fs/littlefs: add block_size_factor mount option

[alexcekay]

Summary

When a board switches storage technology but wants to mount a littlefs filesystem onto this storage it is desired to still use the same firmware.

This almost works out of the box in NuttX with the exception of setting the correct block size, which may be different depending on the used storage. An incorrect block size may lead to suboptimal performance or worse.

To avoid writing a firmware variant that only differs by CONFIG_FS_LITTLEFS_BLOCK_SIZE_FACTOR this adds the option to pass the intended block size at the mount call. To still enable the usage of the existing options this adds a parser for comma-separated mount options, allowing multiple options to be passed simultaneously.

Example: "autoformat,block_size_factor=4",
"autoformat,block_size_factor=1"

This is backwards compatible: single options passed without commas continue to work as before.

Impact

This does not have an impact on existing users. The mount options they currently pass still work as before.

Testing

Tested on a v6s flight-controller (STM32H743VIH6) using the following test mount options:

autoformat
forceformat
block_size_factor=1
block_size_factor=4
autoformat,block_size_factor=1
autoformat,block_size_factor=4

Those were passed at the mount call:

if (nx_mount(lfs_dev, "/fs/microsd", "littlefs", 0, <mount_options>) != 0)

Afterwards a breakpoint was set after the parser code and the values of the variables autoformat, forceformat and block_size_factor were checked for correctness. After that it was ensured that the filesystem mounts correctly.

The obvious edge-case is:

block_size_factor=0

This causes a div by zero in the calculation of fs->cfg.block_count. This however was already the case before with CONFIG_FS_LITTLEFS_BLOCK_SIZE_FACTOR, so I did not add an additional verification as this will directly be visible on the first mount try.

Another edge-case is:

autoformat,forceformat

This semantically does not make much sense but does not cause any real problems as the two options can co-exist. In this case forceformat is executed first. If for some reason the mounting would still fail even after that format with EFAULT then autoformat would format again.

[xiaoxiang781216]

move the variable definition to the begin of function for c89

[xiaoxiang781216]

can we use strchrnul to avoid strdup?

[acassis]

@simbit18 @raiden00pl seems like we are out of space inside our docker image:

docker: failed to register layer: write /tools/zap/node_modules/app-builder-bin/linux/arm64/app-builder: no space left on device