Files
serai/coordinator/src/cosign_evaluator.rs

337 lines
10 KiB
Rust
Raw Normal View History

Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
use core::time::Duration;
use std::{
2023-11-15 20:23:19 -05:00
sync::Arc,
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
collections::{HashSet, HashMap},
};
2023-11-15 20:23:19 -05:00
use tokio::{
sync::{mpsc, Mutex, RwLock},
time::sleep,
};
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
use borsh::BorshSerialize;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
use sp_application_crypto::RuntimePublic;
use serai_client::{
primitives::{NETWORKS, NetworkId, Signature},
validator_sets::primitives::{Session, ValidatorSet},
2023-11-15 20:23:19 -05:00
SeraiError, TemporalSerai, Serai,
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
};
2023-11-15 20:23:19 -05:00
use serai_db::{Get, DbTxn, Db, create_db};
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
use processor_messages::coordinator::cosign_block_msg;
use crate::{
p2p::{CosignedBlock, GossipMessageKind, P2p},
substrate::LatestCosignedBlock,
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
};
2023-11-15 20:23:19 -05:00
create_db! {
CosignDb {
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
ReceivedCosign: (set: ValidatorSet, block: [u8; 32]) -> CosignedBlock,
LatestCosign: (network: NetworkId) -> CosignedBlock,
2023-11-18 20:54:37 -05:00
DistinctChain: (set: ValidatorSet) -> (),
2023-11-15 20:23:19 -05:00
}
}
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
pub struct CosignEvaluator<D: Db> {
db: Mutex<D>,
serai: Arc<Serai>,
stakes: RwLock<Option<HashMap<NetworkId, u64>>>,
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
latest_cosigns: RwLock<HashMap<NetworkId, CosignedBlock>>,
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
}
impl<D: Db> CosignEvaluator<D> {
2023-11-15 20:23:19 -05:00
async fn update_latest_cosign(&self) {
let stakes_lock = self.stakes.read().await;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
// If we haven't gotten the stake data yet, return
let Some(stakes) = stakes_lock.as_ref() else { return };
2023-12-16 20:54:24 -05:00
let total_stake = stakes.values().copied().sum::<u64>();
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
2023-11-15 20:23:19 -05:00
let latest_cosigns = self.latest_cosigns.read().await;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let mut highest_block = 0;
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
for cosign in latest_cosigns.values() {
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let mut networks = HashSet::new();
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
for (network, sub_cosign) in &*latest_cosigns {
if sub_cosign.block_number >= cosign.block_number {
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
networks.insert(network);
}
}
let sum_stake =
networks.into_iter().map(|network| stakes.get(network).unwrap_or(&0)).sum::<u64>();
let needed_stake = ((total_stake * 2) / 3) + 1;
if (total_stake == 0) || (sum_stake > needed_stake) {
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
highest_block = highest_block.max(cosign.block_number);
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
}
}
2023-11-15 20:23:19 -05:00
let mut db_lock = self.db.lock().await;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let mut txn = db_lock.txn();
if highest_block > LatestCosignedBlock::latest_cosigned_block(&txn) {
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
log::info!("setting latest cosigned block to {}", highest_block);
LatestCosignedBlock::set(&mut txn, &highest_block);
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
}
txn.commit();
}
async fn update_stakes(&self) -> Result<(), SeraiError> {
let serai = self.serai.as_of_latest_finalized_block().await?;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let mut stakes = HashMap::new();
for network in NETWORKS {
// Use if this network has published a Batch for a short-circuit of if they've ever set a key
let set_key = serai.in_instructions().last_batch_for_network(network).await?.is_some();
if set_key {
stakes.insert(
network,
serai
.validator_sets()
.total_allocated_stake(network)
.await?
.expect("network which published a batch didn't have a stake set")
.0,
);
}
}
// Since we've successfully built stakes, set it
2023-11-15 20:23:19 -05:00
*self.stakes.write().await = Some(stakes);
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
2023-11-15 20:23:19 -05:00
self.update_latest_cosign().await;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
Ok(())
}
// Uses Err to signify a message should be retried
async fn handle_new_cosign(&self, cosign: CosignedBlock) -> Result<(), SeraiError> {
2023-11-15 20:23:19 -05:00
// If we already have this cosign or a newer cosign, return
if let Some(latest) = self.latest_cosigns.read().await.get(&cosign.network) {
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
if latest.block_number >= cosign.block_number {
2023-11-15 20:23:19 -05:00
return Ok(());
}
}
// If this an old cosign (older than a day), drop it
let latest_block = self.serai.latest_finalized_block().await?;
2023-11-15 20:23:19 -05:00
if (cosign.block_number + (24 * 60 * 60 / 6)) < latest_block.number() {
log::debug!("received old cosign supposedly signed by {:?}", cosign.network);
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
return Ok(());
2023-11-15 20:23:19 -05:00
}
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let Some(block) = self.serai.finalized_block_by_number(cosign.block_number).await? else {
2023-11-15 20:23:19 -05:00
log::warn!("received cosign with a block number which doesn't map to a block");
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
return Ok(());
2023-11-15 20:23:19 -05:00
};
async fn set_with_keys_fn(
serai: &TemporalSerai<'_>,
network: NetworkId,
) -> Result<Option<ValidatorSet>, SeraiError> {
let Some(latest_session) = serai.validator_sets().session(network).await? else {
log::warn!("received cosign from {:?}, which doesn't yet have a session", network);
return Ok(None);
};
let prior_session = Session(latest_session.0.saturating_sub(1));
Ok(Some(
if serai
.validator_sets()
.keys(ValidatorSet { network, session: prior_session })
.await?
.is_some()
{
ValidatorSet { network, session: prior_session }
} else {
ValidatorSet { network, session: latest_session }
},
))
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
}
// Get the key for this network as of the prior block
2023-11-15 20:23:19 -05:00
// If we have two chains, this value may be different across chains depending on if one chain
// included the set_keys and one didn't
// Because set_keys will force a cosign, it will force detection of distinct blocks
// re: set_keys using keys prior to set_keys (assumed amenable to all)
let serai = self.serai.as_of(block.header.parent_hash.into());
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
2023-11-15 20:23:19 -05:00
let Some(set_with_keys) = set_with_keys_fn(&serai, cosign.network).await? else {
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
return Ok(());
};
let Some(keys) = serai.validator_sets().keys(set_with_keys).await? else {
log::warn!("received cosign for a block we didn't have keys for");
return Ok(());
};
2023-11-15 20:23:19 -05:00
if !keys
.0
.verify(&cosign_block_msg(cosign.block_number, cosign.block), &Signature(cosign.signature))
{
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
log::warn!("received cosigned block with an invalid signature");
return Ok(());
}
2023-11-15 20:23:19 -05:00
log::info!(
"received cosign for block {} ({}) by {:?}",
block.number(),
hex::encode(cosign.block),
cosign.network
);
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
2023-11-15 20:23:19 -05:00
// Save this cosign to the DB
{
let mut db = self.db.lock().await;
let mut txn = db.txn();
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
ReceivedCosign::set(&mut txn, set_with_keys, cosign.block, &cosign);
LatestCosign::set(&mut txn, set_with_keys.network, &(cosign));
2023-11-15 20:23:19 -05:00
txn.commit();
}
if cosign.block != block.hash() {
log::error!(
"received cosign for a distinct block at {}. we have {}. cosign had {}",
cosign.block_number,
hex::encode(block.hash()),
hex::encode(cosign.block)
);
let serai = self.serai.as_of(latest_block.hash());
let mut db = self.db.lock().await;
// Save this set as being on a different chain
let mut txn = db.txn();
DistinctChain::set(&mut txn, set_with_keys, &());
txn.commit();
let mut total_stake = 0;
let mut total_on_distinct_chain = 0;
for network in NETWORKS {
if network == NetworkId::Serai {
continue;
}
// Get the current set for this network
let set_with_keys = {
let mut res;
while {
res = set_with_keys_fn(&serai, cosign.network).await;
res.is_err()
} {
log::error!(
"couldn't get the set with keys when checking for a distinct chain: {:?}",
res
);
tokio::time::sleep(core::time::Duration::from_secs(3)).await;
}
res.unwrap()
};
// Get its stake
// Doesn't use the stakes inside self to prevent deadlocks re: multi-lock acquisition
if let Some(set_with_keys) = set_with_keys {
let stake = {
let mut res;
while {
res = serai.validator_sets().total_allocated_stake(set_with_keys.network).await;
res.is_err()
} {
log::error!(
"couldn't get total allocated stake when checking for a distinct chain: {:?}",
res
);
tokio::time::sleep(core::time::Duration::from_secs(3)).await;
}
res.unwrap()
};
if let Some(stake) = stake {
total_stake += stake.0;
if DistinctChain::get(&*db, set_with_keys).is_some() {
total_on_distinct_chain += stake.0;
}
}
}
}
// See https://github.com/serai-dex/serai/issues/339 for the reasoning on 17%
if (total_stake * 17 / 100) <= total_on_distinct_chain {
panic!("17% of validator sets (by stake) have co-signed a distinct chain");
}
} else {
2023-11-16 13:32:15 -05:00
{
let mut latest_cosigns = self.latest_cosigns.write().await;
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
latest_cosigns.insert(cosign.network, cosign);
2023-11-16 13:32:15 -05:00
}
2023-11-15 20:23:19 -05:00
self.update_latest_cosign().await;
}
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
Ok(())
}
#[allow(clippy::new_ret_no_self)]
pub fn new<P: P2p>(db: D, p2p: P, serai: Arc<Serai>) -> mpsc::UnboundedSender<CosignedBlock> {
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
let mut latest_cosigns = HashMap::new();
for network in NETWORKS {
if let Some(cosign) = LatestCosign::get(&db, network) {
latest_cosigns.insert(network, cosign);
}
}
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
let evaluator = Arc::new(Self {
db: Mutex::new(db),
serai,
stakes: RwLock::new(None),
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
latest_cosigns: RwLock::new(latest_cosigns),
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
});
// Spawn a task to update stakes regularly
tokio::spawn({
let evaluator = evaluator.clone();
async move {
loop {
// Run this until it passes
while evaluator.update_stakes().await.is_err() {
log::warn!("couldn't update stakes in the cosign evaluator");
// Try again in 10 seconds
sleep(Duration::from_secs(10)).await;
}
// Run it every 10 minutes as we don't need the exact stake data for this to be valid
sleep(Duration::from_secs(10 * 60)).await;
}
}
});
// Spawn a task to receive cosigns and handle them
let (send, mut recv) = mpsc::unbounded_channel();
tokio::spawn({
let evaluator = evaluator.clone();
async move {
while let Some(msg) = recv.recv().await {
while evaluator.handle_new_cosign(msg).await.is_err() {
// Try again in 10 seconds
sleep(Duration::from_secs(10)).await;
}
}
}
});
// Spawn a task to rebroadcast the most recent cosigns
tokio::spawn({
async move {
loop {
2023-12-16 20:54:24 -05:00
let cosigns = evaluator.latest_cosigns.read().await.values().copied().collect::<Vec<_>>();
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
for cosign in cosigns {
Reattempts (#483) * Schedule re-attempts and add a (not filled out) match statement to actually execute them A comment explains the methodology. To copy it here: """ This is because we *always* re-attempt any protocol which had participation. That doesn't mean we *should* re-attempt this protocol. The alternatives were: 1) Note on-chain we completed a protocol, halting re-attempts upon 34%. 2) Vote on-chain to re-attempt a protocol. This schema doesn't have any additional messages upon the success case (whereas alternative #1 does) and doesn't have overhead (as alternative #2 does, sending votes and then preprocesses. This only sends preprocesses). """ Any signing protocol which reaches sufficient participation will be re-attempted until it no longer does. * Have the Substrate scanner track DKG removals/completions for the Tributary code * Don't keep trying to publish a participant removal if we've already set keys * Pad out the re-attempt match a bit more * Have CosignEvaluator reload from the DB * Correctly schedule cosign re-attempts * Actuall spawn new DKG removal attempts * Use u32 for Batch ID in SubstrateSignableId, finish Batch re-attempt routing The batch ID was an opaque [u8; 5] which also included the network, yet that's redundant and unhelpful. * Clarify a pair of TODOs in the coordinator * Remove old TODO * Final comment cleanup * Correct usage of TARGET_BLOCK_TIME in reattempt scheduler It's in ms and I assumed it was in s. * Have coordinator tests drop BatchReattempts which aren't relevant yet may exist * Bug fix and pointless oddity removal We scheduled a re-attempt upon receiving 2/3rds of preprocesses and upon receiving 2/3rds of shares, so any signing protocol could cause two re-attempts (not one more). The coordinator tests randomly generated the Batch ID since it was prior an opaque byte array. While that didn't break the test, it was pointless and did make the already-succeeded check before re-attempting impossible to hit. * Add log statements, correct dead-lock in coordinator tests * Increase pessimistic timeout on recv_message to compensate for tighter best-case timeouts * Further bump timeout by a minute AFAICT, GH failed by just a few seconds. This also is worst-case in a single instance, making it fine to be decently long. * Further further bump timeout due to lack of distinct error
2023-12-12 12:28:53 -05:00
let mut buf = vec![];
cosign.serialize(&mut buf).unwrap();
P2p::broadcast(&p2p, GossipMessageKind::CosignedBlock, buf).await;
Add a cosigning protocol to ensure finalizations are unique (#433) * Add a function to deterministically decide which Serai blocks should be co-signed Has a 5 minute latency between co-signs, also used as the maximal latency before a co-sign is started. * Get all active tributaries we're in at a specific block * Add and route CosignSubstrateBlock, a new provided TX * Split queued cosigns per network * Rename BatchSignId to SubstrateSignId * Add SubstrateSignableId, a meta-type for either Batch or Block, and modularize around it * Handle the CosignSubstrateBlock provided TX * Revert substrate_signer.rs to develop (and patch to still work) Due to SubstrateSigner moving when the prior multisig closes, yet cosigning occurring with the most recent key, a single SubstrateSigner can be reused. We could manage multiple SubstrateSigners, yet considering the much lower specifications for cosigning, I'd rather treat it distinctly. * Route cosigning through the processor * Add note to rename SubstrateSigner post-PR I don't want to do so now in order to preserve the diff's clarity. * Implement cosign evaluation into the coordinator * Get tests to compile * Bug fixes, mark blocks without cosigners available as cosigned * Correct the ID Batch preprocesses are saved under, add log statements * Create a dedicated function to handle cosigns * Correct the flow around Batch verification/queueing Verifying `Batch`s could stall when a `Batch` was signed before its predecessors/before the block it's contained in was cosigned (the latter being inevitable as we can't sign a block containing a signed batch before signing the batch). Now, Batch verification happens on a distinct async task in order to not block the handling of processor messages. This task is the sole caller of verify in order to ensure last_verified_batch isn't unexpectedly mutated. When the processor message handler needs to access it, or needs to queue a Batch, it associates the DB TXN with a lock preventing the other task from doing so. This lock, as currently implemented, is a poor and inefficient design. It should be modified to the pattern used for cosign management. Additionally, a new primitive of a DB-backed channel may be immensely valuable. Fixes a standing potential deadlock and a deadlock introduced with the cosigning protocol. * Working full-stack tests After the last commit, this only required extending a timeout. * Replace "co-sign" with "cosign" to make finding text easier * Update the coordinator tests to support cosigning * Inline prior_batch calculation to prevent panic on rotation Noticed when doing a final review of the branch.
2023-11-15 16:57:21 -05:00
}
sleep(Duration::from_secs(60)).await;
}
}
});
// Return the channel to send cosigns
send
}
}