How to change the stack size of setup()/loop() in FreeRTOS environment

[masa6161]

I'm a newbie to develop the program with this Arduino-Pico core and FreeRTOS for Raspberry Pi pico 2W board. I'm using the PlatformIO.

What I want to achieve finally

• Using WiFi in FreeRTOS environment (Dual core-processing is better, but single-core now for simplification of the development)
• Time-synchronization from NTP server
• Data sending (JSON format) to Google App Scripts with HTTPS

What is my problem?

The stack size for setup()/loop() might be too small (1024 words)

• Currently, as following the instruction using WiFi with FreeRTOS in the official document, I'm using loop() as a "WiFi manager." Other FreeRTOS tasks are invoked in the setup() function.
• However, the system finally hung-up, and the higher water-mark of loop() acquired with uxTaskGetStackHighWaterMark was about 100.
• Although I'm not sure about concise stack size of loop() when the system hung-up, first I'd like to test the possibility that the stack size is too small to handle the data in my system.
• I assume that this system hung-up might not caused by lacking of heap memory. The free-space by getting with rp2040.getFreeHeap() might be enough (>300 KB).

What I want to know

1. How can I increase the stack size of loop()? (this might be unavailable by my searching, though...)
2. Or, is there an another "best-practice" (e.g., correct implementation) to achieve the problem in this case, or my final objective (data sending via Wifi without system hung-up).

Thank you for your cooperation!

[earlephilhower]

There is no programmatic way of increasing the setup/loop or setup1/loop1 stacks. It's hardcoded to 4K, like the HW stacks, in FreeRTOS variantHooks.cpp.

I would really recommend hooking up a debug probe and watching what happens when it crashes. In GDB you can find the panic cause pretty easily after the fact in many cases, and it may not be related to stack at all.

You can increase that 1024 to something larger or look at moving very large stack objects onto the heap. But I'd get the debugprobe dump first because it's just so nice...

Also, a few notes:

• Unless you manually pin a Task, it can run on any core. So if you have a task started it could run a bit on core0 then a bit on core1. Which is not so good for LWIP or peripherals (IRQs are routed per-core).
• NTP time sync is pretty basic and part of LWIP, no FreeRTOS needed. See the examples
• HTTPS needs two large temporary buffers, allocated at runtime. If you're using a HTTPClient I suggest making sure it's a global and init it once and never delete it to ensure that buffer always exists. OTW the core will need to re-find a couple large blocks which might not exist (i.e. if you have 300KB free but 1 byte allocated every 4KB then there is no way to get those 17KB + 8KB buffers).
• FreeRTOS has lots of limitations on the network stack (and no LeaMDNS, only SimpleMDNS), so if you can do what you're trying to do w/o it then your life will be much simpler. A periodic TLS connect and write is very simple with a Ticker.

[masa6161]

Thank you for your honest and quick reply.

debug probe

Actually, I've already had a debug probe, but I've continued debugging in a legacy/basic way using the serial monitor and Serial.print().
On this occasion, I will try to monitor the program's behavior using the debug probe.

About WiFi-related functions in FreeRTOS must be called only in the setup()/loop(), I'd like to ask an additional question.
The official documentation says that:

FreeRTOS is supported only on core 0 and from within setup and loop, not tasks, due to the requirement for a very different LWIP implementation.

And your comment: #1868 (comment)
I'm wondering that the commonly mentioned here is that "you should call the WiFi-related functions from core0 and the task is pinned to core0." Do you know if my understanding is correct?
If this were correct, can I make a task to execute WiFi-related processing in core0 and change its stack size?
If your answer is No, why should I have used setup()/loop() as a "entry point" of WiFi-related functions?

[earlephilhower]

Yeah, you should probably be able to use a separate task pinned to 0 to do all WiFi/networking. That means all WiFiClients, WiFiServers, WebServers, Wifi.begin, etc.

[masa6161]

That might be good news. I'll try it later and report the result!
At that time, I'm going to test that loop() must shut down after launching the WiFi handling task in core0.