Add explicit assert for minimum block size of 128 bytes #802
Conversation
There was already an assert for this, but because it included the underlying equation for the requirement it was too confusing for users that had no prior knowledge for why the assert could trigger. The math works out such that 128 bytes is a reasonable minimum requirement, so I've added that number as an explicit assert. Hopefully this makes this sort of situation easier to debug. Note that this requirement would need to be increased to 512 bytes if block addresses are ever increased to 64-bits. DESIGN.md goes into more detail why this is.
|
Seems a good idea to make the reason for the assert obvious (although would an explanatory comment suffice?). However, a concern for the future. If you should ever make block addresses 64 bit, I hope this would be an option. The jump from 128 byte blocks to 512 byte blocks could be quite painful for those using small memories (I have just 128K bytes of FRAM, and I think other users have less!). The low overhead of littleFS is currently a key advantage, especially over FAT. |
Well this also makes the rule-of-thumb 128 byte limit a hard requirement, which may be valuable in future changes (though unlikely). It also asserts the math is correct, which is an intimidating piece of code someone might think is broken if they hit this assert. It might reduce the mental load when debugging if you're also calculating your block size a complicated way. Though that may be a bit of a stretch, it could probably go either way. At least if code size is an issue these asserts can always be turned off.
It's just a hypothetical. As is, a 64-bit version. of littlefs would be fundamentally incompatible with the current 32-bit version. And the devices that would need 64-bits addresses would probably be very unlikely to have issues with 512 byte blocks. Another hypothetical would be a 16-bit version of littlefs, which could in theory push the CTZ skip-list requirement down to 32-bytes (24 exactly). But I think at that point you'd really start having issues fitting file metadata into the block. |
See #264 for more info.
There was already an assert for this, but because it included the underlying equation for the requirement it was too confusing for users that had no prior knowledge for why the assert could trigger.
The math works out such that 128 bytes is a reasonable minimum requirement, so I've added that number as an explicit assert. Hopefully this makes this sort of situation easier to debug.
Note that this requirement would need to be increased to 512 bytes if block addresses are ever increased to 64-bits. DESIGN.md#ctz-skip-lists goes into more detail why this is.