LittleFS is still using malloc()

[mcknly]

Even though buffers are provided to littlefs (see storman_service.c in 94b76e4), it still seems to be calling malloc() and causing FreeRTOS to overflow its heap. Increasing FreeRTOS heap in hardware/[platform]/rtos_config.h is a workaround, but we really shouldn't have LFS or any other task calling malloc.

This can be tested by placing a breakpoint at littlefs/littlefs/lfs_util.h line 249 and calling a flash filesystem function via /mnt/flash0, it will only hit the breakpoint if LFS is using malloc.

[fatdollar]

Just getting back to having some free time and was pulling down the latest to get it to compile. Seems a lot of awesome improvements have been made.

I can't seem to get it to compile though. This is a barebones compile. My Cmake settings:

# PROJECT NAME - in quotes, no spaces
set(PROJ_NAME "ALDA")

# PROJECT VERSION - in quotes, no spaces, can contain alphanumeric if necessary
set(PROJ_VER "0.1")

# CLI INTERFACE - 0: use UART for CLI (default), 1: use USB for CLI
set(CLI_IFACE 1)

# MCU PLATFORM - set the MCU platform being used (i.e. the subdir in 'hardware/')
set(PLATFORM rp2xxx)

# BOARD - set the board being used (platform-specific prebuild.cmake contains more information about boards)
set(BOARD pico_w)

# HOSTNAME - hostname will be shown at CLI prompt, and used for network connections
set(HOSTNAME "ALDA")

# BUILD OPTIONS - individual features which can be enabled or disabled
option(ENABLE_MOTD "Enable Message of the Day print at boot" true)
option(ENABLE_WIFI "Enable WiFi support" true)
option(ENABLE_HTTPD "Enable httpd web server" true)`

And the error I'm getting.

[ 52%] Linking CXX executable ALDA.elf
/usr/bin/../lib/gcc/arm-none-eabi/14.3.1/../../../../arm-none-eabi/bin/ld: ALDA.elf section `.heap' will not fit in region `RAM'
/usr/bin/../lib/gcc/arm-none-eabi/14.3.1/../../../../arm-none-eabi/bin/ld: section .stack1_dummy VMA [20040000,200407ff] overlaps section .heap VMA [2003fd80,2004057f]
/usr/bin/../lib/gcc/arm-none-eabi/14.3.1/../../../../arm-none-eabi/bin/ld: region `RAM' overflowed by 1408 bytes
collect2: error: ld returned 1 exit status
make[2]: *** [CMakeFiles/ALDA.dir/build.make:3820: ALDA.elf] Error 1
make[1]: *** [CMakeFiles/Makefile2:1989: CMakeFiles/ALDA.dir/all] Error 2
make: *** [Makefile:91: all] Error 2

I thought that maybe I was enabling features but even with Wi-Fi support disabled and not using USB CLI I get the same heap error and the 1408 bytes is consistent no matter what I change.

I'm using the latest versions of everything. How much did you increase the heap size? My shots in the dark do not seem to be working.

[fatdollar]

#elif defined(USING_RP2040) && defined(HW_USE_WIFI)
#define RTOS_HEAP_SIZE (184*1024) // RP2040 with wireless stack
#endif

I was able to get it to compile by reducing the heap size by 2 was originally 186*1024

I'm at least able to get it to run at this point I'll see about digging into the malloc/heap issue some more

[mcknly]

Hey @fatdollar good to hear from you. Sorry it took me so long to respond. Yes, as you have found, setting that heap size has thus far been a manual trial/error process to try to maximize RAM available to the RTOS on the different variants, my guess is that something moved slightly in the latest SDK or some linked library and caused it to just barely overflow. We could leave some buffer space by shrinking RTOS heap to protect for this, maybe there is a more elegant way to pre-calculate available space but I don't know it. Anyways, let me know if you have any projects cooking! I've barely touched it since the WiFi stuff earlier this year.