Bug in SystemConfig Holocene Upgrade

[tynes]

The Holocene SystemConfig upgrade simply modified the implementations of the SystemConfig contracts rather than calling initialize on them. This means that there are two functions that currently do not work as expected.

• basefeeScalar()
• blobBasefeeScalar()

This is because these values are their own storage slots rather than functions that read from a single source of truth (the scalar). Without calling initialize or calling setGasConfigEcotone, they will return 0.

We have a few options:

• redo the upgrade and call initialize as part of the upgrade
• do a new task that calls setGasConfigEcotone to set the values equal to what they are currently
• leave them as 0 and update the implementation of the SystemConfig to parse from the scalar value when calling these functions

These functions would implement the inverse of this:

scalar = (uint256(0x01) << 248) | (uint256(_blobbasefeeScalar) << 32) | _basefeeScalar;

So the basefeeScalar function would sload the scalar, and mask out the least significant 32 bits. The blobBasefeeScalar function would mask out the next least most significant 32 bits.

[tynes]

Decision: we are going to redo the upgrade and call initialize as part of the upgrade

The post checks should assert specifically on basefeeScalar() and blobBasefeeScalar()

[geoknee]

@tynes what should we do for chains that have a scalar set to something we can't decode? See here https://sepolia.etherscan.io/address/0x5D63A8Dc2737cE771aa4a6510D063b6Ba2c4f6F2#readProxyContract#F33

[tynes]

Can you give more details? I don't understand why you cant decode 684000. Encoded as bytes32, it is version 0 so it should be fine

[sebastianst]

Oh yeah right, that's a pre-Ecotone encoding, so this is a base fee scalar of 684000 (and implicitly blob base fee scalar of 0) because it's a version 0 pre-Ecotone encoding. See specs for more detail @geoknee on how to parse version-0 scalars.

[geoknee]

That makes sense. I had thought there was a problem because I couldn't get the ecotone-scalar to decode it. I think it is a bit buggy, will send a fix.

[geoknee]

ethereum-optimism/optimism#13472

[geoknee]

Just to point out that initialize will reencode with version 1 of the scalar encoder, meaning that chains with a version 0 scalar will get a non trivial change to the raw data of scalar (even if semantically there is no change).

[sebastianst]

Just to point out that initialize will reencode with version 1 of the scalar encoder, meaning that chains with a version 0 scalar will get a non trivial change to the raw data of scalar (even if semantically there is no change).

I discussed this with George. I believe it should be fine because both encodings lead to the same outcome (if blob base fee is 0). We're not emitting a config change event during initialize(), so this just migrates the contract state into a consistent state.

[tynes]

Just to point out that initialize will reencode with version 1 of the scalar encoder, meaning that chains with a version 0 scalar will get a non trivial change to the raw data of scalar (even if semantically there is no change).

What does non trivial mean? From the pov of bytes32, its a single bit. From the pov of uint256, its a massively different number. Is this what you mean? This is expected behavior.