The fix is for the situation when cache disabling mechanism causes
a deadlock with user tasks. Situation is as follows:
1. spi_flash operation is started from low-priority task on CPU0
2. It uses IPC to wake up high-priority IPC1 task on CPU1, preventing
all other tasks on CPU1 from running. This is needed to safely
disable the cache.
3. While the task which started spi_flash operation is waiting for IPC1
task to acknowledge that CPU1 is not using cache anymore, it is
preempted by a higher priority application task ("app0").
4. Task app0 busy-waits for some operation on CPU1 to complete. But
since application tasks are blocked out by IPC1 task, this never
happens. Since app0 is busy-waiting, the task doing spi flash
operation never runs.
The more or less logical soltion to the problem would be to also do
cache disabling on CPU0 and the SPI flash operation itself from IPC0
task. However IPC0 task stack would need to be increased to allow doing
SPI flash operation (and IPC1 stack as well). This would waste some
memory. An alternative approach adopted in this fix is to call FreeRTOS
functions to temporary increase the priority of SPI flash operation task
to the same level as the IPC task.
Fixes https://github.com/espressif/arduino-esp32/issues/740
Fixes https://github.com/espressif/esp-idf/issues/1157
1. When dual core cpu run access DPORT register, must do protection.
2. If access DPORT register, must use DPORT_REG_READ/DPORT_REG_WRITE and DPORT_XXX register operation macro.
Flash operation complete flag was cleared by the core initiating flash
operation. If the other core was running an ISR, then IPC task could
be late to enter the loop to check s_flash_op_complete by the time next
flash operation started. If the flag is cleared on the CPU waiting on
this flag, then the race condition can not happen.
MR !441 (7c155ab) has fixed issue with esp_intr_noniram_{disable,enable}
calls not being properly protected by spi_flash_op_{lock,unlock}.
Unit test was added, but the unit test environment tests only dual-core
config. Similar issue was present in the code path for the single-core
config, where esp_intr_noniram_{disable,enable} calls were unprotected.
This change fixes the protection issue and updates the unit test to
run properly in single core config as well.
The issue with running unit tests for single core config will be
addressed in a separate MR.
spi_flash_enable_interrupts_caches_and_other_cpu function used to enable
non-IRAM interrupts after giving up flash operation lock, which would
cause problems if another task was waiting on the lock to start a flash
operation. In fact, non-IRAM interrupts should be re-enabled before the
task scheduler is resumed. Otherwise non-pinned task can be moved to the
other CPU due to preemption, causing esp_intr_noniram_enable to be
called on the other CPU, causing an abort to be triggered.
Fixes the issue reported in
https://github.com/espressif/esp-idf/pull/258
1) PS is fixed up to allow GDB backtrace to work properly
2) MR!341 discussion: in core dump module: esp_panicPutXXX was replaced by ets_printf.
3) MR!341 discussion: core dump flash magic number was changed.
4) MR!341 discussion: SPI flash access API was redesigned to allow flexible critical section management.
5) test app for core dump feature was added
6) fixed base64 file reading issues on Windows platform
7) now raw bin core file is deleted upon core loader failure by epscoredump.py
Complimentary changes:
1) Partition table definitions files with core dump partition
2) Special sub-type for core dump partition
3) Special version of spi_flash_xxx
4) espcoredump.py is script to get core dump from flash and print useful info
5) FreeRTOS API was extended to get tasks snapshots
Ivan Grokhotkov