-
Notifications
You must be signed in to change notification settings - Fork 976
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
remove gas validations from CL #2699
Conversation
01fa578
to
5870a98
Compare
5870a98
to
dcea5bf
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Generally agree with removing it. We could put a sanity limit in place, but it might be more trouble for what it's worth (if the block data is not too large, it can be passed to execution engine, and dropped as invalid there).
Thanks for the review! Will leave this up a bit for some more eyes before merging |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree with the arguments. My original intention of bringing on these changes was optimisation for the future stateless world. Apparently this optimisation was premature and now I think that even in a stateless world these checks should be done by EL as they are related to the execution.
The block hash check is very important and we must never remove it. This is what enforces mapping of one block tree on another, i.e. one CL block to one EL block (ok, many-to-one to be strict) instead of many-to-many.
The block number check is enshrined into a block header validity that is validated by EL and is redundant. But it's cheap and it's a kind of axiom that isn't a subject to change. So, we can leave it here or remove it for conformance with proposed change. I don't have a strong opinion in that regard.
The proposal looks good to me. One thing, we should remove the gas_limit
check from p2p-interface as well
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think you also can remove GAS_LIMIT_DENOMINATOR
and MIN_GAS_LIMIT
constants
If we assume that EL does a consistency check on the fact that a blockchain is formed -- which is even implicit if using the We are removing gas parent-to-child consistency checks from CL because EL does it. |
just going to chime in that i agree w/ the changes in this PR and put my vote in to remove as much as possible. to that end, i'd support removing the parent hash check and the block number check (i think the only "coupling" we should aim for will be pulling the execution block's header into the beacon state to facilitate efficient light verification and possibly an affordance to communicate "gas/sec" to the EL, but that is another line of work separate to this PR) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
left comment above. generally approve!
Done @mkalinin
Done @terencechain Thank you! |
We can defer all checks to EL except for this one. If there is no |
okay, I see what you are getting at. We want CL to know the blockchain datastructure is respected so that we don't accidentally trigger validation of a disjoint chain CL could point to a disjoint chain and EL would validate from that point and maybe even blindly accept the transition but it was from a different chain We should leave it in |
73ba6d6
to
879bd2f
Compare
I think that having
is_valid_gas_limit
violates the separation of layers for zero gain. By hoisting this logic into CL while retaining the validations in EL, we now create two places where specs need to be maintained, client logic built and tested, and for updates to be made in the event of changes.Sometimes, cross-layers will be unavoidable, but this seems like a place we can and should avoid
I'd also argue that the following can/should be removed as well but I have less of a strong stance because this logic is incredible simple and unlikely to change. That said, there is minimal value in having them in CL.
Edit: As pointed out by @mkalinin we need to ensure that the EL sub-chain is consistent with the outer CL chain so the
parent_hash
validations are left in