feat: increase tree leaf log2 size to 5

[mpernambuco]

This branch updates access logs handling and tests according to the emulator changes introduced in this PR:

cartesi/machine-emulator#259

In the emulator, the Merkle tree word size is increased from 8 to 32 bytes.

The access log format has changed as follows:

• The value field is now 32 bytes long and contains the Merkle tree leaf where the accessed word is located.
• To locate the relevant 64-bit word inside data, we compute the offset of address relative to the leaf-aligned address.
• The sibling_hashes field now has 59 entries.

[GCdePaula]

Hey @mpernambuco, let me know if I can help in any way!

Current implementation has 64-bit words as the unit requested by step, and also the unit that we read from our access log. With the change in this PR, it seems there's a break: step requests words, but internally our memory model has larger blocks.

My suggestion is that we make better distinctions between "words" (64-bit) and "blocks"/"leaves" (currently 256-bit), and that we extend the Memory module to handle "words" vs "blocks" (and move some of the address manipulation there). I can help here if you want!

[diegonehab]

Hey @mpernambuco, let me know if I can help in any way!

Current implementation has 64-bit words as the unit requested by step, and also the unit that we read from our access log. With the change in this PR, it seems there's a break: step requests words, but internally our memory model has larger blocks.

My suggestion is that we make better distinctions between "words" (64-bit) and "blocks"/"leaves" (currently 256-bit), and that we extend the Memory module to handle "words" vs "blocks" (and move some of the address manipulation there). I can help here if you want!

I don't understand the point you are trying to make. We are changing everything to consistently deal with 256-bit words and forget we ever dealt with 64-bit words. The machine emulator is changing, so the log will change, and the memory module will change, and the solidity step will change.

[GCdePaula]

@diegonehab

I don't understand the point you are trying to make.

It's not a point, it's a Solidity code structure suggestion and an offer for help. I didn't want to barge in.

The current Solidity code is such that words are requested from step, words are read from the log, and words are the leafs of the Merkle tree. We're changing some of that. My suggestion is that there's a nicer way to structure this that makes the code clearer.

As an example, there's a function called strideFromWordAddress, which reading the PR appears to be changed to read a 256-bit memory block. Perhaps it's clearer if we renamed it to strideFromBlockAddress or strideFromLeafAddress. This is minor.

A change that perhaps is more important is understanding the meaning of Region to be the size of a leaf (which used to be words, so no difference before). Then AlignedSize becomes a BlockAlignedSize or LeafAlignedSize (which used to be words, so no difference before).

This way, the function strideFromPhysicalAddress should be updated such that the address alignment verification takes into consideration the correct size of the leaf (which used to be words, so no difference before).

The way I'd do it is start by changing what AlignedSize means, to match the leaf. Probably even change the name to indicate that. We can make this "configurable" (uint8 constant LOG2_LEAF = 5;). Then I guess it's just dealing with the cascading changes.

[mpernambuco]

Hey @mpernambuco, let me know if I can help in any way!

Current implementation has 64-bit words as the unit requested by step, and also the unit that we read from our access log. With the change in this PR, it seems there's a break: step requests words, but internally our memory model has larger blocks.

My suggestion is that we make better distinctions between "words" (64-bit) and "blocks"/"leaves" (currently 256-bit), and that we extend the Memory module to handle "words" vs "blocks" (and move some of the address manipulation there). I can help here if you want!

Hey @GCdePaula,
Yes, thank you! I will need your help to improve these changes.
This is still a draft.
I wanted to ensure that the new access logs format can be handled by the Solidity code.

[stephenctw]

This is looking great! I only have a few minor comments.

[GCdePaula]

I think there's an exploit. Let me know what you think.

[GCdePaula]

I've opened #70 (a quick draft) of what I had in mind for reading. Let me know what you think.

@diegonehab @mpernambuco @stephenctw

[stephenctw]

LGTM! @GCdePaula is there anything else you'd like to add/change?

[GCdePaula]

🚀🚀🚀