Missing definition of XTOS_DISABLE_ALL_INTERRUPTS for ESP32 C3

[hallard]

Board

ESP32 C3

Device Description

Lolin ESP32 C3 Mini

Hardware Configuration

No

Version

latest master

IDE Name

PlatformIO

Operating System

Mac OS

Flash frequency

40MHz

PSRAM enabled

yes

Upload speed

921600

Description

When compiling missing some definition of XTOS_DISABLE_ALL_INTERRUPTS or XTOS_RESTORE_INTLEVEL

it's defined for ESP32, ESP32-S2 and ESP32-S3 in xtruntime.h but can't find them for C3

Sketch

void setup() {
  XTOS_DISABLE_ALL_INTERRUPTS;
}

Debug Message

Compiling /Users/charles/piobuild/esp32-c3-mini/src/test.ino.cpp.o
/Users/charles/Library/CloudStorage/OneDrive-Personnel/devt/esp32/_Tests/test_pio/src/test.ino: In function 'void setup()':
/Users/charles/Library/CloudStorage/OneDrive-Personnel/devt/esp32/_Tests/test_pio/src/test.ino:5:3: error: 'XTOS_DISABLE_ALL_INTERRUPTS' was not declared in this scope
   XTOS_DISABLE_ALL_INTERRUPTS;
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/charles/Library/CloudStorage/OneDrive-Personnel/devt/esp32/_Tests/test_pio/src/test.ino:5:3: note: suggested alternative: 'taskDISABLE_INTERRUPTS'
   XTOS_DISABLE_ALL_INTERRUPTS;
   ^~~~~~~~~~~~~~~~~~~~~~~~~~~
   taskDISABLE_INTERRUPTS
*** [/Users/charles/piobuild/esp32-c3-mini/src/test.ino.cpp.o] Error 1
============================================================ [FAILED] Took 1.00 seconds

Other Steps to Reproduce

No response

I have checked existing issues, online documentation and the Troubleshooting Guide

• I confirm I have checked existing issues, online documentation and Troubleshooting guide.

[igrr]

That error makes some sense, xtos* are Xtensa-specific functions, and ESP32-C3 CPU uses RISC-V architecture. I don't think these functions will be provided for C3. Could you please describe the problem you are trying to solve on C3 by using these functions? Perhaps we can suggest another way to solve it.

[hallard]

Yes makes sense that was my 1st thought but macro is defined as follow

#define XTOS_DISABLE_ALL_INTERRUPTS	XTOS_SET_INTLEVEL(15)

and saw somewhere on C3 code some calls to XTOS_SET_INTLEVEL so thought it was existing.

Anyway I'm trying to run excellent radiohead library and error fires on ATOMIC function, a kind of change of IRQ priority I guess, here the latest source for this lib

#elif (RH_PLATFORM == RH_PLATFORM_ESP32)
// jPerotto see hardware/esp32/1.0.4/tools/sdk/include/esp32/xtensa/xruntime.h
 #define ATOMIC_BLOCK_START uint32_t volatile register ilevel = XTOS_DISABLE_ALL_INTERRUPTS;
 #define ATOMIC_BLOCK_END XTOS_RESTORE_INTLEVEL(ilevel);
#else 
 // TO BE DONE:
 #define ATOMIC_BLOCK_START
 #define ATOMIC_BLOCK_END
#endif

Maybe at least define them as nothing for C3 ?

[igrr]

I don't think that defining atomic-related macros to nothing is a good idea. It will give the impression that the code compiles correctly, but it's not actually going to work correctly.

I would suggest making a pull request to RadioHead library, suggesting different functions for ESP32:

#define ATOMIC_BLOCK_START int state = portSET_INTERRUPT_MASK_FROM_ISR();
#define ATOMIC_BLOCK_END portCLEAR_INTERRUPT_MASK_FROM_ISR(state);

This will work on both Xtensa- and RISC-V-based ESP32 chip series.

[hallard]

Awesome thanks,
unfortunately this official lib (and so the latest one) is not on any github like, so PR (also looking on code) is not possible.
Official is here and only way to see is download zip file
https://www.airspayce.com/mikem/arduino/RadioHead/index.html
So I'll try going thru mailing list to ask

[igrr]

@hallard thanks, please report back if there's any issue. I have noticed Adafruit has a fork of this library as well, perhaps they would be open to a PR if it doesn't get accepted upstream.

I would try to avoid defining XTOS_DISABLE_ALL_INTERRUPTS for RISC-V to avoid more confusion. We never know, perhaps someone else does #ifdef XTOS_DISABLE_ALL_INTERRUPTS /* Xtensa specific code follows */...

[hallard]

@PaulStoffregen has also a fork and even I already made a fork 6 years ago here, to get it working without interrupt on ESP target and run on PI (haha long time ago) but the deal is that everyone doing a fork need to maintain by hands on each release comparing each file which is a pain and why so much forks are outdated.
I solved this issue subclassing what I need, like that I can overwrite lib with latest version at no pain, but in this case, it's located on radiohead.h master include file and should be done at source level, but I'm pretty sure Mike will take care of this one.

[PaulStoffregen]

Arduino's official API has only noInterrupts() and interrupts(), but no (officially support API) way to save & restore the state.

Indeed I've ported many libraries over the years, usually ones hard coded for AVR timers. It's not fun, but much less tedious than trying to get people to agree on APIs.

[VojtechBartoska]

Hello @hallard, is this still valid issue?