- Using PMA, the TEE IRAM is marked as R/X while TEE DRAM is marked as R/W.
Moving the internal memory secure service call table from DRAM to IRAM
makes it immutable.
PMP entry 3 (SOC_DROM_MASK_HIGH, TOR mode) in the memprot path
was incorrectly granted RW permission on esp32h21 and esp32c61.
The mask ROM data region is inherently read-only; remove the W bit.
Also added necessary tests to check voilations and re-enabled
tests for ESP32P4
Commit 64704886b7 (refactor(driver_spi): remove some SPI_LL_xxx SOC_xxx macros)
removed #include "soc/spi_periph.h" from hal/spi_hal.h. Out-of-tree code that
included hal/spi_hal.h and used spi_periph_signal then failed with the symbol
undeclared, because that header had been the transitive source of the declaration.
Partially reverts it.
Add a new CMake function esp_partition_flash_binary() that provides a
unified API for registering partition data binaries to be flashed. It
replaces the direct esptool_py_flash_target calls scattered across
components (spiffs, fatfs, nvs_flash) with a single function that:
- Resolves partition offset from the partition table automatically
- Determines encryption requirements (auto-detect or ALWAYS_PLAINTEXT)
- Creates per-partition flash targets (e.g. idf.py <partition>-flash)
- Optionally includes the binary in `idf.py flash` via FLASH_IN_PROJECT
On the linux target, the function registers binaries for pre-loading
into the emulated flash. A build-time manifest (linux_flash_data.txt)
is generated via file(GENERATE), and partition_linux.c reads it at
runtime to copy each binary into the memory-mapped flash buffer at
the correct offset.
The partition_ops example is updated to use the new function and
includes a custom_partition with pre-built data to demonstrate the
full workflow, including on the linux target.
Reorder s_i2c_hw_fsm_reset() so i2c_ll_reset_register and interrupt
masking run before s_i2c_master_clear_bus(). Avoids ISR firing when
i2c_common_set_pins reconnects GPIOs to a stuck FSM (IDFGH-17497).
Closes https://github.com/espressif/esp-idf/issues/18438
Made-with: Cursor
- Set 8.8.8.8 as primary DNS when DHCP provides no DNS server
- Set 8.8.4.4 as backup DNS in both cases
- Avoids intermittent failures in ethernet download test on CI runners
where the DHCP server does not supply a valid DNS address
The condition filtering attributes lacked parentheses, causing the
'not startswith(_)' guard to only apply to str attributes. Python 3.14's
new __firstlineno__ (int) class attribute leaked into the output.
Also includes ruff auto-formatting fixes (imports, trailing commas,
union type annotations).
Rewrite build_lfn_short_entry_name() and add _gen_numname_suffix() helper
to match the gen_numname() algorithm in ff.c. This fixes:
- chr(order) producing raw binary instead of ASCII digits
- Collision for order >= 10 when str(order) makes the name exceed 8 chars
- Hex suffix with dynamic stem shortening (matching C implementation)
- CRC16-CCITT hash for seq > 5 to reduce collision probability
Also fix LDIR_Name2_SIZE typo in long_filename_utils.py (should be
LDIR_Name3_SIZE), which made the assertion guard too permissive.
Add ShortFilenameGenerationTestCase with 9 unit tests covering single-digit,
multi-digit, hash-based, and collision-free generation scenarios.