Configurable option to use IRAM as byte accessible memory (in single core mode) using
load-store (non-word aligned and non-word size IRAM access specific) exception handlers.
This allows to use IRAM for use-cases where certain performance penalty
(upto 170 cpu cycles per load or store operation) is acceptable. Additional configuration
option has been provided to redirect mbedTLS specific in-out content length buffers to
IRAM (in single core mode), allows to save 20KB per TLS connection.
1. Clarify THREADPTR calculation in FreeRTOS code, explaining where
the constant 0x10 offset comes from.
2. On the ESP32-S2, .flash.rodata section had different default
alignment (8 bytes instead of 16), which resulted in different offset
of the TLS sections. Unfortunately I haven’t found a way to query
section alignment from C code, or to use a constant value to define
section alignment in the linker script. The linker scripts are
modified to force a fixed 16 byte alignment for .flash.rodata on the
ESP32 and ESP32-S2beta. Note that the base address of .flash.rodata
was already 16 byte aligned, so this has not changed the actual
memory layout of the application.
Full explanation of the calculation below.
Assume we have the TLS template section base address
(tls_section_vma), the address of a TLS variable in the template
(address), and the final relocation value (offset). The linker
calculates:
offset = address - tls_section_vma + align_up(TCB_SIZE, alignment).
At run time, the TLS section gets copied from _thread_local_start
(in .rodata) to task_thread_local_start. Let’s assume that an address
of a variable in the runtime TLS section is runtime_address.
Access to this address will happen by calculating THREADPTR + offset.
So, by a series of substitutions:
THREADPTR + offset = runtime_address THREADPTR = runtime_address - offset
THREADPTR = runtime_address - (address - tls_section_vma + align_up(TCB_SIZE, alignment)) THREADPTR = (runtime_address - address) + tls_section_vma - align_up(TCB_SIZE, alignment)
The difference between runtime_address and address is same as the
difference between task_thread_local_start and _thread_local_start.
And tls_section_vma is the address of .rodata section, i.e.
_rodata_start. So we arrive to
THREADPTR = task_thread_local_start - _thread_local_start + _rodata_start - align_up(TCB_SIZE, alignment).
The idea with TCB_SIZE being added to the THREADPTR when computing
the relocation was to let the OS save TCB pointer in the TREADPTR
register. The location of the run-time TLS section was assumed to be
immediately after the TCB, aligned to whatever the section alignment
was. However in our case the problem is that the run-time TLS section
is stored not next to the TCB, but at the top of the stack. Plus,
even if it was stored next to the TCB, the size of a FreeRTOS TCB is
not equal to 8 bytes (TCB_SIZE hardcoded in the linker). So we have
to calculate THREADPTR in a slightly obscure way, to compensate for
these differences.
Closes IDF-1239
components/freertos: cleaned up multicore option scheduler.
components/freertos: more cleanup and test optimization to present realistic results
components/freertos: remove unused macros of optimized task selection when multicore is used
freertos/Kconfig: fix trailing space on optimized scheduler option
freertos/tests: moved test context variables inside of test task.
The public variables used on scheduling time test now were packed into a structure allocated on test case task stack and passed to tasks as arguments saving RAM comsumption.
DISABLED_FOR_TARGETS macros are used
Partly revert "ci: disable unavailable tests for esp32s2beta"
This partly reverts commit 76a3a5fb48.
Partly revert "ci: disable UTs for esp32s2beta without runners"
This partly reverts commit eb158e9a22.
Partly revert "fix unit test and examples for s2beta"
This partly reverts commit 9baa7826be.
Partly revert "efuse: Add support for esp32s2beta"
This partly reverts commit db84ba868c.
1. add hal and low-level layer for timer group
2. add callback functions to handle interrupt
3. add timer deinit function
4. add timer spinlock take function
Using xxx_periph.h in whole IDF instead of xxx_reg.h, xxx_struct.h, xxx_channel.h ... .
Cleaned up header files from unnecessary headers (releated to soc/... headers).
This MR removes the common dependency from every IDF components to the SOC component.
Currently, in the ``idf_functions.cmake`` script, we include the header path of SOC component by default for all components.
But for better code organization (or maybe also benifits to the compiling speed), we may remove the dependency to SOC components for most components except the driver and kernel related components.
In CMAKE, we have two kinds of header visibilities (set by include path visibility):
(Assume component A --(depends on)--> B, B is the current component)
1. public (``COMPONENT_ADD_INCLUDEDIRS``): means this path is visible to other depending components (A) (visible to A and B)
2. private (``COMPONENT_PRIV_INCLUDEDIRS``): means this path is only visible to source files inside the component (visible to B only)
and we have two kinds of depending ways:
(Assume component A --(depends on)--> B --(depends on)--> C, B is the current component)
1. public (```COMPONENT_REQUIRES```): means B can access to public include path of C. All other components rely on you (A) will also be available for the public headers. (visible to A, B)
2. private (``COMPONENT_PRIV_REQUIRES``): means B can access to public include path of C, but don't propagate this relation to other components (A). (visible to B)
1. remove the common requirement in ``idf_functions.cmake``, this makes the SOC components invisible to all other components by default.
2. if a component (for example, DRIVER) really needs the dependency to SOC, add a private dependency to SOC for it.
3. some other components that don't really depends on the SOC may still meet some errors saying "can't find header soc/...", this is because it's depended component (DRIVER) incorrectly include the header of SOC in its public headers. Moving all this kind of #include into source files, or private headers
4. Fix the include requirements for some file which miss sufficient #include directives. (Previously they include some headers by the long long long header include link)
This is a breaking change. Previous code may depends on the long include chain.
You may need to include the following headers for some files after this commit:
- soc/soc.h
- soc/soc_memory_layout.h
- driver/gpio.h
- esp_sleep.h
The major broken include chain includes:
1. esp_system.h no longer includes esp_sleep.h. The latter includes driver/gpio.h and driver/touch_pad.h.
2. ets_sys.h no longer includes soc/soc.h
3. freertos/portmacro.h no longer includes soc/soc_memory_layout.h
some peripheral headers no longer includes their hw related headers, e.g. rom/gpio.h no longer includes soc/gpio_pins.h and soc/gpio_reg.h
BREAKING CHANGE
1. separate rom include files and linkscript to esp_rom
2. modefiy "include rom/xxx.h" to "include esp32/rom/xxx.h"
3. Forward compatible
4. update mqtt
New unity component can be used for testing other applications.
Upstream version of Unity is included as a submodule.
Utilities specific to ESP-IDF unit tests (partitions, leak checking
setup/teardown functions, etc) are kept only in unit-test-app.
Kconfig options are added to allow disabling certain Unity features.
This fixes multiple bugs with ring buffers and re-factors the code. The public
API has not changed, however the underlying implementation have various private
functions have been changed. The following behavioral changes have been made
- Size of ring buffers for No-Split/Allow-Split buffers will not be rounded
up to the nearest 32-bit aligned size. This was done to simplify the
implementation
- Item size for No-Split/Allow-Split buffers will also be rounded up to the
nearest 32-bit aligned size.
The following bugs have been fixed
- In copyItemToRingbufAllowSplit(), when copying an item where the aligned
size is smaller than the remaining length, the function does not consider
the case where the true size of the item is less than 4 bytes.
- The copy functions will automatically wrap around the write pointers when
the remaining length of the buffer is not large enough to fit a header, but
does not consider if wrapping around will cause an overlap with the read
pointer. This will make a full buffer be mistaken for an empty buffer
closes#1711
- xRingbufferSend() can get stuck in a infinite loop when the size of the
free memory is larger than the needed_size, but too small to fit in the ring
buffer due to alignment and extra overhead of wrapping around.
closes#1846
- Fixed documentation with ring buffer queue set API
- Adding and removing from queue set does not consider the case where the
read/write semaphores actually hold a value.
The following functions have been deprecated
- xRingbufferIsNextItemWrapped() due to lack of thread safety
- xRingbufferAddToQueueSetWrite() and xRingbufferRemoveFromQueueSetWrite()
as adding the queue sets only work under receive operations.
The following functions have been added
- xRingbufferReceiveSplit() and xRingbufferReceiveSplitFromISR() as a thread
safe way to receive from allow-split buffers
- vRingbufferGetInfo()
Documentation for ring buffers has also been added.
