps) will not affect the minimum free heap size value.
For heaps created before startup, this won't apply and their size will
end up being taken into account in the calculation of the minimum.
On RISC-V chips with SPIRAM support (esp32c5, esp32c61, esp32h4, esp32p4,
esp32s31), esp_restart_noos() was disabling the cache while the current
stack pointer could still be in external RAM. Any stack access after cache
disable (function call, local variable spill) would then fault with
"Cache disabled but cached memory region accessed".
Xtensa chips (esp32, esp32s2, esp32s3) already had this guard via the
SET_STACK macro. Add the equivalent for RISC-V:
- Add rv_utils_set_sp() to riscv/rv_utils.h (plain "mv sp, %0" with a
memory clobber; no window register management needed on RISC-V)
- In each affected system_internal.c, under
CONFIG_FREERTOS_TASK_CREATE_ALLOW_EXT_MEM, check whether the current
SP is in PSRAM and if so switch it to the top of internal BSS before
any cache disable or writeback operation
- Fix xTaskCreateStaticPinnedToCore() stack size argument in the
spiram_stack test (was passing bytes instead of word count)
- Mark the null-pointer write in func_do_exception() as volatile to
prevent the compiler from optimizing it away
- Extend the [spiram_stack] tests to RISC-V by sharing fibonacci() and
the task/finish helpers across architectures, guarding only the
Xtensa-specific WINDOWBASE/WINDOWSTART prints
esp_mem.c in the builtin target via
target_sources(builtin PRIVATE ...) called from the parent CMakeLists.
This cross-directory source injection causes CMake's Ninja generator on
Windows to produce unstable TARGET_PDB/RSP_FILE paths across
reconfigures, changing the ninja command hash and forcing a re-archive
of libmbed-builtin.a on every cmake run — even when no source changed.
This broke test_rebuild_source_files.
Fix by adding esp_mem.c to the IDF mbedtls component library
(mbedtls_srcs) instead. The final ELF link uses --start-group, so
builtin's platform.o resolves esp_mbedtls_mem_calloc/free from the
component library regardless of archive order. esp_mem.c is IDF-specific
code (heap_caps_calloc, sdkconfig.h) and belongs in the port layer, not
in any submodule target.
Add a condition on the definition of the MALLOC_CAP_EXEC macro to
prevent it from being defined if ESP_SYSTEM_MEMPROT_FEATURE or
ESP_SYSTEM_PMP_IDRAM_SPLIT is enabled, thus throwing a compile time
error when using it.
Closes https://github.com/espressif/esp-idf/issues/14837
fix(memory-utils): Check TCM in esp_ptr_internal and esp_ptr_byte_accessible
Closes IDFGH-15339 and IDFGH-15337
See merge request espressif/esp-idf!39388
This feature keeps track of the per task peak memory usage.
- Update the heap_task_tracking example to make use of the new feature
Cleanup the implementation:
- multi_heap_get_free_size() is never used, remove it.
- Minor update in heap_caps_update_per_task_info_xx() funcitons.
- Update settting on block owner in heap_caps.c to work with the
get peak usage feature.
- Update heap_caps_update_per_task_info_free() to detect when it
is called to delete the memory allocated for a task TCB. Mark
the corresponding task in the statistic list as deleted.
- Add a Kconfig option dependant on HEAP_TASK_TRACKING being enabled
that force the deletion of the statistics related to deleted task
when set to true.
- In task tracking feature, add a current and peak memory usage
to the heap_stat_t structure to keep track of the current and
peak memory usage of the given task across all heaps.
- Fix missing block owner when allocating memory for heaps_array
in heap_caps_init.
- Keep the original implementation of the task tracking
for backward compatibility reasons.
esp_ptr_in_rtc_iram_fast logic should be executed if
SOC_RTC_FAST_MEM_SUPPORTED is set but it should also be executed
if IRAM and DRAM region mapping is the same. Remove the
SOC_RTC_IRAM_LOW != SOC_RTC_DRAM_LOW part of the check.
Update heap component to use the modify function appropriately.
esp_ptr_executable does not consider that the PSRAM and FLASH
memory mapping are not always matching.
Added a "pointer is in PSRAM" specific check in the logic to
fix the issue.
In non blocking mode, the read function is expected
to return weather data is available for reading or not.
In case data are available but the size does not match
the expected size, the function read should return whatever
data is available.
Previously, the function was returning -1 with errno set
to EWOULDBLOCK even if the size of data in the buffer was
less than the requested size. It would only return the
available data if the size in the buffer was greater or equal
to the requested size.
The implementation of cdcacm_read is modified to return the avilable
data from the buffer even is the size is lesser than the requested
size.
when esp_event_handler_unregister_with_internal cannot take
the loop mutex (e.g., when the handler unregisters itself),
create an event with a special base identifier and add it to
the queue of the corresponding loop to postpone the removal
of the handler from the list at a time when the loop mutex can be
successfully taken.
handler_execute function is looking to match the handler only in the
list of loop events but does not look in the base event handler list
nor the id event handler list. So unless the event handler is
registered to be triggered for all event bases and all event ids of
an event loop, its profiling fields (invoked and time) are not updated
when it is called.
This commit updates the search for the matching handler to also look
in base event list and ID event list.
Closes https://github.com/espressif/esp-idf/issues/15041
This commit fixes the issue when trying to allocate memory
with the MALLOC_CAP_EXEC in RTC memory. Prior to the fix,
the heap allocator was returning an address in RTC DRAM.
To fix this issue:
- modified memory_layout.c of the concerned targets to fill the iram_address
field in the rtc entry of the soc_memory_region array properly.
- modified heap component related functions to return IRAM address when
an allocation in RTC memory with MALLOC_CAP_EXEC is requested.
Closes https://github.com/espressif/esp-idf/issues/14835
In esp32c2 and esp32c61 memory_layout.c files, the config used to allow
MALLOC_CAP_EXEC was CONFIG_ESP_SYSTEM_MEMPROT_FEATURE when
CONFIG_ESP_SYSTEM_PMP_IDRAM_SPLIT should be used.
Closes https://github.com/espressif/esp-idf/issues/14836
For a given group of heaps sharing the same capabilities, it is
best to use smaller heaps first when performing "small" allocations
to keep the bigger heaps untouched. This prevents the scenario where
a series of small allocations would crowd the bigger heaps making
any bigger allocation fail for lack of space.
Closes https://github.com/espressif/esp-idf/issues/13588
- Place the section on alloc failed hook at a better place
- Add reference to the different poisoning configs at the beginning
of the heap corruption detectino section
- Update the information concerning heap tracing
- Update the heap tracing log examples
- Fix heap tracing standalone code:
- Update the logging format in heap_trace_dump_base
- Add freed field in trace stucture to keep this info even
when no call stack is available
Create a specific group "psram" to only run meaningful
tests on the sdkconfig.ci.psram and sdkconfig.ci.psram_all_ext
configuration.
Add test testing that external memory is prioritized when using
MALLOC_CAP_DEFAULT.
The new TLSF architecture has changed. tlsf.h public API
header is now moved into an include folder. tlsf_common.h
is removed from the repo.
This commit updates the heap component and respective
esp_rom patches to take into account this new
architecture.
When data was sent through USB Serial JTAG before the
driver was installed, the bus was malfunctioning. This
was because the interrupt bit for data reception was cleared
regardless of whether data was received or not. Consequently,
usb_serial_jtag_isr_handler_default was not triggered and the
data was never read causing the bus to malfunction.
This commit is modifying usb_serial_jtag_driver_install to
prevent clearing USB_SERIAL_JTAG_INTR_SERIAL_OUT_RECV_PKT and
USB_SERIAL_JTAG_INTR_SERIAL_IN_EMPTY thus allowing the callback
usb_serial_jtag_isr_handler_default to trigger for possible data
exchanged prior to the call to usb_serial_jtag_driver_install.
This commit also modified the while logic in linenoiseProbe to
discard any data that doesn't match the expected chaaracter sequences
to prevent random input from interfering with evaluating whether the
terminal supports escape sequences or not.
See https://github.com/espressif/esp-idf/issues/13940
This commit adds a feature to pause the heap tracing in the
sense that in this state, the heap tracing will no longer
record the new allocations but will continue to monitor
the free() in order to keep track of the status of the
allocations present in the list of records.
See https://github.com/espressif/esp-idf/issues/13803
in speed.rst:
- add startup time increase info when spiram test is enabled
- add startup time increase info when spiram is enabled and
poisoning comprehensive is enabled
- add L2 cache variable size info to optimize IRAM space / cache misses
- update sections refencing bluetooth/wifi built-in tasks to not show
related info for p4 targets.
- Add IDF_TARGET_RF_TYPE for esp32c5
in ram-usage.rst:
- add L2 cache variable size info to maximize RAM space
Remove the files from esp32c5.txt and esp32p4.txt
that are no longer in need of update.
Guillaume Souchere