FM-407: Insert 0x00..00 for system account #409
Conversation
|
That's interesting. Doing this wouldn't work in Lotus, since I think we validate that the sender of an Eth message is nothing else but an Eth Account (EOA). It might just work here. You don't foresee any complications down the line in merging these types? You mentioned something about doing this to disable gas accounting. We do want gas accounting to run though, since I think some apps may call eth_estimateGas with the null sender. |
|
I'm not sure but intuitively the masked ID address of the system actor also didn't come with an ethereum account actor attached to it, and it worked in calls. As far as I can see as long as the address can be turned into an ID it should be equivalent. The implicit execution still counts the gas and respects the gas limit, it just doesn't check that the sender has enough balance to cover it up front, so it's one less thing to worry about with calls. If you send a call in your own name, then it's an explicit execution and you need to have the funds, however currently there is no miner penalty for including transactions with 0 gas price, so that's another way to circumvent this. But tools often estimate the price and set it, which sometimes means a lot of stars have to align. |
|
It looks like you are right about there being some validation, the tests failed with: 2023-11-08T01:23:36.415823Z INFO ethers: example transfer height=1088 tx_hash=0xb338e8cd3676f2fa6e49b3315199cf6baee9fc1599df03adc69752ede5e7b939
Error: failed to call eth_call: (code: 1, message: pre-validation failed: Send not from valid sender
, data: Some(String(""))) |
|
|
||
| define_singleton!(SYSTEM { id: 0, code_id: 1 }); | ||
|
|
||
| lazy_static! { | ||
| /// The Ethereum null-address 0x00..00 can also be used to identify the system actor. |
There was a problem hiding this comment.
@raulk, is this also the case in lotus?
There was a problem hiding this comment.
Based on this I think they just create Ethereum account actors with these addresses, they aren't the same as the system address. It is only me who thought since we have this notion of sending these calls as the system actor, and that's actually what Lotus does too, then why not make it official?
| @@ -190,7 +190,7 @@ where | |||
| msg.gas_limit = fvm_shared::BLOCK_GAS_LIMIT; | |||
| } | |||
|
|
|||
| if msg.from == SYSTEM_ACTOR_ADDR { | |||
| if is_system_addr(&msg.from) { | |||
There was a problem hiding this comment.
Maybe a dumb question, but what happens if a user sends an external transaction with the zero address in the from? I guess that it will have no signature and it will be rejected, right? I am just wondering if we are opening the door here to users being able to send free transactions that are executed implicitly by just setting from = 0x00..
There was a problem hiding this comment.
You are right, they would have to sign these transactions as well, which is impossible. It's the same as them setting an FVM transaction from to the f0 address of the system actor.
Fixes consensus-shipyard/ipc#166
Inserts into the
Initactor an association from the "null"f410address0x00..00to theSYSTEM_ACTOR_ID. This way calls that do not set afromwill continue to work with implicit semantics FWIW.