302 Commits

Author SHA1 Message Date
Luke Parker
ce3b90541e Make transactions undroppable
coordinator/cosign/src/delay.rs literally demonstrates how we'd need to rewrite
our handling of transactions with this change. It can be cleaned up a bit but
already identifies ergonomic issues. It also doesn't model passing an &mut txn
to an async function, which would also require using the droppable wrapper
struct.

To locally see this build, run

RUSTFLAGS="-Zpanic_abort_tests -C panic=abort" cargo +nightly build -p serai-cosign --all-targets

To locally see this fail to build, run

cargo build -p serai-cosign --all-targets

While it doesn't say which line causes it fail to build, the only distinction
is panic=unwind.

For more context, please see #578.
2025-01-15 03:56:59 -05:00
Luke Parker
cb410cc4e0 Correct how we handle rounding errors within the penalty fn
We explicitly no longer slash stakes but we still set the maximum slash to the
allocated stake + the rewards. Now, the reward slash is bound to the rewards
and the stake slash is bound to the stake. This prevents an improperly rounded
reward slash from effecting a stake slash.
2025-01-15 02:46:31 -05:00
Luke Parker
6c145a5ec3 Disable offline, disruptive slashes
Reasoning commented in codebase
2025-01-14 11:44:13 -05:00
Luke Parker
a7fef2ba7a Redesign Slash/SlashReport types with a function to calculate the penalty 2025-01-14 07:51:39 -05:00
Luke Parker
291ebf5e24 Have serai-task warnings print with the name of the task 2025-01-14 02:52:26 -05:00
Luke Parker
5e0e91c85d Add tasks to publish data onto Serai 2025-01-14 01:58:26 -05:00
Luke Parker
b5a6b0693e Add a proper error type to ContinuallyRan
This isn't necessary. Because we just log the error, we never match off of it,
we don't need any structure beyond String (or now Debug, which still gives us
a way to print the error). This is for the ergonomics of not having to
constantly write `.map_err(|e| format!("{e:?}"))`.
2025-01-12 18:29:08 -05:00
Luke Parker
3cc2abfedc Add a task to publish slash reports 2025-01-12 17:47:48 -05:00
Luke Parker
0ce9aad9b2 Add flow to add transactions onto Tributaries 2025-01-12 07:32:45 -05:00
Luke Parker
e35aa04afb Start handling messages from the processor
Does route ProcessorMessage::CosignedBlock. Rest are stubbed with TODO.
2025-01-12 06:07:55 -05:00
Luke Parker
e7de5125a2 Have processor-messages use CosignIntent/SignedCosign, not the historic cosign format
Has yet to update the processor accordingly.
2025-01-12 05:52:33 -05:00
Luke Parker
158140c3a7 Add a proper error for intake_cosign 2025-01-12 05:49:17 -05:00
Luke Parker
df9a9adaa8 Remove direct dependencies of void, async-trait 2025-01-12 03:48:43 -05:00
Luke Parker
d854807edd Make message_queue::client::Client::send fallible
Allows tasks to report the errors themselves and handle retry in our
standardized way.
2025-01-11 21:57:58 -05:00
Luke Parker
f501d46d44 Correct disabling of Nagle's algorithm 2025-01-11 06:54:43 -05:00
Luke Parker
74106b025f Publish SlashReport onto the Tributary 2025-01-11 06:51:55 -05:00
Luke Parker
e731b546ab Update documentation 2025-01-11 05:13:43 -05:00
Luke Parker
77d60660d2 Move spawn_cosign from main.rs into tributary.rs
Also refines the tasks within tributary.rs a good bit.
2025-01-11 05:12:56 -05:00
Luke Parker
3c664ff05f Re-arrange coordinator/
coordinator/tributary was tributary-chain. This crate has been renamed
tributary-sdk and moved to coordinator/tributary-sdk.

coordinator/src/tributary was our instantion of a Tributary, the Transaction
type and scan task. This has been moved to coordinator/tributary.

The main reason for this was due to coordinator/main.rs becoming untidy. There
is now a collection of clean, independent APIs present in the codebase.
coordinator/main.rs is to compose them. Sometimes, these compositions are a bit
silly (reading from a channel just to forward the message to a distinct
channel). That's more than fine as the code is still readable and the value
from the cleanliness of the APIs composed far exceeds the nits from having
these odd compositions.

This breaks down a bit as we now define a global database, and have some APIs
interact with multiple other APIs.

coordinator/src/tributary was a self-contained, clean API. The recently added
task present in coordinator/tributary/mod.rs, which bound it to the rest of the
Coordinator, wasn't.

Now, coordinator/src is solely the API compositions, and all self-contained
APIs are their own crates.
2025-01-11 04:14:21 -05:00
Luke Parker
c05b0c9eba Handle Canonical, NewSet from serai-coordinator-substrate 2025-01-11 03:07:15 -05:00
Luke Parker
6d5049cab2 Move the task providing transactions onto the Tributary to the Tributary module
Slims down the main file a bit
2025-01-11 02:13:23 -05:00
Luke Parker
1419ba570a Route from tributary scanner to message-queue 2025-01-11 01:55:36 -05:00
Luke Parker
542bf2170a Provide Cosign/CosignIntent for Tributaries 2025-01-11 01:31:28 -05:00
Luke Parker
378d6b90cf Delete old Tributaries on reboot 2025-01-10 20:10:05 -05:00
Luke Parker
cbe83956aa Flesh out Coordinator main
Lot of TODOs as the APIs are all being routed together.
2025-01-10 02:24:24 -05:00
Luke Parker
091d485fd8 Have the Tributary scanner DB be distinct from the cosign DB
Allows deleting the entire Tributary scanner DB upon retiry.
2025-01-10 02:22:58 -05:00
Luke Parker
2a3eaf4d7e Wrap the entire Libp2p object in an Arc
Makes `Clone` calls significantly cheaper as now only the outer Arc is cloned
(the inner ones have been removed). Also wraps uses of Serai in an Arc as we
shouldn't actually need/want multiple caller connection pools.
2025-01-10 01:26:07 -05:00
Luke Parker
23122712cb Document validator jailing upon participation failures and slash report determination
These are TODOs. I just wanted to ensure this was written down and each seemed
too small for GH issues.
2025-01-09 19:50:39 -05:00
Luke Parker
47eb793ce9 Slash upon Tendermint evidence
Decoding slash evidence requires specifying the instantiated generic
`TendermintNetwork`. While irrelevant, that generic includes a type satisfying
`tributary::P2p`. It was only possible to route now that we've redone the P2P
API.
2025-01-09 06:58:00 -05:00
Luke Parker
9b0b5fd1e2 Have serai-cosign index finalized blocks' numbers 2025-01-09 06:57:26 -05:00
Luke Parker
893a24a1cc Better document bounds in serai-coordinator-p2p 2025-01-09 06:57:12 -05:00
Luke Parker
b101e2211a Complete serai-coordinator-p2p 2025-01-09 06:23:14 -05:00
Luke Parker
201a444e89 Remove tokio dependency from serai-coordinator-p2p
Re-implements tokio::mpsc::oneshot with a thin wrapper around async-channel.

Also replaces futures-util with futures-lite.
2025-01-09 02:16:05 -05:00
Luke Parker
9833911e06 Promote Request::Heartbeat from an enum variant to a struct 2025-01-09 01:41:42 -05:00
Luke Parker
465e8498c4 Make the coordinator's P2P modules their own crates 2025-01-09 01:26:25 -05:00
Luke Parker
adf20773ac Add libp2p module documentation 2025-01-09 00:40:07 -05:00
Luke Parker
295c1bd044 Document improper handling of session rotation in P2P allow list 2025-01-09 00:16:45 -05:00
Luke Parker
dda6e3e899 Limit each peer to one connection
Prevents dialing the same peer multiple times (successfully).
2025-01-09 00:06:51 -05:00
Luke Parker
75a00f2a1a Add allow_block_list to libp2p
The check in validators prevented connections from non-validators.
Non-validators could still participate in the network if they laundered their
connection through a malicious validator. allow_block_list ensures that peers,
not connections, are explicitly limited to validators.
2025-01-08 23:54:27 -05:00
Luke Parker
6cde2bb6ef Correct and document topic subscription 2025-01-08 23:16:04 -05:00
Luke Parker
20326bba73 Replace KeepAlive with ping
This is more standard and allows measuring latency.
2025-01-08 23:01:36 -05:00
Luke Parker
ce83b41712 Finish mapping Libp2p to the P2p trait API 2025-01-08 19:39:09 -05:00
Luke Parker
b2bd5d3a44 Remove Debug bound on tributary::P2p 2025-01-08 17:40:32 -05:00
Luke Parker
de2d6568a4 Actually implement the Peer abstraction for Libp2p 2025-01-08 17:40:08 -05:00
Luke Parker
fd9b464b35 Add a trait for the P2p network used in the coordinator
Moves all of the Libp2p code to a dedicated directory. Makes the Heartbeat task
abstract over any P2p network.
2025-01-08 17:01:37 -05:00
Luke Parker
376a66b000 Remove async-trait from tendermint-machine, tributary-chain 2025-01-08 16:41:11 -05:00
Luke Parker
2121a9b131 Spawn the task to select validators to dial 2025-01-07 18:17:36 -05:00
Luke Parker
419223c54e Build the swarm
Moves UpdateSharedValidatorsTask to validators.rs. While prior planned to
re-use a validators object across connecting and peer state management, the
current plan is to use an independent validators object for each to minimize
any contention. They should be built infrequently enough, and cheap enough to
update in the majority case (due to quickly checking if an update is needed),
that this is fine.
2025-01-07 18:09:25 -05:00
Luke Parker
a731c0005d Finish routing our own channel abstraction around the Swarm event stream 2025-01-07 16:51:56 -05:00
Luke Parker
f27e4e3202 Move the WIP SwarmTask to its own file 2025-01-07 16:34:19 -05:00
Luke Parker
f55165e016 Add channels to send requests/recv responses 2025-01-07 15:51:15 -05:00
Luke Parker
d9e9887d34 Run the dial task whenever we have a peer disconnect 2025-01-07 15:36:42 -05:00
Luke Parker
82e753db30 Document risk of eclipse in the dial task 2025-01-07 15:35:34 -05:00
Luke Parker
052388285b Remove TaskHandle::close
TaskHandle::close meant run_now may panic if the task was closed. Now, tasks
are only closed when all handles are dropped, causing all handles to point to
running tasks (ensuring run_now won't panic).
2025-01-07 15:26:41 -05:00
Luke Parker
47a4e534ef Update serai-processor-signers to VariantSignid::Batch([u8; 32]) 2025-01-07 15:26:23 -05:00
Luke Parker
257f691277 Start filling out message handling in SwarmTask 2025-01-05 01:23:28 -05:00
Luke Parker
c6d0fb477c Inline noise into OnlyValidators
libp2p does support (noise, OnlyValidators) but it'll interpret it as either,
not a chain. This will act as the desired chain.
2025-01-05 00:55:25 -05:00
Luke Parker
96518500b1 Don't hold the shared Validators write lock while making requests to Serai 2025-01-05 00:29:11 -05:00
Luke Parker
2b8f481364 Parallelize requests within Validators::update 2025-01-05 00:17:05 -05:00
Luke Parker
479ca0410a Add commentary on the use of FuturesOrdered 2025-01-04 23:28:54 -05:00
Luke Parker
9a5a661d04 Start on the task to manage the swarm 2025-01-04 23:28:29 -05:00
Luke Parker
3daeea09e6 Only let active Serai validators connect over P2P 2025-01-04 22:21:23 -05:00
Luke Parker
a64e2004ab Dial new peers when we don't have the target amount 2025-01-04 18:04:24 -05:00
Luke Parker
f9f6d40695 Use Serai validator keys as PeerIds 2025-01-04 18:03:37 -05:00
Luke Parker
4836c1676b Don't consider the Serai set in the cosigning protocol
The Serai set SHOULD be banned from setting keys so this SHOULD be unreachable.
It's now explicitly unreachable.
2025-01-04 13:52:17 -05:00
Luke Parker
985261574c Add gossip behavior back to the coordinator 2025-01-03 14:00:20 -05:00
Luke Parker
3f3b0255f8 Tweak heartbeat task to run less often if there's no progress to be made 2025-01-03 13:59:14 -05:00
Luke Parker
5fc8500f8d Add task to heartbeat a tributary to the P2P code 2025-01-03 13:04:27 -05:00
Luke Parker
49c221cca2 Restore request-response code to the coordinator 2025-01-03 13:02:50 -05:00
Luke Parker
906e2fb669 Start cosigning on Cosign or Cosigned, not just on Cosigned 2025-01-03 10:30:39 -05:00
Luke Parker
ce676efb1f cargo update 2025-01-03 07:01:06 -05:00
Luke Parker
0a611cb155 Further flesh out tributary scanning
Renames `label` to `round` since `Label` was renamed to `SigningProtocolRound`.

Adds some more context-less validation to transactions which used to be done
within the custom decode function which was simplified via the usage of borsh.

Documents in processor-messages where the Coordinator sends each of its
messages.
2025-01-03 06:57:28 -05:00
Luke Parker
bcd3f14f4f Start work on cleaning up the coordinator's tributary handling 2025-01-02 09:11:04 -05:00
Luke Parker
6272c40561 Restore block_hash to Batch
It's not only helpful (to easily check where Serai's view of the external
network is) but it's necessary in case of a non-trivial chain fork to determine
which blockchain Serai considers canonical.
2024-12-31 18:10:47 -05:00
Luke Parker
2240a50a0c Rebroadcast cosigns for the currently evaluated session, not the latest intended
If Substrate has a block 500 with a key gen, and a block 600 with a key gen,
and the session starting on 500 never cosigns everything, everyone up-to-date
will want the cosigns for the session starting on block 500. Everyone
up-to-date will also be rebroadcasting the non-existent cosigns for the session
which has yet to start. This wouldn't cause a stall as eventually, each
individual set would cosign the latest notable block, and then that would be
explicitly synced, but it's still not the intended behavior.

We also won't even intake the cosigns for the latest intended session if it
exceeds the session we're currently evaluating. This does mean those behind on
the cosigning protocol wouldn't have rebroadcasted their historical cosigns,
and now will, but that's valuable as we don't actually know if we're behind or
up-to-date (per above posited issue).
2024-12-31 17:17:12 -05:00
Luke Parker
7e2b31e5da Clean the transaction definitions in the coordinator
Moves to borsh for serialization. No longer includes nonces anywhere in the TX.
2024-12-31 12:14:32 -05:00
Luke Parker
8c9441a1a5 Redo coordinator's Substrate scanner 2024-12-31 10:37:19 -05:00
Luke Parker
5a42f66dc2 alloy 0.9 2024-12-30 11:09:09 -05:00
Luke Parker
b584a2beab Remove old DB entry from the scanner
We read from it but never writ to it.

It was used to check we didn't flag a block as notable after reporting it, but
it was called by the scan task as it scanned a block. We only batch/report
blocks after the scan task after scanning them, so it was very redundant.
2024-12-30 11:07:05 -05:00
Luke Parker
26ccff25a1 Split reporting Batches to the signer from the Batch test 2024-12-30 11:03:52 -05:00
Luke Parker
f0094b3c7c Rename Report task to Batch task 2024-12-30 10:49:35 -05:00
Luke Parker
458f4fe170 Move where we check if we should delay reporting of Batches 2024-12-30 10:18:38 -05:00
Luke Parker
1de8136739 Remove Session from VariantSignId::SlashReport
It's only there to make the VariantSignid unique across Sessions. By localizing
the VariantSignid to a Session, we avoid this, and can better ensure we don't
queue work for historic sessions.
2024-12-30 06:16:03 -05:00
Luke Parker
445c49f030 Have the scanner's report task ensure handovers only occur if Batchs are valid
This is incomplete at this time. The logic is fine, but needs to be moved to a
distinct location to handle singular blocks which produce multiple Batches.
2024-12-30 06:11:47 -05:00
Luke Parker
5b74fc8ac1 Merge ExternalKeyForSessionToSignBatch into InfoForBatch 2024-12-30 05:34:13 -05:00
Luke Parker
e67e301fc2 Have the processor verify the published Batches match expectations 2024-12-30 05:21:26 -05:00
Luke Parker
1d50792eed Document serai-db with bounds and intent 2024-12-26 02:35:32 -05:00
Luke Parker
9c92709e62 Delay cosign acknowledgments 2024-12-26 01:04:20 -05:00
Luke Parker
3d15710a43 Only check the cosign is after its start block if faulty
We don't have consensus on the session's last block, so we shouldn't check if
the cosign is before the session ends. What matters is that network, within its
set, claims it's still active at that block (on its view of the blockchain).
2024-12-26 00:26:48 -05:00
Luke Parker
df06da5552 Only check if the cosign is stale if it isn't faulty
If it is faulty, we want to archive it regardless.
2024-12-26 00:24:48 -05:00
Luke Parker
cef5bc95b0 Revert prior commit
An archive of all GlobalSessions is necessary to check for faults. The storage
cost is also minimal. While it should be avoided if it can be, it can't be
here.
2024-12-26 00:15:49 -05:00
Luke Parker
f336ab1ece Remove GlobalSessions DB entry
If we read the currently-being-evaluated session from the evaluator, we can
avoid paying the storage costs on all sessions ad-infinitum.
2024-12-25 23:57:51 -05:00
Luke Parker
2aebfb21af Remove serai from the cosign evaluator 2024-12-25 23:47:21 -05:00
Luke Parker
56af6c44eb Remove usage of serai from intake_cosign 2024-12-25 21:19:04 -05:00
Luke Parker
4b34be05bf rocksdb 0.23 2024-12-25 19:48:48 -05:00
Luke Parker
5b337c3ce8 Prevent a malicious validator set from overwriting a notable cosign
Also prevents panics from an invalid Serai node (removing the assumption of an
honest Serai node).
2024-12-25 02:11:05 -05:00
Luke Parker
e119fb4c16 Replace Cosigns by extending NetworksLatestCosignedBlock
Cosigns was an archive of every single cosign ever received. By scoping
NetworksLatestCosignedBlock to be by the global session, we have the latest
cosign for each network in a session (valid to replace all prior cosigns by
that network within that session, even for the purposes of fault) and
automatically have the notable cosigns indexed (as they are the latest ones
within their session). This not only saves space yet also allows optimizing
evaluation a bit.
2024-12-25 01:45:37 -05:00
Luke Parker
ef972b2658 Add cosign signature verification 2024-12-25 00:06:46 -05:00
Luke Parker
4de1a5804d Dedicated library for intending and evaluating cosigns
Not only cleans the existing cosign code but enables non-Serai-coordinators to
evaluate cosigns if they gain access to a feed of them (such as over an RPC).
This would let centralized services not only track the finalized chain yet the
cosigned chain without directly running a coordinator.

Still being wrapped up.
2024-12-22 06:41:55 -05:00
Luke Parker
147a6e43d0 Split task from serai-processor-primitives into serai-task 2024-12-19 10:08:13 -05:00
Luke Parker
066aa9eda4 cargo update
Resolves RUSTSEC-2024-0421
2024-12-12 00:45:19 -05:00
Luke Parker
9593a428e3 alloy 0.8 2024-12-11 01:02:58 -05:00
Luke Parker
5b3c5ec02b Basic Ethereum escapeHatch test 2024-12-09 02:00:17 -05:00
Luke Parker
9ccfa8a9f5 Fix deny 2024-12-08 22:01:43 -05:00
Luke Parker
18897978d0 thiserror 2.0, cargo update 2024-12-08 21:55:37 -05:00
Luke Parker
3192370484 Add Serai key confirmation to prevent rotating to an unusable key
Also updates alloy to the latest version
2024-12-08 20:42:37 -05:00
Luke Parker
8013c56195 Add/correct msrv labels 2024-12-08 18:27:15 -05:00
Luke Parker
834c16930b Add a bitmask of OutInstruction events to Executed
Allows explorers to provide clarity on what occurred.
2024-11-02 21:00:01 -04:00
Luke Parker
2920987173 Add a re-entrancy guard to Router.execute 2024-11-02 20:12:48 -04:00
Luke Parker
26230377b0 Define IRouterWithoutCollisions which Router inherits from
This ensures Router implements most of IRouterWithoutCollisions. It solely
leaves us to confirm Router implements the extensions defined in IRouter.
2024-11-02 19:10:39 -04:00
Luke Parker
2f5c0c68d0 Add selector collisions to Router to make it IRouter compatible 2024-11-02 18:13:02 -04:00
Luke Parker
8de42cc2d4 Add IRouter 2024-11-02 13:19:07 -04:00
Luke Parker
cf4123b0f8 Update how signatures are handled by the Router 2024-11-02 10:47:09 -04:00
Luke Parker
6a520a7412 Work on testing the Router 2024-10-31 02:23:59 -04:00
Luke Parker
b2ec58a445 Update serai-ethereum-processor to compile 2024-10-30 21:48:40 -04:00
Luke Parker
8e800885fb Simplify deterministic signing process in serai-processor-ethereum-primitives
This should be easier to specify/do an alternative implementation of.
2024-10-30 21:36:31 -04:00
Luke Parker
2a427382f1 Natspec, slither Deployer, Router 2024-10-30 21:35:43 -04:00
Luke Parker
ce1689b325 Expand tests for ethereum-schnorr-contract 2024-10-28 18:08:31 -04:00
Luke Parker
0b61a75afc Add lint against string slicing
These are tricky as it panics if the slice doesn't hit a UTF-8 codepoint
boundary.
2024-10-02 21:58:48 -04:00
Luke Parker
2aee21e507 Fix decomposition -> divisor points vartime due to branch prediction/cache rules 2024-09-29 04:19:16 -04:00
Luke Parker
b3e003bd5d cargo +nightly fmt 2024-09-25 10:22:49 -04:00
Luke Parker
251a6e96e8 Constant-time divisors (#617)
* WIP constant-time implementation of the ec-divisors library

* Fix misc logic errors in poly.rs

* Remove accidentally committed test statements

* Fix ConstantTimeEq for CoefficientIndex

* Correct the iterations formula

x**3 / (0 y + x**1) would prior be considered indivisible with iterations = 0.
It is divisible however. The amount of iterations should be the amount of
coefficients within the numerator *excluding the coefficient for y**0 x**0*.

* Poly PartialEq, conditional_select_poly which checks poly structure equivalence

If the first passed argument is smaller than the latter, it's padded to the
necessary length.

Also adds code to trim the remainder as the remainder is the value modulo, so
it's very important it remains concise and workable.

* Fix the line function

It selected the case if both were identity before selecting the case if either
were identity, the latter overwriting the former.

* Final fixes re: ct_get

1) Our quotient structure does need to be of size equal to the numerator
   entirely to prevent out-of-bounds reads on it
2) We need to get from yx_coefficients if of length >=, so if the length is 1
   we can read y_pow=1 from it. If y_pow=0, and its length is 0 so it has no
   inner Vecs, we need to fall back with the guard y_pow != 0.

* Add a trim algorithm to lib.rs to prevent Polys from becoming unbearably gigantic

Our Poly algorithm is incredibly leaky. While it presumably should be improved,
we can take advantage of our known structure while constructing divisors (and
the small modulus) to simply trim out the zero coefficients leaked. This
maintains Polys in a manageable size.

* Move constant-time scalar mul gadget divisor creation from dkg to ec-divisors

Anyone creating a divisor for the scalar mul gadget should use constant time
code, so this code should at least be in the EC gadgets crate It's of
non-trivial complexity to deal with otherwise.

* Remove unsafe, cache timing attacks from ec-divisors
2024-09-24 17:27:05 -04:00
Luke Parker
2c8af04781 machete, drain > mem::swap for clarity reasons 2024-09-19 23:36:32 -07:00
Luke Parker
a0ed043372 Move old processor/src directory to processor/TODO 2024-09-19 23:36:32 -07:00
Luke Parker
2984d2f8cf Misc comments 2024-09-19 23:36:32 -07:00
Luke Parker
554c5778e4 Don't track deployment block in the Router
This technically has a TOCTOU where we sync an Epoch's metadata (signifying we
did sync to that point), then check if the Router was deployed, yet at that
very moment the node resets to genesis. By ensuring the Router is deployed, we
avoid this (and don't need to track the deployment block in-contract).

Also uses a JoinSet to sync the 32 blocks in parallel.
2024-09-19 23:36:32 -07:00
Luke Parker
7e4c59a0a3 Have the Router track its deployment block
Prevents a consensus split where some nodes would drop transfers if their node
didn't think the Router was deployed, and some would handle them.
2024-09-19 23:36:32 -07:00
Luke Parker
294462641e Don't have the ERC20 collapse the top-level transfer ID to the transaction ID
Uses the ID of the transfer event associated with the top-level transfer.
2024-09-19 23:36:32 -07:00
Luke Parker
ae76749513 Transfer ETH with CREATE, not prior to CREATE
Saves a few thousand gas.
2024-09-19 23:36:32 -07:00
Luke Parker
1e1b821d34 Report a Change Output with every Eventuality to ensure we don't fall out of synchrony 2024-09-19 23:36:32 -07:00
Luke Parker
702b4c860c Add dummy fee values to the scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
bc1bbf9951 Set a fixed fee transferred to the caller for publication
Avoids the risk of the gas used by the contract exceeding the gas presumed to
be used (causing an insolvency).
2024-09-19 23:36:32 -07:00
Luke Parker
ec9211fd84 Remove accidentally included bitcoin feature from processor-bin 2024-09-19 23:36:32 -07:00
Luke Parker
4292660eda Have the Ethereum scheduler create Batches as necessary
Also introduces the fee logic, despite it being stubbed.
2024-09-19 23:36:32 -07:00
Luke Parker
8ea5acbacb Update the Router smart contract to pay fees to the caller
The caller is paid a fixed fee per unit of gas spent. That arguably
incentivizes the publisher to raise the gas used by internal calls, yet this
doesn't effect the user UX as they'll have flatly paid the worst-case fee
already. It does pose a risk where callers are arguably incentivized to cause
transaction failures which consume all the gas, not just increased gas, yet:

1) Modern smart contracts don't error by consuming all the gas
2) This is presumably infeasible
3) Even if it was feasible, the gas fees gained presumably exceed the gas fees
   spent causing the failure

The benefit to only paying the callers for the gas used, not the gas alotted,
is it allows Serai to build up a buffer. While this should be minor, a few
cents on every transaction at best, if we ever do have any costs slip through
the cracks, it ideally is sufficient to handle those.
2024-09-19 23:36:32 -07:00
Luke Parker
1b1aa74770 Correct forge fmt config 2024-09-19 23:36:32 -07:00
Luke Parker
861a8352e5 Update to the latest bitcoin-serai 2024-09-19 23:36:32 -07:00
Luke Parker
e64827b6d7 Mark files in TODO/ with "TODO" to ensure it pops up on search 2024-09-19 23:36:32 -07:00
Luke Parker
c27aaf8658 Merge BlockWithAcknowledgedBatch and BatchWithoutAcknowledgeBatch
Offers a simpler API to the coordinator.
2024-09-19 23:36:32 -07:00
Luke Parker
53567e91c8 Read NetworkId from ScannerFeed trait, not env 2024-09-19 23:36:32 -07:00
Luke Parker
1a08d50e16 Remove unused code in the Ethereum processor 2024-09-19 23:36:32 -07:00
Luke Parker
855e53164e Finish Ethereum ScannerFeed 2024-09-19 23:36:32 -07:00
Luke Parker
1367e41510 Add hooks to the main loop
Lets the Ethereum processor track the first key set as soon as it's set.
2024-09-19 23:36:32 -07:00
Luke Parker
a691be21c8 Call tidy_keys upon queue_key
Prevents the potential case of the substrate task and the scan task writing to
the same storage slot at once.
2024-09-19 23:36:32 -07:00
Luke Parker
673cf8fd47 Pass the latest active key to the Block's scan function
Effectively necessary for networks on which we utilize account abstraction in
order to know what key to associate the received coins with.
2024-09-19 23:36:32 -07:00
Luke Parker
118d81bc90 Finish the Ethereum TX publishing code 2024-09-19 23:36:32 -07:00
Luke Parker
e75c4ec6ed Explicitly add an unspendable script path to the processor's generated keys 2024-09-19 23:36:32 -07:00
Luke Parker
9e628d217f cargo fmt, move ScannerFeed from String to the RPC error 2024-09-19 23:36:32 -07:00
Luke Parker
a717ae9ea7 Have the TransactionPublisher build a TxLegacy from Transaction 2024-09-19 23:36:32 -07:00
Luke Parker
98c3f75fa2 Move the Ethereum Action machine to its own file 2024-09-19 23:36:32 -07:00
Luke Parker
18178f3764 Add note on the returned top-level transfers being unordered 2024-09-19 23:36:32 -07:00
Luke Parker
bdc3bda04a Remove ethereum-serai/serai-processor-ethereum-contracts
contracts was smashed out of ethereum-serai. Both have now been smashed into
individual crates.

Creates a TODO directory with left-over test code yet to be moved.
2024-09-19 23:36:32 -07:00
Luke Parker
433beac93a Ethereum SignableTransaction, Eventuality 2024-09-19 23:36:32 -07:00
Luke Parker
8f2a9301cf Don't have the router drop transactions which may have top-level transfers
The router will now match the top-level transfer so it isn't used as the
justification for the InInstruction it's handling. This allows the theoretical
case where a top-level transfer occurs (to any entity) and an internal call
performs a transfer to Serai.

Also uses a JoinSet for fetching transactions' top-level transfers in the ERC20
crate. This does add a dependency on tokio yet improves performance, and it's
scoped under serai-processor (which is always presumed to be tokio-based).
While we could instead import futures for join_all,
https://github.com/smol-rs/futures-lite/issues/6 summarizes why that wouldn't
be a good idea. While we could prefer async-executor over tokio's JoinSet,
JoinSet doesn't share the same issues as FuturesUnordered. That means our
question is solely if we want the async-executor executor or the tokio
executor, when we've already established the Serai processor is always presumed
to be tokio-based.
2024-09-19 23:36:32 -07:00
Luke Parker
d21034c349 Add calls to get the messages to sign for the router 2024-09-19 23:36:32 -07:00
Luke Parker
381495618c Trim dead code 2024-09-19 23:36:32 -07:00
Luke Parker
ee0efe7cde Don't have the Deployer store the deployment block
Also updates how re-entrancy is handled to a more efficient and portable
mechanism.
2024-09-19 23:36:32 -07:00
Luke Parker
7feb7aed22 Hash the message before the challenge function in the Schnorr contract
Slightly more efficient.
2024-09-19 23:36:32 -07:00
Luke Parker
cc75a92641 Smash out the router library 2024-09-19 23:36:32 -07:00
Luke Parker
a7d5640642 Smash ERC20 into its own library 2024-09-19 23:36:32 -07:00
Luke Parker
ae61f3d359 forge fmt 2024-09-19 23:36:32 -07:00
Luke Parker
4bcea31c2a Break Ethereum Deployer into crate 2024-09-19 23:36:32 -07:00
Luke Parker
eb9bce6862 Remove OutInstruction's data field
It makes sense for networks which support arbitrary data to do as part of their
address. This reduces the ability to perform DoSs, achieves better performance,
and better uses the type system (as now networks we don't support data on don't
have a data field).

Updates the Ethereum address definition in serai-client accordingly
2024-09-19 23:36:32 -07:00
Luke Parker
39be23d807 Remove artifacts for serai-processor-ethereum-contracts 2024-09-19 23:36:32 -07:00
Luke Parker
3f0f4d520d Remove the Sandbox contract
If instead of intaking calls, we intake code, we can deploy a fresh contract
which makes arbitrary calls *without* attempting to build our abstraction
layer over the concept.

This should have the same gas costs, as we still have one contract deployment.
The new contract only has a constructor, so it should have no actual code and
beat the Sandbox in that regard? We do have to call into ourselves to meter the
gas, yet we already had to call into the deployed Sandbox to achieve that.

Also re-defines the OutInstruction to include tokens, implements
OutInstruction-specified gas amounts, bumps the Solidity version, and other
such misc changes.
2024-09-19 23:36:32 -07:00
Luke Parker
80ca2b780a Add tests for the premise of the Schnorr contract to the Schnorr crate 2024-09-19 23:36:32 -07:00
Luke Parker
0813351f1f OUT_DIR > artifacts 2024-09-19 23:36:32 -07:00
Luke Parker
a38d135059 rust-toolchain 1.81 2024-09-19 23:36:32 -07:00
Luke Parker
67f9f76fdf Remove publish = false 2024-09-19 23:36:32 -07:00
Luke Parker
1c5bc2259e Dedicated crate for the Schnorr contract 2024-09-19 23:36:32 -07:00
Luke Parker
bdf89f5350 Add dedicated crate for building Solidity contracts 2024-09-19 23:36:32 -07:00
Luke Parker
239127aae5 Add crate for the Ethereum contracts 2024-09-19 23:36:32 -07:00
Luke Parker
d9543bee40 Move ethereum-serai under the processor
It isn't generally usable and should be directly integrated at this point.
2024-09-19 23:36:32 -07:00
Luke Parker
8746b54a43 Don't use a different address for DAI in test
anvil will let us deploy to the existing address.
2024-09-19 23:36:32 -07:00
Luke Parker
7761798a78 Outline the Ethereum processor
This was only half-finished to begin with, unfortunately...
2024-09-19 23:36:32 -07:00
Luke Parker
72a18bf8bb Smart Contract Scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
0616085109 Monero Planner
Finishes the Monero processor.
2024-09-19 23:36:32 -07:00
Luke Parker
e23176deeb Change dummy payment ID behavior on 2-output, no change
This reduces the ability to fingerprint from any observer of the blockchain to
just one of the two recipients.
2024-09-19 23:36:32 -07:00
Luke Parker
5551521e58 Tighten documentation on Block::number 2024-09-19 23:36:32 -07:00
Luke Parker
a2d9aeaed7 Stub out Scheduler in the Monero processor 2024-09-19 23:36:32 -07:00
Luke Parker
e1ad897f7e Allow scheduler's creation of transactions to be async and error
I don't love this, but it's the only way to select decoys without using a local
database. While the prior commit added such a databse, the performance of it
presumably wasn't viable, and while TODOs marked the needed improvements, it
was still messy with an immense scope re: any auditing.

The relevant scheduler functions now take `&self` (intentional, as all
mutations should be via the `&mut impl DbTxn` passed). The calls to `&self` are
expected to be completely deterministic (as usual).
2024-09-19 23:36:32 -07:00
Luke Parker
2edc2f3612 Add a database of all Monero outs into the processor
Enables synchronous transaction creation (which requires synchronous decoy
selection).
2024-09-19 23:36:32 -07:00
Luke Parker
e56af7fc51 Monero time_for_block, dust 2024-09-19 23:36:32 -07:00
Luke Parker
947e1067d9 Monero Processor scan, check_for_eventuality_resolutions 2024-09-19 23:36:32 -07:00
Luke Parker
b4e94f3d51 cargo fmt signers/scanner 2024-09-19 23:36:32 -07:00
Luke Parker
1b39138472 Define subaddress indexes to use
(1, 0) is the external address. (2, *) are the internal addresses.
2024-09-19 23:36:32 -07:00
Luke Parker
e78236276a Remove async-trait from processor/
Part of https://github.com/serai-dex/issues/607.
2024-09-19 23:36:32 -07:00
Luke Parker
2c4c33e632 Misc continuances on the Monero processor 2024-09-19 23:36:32 -07:00
Luke Parker
02409c5735 Correct Multisig Rotation to use WINDOW_LENGTH where proper 2024-09-19 23:36:32 -07:00
Luke Parker
f2cf03cedf Monero processor primitives 2024-09-19 23:36:32 -07:00
Luke Parker
0d4c8cf032 Use a local DB channel for sending to the message-queue
The provided message-queue queue functions runs unti it succeeds. This means
sending to the message-queue will no longer potentially block for arbitrary
amount of times as sending messages is just writing them to a DB.
2024-09-19 23:36:32 -07:00
Luke Parker
b6811f9015 serai-processor-bin
Moves the coordinator loop out of serai-bitcoin-processor, completing it.

Fixes a potential race condition in the message-queue regarding multiple
sockets sending messages at once.
2024-09-19 23:36:32 -07:00
Luke Parker
fcd5fb85df Add binary search to find the block to start scanning from 2024-09-19 23:36:32 -07:00
Luke Parker
3ac0265f07 Add section documenting the safety of txindex upon reorganizations 2024-09-19 23:36:32 -07:00
Luke Parker
9b8c8f8231 Misc tidying of serai-db calls 2024-09-19 23:36:32 -07:00
Luke Parker
59fa49f750 Continue filling out main loop
Adds generics to the db_channel macro, fixes the bug where it needed at least
one key.
2024-09-19 23:36:32 -07:00
Luke Parker
723f529659 Note better message structure in messages 2024-09-19 23:36:32 -07:00
Luke Parker
73af09effb Add note to signers on reducing disk IO 2024-09-19 23:36:32 -07:00
Luke Parker
4054e44471 Start on the new processor main loop 2024-09-19 23:36:32 -07:00
Luke Parker
a8159e9070 Bitcoin Key Gen 2024-09-19 23:36:32 -07:00
Luke Parker
b61ba9d1bb Adjust Bitcoin processor layout 2024-09-19 23:36:32 -07:00
Luke Parker
776cbbb9a4 Misc changes in response to prior two commits 2024-09-19 23:36:32 -07:00
Luke Parker
76a3f3ec4b Add an anyone-can-pay output to every Bitcoin transaction
Resolves #284.
2024-09-19 23:36:32 -07:00
Luke Parker
93c7d06684 Implement presumed_origin
Before we yield a block for scanning, we save all of the contained script
public keys. Then, when we want the address credited for creating an output,
we read the script public key of the spent output from the database.

Fixes #559.
2024-09-19 23:36:32 -07:00
Luke Parker
4cb838e248 Bitcoin processor lib.rs -> main.rs 2024-09-19 23:36:32 -07:00
Luke Parker
c988b7cdb0 Bitcoin TransactionPublisher 2024-09-19 23:36:32 -07:00
Luke Parker
017aab2258 Satisfy Scheduler for Bitcoin 2024-09-19 23:36:32 -07:00
Luke Parker
ba3a6f9e91 Bitcoin ScannerFeed 2024-09-19 23:36:32 -07:00
Luke Parker
e36b671f37 Remove bound that WINDOW_LENGTH < CONFIRMATIONS
It's unnecessary and not valuable.
2024-09-19 23:36:32 -07:00
Luke Parker
2d4b775b6e Add bitcoin Block trait impl 2024-09-19 23:36:32 -07:00
Luke Parker
247cc8f0cc Bitcoin Output/Transaction definitions 2024-09-19 23:36:32 -07:00
Luke Parker
0ccf71df1e Remove old signer impls 2024-09-19 23:36:32 -07:00
Luke Parker
8aba71b9c4 Add CosignerTask to signers, completing it 2024-09-19 23:36:32 -07:00
Luke Parker
46c12c0e66 SlashReport signing and signature publication 2024-09-19 23:36:32 -07:00
Luke Parker
3cc7b49492 Strongly type SlashReport, populate cosign/slash report tasks with work 2024-09-19 23:36:32 -07:00
Luke Parker
0078858c1c Tidy messages, publish all Batches to the coordinator
Prior, we published SignedBatches, yet Batches are necessary for auditing
purposes.
2024-09-19 23:36:32 -07:00
Luke Parker
a3cb514400 Have the coordinator task publish Batches 2024-09-19 23:36:32 -07:00
Luke Parker
ed0221d804 Add BatchSignerTask
Uses a wrapper around AlgorithmMachine Schnorrkel to let the message be &[].
2024-09-19 23:36:32 -07:00
Luke Parker
4152bcacb2 Replace scanner's BatchPublisher with a pair of DB channels 2024-09-19 23:36:32 -07:00
Luke Parker
f07ec7bee0 Route the coordinator, fix race conditions in the signers library 2024-09-19 23:36:32 -07:00
Luke Parker
7484eadbbb Expand task management
These extensions are necessary for the signers task management.
2024-09-19 23:36:32 -07:00
Luke Parker
59ff944152 Work on the higher-level signers API 2024-09-19 23:36:32 -07:00
Luke Parker
8f848b1abc Tidy transaction signing task 2024-09-19 23:36:32 -07:00
Luke Parker
100c80be9f Finish transaction signing task with TX rebroadcast code 2024-09-19 23:36:32 -07:00
Luke Parker
a353f9e2da Further work on transaction signing 2024-09-19 23:36:32 -07:00
Luke Parker
b62fc3a1fa Minor work on the transaction signing task 2024-09-19 23:36:32 -07:00
Luke Parker
8380653855 Add empty serai-processor-signers library
This will replace the signers still in the monolithic Processor binary.
2024-09-19 23:36:32 -07:00
Luke Parker
b50b889918 Split processor into bitcoin-processor, ethereum-processor, monero-processor 2024-09-19 23:36:32 -07:00
Luke Parker
d570c1d277 Move additional_key.rs to serai-processor-view-keys
I don't love this. I wanted to simply add this function to `processor/key-gen`,
but then anyone who wants a view key needs to pull in Bulletproofs which is a
mess of code. They'd also be subject to an AGPL licensed library.

This is so small it should be a primitive elsewhere, yet there is no primitives
library eligible. Maybe serai-client since that has the code to make
transactions to Serai (and will have this as a dependency)? Except then the
processor has to import serai-client when this rewrite removed it as a
dependency.
2024-09-19 23:36:32 -07:00
Luke Parker
2da24506a2 Remove vast swaths of legacy code in the processor 2024-09-19 23:36:32 -07:00
Luke Parker
6e9cb74022 Add non-transaction-chaining scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
0c1aec29bb Finish routing output flushing
Completes the transaction-chaining scheduler.
2024-09-19 23:36:32 -07:00
Luke Parker
653ead1e8c Finish the tree logic in the transaction-chaining scheduler
Also completes the DB functions, makes Scheduler never instantiated, and
ensures tree roots have change outputs.
2024-09-19 23:36:32 -07:00
Luke Parker
8ff019265f Near-complete version of the tree algorithm in the transaction-chaining scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
0601d47789 Work on the tree logic in the transaction-chaining scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
ebef38d93b Ensure the transaction-chaining scheduler doesn't accumulate the same output multiple times 2024-09-19 23:36:32 -07:00
Luke Parker
75b4707002 Add input aggregation in the transaction-chaining scheduler
Also handles some other misc in it.
2024-09-19 23:36:32 -07:00
Luke Parker
3c787e005f Fix bug in the scanner regarding forwarded output amounts
We'd report the amount originally received, minus 2x the cost to aggregate,
regardless the amount successfully forwarded. We should've reduced to the
amount successfully forwarded, if it was smaller, in case the cost to
forward exceeded the aggregation cost.
2024-09-19 23:36:32 -07:00
Luke Parker
f11a6b4ff1 Better document the forwarded output flow 2024-09-19 23:36:32 -07:00
Luke Parker
fadc88d2ad Add scheduler-primitives
The main benefit is whatever scheduler is in use, we now have a single API to
receive TXs to sign (which is of value to the TX signer crate we'll inevitably
build).
2024-09-19 23:36:32 -07:00
Luke Parker
c88ebe985e Outline of the transaction-chaining scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
6deb60513c Expand primitives/scanner with niceties needed for the scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
bd277e7032 Add processor/scheduler/utxo/primitives
Includes the necessary signing functions and the fee amortization logic.

Moves transaction-chaining to utxo/transaction-chaining.
2024-09-19 23:36:32 -07:00
Luke Parker
fc765bb9e0 Add crate for the transaction-chaining Scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
13b74195f7 Don't have acknowledge_batch immediately run
`acknowledge_batch` can only be run if we know what the Batch should be. If we
don't know what the Batch should be, we have to block until we do.
Specifically, we need the block number associated with the Batch.

Instead of blocking over the Scanner API, the Scanner API now solely queues
actions. A new task intakes those actions once we can. This ensures we can
intake the entire Substrate chain, even if our daemon for the external network
is stalled at its genesis block.

All of this for the block number alone seems ridiculous. To go from the block
hash in the Batch to the block number without this task, we'd at least need the
index task to be up to date (still requiring blocking or an API returning
ephemeral errors).
2024-09-19 23:36:32 -07:00
Luke Parker
f21838e0d5 Replace acknowledge_block with acknowledge_batch 2024-09-19 23:36:32 -07:00
Luke Parker
76cbe6cf1e Have acknowledge_block take in the results of the InInstructions executed
If any failed, the scanner now creates a Burn for the return.
2024-09-19 23:36:32 -07:00
Luke Parker
5999f5d65a Route the DB w.r.t. forwarded outputs' information 2024-09-19 23:36:32 -07:00
Luke Parker
d429a0bae6 Remove unused ID -> number lookup 2024-09-19 23:36:32 -07:00
Luke Parker
775824f373 Impl ScanData serialization in the DB 2024-09-19 23:36:32 -07:00
Luke Parker
41a74cb513 Check a queued key has never been queued before
Re-queueing should only happen with a malicious supermajority and breaks
indexing by the key.
2024-09-19 23:36:32 -07:00
Luke Parker
e26da1ec34 Have the Eventuality task drop outputs which aren't ours and aren't worth it to aggregate
We could drop these entirely, yet there's some degree of utility to be able to
add coins to Serai in this manner.
2024-09-19 23:36:32 -07:00
Luke Parker
7266e7f7ea Add note on why LifetimeStage is monotonic 2024-09-19 23:36:32 -07:00
Luke Parker
a8b9b7bad3 Add sanity checks we haven't prior reported an InInstruction for/accumulated an output 2024-09-19 23:36:32 -07:00
Luke Parker
2ca7fccb08 Pass the lifetime information to the scheduler
Enables it to decide which keys to use for fulfillment/change.
2024-09-19 23:36:32 -07:00
Luke Parker
4f6d91037e Call flush_key 2024-09-19 23:36:32 -07:00
Luke Parker
8db76ed67c Add key management to the scheduler 2024-09-19 23:36:32 -07:00
Luke Parker
920303e1b4 Add helper to intake Eventualities 2024-09-19 23:36:32 -07:00
Luke Parker
9f4b28e5ae Clarify output-to-self to output-to-Serai
There's only the requirement it's to an active key which is being reported for.
2024-09-19 23:36:32 -07:00
Luke Parker
f9d02d43c2 Route burns through the scanner 2024-09-19 23:36:32 -07:00
Luke Parker
8ac501028d Add API to publish Batches with
This doesn't have to be abstract, we can generate the message and use the
message-queue API, yet this should help with testing.
2024-09-19 23:36:32 -07:00
Luke Parker
612c67c537 Cache the cost to aggregate 2024-09-19 23:36:32 -07:00
Luke Parker
04a971a024 Fill in various DB functions 2024-09-19 23:36:32 -07:00
Luke Parker
738636c238 Have Scanner::new spawn tasks 2024-09-19 23:36:32 -07:00
Luke Parker
65f3f48517 Add ReportDb 2024-09-19 23:36:32 -07:00
Luke Parker
7cc07d64d1 Make report.rs a folder, not a file 2024-09-19 23:36:32 -07:00
Luke Parker
fdfe520f9d Add ScanDb 2024-09-19 23:36:32 -07:00
Luke Parker
77ef25416b Make scan.rs a folder, not a file 2024-09-19 23:36:32 -07:00
Luke Parker
7c1025dbcb Implement key retiry 2024-09-19 23:36:32 -07:00
Luke Parker
a771fbe1c6 Logs, documentation, misc 2024-09-19 23:36:32 -07:00
Luke Parker
9cebdf7c68 Add sorts for safety even upon non-determinism 2024-09-19 23:36:32 -07:00
Luke Parker
75251f04b4 Use a channel for the InInstructions
It's still unclear how we'll handle refunding failed InInstructions at this
time. Presumably, extending the InInstruction channel with the associated
output ID?
2024-09-19 23:36:32 -07:00
Luke Parker
6196642beb Add a DbChannel between scan and eventuality task 2024-09-19 23:36:32 -07:00
Luke Parker
2bddf00222 Don't expose IndexDb throughout the crate 2024-09-19 23:36:32 -07:00
Luke Parker
9ab8ba0215 Add dedicated Eventuality DB and stub missing fns 2024-09-19 23:36:32 -07:00
Luke Parker
33e0c85f34 Make Eventuality a folder, not a file 2024-09-19 23:36:32 -07:00
Luke Parker
1e8f4e6156 Make a dedicated IndexDb 2024-09-19 23:36:32 -07:00
Luke Parker
66f3428051 Make index a folder, not a file 2024-09-19 23:36:32 -07:00
Luke Parker
7e71840822 Add helper methods
Has fetched blocks checked to be the indexed blocks. Has scanned outputs be
sorted, meaning they aren't subject to implicit order/may be non-deterministic
(such as if handled by a threadpool).
2024-09-19 23:36:32 -07:00
Luke Parker
b65dbacd6a Move ContinuallyRan into primitives
I'm unsure where else it'll be used within the processor, yet it's generally
useful and I don't want to make a dedicated crate yet.
2024-09-19 23:36:32 -07:00
Luke Parker
2fcd9530dd Add a callback to accumulate outputs and return the new Eventualities 2024-09-19 23:36:32 -07:00
Luke Parker
379780a3c9 Flesh out eventuality task 2024-09-19 23:36:32 -07:00
Luke Parker
945f31dfc7 Have the scan flag blocks with change/branch/forwarded as notable 2024-09-19 23:36:32 -07:00
Luke Parker
d5d1fc3eea Flesh out report task 2024-09-19 23:36:32 -07:00
Luke Parker
fd12cc0213 Finish scan task 2024-09-19 23:36:32 -07:00
Luke Parker
ce805c8cc8 Correct compilation errors 2024-09-19 23:36:32 -07:00
Luke Parker
bc0cc5a754 Decide flow between scan/eventuality/report
Scan now only handles External outputs, with an associated essay going over
why. Scan directly creates the InInstruction (prior planned to be done in
Report), and Eventuality is declared to end up yielding the outputs.

That will require making the Eventuality flow two-stage. One stage to evaluate
existing Eventualities and yield outputs, and one stage to incorporate new
Eventualities before advancing the scan window.
2024-09-19 23:36:32 -07:00
Luke Parker
f2ee4daf43 Add Eventuality back to processor primitives
Also splits crate into modules.
2024-09-19 23:36:32 -07:00
Luke Parker
4e29678799 Add bounds for the eventuality task 2024-09-19 23:36:32 -07:00
Luke Parker
74d3075dae Document expectations on Eventuality task and correct code determining the block safe to scan/report 2024-09-19 23:36:32 -07:00
Luke Parker
155ad48f4c Handle dust 2024-09-19 23:36:32 -07:00
Luke Parker
951872b026 Differentiate BlockHeader from Block 2024-09-19 23:36:32 -07:00
Luke Parker
2b47feafed Correct misc compilation errors 2024-09-19 23:36:32 -07:00
Luke Parker
a2717d73f0 Flesh out new scanner a bit more
Adds the task to mark blocks safe to scan, and outlines the task to report
blocks.
2024-09-19 23:36:32 -07:00
Luke Parker
8763ef23ed Definition and delineation of tasks within the scanner
Also defines primitives for the processor.
2024-09-19 23:36:32 -07:00
Luke Parker
57a0ba966b Extend serai-db with support for generic keys/values 2024-09-19 23:36:32 -07:00
Luke Parker
e843b4a2a0 Move scanner.rs to scanner/lib.rs 2024-09-19 23:36:32 -07:00
Luke Parker
2f3bd7a02a Cleanup DB handling a bit in key-gen/attempt-manager 2024-09-19 23:36:32 -07:00
Luke Parker
1e8a9ec5bd Smash out the signer
Abstract, to be done for the transactions, the batches, the cosigns, the slash
reports, everything. It has a minimal API itself, intending to be as clear as
possible.
2024-09-19 23:36:32 -07:00
Luke Parker
2f29c91d30 Smash key-gen out of processor
Resolves some bad assumptions made regarding keys being unique or not.
2024-09-19 23:36:32 -07:00
Luke Parker
f3b91bd44f Smash key-gen into independent crate 2024-09-19 23:36:32 -07:00
Luke Parker
e4e4245ee3 One Round DKG (#589)
* Upstream GBP, divisor, circuit abstraction, and EC gadgets from FCMP++

* Initial eVRF implementation

Not quite done yet. It needs to communicate the resulting points and proofs to
extract them from the Pedersen Commitments in order to return those, and then
be tested.

* Add the openings of the PCs to the eVRF as necessary

* Add implementation of secq256k1

* Make DKG Encryption a bit more flexible

No longer requires the use of an EncryptionKeyMessage, and allows pre-defined
keys for encryption.

* Make NUM_BITS an argument for the field macro

* Have the eVRF take a Zeroizing private key

* Initial eVRF-based DKG

* Add embedwards25519 curve

* Inline the eVRF into the DKG library

Due to how we're handling share encryption, we'd either need two circuits or to
dedicate this circuit to the DKG. The latter makes sense at this time.

* Add documentation to the eVRF-based DKG

* Add paragraph claiming robustness

* Update to the new eVRF proof

* Finish routing the eVRF functionality

Still needs errors and serialization, along with a few other TODOs.

* Add initial eVRF DKG test

* Improve eVRF DKG

Updates how we calculcate verification shares, improves performance when
extracting multiple sets of keys, and adds more to the test for it.

* Start using a proper error for the eVRF DKG

* Resolve various TODOs

Supports recovering multiple key shares from the eVRF DKG.

Inlines two loops to save 2**16 iterations.

Adds support for creating a constant time representation of scalars < NUM_BITS.

* Ban zero ECDH keys, document non-zero requirements

* Implement eVRF traits, all the way up to the DKG, for secp256k1/ed25519

* Add Ristretto eVRF trait impls

* Support participating multiple times in the eVRF DKG

* Only participate once per key, not once per key share

* Rewrite processor key-gen around the eVRF DKG

Still a WIP.

* Finish routing the new key gen in the processor

Doesn't touch the tests, coordinator, nor Substrate yet.
`cargo +nightly fmt && cargo +nightly-2024-07-01 clippy --all-features -p serai-processor`
does pass.

* Deduplicate and better document in processor key_gen

* Update serai-processor tests to the new key gen

* Correct amount of yx coefficients, get processor key gen test to pass

* Add embedded elliptic curve keys to Substrate

* Update processor key gen tests to the eVRF DKG

* Have set_keys take signature_participants, not removed_participants

Now no one is removed from the DKG. Only `t` people publish the key however.

Uses a BitVec for an efficient encoding of the participants.

* Update the coordinator binary for the new DKG

This does not yet update any tests.

* Add sensible Debug to key_gen::[Processor, Coordinator]Message

* Have the DKG explicitly declare how to interpolate its shares

Removes the hack for MuSig where we multiply keys by the inverse of their
lagrange interpolation factor.

* Replace Interpolation::None with Interpolation::Constant

Allows the MuSig DKG to keep the secret share as the original private key,
enabling deriving FROST nonces consistently regardless of the MuSig context.

* Get coordinator tests to pass

* Update spec to the new DKG

* Get clippy to pass across the repo

* cargo machete

* Add an extra sleep to ensure expected ordering of `Participation`s

* Update orchestration

* Remove bad panic in coordinator

It expected ConfirmationShare to be n-of-n, not t-of-n.

* Improve documentation on  functions

* Update TX size limit

We now no longer have to support the ridiculous case of having 49 DKG
participations within a 101-of-150 DKG. It does remain quite high due to
needing to _sign_ so many times. It'd may be optimal for parties with multiple
key shares to independently send their preprocesses/shares (despite the
overhead that'll cause with signatures and the transaction structure).

* Correct error in the Processor spec document

* Update a few comments in the validator-sets pallet

* Send/Recv Participation one at a time

Sending all, then attempting to receive all in an expected order, wasn't working
even with notable delays between sending messages. This points to the mempool
not working as expected...

* Correct ThresholdKeys serialization in modular-frost test

* Updating existing TX size limit test for the new DKG parameters

* Increase time allowed for the DKG on the GH CI

* Correct construction of signature_participants in serai-client tests

Fault identified by akil.

* Further contextualize DkgConfirmer by ValidatorSet

Caught by a safety check we wouldn't reuse preprocesses across messages. That
raises the question of we were prior reusing preprocesses (reusing keys)?
Except that'd have caused a variety of signing failures (suggesting we had some
staggered timing avoiding it in practice but yes, this was possible in theory).

* Add necessary calls to set_embedded_elliptic_curve_key in coordinator set rotation tests

* Correct shimmed setting of a secq256k1 key

* cargo fmt

* Don't use `[0; 32]` for the embedded keys in the coordinator rotation test

The key_gen function expects the random values already decided.

* Big-endian secq256k1 scalars

Also restores the prior, safer, Encryption::register function.
2024-09-19 21:43:26 -04:00
542 changed files with 37971 additions and 23229 deletions

View File

@@ -37,4 +37,4 @@ runs:
- name: Bitcoin Regtest Daemon
shell: bash
run: PATH=$PATH:/usr/bin ./orchestration/dev/networks/bitcoin/run.sh -daemon
run: PATH=$PATH:/usr/bin ./orchestration/dev/networks/bitcoin/run.sh -txindex -daemon

View File

@@ -42,8 +42,8 @@ runs:
shell: bash
run: |
cargo install svm-rs
svm install 0.8.25
svm use 0.8.25
svm install 0.8.26
svm use 0.8.26
# - name: Cache Rust
# uses: Swatinem/rust-cache@a95ba195448af2da9b00fb742d14ffaaf3c21f43

View File

@@ -30,4 +30,5 @@ jobs:
-p patchable-async-sleep \
-p serai-db \
-p serai-env \
-p serai-task \
-p simple-request

View File

@@ -35,6 +35,10 @@ jobs:
-p multiexp \
-p schnorr-signatures \
-p dleq \
-p generalized-bulletproofs \
-p generalized-bulletproofs-circuit-abstraction \
-p ec-divisors \
-p generalized-bulletproofs-ec-gadgets \
-p dkg \
-p modular-frost \
-p frost-schnorrkel

View File

@@ -73,6 +73,15 @@ jobs:
- name: Run rustfmt
run: cargo +${{ steps.nightly.outputs.version }} fmt -- --check
- name: Install foundry
uses: foundry-rs/foundry-toolchain@8f1998e9878d786675189ef566a2e4bf24869773
with:
version: nightly-41d4e5437107f6f42c7711123890147bc736a609
cache: false
- name: Run forge fmt
run: FOUNDRY_FMT_SORT_INPUTS=false FOUNDRY_FMT_LINE_LENGTH=100 FOUNDRY_FMT_TAB_WIDTH=2 FOUNDRY_FMT_BRACKET_SPACING=true FOUNDRY_FMT_INT_TYPES=preserve forge fmt --check $(find . -iname "*.sol")
machete:
runs-on: ubuntu-latest
steps:
@@ -81,3 +90,25 @@ jobs:
run: |
cargo install cargo-machete
cargo machete
slither:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Slither
run: |
python3 -m pip install solc-select
solc-select install 0.8.26
solc-select use 0.8.26
python3 -m pip install slither-analyzer
slither --include-paths ./networks/ethereum/schnorr/contracts/Schnorr.sol
slither --include-paths ./networks/ethereum/schnorr/contracts ./networks/ethereum/schnorr/contracts/tests/Schnorr.sol
slither processor/ethereum/deployer/contracts/Deployer.sol
slither processor/ethereum/erc20/contracts/IERC20.sol
cp networks/ethereum/schnorr/contracts/Schnorr.sol processor/ethereum/router/contracts/
cp processor/ethereum/erc20/contracts/IERC20.sol processor/ethereum/router/contracts/
cd processor/ethereum/router/contracts
slither Router.sol

259
.github/workflows/msrv.yml vendored Normal file
View File

@@ -0,0 +1,259 @@
name: Weekly MSRV Check
on:
schedule:
- cron: "0 0 * * 0"
workflow_dispatch:
jobs:
msrv-common:
name: Run cargo msrv on common
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on common
run: |
cargo msrv verify --manifest-path common/zalloc/Cargo.toml
cargo msrv verify --manifest-path common/std-shims/Cargo.toml
cargo msrv verify --manifest-path common/env/Cargo.toml
cargo msrv verify --manifest-path common/db/Cargo.toml
cargo msrv verify --manifest-path common/task/Cargo.toml
cargo msrv verify --manifest-path common/request/Cargo.toml
cargo msrv verify --manifest-path common/patchable-async-sleep/Cargo.toml
msrv-crypto:
name: Run cargo msrv on crypto
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on crypto
run: |
cargo msrv verify --manifest-path crypto/transcript/Cargo.toml
cargo msrv verify --manifest-path crypto/ff-group-tests/Cargo.toml
cargo msrv verify --manifest-path crypto/dalek-ff-group/Cargo.toml
cargo msrv verify --manifest-path crypto/ed448/Cargo.toml
cargo msrv verify --manifest-path crypto/multiexp/Cargo.toml
cargo msrv verify --manifest-path crypto/dleq/Cargo.toml
cargo msrv verify --manifest-path crypto/ciphersuite/Cargo.toml
cargo msrv verify --manifest-path crypto/schnorr/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/generalized-bulletproofs/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/circuit-abstraction/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/divisors/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/ec-gadgets/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/embedwards25519/Cargo.toml
cargo msrv verify --manifest-path crypto/evrf/secq256k1/Cargo.toml
cargo msrv verify --manifest-path crypto/dkg/Cargo.toml
cargo msrv verify --manifest-path crypto/frost/Cargo.toml
cargo msrv verify --manifest-path crypto/schnorrkel/Cargo.toml
msrv-networks:
name: Run cargo msrv on networks
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on networks
run: |
cargo msrv verify --manifest-path networks/bitcoin/Cargo.toml
cargo msrv verify --manifest-path networks/ethereum/build-contracts/Cargo.toml
cargo msrv verify --manifest-path networks/ethereum/schnorr/Cargo.toml
cargo msrv verify --manifest-path networks/ethereum/alloy-simple-request-transport/Cargo.toml
cargo msrv verify --manifest-path networks/ethereum/relayer/Cargo.toml --features parity-db
cargo msrv verify --manifest-path networks/monero/io/Cargo.toml
cargo msrv verify --manifest-path networks/monero/generators/Cargo.toml
cargo msrv verify --manifest-path networks/monero/primitives/Cargo.toml
cargo msrv verify --manifest-path networks/monero/ringct/mlsag/Cargo.toml
cargo msrv verify --manifest-path networks/monero/ringct/clsag/Cargo.toml
cargo msrv verify --manifest-path networks/monero/ringct/borromean/Cargo.toml
cargo msrv verify --manifest-path networks/monero/ringct/bulletproofs/Cargo.toml
cargo msrv verify --manifest-path networks/monero/Cargo.toml
cargo msrv verify --manifest-path networks/monero/rpc/Cargo.toml
cargo msrv verify --manifest-path networks/monero/rpc/simple-request/Cargo.toml
cargo msrv verify --manifest-path networks/monero/wallet/address/Cargo.toml
cargo msrv verify --manifest-path networks/monero/wallet/Cargo.toml
cargo msrv verify --manifest-path networks/monero/verify-chain/Cargo.toml
msrv-message-queue:
name: Run cargo msrv on message-queue
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on message-queue
run: |
cargo msrv verify --manifest-path message-queue/Cargo.toml --features parity-db
msrv-processor:
name: Run cargo msrv on processor
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on processor
run: |
cargo msrv verify --manifest-path processor/view-keys/Cargo.toml
cargo msrv verify --manifest-path processor/primitives/Cargo.toml
cargo msrv verify --manifest-path processor/messages/Cargo.toml
cargo msrv verify --manifest-path processor/scanner/Cargo.toml
cargo msrv verify --manifest-path processor/scheduler/primitives/Cargo.toml
cargo msrv verify --manifest-path processor/scheduler/smart-contract/Cargo.toml
cargo msrv verify --manifest-path processor/scheduler/utxo/primitives/Cargo.toml
cargo msrv verify --manifest-path processor/scheduler/utxo/standard/Cargo.toml
cargo msrv verify --manifest-path processor/scheduler/utxo/transaction-chaining/Cargo.toml
cargo msrv verify --manifest-path processor/key-gen/Cargo.toml
cargo msrv verify --manifest-path processor/frost-attempt-manager/Cargo.toml
cargo msrv verify --manifest-path processor/signers/Cargo.toml
cargo msrv verify --manifest-path processor/bin/Cargo.toml --features parity-db
cargo msrv verify --manifest-path processor/bitcoin/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/primitives/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/test-primitives/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/erc20/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/deployer/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/router/Cargo.toml
cargo msrv verify --manifest-path processor/ethereum/Cargo.toml
cargo msrv verify --manifest-path processor/monero/Cargo.toml
msrv-coordinator:
name: Run cargo msrv on coordinator
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on coordinator
run: |
cargo msrv verify --manifest-path coordinator/tributary-sdk/tendermint/Cargo.toml
cargo msrv verify --manifest-path coordinator/tributary-sdk/Cargo.toml
cargo msrv verify --manifest-path coordinator/cosign/Cargo.toml
cargo msrv verify --manifest-path coordinator/substrate/Cargo.toml
cargo msrv verify --manifest-path coordinator/tributary/Cargo.toml
cargo msrv verify --manifest-path coordinator/p2p/Cargo.toml
cargo msrv verify --manifest-path coordinator/p2p/libp2p/Cargo.toml
cargo msrv verify --manifest-path coordinator/Cargo.toml
msrv-substrate:
name: Run cargo msrv on substrate
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on substrate
run: |
cargo msrv verify --manifest-path substrate/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/coins/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/coins/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/dex/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/economic-security/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/genesis-liquidity/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/genesis-liquidity/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/in-instructions/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/in-instructions/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/validator-sets/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/validator-sets/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/emissions/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/emissions/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/signals/primitives/Cargo.toml
cargo msrv verify --manifest-path substrate/signals/pallet/Cargo.toml
cargo msrv verify --manifest-path substrate/abi/Cargo.toml
cargo msrv verify --manifest-path substrate/client/Cargo.toml
cargo msrv verify --manifest-path substrate/runtime/Cargo.toml
cargo msrv verify --manifest-path substrate/node/Cargo.toml
msrv-orchestration:
name: Run cargo msrv on orchestration
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on message-queue
run: |
cargo msrv verify --manifest-path orchestration/Cargo.toml
msrv-mini:
name: Run cargo msrv on mini
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@3df4ab11eba7bda6032a0b82a6bb43b11571feac
- name: Install Build Dependencies
uses: ./.github/actions/build-dependencies
- name: Install cargo msrv
run: cargo install --locked cargo-msrv
- name: Run cargo msrv on mini
run: |
cargo msrv verify --manifest-path mini/Cargo.toml

View File

@@ -30,8 +30,9 @@ jobs:
run: |
GITHUB_CI=true RUST_BACKTRACE=1 cargo test --all-features \
-p bitcoin-serai \
-p build-solidity-contracts \
-p ethereum-schnorr-contract \
-p alloy-simple-request-transport \
-p ethereum-serai \
-p serai-ethereum-relayer \
-p monero-io \
-p monero-generators \

View File

@@ -39,9 +39,33 @@ jobs:
GITHUB_CI=true RUST_BACKTRACE=1 cargo test --all-features \
-p serai-message-queue \
-p serai-processor-messages \
-p serai-processor \
-p serai-processor-key-gen \
-p serai-processor-view-keys \
-p serai-processor-frost-attempt-manager \
-p serai-processor-primitives \
-p serai-processor-scanner \
-p serai-processor-scheduler-primitives \
-p serai-processor-utxo-scheduler-primitives \
-p serai-processor-utxo-scheduler \
-p serai-processor-transaction-chaining-scheduler \
-p serai-processor-smart-contract-scheduler \
-p serai-processor-signers \
-p serai-processor-bin \
-p serai-bitcoin-processor \
-p serai-processor-ethereum-primitives \
-p serai-processor-ethereum-test-primitives \
-p serai-processor-ethereum-deployer \
-p serai-processor-ethereum-router \
-p serai-processor-ethereum-erc20 \
-p serai-ethereum-processor \
-p serai-monero-processor \
-p tendermint-machine \
-p tributary-chain \
-p tributary-sdk \
-p serai-cosign \
-p serai-coordinator-substrate \
-p serai-coordinator-tributary \
-p serai-coordinator-p2p \
-p serai-coordinator-libp2p-p2p \
-p serai-coordinator \
-p serai-orchestrator \
-p serai-docker-tests

2293
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -20,6 +20,7 @@ members = [
"common/patchable-async-sleep",
"common/db",
"common/env",
"common/task",
"common/request",
"crypto/transcript",
@@ -30,17 +31,25 @@ members = [
"crypto/ciphersuite",
"crypto/multiexp",
"crypto/schnorr",
"crypto/dleq",
"crypto/evrf/secq256k1",
"crypto/evrf/embedwards25519",
"crypto/evrf/generalized-bulletproofs",
"crypto/evrf/circuit-abstraction",
"crypto/evrf/divisors",
"crypto/evrf/ec-gadgets",
"crypto/dkg",
"crypto/frost",
"crypto/schnorrkel",
"networks/bitcoin",
"networks/ethereum/build-contracts",
"networks/ethereum/schnorr",
"networks/ethereum/alloy-simple-request-transport",
"networks/ethereum",
"networks/ethereum/relayer",
"networks/monero/io",
@@ -63,10 +72,37 @@ members = [
"message-queue",
"processor/messages",
"processor",
"coordinator/tributary/tendermint",
"processor/key-gen",
"processor/view-keys",
"processor/frost-attempt-manager",
"processor/primitives",
"processor/scanner",
"processor/scheduler/primitives",
"processor/scheduler/utxo/primitives",
"processor/scheduler/utxo/standard",
"processor/scheduler/utxo/transaction-chaining",
"processor/scheduler/smart-contract",
"processor/signers",
"processor/bin",
"processor/bitcoin",
"processor/ethereum/primitives",
"processor/ethereum/test-primitives",
"processor/ethereum/deployer",
"processor/ethereum/router",
"processor/ethereum/erc20",
"processor/ethereum",
"processor/monero",
"coordinator/tributary-sdk/tendermint",
"coordinator/tributary-sdk",
"coordinator/cosign",
"coordinator/substrate",
"coordinator/tributary",
"coordinator/p2p",
"coordinator/p2p/libp2p",
"coordinator",
"substrate/primitives",
@@ -118,18 +154,32 @@ members = [
# to the extensive operations required for Bulletproofs
[profile.dev.package]
subtle = { opt-level = 3 }
curve25519-dalek = { opt-level = 3 }
ff = { opt-level = 3 }
group = { opt-level = 3 }
crypto-bigint = { opt-level = 3 }
secp256k1 = { opt-level = 3 }
curve25519-dalek = { opt-level = 3 }
dalek-ff-group = { opt-level = 3 }
minimal-ed448 = { opt-level = 3 }
multiexp = { opt-level = 3 }
monero-serai = { opt-level = 3 }
secq256k1 = { opt-level = 3 }
embedwards25519 = { opt-level = 3 }
generalized-bulletproofs = { opt-level = 3 }
generalized-bulletproofs-circuit-abstraction = { opt-level = 3 }
ec-divisors = { opt-level = 3 }
generalized-bulletproofs-ec-gadgets = { opt-level = 3 }
dkg = { opt-level = 3 }
monero-generators = { opt-level = 3 }
monero-borromean = { opt-level = 3 }
monero-bulletproofs = { opt-level = 3 }
monero-mlsag = { opt-level = 3 }
monero-clsag = { opt-level = 3 }
[profile.release]
panic = "unwind"
@@ -158,11 +208,12 @@ matches = { path = "patches/matches" }
option-ext = { path = "patches/option-ext" }
directories-next = { path = "patches/directories-next" }
# https://github.com/alloy-rs/core/issues/717
alloy-sol-type-parser = { git = "https://github.com/alloy-rs/core", rev = "446b9d2fbce12b88456152170709a3eaac929af0" }
# The official pasta_curves repo doesn't support Zeroize
pasta_curves = { git = "https://github.com/kayabaNerve/pasta_curves", rev = "a46b5be95cacbff54d06aad8d3bbcba42e05d616" }
[workspace.lints.clippy]
unwrap_or_default = "allow"
map_unwrap_or = "allow"
borrow_as_ptr = "deny"
cast_lossless = "deny"
cast_possible_truncation = "deny"
@@ -190,7 +241,6 @@ manual_instant_elapsed = "deny"
manual_let_else = "deny"
manual_ok_or = "deny"
manual_string_new = "deny"
map_unwrap_or = "deny"
match_bool = "deny"
match_same_arms = "deny"
missing_fields_in_debug = "deny"
@@ -202,6 +252,7 @@ range_plus_one = "deny"
redundant_closure_for_method_calls = "deny"
redundant_else = "deny"
string_add_assign = "deny"
string_slice = "deny"
unchecked_duration_subtraction = "deny"
uninlined_format_args = "deny"
unnecessary_box_returns = "deny"

View File

@@ -1,13 +1,13 @@
[package]
name = "serai-db"
version = "0.1.0"
version = "0.1.1"
description = "A simple database trait and backends for it"
license = "MIT"
repository = "https://github.com/serai-dex/serai/tree/develop/common/db"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
rust-version = "1.65"
rust-version = "1.71"
[package.metadata.docs.rs]
all-features = true
@@ -18,7 +18,7 @@ workspace = true
[dependencies]
parity-db = { version = "0.4", default-features = false, optional = true }
rocksdb = { version = "0.21", default-features = false, features = ["zstd"], optional = true }
rocksdb = { version = "0.23", default-features = false, features = ["zstd"], optional = true }
[features]
parity-db = ["dep:parity-db"]

8
common/db/README.md Normal file
View File

@@ -0,0 +1,8 @@
# Serai DB
An inefficient, minimal abstraction around databases.
The abstraction offers `get`, `put`, and `del` with helper functions and macros
built on top. Database iteration is not offered, forcing the caller to manually
implement indexing schemes. This ensures wide compatibility across abstracted
databases.

View File

@@ -38,12 +38,21 @@ pub fn serai_db_key(
#[macro_export]
macro_rules! create_db {
($db_name: ident {
$($field_name: ident: ($($arg: ident: $arg_type: ty),*) -> $field_type: ty$(,)?)*
$(
$field_name: ident:
$(<$($generic_name: tt: $generic_type: tt),+>)?(
$($arg: ident: $arg_type: ty),*
) -> $field_type: ty$(,)?
)*
}) => {
$(
#[derive(Clone, Debug)]
pub(crate) struct $field_name;
impl $field_name {
pub(crate) struct $field_name$(
<$($generic_name: $generic_type),+>
)?$(
(core::marker::PhantomData<($($generic_name),+)>)
)?;
impl$(<$($generic_name: $generic_type),+>)? $field_name$(<$($generic_name),+>)? {
pub(crate) fn key($($arg: $arg_type),*) -> Vec<u8> {
use scale::Encode;
$crate::serai_db_key(
@@ -52,18 +61,43 @@ macro_rules! create_db {
($($arg),*).encode()
)
}
pub(crate) fn set(txn: &mut impl DbTxn $(, $arg: $arg_type)*, data: &$field_type) {
let key = $field_name::key($($arg),*);
pub(crate) fn set(
txn: &mut impl DbTxn
$(, $arg: $arg_type)*,
data: &$field_type
) {
let key = Self::key($($arg),*);
txn.put(&key, borsh::to_vec(data).unwrap());
}
pub(crate) fn get(getter: &impl Get, $($arg: $arg_type),*) -> Option<$field_type> {
getter.get($field_name::key($($arg),*)).map(|data| {
pub(crate) fn get(
getter: &impl Get,
$($arg: $arg_type),*
) -> Option<$field_type> {
getter.get(Self::key($($arg),*)).map(|data| {
borsh::from_slice(data.as_ref()).unwrap()
})
}
// Returns a PhantomData of all generic types so if the generic was only used in the value,
// not the keys, this doesn't have unused generic types
#[allow(dead_code)]
pub(crate) fn del(txn: &mut impl DbTxn $(, $arg: $arg_type)*) {
txn.del(&$field_name::key($($arg),*))
pub(crate) fn del(
txn: &mut impl DbTxn
$(, $arg: $arg_type)*
) -> core::marker::PhantomData<($($($generic_name),+)?)> {
txn.del(&Self::key($($arg),*));
core::marker::PhantomData
}
pub(crate) fn take(
txn: &mut impl DbTxn
$(, $arg: $arg_type)*
) -> Option<$field_type> {
let key = Self::key($($arg),*);
let res = txn.get(&key).map(|data| borsh::from_slice(data.as_ref()).unwrap());
if res.is_some() {
txn.del(key);
}
res
}
}
)*
@@ -73,19 +107,30 @@ macro_rules! create_db {
#[macro_export]
macro_rules! db_channel {
($db_name: ident {
$($field_name: ident: ($($arg: ident: $arg_type: ty),*) -> $field_type: ty$(,)?)*
$($field_name: ident:
$(<$($generic_name: tt: $generic_type: tt),+>)?(
$($arg: ident: $arg_type: ty),*
) -> $field_type: ty$(,)?
)*
}) => {
$(
create_db! {
$db_name {
$field_name: ($($arg: $arg_type,)* index: u32) -> $field_type,
$field_name: $(<$($generic_name: $generic_type),+>)?(
$($arg: $arg_type,)*
index: u32
) -> $field_type
}
}
impl $field_name {
pub(crate) fn send(txn: &mut impl DbTxn $(, $arg: $arg_type)*, value: &$field_type) {
impl$(<$($generic_name: $generic_type),+>)? $field_name$(<$($generic_name),+>)? {
pub(crate) fn send(
txn: &mut impl DbTxn
$(, $arg: $arg_type)*
, value: &$field_type
) {
// Use index 0 to store the amount of messages
let messages_sent_key = $field_name::key($($arg),*, 0);
let messages_sent_key = Self::key($($arg,)* 0);
let messages_sent = txn.get(&messages_sent_key).map(|counter| {
u32::from_le_bytes(counter.try_into().unwrap())
}).unwrap_or(0);
@@ -96,19 +141,35 @@ macro_rules! db_channel {
// at the same time
let index_to_use = messages_sent + 2;
$field_name::set(txn, $($arg),*, index_to_use, value);
Self::set(txn, $($arg,)* index_to_use, value);
}
pub(crate) fn try_recv(txn: &mut impl DbTxn $(, $arg: $arg_type)*) -> Option<$field_type> {
let messages_recvd_key = $field_name::key($($arg),*, 1);
pub(crate) fn peek(
getter: &impl Get
$(, $arg: $arg_type)*
) -> Option<$field_type> {
let messages_recvd_key = Self::key($($arg,)* 1);
let messages_recvd = getter.get(&messages_recvd_key).map(|counter| {
u32::from_le_bytes(counter.try_into().unwrap())
}).unwrap_or(0);
let index_to_read = messages_recvd + 2;
Self::get(getter, $($arg,)* index_to_read)
}
pub(crate) fn try_recv(
txn: &mut impl DbTxn
$(, $arg: $arg_type)*
) -> Option<$field_type> {
let messages_recvd_key = Self::key($($arg,)* 1);
let messages_recvd = txn.get(&messages_recvd_key).map(|counter| {
u32::from_le_bytes(counter.try_into().unwrap())
}).unwrap_or(0);
let index_to_read = messages_recvd + 2;
let res = $field_name::get(txn, $($arg),*, index_to_read);
let res = Self::get(txn, $($arg,)* index_to_read);
if res.is_some() {
$field_name::del(txn, $($arg),*, index_to_read);
Self::del(txn, $($arg,)* index_to_read);
txn.put(&messages_recvd_key, (messages_recvd + 1).to_le_bytes());
}
res

View File

@@ -14,26 +14,87 @@ mod parity_db;
#[cfg(feature = "parity-db")]
pub use parity_db::{ParityDb, new_parity_db};
/// An object implementing get.
/// An object implementing `get`.
pub trait Get {
/// Get a value from the database.
fn get(&self, key: impl AsRef<[u8]>) -> Option<Vec<u8>>;
}
/// An atomic database operation.
/// An atomic database transaction.
///
/// A transaction is only required to atomically commit. It is not required that two `Get` calls
/// made with the same transaction return the same result, if another transaction wrote to that
/// key.
///
/// If two transactions are created, and both write (including deletions) to the same key, behavior
/// is undefined. The transaction may block, deadlock, panic, overwrite one of the two values
/// randomly, or any other action, at time of write or at time of commit.
#[must_use]
pub trait DbTxn: Send + Get {
pub trait DbTxn: Sized + Send + Get {
/// Write a value to this key.
fn put(&mut self, key: impl AsRef<[u8]>, value: impl AsRef<[u8]>);
/// Delete the value from this key.
fn del(&mut self, key: impl AsRef<[u8]>);
/// Commit this transaction.
fn commit(self);
/// Close this transaction.
///
/// This is equivalent to `Drop` on transactions which can be dropped. This is explicit and works
/// with transactions which can't be dropped.
fn close(self) {
drop(self);
}
}
/// A database supporting atomic operations.
// Credit for the idea goes to https://jack.wrenn.fyi/blog/undroppable
pub struct Undroppable<T>(Option<T>);
impl<T> Drop for Undroppable<T> {
fn drop(&mut self) {
// Use an assertion at compile time to prevent this code from compiling if generated
#[allow(clippy::assertions_on_constants)]
const {
assert!(false, "Undroppable DbTxn was dropped. Ensure all code paths call commit or close");
}
}
}
impl<T: DbTxn> Get for Undroppable<T> {
fn get(&self, key: impl AsRef<[u8]>) -> Option<Vec<u8>> {
self.0.as_ref().unwrap().get(key)
}
}
impl<T: DbTxn> DbTxn for Undroppable<T> {
fn put(&mut self, key: impl AsRef<[u8]>, value: impl AsRef<[u8]>) {
self.0.as_mut().unwrap().put(key, value);
}
fn del(&mut self, key: impl AsRef<[u8]>) {
self.0.as_mut().unwrap().del(key);
}
fn commit(mut self) {
self.0.take().unwrap().commit();
let _ = core::mem::ManuallyDrop::new(self);
}
fn close(mut self) {
drop(self.0.take().unwrap());
let _ = core::mem::ManuallyDrop::new(self);
}
}
/// A database supporting atomic transaction.
pub trait Db: 'static + Send + Sync + Clone + Get {
/// The type representing a database transaction.
type Transaction<'a>: DbTxn;
/// Calculate a key for a database entry.
///
/// Keys are separated by the database, the item within the database, and the item's key itself.
fn key(db_dst: &'static [u8], item_dst: &'static [u8], key: impl AsRef<[u8]>) -> Vec<u8> {
let db_len = u8::try_from(db_dst.len()).unwrap();
let dst_len = u8::try_from(item_dst.len()).unwrap();
[[db_len].as_ref(), db_dst, [dst_len].as_ref(), item_dst, key.as_ref()].concat()
}
fn txn(&mut self) -> Self::Transaction<'_>;
/// Open a new transaction which may be dropped.
fn unsafe_txn(&mut self) -> Self::Transaction<'_>;
/// Open a new transaction which must be committed or closed.
fn txn(&mut self) -> Undroppable<Self::Transaction<'_>> {
Undroppable(Some(self.unsafe_txn()))
}
}

View File

@@ -11,7 +11,7 @@ use crate::*;
#[derive(PartialEq, Eq, Debug)]
pub struct MemDbTxn<'a>(&'a MemDb, HashMap<Vec<u8>, Vec<u8>>, HashSet<Vec<u8>>);
impl<'a> Get for MemDbTxn<'a> {
impl Get for MemDbTxn<'_> {
fn get(&self, key: impl AsRef<[u8]>) -> Option<Vec<u8>> {
if self.2.contains(key.as_ref()) {
return None;
@@ -23,7 +23,7 @@ impl<'a> Get for MemDbTxn<'a> {
.or_else(|| self.0 .0.read().unwrap().get(key.as_ref()).cloned())
}
}
impl<'a> DbTxn for MemDbTxn<'a> {
impl DbTxn for MemDbTxn<'_> {
fn put(&mut self, key: impl AsRef<[u8]>, value: impl AsRef<[u8]>) {
self.2.remove(key.as_ref());
self.1.insert(key.as_ref().to_vec(), value.as_ref().to_vec());
@@ -74,7 +74,7 @@ impl Get for MemDb {
}
impl Db for MemDb {
type Transaction<'a> = MemDbTxn<'a>;
fn txn(&mut self) -> MemDbTxn<'_> {
fn unsafe_txn(&mut self) -> MemDbTxn<'_> {
MemDbTxn(self, HashMap::new(), HashSet::new())
}
}

View File

@@ -37,7 +37,7 @@ impl Get for Arc<ParityDb> {
}
impl Db for Arc<ParityDb> {
type Transaction<'a> = Transaction<'a>;
fn txn(&mut self) -> Self::Transaction<'_> {
fn unsafe_txn(&mut self) -> Self::Transaction<'_> {
Transaction(self, vec![])
}
}

View File

@@ -39,7 +39,7 @@ impl<T: ThreadMode> Get for Arc<OptimisticTransactionDB<T>> {
}
impl<T: Send + ThreadMode + 'static> Db for Arc<OptimisticTransactionDB<T>> {
type Transaction<'a> = Transaction<'a, T>;
fn txn(&mut self) -> Self::Transaction<'_> {
fn unsafe_txn(&mut self) -> Self::Transaction<'_> {
let mut opts = WriteOptions::default();
opts.set_sync(true);
Transaction(self.transaction_opt(&opts, &Default::default()), &**self)

View File

@@ -7,7 +7,7 @@ repository = "https://github.com/serai-dex/serai/tree/develop/common/env"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
rust-version = "1.60"
rust-version = "1.71"
[package.metadata.docs.rs]
all-features = true

View File

@@ -7,6 +7,7 @@ repository = "https://github.com/serai-dex/serai/tree/develop/common/patchable-a
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = ["async", "sleep", "tokio", "smol", "async-std"]
edition = "2021"
rust-version = "1.71"
[package.metadata.docs.rs]
all-features = true

View File

@@ -7,7 +7,7 @@ repository = "https://github.com/serai-dex/serai/tree/develop/common/simple-requ
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = ["http", "https", "async", "request", "ssl"]
edition = "2021"
rust-version = "1.64"
rust-version = "1.71"
[package.metadata.docs.rs]
all-features = true

View File

@@ -7,7 +7,7 @@ repository = "https://github.com/serai-dex/serai/tree/develop/common/std-shims"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = ["nostd", "no_std", "alloc", "io"]
edition = "2021"
rust-version = "1.70"
rust-version = "1.80"
[package.metadata.docs.rs]
all-features = true
@@ -18,7 +18,7 @@ workspace = true
[dependencies]
spin = { version = "0.9", default-features = false, features = ["use_ticket_mutex", "lazy"] }
hashbrown = { version = "0.14", default-features = false, features = ["ahash", "inline-more"] }
hashbrown = { version = "0.15", default-features = false, features = ["default-hasher", "inline-more"] }
[features]
std = []

22
common/task/Cargo.toml Normal file
View File

@@ -0,0 +1,22 @@
[package]
name = "serai-task"
version = "0.1.0"
description = "A task schema for Serai services"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/common/task"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.75"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
log = { version = "0.4", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false, features = ["macros", "sync", "time"] }

View File

@@ -1,6 +1,6 @@
AGPL-3.0-only license
Copyright (c) 2022-2023 Luke Parker
Copyright (c) 2022-2024 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as

3
common/task/README.md Normal file
View File

@@ -0,0 +1,3 @@
# Task
A schema to define tasks to be run ad infinitum.

161
common/task/src/lib.rs Normal file
View File

@@ -0,0 +1,161 @@
#![cfg_attr(docsrs, feature(doc_auto_cfg))]
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
use core::{
fmt::{self, Debug},
future::Future,
time::Duration,
};
use tokio::sync::mpsc;
mod type_name;
/// A handle for a task.
///
/// The task will only stop running once all handles for it are dropped.
//
// `run_now` isn't infallible if the task may have been closed. `run_now` on a closed task would
// either need to panic (historic behavior), silently drop the fact the task can't be run, or
// return an error. Instead of having a potential panic, and instead of modeling the error
// behavior, this task can't be closed unless all handles are dropped, ensuring calls to `run_now`
// are infallible.
#[derive(Clone)]
pub struct TaskHandle {
run_now: mpsc::Sender<()>,
#[allow(dead_code)] // This is used to track if all handles have been dropped
close: mpsc::Sender<()>,
}
/// A task's internal structures.
pub struct Task {
run_now: mpsc::Receiver<()>,
close: mpsc::Receiver<()>,
}
impl Task {
/// Create a new task definition.
pub fn new() -> (Self, TaskHandle) {
// Uses a capacity of 1 as any call to run as soon as possible satisfies all calls to run as
// soon as possible
let (run_now_send, run_now_recv) = mpsc::channel(1);
// And any call to close satisfies all calls to close
let (close_send, close_recv) = mpsc::channel(1);
(
Self { run_now: run_now_recv, close: close_recv },
TaskHandle { run_now: run_now_send, close: close_send },
)
}
}
impl TaskHandle {
/// Tell the task to run now (and not whenever its next iteration on a timer is).
pub fn run_now(&self) {
#[allow(clippy::match_same_arms)]
match self.run_now.try_send(()) {
Ok(()) => {}
// NOP on full, as this task will already be ran as soon as possible
Err(mpsc::error::TrySendError::Full(())) => {}
Err(mpsc::error::TrySendError::Closed(())) => {
// The task should only be closed if all handles are dropped, and this one hasn't been
panic!("task was unexpectedly closed when calling run_now")
}
}
}
}
/// An enum which can't be constructed, representing that the task does not error.
pub enum DoesNotError {}
impl Debug for DoesNotError {
fn fmt(&self, _: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
// This type can't be constructed so we'll never have a `&self` to call this fn with
unreachable!()
}
}
/// A task to be continually ran.
pub trait ContinuallyRan: Sized + Send {
/// The amount of seconds before this task should be polled again.
const DELAY_BETWEEN_ITERATIONS: u64 = 5;
/// The maximum amount of seconds before this task should be run again.
///
/// Upon error, the amount of time waited will be linearly increased until this limit.
const MAX_DELAY_BETWEEN_ITERATIONS: u64 = 120;
/// The error potentially yielded upon running an iteration of this task.
type Error: Debug;
/// Run an iteration of the task.
///
/// If this returns `true`, all dependents of the task will immediately have a new iteration ran
/// (without waiting for whatever timer they were already on).
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>>;
/// Continually run the task.
fn continually_run(
mut self,
mut task: Task,
dependents: Vec<TaskHandle>,
) -> impl Send + Future<Output = ()> {
async move {
// The default number of seconds to sleep before running the task again
let default_sleep_before_next_task = Self::DELAY_BETWEEN_ITERATIONS;
// The current number of seconds to sleep before running the task again
// We increment this upon errors in order to not flood the logs with errors
let mut current_sleep_before_next_task = default_sleep_before_next_task;
let increase_sleep_before_next_task = |current_sleep_before_next_task: &mut u64| {
let new_sleep = *current_sleep_before_next_task + default_sleep_before_next_task;
// Set a limit of sleeping for two minutes
*current_sleep_before_next_task = new_sleep.max(Self::MAX_DELAY_BETWEEN_ITERATIONS);
};
loop {
// If we were told to close/all handles were dropped, drop it
{
let should_close = task.close.try_recv();
match should_close {
Ok(()) | Err(mpsc::error::TryRecvError::Disconnected) => break,
Err(mpsc::error::TryRecvError::Empty) => {}
}
}
match self.run_iteration().await {
Ok(run_dependents) => {
// Upon a successful (error-free) loop iteration, reset the amount of time we sleep
current_sleep_before_next_task = default_sleep_before_next_task;
if run_dependents {
for dependent in &dependents {
dependent.run_now();
}
}
}
Err(e) => {
// Get the type name
let type_name = type_name::strip_type_name(core::any::type_name::<Self>());
// Print the error as a warning, prefixed by the task's type
log::warn!("{type_name}: {e:?}");
increase_sleep_before_next_task(&mut current_sleep_before_next_task);
}
}
// Don't run the task again for another few seconds UNLESS told to run now
/*
We could replace tokio::mpsc with async_channel, tokio::time::sleep with
patchable_async_sleep::sleep, and tokio::select with futures_lite::future::or
It isn't worth the effort when patchable_async_sleep::sleep will still resolve to tokio
*/
tokio::select! {
() = tokio::time::sleep(Duration::from_secs(current_sleep_before_next_task)) => {},
msg = task.run_now.recv() => {
// Check if this is firing because the handle was dropped
if msg.is_none() {
break;
}
},
}
}
}
}
}

View File

@@ -0,0 +1,31 @@
/// Strip the modules from a type name.
// This may be of the form `a::b::C`, in which case we only want `C`
pub(crate) fn strip_type_name(full_type_name: &'static str) -> String {
// It also may be `a::b::C<d::e::F>`, in which case, we only attempt to strip `a::b`
let mut by_generics = full_type_name.split('<');
// Strip to just `C`
let full_outer_object_name = by_generics.next().unwrap();
let mut outer_object_name_parts = full_outer_object_name.split("::");
let mut last_part_in_outer_object_name = outer_object_name_parts.next().unwrap();
for part in outer_object_name_parts {
last_part_in_outer_object_name = part;
}
// Push back on the generic terms
let mut type_name = last_part_in_outer_object_name.to_string();
for generic in by_generics {
type_name.push('<');
type_name.push_str(generic);
}
type_name
}
#[test]
fn test_strip_type_name() {
assert_eq!(strip_type_name("core::option::Option"), "Option");
assert_eq!(
strip_type_name("core::option::Option<alloc::string::String>"),
"Option<alloc::string::String>"
);
}

View File

@@ -7,7 +7,7 @@ repository = "https://github.com/serai-dex/serai/tree/develop/common/zalloc"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
rust-version = "1.77.0"
rust-version = "1.77"
[package.metadata.docs.rs]
all-features = true

View File

@@ -8,6 +8,7 @@ authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
@@ -17,30 +18,29 @@ rustdoc-args = ["--cfg", "docsrs"]
workspace = true
[dependencies]
async-trait = { version = "0.1", default-features = false }
zeroize = { version = "^1.5", default-features = false, features = ["std"] }
bitvec = { version = "1", default-features = false, features = ["std"] }
rand_core = { version = "0.6", default-features = false, features = ["std"] }
blake2 = { version = "0.10", default-features = false, features = ["std"] }
schnorrkel = { version = "0.11", default-features = false, features = ["std"] }
transcript = { package = "flexible-transcript", path = "../crypto/transcript", default-features = false, features = ["std", "recommended"] }
ciphersuite = { path = "../crypto/ciphersuite", default-features = false, features = ["std"] }
schnorr = { package = "schnorr-signatures", path = "../crypto/schnorr", default-features = false, features = ["std"] }
frost = { package = "modular-frost", path = "../crypto/frost" }
frost-schnorrkel = { path = "../crypto/schnorrkel" }
scale = { package = "parity-scale-codec", version = "3", default-features = false, features = ["std", "derive"] }
scale = { package = "parity-scale-codec", version = "3", default-features = false, features = ["std", "derive", "bit-vec"] }
zalloc = { path = "../common/zalloc" }
serai-db = { path = "../common/db" }
serai-env = { path = "../common/env" }
serai-task = { path = "../common/task", version = "0.1" }
processor-messages = { package = "serai-processor-messages", path = "../processor/messages" }
messages = { package = "serai-processor-messages", path = "../processor/messages" }
message-queue = { package = "serai-message-queue", path = "../message-queue" }
tributary = { package = "tributary-chain", path = "./tributary" }
tributary-sdk = { path = "./tributary-sdk" }
sp-application-crypto = { git = "https://github.com/serai-dex/substrate", default-features = false, features = ["std"] }
serai-client = { path = "../substrate/client", default-features = false, features = ["serai", "borsh"] }
hex = { version = "0.4", default-features = false, features = ["std"] }
@@ -49,16 +49,15 @@ borsh = { version = "1", default-features = false, features = ["std", "derive",
log = { version = "0.4", default-features = false, features = ["std"] }
env_logger = { version = "0.10", default-features = false, features = ["humantime"] }
futures-util = { version = "0.3", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false, features = ["rt-multi-thread", "sync", "time", "macros"] }
libp2p = { version = "0.52", default-features = false, features = ["tokio", "tcp", "noise", "yamux", "request-response", "gossipsub", "macros"] }
tokio = { version = "1", default-features = false, features = ["time", "sync", "macros", "rt-multi-thread"] }
[dev-dependencies]
tributary = { package = "tributary-chain", path = "./tributary", features = ["tests"] }
sp-application-crypto = { git = "https://github.com/serai-dex/substrate", default-features = false, features = ["std"] }
sp-runtime = { git = "https://github.com/serai-dex/substrate", default-features = false, features = ["std"] }
serai-cosign = { path = "./cosign" }
serai-coordinator-substrate = { path = "./substrate" }
serai-coordinator-tributary = { path = "./tributary" }
serai-coordinator-p2p = { path = "./p2p" }
serai-coordinator-libp2p-p2p = { path = "./p2p/libp2p" }
[features]
longer-reattempts = []
longer-reattempts = ["serai-coordinator-tributary/longer-reattempts"]
parity-db = ["serai-db/parity-db"]
rocksdb = ["serai-db/rocksdb"]

View File

@@ -1,6 +1,6 @@
AGPL-3.0-only license
Copyright (c) 2023 Luke Parker
Copyright (c) 2023-2025 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as

View File

@@ -1,7 +1,29 @@
# Coordinator
The Serai coordinator communicates with other coordinators to prepare batches
for Serai and sign transactions.
- [`tendermint`](/tributary/tendermint) is an implementation of the Tendermint
BFT algorithm.
In order to achieve consensus over gossip, and order certain events, a
micro-blockchain is instantiated.
- [`tributary-sdk`](./tributary-sdk) is a micro-blockchain framework. Instead
of a producing a blockchain daemon like the Polkadot SDK or Cosmos SDK intend
to, `tributary` is solely intended to be an embedded asynchronous task within
an application.
The Serai coordinator spawns a tributary for each validator set it's
coordinating. This allows the participating validators to communicate in a
byzantine-fault-tolerant manner (relying on Tendermint for consensus).
- [`cosign`](./cosign) contains a library to decide which Substrate blocks
should be cosigned and to evaluate cosigns.
- [`substrate`](./substrate) contains a library to index the Substrate
blockchain and handle its events.
- [`tributary`](./tributary) is our instantiation of the Tributary SDK for the
Serai processor. It includes the `Transaction` definition and deferred
execution logic.
- [`p2p`](./p2p) is our abstract P2P API to service the Coordinator.
- [`libp2p`](./p2p/libp2p) is our libp2p-backed implementation of the P2P API.
- [`src`](./src) contains the source code for the Coordinator binary itself.

View File

@@ -0,0 +1,33 @@
[package]
name = "serai-cosign"
version = "0.1.0"
description = "Evaluator of cosigns for the Serai network"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/coordinator/cosign"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
blake2 = { version = "0.10", default-features = false, features = ["std"] }
schnorrkel = { version = "0.11", default-features = false, features = ["std"] }
scale = { package = "parity-scale-codec", version = "3", default-features = false, features = ["std", "derive"] }
borsh = { version = "1", default-features = false, features = ["std", "derive", "de_strict_order"] }
serai-client = { path = "../../substrate/client", default-features = false, features = ["serai", "borsh"] }
log = { version = "0.4", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false }
serai-db = { path = "../../common/db", version = "0.1.1" }
serai-task = { path = "../../common/task", version = "0.1" }

View File

@@ -1,6 +1,6 @@
AGPL-3.0-only license
Copyright (c) 2022-2023 Luke Parker
Copyright (c) 2023-2024 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as

View File

@@ -0,0 +1,121 @@
# Serai Cosign
The Serai blockchain is controlled by a set of validators referred to as the
Serai validators. These validators could attempt to double-spend, even if every
node on the network is a full node, via equivocating.
Posit:
- The Serai validators control X SRI
- The Serai validators produce block A swapping X SRI to Y XYZ
- The Serai validators produce block B swapping X SRI to Z ABC
- The Serai validators finalize block A and send to the validators for XYZ
- The Serai validators finalize block B and send to the validators for ABC
This is solved via the cosigning protocol. The validators for XYZ and the
validators for ABC each sign their view of the Serai blockchain, communicating
amongst each other to ensure consistency.
The security of the cosigning protocol is not formally proven, and there are no
claims it achieves Byzantine Fault Tolerance. This protocol is meant to be
practical and make such attacks infeasible, when they could already be argued
difficult to perform.
### Definitions
- Cosign: A signature from a non-Serai validator set for a Serai block
- Cosign Commit: A collection of cosigns which achieve the necessary weight
### Methodology
Finalized blocks from the Serai network are intended to be cosigned if they
contain burn events. Only once cosigned should non-Serai validators process
them.
Cosigning occurs by a non-Serai validator set, using their threshold keys
declared on the Serai blockchain. Once 83% of non-Serai validator sets, by
weight, cosign a block, a cosign commit is formed. A cosign commit for a block
is considered to also cosign for all blocks preceding it.
### Bounds Under Asynchrony
Assuming an asynchronous environment fully controlled by the adversary, 34% of
a validator set may cause an equivocation. Control of 67% of non-Serai
validator sets, by weight, is sufficient to produce two distinct cosign commits
at the same position. This is due to the honest stake, 33%, being split across
the two candidates (67% + 16.5% = 83.5%, just over the threshold). This means
the cosigning protocol may produce multiple cosign commits if 34% of 67%, just
22.78%, of the non-Serai validator sets, is malicious. This would be in
conjunction with 34% of the Serai validator set (assumed 20% of total stake),
for a total stake requirement of 34% of 20% + 22.78% of 80% (25.024%). This is
an increase from the 6.8% required without the cosigning protocol.
### Bounds Under Synchrony
Assuming the honest stake within the non-Serai validator sets detect the
malicious stake within their set prior to assisting in producing a cosign for
their set, for which there is a multi-second window, 67% of 67% of non-Serai
validator sets is required to produce cosigns for those sets. This raises the
total stake requirement to 42.712% (past the usual 34% threshold).
### Behavior Reliant on Synchrony
If the Serai blockchain node detects an equivocation, it will stop responding
to all RPC requests and stop participating in finalizing further blocks. This
lets the node communicate the equivocating commits to other nodes (causing them
to exhibit the same behavior), yet prevents interaction with it.
If cosigns representing 17% of the non-Serai validators sets by weight are
detected for distinct blocks at the same position, the protocol halts. An
explicit latency period of seventy seconds is enacted after receiving a cosign
commit for the detection of such an equivocation. This is largely redundant
given how the Serai blockchain node will presumably have halted itself by this
time.
### Equivocation-Detection Avoidance
Malicious Serai validators could avoid detection of their equivocating if they
produced two distinct blockchains, A and B, with different keys declared for
the same non-Serai validator set. While the validators following A may detect
the cosigns for distinct blocks by validators following B, the cosigns would be
assumed invalid due to their signatures being verified against distinct keys.
This is prevented by requiring cosigns on the blocks which declare new keys,
ensuring all validators have a consistent view of the keys used within the
cosigning protocol (per the bounds of the cosigning protocol). These blocks are
exempt from the general policy of cosign commits cosigning all prior blocks,
preventing the newly declared keys (which aren't yet cosigned) from being used
to cosign themselves. These cosigns are flagged as "notable", are permanently
archived, and must be synced before a validator will move forward.
Cosigning the block which declares new keys also ensures agreement on the
preceding block which declared the new set, with an exact specification of the
participants and their weight, before it impacts the cosigning protocol.
### Denial of Service Concerns
Any historical Serai validator set may trigger a chain halt by producing an
equivocation after their retiry. This requires 67% to be malicious. 34% of the
active Serai validator set may also trigger a chain halt.
17% of non-Serai validator sets equivocating causing a halt means 5.67% of
non-Serai validator sets' stake may cause a halt (in an asynchronous
environment fully controlled by the adversary). In a synchronous environment
where the honest stake cannot be split across two candidates, 11.33% of
non-Serai validator sets' stake is required.
The more practical attack is for one to obtain 5.67% of non-Serai validator
sets' stake, under any network conditions, and simply go offline. This will
take 17% of validator sets offline with it, preventing any cosign commits
from being performed. A fallback protocol where validators individually produce
cosigns, removing the network's horizontal scalability but ensuring liveness,
prevents this, restoring the additional requirements for control of an
asynchronous network or 11.33% of non-Serai validator sets' stake.
### TODO
The Serai node no longer responding to RPC requests upon detecting any
equivocation, and the fallback protocol where validators individually produce
signatures, are not implemented at this time. The former means the detection of
equivocating cosigns is not redundant and the latter makes 5.67% of non-Serai
validator sets' stake the DoS threshold, even without control of an
asynchronous network.

View File

@@ -0,0 +1,70 @@
use core::future::Future;
use std::time::{Duration, SystemTime};
use serai_db::*;
use serai_task::{DoesNotError, ContinuallyRan};
use crate::evaluator::CosignedBlocks;
/// How often callers should broadcast the cosigns flagged for rebroadcasting.
pub const BROADCAST_FREQUENCY: Duration = Duration::from_secs(60);
const SYNCHRONY_EXPECTATION: Duration = Duration::from_secs(10);
const ACKNOWLEDGEMENT_DELAY: Duration =
Duration::from_secs(BROADCAST_FREQUENCY.as_secs() + SYNCHRONY_EXPECTATION.as_secs());
create_db!(
SubstrateCosignDelay {
// The latest cosigned block number.
LatestCosignedBlockNumber: () -> u64,
}
);
/// A task to delay acknowledgement of cosigns.
pub(crate) struct CosignDelayTask<D: Db> {
pub(crate) db: D,
}
struct AwaitUndroppable<T: DbTxn>(Option<core::mem::ManuallyDrop<Undroppable<T>>>);
impl<T: DbTxn> Drop for AwaitUndroppable<T> {
fn drop(&mut self) {
if let Some(mut txn) = self.0.take() {
(unsafe { core::mem::ManuallyDrop::take(&mut txn) }).close();
}
}
}
impl<D: Db> ContinuallyRan for CosignDelayTask<D> {
type Error = DoesNotError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
loop {
let mut txn = self.db.txn();
// Receive the next block to mark as cosigned
let Some((block_number, time_evaluated)) = CosignedBlocks::try_recv(&mut txn) else {
txn.close();
break;
};
// Calculate when we should mark it as valid
let time_valid =
SystemTime::UNIX_EPOCH + Duration::from_secs(time_evaluated) + ACKNOWLEDGEMENT_DELAY;
// Sleep until then
let mut txn = AwaitUndroppable(Some(core::mem::ManuallyDrop::new(txn)));
tokio::time::sleep(SystemTime::now().duration_since(time_valid).unwrap_or(Duration::ZERO))
.await;
let mut txn = core::mem::ManuallyDrop::into_inner(txn.0.take().unwrap());
// Set the cosigned block
LatestCosignedBlockNumber::set(&mut txn, &block_number);
txn.commit();
made_progress = true;
}
Ok(made_progress)
}
}
}

View File

@@ -0,0 +1,233 @@
use core::future::Future;
use std::time::{Duration, SystemTime};
use serai_db::*;
use serai_task::ContinuallyRan;
use crate::{
HasEvents, GlobalSession, NetworksLatestCosignedBlock, RequestNotableCosigns,
intend::{GlobalSessionsChannel, BlockEventData, BlockEvents},
};
create_db!(
SubstrateCosignEvaluator {
// The global session currently being evaluated.
CurrentlyEvaluatedGlobalSession: () -> ([u8; 32], GlobalSession),
}
);
db_channel!(
SubstrateCosignEvaluatorChannels {
// (cosigned block, time cosign was evaluated)
CosignedBlocks: () -> (u64, u64),
}
);
// This is a strict function which won't panic, even with a malicious Serai node, so long as:
// - It's called incrementally (with an increment of 1)
// - It's only called for block numbers we've completed indexing on within the intend task
// - It's only called for block numbers after a global session has started
// - The global sessions channel is populated as the block declaring the session is indexed
// Which all hold true within the context of this task and the intend task.
//
// This function will also ensure the currently evaluated global session is incremented once we
// finish evaluation of the prior session.
fn currently_evaluated_global_session_strict(
txn: &mut impl DbTxn,
block_number: u64,
) -> ([u8; 32], GlobalSession) {
let mut res = {
let existing = match CurrentlyEvaluatedGlobalSession::get(txn) {
Some(existing) => existing,
None => {
let first = GlobalSessionsChannel::try_recv(txn)
.expect("fetching latest global session yet none declared");
CurrentlyEvaluatedGlobalSession::set(txn, &first);
first
}
};
assert!(
existing.1.start_block_number <= block_number,
"candidate's start block number exceeds our block number"
);
existing
};
if let Some(next) = GlobalSessionsChannel::peek(txn) {
assert!(
block_number <= next.1.start_block_number,
"currently_evaluated_global_session_strict wasn't called incrementally"
);
// If it's time for this session to activate, take it from the channel and set it
if block_number == next.1.start_block_number {
GlobalSessionsChannel::try_recv(txn).unwrap();
CurrentlyEvaluatedGlobalSession::set(txn, &next);
res = next;
}
}
res
}
pub(crate) fn currently_evaluated_global_session(getter: &impl Get) -> Option<[u8; 32]> {
CurrentlyEvaluatedGlobalSession::get(getter).map(|(id, _info)| id)
}
/// A task to determine if a block has been cosigned and we should handle it.
pub(crate) struct CosignEvaluatorTask<D: Db, R: RequestNotableCosigns> {
pub(crate) db: D,
pub(crate) request: R,
}
impl<D: Db, R: RequestNotableCosigns> ContinuallyRan for CosignEvaluatorTask<D, R> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut known_cosign = None;
let mut made_progress = false;
loop {
let mut txn = self.db.unsafe_txn();
let Some(BlockEventData { block_number, has_events }) = BlockEvents::try_recv(&mut txn)
else {
break;
};
// Fetch the global session information
let (global_session, global_session_info) =
currently_evaluated_global_session_strict(&mut txn, block_number);
match has_events {
// Because this had notable events, we require an explicit cosign for this block by a
// supermajority of the prior block's validator sets
HasEvents::Notable => {
let mut weight_cosigned = 0;
for set in global_session_info.sets {
// Check if we have the cosign from this set
if NetworksLatestCosignedBlock::get(&txn, global_session, set.network)
.map(|signed_cosign| signed_cosign.cosign.block_number) ==
Some(block_number)
{
// Since have this cosign, add the set's weight to the weight which has cosigned
weight_cosigned +=
global_session_info.stakes.get(&set.network).ok_or_else(|| {
"ValidatorSet in global session yet didn't have its stake".to_string()
})?;
}
}
// Check if the sum weight doesn't cross the required threshold
if weight_cosigned < (((global_session_info.total_stake * 83) / 100) + 1) {
// Request the necessary cosigns over the network
// TODO: Add a timer to ensure this isn't called too often
self
.request
.request_notable_cosigns(global_session)
.await
.map_err(|e| format!("{e:?}"))?;
// We return an error so the delay before this task is run again increases
return Err(format!(
"notable block (#{block_number}) wasn't yet cosigned. this should resolve shortly",
));
}
log::info!("marking notable block #{block_number} as cosigned");
}
// Since this block didn't have any notable events, we simply require a cosign for this
// block or a greater block by the current validator sets
HasEvents::NonNotable => {
// Check if this was satisfied by a cached result which wasn't calculated incrementally
let known_cosigned = if let Some(known_cosign) = known_cosign {
known_cosign >= block_number
} else {
// Clear `known_cosign` which is no longer helpful
known_cosign = None;
false
};
// If it isn't already known to be cosigned, evaluate the latest cosigns
if !known_cosigned {
/*
LatestCosign is populated with the latest cosigns for each network which don't
exceed the latest global session we've evaluated the start of. This current block
is during the latest global session we've evaluated the start of.
*/
let mut weight_cosigned = 0;
let mut lowest_common_block: Option<u64> = None;
for set in global_session_info.sets {
// Check if this set cosigned this block or not
let Some(cosign) =
NetworksLatestCosignedBlock::get(&txn, global_session, set.network)
else {
continue;
};
if cosign.cosign.block_number >= block_number {
weight_cosigned +=
global_session_info.stakes.get(&set.network).ok_or_else(|| {
"ValidatorSet in global session yet didn't have its stake".to_string()
})?;
}
// Update the lowest block common to all of these cosigns
lowest_common_block = lowest_common_block
.map(|existing| existing.min(cosign.cosign.block_number))
.or(Some(cosign.cosign.block_number));
}
// Check if the sum weight doesn't cross the required threshold
if weight_cosigned < (((global_session_info.total_stake * 83) / 100) + 1) {
// Request the superseding notable cosigns over the network
// If this session hasn't yet produced notable cosigns, then we presume we'll see
// the desired non-notable cosigns as part of normal operations, without needing to
// explicitly request them
self
.request
.request_notable_cosigns(global_session)
.await
.map_err(|e| format!("{e:?}"))?;
// We return an error so the delay before this task is run again increases
return Err(format!(
"block (#{block_number}) wasn't yet cosigned. this should resolve shortly",
));
}
// Update the cached result for the block we know is cosigned
/*
There may be a higher block which was cosigned, but once we get to this block,
we'll re-evaluate and find it then. The alternative would be an optimistic
re-evaluation now. Both are fine, so the lower-complexity option is preferred.
*/
known_cosign = lowest_common_block;
}
log::debug!("marking non-notable block #{block_number} as cosigned");
}
// If this block has no events necessitating cosigning, we can immediately consider the
// block cosigned (making this block a NOP)
HasEvents::No => {}
}
// Since we checked we had the necessary cosigns, send it for delay before acknowledgement
CosignedBlocks::send(
&mut txn,
&(
block_number,
SystemTime::now()
.duration_since(SystemTime::UNIX_EPOCH)
.unwrap_or(Duration::ZERO)
.as_secs(),
),
);
txn.commit();
if (block_number % 500) == 0 {
log::info!("marking block #{block_number} as cosigned");
}
made_progress = true;
}
Ok(made_progress)
}
}
}

View File

@@ -0,0 +1,184 @@
use core::future::Future;
use std::{sync::Arc, collections::HashMap};
use serai_client::{
primitives::{SeraiAddress, Amount},
validator_sets::primitives::ValidatorSet,
Serai,
};
use serai_db::*;
use serai_task::ContinuallyRan;
use crate::*;
create_db!(
CosignIntend {
ScanCosignFrom: () -> u64,
}
);
#[derive(Debug, BorshSerialize, BorshDeserialize)]
pub(crate) struct BlockEventData {
pub(crate) block_number: u64,
pub(crate) has_events: HasEvents,
}
db_channel! {
CosignIntendChannels {
GlobalSessionsChannel: () -> ([u8; 32], GlobalSession),
BlockEvents: () -> BlockEventData,
IntendedCosigns: (set: ValidatorSet) -> CosignIntent,
}
}
async fn block_has_events_justifying_a_cosign(
serai: &Serai,
block_number: u64,
) -> Result<(Block, HasEvents), String> {
let block = serai
.finalized_block_by_number(block_number)
.await
.map_err(|e| format!("{e:?}"))?
.ok_or_else(|| "couldn't get block which should've been finalized".to_string())?;
let serai = serai.as_of(block.hash());
if !serai.validator_sets().key_gen_events().await.map_err(|e| format!("{e:?}"))?.is_empty() {
return Ok((block, HasEvents::Notable));
}
if !serai.coins().burn_with_instruction_events().await.map_err(|e| format!("{e:?}"))?.is_empty() {
return Ok((block, HasEvents::NonNotable));
}
Ok((block, HasEvents::No))
}
/// A task to determine which blocks we should intend to cosign.
pub(crate) struct CosignIntendTask<D: Db> {
pub(crate) db: D,
pub(crate) serai: Arc<Serai>,
}
impl<D: Db> ContinuallyRan for CosignIntendTask<D> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let start_block_number = ScanCosignFrom::get(&self.db).unwrap_or(1);
let latest_block_number =
self.serai.latest_finalized_block().await.map_err(|e| format!("{e:?}"))?.number();
for block_number in start_block_number ..= latest_block_number {
let mut txn = self.db.unsafe_txn();
let (block, mut has_events) =
block_has_events_justifying_a_cosign(&self.serai, block_number)
.await
.map_err(|e| format!("{e:?}"))?;
// Check we are indexing a linear chain
if (block_number > 1) &&
(<[u8; 32]>::from(block.header.parent_hash) !=
SubstrateBlockHash::get(&txn, block_number - 1)
.expect("indexing a block but haven't indexed its parent"))
{
Err(format!(
"node's block #{block_number} doesn't build upon the block #{} prior indexed",
block_number - 1
))?;
}
let block_hash = block.hash();
SubstrateBlockHash::set(&mut txn, block_number, &block_hash);
let global_session_for_this_block = LatestGlobalSessionIntended::get(&txn);
// If this is notable, it creates a new global session, which we index into the database
// now
if has_events == HasEvents::Notable {
let serai = self.serai.as_of(block_hash);
let sets_and_keys = cosigning_sets(&serai).await?;
let global_session =
GlobalSession::id(sets_and_keys.iter().map(|(set, _key)| *set).collect());
let mut sets = Vec::with_capacity(sets_and_keys.len());
let mut keys = HashMap::with_capacity(sets_and_keys.len());
let mut stakes = HashMap::with_capacity(sets_and_keys.len());
let mut total_stake = 0;
for (set, key) in &sets_and_keys {
sets.push(*set);
keys.insert(set.network, SeraiAddress::from(*key));
let stake = serai
.validator_sets()
.total_allocated_stake(set.network)
.await
.map_err(|e| format!("{e:?}"))?
.unwrap_or(Amount(0))
.0;
stakes.insert(set.network, stake);
total_stake += stake;
}
if total_stake == 0 {
Err(format!("cosigning sets for block #{block_number} had 0 stake in total"))?;
}
let global_session_info = GlobalSession {
// This session starts cosigning after this block, as this block must be cosigned by
// the existing validators
start_block_number: block_number + 1,
sets,
keys,
stakes,
total_stake,
};
GlobalSessions::set(&mut txn, global_session, &global_session_info);
if let Some(ending_global_session) = global_session_for_this_block {
GlobalSessionsLastBlock::set(&mut txn, ending_global_session, &block_number);
}
LatestGlobalSessionIntended::set(&mut txn, &global_session);
GlobalSessionsChannel::send(&mut txn, &(global_session, global_session_info));
}
// If there isn't anyone available to cosign this block, meaning it'll never be cosigned,
// we flag it as not having any events requiring cosigning so we don't attempt to
// sign/require a cosign for it
if global_session_for_this_block.is_none() {
has_events = HasEvents::No;
}
match has_events {
HasEvents::Notable | HasEvents::NonNotable => {
let global_session_for_this_block = global_session_for_this_block
.expect("global session for this block was None but still attempting to cosign it");
let global_session_info = GlobalSessions::get(&txn, global_session_for_this_block)
.expect("last global session intended wasn't saved to the database");
// Tell each set of their expectation to cosign this block
for set in global_session_info.sets {
log::debug!("{:?} will be cosigning block #{block_number}", set);
IntendedCosigns::send(
&mut txn,
set,
&CosignIntent {
global_session: global_session_for_this_block,
block_number,
block_hash,
notable: has_events == HasEvents::Notable,
},
);
}
}
HasEvents::No => {}
}
// Populate a singular feed with every block's status for the evluator to work off of
BlockEvents::send(&mut txn, &(BlockEventData { block_number, has_events }));
// Mark this block as handled, meaning we should scan from the next block moving on
ScanCosignFrom::set(&mut txn, &(block_number + 1));
txn.commit();
}
Ok(start_block_number <= latest_block_number)
}
}
}

View File

@@ -0,0 +1,509 @@
#![cfg_attr(docsrs, feature(doc_auto_cfg))]
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
use core::{fmt::Debug, future::Future};
use std::{sync::Arc, collections::HashMap};
use blake2::{Digest, Blake2s256};
use scale::{Encode, Decode};
use borsh::{BorshSerialize, BorshDeserialize};
use serai_client::{
primitives::{NetworkId, SeraiAddress},
validator_sets::primitives::{Session, ValidatorSet, KeyPair},
Public, Block, Serai, TemporalSerai,
};
use serai_db::*;
use serai_task::*;
/// The cosigns which are intended to be performed.
mod intend;
/// The evaluator of the cosigns.
mod evaluator;
/// The task to delay acknowledgement of the cosigns.
mod delay;
pub use delay::BROADCAST_FREQUENCY;
use delay::LatestCosignedBlockNumber;
/// The schnorrkel context to used when signing a cosign.
pub const COSIGN_CONTEXT: &[u8] = b"/serai/coordinator/cosign";
/// A 'global session', defined as all validator sets used for cosigning at a given moment.
///
/// We evaluate cosign faults within a global session. This ensures even if cosigners cosign
/// distinct blocks at distinct positions within a global session, we still identify the faults.
/*
There is the attack where a validator set is given an alternate blockchain with a key generation
event at block #n, while most validator sets are given a blockchain with a key generation event
at block number #(n+1). This prevents whoever has the alternate blockchain from verifying the
cosigns on the primary blockchain, and detecting the faults, if they use the keys as of the block
prior to the block being cosigned.
We solve this by binding cosigns to a global session ID, which has a specific start block, and
reading the keys from the start block. This means that so long as all validator sets agree on the
start of a global session, they can verify all cosigns produced by that session, regardless of
how it advances. Since agreeing on the start of a global session is mandated, there's no way to
have validator sets follow two distinct global sessions without breaking the bounds of the
cosigning protocol.
*/
#[derive(Debug, BorshSerialize, BorshDeserialize)]
pub(crate) struct GlobalSession {
pub(crate) start_block_number: u64,
pub(crate) sets: Vec<ValidatorSet>,
pub(crate) keys: HashMap<NetworkId, SeraiAddress>,
pub(crate) stakes: HashMap<NetworkId, u64>,
pub(crate) total_stake: u64,
}
impl GlobalSession {
fn id(mut cosigners: Vec<ValidatorSet>) -> [u8; 32] {
cosigners.sort_by_key(|a| borsh::to_vec(a).unwrap());
Blake2s256::digest(borsh::to_vec(&cosigners).unwrap()).into()
}
}
/// If the block has events.
#[derive(Clone, Copy, PartialEq, Eq, Debug, BorshSerialize, BorshDeserialize)]
enum HasEvents {
/// The block had a notable event.
///
/// This is a special case as blocks with key gen events change the keys used for cosigning, and
/// accordingly must be cosigned before we advance past them.
Notable,
/// The block had an non-notable event justifying a cosign.
NonNotable,
/// The block didn't have an event justifying a cosign.
No,
}
/// An intended cosign.
#[derive(Clone, Copy, PartialEq, Eq, Debug, BorshSerialize, BorshDeserialize)]
pub struct CosignIntent {
/// The global session this cosign is being performed under.
pub global_session: [u8; 32],
/// The number of the block to cosign.
pub block_number: u64,
/// The hash of the block to cosign.
pub block_hash: [u8; 32],
/// If this cosign must be handled before further cosigns are.
pub notable: bool,
}
/// A cosign.
#[derive(Clone, PartialEq, Eq, Debug, Encode, Decode, BorshSerialize, BorshDeserialize)]
pub struct Cosign {
/// The global session this cosign is being performed under.
pub global_session: [u8; 32],
/// The number of the block to cosign.
pub block_number: u64,
/// The hash of the block to cosign.
pub block_hash: [u8; 32],
/// The actual cosigner.
pub cosigner: NetworkId,
}
/// A signed cosign.
#[derive(Clone, Debug, BorshSerialize, BorshDeserialize)]
pub struct SignedCosign {
/// The cosign.
pub cosign: Cosign,
/// The signature for the cosign.
pub signature: [u8; 64],
}
impl SignedCosign {
fn verify_signature(&self, signer: serai_client::Public) -> bool {
let Ok(signer) = schnorrkel::PublicKey::from_bytes(&signer.0) else { return false };
let Ok(signature) = schnorrkel::Signature::from_bytes(&self.signature) else { return false };
signer.verify_simple(COSIGN_CONTEXT, &self.cosign.encode(), &signature).is_ok()
}
}
create_db! {
Cosign {
// The following are populated by the intend task and used throughout the library
// An index of Substrate blocks
SubstrateBlockHash: (block_number: u64) -> [u8; 32],
// A mapping from a global session's ID to its relevant information.
GlobalSessions: (global_session: [u8; 32]) -> GlobalSession,
// The last block to be cosigned by a global session.
GlobalSessionsLastBlock: (global_session: [u8; 32]) -> u64,
// The latest global session intended.
//
// This is distinct from the latest global session for which we've evaluated the cosigns for.
LatestGlobalSessionIntended: () -> [u8; 32],
// The following are managed by the `intake_cosign` function present in this file
// The latest cosigned block for each network.
//
// This will only be populated with cosigns predating or during the most recent global session
// to have its start cosigned.
//
// The global session changes upon a notable block, causing each global session to have exactly
// one notable block. All validator sets will explicitly produce a cosign for their notable
// block, causing the latest cosigned block for a global session to either be the global
// session's notable cosigns or the network's latest cosigns.
NetworksLatestCosignedBlock: (global_session: [u8; 32], network: NetworkId) -> SignedCosign,
// Cosigns received for blocks not locally recognized as finalized.
Faults: (global_session: [u8; 32]) -> Vec<SignedCosign>,
// The global session which faulted.
FaultedSession: () -> [u8; 32],
}
}
/// Fetch the keys used for cosigning by a specific network.
async fn keys_for_network(
serai: &TemporalSerai<'_>,
network: NetworkId,
) -> Result<Option<(Session, KeyPair)>, String> {
// The Serai network never cosigns so it has no keys for cosigning
if network == NetworkId::Serai {
return Ok(None);
}
let Some(latest_session) =
serai.validator_sets().session(network).await.map_err(|e| format!("{e:?}"))?
else {
// If this network hasn't had a session declared, move on
return Ok(None);
};
// Get the keys for the latest session
if let Some(keys) = serai
.validator_sets()
.keys(ValidatorSet { network, session: latest_session })
.await
.map_err(|e| format!("{e:?}"))?
{
return Ok(Some((latest_session, keys)));
}
// If the latest session has yet to set keys, use the prior session
if let Some(prior_session) = latest_session.0.checked_sub(1).map(Session) {
if let Some(keys) = serai
.validator_sets()
.keys(ValidatorSet { network, session: prior_session })
.await
.map_err(|e| format!("{e:?}"))?
{
return Ok(Some((prior_session, keys)));
}
}
Ok(None)
}
/// Fetch the `ValidatorSet`s, and their associated keys, used for cosigning as of this block.
async fn cosigning_sets(serai: &TemporalSerai<'_>) -> Result<Vec<(ValidatorSet, Public)>, String> {
let mut sets = Vec::with_capacity(serai_client::primitives::NETWORKS.len());
for network in serai_client::primitives::NETWORKS {
let Some((session, keys)) = keys_for_network(serai, network).await? else {
// If this network doesn't have usable keys, move on
continue;
};
sets.push((ValidatorSet { network, session }, keys.0));
}
Ok(sets)
}
/// An object usable to request notable cosigns for a block.
pub trait RequestNotableCosigns: 'static + Send {
/// The error type which may be encountered when requesting notable cosigns.
type Error: Debug;
/// Request the notable cosigns for this global session.
fn request_notable_cosigns(
&self,
global_session: [u8; 32],
) -> impl Send + Future<Output = Result<(), Self::Error>>;
}
/// An error used to indicate the cosigning protocol has faulted.
#[derive(Debug)]
pub struct Faulted;
/// An error incurred while intaking a cosign.
#[derive(Debug)]
pub enum IntakeCosignError {
/// Cosign is for a not-yet-indexed block
NotYetIndexedBlock,
/// A later cosign for this cosigner has already been handled
StaleCosign,
/// The cosign's global session isn't recognized
UnrecognizedGlobalSession,
/// The cosign is for a block before its global session starts
BeforeGlobalSessionStart,
/// The cosign is for a block after its global session ends
AfterGlobalSessionEnd,
/// The cosign's signing network wasn't a participant in this global session
NonParticipatingNetwork,
/// The cosign had an invalid signature
InvalidSignature,
/// The cosign is for a global session which has yet to have its declaration block cosigned
FutureGlobalSession,
}
impl IntakeCosignError {
/// If this error is temporal to the local view
pub fn temporal(&self) -> bool {
match self {
IntakeCosignError::NotYetIndexedBlock |
IntakeCosignError::StaleCosign |
IntakeCosignError::UnrecognizedGlobalSession |
IntakeCosignError::FutureGlobalSession => true,
IntakeCosignError::BeforeGlobalSessionStart |
IntakeCosignError::AfterGlobalSessionEnd |
IntakeCosignError::NonParticipatingNetwork |
IntakeCosignError::InvalidSignature => false,
}
}
}
/// The interface to manage cosigning with.
pub struct Cosigning<D: Db> {
db: D,
}
impl<D: Db> Cosigning<D> {
/// Spawn the tasks to intend and evaluate cosigns.
///
/// The database specified must only be used with a singular instance of the Serai network, and
/// only used once at any given time.
pub fn spawn<R: RequestNotableCosigns>(
db: D,
serai: Arc<Serai>,
request: R,
tasks_to_run_upon_cosigning: Vec<TaskHandle>,
) -> Self {
let (intend_task, _intend_task_handle) = Task::new();
let (evaluator_task, evaluator_task_handle) = Task::new();
let (delay_task, delay_task_handle) = Task::new();
tokio::spawn(
(intend::CosignIntendTask { db: db.clone(), serai })
.continually_run(intend_task, vec![evaluator_task_handle]),
);
tokio::spawn(
(evaluator::CosignEvaluatorTask { db: db.clone(), request })
.continually_run(evaluator_task, vec![delay_task_handle]),
);
tokio::spawn(
(delay::CosignDelayTask { db: db.clone() })
.continually_run(delay_task, tasks_to_run_upon_cosigning),
);
Self { db }
}
/// The latest cosigned block number.
pub fn latest_cosigned_block_number(getter: &impl Get) -> Result<u64, Faulted> {
if FaultedSession::get(getter).is_some() {
Err(Faulted)?;
}
Ok(LatestCosignedBlockNumber::get(getter).unwrap_or(0))
}
/// Fetch a cosigned Substrate block's hash by its block number.
pub fn cosigned_block(getter: &impl Get, block_number: u64) -> Result<Option<[u8; 32]>, Faulted> {
if block_number > Self::latest_cosigned_block_number(getter)? {
return Ok(None);
}
Ok(Some(
SubstrateBlockHash::get(getter, block_number).expect("cosigned block but didn't index it"),
))
}
/// Fetch the notable cosigns for a global session in order to respond to requests.
///
/// If this global session hasn't produced any notable cosigns, this will return the latest
/// cosigns for this session.
pub fn notable_cosigns(getter: &impl Get, global_session: [u8; 32]) -> Vec<SignedCosign> {
let mut cosigns = Vec::with_capacity(serai_client::primitives::NETWORKS.len());
for network in serai_client::primitives::NETWORKS {
if let Some(cosign) = NetworksLatestCosignedBlock::get(getter, global_session, network) {
cosigns.push(cosign);
}
}
cosigns
}
/// The cosigns to rebroadcast every `BROADCAST_FREQUENCY` seconds.
///
/// This will be the most recent cosigns, in case the initial broadcast failed, or the faulty
/// cosigns, in case of a fault, to induce identification of the fault by others.
pub fn cosigns_to_rebroadcast(&self) -> Vec<SignedCosign> {
if let Some(faulted) = FaultedSession::get(&self.db) {
let mut cosigns = Faults::get(&self.db, faulted).expect("faulted with no faults");
// Also include all of our recognized-as-honest cosigns in an attempt to induce fault
// identification in those who see the faulty cosigns as honest
for network in serai_client::primitives::NETWORKS {
if let Some(cosign) = NetworksLatestCosignedBlock::get(&self.db, faulted, network) {
if cosign.cosign.global_session == faulted {
cosigns.push(cosign);
}
}
}
cosigns
} else {
let Some(global_session) = evaluator::currently_evaluated_global_session(&self.db) else {
return vec![];
};
let mut cosigns = Vec::with_capacity(serai_client::primitives::NETWORKS.len());
for network in serai_client::primitives::NETWORKS {
if let Some(cosign) = NetworksLatestCosignedBlock::get(&self.db, global_session, network) {
cosigns.push(cosign);
}
}
cosigns
}
}
/// Intake a cosign.
//
// Takes `&mut self` as this should only be called once at any given moment.
pub fn intake_cosign(&mut self, signed_cosign: &SignedCosign) -> Result<(), IntakeCosignError> {
let cosign = &signed_cosign.cosign;
let network = cosign.cosigner;
// Check our indexed blockchain includes a block with this block number
let Some(our_block_hash) = SubstrateBlockHash::get(&self.db, cosign.block_number) else {
Err(IntakeCosignError::NotYetIndexedBlock)?
};
let faulty = cosign.block_hash != our_block_hash;
// Check this isn't a dated cosign within its global session (as it would be if rebroadcasted)
if !faulty {
if let Some(existing) =
NetworksLatestCosignedBlock::get(&self.db, cosign.global_session, network)
{
if existing.cosign.block_number >= cosign.block_number {
Err(IntakeCosignError::StaleCosign)?;
}
}
}
let Some(global_session) = GlobalSessions::get(&self.db, cosign.global_session) else {
Err(IntakeCosignError::UnrecognizedGlobalSession)?
};
// Check the cosigned block number is in range to the global session
if cosign.block_number < global_session.start_block_number {
// Cosign is for a block predating the global session
Err(IntakeCosignError::BeforeGlobalSessionStart)?;
}
if !faulty {
// This prevents a malicious validator set, on the same chain, from producing a cosign after
// their final block, replacing their notable cosign
if let Some(last_block) = GlobalSessionsLastBlock::get(&self.db, cosign.global_session) {
if cosign.block_number > last_block {
// Cosign is for a block after the last block this global session should have signed
Err(IntakeCosignError::AfterGlobalSessionEnd)?;
}
}
}
// Check the cosign's signature
{
let key = Public::from({
let Some(key) = global_session.keys.get(&network) else {
Err(IntakeCosignError::NonParticipatingNetwork)?
};
*key
});
if !signed_cosign.verify_signature(key) {
Err(IntakeCosignError::InvalidSignature)?;
}
}
// Since we verified this cosign's signature, and have a chain sufficiently long, handle the
// cosign
let mut txn = self.db.unsafe_txn();
if !faulty {
// If this is for a future global session, we don't acknowledge this cosign at this time
let latest_cosigned_block_number = LatestCosignedBlockNumber::get(&txn).unwrap_or(0);
// This global session starts the block *after* its declaration, so we want to check if the
// block declaring it was cosigned
if (global_session.start_block_number - 1) > latest_cosigned_block_number {
drop(txn);
return Err(IntakeCosignError::FutureGlobalSession);
}
// This is safe as it's in-range and newer, as prior checked since it isn't faulty
NetworksLatestCosignedBlock::set(&mut txn, cosign.global_session, network, signed_cosign);
} else {
let mut faults = Faults::get(&txn, cosign.global_session).unwrap_or(vec![]);
// Only handle this as a fault if this set wasn't prior faulty
if !faults.iter().any(|cosign| cosign.cosign.cosigner == network) {
faults.push(signed_cosign.clone());
Faults::set(&mut txn, cosign.global_session, &faults);
let mut weight_cosigned = 0;
for fault in &faults {
let stake = global_session
.stakes
.get(&fault.cosign.cosigner)
.expect("cosigner with recognized key didn't have a stake entry saved");
weight_cosigned += stake;
}
// Check if the sum weight means a fault has occurred
if weight_cosigned >= ((global_session.total_stake * 17) / 100) {
FaultedSession::set(&mut txn, &cosign.global_session);
}
}
}
txn.commit();
Ok(())
}
/// Receive intended cosigns to produce for this ValidatorSet.
///
/// All cosigns intended, up to and including the next notable cosign, are returned.
///
/// This will drain the internal channel and not re-yield these intentions again.
pub fn intended_cosigns(txn: &mut impl DbTxn, set: ValidatorSet) -> Vec<CosignIntent> {
let mut res: Vec<CosignIntent> = vec![];
// While we have yet to find a notable cosign...
while !res.last().map(|cosign| cosign.notable).unwrap_or(false) {
let Some(intent) = intend::IntendedCosigns::try_recv(txn, set) else { break };
res.push(intent);
}
res
}
}
mod tests {
use super::*;
struct RNC;
impl RequestNotableCosigns for RNC {
/// The error type which may be encountered when requesting notable cosigns.
type Error = ();
/// Request the notable cosigns for this global session.
fn request_notable_cosigns(
&self,
global_session: [u8; 32],
) -> impl Send + Future<Output = Result<(), Self::Error>> {
async move { Ok(()) }
}
}
#[tokio::test]
async fn test() {
let db: serai_db::MemDb = serai_db::MemDb::new();
let serai = unsafe { core::mem::transmute(0u64) };
let request = RNC;
let tasks = vec![];
let _ = Cosigning::spawn(db, serai, request, tasks);
core::future::pending().await
}
}

View File

@@ -0,0 +1,33 @@
[package]
name = "serai-coordinator-p2p"
version = "0.1.0"
description = "Serai coordinator's P2P abstraction"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/coordinator/p2p"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
borsh = { version = "1", default-features = false, features = ["std", "derive", "de_strict_order"] }
serai-db = { path = "../../common/db", version = "0.1" }
serai-client = { path = "../../substrate/client", default-features = false, features = ["serai", "borsh"] }
serai-cosign = { path = "../cosign" }
tributary-sdk = { path = "../tributary-sdk" }
futures-lite = { version = "2", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false, features = ["sync", "macros"] }
log = { version = "0.4", default-features = false, features = ["std"] }
serai-task = { path = "../../common/task", version = "0.1" }

15
coordinator/p2p/LICENSE Normal file
View File

@@ -0,0 +1,15 @@
AGPL-3.0-only license
Copyright (c) 2023-2025 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.

View File

@@ -0,0 +1,3 @@
# Serai Coordinator P2P
The P2P abstraction used by Serai's coordinator, and tasks over it.

View File

@@ -0,0 +1,42 @@
[package]
name = "serai-coordinator-libp2p-p2p"
version = "0.1.0"
description = "Serai coordinator's libp2p-based P2P backend"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/coordinator/p2p/libp2p"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
async-trait = { version = "0.1", default-features = false }
rand_core = { version = "0.6", default-features = false, features = ["std"] }
zeroize = { version = "^1.5", default-features = false, features = ["std"] }
blake2 = { version = "0.10", default-features = false, features = ["std"] }
schnorrkel = { version = "0.11", default-features = false, features = ["std"] }
hex = { version = "0.4", default-features = false, features = ["std"] }
borsh = { version = "1", default-features = false, features = ["std", "derive", "de_strict_order"] }
serai-client = { path = "../../../substrate/client", default-features = false, features = ["serai", "borsh"] }
serai-cosign = { path = "../../cosign" }
tributary-sdk = { path = "../../tributary-sdk" }
futures-util = { version = "0.3", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false, features = ["sync"] }
libp2p = { version = "0.52", default-features = false, features = ["tokio", "tcp", "noise", "yamux", "ping", "request-response", "gossipsub", "macros"] }
log = { version = "0.4", default-features = false, features = ["std"] }
serai-task = { path = "../../../common/task", version = "0.1" }
serai-coordinator-p2p = { path = "../" }

View File

@@ -0,0 +1,15 @@
AGPL-3.0-only license
Copyright (c) 2023-2025 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.

View File

@@ -0,0 +1,14 @@
# Serai Coordinator libp2p P2P
A libp2p-backed P2P instantiation for Serai's coordinator.
The libp2p swarm is limited to validators from the Serai network. The swarm
does not maintain any of its own peer finding/routing infrastructure, instead
relying on the Serai network's connection information to dial peers. This does
limit the listening peers to only the peers immediately reachable via the same
IP address (despite the two distinct services), not hidden behind a NAT, yet is
also quite simple and gives full control of who to connect to to us.
Peers are decided via the internal `DialTask` which aims to maintain a target
amount of peers for each external network. This ensures cosigns are able to
propagate across the external networks which sign them.

View File

@@ -0,0 +1,176 @@
use core::{pin::Pin, future::Future};
use std::io;
use zeroize::Zeroizing;
use rand_core::{RngCore, OsRng};
use blake2::{Digest, Blake2s256};
use schnorrkel::{Keypair, PublicKey, Signature};
use serai_client::primitives::PublicKey as Public;
use futures_util::{AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt};
use libp2p::{
core::UpgradeInfo,
InboundUpgrade, OutboundUpgrade,
identity::{self, PeerId},
noise,
};
use crate::peer_id_from_public;
const PROTOCOL: &str = "/serai/coordinator/validators";
#[derive(Clone)]
pub(crate) struct OnlyValidators {
pub(crate) serai_key: Zeroizing<Keypair>,
pub(crate) noise_keypair: identity::Keypair,
}
impl OnlyValidators {
/// The ephemeral challenge protocol for authentication.
///
/// We use ephemeral challenges to prevent replaying signatures from historic sessions.
///
/// We don't immediately send the challenge. We only send a commitment to it. This prevents our
/// remote peer from choosing their challenge in response to our challenge, in case there was any
/// benefit to doing so.
async fn challenges<S: 'static + Send + Unpin + AsyncRead + AsyncWrite>(
socket: &mut noise::Output<S>,
) -> io::Result<([u8; 32], [u8; 32])> {
let mut our_challenge = [0; 32];
OsRng.fill_bytes(&mut our_challenge);
// Write the hash of our challenge
socket.write_all(&Blake2s256::digest(our_challenge)).await?;
// Read the hash of their challenge
let mut their_challenge_commitment = [0; 32];
socket.read_exact(&mut their_challenge_commitment).await?;
// Reveal our challenge
socket.write_all(&our_challenge).await?;
// Read their challenge
let mut their_challenge = [0; 32];
socket.read_exact(&mut their_challenge).await?;
// Verify their challenge
if <[u8; 32]>::from(Blake2s256::digest(their_challenge)) != their_challenge_commitment {
Err(io::Error::other("challenge didn't match challenge commitment"))?;
}
Ok((our_challenge, their_challenge))
}
// We sign the two noise peer IDs and the ephemeral challenges.
//
// Signing the noise peer IDs ensures we're authenticating this noise connection. The only
// expectations placed on noise are for it to prevent a MITM from impersonating the other end or
// modifying any messages sent.
//
// Signing the ephemeral challenges prevents any replays. While that should be unnecessary, as
// noise MAY prevent replays across sessions (even when the same key is used), and noise IDs
// shouldn't be reused (so it should be fine to reuse an existing signature for these noise IDs),
// it doesn't hurt.
async fn authenticate<S: 'static + Send + Unpin + AsyncRead + AsyncWrite>(
&self,
socket: &mut noise::Output<S>,
dialer_peer_id: PeerId,
dialer_challenge: [u8; 32],
listener_peer_id: PeerId,
listener_challenge: [u8; 32],
) -> io::Result<PeerId> {
// Write our public key
socket.write_all(&self.serai_key.public.to_bytes()).await?;
let msg = borsh::to_vec(&(
dialer_peer_id.to_bytes(),
dialer_challenge,
listener_peer_id.to_bytes(),
listener_challenge,
))
.unwrap();
let signature = self.serai_key.sign_simple(PROTOCOL.as_bytes(), &msg);
socket.write_all(&signature.to_bytes()).await?;
let mut public_key_and_sig = [0; 96];
socket.read_exact(&mut public_key_and_sig).await?;
let public_key = PublicKey::from_bytes(&public_key_and_sig[.. 32])
.map_err(|_| io::Error::other("invalid public key"))?;
let sig = Signature::from_bytes(&public_key_and_sig[32 ..])
.map_err(|_| io::Error::other("invalid signature serialization"))?;
public_key
.verify_simple(PROTOCOL.as_bytes(), &msg, &sig)
.map_err(|_| io::Error::other("invalid signature"))?;
Ok(peer_id_from_public(Public::from_raw(public_key.to_bytes())))
}
}
impl UpgradeInfo for OnlyValidators {
type Info = <noise::Config as UpgradeInfo>::Info;
type InfoIter = <noise::Config as UpgradeInfo>::InfoIter;
fn protocol_info(&self) -> Self::InfoIter {
// A keypair only causes an error if its sign operation fails, which is only possible with RSA,
// which isn't used within this codebase
noise::Config::new(&self.noise_keypair).unwrap().protocol_info()
}
}
impl<S: 'static + Send + Unpin + AsyncRead + AsyncWrite> InboundUpgrade<S> for OnlyValidators {
type Output = (PeerId, noise::Output<S>);
type Error = io::Error;
type Future = Pin<Box<dyn Send + Future<Output = Result<Self::Output, Self::Error>>>>;
fn upgrade_inbound(self, socket: S, info: Self::Info) -> Self::Future {
Box::pin(async move {
let (dialer_noise_peer_id, mut socket) = noise::Config::new(&self.noise_keypair)
.unwrap()
.upgrade_inbound(socket, info)
.await
.map_err(io::Error::other)?;
let (our_challenge, dialer_challenge) = OnlyValidators::challenges(&mut socket).await?;
let dialer_serai_validator = self
.authenticate(
&mut socket,
dialer_noise_peer_id,
dialer_challenge,
PeerId::from_public_key(&self.noise_keypair.public()),
our_challenge,
)
.await?;
Ok((dialer_serai_validator, socket))
})
}
}
impl<S: 'static + Send + Unpin + AsyncRead + AsyncWrite> OutboundUpgrade<S> for OnlyValidators {
type Output = (PeerId, noise::Output<S>);
type Error = io::Error;
type Future = Pin<Box<dyn Send + Future<Output = Result<Self::Output, Self::Error>>>>;
fn upgrade_outbound(self, socket: S, info: Self::Info) -> Self::Future {
Box::pin(async move {
let (listener_noise_peer_id, mut socket) = noise::Config::new(&self.noise_keypair)
.unwrap()
.upgrade_outbound(socket, info)
.await
.map_err(io::Error::other)?;
let (our_challenge, listener_challenge) = OnlyValidators::challenges(&mut socket).await?;
let listener_serai_validator = self
.authenticate(
&mut socket,
PeerId::from_public_key(&self.noise_keypair.public()),
our_challenge,
listener_noise_peer_id,
listener_challenge,
)
.await?;
Ok((listener_serai_validator, socket))
})
}
}

View File

@@ -0,0 +1,127 @@
use core::future::Future;
use std::{sync::Arc, collections::HashSet};
use rand_core::{RngCore, OsRng};
use tokio::sync::mpsc;
use serai_client::{SeraiError, Serai};
use libp2p::{
core::multiaddr::{Protocol, Multiaddr},
swarm::dial_opts::DialOpts,
};
use serai_task::ContinuallyRan;
use crate::{PORT, Peers, validators::Validators};
const TARGET_PEERS_PER_NETWORK: usize = 5;
/*
If we only tracked the target amount of peers per network, we'd risk being eclipsed by an
adversary who immediately connects to us with their array of validators upon our boot. Their
array would satisfy our target amount of peers, so we'd never seek more, enabling the adversary
to be the only entity we peered with.
We solve this by additionally requiring an explicit amount of peers we dialed. That means we
randomly chose to connect to these peers.
*/
// TODO const TARGET_DIALED_PEERS_PER_NETWORK: usize = 3;
pub(crate) struct DialTask {
serai: Arc<Serai>,
validators: Validators,
peers: Peers,
to_dial: mpsc::UnboundedSender<DialOpts>,
}
impl DialTask {
pub(crate) fn new(
serai: Arc<Serai>,
peers: Peers,
to_dial: mpsc::UnboundedSender<DialOpts>,
) -> Self {
DialTask { serai: serai.clone(), validators: Validators::new(serai).0, peers, to_dial }
}
}
impl ContinuallyRan for DialTask {
// Only run every five minutes, not the default of every five seconds
const DELAY_BETWEEN_ITERATIONS: u64 = 5 * 60;
const MAX_DELAY_BETWEEN_ITERATIONS: u64 = 10 * 60;
type Error = SeraiError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
self.validators.update().await?;
// If any of our peers is lacking, try to connect to more
let mut dialed = false;
let peer_counts = self
.peers
.peers
.read()
.await
.iter()
.map(|(network, peers)| (*network, peers.len()))
.collect::<Vec<_>>();
for (network, peer_count) in peer_counts {
/*
If we don't have the target amount of peers, and we don't have all the validators in the
set but one, attempt to connect to more validators within this set.
The latter clause is so if there's a set with only 3 validators, we don't infinitely try
to connect to the target amount of peers for this network as we never will. Instead, we
only try to connect to most of the validators actually present.
*/
if (peer_count < TARGET_PEERS_PER_NETWORK) &&
(peer_count <
self
.validators
.by_network()
.get(&network)
.map(HashSet::len)
.unwrap_or(0)
.saturating_sub(1))
{
let mut potential_peers = self.serai.p2p_validators(network).await?;
for _ in 0 .. (TARGET_PEERS_PER_NETWORK - peer_count) {
if potential_peers.is_empty() {
break;
}
let index_to_dial =
usize::try_from(OsRng.next_u64() % u64::try_from(potential_peers.len()).unwrap())
.unwrap();
let randomly_selected_peer = potential_peers.swap_remove(index_to_dial);
log::info!("found peer from substrate: {randomly_selected_peer}");
// Map the peer from a Substrate P2P network peer to a Coordinator P2P network peer
let mapped_peer = randomly_selected_peer
.into_iter()
.filter_map(|protocol| match protocol {
// Drop PeerIds from the Substrate P2p network
Protocol::P2p(_) => None,
// Use our own TCP port
Protocol::Tcp(_) => Some(Protocol::Tcp(PORT)),
// Pass-through any other specifications (IPv4, IPv6, etc)
other => Some(other),
})
.collect::<Multiaddr>();
log::debug!("mapped found peer: {mapped_peer}");
self
.to_dial
.send(DialOpts::unknown_peer_id().address(mapped_peer).build())
.expect("dial receiver closed?");
dialed = true;
}
}
}
Ok(dialed)
}
}
}

View File

@@ -0,0 +1,75 @@
use core::time::Duration;
use blake2::{Digest, Blake2s256};
use borsh::{BorshSerialize, BorshDeserialize};
use libp2p::gossipsub::{
IdentTopic, MessageId, MessageAuthenticity, ValidationMode, ConfigBuilder, IdentityTransform,
AllowAllSubscriptionFilter, Behaviour,
};
pub use libp2p::gossipsub::Event;
use serai_cosign::SignedCosign;
// Block size limit + 16 KB of space for signatures/metadata
pub(crate) const MAX_LIBP2P_GOSSIP_MESSAGE_SIZE: usize = tributary_sdk::BLOCK_SIZE_LIMIT + 16384;
const LIBP2P_PROTOCOL: &str = "/serai/coordinator/gossip/1.0.0";
const BASE_TOPIC: &str = "/";
fn topic_for_tributary(tributary: [u8; 32]) -> IdentTopic {
IdentTopic::new(format!("/tributary/{}", hex::encode(tributary)))
}
#[derive(Clone, BorshSerialize, BorshDeserialize)]
pub(crate) enum Message {
Tributary { tributary: [u8; 32], message: Vec<u8> },
Cosign(SignedCosign),
}
impl Message {
pub(crate) fn topic(&self) -> IdentTopic {
match self {
Message::Tributary { tributary, .. } => topic_for_tributary(*tributary),
Message::Cosign(_) => IdentTopic::new(BASE_TOPIC),
}
}
}
pub(crate) type Behavior = Behaviour<IdentityTransform, AllowAllSubscriptionFilter>;
pub(crate) fn new_behavior() -> Behavior {
// The latency used by the Tendermint protocol, used here as the gossip epoch duration
// libp2p-rs defaults to 1 second, whereas ours will be ~2
let heartbeat_interval = tributary_sdk::tendermint::LATENCY_TIME;
// The amount of heartbeats which will occur within a single Tributary block
let heartbeats_per_block =
tributary_sdk::tendermint::TARGET_BLOCK_TIME.div_ceil(heartbeat_interval);
// libp2p-rs defaults to 5, whereas ours will be ~8
let heartbeats_to_keep = 2 * heartbeats_per_block;
// libp2p-rs defaults to 3 whereas ours will be ~4
let heartbeats_to_gossip = heartbeats_per_block;
let config = ConfigBuilder::default()
.protocol_id_prefix(LIBP2P_PROTOCOL)
.history_length(usize::try_from(heartbeats_to_keep).unwrap())
.history_gossip(usize::try_from(heartbeats_to_gossip).unwrap())
.heartbeat_interval(Duration::from_millis(heartbeat_interval.into()))
.max_transmit_size(MAX_LIBP2P_GOSSIP_MESSAGE_SIZE)
.duplicate_cache_time(Duration::from_millis((heartbeats_to_keep * heartbeat_interval).into()))
.validation_mode(ValidationMode::Anonymous)
// Uses a content based message ID to avoid duplicates as much as possible
.message_id_fn(|msg| {
MessageId::new(&Blake2s256::digest([msg.topic.as_str().as_bytes(), &msg.data].concat()))
})
.build();
let mut gossip = Behavior::new(MessageAuthenticity::Anonymous, config.unwrap()).unwrap();
// Subscribe to the base topic
let topic = IdentTopic::new(BASE_TOPIC);
let _ = gossip.subscribe(&topic);
gossip
}

View File

@@ -0,0 +1,433 @@
#![cfg_attr(docsrs, feature(doc_auto_cfg))]
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
use core::{future::Future, time::Duration};
use std::{
sync::Arc,
collections::{HashSet, HashMap},
};
use rand_core::{RngCore, OsRng};
use zeroize::Zeroizing;
use schnorrkel::Keypair;
use serai_client::{
primitives::{NetworkId, PublicKey},
validator_sets::primitives::ValidatorSet,
Serai,
};
use tokio::sync::{mpsc, oneshot, Mutex, RwLock};
use serai_task::{Task, ContinuallyRan};
use serai_cosign::SignedCosign;
use libp2p::{
multihash::Multihash,
identity::{self, PeerId},
tcp::Config as TcpConfig,
yamux, allow_block_list,
connection_limits::{self, ConnectionLimits},
swarm::NetworkBehaviour,
SwarmBuilder,
};
use serai_coordinator_p2p::{Heartbeat, TributaryBlockWithCommit};
/// A struct to sync the validators from the Serai node in order to keep track of them.
mod validators;
use validators::UpdateValidatorsTask;
/// The authentication protocol upgrade to limit the P2P network to active validators.
mod authenticate;
use authenticate::OnlyValidators;
/// The ping behavior, used to ensure connection latency is below the limit
mod ping;
/// The request-response messages and behavior
mod reqres;
use reqres::{RequestId, Request, Response};
/// The gossip messages and behavior
mod gossip;
use gossip::Message;
/// The swarm task, running it and dispatching to/from it
mod swarm;
use swarm::SwarmTask;
/// The dial task, to find new peers to connect to
mod dial;
use dial::DialTask;
const PORT: u16 = 30563; // 5132 ^ (('c' << 8) | 'o')
// usize::max, manually implemented, as max isn't a const fn
const MAX_LIBP2P_MESSAGE_SIZE: usize =
if gossip::MAX_LIBP2P_GOSSIP_MESSAGE_SIZE > reqres::MAX_LIBP2P_REQRES_MESSAGE_SIZE {
gossip::MAX_LIBP2P_GOSSIP_MESSAGE_SIZE
} else {
reqres::MAX_LIBP2P_REQRES_MESSAGE_SIZE
};
fn peer_id_from_public(public: PublicKey) -> PeerId {
// 0 represents the identity Multihash, that no hash was performed
// It's an internal constant so we can't refer to the constant inside libp2p
PeerId::from_multihash(Multihash::wrap(0, &public.0).unwrap()).unwrap()
}
/// The representation of a peer.
pub struct Peer<'a> {
outbound_requests: &'a mpsc::UnboundedSender<(PeerId, Request, oneshot::Sender<Response>)>,
id: PeerId,
}
impl serai_coordinator_p2p::Peer<'_> for Peer<'_> {
fn send_heartbeat(
&self,
heartbeat: Heartbeat,
) -> impl Send + Future<Output = Option<Vec<TributaryBlockWithCommit>>> {
async move {
const HEARTBEAT_TIMEOUT: Duration = Duration::from_secs(5);
let request = Request::Heartbeat(heartbeat);
let (sender, receiver) = oneshot::channel();
self
.outbound_requests
.send((self.id, request, sender))
.expect("outbound requests recv channel was dropped?");
if let Ok(Ok(Response::Blocks(blocks))) =
tokio::time::timeout(HEARTBEAT_TIMEOUT, receiver).await
{
Some(blocks)
} else {
None
}
}
}
}
#[derive(Clone)]
struct Peers {
peers: Arc<RwLock<HashMap<NetworkId, HashSet<PeerId>>>>,
}
// Consider adding identify/kad/autonat/rendevous/(relay + dcutr). While we currently use the Serai
// network for peers, we could use it solely for bootstrapping/as a fallback.
#[derive(NetworkBehaviour)]
struct Behavior {
// Used to only allow Serai validators as peers
allow_list: allow_block_list::Behaviour<allow_block_list::AllowedPeers>,
// Used to limit each peer to a single connection
connection_limits: connection_limits::Behaviour,
// Used to ensure connection latency is within tolerances
ping: ping::Behavior,
// Used to request data from specific peers
reqres: reqres::Behavior,
// Used to broadcast messages to all other peers subscribed to a topic
gossip: gossip::Behavior,
}
#[allow(clippy::type_complexity)]
struct Libp2pInner {
peers: Peers,
gossip: mpsc::UnboundedSender<Message>,
outbound_requests: mpsc::UnboundedSender<(PeerId, Request, oneshot::Sender<Response>)>,
tributary_gossip: Mutex<mpsc::UnboundedReceiver<([u8; 32], Vec<u8>)>>,
signed_cosigns: Mutex<mpsc::UnboundedReceiver<SignedCosign>>,
signed_cosigns_send: mpsc::UnboundedSender<SignedCosign>,
heartbeat_requests: Mutex<mpsc::UnboundedReceiver<(RequestId, ValidatorSet, [u8; 32])>>,
notable_cosign_requests: Mutex<mpsc::UnboundedReceiver<(RequestId, [u8; 32])>>,
inbound_request_responses: mpsc::UnboundedSender<(RequestId, Response)>,
}
/// The libp2p-backed P2P implementation.
///
/// The P2p trait implementation does not support backpressure and is expected to be fully
/// utilized. Failure to poll the entire API will cause unbounded memory growth.
#[derive(Clone)]
pub struct Libp2p(Arc<Libp2pInner>);
impl Libp2p {
/// Create a new libp2p-backed P2P instance.
///
/// This will spawn all of the internal tasks necessary for functioning.
pub fn new(serai_key: &Zeroizing<Keypair>, serai: Arc<Serai>) -> Libp2p {
// Define the object we track peers with
let peers = Peers { peers: Arc::new(RwLock::new(HashMap::new())) };
// Define the dial task
let (dial_task_def, dial_task) = Task::new();
let (to_dial_send, to_dial_recv) = mpsc::unbounded_channel();
tokio::spawn(
DialTask::new(serai.clone(), peers.clone(), to_dial_send)
.continually_run(dial_task_def, vec![]),
);
let swarm = {
let new_only_validators = |noise_keypair: &identity::Keypair| -> Result<_, ()> {
Ok(OnlyValidators { serai_key: serai_key.clone(), noise_keypair: noise_keypair.clone() })
};
let new_yamux = || {
let mut config = yamux::Config::default();
// 1 MiB default + max message size
config.set_max_buffer_size((1024 * 1024) + MAX_LIBP2P_MESSAGE_SIZE);
// 256 KiB default + max message size
config
.set_receive_window_size(((256 * 1024) + MAX_LIBP2P_MESSAGE_SIZE).try_into().unwrap());
config
};
let mut swarm = SwarmBuilder::with_existing_identity(identity::Keypair::generate_ed25519())
.with_tokio()
.with_tcp(TcpConfig::default().nodelay(true), new_only_validators, new_yamux)
.unwrap()
.with_behaviour(|_| Behavior {
allow_list: allow_block_list::Behaviour::default(),
// Limit each per to a single connection
connection_limits: connection_limits::Behaviour::new(
ConnectionLimits::default().with_max_established_per_peer(Some(1)),
),
ping: ping::new_behavior(),
reqres: reqres::new_behavior(),
gossip: gossip::new_behavior(),
})
.unwrap()
.with_swarm_config(|config| {
config
.with_idle_connection_timeout(ping::INTERVAL + ping::TIMEOUT + Duration::from_secs(5))
})
.build();
swarm.listen_on(format!("/ip4/0.0.0.0/tcp/{PORT}").parse().unwrap()).unwrap();
swarm.listen_on(format!("/ip6/::/tcp/{PORT}").parse().unwrap()).unwrap();
swarm
};
let (swarm_validators, validator_changes) = UpdateValidatorsTask::spawn(serai);
let (gossip_send, gossip_recv) = mpsc::unbounded_channel();
let (signed_cosigns_send, signed_cosigns_recv) = mpsc::unbounded_channel();
let (tributary_gossip_send, tributary_gossip_recv) = mpsc::unbounded_channel();
let (outbound_requests_send, outbound_requests_recv) = mpsc::unbounded_channel();
let (heartbeat_requests_send, heartbeat_requests_recv) = mpsc::unbounded_channel();
let (notable_cosign_requests_send, notable_cosign_requests_recv) = mpsc::unbounded_channel();
let (inbound_request_responses_send, inbound_request_responses_recv) =
mpsc::unbounded_channel();
// Create the swarm task
SwarmTask::spawn(
dial_task,
to_dial_recv,
swarm_validators,
validator_changes,
peers.clone(),
swarm,
gossip_recv,
signed_cosigns_send.clone(),
tributary_gossip_send,
outbound_requests_recv,
heartbeat_requests_send,
notable_cosign_requests_send,
inbound_request_responses_recv,
);
Libp2p(Arc::new(Libp2pInner {
peers,
gossip: gossip_send,
outbound_requests: outbound_requests_send,
tributary_gossip: Mutex::new(tributary_gossip_recv),
signed_cosigns: Mutex::new(signed_cosigns_recv),
signed_cosigns_send,
heartbeat_requests: Mutex::new(heartbeat_requests_recv),
notable_cosign_requests: Mutex::new(notable_cosign_requests_recv),
inbound_request_responses: inbound_request_responses_send,
}))
}
}
impl tributary_sdk::P2p for Libp2p {
fn broadcast(&self, tributary: [u8; 32], message: Vec<u8>) -> impl Send + Future<Output = ()> {
async move {
self
.0
.gossip
.send(Message::Tributary { tributary, message })
.expect("gossip recv channel was dropped?");
}
}
}
impl serai_cosign::RequestNotableCosigns for Libp2p {
type Error = ();
fn request_notable_cosigns(
&self,
global_session: [u8; 32],
) -> impl Send + Future<Output = Result<(), Self::Error>> {
async move {
const AMOUNT_OF_PEERS_TO_REQUEST_FROM: usize = 3;
const NOTABLE_COSIGNS_TIMEOUT: Duration = Duration::from_secs(5);
let request = Request::NotableCosigns { global_session };
let peers = self.0.peers.peers.read().await.clone();
// HashSet of all peers
let peers = peers.into_values().flat_map(<_>::into_iter).collect::<HashSet<_>>();
// Vec of all peers
let mut peers = peers.into_iter().collect::<Vec<_>>();
let mut channels = Vec::with_capacity(AMOUNT_OF_PEERS_TO_REQUEST_FROM);
for _ in 0 .. AMOUNT_OF_PEERS_TO_REQUEST_FROM {
if peers.is_empty() {
break;
}
let i = usize::try_from(OsRng.next_u64() % u64::try_from(peers.len()).unwrap()).unwrap();
let peer = peers.swap_remove(i);
let (sender, receiver) = oneshot::channel();
self
.0
.outbound_requests
.send((peer, request, sender))
.expect("outbound requests recv channel was dropped?");
channels.push(receiver);
}
// We could reduce our latency by using FuturesUnordered here but the latency isn't a concern
for channel in channels {
if let Ok(Ok(Response::NotableCosigns(cosigns))) =
tokio::time::timeout(NOTABLE_COSIGNS_TIMEOUT, channel).await
{
for cosign in cosigns {
self
.0
.signed_cosigns_send
.send(cosign)
.expect("signed_cosigns recv in this object was dropped?");
}
}
}
Ok(())
}
}
}
impl serai_coordinator_p2p::P2p for Libp2p {
type Peer<'a> = Peer<'a>;
fn peers(&self, network: NetworkId) -> impl Send + Future<Output = Vec<Self::Peer<'_>>> {
async move {
let Some(peer_ids) = self.0.peers.peers.read().await.get(&network).cloned() else {
return vec![];
};
let mut res = vec![];
for id in peer_ids {
res.push(Peer { outbound_requests: &self.0.outbound_requests, id });
}
res
}
}
fn publish_cosign(&self, cosign: SignedCosign) -> impl Send + Future<Output = ()> {
async move {
self.0.gossip.send(Message::Cosign(cosign)).expect("gossip recv channel was dropped?");
}
}
fn heartbeat(
&self,
) -> impl Send + Future<Output = (Heartbeat, oneshot::Sender<Vec<TributaryBlockWithCommit>>)> {
async move {
let (request_id, set, latest_block_hash) = self
.0
.heartbeat_requests
.lock()
.await
.recv()
.await
.expect("heartbeat_requests_send was dropped?");
let (sender, receiver) = oneshot::channel();
tokio::spawn({
let respond = self.0.inbound_request_responses.clone();
async move {
// The swarm task expects us to respond to every request. If the caller drops this
// channel, we'll receive `Err` and respond with `vec![]`, safely satisfying that bound
// without requiring the caller send a value down this channel
let response = if let Ok(blocks) = receiver.await {
Response::Blocks(blocks)
} else {
Response::Blocks(vec![])
};
respond
.send((request_id, response))
.expect("inbound_request_responses_recv was dropped?");
}
});
(Heartbeat { set, latest_block_hash }, sender)
}
}
fn notable_cosigns_request(
&self,
) -> impl Send + Future<Output = ([u8; 32], oneshot::Sender<Vec<SignedCosign>>)> {
async move {
let (request_id, global_session) = self
.0
.notable_cosign_requests
.lock()
.await
.recv()
.await
.expect("notable_cosign_requests_send was dropped?");
let (sender, receiver) = oneshot::channel();
tokio::spawn({
let respond = self.0.inbound_request_responses.clone();
async move {
let response = if let Ok(notable_cosigns) = receiver.await {
Response::NotableCosigns(notable_cosigns)
} else {
Response::NotableCosigns(vec![])
};
respond
.send((request_id, response))
.expect("inbound_request_responses_recv was dropped?");
}
});
(global_session, sender)
}
}
fn tributary_message(&self) -> impl Send + Future<Output = ([u8; 32], Vec<u8>)> {
async move {
self.0.tributary_gossip.lock().await.recv().await.expect("tributary_gossip send was dropped?")
}
}
fn cosign(&self) -> impl Send + Future<Output = SignedCosign> {
async move {
self
.0
.signed_cosigns
.lock()
.await
.recv()
.await
.expect("signed_cosigns couldn't recv despite send in same object?")
}
}
}

View File

@@ -0,0 +1,17 @@
use core::time::Duration;
use tributary_sdk::tendermint::LATENCY_TIME;
use libp2p::ping::{self, Config, Behaviour};
pub use ping::Event;
pub(crate) const INTERVAL: Duration = Duration::from_secs(30);
// LATENCY_TIME represents the maximum latency for message delivery. Sending the ping, and
// receiving the pong, each have to occur within this time bound to validate the connection. We
// enforce that, as best we can, by requiring the round-trip be within twice the allowed latency.
pub(crate) const TIMEOUT: Duration = Duration::from_millis((2 * LATENCY_TIME) as u64);
pub(crate) type Behavior = Behaviour;
pub(crate) fn new_behavior() -> Behavior {
Behavior::new(Config::default().with_interval(INTERVAL).with_timeout(TIMEOUT))
}

View File

@@ -0,0 +1,135 @@
use core::{fmt, time::Duration};
use std::io;
use async_trait::async_trait;
use borsh::{BorshSerialize, BorshDeserialize};
use futures_util::{AsyncRead, AsyncReadExt, AsyncWrite, AsyncWriteExt};
use libp2p::request_response::{
self, Codec as CodecTrait, Event as GenericEvent, Config, Behaviour, ProtocolSupport,
};
pub use request_response::{RequestId, Message};
use serai_cosign::SignedCosign;
use serai_coordinator_p2p::{Heartbeat, TributaryBlockWithCommit};
/// The maximum message size for the request-response protocol
// This is derived from the heartbeat message size as it's our largest message
pub(crate) const MAX_LIBP2P_REQRES_MESSAGE_SIZE: usize =
1024 + serai_coordinator_p2p::heartbeat::BATCH_SIZE_LIMIT;
const PROTOCOL: &str = "/serai/coordinator/reqres/1.0.0";
/// Requests which can be made via the request-response protocol.
#[derive(Clone, Copy, Debug, BorshSerialize, BorshDeserialize)]
pub(crate) enum Request {
/// A heartbeat informing our peers of our latest block, for the specified blockchain, on regular
/// intervals.
///
/// If our peers have more blocks than us, they're expected to respond with those blocks.
Heartbeat(Heartbeat),
/// A request for the notable cosigns for a global session.
NotableCosigns { global_session: [u8; 32] },
}
/// Responses which can be received via the request-response protocol.
#[derive(Clone, BorshSerialize, BorshDeserialize)]
pub(crate) enum Response {
None,
Blocks(Vec<TributaryBlockWithCommit>),
NotableCosigns(Vec<SignedCosign>),
}
impl fmt::Debug for Response {
fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
Response::None => fmt.debug_struct("Response::None").finish(),
Response::Blocks(_) => fmt.debug_struct("Response::Block").finish_non_exhaustive(),
Response::NotableCosigns(_) => {
fmt.debug_struct("Response::NotableCosigns").finish_non_exhaustive()
}
}
}
}
/// The codec used for the request-response protocol.
///
/// We don't use CBOR or JSON, but use borsh to create `Vec<u8>`s we then length-prefix. While
/// ideally, we'd use borsh directly with the `io` traits defined here, they're async and there
/// isn't an amenable API within borsh for incremental deserialization.
#[derive(Default, Clone, Copy, Debug)]
pub(crate) struct Codec;
impl Codec {
async fn read<M: BorshDeserialize>(io: &mut (impl Unpin + AsyncRead)) -> io::Result<M> {
let mut len = [0; 4];
io.read_exact(&mut len).await?;
let len = usize::try_from(u32::from_le_bytes(len)).expect("not at least a 32-bit platform?");
if len > MAX_LIBP2P_REQRES_MESSAGE_SIZE {
Err(io::Error::other("request length exceeded MAX_LIBP2P_REQRES_MESSAGE_SIZE"))?;
}
// This may be a non-trivial allocation easily causable
// While we could chunk the read, meaning we only perform the allocation as bandwidth is used,
// the max message size should be sufficiently sane
let mut buf = vec![0; len];
io.read_exact(&mut buf).await?;
let mut buf = buf.as_slice();
let res = M::deserialize(&mut buf)?;
if !buf.is_empty() {
Err(io::Error::other("p2p message had extra data appended to it"))?;
}
Ok(res)
}
async fn write(io: &mut (impl Unpin + AsyncWrite), msg: &impl BorshSerialize) -> io::Result<()> {
let msg = borsh::to_vec(msg).unwrap();
io.write_all(&u32::try_from(msg.len()).unwrap().to_le_bytes()).await?;
io.write_all(&msg).await
}
}
#[async_trait]
impl CodecTrait for Codec {
type Protocol = &'static str;
type Request = Request;
type Response = Response;
async fn read_request<R: Send + Unpin + AsyncRead>(
&mut self,
_: &Self::Protocol,
io: &mut R,
) -> io::Result<Request> {
Self::read(io).await
}
async fn read_response<R: Send + Unpin + AsyncRead>(
&mut self,
_: &Self::Protocol,
io: &mut R,
) -> io::Result<Response> {
Self::read(io).await
}
async fn write_request<W: Send + Unpin + AsyncWrite>(
&mut self,
_: &Self::Protocol,
io: &mut W,
req: Request,
) -> io::Result<()> {
Self::write(io, &req).await
}
async fn write_response<W: Send + Unpin + AsyncWrite>(
&mut self,
_: &Self::Protocol,
io: &mut W,
res: Response,
) -> io::Result<()> {
Self::write(io, &res).await
}
}
pub(crate) type Event = GenericEvent<Request, Response>;
pub(crate) type Behavior = Behaviour<Codec>;
pub(crate) fn new_behavior() -> Behavior {
let mut config = Config::default();
config.set_request_timeout(Duration::from_secs(5));
Behavior::new([(PROTOCOL, ProtocolSupport::Full)], config)
}

View File

@@ -0,0 +1,356 @@
use std::{
sync::Arc,
collections::{HashSet, HashMap},
time::{Duration, Instant},
};
use borsh::BorshDeserialize;
use serai_client::validator_sets::primitives::ValidatorSet;
use tokio::sync::{mpsc, oneshot, RwLock};
use serai_task::TaskHandle;
use serai_cosign::SignedCosign;
use futures_util::StreamExt;
use libp2p::{
identity::PeerId,
request_response::{RequestId, ResponseChannel},
swarm::{dial_opts::DialOpts, SwarmEvent, Swarm},
};
use serai_coordinator_p2p::Heartbeat;
use crate::{
Peers, BehaviorEvent, Behavior,
validators::{self, Validators},
ping,
reqres::{self, Request, Response},
gossip,
};
const TIME_BETWEEN_REBUILD_PEERS: Duration = Duration::from_secs(10 * 60);
/*
`SwarmTask` handles everything we need the `Swarm` object for. The goal is to minimize the
contention on this task. Unfortunately, the `Swarm` object itself is needed for a variety of
purposes making this a rather large task.
Responsibilities include:
- Actually dialing new peers (the selection process occurs in another task)
- Maintaining the peers structure (as we need the Swarm object to see who our peers are)
- Gossiping messages
- Dispatching gossiped messages
- Sending requests
- Dispatching responses to requests
- Dispatching received requests
- Sending responses
*/
pub(crate) struct SwarmTask {
dial_task: TaskHandle,
to_dial: mpsc::UnboundedReceiver<DialOpts>,
last_dial_task_run: Instant,
validators: Arc<RwLock<Validators>>,
validator_changes: mpsc::UnboundedReceiver<validators::Changes>,
peers: Peers,
rebuild_peers_at: Instant,
swarm: Swarm<Behavior>,
gossip: mpsc::UnboundedReceiver<gossip::Message>,
signed_cosigns: mpsc::UnboundedSender<SignedCosign>,
tributary_gossip: mpsc::UnboundedSender<([u8; 32], Vec<u8>)>,
outbound_requests: mpsc::UnboundedReceiver<(PeerId, Request, oneshot::Sender<Response>)>,
outbound_request_responses: HashMap<RequestId, oneshot::Sender<Response>>,
inbound_request_response_channels: HashMap<RequestId, ResponseChannel<Response>>,
heartbeat_requests: mpsc::UnboundedSender<(RequestId, ValidatorSet, [u8; 32])>,
notable_cosign_requests: mpsc::UnboundedSender<(RequestId, [u8; 32])>,
inbound_request_responses: mpsc::UnboundedReceiver<(RequestId, Response)>,
}
impl SwarmTask {
fn handle_gossip(&mut self, event: gossip::Event) {
match event {
gossip::Event::Message { message, .. } => {
let Ok(message) = gossip::Message::deserialize(&mut message.data.as_slice()) else {
// TODO: Penalize the PeerId which created this message, which requires authenticating
// each message OR moving to explicit acknowledgement before re-gossiping
return;
};
match message {
gossip::Message::Tributary { tributary, message } => {
let _: Result<_, _> = self.tributary_gossip.send((tributary, message));
}
gossip::Message::Cosign(signed_cosign) => {
let _: Result<_, _> = self.signed_cosigns.send(signed_cosign);
}
}
}
gossip::Event::Subscribed { .. } | gossip::Event::Unsubscribed { .. } => {}
gossip::Event::GossipsubNotSupported { peer_id } => {
let _: Result<_, _> = self.swarm.disconnect_peer_id(peer_id);
}
}
}
fn handle_reqres(&mut self, event: reqres::Event) {
match event {
reqres::Event::Message { message, .. } => match message {
reqres::Message::Request { request_id, request, channel } => match request {
reqres::Request::Heartbeat(Heartbeat { set, latest_block_hash }) => {
self.inbound_request_response_channels.insert(request_id, channel);
let _: Result<_, _> =
self.heartbeat_requests.send((request_id, set, latest_block_hash));
}
reqres::Request::NotableCosigns { global_session } => {
self.inbound_request_response_channels.insert(request_id, channel);
let _: Result<_, _> = self.notable_cosign_requests.send((request_id, global_session));
}
},
reqres::Message::Response { request_id, response } => {
if let Some(channel) = self.outbound_request_responses.remove(&request_id) {
let _: Result<_, _> = channel.send(response);
}
}
},
reqres::Event::OutboundFailure { request_id, .. } => {
// Send None as the response for the request
if let Some(channel) = self.outbound_request_responses.remove(&request_id) {
let _: Result<_, _> = channel.send(Response::None);
}
}
reqres::Event::InboundFailure { .. } | reqres::Event::ResponseSent { .. } => {}
}
}
async fn run(mut self) {
loop {
let time_till_rebuild_peers = self.rebuild_peers_at.saturating_duration_since(Instant::now());
tokio::select! {
// If the validators have changed, update the allow list
validator_changes = self.validator_changes.recv() => {
let validator_changes = validator_changes.expect("validators update task shut down?");
let behavior = &mut self.swarm.behaviour_mut().allow_list;
for removed in validator_changes.removed {
behavior.disallow_peer(removed);
}
for added in validator_changes.added {
behavior.allow_peer(added);
}
}
// Dial peers we're instructed to
dial_opts = self.to_dial.recv() => {
let dial_opts = dial_opts.expect("DialTask was closed?");
let _: Result<_, _> = self.swarm.dial(dial_opts);
}
/*
Rebuild the peers every 10 minutes.
This protects against any race conditions/edge cases we have in our logic to track peers,
along with unrepresented behavior such as when a peer changes the networks they're active
in. This lets the peer tracking logic simply be 'good enough' to not become horribly
corrupt over the span of `TIME_BETWEEN_REBUILD_PEERS`.
We also use this to disconnect all peers who are no longer active in any network.
*/
() = tokio::time::sleep(time_till_rebuild_peers) => {
let validators_by_network = self.validators.read().await.by_network().clone();
let connected_peers = self.swarm.connected_peers().copied().collect::<HashSet<_>>();
// Build the new peers object
let mut peers = HashMap::new();
for (network, validators) in validators_by_network {
peers.insert(network, validators.intersection(&connected_peers).copied().collect());
}
// Write the new peers object
*self.peers.peers.write().await = peers;
self.rebuild_peers_at = Instant::now() + TIME_BETWEEN_REBUILD_PEERS;
}
// Handle swarm events
event = self.swarm.next() => {
// `Swarm::next` will never return `Poll::Ready(None)`
// https://docs.rs/
// libp2p/0.54.1/libp2p/struct.Swarm.html#impl-Stream-for-Swarm%3CTBehaviour%3E
let event = event.unwrap();
match event {
// New connection, so update peers
SwarmEvent::ConnectionEstablished { peer_id, .. } => {
let Some(networks) =
self.validators.read().await.networks(&peer_id).cloned() else { continue };
let mut peers = self.peers.peers.write().await;
for network in networks {
peers.entry(network).or_insert_with(HashSet::new).insert(peer_id);
}
}
// Connection closed, so update peers
SwarmEvent::ConnectionClosed { peer_id, .. } => {
let Some(networks) =
self.validators.read().await.networks(&peer_id).cloned() else { continue };
let mut peers = self.peers.peers.write().await;
for network in networks {
peers.entry(network).or_insert_with(HashSet::new).remove(&peer_id);
}
/*
We want to re-run the dial task, since we lost a peer, in case we should find new
peers. This opens a DoS where a validator repeatedly opens/closes connections to
force iterations of the dial task. We prevent this by setting a minimum distance
since the last explicit iteration.
This is suboptimal. If we have several disconnects in immediate proximity, we'll
trigger the dial task upon the first (where we may still have enough peers we
shouldn't dial more) but not the last (where we may have so few peers left we
should dial more). This is accepted as the dial task will eventually run on its
natural timer.
*/
const MINIMUM_TIME_SINCE_LAST_EXPLICIT_DIAL: Duration = Duration::from_secs(60);
let now = Instant::now();
if (self.last_dial_task_run + MINIMUM_TIME_SINCE_LAST_EXPLICIT_DIAL) < now {
self.dial_task.run_now();
self.last_dial_task_run = now;
}
}
SwarmEvent::Behaviour(
BehaviorEvent::AllowList(event) | BehaviorEvent::ConnectionLimits(event)
) => {
// This *is* an exhaustive match as these events are empty enums
match event {}
}
SwarmEvent::Behaviour(
BehaviorEvent::Ping(ping::Event { peer: _, connection, result, })
) => {
if result.is_err() {
self.swarm.close_connection(connection);
}
}
SwarmEvent::Behaviour(BehaviorEvent::Reqres(event)) => {
self.handle_reqres(event)
}
SwarmEvent::Behaviour(BehaviorEvent::Gossip(event)) => {
self.handle_gossip(event)
}
// We don't handle any of these
SwarmEvent::IncomingConnection { .. } |
SwarmEvent::IncomingConnectionError { .. } |
SwarmEvent::OutgoingConnectionError { .. } |
SwarmEvent::NewListenAddr { .. } |
SwarmEvent::ExpiredListenAddr { .. } |
SwarmEvent::ListenerClosed { .. } |
SwarmEvent::ListenerError { .. } |
SwarmEvent::Dialing { .. } => {}
}
}
message = self.gossip.recv() => {
let message = message.expect("channel for messages to gossip was closed?");
let topic = message.topic();
let message = borsh::to_vec(&message).unwrap();
/*
If we're sending a message for this topic, it's because this topic is relevant to us.
Subscribe to it.
We create topics roughly weekly, one per validator set/session. Once present in a
topic, we're interested in all messages for it until the validator set/session retires.
Then there should no longer be any messages for the topic as we should drop the
Tributary which creates the messages.
We use this as an argument to not bother implement unsubscribing from topics. They're
incredibly infrequently created and old topics shouldn't still have messages published
to them. Having the coordinator reboot being our method of unsubscribing is fine.
Alternatively, we could route an API to determine when a topic is retired, or retire
any topics we haven't sent messages on in the past hour.
*/
let behavior = self.swarm.behaviour_mut();
let _: Result<_, _> = behavior.gossip.subscribe(&topic);
/*
This may be an error of `InsufficientPeers`. If so, we could ask DialTask to dial more
peers for this network. We don't as we assume DialTask will detect the lack of peers
for this network, and will already successfully handle this.
*/
let _: Result<_, _> = behavior.gossip.publish(topic.hash(), message);
}
request = self.outbound_requests.recv() => {
let (peer, request, response_channel) =
request.expect("channel for requests was closed?");
let request_id = self.swarm.behaviour_mut().reqres.send_request(&peer, request);
self.outbound_request_responses.insert(request_id, response_channel);
}
response = self.inbound_request_responses.recv() => {
let (request_id, response) =
response.expect("channel for inbound request responses was closed?");
if let Some(channel) = self.inbound_request_response_channels.remove(&request_id) {
let _: Result<_, _> =
self.swarm.behaviour_mut().reqres.send_response(channel, response);
}
}
}
}
}
#[allow(clippy::too_many_arguments)]
pub(crate) fn spawn(
dial_task: TaskHandle,
to_dial: mpsc::UnboundedReceiver<DialOpts>,
validators: Arc<RwLock<Validators>>,
validator_changes: mpsc::UnboundedReceiver<validators::Changes>,
peers: Peers,
swarm: Swarm<Behavior>,
gossip: mpsc::UnboundedReceiver<gossip::Message>,
signed_cosigns: mpsc::UnboundedSender<SignedCosign>,
tributary_gossip: mpsc::UnboundedSender<([u8; 32], Vec<u8>)>,
outbound_requests: mpsc::UnboundedReceiver<(PeerId, Request, oneshot::Sender<Response>)>,
heartbeat_requests: mpsc::UnboundedSender<(RequestId, ValidatorSet, [u8; 32])>,
notable_cosign_requests: mpsc::UnboundedSender<(RequestId, [u8; 32])>,
inbound_request_responses: mpsc::UnboundedReceiver<(RequestId, Response)>,
) {
tokio::spawn(
SwarmTask {
dial_task,
to_dial,
last_dial_task_run: Instant::now(),
validators,
validator_changes,
peers,
rebuild_peers_at: Instant::now() + TIME_BETWEEN_REBUILD_PEERS,
swarm,
gossip,
signed_cosigns,
tributary_gossip,
outbound_requests,
outbound_request_responses: HashMap::new(),
inbound_request_response_channels: HashMap::new(),
heartbeat_requests,
notable_cosign_requests,
inbound_request_responses,
}
.run(),
);
}
}

View File

@@ -0,0 +1,214 @@
use core::{borrow::Borrow, future::Future};
use std::{
sync::Arc,
collections::{HashSet, HashMap},
};
use serai_client::{primitives::NetworkId, validator_sets::primitives::Session, SeraiError, Serai};
use serai_task::{Task, ContinuallyRan};
use libp2p::PeerId;
use futures_util::stream::{StreamExt, FuturesUnordered};
use tokio::sync::{mpsc, RwLock};
use crate::peer_id_from_public;
pub(crate) struct Changes {
pub(crate) removed: HashSet<PeerId>,
pub(crate) added: HashSet<PeerId>,
}
pub(crate) struct Validators {
serai: Arc<Serai>,
// A cache for which session we're populated with the validators of
sessions: HashMap<NetworkId, Session>,
// The validators by network
by_network: HashMap<NetworkId, HashSet<PeerId>>,
// The validators and their networks
validators: HashMap<PeerId, HashSet<NetworkId>>,
// The channel to send the changes down
changes: mpsc::UnboundedSender<Changes>,
}
impl Validators {
pub(crate) fn new(serai: Arc<Serai>) -> (Self, mpsc::UnboundedReceiver<Changes>) {
let (send, recv) = mpsc::unbounded_channel();
let validators = Validators {
serai,
sessions: HashMap::new(),
by_network: HashMap::new(),
validators: HashMap::new(),
changes: send,
};
(validators, recv)
}
async fn session_changes(
serai: impl Borrow<Serai>,
sessions: impl Borrow<HashMap<NetworkId, Session>>,
) -> Result<Vec<(NetworkId, Session, HashSet<PeerId>)>, SeraiError> {
let temporal_serai = serai.borrow().as_of_latest_finalized_block().await?;
let temporal_serai = temporal_serai.validator_sets();
let mut session_changes = vec![];
{
// FuturesUnordered can be bad practice as it'll cause timeouts if infrequently polled, but
// we poll it till it yields all futures with the most minimal processing possible
let mut futures = FuturesUnordered::new();
for network in serai_client::primitives::NETWORKS {
if network == NetworkId::Serai {
continue;
}
let sessions = sessions.borrow();
futures.push(async move {
let session = match temporal_serai.session(network).await {
Ok(Some(session)) => session,
Ok(None) => return Ok(None),
Err(e) => return Err(e),
};
if sessions.get(&network) == Some(&session) {
Ok(None)
} else {
match temporal_serai.active_network_validators(network).await {
Ok(validators) => Ok(Some((
network,
session,
validators.into_iter().map(peer_id_from_public).collect(),
))),
Err(e) => Err(e),
}
}
});
}
while let Some(session_change) = futures.next().await {
if let Some(session_change) = session_change? {
session_changes.push(session_change);
}
}
}
Ok(session_changes)
}
fn incorporate_session_changes(
&mut self,
session_changes: Vec<(NetworkId, Session, HashSet<PeerId>)>,
) {
let mut removed = HashSet::new();
let mut added = HashSet::new();
for (network, session, validators) in session_changes {
// Remove the existing validators
for validator in self.by_network.remove(&network).unwrap_or_else(HashSet::new) {
// Get all networks this validator is in
let mut networks = self.validators.remove(&validator).unwrap();
// Remove this one
networks.remove(&network);
if !networks.is_empty() {
// Insert the networks back if the validator was present in other networks
self.validators.insert(validator, networks);
} else {
// Because this validator is no longer present in any network, mark them as removed
/*
This isn't accurate. The validator isn't present in the latest session for this
network. The validator was present in the prior session which has yet to retire. Our
lack of explicit inclusion for both the prior session and the current session causes
only the validators mutually present in both sessions to be responsible for all actions
still ongoing as the prior validator set retires.
TODO: Fix this
*/
removed.insert(validator);
}
}
// Add the new validators
for validator in validators.iter().copied() {
self.validators.entry(validator).or_insert_with(HashSet::new).insert(network);
added.insert(validator);
}
self.by_network.insert(network, validators);
// Update the session we have populated
self.sessions.insert(network, session);
}
// Only flag validators for removal if they weren't simultaneously added by these changes
removed.retain(|validator| !added.contains(validator));
// Send the changes, dropping the error
// This lets the caller opt-out of change notifications by dropping the receiver
let _: Result<_, _> = self.changes.send(Changes { removed, added });
}
/// Update the view of the validators.
pub(crate) async fn update(&mut self) -> Result<(), SeraiError> {
let session_changes = Self::session_changes(&*self.serai, &self.sessions).await?;
self.incorporate_session_changes(session_changes);
Ok(())
}
pub(crate) fn by_network(&self) -> &HashMap<NetworkId, HashSet<PeerId>> {
&self.by_network
}
pub(crate) fn networks(&self, peer_id: &PeerId) -> Option<&HashSet<NetworkId>> {
self.validators.get(peer_id)
}
}
/// A task which updates a set of validators.
///
/// The validators managed by this tak will have their exclusive lock held for a minimal amount of
/// time while the update occurs to minimize the disruption to the services relying on it.
pub(crate) struct UpdateValidatorsTask {
validators: Arc<RwLock<Validators>>,
}
impl UpdateValidatorsTask {
/// Spawn a new instance of the UpdateValidatorsTask.
///
/// This returns a reference to the Validators it updates after spawning itself.
pub(crate) fn spawn(
serai: Arc<Serai>,
) -> (Arc<RwLock<Validators>>, mpsc::UnboundedReceiver<Changes>) {
// The validators which will be updated
let (validators, changes) = Validators::new(serai);
let validators = Arc::new(RwLock::new(validators));
// Define the task
let (update_validators_task, update_validators_task_handle) = Task::new();
// Forget the handle, as dropping the handle would stop the task
core::mem::forget(update_validators_task_handle);
// Spawn the task
tokio::spawn(
(Self { validators: validators.clone() }).continually_run(update_validators_task, vec![]),
);
// Return the validators
(validators, changes)
}
}
impl ContinuallyRan for UpdateValidatorsTask {
// Only run every minute, not the default of every five seconds
const DELAY_BETWEEN_ITERATIONS: u64 = 60;
const MAX_DELAY_BETWEEN_ITERATIONS: u64 = 5 * 60;
type Error = SeraiError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let session_changes = {
let validators = self.validators.read().await;
Validators::session_changes(validators.serai.clone(), validators.sessions.clone()).await?
};
self.validators.write().await.incorporate_session_changes(session_changes);
Ok(true)
}
}
}

View File

@@ -0,0 +1,151 @@
use core::future::Future;
use std::time::{Duration, SystemTime};
use serai_client::validator_sets::primitives::{MAX_KEY_SHARES_PER_SET, ValidatorSet};
use futures_lite::FutureExt;
use tributary_sdk::{ReadWrite, TransactionTrait, Block, Tributary, TributaryReader};
use serai_db::*;
use serai_task::ContinuallyRan;
use crate::{Heartbeat, Peer, P2p};
// Amount of blocks in a minute
const BLOCKS_PER_MINUTE: usize =
(60 / (tributary_sdk::tendermint::TARGET_BLOCK_TIME / 1000)) as usize;
/// The minimum amount of blocks to include/included within a batch, assuming there's blocks to
/// include in the batch.
///
/// This decides the size limit of the Batch (the Block size limit multiplied by the minimum amount
/// of blocks we'll send). The actual amount of blocks sent will be the amount which fits within
/// the size limit.
pub const MIN_BLOCKS_PER_BATCH: usize = BLOCKS_PER_MINUTE + 1;
/// The size limit for a batch of blocks sent in response to a Heartbeat.
///
/// This estimates the size of a commit as `32 + (MAX_VALIDATORS * 128)`. At the time of writing, a
/// commit is `8 + (validators * 32) + (32 + (validators * 32))` (for the time, list of validators,
/// and aggregate signature). Accordingly, this should be a safe over-estimate.
pub const BATCH_SIZE_LIMIT: usize = MIN_BLOCKS_PER_BATCH *
(tributary_sdk::BLOCK_SIZE_LIMIT + 32 + ((MAX_KEY_SHARES_PER_SET as usize) * 128));
/// Sends a heartbeat to other validators on regular intervals informing them of our Tributary's
/// tip.
///
/// If the other validator has more blocks then we do, they're expected to inform us. This forms
/// the sync protocol for our Tributaries.
pub(crate) struct HeartbeatTask<TD: Db, Tx: TransactionTrait, P: P2p> {
pub(crate) set: ValidatorSet,
pub(crate) tributary: Tributary<TD, Tx, P>,
pub(crate) reader: TributaryReader<TD, Tx>,
pub(crate) p2p: P,
}
impl<TD: Db, Tx: TransactionTrait, P: P2p> ContinuallyRan for HeartbeatTask<TD, Tx, P> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
// If our blockchain hasn't had a block in the past minute, trigger the heartbeat protocol
const TIME_TO_TRIGGER_SYNCING: Duration = Duration::from_secs(60);
let mut tip = self.reader.tip();
let time_since = {
let block_time = if let Some(time_of_block) = self.reader.time_of_block(&tip) {
SystemTime::UNIX_EPOCH + Duration::from_secs(time_of_block)
} else {
// If we couldn't fetch this block's time, assume it's old
// We don't want to declare its unix time as 0 and claim it's 50+ years old though
log::warn!(
"heartbeat task couldn't fetch the time of a block, flagging it as a minute old"
);
SystemTime::now() - TIME_TO_TRIGGER_SYNCING
};
SystemTime::now().duration_since(block_time).unwrap_or(Duration::ZERO)
};
let mut tip_is_stale = false;
let mut synced_block = false;
if TIME_TO_TRIGGER_SYNCING <= time_since {
log::warn!(
"last known tributary block for {:?} was {} seconds ago",
self.set,
time_since.as_secs()
);
// This requests all peers for this network, without differentiating by session
// This should be fine as most validators should overlap across sessions
'peer: for peer in self.p2p.peers(self.set.network).await {
loop {
// Create the request for blocks
if tip_is_stale {
tip = self.reader.tip();
tip_is_stale = false;
}
// Necessary due to https://github.com/rust-lang/rust/issues/100013
let Some(blocks) = peer
.send_heartbeat(Heartbeat { set: self.set, latest_block_hash: tip })
.boxed()
.await
else {
continue 'peer;
};
// This is the final batch if it has less than the maximum amount of blocks
// (signifying there weren't more blocks after this to fill the batch with)
let final_batch = blocks.len() < MIN_BLOCKS_PER_BATCH;
// Sync each block
for block_with_commit in blocks {
let Ok(block) = Block::read(&mut block_with_commit.block.as_slice()) else {
// TODO: Disconnect/slash this peer
log::warn!("received invalid Block inside response to heartbeat");
continue 'peer;
};
// Attempt to sync the block
if !self.tributary.sync_block(block, block_with_commit.commit).await {
// The block may be invalid or stale if we added a block elsewhere
if (!tip_is_stale) && (tip != self.reader.tip()) {
// Since the Tributary's tip advanced on its own, return
return Ok(false);
}
// Since this block was invalid or stale in a way non-trivial to detect, try to
// sync with the next peer
continue 'peer;
}
// Because we synced a block, flag the tip as stale
tip_is_stale = true;
// And that we did sync a block
synced_block = true;
}
// If this was the final batch, move on from this peer
// We could assume they were honest and we are done syncing the chain, but this is a
// bit more robust
if final_batch {
continue 'peer;
}
}
}
// This will cause the tak to be run less and less often, ensuring we aren't spamming the
// net if we legitimately aren't making progress
if !synced_block {
Err(format!(
"tried to sync blocks for {:?} since we haven't seen one in {} seconds but didn't",
self.set,
time_since.as_secs(),
))?;
}
}
Ok(synced_block)
}
}
}

204
coordinator/p2p/src/lib.rs Normal file
View File

@@ -0,0 +1,204 @@
#![cfg_attr(docsrs, feature(doc_auto_cfg))]
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
use core::future::Future;
use std::collections::HashMap;
use borsh::{BorshSerialize, BorshDeserialize};
use serai_client::{primitives::NetworkId, validator_sets::primitives::ValidatorSet};
use serai_db::Db;
use tributary_sdk::{ReadWrite, TransactionTrait, Tributary, TributaryReader};
use serai_cosign::{SignedCosign, Cosigning};
use tokio::sync::{mpsc, oneshot};
use serai_task::{Task, ContinuallyRan};
/// The heartbeat task, effecting sync of Tributaries
pub mod heartbeat;
use crate::heartbeat::HeartbeatTask;
/// A heartbeat for a Tributary.
#[derive(Clone, Copy, BorshSerialize, BorshDeserialize, Debug)]
pub struct Heartbeat {
/// The Tributary this is the heartbeat of.
pub set: ValidatorSet,
/// The hash of the latest block added to the Tributary.
pub latest_block_hash: [u8; 32],
}
/// A tributary block and its commit.
#[derive(Clone, BorshSerialize, BorshDeserialize)]
pub struct TributaryBlockWithCommit {
/// The serialized block.
pub block: Vec<u8>,
/// The serialized commit.
pub commit: Vec<u8>,
}
/// A representation of a peer.
pub trait Peer<'a>: Send {
/// Send a heartbeat to this peer.
fn send_heartbeat(
&self,
heartbeat: Heartbeat,
) -> impl Send + Future<Output = Option<Vec<TributaryBlockWithCommit>>>;
}
/// The representation of the P2P network.
pub trait P2p:
Send + Sync + Clone + tributary_sdk::P2p + serai_cosign::RequestNotableCosigns
{
/// The representation of a peer.
type Peer<'a>: Peer<'a>;
/// Fetch the peers for this network.
fn peers(&self, network: NetworkId) -> impl Send + Future<Output = Vec<Self::Peer<'_>>>;
/// Broadcast a cosign.
fn publish_cosign(&self, cosign: SignedCosign) -> impl Send + Future<Output = ()>;
/// A cancel-safe future for the next heartbeat received over the P2P network.
///
/// Yields the validator set its for, the latest block hash observed, and a channel to return the
/// descending blocks. This channel MUST NOT and will not have its receiver dropped before a
/// message is sent.
fn heartbeat(
&self,
) -> impl Send + Future<Output = (Heartbeat, oneshot::Sender<Vec<TributaryBlockWithCommit>>)>;
/// A cancel-safe future for the next request for the notable cosigns of a gloabl session.
///
/// Yields the global session the request is for and a channel to return the notable cosigns.
/// This channel MUST NOT and will not have its receiver dropped before a message is sent.
fn notable_cosigns_request(
&self,
) -> impl Send + Future<Output = ([u8; 32], oneshot::Sender<Vec<SignedCosign>>)>;
/// A cancel-safe future for the next message regarding a Tributary.
///
/// Yields the message's Tributary's genesis block hash and the message.
fn tributary_message(&self) -> impl Send + Future<Output = ([u8; 32], Vec<u8>)>;
/// A cancel-safe future for the next cosign received.
fn cosign(&self) -> impl Send + Future<Output = SignedCosign>;
}
fn handle_notable_cosigns_request<D: Db>(
db: &D,
global_session: [u8; 32],
channel: oneshot::Sender<Vec<SignedCosign>>,
) {
let cosigns = Cosigning::<D>::notable_cosigns(db, global_session);
channel.send(cosigns).expect("channel listening for cosign oneshot response was dropped?");
}
fn handle_heartbeat<D: Db, T: TransactionTrait>(
reader: &TributaryReader<D, T>,
mut latest_block_hash: [u8; 32],
channel: oneshot::Sender<Vec<TributaryBlockWithCommit>>,
) {
let mut res_size = 8;
let mut res = vec![];
// This former case should be covered by this latter case
while (res.len() < heartbeat::MIN_BLOCKS_PER_BATCH) || (res_size < heartbeat::BATCH_SIZE_LIMIT) {
let Some(block_after) = reader.block_after(&latest_block_hash) else { break };
// These `break` conditions should only occur under edge cases, such as if we're actively
// deleting this Tributary due to being done with it
let Some(block) = reader.block(&block_after) else { break };
let block = block.serialize();
let Some(commit) = reader.commit(&block_after) else { break };
res_size += 8 + block.len() + 8 + commit.len();
res.push(TributaryBlockWithCommit { block, commit });
latest_block_hash = block_after;
}
channel
.send(res)
.map_err(|_| ())
.expect("channel listening for heartbeat oneshot response was dropped?");
}
/// Run the P2P instance.
///
/// `add_tributary`'s and `retire_tributary's senders, along with `send_cosigns`'s receiver, must
/// never be dropped. `retire_tributary` is not required to only be instructed with added
/// Tributaries.
pub async fn run<TD: Db, Tx: TransactionTrait, P: P2p>(
db: impl Db,
p2p: P,
mut add_tributary: mpsc::UnboundedReceiver<(ValidatorSet, Tributary<TD, Tx, P>)>,
mut retire_tributary: mpsc::UnboundedReceiver<ValidatorSet>,
send_cosigns: mpsc::UnboundedSender<SignedCosign>,
) {
let mut readers = HashMap::<ValidatorSet, TributaryReader<TD, Tx>>::new();
let mut tributaries = HashMap::<[u8; 32], mpsc::UnboundedSender<Vec<u8>>>::new();
let mut heartbeat_tasks = HashMap::<ValidatorSet, _>::new();
loop {
tokio::select! {
tributary = add_tributary.recv() => {
let (set, tributary) = tributary.expect("add_tributary send was dropped");
let reader = tributary.reader();
readers.insert(set, reader.clone());
let (heartbeat_task_def, heartbeat_task) = Task::new();
tokio::spawn(
(HeartbeatTask {
set,
tributary: tributary.clone(),
reader: reader.clone(),
p2p: p2p.clone(),
}).continually_run(heartbeat_task_def, vec![])
);
heartbeat_tasks.insert(set, heartbeat_task);
let (tributary_message_send, mut tributary_message_recv) = mpsc::unbounded_channel();
tributaries.insert(tributary.genesis(), tributary_message_send);
// For as long as this sender exists, handle the messages from it on a dedicated task
tokio::spawn(async move {
while let Some(message) = tributary_message_recv.recv().await {
tributary.handle_message(&message).await;
}
});
}
set = retire_tributary.recv() => {
let set = set.expect("retire_tributary send was dropped");
let Some(reader) = readers.remove(&set) else { continue };
tributaries.remove(&reader.genesis()).expect("tributary reader but no tributary");
heartbeat_tasks.remove(&set).expect("tributary but no heartbeat task");
}
(heartbeat, channel) = p2p.heartbeat() => {
if let Some(reader) = readers.get(&heartbeat.set) {
let reader = reader.clone(); // This is a cheap clone
// We spawn this on a task due to the DB reads needed
tokio::spawn(async move {
handle_heartbeat(&reader, heartbeat.latest_block_hash, channel)
});
}
}
(global_session, channel) = p2p.notable_cosigns_request() => {
tokio::spawn({
let db = db.clone();
async move { handle_notable_cosigns_request(&db, global_session, channel) }
});
}
(tributary, message) = p2p.tributary_message() => {
if let Some(tributary) = tributaries.get(&tributary) {
tributary.send(message).expect("tributary message recv was dropped?");
}
}
cosign = p2p.cosign() => {
// We don't call `Cosigning::intake_cosign` here as that can only be called from a single
// location. We also need to intake the cosigns we produce, which means we need to merge
// these streams (signing, network) somehow. That's done with this mpsc channel
send_cosigns.send(cosign).expect("channel receiving cosigns was dropped");
}
}
}
}

View File

@@ -1,336 +0,0 @@
use core::time::Duration;
use std::{
sync::Arc,
collections::{HashSet, HashMap},
};
use tokio::{
sync::{mpsc, Mutex, RwLock},
time::sleep,
};
use borsh::BorshSerialize;
use sp_application_crypto::RuntimePublic;
use serai_client::{
primitives::{NETWORKS, NetworkId, Signature},
validator_sets::primitives::{Session, ValidatorSet},
SeraiError, TemporalSerai, Serai,
};
use serai_db::{Get, DbTxn, Db, create_db};
use processor_messages::coordinator::cosign_block_msg;
use crate::{
p2p::{CosignedBlock, GossipMessageKind, P2p},
substrate::LatestCosignedBlock,
};
create_db! {
CosignDb {
ReceivedCosign: (set: ValidatorSet, block: [u8; 32]) -> CosignedBlock,
LatestCosign: (network: NetworkId) -> CosignedBlock,
DistinctChain: (set: ValidatorSet) -> (),
}
}
pub struct CosignEvaluator<D: Db> {
db: Mutex<D>,
serai: Arc<Serai>,
stakes: RwLock<Option<HashMap<NetworkId, u64>>>,
latest_cosigns: RwLock<HashMap<NetworkId, CosignedBlock>>,
}
impl<D: Db> CosignEvaluator<D> {
async fn update_latest_cosign(&self) {
let stakes_lock = self.stakes.read().await;
// If we haven't gotten the stake data yet, return
let Some(stakes) = stakes_lock.as_ref() else { return };
let total_stake = stakes.values().copied().sum::<u64>();
let latest_cosigns = self.latest_cosigns.read().await;
let mut highest_block = 0;
for cosign in latest_cosigns.values() {
let mut networks = HashSet::new();
for (network, sub_cosign) in &*latest_cosigns {
if sub_cosign.block_number >= cosign.block_number {
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) {
highest_block = highest_block.max(cosign.block_number);
}
}
let mut db_lock = self.db.lock().await;
let mut txn = db_lock.txn();
if highest_block > LatestCosignedBlock::latest_cosigned_block(&txn) {
log::info!("setting latest cosigned block to {}", highest_block);
LatestCosignedBlock::set(&mut txn, &highest_block);
}
txn.commit();
}
async fn update_stakes(&self) -> Result<(), SeraiError> {
let serai = self.serai.as_of_latest_finalized_block().await?;
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
*self.stakes.write().await = Some(stakes);
self.update_latest_cosign().await;
Ok(())
}
// Uses Err to signify a message should be retried
async fn handle_new_cosign(&self, cosign: CosignedBlock) -> Result<(), SeraiError> {
// If we already have this cosign or a newer cosign, return
if let Some(latest) = self.latest_cosigns.read().await.get(&cosign.network) {
if latest.block_number >= cosign.block_number {
return Ok(());
}
}
// If this an old cosign (older than a day), drop it
let latest_block = self.serai.latest_finalized_block().await?;
if (cosign.block_number + (24 * 60 * 60 / 6)) < latest_block.number() {
log::debug!("received old cosign supposedly signed by {:?}", cosign.network);
return Ok(());
}
let Some(block) = self.serai.finalized_block_by_number(cosign.block_number).await? else {
log::warn!("received cosign with a block number which doesn't map to a block");
return Ok(());
};
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 }
},
))
}
// Get the key for this network as of the prior block
// 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());
let Some(set_with_keys) = set_with_keys_fn(&serai, cosign.network).await? else {
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(());
};
if !keys
.0
.verify(&cosign_block_msg(cosign.block_number, cosign.block), &Signature(cosign.signature))
{
log::warn!("received cosigned block with an invalid signature");
return Ok(());
}
log::info!(
"received cosign for block {} ({}) by {:?}",
block.number(),
hex::encode(cosign.block),
cosign.network
);
// Save this cosign to the DB
{
let mut db = self.db.lock().await;
let mut txn = db.txn();
ReceivedCosign::set(&mut txn, set_with_keys, cosign.block, &cosign);
LatestCosign::set(&mut txn, set_with_keys.network, &(cosign));
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 {
{
let mut latest_cosigns = self.latest_cosigns.write().await;
latest_cosigns.insert(cosign.network, cosign);
}
self.update_latest_cosign().await;
}
Ok(())
}
#[allow(clippy::new_ret_no_self)]
pub fn new<P: P2p>(db: D, p2p: P, serai: Arc<Serai>) -> mpsc::UnboundedSender<CosignedBlock> {
let mut latest_cosigns = HashMap::new();
for network in NETWORKS {
if let Some(cosign) = LatestCosign::get(&db, network) {
latest_cosigns.insert(network, cosign);
}
}
let evaluator = Arc::new(Self {
db: Mutex::new(db),
serai,
stakes: RwLock::new(None),
latest_cosigns: RwLock::new(latest_cosigns),
});
// 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 {
let cosigns = evaluator.latest_cosigns.read().await.values().copied().collect::<Vec<_>>();
for cosign in cosigns {
let mut buf = vec![];
cosign.serialize(&mut buf).unwrap();
P2p::broadcast(&p2p, GossipMessageKind::CosignedBlock, buf).await;
}
sleep(Duration::from_secs(60)).await;
}
}
});
// Return the channel to send cosigns
send
}
}

View File

@@ -1,134 +1,113 @@
use blake2::{
digest::{consts::U32, Digest},
Blake2b,
};
use std::{path::Path, fs};
pub(crate) use serai_db::{Get, DbTxn, Db as DbTrait};
use serai_db::{create_db, db_channel};
use scale::Encode;
use borsh::{BorshSerialize, BorshDeserialize};
use serai_client::{
primitives::NetworkId,
validator_sets::primitives::{Session, ValidatorSet},
in_instructions::primitives::{Batch, SignedBatch},
};
pub use serai_db::*;
use serai_cosign::SignedCosign;
use serai_coordinator_substrate::NewSetInformation;
use serai_coordinator_tributary::Transaction;
use ::tributary::ReadWrite;
use crate::tributary::{TributarySpec, Transaction, scanner::RecognizedIdType};
#[cfg(all(feature = "parity-db", not(feature = "rocksdb")))]
pub(crate) type Db = serai_db::ParityDb;
#[cfg(feature = "rocksdb")]
pub(crate) type Db = serai_db::RocksDB;
create_db!(
MainDb {
HandledMessageDb: (network: NetworkId) -> u64,
ActiveTributaryDb: () -> Vec<u8>,
RetiredTributaryDb: (set: ValidatorSet) -> (),
FirstPreprocessDb: (
network: NetworkId,
id_type: RecognizedIdType,
id: &[u8]
) -> Vec<Vec<u8>>,
LastReceivedBatchDb: (network: NetworkId) -> u32,
ExpectedBatchDb: (network: NetworkId, id: u32) -> [u8; 32],
BatchDb: (network: NetworkId, id: u32) -> SignedBatch,
LastVerifiedBatchDb: (network: NetworkId) -> u32,
HandoverBatchDb: (set: ValidatorSet) -> u32,
LookupHandoverBatchDb: (network: NetworkId, batch: u32) -> Session,
QueuedBatchesDb: (set: ValidatorSet) -> Vec<u8>
}
);
impl ActiveTributaryDb {
pub fn active_tributaries<G: Get>(getter: &G) -> (Vec<u8>, Vec<TributarySpec>) {
let bytes = Self::get(getter).unwrap_or_default();
let mut bytes_ref: &[u8] = bytes.as_ref();
let mut tributaries = vec![];
while !bytes_ref.is_empty() {
tributaries.push(TributarySpec::deserialize_reader(&mut bytes_ref).unwrap());
}
(bytes, tributaries)
#[allow(unused_variables, unreachable_code)]
fn db(path: &str) -> Db {
{
let path: &Path = path.as_ref();
// This may error if this path already exists, which we shouldn't propagate/panic on. If this
// is a problem (such as we don't have the necessary permissions to write to this path), we
// expect the following DB opening to error.
let _: Result<_, _> = fs::create_dir_all(path.parent().unwrap());
}
pub fn add_participating_in_tributary(txn: &mut impl DbTxn, spec: &TributarySpec) {
let (mut existing_bytes, existing) = ActiveTributaryDb::active_tributaries(txn);
for tributary in &existing {
if tributary == spec {
return;
}
}
#[cfg(all(feature = "parity-db", feature = "rocksdb"))]
panic!("built with parity-db and rocksdb");
#[cfg(all(feature = "parity-db", not(feature = "rocksdb")))]
let db = serai_db::new_parity_db(path);
#[cfg(feature = "rocksdb")]
let db = serai_db::new_rocksdb(path);
db
}
spec.serialize(&mut existing_bytes).unwrap();
ActiveTributaryDb::set(txn, &existing_bytes);
}
pub(crate) fn coordinator_db() -> Db {
let root_path = serai_env::var("DB_PATH").expect("path to DB wasn't specified");
db(&format!("{root_path}/coordinator/db"))
}
pub fn retire_tributary(txn: &mut impl DbTxn, set: ValidatorSet) {
let mut active = Self::active_tributaries(txn).1;
for i in 0 .. active.len() {
if active[i].set() == set {
active.remove(i);
break;
}
}
fn tributary_db_folder(set: ValidatorSet) -> String {
let root_path = serai_env::var("DB_PATH").expect("path to DB wasn't specified");
let network = match set.network {
NetworkId::Serai => panic!("creating Tributary for the Serai network"),
NetworkId::Bitcoin => "Bitcoin",
NetworkId::Ethereum => "Ethereum",
NetworkId::Monero => "Monero",
};
format!("{root_path}/tributary-{network}-{}", set.session.0)
}
let mut bytes = vec![];
for active in active {
active.serialize(&mut bytes).unwrap();
}
Self::set(txn, &bytes);
RetiredTributaryDb::set(txn, set, &());
pub(crate) fn tributary_db(set: ValidatorSet) -> Db {
db(&format!("{}/db", tributary_db_folder(set)))
}
pub(crate) fn prune_tributary_db(set: ValidatorSet) {
log::info!("pruning data directory for tributary {set:?}");
let db = tributary_db_folder(set);
if fs::exists(&db).expect("couldn't check if tributary DB exists") {
fs::remove_dir_all(db).unwrap();
}
}
impl FirstPreprocessDb {
pub fn save_first_preprocess(
txn: &mut impl DbTxn,
network: NetworkId,
id_type: RecognizedIdType,
id: &[u8],
preprocess: &Vec<Vec<u8>>,
) {
if let Some(existing) = FirstPreprocessDb::get(txn, network, id_type, id) {
assert_eq!(&existing, preprocess, "saved a distinct first preprocess");
return;
create_db! {
Coordinator {
// The currently active Tributaries
ActiveTributaries: () -> Vec<NewSetInformation>,
// The latest Tributary to have been retired for a network
// Since Tributaries are retired sequentially, this is informative to if any Tributary has been
// retired
RetiredTributary: (network: NetworkId) -> Session,
// The last handled message from a Processor
LastProcessorMessage: (network: NetworkId) -> u64,
// Cosigns we produced and tried to intake yet incurred an error while doing so
ErroneousCosigns: () -> Vec<SignedCosign>,
}
}
db_channel! {
Coordinator {
// Cosigns we produced
SignedCosigns: () -> SignedCosign,
// Tributaries to clean up upon reboot
TributaryCleanup: () -> ValidatorSet,
}
}
mod _internal_db {
use super::*;
db_channel! {
Coordinator {
// Tributary transactions to publish
TributaryTransactions: (set: ValidatorSet) -> Transaction,
}
FirstPreprocessDb::set(txn, network, id_type, id, preprocess);
}
}
impl ExpectedBatchDb {
pub fn save_expected_batch(txn: &mut impl DbTxn, batch: &Batch) {
LastReceivedBatchDb::set(txn, batch.network, &batch.id);
Self::set(
txn,
batch.network,
batch.id,
&Blake2b::<U32>::digest(batch.instructions.encode()).into(),
);
}
}
impl HandoverBatchDb {
pub fn set_handover_batch(txn: &mut impl DbTxn, set: ValidatorSet, batch: u32) {
Self::set(txn, set, &batch);
LookupHandoverBatchDb::set(txn, set.network, batch, &set.session);
}
}
impl QueuedBatchesDb {
pub fn queue(txn: &mut impl DbTxn, set: ValidatorSet, batch: &Transaction) {
let mut batches = Self::get(txn, set).unwrap_or_default();
batch.write(&mut batches).unwrap();
Self::set(txn, set, &batches);
}
pub fn take(txn: &mut impl DbTxn, set: ValidatorSet) -> Vec<Transaction> {
let batches_vec = Self::get(txn, set).unwrap_or_default();
txn.del(Self::key(set));
let mut batches: &[u8] = &batches_vec;
let mut res = vec![];
while !batches.is_empty() {
res.push(Transaction::read(&mut batches).unwrap());
pub(crate) struct TributaryTransactions;
impl TributaryTransactions {
pub(crate) fn send(txn: &mut impl DbTxn, set: ValidatorSet, tx: &Transaction) {
// If this set has yet to be retired, send this transaction
if RetiredTributary::get(txn, set.network).map(|session| session.0) < Some(set.session.0) {
_internal_db::TributaryTransactions::send(txn, set, tx);
}
res
}
pub(crate) fn try_recv(txn: &mut impl DbTxn, set: ValidatorSet) -> Option<Transaction> {
_internal_db::TributaryTransactions::try_recv(txn, set)
}
}

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@@ -1,46 +0,0 @@
use std::sync::Arc;
use serai_client::primitives::NetworkId;
use processor_messages::{ProcessorMessage, CoordinatorMessage};
use message_queue::{Service, Metadata, client::MessageQueue};
#[derive(Clone, PartialEq, Eq, Debug)]
pub struct Message {
pub id: u64,
pub network: NetworkId,
pub msg: ProcessorMessage,
}
#[async_trait::async_trait]
pub trait Processors: 'static + Send + Sync + Clone {
async fn send(&self, network: NetworkId, msg: impl Send + Into<CoordinatorMessage>);
async fn recv(&self, network: NetworkId) -> Message;
async fn ack(&self, msg: Message);
}
#[async_trait::async_trait]
impl Processors for Arc<MessageQueue> {
async fn send(&self, network: NetworkId, msg: impl Send + Into<CoordinatorMessage>) {
let msg: CoordinatorMessage = msg.into();
let metadata =
Metadata { from: self.service, to: Service::Processor(network), intent: msg.intent() };
let msg = borsh::to_vec(&msg).unwrap();
self.queue(metadata, msg).await;
}
async fn recv(&self, network: NetworkId) -> Message {
let msg = self.next(Service::Processor(network)).await;
assert_eq!(msg.from, Service::Processor(network));
let id = msg.id;
// Deserialize it into a ProcessorMessage
let msg: ProcessorMessage =
borsh::from_slice(&msg.msg).expect("message wasn't a borsh-encoded ProcessorMessage");
return Message { id, network, msg };
}
async fn ack(&self, msg: Message) {
MessageQueue::ack(self, Service::Processor(msg.network), msg.id).await
}
}

View File

@@ -0,0 +1,159 @@
use core::future::Future;
use std::sync::Arc;
use zeroize::Zeroizing;
use ciphersuite::{Ciphersuite, Ristretto};
use tokio::sync::mpsc;
use serai_db::{DbTxn, Db as DbTrait};
use serai_client::validator_sets::primitives::{Session, ValidatorSet};
use message_queue::{Service, Metadata, client::MessageQueue};
use tributary_sdk::Tributary;
use serai_task::ContinuallyRan;
use serai_coordinator_tributary::Transaction;
use serai_coordinator_p2p::P2p;
use crate::Db;
pub(crate) struct SubstrateTask<P: P2p> {
pub(crate) serai_key: Zeroizing<<Ristretto as Ciphersuite>::F>,
pub(crate) db: Db,
pub(crate) message_queue: Arc<MessageQueue>,
pub(crate) p2p: P,
pub(crate) p2p_add_tributary:
mpsc::UnboundedSender<(ValidatorSet, Tributary<Db, Transaction, P>)>,
pub(crate) p2p_retire_tributary: mpsc::UnboundedSender<ValidatorSet>,
}
impl<P: P2p> ContinuallyRan for SubstrateTask<P> {
type Error = String; // TODO
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
// Handle the Canonical events
for network in serai_client::primitives::NETWORKS {
loop {
let mut txn = self.db.txn();
let Some(msg) = serai_coordinator_substrate::Canonical::try_recv(&mut txn, network)
else {
break;
};
match msg {
// TODO: Stop trying to confirm the DKG
messages::substrate::CoordinatorMessage::SetKeys { .. } => todo!("TODO"),
messages::substrate::CoordinatorMessage::SlashesReported { session } => {
let prior_retired = crate::db::RetiredTributary::get(&txn, network);
let next_to_be_retired =
prior_retired.map(|session| Session(session.0 + 1)).unwrap_or(Session(0));
assert_eq!(session, next_to_be_retired);
crate::db::RetiredTributary::set(&mut txn, network, &session);
self
.p2p_retire_tributary
.send(ValidatorSet { network, session })
.expect("p2p retire_tributary channel dropped?");
}
messages::substrate::CoordinatorMessage::Block { .. } => {}
}
let msg = messages::CoordinatorMessage::from(msg);
let metadata = Metadata {
from: Service::Coordinator,
to: Service::Processor(network),
intent: msg.intent(),
};
let msg = borsh::to_vec(&msg).unwrap();
self.message_queue.queue(metadata, msg).await?;
txn.commit();
made_progress = true;
}
}
// Handle the NewSet events
loop {
let mut txn = self.db.txn();
let Some(new_set) = serai_coordinator_substrate::NewSet::try_recv(&mut txn) else { break };
if let Some(historic_session) = new_set.set.session.0.checked_sub(2) {
// We should have retired this session if we're here
if crate::db::RetiredTributary::get(&txn, new_set.set.network).map(|session| session.0) <
Some(historic_session)
{
/*
If we haven't, it's because we're processing the NewSet event before the retiry
event from the Canonical event stream. This happens if the Canonical event, and
then the NewSet event, is fired while we're already iterating over NewSet events.
We break, dropping the txn, restoring this NewSet to the database, so we'll only
handle it once a future iteration of this loop handles the retiry event.
*/
break;
}
/*
Queue this historical Tributary for deletion.
We explicitly don't queue this upon Tributary retire, instead here, to give time to
investigate retired Tributaries if questions are raised post-retiry. This gives a
week (the duration of the following session) after the Tributary has been retired to
make a backup of the data directory for any investigations.
*/
crate::db::TributaryCleanup::send(
&mut txn,
&ValidatorSet { network: new_set.set.network, session: Session(historic_session) },
);
}
// Save this Tributary as active to the database
{
let mut active_tributaries =
crate::db::ActiveTributaries::get(&txn).unwrap_or(Vec::with_capacity(1));
active_tributaries.push(new_set.clone());
crate::db::ActiveTributaries::set(&mut txn, &active_tributaries);
}
// Send GenerateKey to the processor
let msg = messages::key_gen::CoordinatorMessage::GenerateKey {
session: new_set.set.session,
threshold: new_set.threshold,
evrf_public_keys: new_set.evrf_public_keys.clone(),
};
let msg = messages::CoordinatorMessage::from(msg);
let metadata = Metadata {
from: Service::Coordinator,
to: Service::Processor(new_set.set.network),
intent: msg.intent(),
};
let msg = borsh::to_vec(&msg).unwrap();
self.message_queue.queue(metadata, msg).await?;
// Commit the transaction for all of this
txn.commit();
// Now spawn the Tributary
// If we reboot after committing the txn, but before this is called, this will be called
// on boot
crate::tributary::spawn_tributary(
self.db.clone(),
self.message_queue.clone(),
self.p2p.clone(),
&self.p2p_add_tributary,
new_set,
self.serai_key.clone(),
)
.await;
made_progress = true;
}
Ok(made_progress)
}
}
}

View File

@@ -1,332 +0,0 @@
/*
If:
A) This block has events and it's been at least X blocks since the last cosign or
B) This block doesn't have events but it's been X blocks since a skipped block which did
have events or
C) This block key gens (which changes who the cosigners are)
cosign this block.
This creates both a minimum and maximum delay of X blocks before a block's cosigning begins,
barring key gens which are exceptional. The minimum delay is there to ensure we don't constantly
spawn new protocols every 6 seconds, overwriting the old ones. The maximum delay is there to
ensure any block needing cosigned is consigned within a reasonable amount of time.
*/
use zeroize::Zeroizing;
use ciphersuite::{Ciphersuite, Ristretto};
use borsh::{BorshSerialize, BorshDeserialize};
use serai_client::{
SeraiError, Serai,
primitives::NetworkId,
validator_sets::primitives::{Session, ValidatorSet},
};
use serai_db::*;
use crate::{Db, substrate::in_set, tributary::SeraiBlockNumber};
// 5 minutes, expressed in blocks
// TODO: Pull a constant for block time
const COSIGN_DISTANCE: u64 = 5 * 60 / 6;
#[derive(Clone, Copy, PartialEq, Eq, Debug, BorshSerialize, BorshDeserialize)]
enum HasEvents {
KeyGen,
Yes,
No,
}
create_db!(
SubstrateCosignDb {
ScanCosignFrom: () -> u64,
IntendedCosign: () -> (u64, Option<u64>),
BlockHasEventsCache: (block: u64) -> HasEvents,
LatestCosignedBlock: () -> u64,
}
);
impl IntendedCosign {
// Sets the intended to cosign block, clearing the prior value entirely.
pub fn set_intended_cosign(txn: &mut impl DbTxn, intended: u64) {
Self::set(txn, &(intended, None::<u64>));
}
// Sets the cosign skipped since the last intended to cosign block.
pub fn set_skipped_cosign(txn: &mut impl DbTxn, skipped: u64) {
let (intended, prior_skipped) = Self::get(txn).unwrap();
assert!(prior_skipped.is_none());
Self::set(txn, &(intended, Some(skipped)));
}
}
impl LatestCosignedBlock {
pub fn latest_cosigned_block(getter: &impl Get) -> u64 {
Self::get(getter).unwrap_or_default().max(1)
}
}
db_channel! {
SubstrateDbChannels {
CosignTransactions: (network: NetworkId) -> (Session, u64, [u8; 32]),
}
}
impl CosignTransactions {
// Append a cosign transaction.
pub fn append_cosign(txn: &mut impl DbTxn, set: ValidatorSet, number: u64, hash: [u8; 32]) {
CosignTransactions::send(txn, set.network, &(set.session, number, hash))
}
}
async fn block_has_events(
txn: &mut impl DbTxn,
serai: &Serai,
block: u64,
) -> Result<HasEvents, SeraiError> {
let cached = BlockHasEventsCache::get(txn, block);
match cached {
None => {
let serai = serai.as_of(
serai
.finalized_block_by_number(block)
.await?
.expect("couldn't get block which should've been finalized")
.hash(),
);
if !serai.validator_sets().key_gen_events().await?.is_empty() {
return Ok(HasEvents::KeyGen);
}
let has_no_events = serai.coins().burn_with_instruction_events().await?.is_empty() &&
serai.in_instructions().batch_events().await?.is_empty() &&
serai.validator_sets().new_set_events().await?.is_empty() &&
serai.validator_sets().set_retired_events().await?.is_empty();
let has_events = if has_no_events { HasEvents::No } else { HasEvents::Yes };
BlockHasEventsCache::set(txn, block, &has_events);
Ok(has_events)
}
Some(code) => Ok(code),
}
}
async fn potentially_cosign_block(
txn: &mut impl DbTxn,
serai: &Serai,
block: u64,
skipped_block: Option<u64>,
window_end_exclusive: u64,
) -> Result<bool, SeraiError> {
// The following code regarding marking cosigned if prior block is cosigned expects this block to
// not be zero
// While we could perform this check there, there's no reason not to optimize the entire function
// as such
if block == 0 {
return Ok(false);
}
let block_has_events = block_has_events(txn, serai, block).await?;
// If this block had no events and immediately follows a cosigned block, mark it as cosigned
if (block_has_events == HasEvents::No) &&
(LatestCosignedBlock::latest_cosigned_block(txn) == (block - 1))
{
log::debug!("automatically co-signing next block ({block}) since it has no events");
LatestCosignedBlock::set(txn, &block);
}
// If we skipped a block, we're supposed to sign it plus the COSIGN_DISTANCE if no other blocks
// trigger a cosigning protocol covering it
// This means there will be the maximum delay allowed from a block needing cosigning occurring
// and a cosign for it triggering
let maximally_latent_cosign_block =
skipped_block.map(|skipped_block| skipped_block + COSIGN_DISTANCE);
// If this block is within the window,
if block < window_end_exclusive {
// and set a key, cosign it
if block_has_events == HasEvents::KeyGen {
IntendedCosign::set_intended_cosign(txn, block);
// Carry skipped if it isn't included by cosigning this block
if let Some(skipped) = skipped_block {
if skipped > block {
IntendedCosign::set_skipped_cosign(txn, block);
}
}
return Ok(true);
}
} else if (Some(block) == maximally_latent_cosign_block) || (block_has_events != HasEvents::No) {
// Since this block was outside the window and had events/was maximally latent, cosign it
IntendedCosign::set_intended_cosign(txn, block);
return Ok(true);
}
Ok(false)
}
/*
Advances the cosign protocol as should be done per the latest block.
A block is considered cosigned if:
A) It was cosigned
B) It's the parent of a cosigned block
C) It immediately follows a cosigned block and has no events requiring cosigning
This only actually performs advancement within a limited bound (generally until it finds a block
which should be cosigned). Accordingly, it is necessary to call multiple times even if
`latest_number` doesn't change.
*/
async fn advance_cosign_protocol_inner(
db: &mut impl Db,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
serai: &Serai,
latest_number: u64,
) -> Result<(), SeraiError> {
let mut txn = db.txn();
const INITIAL_INTENDED_COSIGN: u64 = 1;
let (last_intended_to_cosign_block, mut skipped_block) = {
let intended_cosign = IntendedCosign::get(&txn);
// If we haven't prior intended to cosign a block, set the intended cosign to 1
if let Some(intended_cosign) = intended_cosign {
intended_cosign
} else {
IntendedCosign::set_intended_cosign(&mut txn, INITIAL_INTENDED_COSIGN);
IntendedCosign::get(&txn).unwrap()
}
};
// "windows" refers to the window of blocks where even if there's a block which should be
// cosigned, it won't be due to proximity due to the prior cosign
let mut window_end_exclusive = last_intended_to_cosign_block + COSIGN_DISTANCE;
// If we've never triggered a cosign, don't skip any cosigns based on proximity
if last_intended_to_cosign_block == INITIAL_INTENDED_COSIGN {
window_end_exclusive = 1;
}
// The consensus rules for this are `last_intended_to_cosign_block + 1`
let scan_start_block = last_intended_to_cosign_block + 1;
// As a practical optimization, we don't re-scan old blocks since old blocks are independent to
// new state
let scan_start_block = scan_start_block.max(ScanCosignFrom::get(&txn).unwrap_or(1));
// Check all blocks within the window to see if they should be cosigned
// If so, we're skipping them and need to flag them as skipped so that once the window closes, we
// do cosign them
// We only perform this check if we haven't already marked a block as skipped since the cosign
// the skipped block will cause will cosign all other blocks within this window
if skipped_block.is_none() {
let window_end_inclusive = window_end_exclusive - 1;
for b in scan_start_block ..= window_end_inclusive.min(latest_number) {
if block_has_events(&mut txn, serai, b).await? == HasEvents::Yes {
skipped_block = Some(b);
log::debug!("skipping cosigning {b} due to proximity to prior cosign");
IntendedCosign::set_skipped_cosign(&mut txn, b);
break;
}
}
}
// A block which should be cosigned
let mut to_cosign = None;
// A list of sets which are cosigning, along with a boolean of if we're in the set
let mut cosigning = vec![];
for block in scan_start_block ..= latest_number {
let actual_block = serai
.finalized_block_by_number(block)
.await?
.expect("couldn't get block which should've been finalized");
// Save the block number for this block, as needed by the cosigner to perform cosigning
SeraiBlockNumber::set(&mut txn, actual_block.hash(), &block);
if potentially_cosign_block(&mut txn, serai, block, skipped_block, window_end_exclusive).await?
{
to_cosign = Some((block, actual_block.hash()));
// Get the keys as of the prior block
// If this key sets new keys, the coordinator won't acknowledge so until we process this
// block
// We won't process this block until its co-signed
// Using the keys of the prior block ensures this deadlock isn't reached
let serai = serai.as_of(actual_block.header.parent_hash.into());
for network in serai_client::primitives::NETWORKS {
// Get the latest session to have set keys
let set_with_keys = {
let Some(latest_session) = serai.validator_sets().session(network).await? else {
continue;
};
let prior_session = Session(latest_session.0.saturating_sub(1));
if serai
.validator_sets()
.keys(ValidatorSet { network, session: prior_session })
.await?
.is_some()
{
ValidatorSet { network, session: prior_session }
} else {
let set = ValidatorSet { network, session: latest_session };
if serai.validator_sets().keys(set).await?.is_none() {
continue;
}
set
}
};
log::debug!("{:?} will be cosigning {block}", set_with_keys.network);
cosigning.push((set_with_keys, in_set(key, &serai, set_with_keys).await?.unwrap()));
}
break;
}
// If this TX is committed, always start future scanning from the next block
ScanCosignFrom::set(&mut txn, &(block + 1));
// Since we're scanning *from* the next block, tidy the cache
BlockHasEventsCache::del(&mut txn, block);
}
if let Some((number, hash)) = to_cosign {
// If this block doesn't have cosigners, yet does have events, automatically mark it as
// cosigned
if cosigning.is_empty() {
log::debug!("{} had no cosigners available, marking as cosigned", number);
LatestCosignedBlock::set(&mut txn, &number);
} else {
for (set, in_set) in cosigning {
if in_set {
log::debug!("cosigning {number} with {:?} {:?}", set.network, set.session);
CosignTransactions::append_cosign(&mut txn, set, number, hash);
}
}
}
}
txn.commit();
Ok(())
}
pub async fn advance_cosign_protocol(
db: &mut impl Db,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
serai: &Serai,
latest_number: u64,
) -> Result<(), SeraiError> {
loop {
let scan_from = ScanCosignFrom::get(db).unwrap_or(1);
// Only scan 1000 blocks at a time to limit a massive txn from forming
let scan_to = latest_number.min(scan_from + 1000);
advance_cosign_protocol_inner(db, key, serai, scan_to).await?;
// If we didn't limit the scan_to, break
if scan_to == latest_number {
break;
}
}
Ok(())
}

View File

@@ -1,32 +0,0 @@
use serai_client::primitives::NetworkId;
pub use serai_db::*;
mod inner_db {
use super::*;
create_db!(
SubstrateDb {
NextBlock: () -> u64,
HandledEvent: (block: [u8; 32]) -> u32,
BatchInstructionsHashDb: (network: NetworkId, id: u32) -> [u8; 32]
}
);
}
pub(crate) use inner_db::{NextBlock, BatchInstructionsHashDb};
pub struct HandledEvent;
impl HandledEvent {
fn next_to_handle_event(getter: &impl Get, block: [u8; 32]) -> u32 {
inner_db::HandledEvent::get(getter, block).map_or(0, |last| last + 1)
}
pub fn is_unhandled(getter: &impl Get, block: [u8; 32], event_id: u32) -> bool {
let next = Self::next_to_handle_event(getter, block);
assert!(next >= event_id);
next == event_id
}
pub fn handle_event(txn: &mut impl DbTxn, block: [u8; 32], index: u32) {
assert!(Self::next_to_handle_event(txn, block) == index);
inner_db::HandledEvent::set(txn, block, &index);
}
}

View File

@@ -1,550 +0,0 @@
use core::{ops::Deref, time::Duration};
use std::{
sync::Arc,
collections::{HashSet, HashMap},
};
use zeroize::Zeroizing;
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use serai_client::{
SeraiError, Block, Serai, TemporalSerai,
primitives::{BlockHash, NetworkId},
validator_sets::{primitives::ValidatorSet, ValidatorSetsEvent},
in_instructions::InInstructionsEvent,
coins::CoinsEvent,
};
use serai_db::DbTxn;
use processor_messages::SubstrateContext;
use tokio::{sync::mpsc, time::sleep};
use crate::{
Db,
processors::Processors,
tributary::{TributarySpec, SeraiDkgCompleted},
};
mod db;
pub use db::*;
mod cosign;
pub use cosign::*;
async fn in_set(
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
serai: &TemporalSerai<'_>,
set: ValidatorSet,
) -> Result<Option<bool>, SeraiError> {
let Some(participants) = serai.validator_sets().participants(set.network).await? else {
return Ok(None);
};
let key = (Ristretto::generator() * key.deref()).to_bytes();
Ok(Some(participants.iter().any(|(participant, _)| participant.0 == key)))
}
async fn handle_new_set<D: Db>(
txn: &mut D::Transaction<'_>,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
new_tributary_spec: &mpsc::UnboundedSender<TributarySpec>,
serai: &Serai,
block: &Block,
set: ValidatorSet,
) -> Result<(), SeraiError> {
if in_set(key, &serai.as_of(block.hash()), set)
.await?
.expect("NewSet for set which doesn't exist")
{
log::info!("present in set {:?}", set);
let set_data = {
let serai = serai.as_of(block.hash());
let serai = serai.validator_sets();
let set_participants =
serai.participants(set.network).await?.expect("NewSet for set which doesn't exist");
set_participants.into_iter().map(|(k, w)| (k, u16::try_from(w).unwrap())).collect::<Vec<_>>()
};
let time = if let Ok(time) = block.time() {
time
} else {
assert_eq!(block.number(), 0);
// Use the next block's time
loop {
let Ok(Some(res)) = serai.finalized_block_by_number(1).await else {
sleep(Duration::from_secs(5)).await;
continue;
};
break res.time().unwrap();
}
};
// The block time is in milliseconds yet the Tributary is in seconds
let time = time / 1000;
// Since this block is in the past, and Tendermint doesn't play nice with starting chains after
// their start time (though it does eventually work), delay the start time by 120 seconds
// This is meant to handle ~20 blocks of lack of finalization for this first block
const SUBSTRATE_TO_TRIBUTARY_TIME_DELAY: u64 = 120;
let time = time + SUBSTRATE_TO_TRIBUTARY_TIME_DELAY;
let spec = TributarySpec::new(block.hash(), time, set, set_data);
log::info!("creating new tributary for {:?}", spec.set());
// Save it to the database now, not on the channel receiver's side, so this is safe against
// reboots
// If this txn finishes, and we reboot, then this'll be reloaded from active Tributaries
// If this txn doesn't finish, this will be re-fired
// If we waited to save to the DB, this txn may be finished, preventing re-firing, yet the
// prior fired event may have not been received yet
crate::ActiveTributaryDb::add_participating_in_tributary(txn, &spec);
new_tributary_spec.send(spec).unwrap();
} else {
log::info!("not present in new set {:?}", set);
}
Ok(())
}
async fn handle_batch_and_burns<Pro: Processors>(
txn: &mut impl DbTxn,
processors: &Pro,
serai: &Serai,
block: &Block,
) -> Result<(), SeraiError> {
// Track which networks had events with a Vec in ordr to preserve the insertion order
// While that shouldn't be needed, ensuring order never hurts, and may enable design choices
// with regards to Processor <-> Coordinator message passing
let mut networks_with_event = vec![];
let mut network_had_event = |burns: &mut HashMap<_, _>, batches: &mut HashMap<_, _>, network| {
// Don't insert this network multiple times
// A Vec is still used in order to maintain the insertion order
if !networks_with_event.contains(&network) {
networks_with_event.push(network);
burns.insert(network, vec![]);
batches.insert(network, vec![]);
}
};
let mut batch_block = HashMap::new();
let mut batches = HashMap::<NetworkId, Vec<u32>>::new();
let mut burns = HashMap::new();
let serai = serai.as_of(block.hash());
for batch in serai.in_instructions().batch_events().await? {
if let InInstructionsEvent::Batch { network, id, block: network_block, instructions_hash } =
batch
{
network_had_event(&mut burns, &mut batches, network);
BatchInstructionsHashDb::set(txn, network, id, &instructions_hash);
// Make sure this is the only Batch event for this network in this Block
assert!(batch_block.insert(network, network_block).is_none());
// Add the batch included by this block
batches.get_mut(&network).unwrap().push(id);
} else {
panic!("Batch event wasn't Batch: {batch:?}");
}
}
for burn in serai.coins().burn_with_instruction_events().await? {
if let CoinsEvent::BurnWithInstruction { from: _, instruction } = burn {
let network = instruction.balance.coin.network();
network_had_event(&mut burns, &mut batches, network);
// network_had_event should register an entry in burns
burns.get_mut(&network).unwrap().push(instruction);
} else {
panic!("Burn event wasn't Burn: {burn:?}");
}
}
assert_eq!(HashSet::<&_>::from_iter(networks_with_event.iter()).len(), networks_with_event.len());
for network in networks_with_event {
let network_latest_finalized_block = if let Some(block) = batch_block.remove(&network) {
block
} else {
// If it's had a batch or a burn, it must have had a block acknowledged
serai
.in_instructions()
.latest_block_for_network(network)
.await?
.expect("network had a batch/burn yet never set a latest block")
};
processors
.send(
network,
processor_messages::substrate::CoordinatorMessage::SubstrateBlock {
context: SubstrateContext {
serai_time: block.time().unwrap() / 1000,
network_latest_finalized_block,
},
block: block.number(),
burns: burns.remove(&network).unwrap(),
batches: batches.remove(&network).unwrap(),
},
)
.await;
}
Ok(())
}
// Handle a specific Substrate block, returning an error when it fails to get data
// (not blocking / holding)
#[allow(clippy::too_many_arguments)]
async fn handle_block<D: Db, Pro: Processors>(
db: &mut D,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
new_tributary_spec: &mpsc::UnboundedSender<TributarySpec>,
perform_slash_report: &mpsc::UnboundedSender<ValidatorSet>,
tributary_retired: &mpsc::UnboundedSender<ValidatorSet>,
processors: &Pro,
serai: &Serai,
block: Block,
) -> Result<(), SeraiError> {
let hash = block.hash();
// Define an indexed event ID.
let mut event_id = 0;
// If a new validator set was activated, create tributary/inform processor to do a DKG
for new_set in serai.as_of(hash).validator_sets().new_set_events().await? {
// Individually mark each event as handled so on reboot, we minimize duplicates
// Additionally, if the Serai connection also fails 1/100 times, this means a block with 1000
// events will successfully be incrementally handled
// (though the Serai connection should be stable, making this unnecessary)
let ValidatorSetsEvent::NewSet { set } = new_set else {
panic!("NewSet event wasn't NewSet: {new_set:?}");
};
// If this is Serai, do nothing
// We only coordinate/process external networks
if set.network == NetworkId::Serai {
continue;
}
if HandledEvent::is_unhandled(db, hash, event_id) {
log::info!("found fresh new set event {:?}", new_set);
let mut txn = db.txn();
handle_new_set::<D>(&mut txn, key, new_tributary_spec, serai, &block, set).await?;
HandledEvent::handle_event(&mut txn, hash, event_id);
txn.commit();
}
event_id += 1;
}
// If a key pair was confirmed, inform the processor
for key_gen in serai.as_of(hash).validator_sets().key_gen_events().await? {
if HandledEvent::is_unhandled(db, hash, event_id) {
log::info!("found fresh key gen event {:?}", key_gen);
let ValidatorSetsEvent::KeyGen { set, key_pair } = key_gen else {
panic!("KeyGen event wasn't KeyGen: {key_gen:?}");
};
let substrate_key = key_pair.0 .0;
processors
.send(
set.network,
processor_messages::substrate::CoordinatorMessage::ConfirmKeyPair {
context: SubstrateContext {
serai_time: block.time().unwrap() / 1000,
network_latest_finalized_block: serai
.as_of(block.hash())
.in_instructions()
.latest_block_for_network(set.network)
.await?
// The processor treats this as a magic value which will cause it to find a network
// block which has a time greater than or equal to the Serai time
.unwrap_or(BlockHash([0; 32])),
},
session: set.session,
key_pair,
},
)
.await;
// TODO: If we were in the set, yet were removed, drop the tributary
let mut txn = db.txn();
SeraiDkgCompleted::set(&mut txn, set, &substrate_key);
HandledEvent::handle_event(&mut txn, hash, event_id);
txn.commit();
}
event_id += 1;
}
for accepted_handover in serai.as_of(hash).validator_sets().accepted_handover_events().await? {
let ValidatorSetsEvent::AcceptedHandover { set } = accepted_handover else {
panic!("AcceptedHandover event wasn't AcceptedHandover: {accepted_handover:?}");
};
if set.network == NetworkId::Serai {
continue;
}
if HandledEvent::is_unhandled(db, hash, event_id) {
log::info!("found fresh accepted handover event {:?}", accepted_handover);
// TODO: This isn't atomic with the event handling
// Send a oneshot receiver so we can await the response?
perform_slash_report.send(set).unwrap();
let mut txn = db.txn();
HandledEvent::handle_event(&mut txn, hash, event_id);
txn.commit();
}
event_id += 1;
}
for retired_set in serai.as_of(hash).validator_sets().set_retired_events().await? {
let ValidatorSetsEvent::SetRetired { set } = retired_set else {
panic!("SetRetired event wasn't SetRetired: {retired_set:?}");
};
if set.network == NetworkId::Serai {
continue;
}
if HandledEvent::is_unhandled(db, hash, event_id) {
log::info!("found fresh set retired event {:?}", retired_set);
let mut txn = db.txn();
crate::ActiveTributaryDb::retire_tributary(&mut txn, set);
tributary_retired.send(set).unwrap();
HandledEvent::handle_event(&mut txn, hash, event_id);
txn.commit();
}
event_id += 1;
}
// Finally, tell the processor of acknowledged blocks/burns
// This uses a single event as unlike prior events which individually executed code, all
// following events share data collection
if HandledEvent::is_unhandled(db, hash, event_id) {
let mut txn = db.txn();
handle_batch_and_burns(&mut txn, processors, serai, &block).await?;
HandledEvent::handle_event(&mut txn, hash, event_id);
txn.commit();
}
Ok(())
}
#[allow(clippy::too_many_arguments)]
async fn handle_new_blocks<D: Db, Pro: Processors>(
db: &mut D,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
new_tributary_spec: &mpsc::UnboundedSender<TributarySpec>,
perform_slash_report: &mpsc::UnboundedSender<ValidatorSet>,
tributary_retired: &mpsc::UnboundedSender<ValidatorSet>,
processors: &Pro,
serai: &Serai,
next_block: &mut u64,
) -> Result<(), SeraiError> {
// Check if there's been a new Substrate block
let latest_number = serai.latest_finalized_block().await?.number();
// Advance the cosigning protocol
advance_cosign_protocol(db, key, serai, latest_number).await?;
// Reduce to the latest cosigned block
let latest_number = latest_number.min(LatestCosignedBlock::latest_cosigned_block(db));
if latest_number < *next_block {
return Ok(());
}
for b in *next_block ..= latest_number {
let block = serai
.finalized_block_by_number(b)
.await?
.expect("couldn't get block before the latest finalized block");
log::info!("handling substrate block {b}");
handle_block(
db,
key,
new_tributary_spec,
perform_slash_report,
tributary_retired,
processors,
serai,
block,
)
.await?;
*next_block += 1;
let mut txn = db.txn();
NextBlock::set(&mut txn, next_block);
txn.commit();
log::info!("handled substrate block {b}");
}
Ok(())
}
pub async fn scan_task<D: Db, Pro: Processors>(
mut db: D,
key: Zeroizing<<Ristretto as Ciphersuite>::F>,
processors: Pro,
serai: Arc<Serai>,
new_tributary_spec: mpsc::UnboundedSender<TributarySpec>,
perform_slash_report: mpsc::UnboundedSender<ValidatorSet>,
tributary_retired: mpsc::UnboundedSender<ValidatorSet>,
) {
log::info!("scanning substrate");
let mut next_substrate_block = NextBlock::get(&db).unwrap_or_default();
/*
let new_substrate_block_notifier = {
let serai = &serai;
move || async move {
loop {
match serai.newly_finalized_block().await {
Ok(sub) => return sub,
Err(e) => {
log::error!("couldn't communicate with serai node: {e}");
sleep(Duration::from_secs(5)).await;
}
}
}
}
};
*/
// TODO: Restore the above subscription-based system
// That would require moving serai-client from HTTP to websockets
let new_substrate_block_notifier = {
let serai = &serai;
move |next_substrate_block| async move {
loop {
match serai.latest_finalized_block().await {
Ok(latest) => {
if latest.header.number >= next_substrate_block {
return latest;
}
sleep(Duration::from_secs(3)).await;
}
Err(e) => {
log::error!("couldn't communicate with serai node: {e}");
sleep(Duration::from_secs(5)).await;
}
}
}
}
};
loop {
// await the next block, yet if our notifier had an error, re-create it
{
let Ok(_) = tokio::time::timeout(
Duration::from_secs(60),
new_substrate_block_notifier(next_substrate_block),
)
.await
else {
// Timed out, which may be because Serai isn't finalizing or may be some issue with the
// notifier
if serai.latest_finalized_block().await.map(|block| block.number()).ok() ==
Some(next_substrate_block.saturating_sub(1))
{
log::info!("serai hasn't finalized a block in the last 60s...");
}
continue;
};
/*
// next_block is a Option<Result>
if next_block.and_then(Result::ok).is_none() {
substrate_block_notifier = new_substrate_block_notifier(next_substrate_block);
continue;
}
*/
}
match handle_new_blocks(
&mut db,
&key,
&new_tributary_spec,
&perform_slash_report,
&tributary_retired,
&processors,
&serai,
&mut next_substrate_block,
)
.await
{
Ok(()) => {}
Err(e) => {
log::error!("couldn't communicate with serai node: {e}");
sleep(Duration::from_secs(5)).await;
}
}
}
}
/// Gets the expected ID for the next Batch.
///
/// Will log an error and apply a slight sleep on error, letting the caller simply immediately
/// retry.
pub(crate) async fn expected_next_batch(
serai: &Serai,
network: NetworkId,
) -> Result<u32, SeraiError> {
async fn expected_next_batch_inner(serai: &Serai, network: NetworkId) -> Result<u32, SeraiError> {
let serai = serai.as_of_latest_finalized_block().await?;
let last = serai.in_instructions().last_batch_for_network(network).await?;
Ok(if let Some(last) = last { last + 1 } else { 0 })
}
match expected_next_batch_inner(serai, network).await {
Ok(next) => Ok(next),
Err(e) => {
log::error!("couldn't get the expected next batch from substrate: {e:?}");
sleep(Duration::from_millis(100)).await;
Err(e)
}
}
}
/// Verifies `Batch`s which have already been indexed from Substrate.
///
/// Spins if a distinct `Batch` is detected on-chain.
///
/// This has a slight malleability in that doesn't verify *who* published a `Batch` is as expected.
/// This is deemed fine.
pub(crate) async fn verify_published_batches<D: Db>(
txn: &mut D::Transaction<'_>,
network: NetworkId,
optimistic_up_to: u32,
) -> Option<u32> {
// TODO: Localize from MainDb to SubstrateDb
let last = crate::LastVerifiedBatchDb::get(txn, network);
for id in last.map_or(0, |last| last + 1) ..= optimistic_up_to {
let Some(on_chain) = BatchInstructionsHashDb::get(txn, network, id) else {
break;
};
let off_chain = crate::ExpectedBatchDb::get(txn, network, id).unwrap();
if on_chain != off_chain {
// Halt operations on this network and spin, as this is a critical fault
loop {
log::error!(
"{}! network: {:?} id: {} off-chain: {} on-chain: {}",
"on-chain batch doesn't match off-chain",
network,
id,
hex::encode(off_chain),
hex::encode(on_chain),
);
sleep(Duration::from_secs(60)).await;
}
}
crate::LastVerifiedBatchDb::set(txn, network, &id);
}
crate::LastVerifiedBatchDb::get(txn, network)
}

View File

@@ -1,125 +0,0 @@
use core::fmt::Debug;
use std::{
sync::Arc,
collections::{VecDeque, HashSet, HashMap},
};
use serai_client::{primitives::NetworkId, validator_sets::primitives::ValidatorSet};
use processor_messages::CoordinatorMessage;
use async_trait::async_trait;
use tokio::sync::RwLock;
use crate::{
processors::{Message, Processors},
TributaryP2p, ReqResMessageKind, GossipMessageKind, P2pMessageKind, Message as P2pMessage, P2p,
};
pub mod tributary;
#[derive(Clone)]
pub struct MemProcessors(pub Arc<RwLock<HashMap<NetworkId, VecDeque<CoordinatorMessage>>>>);
impl MemProcessors {
#[allow(clippy::new_without_default)]
pub fn new() -> MemProcessors {
MemProcessors(Arc::new(RwLock::new(HashMap::new())))
}
}
#[async_trait::async_trait]
impl Processors for MemProcessors {
async fn send(&self, network: NetworkId, msg: impl Send + Into<CoordinatorMessage>) {
let mut processors = self.0.write().await;
let processor = processors.entry(network).or_insert_with(VecDeque::new);
processor.push_back(msg.into());
}
async fn recv(&self, _: NetworkId) -> Message {
todo!()
}
async fn ack(&self, _: Message) {
todo!()
}
}
#[allow(clippy::type_complexity)]
#[derive(Clone, Debug)]
pub struct LocalP2p(
usize,
pub Arc<RwLock<(HashSet<Vec<u8>>, Vec<VecDeque<(usize, P2pMessageKind, Vec<u8>)>>)>>,
);
impl LocalP2p {
pub fn new(validators: usize) -> Vec<LocalP2p> {
let shared = Arc::new(RwLock::new((HashSet::new(), vec![VecDeque::new(); validators])));
let mut res = vec![];
for i in 0 .. validators {
res.push(LocalP2p(i, shared.clone()));
}
res
}
}
#[async_trait]
impl P2p for LocalP2p {
type Id = usize;
async fn subscribe(&self, _set: ValidatorSet, _genesis: [u8; 32]) {}
async fn unsubscribe(&self, _set: ValidatorSet, _genesis: [u8; 32]) {}
async fn send_raw(&self, to: Self::Id, msg: Vec<u8>) {
let mut msg_ref = msg.as_slice();
let kind = ReqResMessageKind::read(&mut msg_ref).unwrap();
self.1.write().await.1[to].push_back((self.0, P2pMessageKind::ReqRes(kind), msg_ref.to_vec()));
}
async fn broadcast_raw(&self, kind: P2pMessageKind, msg: Vec<u8>) {
// Content-based deduplication
let mut lock = self.1.write().await;
{
let already_sent = &mut lock.0;
if already_sent.contains(&msg) {
return;
}
already_sent.insert(msg.clone());
}
let queues = &mut lock.1;
let kind_len = (match kind {
P2pMessageKind::ReqRes(kind) => kind.serialize(),
P2pMessageKind::Gossip(kind) => kind.serialize(),
})
.len();
let msg = msg[kind_len ..].to_vec();
for (i, msg_queue) in queues.iter_mut().enumerate() {
if i == self.0 {
continue;
}
msg_queue.push_back((self.0, kind, msg.clone()));
}
}
async fn receive(&self) -> P2pMessage<Self> {
// This is a cursed way to implement an async read from a Vec
loop {
if let Some((sender, kind, msg)) = self.1.write().await.1[self.0].pop_front() {
return P2pMessage { sender, kind, msg };
}
tokio::time::sleep(std::time::Duration::from_millis(100)).await;
}
}
}
#[async_trait]
impl TributaryP2p for LocalP2p {
async fn broadcast(&self, genesis: [u8; 32], msg: Vec<u8>) {
<Self as P2p>::broadcast(
self,
P2pMessageKind::Gossip(GossipMessageKind::Tributary(genesis)),
msg,
)
.await
}
}

View File

@@ -1,237 +0,0 @@
use std::{
time::{Duration, SystemTime},
collections::HashSet,
};
use zeroize::Zeroizing;
use rand_core::{RngCore, CryptoRng, OsRng};
use futures_util::{task::Poll, poll};
use ciphersuite::{
group::{ff::Field, GroupEncoding},
Ciphersuite, Ristretto,
};
use sp_application_crypto::sr25519;
use borsh::BorshDeserialize;
use serai_client::{
primitives::NetworkId,
validator_sets::primitives::{Session, ValidatorSet},
};
use tokio::time::sleep;
use serai_db::MemDb;
use tributary::Tributary;
use crate::{
GossipMessageKind, P2pMessageKind, P2p,
tributary::{Transaction, TributarySpec},
tests::LocalP2p,
};
pub fn new_keys<R: RngCore + CryptoRng>(
rng: &mut R,
) -> Vec<Zeroizing<<Ristretto as Ciphersuite>::F>> {
let mut keys = vec![];
for _ in 0 .. 5 {
keys.push(Zeroizing::new(<Ristretto as Ciphersuite>::F::random(&mut *rng)));
}
keys
}
pub fn new_spec<R: RngCore + CryptoRng>(
rng: &mut R,
keys: &[Zeroizing<<Ristretto as Ciphersuite>::F>],
) -> TributarySpec {
let mut serai_block = [0; 32];
rng.fill_bytes(&mut serai_block);
let start_time = SystemTime::now().duration_since(SystemTime::UNIX_EPOCH).unwrap().as_secs();
let set = ValidatorSet { session: Session(0), network: NetworkId::Bitcoin };
let set_participants = keys
.iter()
.map(|key| (sr25519::Public((<Ristretto as Ciphersuite>::generator() * **key).to_bytes()), 1))
.collect::<Vec<_>>();
let res = TributarySpec::new(serai_block, start_time, set, set_participants);
assert_eq!(
TributarySpec::deserialize_reader(&mut borsh::to_vec(&res).unwrap().as_slice()).unwrap(),
res,
);
res
}
pub async fn new_tributaries(
keys: &[Zeroizing<<Ristretto as Ciphersuite>::F>],
spec: &TributarySpec,
) -> Vec<(MemDb, LocalP2p, Tributary<MemDb, Transaction, LocalP2p>)> {
let p2p = LocalP2p::new(keys.len());
let mut res = vec![];
for (i, key) in keys.iter().enumerate() {
let db = MemDb::new();
res.push((
db.clone(),
p2p[i].clone(),
Tributary::<_, Transaction, _>::new(
db,
spec.genesis(),
spec.start_time(),
key.clone(),
spec.validators(),
p2p[i].clone(),
)
.await
.unwrap(),
));
}
res
}
pub async fn run_tributaries(
mut tributaries: Vec<(LocalP2p, Tributary<MemDb, Transaction, LocalP2p>)>,
) {
loop {
for (p2p, tributary) in &mut tributaries {
while let Poll::Ready(msg) = poll!(p2p.receive()) {
match msg.kind {
P2pMessageKind::Gossip(GossipMessageKind::Tributary(genesis)) => {
assert_eq!(genesis, tributary.genesis());
if tributary.handle_message(&msg.msg).await {
p2p.broadcast(msg.kind, msg.msg).await;
}
}
_ => panic!("unexpected p2p message found"),
}
}
}
sleep(Duration::from_millis(100)).await;
}
}
pub async fn wait_for_tx_inclusion(
tributary: &Tributary<MemDb, Transaction, LocalP2p>,
mut last_checked: [u8; 32],
hash: [u8; 32],
) -> [u8; 32] {
let reader = tributary.reader();
loop {
let tip = tributary.tip().await;
if tip == last_checked {
sleep(Duration::from_secs(1)).await;
continue;
}
let mut queue = vec![reader.block(&tip).unwrap()];
let mut block = None;
while {
let parent = queue.last().unwrap().parent();
if parent == tributary.genesis() {
false
} else {
block = Some(reader.block(&parent).unwrap());
block.as_ref().unwrap().hash() != last_checked
}
} {
queue.push(block.take().unwrap());
}
while let Some(block) = queue.pop() {
for tx in &block.transactions {
if tx.hash() == hash {
return block.hash();
}
}
}
last_checked = tip;
}
}
#[tokio::test]
async fn tributary_test() {
let keys = new_keys(&mut OsRng);
let spec = new_spec(&mut OsRng, &keys);
let mut tributaries = new_tributaries(&keys, &spec)
.await
.into_iter()
.map(|(_, p2p, tributary)| (p2p, tributary))
.collect::<Vec<_>>();
let mut blocks = 0;
let mut last_block = spec.genesis();
// Doesn't use run_tributaries as we want to wind these down at a certain point
// run_tributaries will run them ad infinitum
let timeout = SystemTime::now() + Duration::from_secs(65);
while (blocks < 10) && (SystemTime::now().duration_since(timeout).is_err()) {
for (p2p, tributary) in &mut tributaries {
while let Poll::Ready(msg) = poll!(p2p.receive()) {
match msg.kind {
P2pMessageKind::Gossip(GossipMessageKind::Tributary(genesis)) => {
assert_eq!(genesis, tributary.genesis());
tributary.handle_message(&msg.msg).await;
}
_ => panic!("unexpected p2p message found"),
}
}
}
let tip = tributaries[0].1.tip().await;
if tip != last_block {
last_block = tip;
blocks += 1;
}
sleep(Duration::from_millis(100)).await;
}
if blocks != 10 {
panic!("tributary chain test hit timeout");
}
// Handle all existing messages
for (p2p, tributary) in &mut tributaries {
while let Poll::Ready(msg) = poll!(p2p.receive()) {
match msg.kind {
P2pMessageKind::Gossip(GossipMessageKind::Tributary(genesis)) => {
assert_eq!(genesis, tributary.genesis());
tributary.handle_message(&msg.msg).await;
}
_ => panic!("unexpected p2p message found"),
}
}
}
// handle_message informed the Tendermint machine, yet it still has to process it
// Sleep for a second accordingly
// TODO: Is there a better way to handle this?
sleep(Duration::from_secs(1)).await;
// All tributaries should agree on the tip, within a block
let mut tips = HashSet::new();
for (_, tributary) in &tributaries {
tips.insert(tributary.tip().await);
}
assert!(tips.len() <= 2);
if tips.len() == 2 {
for tip in &tips {
// Find a Tributary where this isn't the tip
for (_, tributary) in &tributaries {
let Some(after) = tributary.reader().block_after(tip) else { continue };
// Make sure the block after is the other tip
assert!(tips.contains(&after));
return;
}
}
} else {
assert_eq!(tips.len(), 1);
return;
}
panic!("tributary had different tip with a variance exceeding one block");
}

View File

@@ -1,392 +0,0 @@
use core::time::Duration;
use std::collections::HashMap;
use zeroize::Zeroizing;
use rand_core::{RngCore, OsRng};
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use frost::Participant;
use sp_runtime::traits::Verify;
use serai_client::{
primitives::{SeraiAddress, Signature},
validator_sets::primitives::{ValidatorSet, KeyPair},
};
use tokio::time::sleep;
use serai_db::{Get, DbTxn, Db, MemDb};
use processor_messages::{
key_gen::{self, KeyGenId},
CoordinatorMessage,
};
use tributary::{TransactionTrait, Tributary};
use crate::{
tributary::{
Transaction, TributarySpec,
scanner::{PublishSeraiTransaction, handle_new_blocks},
},
tests::{
MemProcessors, LocalP2p,
tributary::{new_keys, new_spec, new_tributaries, run_tributaries, wait_for_tx_inclusion},
},
};
#[tokio::test]
async fn dkg_test() {
env_logger::init();
let keys = new_keys(&mut OsRng);
let spec = new_spec(&mut OsRng, &keys);
let full_tributaries = new_tributaries(&keys, &spec).await;
let mut dbs = vec![];
let mut tributaries = vec![];
for (db, p2p, tributary) in full_tributaries {
dbs.push(db);
tributaries.push((p2p, tributary));
}
// Run the tributaries in the background
tokio::spawn(run_tributaries(tributaries.clone()));
let mut txs = vec![];
// Create DKG commitments for each key
for key in &keys {
let attempt = 0;
let mut commitments = vec![0; 256];
OsRng.fill_bytes(&mut commitments);
let mut tx = Transaction::DkgCommitments {
attempt,
commitments: vec![commitments],
signed: Transaction::empty_signed(),
};
tx.sign(&mut OsRng, spec.genesis(), key);
txs.push(tx);
}
let block_before_tx = tributaries[0].1.tip().await;
// Publish all commitments but one
for (i, tx) in txs.iter().enumerate().skip(1) {
assert_eq!(tributaries[i].1.add_transaction(tx.clone()).await, Ok(true));
}
// Wait until these are included
for tx in txs.iter().skip(1) {
wait_for_tx_inclusion(&tributaries[0].1, block_before_tx, tx.hash()).await;
}
let expected_commitments: HashMap<_, _> = txs
.iter()
.enumerate()
.map(|(i, tx)| {
if let Transaction::DkgCommitments { commitments, .. } = tx {
(Participant::new((i + 1).try_into().unwrap()).unwrap(), commitments[0].clone())
} else {
panic!("txs had non-commitments");
}
})
.collect();
async fn new_processors(
db: &mut MemDb,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
spec: &TributarySpec,
tributary: &Tributary<MemDb, Transaction, LocalP2p>,
) -> MemProcessors {
let processors = MemProcessors::new();
handle_new_blocks::<_, _, _, _, _, LocalP2p>(
db,
key,
&|_, _, _, _| async {
panic!("provided TX caused recognized_id to be called in new_processors")
},
&processors,
&(),
&|_| async {
panic!(
"test tried to publish a new Tributary TX from handle_application_tx in new_processors"
)
},
spec,
&tributary.reader(),
)
.await;
processors
}
// Instantiate a scanner and verify it has nothing to report
let processors = new_processors(&mut dbs[0], &keys[0], &spec, &tributaries[0].1).await;
assert!(processors.0.read().await.is_empty());
// Publish the last commitment
let block_before_tx = tributaries[0].1.tip().await;
assert_eq!(tributaries[0].1.add_transaction(txs[0].clone()).await, Ok(true));
wait_for_tx_inclusion(&tributaries[0].1, block_before_tx, txs[0].hash()).await;
sleep(Duration::from_secs(Tributary::<MemDb, Transaction, LocalP2p>::block_time().into())).await;
// Verify the scanner emits a KeyGen::Commitments message
handle_new_blocks::<_, _, _, _, _, LocalP2p>(
&mut dbs[0],
&keys[0],
&|_, _, _, _| async {
panic!("provided TX caused recognized_id to be called after Commitments")
},
&processors,
&(),
&|_| async {
panic!(
"test tried to publish a new Tributary TX from handle_application_tx after Commitments"
)
},
&spec,
&tributaries[0].1.reader(),
)
.await;
{
let mut msgs = processors.0.write().await;
assert_eq!(msgs.len(), 1);
let msgs = msgs.get_mut(&spec.set().network).unwrap();
let mut expected_commitments = expected_commitments.clone();
expected_commitments.remove(&Participant::new((1).try_into().unwrap()).unwrap());
assert_eq!(
msgs.pop_front().unwrap(),
CoordinatorMessage::KeyGen(key_gen::CoordinatorMessage::Commitments {
id: KeyGenId { session: spec.set().session, attempt: 0 },
commitments: expected_commitments
})
);
assert!(msgs.is_empty());
}
// Verify all keys exhibit this scanner behavior
for (i, key) in keys.iter().enumerate().skip(1) {
let processors = new_processors(&mut dbs[i], key, &spec, &tributaries[i].1).await;
let mut msgs = processors.0.write().await;
assert_eq!(msgs.len(), 1);
let msgs = msgs.get_mut(&spec.set().network).unwrap();
let mut expected_commitments = expected_commitments.clone();
expected_commitments.remove(&Participant::new((i + 1).try_into().unwrap()).unwrap());
assert_eq!(
msgs.pop_front().unwrap(),
CoordinatorMessage::KeyGen(key_gen::CoordinatorMessage::Commitments {
id: KeyGenId { session: spec.set().session, attempt: 0 },
commitments: expected_commitments
})
);
assert!(msgs.is_empty());
}
// Now do shares
let mut txs = vec![];
for (k, key) in keys.iter().enumerate() {
let attempt = 0;
let mut shares = vec![vec![]];
for i in 0 .. keys.len() {
if i != k {
let mut share = vec![0; 256];
OsRng.fill_bytes(&mut share);
shares.last_mut().unwrap().push(share);
}
}
let mut txn = dbs[k].txn();
let mut tx = Transaction::DkgShares {
attempt,
shares,
confirmation_nonces: crate::tributary::dkg_confirmation_nonces(key, &spec, &mut txn, 0),
signed: Transaction::empty_signed(),
};
txn.commit();
tx.sign(&mut OsRng, spec.genesis(), key);
txs.push(tx);
}
let block_before_tx = tributaries[0].1.tip().await;
for (i, tx) in txs.iter().enumerate().skip(1) {
assert_eq!(tributaries[i].1.add_transaction(tx.clone()).await, Ok(true));
}
for tx in txs.iter().skip(1) {
wait_for_tx_inclusion(&tributaries[0].1, block_before_tx, tx.hash()).await;
}
// With just 4 sets of shares, nothing should happen yet
handle_new_blocks::<_, _, _, _, _, LocalP2p>(
&mut dbs[0],
&keys[0],
&|_, _, _, _| async {
panic!("provided TX caused recognized_id to be called after some shares")
},
&processors,
&(),
&|_| async {
panic!(
"test tried to publish a new Tributary TX from handle_application_tx after some shares"
)
},
&spec,
&tributaries[0].1.reader(),
)
.await;
assert_eq!(processors.0.read().await.len(), 1);
assert!(processors.0.read().await[&spec.set().network].is_empty());
// Publish the final set of shares
let block_before_tx = tributaries[0].1.tip().await;
assert_eq!(tributaries[0].1.add_transaction(txs[0].clone()).await, Ok(true));
wait_for_tx_inclusion(&tributaries[0].1, block_before_tx, txs[0].hash()).await;
sleep(Duration::from_secs(Tributary::<MemDb, Transaction, LocalP2p>::block_time().into())).await;
// Each scanner should emit a distinct shares message
let shares_for = |i: usize| {
CoordinatorMessage::KeyGen(key_gen::CoordinatorMessage::Shares {
id: KeyGenId { session: spec.set().session, attempt: 0 },
shares: vec![txs
.iter()
.enumerate()
.filter_map(|(l, tx)| {
if let Transaction::DkgShares { shares, .. } = tx {
if i == l {
None
} else {
let relative_i = i - (if i > l { 1 } else { 0 });
Some((
Participant::new((l + 1).try_into().unwrap()).unwrap(),
shares[0][relative_i].clone(),
))
}
} else {
panic!("txs had non-shares");
}
})
.collect::<HashMap<_, _>>()],
})
};
// Any scanner which has handled the prior blocks should only emit the new event
for (i, key) in keys.iter().enumerate() {
handle_new_blocks::<_, _, _, _, _, LocalP2p>(
&mut dbs[i],
key,
&|_, _, _, _| async { panic!("provided TX caused recognized_id to be called after shares") },
&processors,
&(),
&|_| async { panic!("test tried to publish a new Tributary TX from handle_application_tx") },
&spec,
&tributaries[i].1.reader(),
)
.await;
{
let mut msgs = processors.0.write().await;
assert_eq!(msgs.len(), 1);
let msgs = msgs.get_mut(&spec.set().network).unwrap();
assert_eq!(msgs.pop_front().unwrap(), shares_for(i));
assert!(msgs.is_empty());
}
}
// Yet new scanners should emit all events
for (i, key) in keys.iter().enumerate() {
let processors = new_processors(&mut MemDb::new(), key, &spec, &tributaries[i].1).await;
let mut msgs = processors.0.write().await;
assert_eq!(msgs.len(), 1);
let msgs = msgs.get_mut(&spec.set().network).unwrap();
let mut expected_commitments = expected_commitments.clone();
expected_commitments.remove(&Participant::new((i + 1).try_into().unwrap()).unwrap());
assert_eq!(
msgs.pop_front().unwrap(),
CoordinatorMessage::KeyGen(key_gen::CoordinatorMessage::Commitments {
id: KeyGenId { session: spec.set().session, attempt: 0 },
commitments: expected_commitments
})
);
assert_eq!(msgs.pop_front().unwrap(), shares_for(i));
assert!(msgs.is_empty());
}
// Send DkgConfirmed
let mut substrate_key = [0; 32];
OsRng.fill_bytes(&mut substrate_key);
let mut network_key = vec![0; usize::try_from((OsRng.next_u64() % 32) + 32).unwrap()];
OsRng.fill_bytes(&mut network_key);
let key_pair = KeyPair(serai_client::Public(substrate_key), network_key.try_into().unwrap());
let mut txs = vec![];
for (i, key) in keys.iter().enumerate() {
let attempt = 0;
let mut txn = dbs[i].txn();
let share =
crate::tributary::generated_key_pair::<MemDb>(&mut txn, key, &spec, &key_pair, 0).unwrap();
txn.commit();
let mut tx = Transaction::DkgConfirmed {
attempt,
confirmation_share: share,
signed: Transaction::empty_signed(),
};
tx.sign(&mut OsRng, spec.genesis(), key);
txs.push(tx);
}
let block_before_tx = tributaries[0].1.tip().await;
for (i, tx) in txs.iter().enumerate() {
assert_eq!(tributaries[i].1.add_transaction(tx.clone()).await, Ok(true));
}
for tx in &txs {
wait_for_tx_inclusion(&tributaries[0].1, block_before_tx, tx.hash()).await;
}
struct CheckPublishSetKeys {
spec: TributarySpec,
key_pair: KeyPair,
}
#[async_trait::async_trait]
impl PublishSeraiTransaction for CheckPublishSetKeys {
async fn publish_set_keys(
&self,
_db: &(impl Sync + Get),
set: ValidatorSet,
removed: Vec<SeraiAddress>,
key_pair: KeyPair,
signature: Signature,
) {
assert_eq!(set, self.spec.set());
assert!(removed.is_empty());
assert_eq!(self.key_pair, key_pair);
assert!(signature.verify(
&*serai_client::validator_sets::primitives::set_keys_message(&set, &[], &key_pair),
&serai_client::Public(
frost::dkg::musig::musig_key::<Ristretto>(
&serai_client::validator_sets::primitives::musig_context(set),
&self.spec.validators().into_iter().map(|(validator, _)| validator).collect::<Vec<_>>()
)
.unwrap()
.to_bytes()
),
));
}
}
// The scanner should successfully try to publish a transaction with a validly signed signature
handle_new_blocks::<_, _, _, _, _, LocalP2p>(
&mut dbs[0],
&keys[0],
&|_, _, _, _| async {
panic!("provided TX caused recognized_id to be called after DKG confirmation")
},
&processors,
&CheckPublishSetKeys { spec: spec.clone(), key_pair: key_pair.clone() },
&|_| async { panic!("test tried to publish a new Tributary TX from handle_application_tx") },
&spec,
&tributaries[0].1.reader(),
)
.await;
{
assert!(processors.0.read().await.get(&spec.set().network).unwrap().is_empty());
}
}

View File

@@ -1,74 +0,0 @@
use core::time::Duration;
use std::sync::Arc;
use rand_core::OsRng;
use tokio::{
sync::{mpsc, broadcast},
time::sleep,
};
use serai_db::MemDb;
use tributary::Tributary;
use crate::{
tributary::Transaction,
ActiveTributary, TributaryEvent,
p2p::handle_p2p_task,
tests::{
LocalP2p,
tributary::{new_keys, new_spec, new_tributaries},
},
};
#[tokio::test]
async fn handle_p2p_test() {
let keys = new_keys(&mut OsRng);
let spec = new_spec(&mut OsRng, &keys);
let mut tributaries = new_tributaries(&keys, &spec)
.await
.into_iter()
.map(|(_, p2p, tributary)| (p2p, tributary))
.collect::<Vec<_>>();
let mut tributary_senders = vec![];
let mut tributary_arcs = vec![];
for (p2p, tributary) in tributaries.drain(..) {
let tributary = Arc::new(tributary);
tributary_arcs.push(tributary.clone());
let (new_tributary_send, new_tributary_recv) = broadcast::channel(5);
let (cosign_send, _) = mpsc::unbounded_channel();
tokio::spawn(handle_p2p_task(p2p, cosign_send, new_tributary_recv));
new_tributary_send
.send(TributaryEvent::NewTributary(ActiveTributary { spec: spec.clone(), tributary }))
.map_err(|_| "failed to send ActiveTributary")
.unwrap();
tributary_senders.push(new_tributary_send);
}
let tributaries = tributary_arcs;
// After two blocks of time, we should have a new block
// We don't wait one block of time as we may have missed the chance for this block
sleep(Duration::from_secs((2 * Tributary::<MemDb, Transaction, LocalP2p>::block_time()).into()))
.await;
let tip = tributaries[0].tip().await;
assert!(tip != spec.genesis());
// Sleep one second to make sure this block propagates
sleep(Duration::from_secs(1)).await;
// Make sure every tributary has it
for tributary in &tributaries {
assert!(tributary.reader().block(&tip).is_some());
}
// Then after another block of time, we should have yet another new block
sleep(Duration::from_secs(Tributary::<MemDb, Transaction, LocalP2p>::block_time().into())).await;
let new_tip = tributaries[0].tip().await;
assert!(new_tip != tip);
sleep(Duration::from_secs(1)).await;
for tributary in tributaries {
assert!(tributary.reader().block(&new_tip).is_some());
}
}

View File

@@ -1,293 +0,0 @@
use core::fmt::Debug;
use rand_core::{RngCore, OsRng};
use ciphersuite::{group::Group, Ciphersuite, Ristretto};
use scale::{Encode, Decode};
use serai_client::{
primitives::{SeraiAddress, Signature},
validator_sets::primitives::{MAX_KEY_SHARES_PER_SET, ValidatorSet, KeyPair},
};
use processor_messages::coordinator::SubstrateSignableId;
use tributary::{ReadWrite, tests::random_signed_with_nonce};
use crate::tributary::{Label, SignData, Transaction, scanner::PublishSeraiTransaction};
mod chain;
pub use chain::*;
mod tx;
mod dkg;
// TODO: Test the other transactions
mod handle_p2p;
mod sync;
#[async_trait::async_trait]
impl PublishSeraiTransaction for () {
async fn publish_set_keys(
&self,
_db: &(impl Sync + serai_db::Get),
_set: ValidatorSet,
_removed: Vec<SeraiAddress>,
_key_pair: KeyPair,
_signature: Signature,
) {
panic!("publish_set_keys was called in test")
}
}
fn random_u32<R: RngCore>(rng: &mut R) -> u32 {
u32::try_from(rng.next_u64() >> 32).unwrap()
}
fn random_vec<R: RngCore>(rng: &mut R, limit: usize) -> Vec<u8> {
let len = usize::try_from(rng.next_u64() % u64::try_from(limit).unwrap()).unwrap();
let mut res = vec![0; len];
rng.fill_bytes(&mut res);
res
}
fn random_sign_data<R: RngCore, Id: Clone + PartialEq + Eq + Debug + Encode + Decode>(
rng: &mut R,
plan: Id,
label: Label,
) -> SignData<Id> {
SignData {
plan,
attempt: random_u32(&mut OsRng),
label,
data: {
let mut res = vec![];
for _ in 0 ..= (rng.next_u64() % 255) {
res.push(random_vec(&mut OsRng, 512));
}
res
},
signed: random_signed_with_nonce(&mut OsRng, label.nonce()),
}
}
fn test_read_write<RW: Eq + Debug + ReadWrite>(value: &RW) {
assert_eq!(value, &RW::read::<&[u8]>(&mut value.serialize().as_ref()).unwrap());
}
#[test]
fn tx_size_limit() {
use serai_client::validator_sets::primitives::MAX_KEY_LEN;
use tributary::TRANSACTION_SIZE_LIMIT;
let max_dkg_coefficients = (MAX_KEY_SHARES_PER_SET * 2).div_ceil(3) + 1;
let max_key_shares_per_individual = MAX_KEY_SHARES_PER_SET - max_dkg_coefficients;
// Handwave the DKG Commitments size as the size of the commitments to the coefficients and
// 1024 bytes for all overhead
let handwaved_dkg_commitments_size = (max_dkg_coefficients * MAX_KEY_LEN) + 1024;
assert!(
u32::try_from(TRANSACTION_SIZE_LIMIT).unwrap() >=
(handwaved_dkg_commitments_size * max_key_shares_per_individual)
);
// Encryption key, PoP (2 elements), message
let elements_per_share = 4;
let handwaved_dkg_shares_size =
(elements_per_share * MAX_KEY_LEN * MAX_KEY_SHARES_PER_SET) + 1024;
assert!(
u32::try_from(TRANSACTION_SIZE_LIMIT).unwrap() >=
(handwaved_dkg_shares_size * max_key_shares_per_individual)
);
}
#[test]
fn serialize_sign_data() {
fn test_read_write<Id: Clone + PartialEq + Eq + Debug + Encode + Decode>(value: &SignData<Id>) {
let mut buf = vec![];
value.write(&mut buf).unwrap();
assert_eq!(value, &SignData::read(&mut buf.as_slice()).unwrap())
}
let mut plan = [0; 3];
OsRng.fill_bytes(&mut plan);
test_read_write(&random_sign_data::<_, _>(
&mut OsRng,
plan,
if (OsRng.next_u64() % 2) == 0 { Label::Preprocess } else { Label::Share },
));
let mut plan = [0; 5];
OsRng.fill_bytes(&mut plan);
test_read_write(&random_sign_data::<_, _>(
&mut OsRng,
plan,
if (OsRng.next_u64() % 2) == 0 { Label::Preprocess } else { Label::Share },
));
let mut plan = [0; 8];
OsRng.fill_bytes(&mut plan);
test_read_write(&random_sign_data::<_, _>(
&mut OsRng,
plan,
if (OsRng.next_u64() % 2) == 0 { Label::Preprocess } else { Label::Share },
));
let mut plan = [0; 24];
OsRng.fill_bytes(&mut plan);
test_read_write(&random_sign_data::<_, _>(
&mut OsRng,
plan,
if (OsRng.next_u64() % 2) == 0 { Label::Preprocess } else { Label::Share },
));
}
#[test]
fn serialize_transaction() {
test_read_write(&Transaction::RemoveParticipantDueToDkg {
participant: <Ristretto as Ciphersuite>::G::random(&mut OsRng),
signed: random_signed_with_nonce(&mut OsRng, 0),
});
{
let mut commitments = vec![random_vec(&mut OsRng, 512)];
for _ in 0 .. (OsRng.next_u64() % 100) {
let mut temp = commitments[0].clone();
OsRng.fill_bytes(&mut temp);
commitments.push(temp);
}
test_read_write(&Transaction::DkgCommitments {
attempt: random_u32(&mut OsRng),
commitments,
signed: random_signed_with_nonce(&mut OsRng, 0),
});
}
{
// This supports a variable share length, and variable amount of sent shares, yet share length
// and sent shares is expected to be constant among recipients
let share_len = usize::try_from((OsRng.next_u64() % 512) + 1).unwrap();
let amount_of_shares = usize::try_from((OsRng.next_u64() % 3) + 1).unwrap();
// Create a valid vec of shares
let mut shares = vec![];
// Create up to 150 participants
for _ in 0 ..= (OsRng.next_u64() % 150) {
// Give each sender multiple shares
let mut sender_shares = vec![];
for _ in 0 .. amount_of_shares {
let mut share = vec![0; share_len];
OsRng.fill_bytes(&mut share);
sender_shares.push(share);
}
shares.push(sender_shares);
}
test_read_write(&Transaction::DkgShares {
attempt: random_u32(&mut OsRng),
shares,
confirmation_nonces: {
let mut nonces = [0; 64];
OsRng.fill_bytes(&mut nonces);
nonces
},
signed: random_signed_with_nonce(&mut OsRng, 1),
});
}
for i in 0 .. 2 {
test_read_write(&Transaction::InvalidDkgShare {
attempt: random_u32(&mut OsRng),
accuser: frost::Participant::new(
u16::try_from(OsRng.next_u64() >> 48).unwrap().saturating_add(1),
)
.unwrap(),
faulty: frost::Participant::new(
u16::try_from(OsRng.next_u64() >> 48).unwrap().saturating_add(1),
)
.unwrap(),
blame: if i == 0 {
None
} else {
Some(random_vec(&mut OsRng, 500)).filter(|blame| !blame.is_empty())
},
signed: random_signed_with_nonce(&mut OsRng, 2),
});
}
test_read_write(&Transaction::DkgConfirmed {
attempt: random_u32(&mut OsRng),
confirmation_share: {
let mut share = [0; 32];
OsRng.fill_bytes(&mut share);
share
},
signed: random_signed_with_nonce(&mut OsRng, 2),
});
{
let mut block = [0; 32];
OsRng.fill_bytes(&mut block);
test_read_write(&Transaction::CosignSubstrateBlock(block));
}
{
let mut block = [0; 32];
OsRng.fill_bytes(&mut block);
let batch = u32::try_from(OsRng.next_u64() >> 32).unwrap();
test_read_write(&Transaction::Batch { block, batch });
}
test_read_write(&Transaction::SubstrateBlock(OsRng.next_u64()));
{
let batch = u32::try_from(OsRng.next_u64() >> 32).unwrap();
test_read_write(&Transaction::SubstrateSign(random_sign_data(
&mut OsRng,
SubstrateSignableId::Batch(batch),
Label::Preprocess,
)));
}
{
let batch = u32::try_from(OsRng.next_u64() >> 32).unwrap();
test_read_write(&Transaction::SubstrateSign(random_sign_data(
&mut OsRng,
SubstrateSignableId::Batch(batch),
Label::Share,
)));
}
{
let mut plan = [0; 32];
OsRng.fill_bytes(&mut plan);
test_read_write(&Transaction::Sign(random_sign_data(&mut OsRng, plan, Label::Preprocess)));
}
{
let mut plan = [0; 32];
OsRng.fill_bytes(&mut plan);
test_read_write(&Transaction::Sign(random_sign_data(&mut OsRng, plan, Label::Share)));
}
{
let mut plan = [0; 32];
OsRng.fill_bytes(&mut plan);
let mut tx_hash = vec![0; (OsRng.next_u64() % 64).try_into().unwrap()];
OsRng.fill_bytes(&mut tx_hash);
test_read_write(&Transaction::SignCompleted {
plan,
tx_hash,
first_signer: random_signed_with_nonce(&mut OsRng, 2).signer,
signature: random_signed_with_nonce(&mut OsRng, 2).signature,
});
}
test_read_write(&Transaction::SlashReport(
{
let amount =
usize::try_from(OsRng.next_u64() % u64::from(MAX_KEY_SHARES_PER_SET - 1)).unwrap();
let mut points = vec![];
for _ in 0 .. amount {
points.push((OsRng.next_u64() >> 32).try_into().unwrap());
}
points
},
random_signed_with_nonce(&mut OsRng, 0),
));
}

View File

@@ -1,165 +0,0 @@
use core::time::Duration;
use std::{sync::Arc, collections::HashSet};
use rand_core::OsRng;
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use tokio::{
sync::{mpsc, broadcast},
time::sleep,
};
use serai_db::MemDb;
use tributary::Tributary;
use crate::{
tributary::Transaction,
ActiveTributary, TributaryEvent,
p2p::{heartbeat_tributaries_task, handle_p2p_task},
tests::{
LocalP2p,
tributary::{new_keys, new_spec, new_tributaries},
},
};
#[tokio::test]
async fn sync_test() {
let mut keys = new_keys(&mut OsRng);
let spec = new_spec(&mut OsRng, &keys);
// Ensure this can have a node fail
assert!(spec.n(&[]) > spec.t());
let mut tributaries = new_tributaries(&keys, &spec)
.await
.into_iter()
.map(|(_, p2p, tributary)| (p2p, tributary))
.collect::<Vec<_>>();
// Keep a Tributary back, effectively having it offline
let syncer_key = keys.pop().unwrap();
let (syncer_p2p, syncer_tributary) = tributaries.pop().unwrap();
// Have the rest form a P2P net
let mut tributary_senders = vec![];
let mut tributary_arcs = vec![];
let mut p2p_threads = vec![];
for (p2p, tributary) in tributaries.drain(..) {
let tributary = Arc::new(tributary);
tributary_arcs.push(tributary.clone());
let (new_tributary_send, new_tributary_recv) = broadcast::channel(5);
let (cosign_send, _) = mpsc::unbounded_channel();
let thread = tokio::spawn(handle_p2p_task(p2p, cosign_send, new_tributary_recv));
new_tributary_send
.send(TributaryEvent::NewTributary(ActiveTributary { spec: spec.clone(), tributary }))
.map_err(|_| "failed to send ActiveTributary")
.unwrap();
tributary_senders.push(new_tributary_send);
p2p_threads.push(thread);
}
let tributaries = tributary_arcs;
// After four blocks of time, we should have a new block
// We don't wait one block of time as we may have missed the chance for the first block
// We don't wait two blocks because we may have missed the chance, and then had a failure to
// propose by our 'offline' validator, which would cause the Tendermint round time to increase,
// requiring a longer delay
let block_time = u64::from(Tributary::<MemDb, Transaction, LocalP2p>::block_time());
sleep(Duration::from_secs(4 * block_time)).await;
let tip = tributaries[0].tip().await;
assert!(tip != spec.genesis());
// Sleep one second to make sure this block propagates
sleep(Duration::from_secs(1)).await;
// Make sure every tributary has it
for tributary in &tributaries {
assert!(tributary.reader().block(&tip).is_some());
}
// Now that we've confirmed the other tributaries formed a net without issue, drop the syncer's
// pending P2P messages
syncer_p2p.1.write().await.1.last_mut().unwrap().clear();
// Have it join the net
let syncer_key = Ristretto::generator() * *syncer_key;
let syncer_tributary = Arc::new(syncer_tributary);
let (syncer_tributary_send, syncer_tributary_recv) = broadcast::channel(5);
let (cosign_send, _) = mpsc::unbounded_channel();
tokio::spawn(handle_p2p_task(syncer_p2p.clone(), cosign_send, syncer_tributary_recv));
syncer_tributary_send
.send(TributaryEvent::NewTributary(ActiveTributary {
spec: spec.clone(),
tributary: syncer_tributary.clone(),
}))
.map_err(|_| "failed to send ActiveTributary to syncer")
.unwrap();
// It shouldn't automatically catch up. If it somehow was, our test would be broken
// Sanity check this
let tip = tributaries[0].tip().await;
// Wait until a new block occurs
sleep(Duration::from_secs(3 * block_time)).await;
// Make sure a new block actually occurred
assert!(tributaries[0].tip().await != tip);
// Make sure the new block alone didn't trigger catching up
assert_eq!(syncer_tributary.tip().await, spec.genesis());
// Start the heartbeat protocol
let (syncer_heartbeat_tributary_send, syncer_heartbeat_tributary_recv) = broadcast::channel(5);
tokio::spawn(heartbeat_tributaries_task(syncer_p2p, syncer_heartbeat_tributary_recv));
syncer_heartbeat_tributary_send
.send(TributaryEvent::NewTributary(ActiveTributary {
spec: spec.clone(),
tributary: syncer_tributary.clone(),
}))
.map_err(|_| "failed to send ActiveTributary to heartbeat")
.unwrap();
// The heartbeat is once every 10 blocks, with some limitations
sleep(Duration::from_secs(20 * block_time)).await;
assert!(syncer_tributary.tip().await != spec.genesis());
// Verify it synced to the tip
let syncer_tip = {
let tributary = &tributaries[0];
let tip = tributary.tip().await;
let syncer_tip = syncer_tributary.tip().await;
// Allow a one block tolerance in case of race conditions
assert!(
HashSet::from([tip, tributary.reader().block(&tip).unwrap().parent()]).contains(&syncer_tip)
);
syncer_tip
};
sleep(Duration::from_secs(block_time)).await;
// Verify it's now keeping up
assert!(syncer_tributary.tip().await != syncer_tip);
// Verify it's now participating in consensus
// Because only `t` validators are used in a commit, take n - t nodes offline
// leaving only `t` nodes. Which should force it to participate in the consensus
// of next blocks.
let spares = usize::from(spec.n(&[]) - spec.t());
for thread in p2p_threads.iter().take(spares) {
thread.abort();
}
// wait for a block
sleep(Duration::from_secs(block_time)).await;
if syncer_tributary
.reader()
.parsed_commit(&syncer_tributary.tip().await)
.unwrap()
.validators
.iter()
.any(|signer| signer == &syncer_key.to_bytes())
{
return;
}
panic!("synced tributary didn't start participating in consensus");
}

View File

@@ -1,63 +0,0 @@
use core::time::Duration;
use rand_core::{RngCore, OsRng};
use tokio::time::sleep;
use serai_db::MemDb;
use tributary::{
transaction::Transaction as TransactionTrait, Transaction as TributaryTransaction, Tributary,
};
use crate::{
tributary::Transaction,
tests::{
LocalP2p,
tributary::{new_keys, new_spec, new_tributaries, run_tributaries, wait_for_tx_inclusion},
},
};
#[tokio::test]
async fn tx_test() {
let keys = new_keys(&mut OsRng);
let spec = new_spec(&mut OsRng, &keys);
let tributaries = new_tributaries(&keys, &spec)
.await
.into_iter()
.map(|(_, p2p, tributary)| (p2p, tributary))
.collect::<Vec<_>>();
// Run the tributaries in the background
tokio::spawn(run_tributaries(tributaries.clone()));
// Send a TX from a random Tributary
let sender =
usize::try_from(OsRng.next_u64() % u64::try_from(tributaries.len()).unwrap()).unwrap();
let key = keys[sender].clone();
let attempt = 0;
let mut commitments = vec![0; 256];
OsRng.fill_bytes(&mut commitments);
// Create the TX with a null signature so we can get its sig hash
let block_before_tx = tributaries[sender].1.tip().await;
let mut tx = Transaction::DkgCommitments {
attempt,
commitments: vec![commitments.clone()],
signed: Transaction::empty_signed(),
};
tx.sign(&mut OsRng, spec.genesis(), &key);
assert_eq!(tributaries[sender].1.add_transaction(tx.clone()).await, Ok(true));
let included_in = wait_for_tx_inclusion(&tributaries[sender].1, block_before_tx, tx.hash()).await;
// Also sleep for the block time to ensure the block is synced around before we run checks on it
sleep(Duration::from_secs(Tributary::<MemDb, Transaction, LocalP2p>::block_time().into())).await;
// All tributaries should have acknowledged this transaction in a block
for (_, tributary) in tributaries {
let block = tributary.reader().block(&included_in).unwrap();
assert_eq!(block.transactions, vec![TributaryTransaction::Application(tx.clone())]);
}
}

View File

@@ -0,0 +1,453 @@
use core::{future::Future, time::Duration};
use std::sync::Arc;
use zeroize::Zeroizing;
use rand_core::OsRng;
use blake2::{digest::typenum::U32, Digest, Blake2s};
use ciphersuite::{Ciphersuite, Ristretto};
use tokio::sync::mpsc;
use serai_db::{Get, DbTxn, Db as DbTrait, create_db, db_channel};
use scale::Encode;
use serai_client::validator_sets::primitives::ValidatorSet;
use tributary_sdk::{TransactionKind, TransactionError, ProvidedError, TransactionTrait, Tributary};
use serai_task::{Task, TaskHandle, DoesNotError, ContinuallyRan};
use message_queue::{Service, Metadata, client::MessageQueue};
use serai_cosign::{Faulted, CosignIntent, Cosigning};
use serai_coordinator_substrate::{NewSetInformation, SignSlashReport};
use serai_coordinator_tributary::{Transaction, ProcessorMessages, CosignIntents, ScanTributaryTask};
use serai_coordinator_p2p::P2p;
use crate::{Db, TributaryTransactions};
db_channel! {
Coordinator {
PendingCosigns: (set: ValidatorSet) -> CosignIntent,
}
}
/// Provide a Provided Transaction to the Tributary.
///
/// This is not a well-designed function. This is specific to the context in which its called,
/// within this file. It should only be considered an internal helper for this domain alone.
async fn provide_transaction<TD: DbTrait, P: P2p>(
set: ValidatorSet,
tributary: &Tributary<TD, Transaction, P>,
tx: Transaction,
) {
match tributary.provide_transaction(tx.clone()).await {
// The Tributary uses its own DB, so we may provide this multiple times if we reboot before
// committing the txn which provoked this
Ok(()) | Err(ProvidedError::AlreadyProvided) => {}
Err(ProvidedError::NotProvided) => {
panic!("providing a Transaction which wasn't a Provided transaction: {tx:?}");
}
Err(ProvidedError::InvalidProvided(e)) => {
panic!("providing an invalid Provided transaction, tx: {tx:?}, error: {e:?}")
}
// The Tributary's scan task won't advance if we don't have the Provided transactions
// present on-chain, and this enters an infinite loop to block the calling task from
// advancing
Err(ProvidedError::LocalMismatchesOnChain) => loop {
log::error!(
"Tributary {:?} was supposed to provide {:?} but peers disagree, halting Tributary",
set,
tx,
);
// Print this every five minutes as this does need to be handled
tokio::time::sleep(Duration::from_secs(5 * 60)).await;
},
}
}
/// Provides Cosign/Cosigned Transactions onto the Tributary.
pub(crate) struct ProvideCosignCosignedTransactionsTask<CD: DbTrait, TD: DbTrait, P: P2p> {
db: CD,
tributary_db: TD,
set: NewSetInformation,
tributary: Tributary<TD, Transaction, P>,
}
impl<CD: DbTrait, TD: DbTrait, P: P2p> ContinuallyRan
for ProvideCosignCosignedTransactionsTask<CD, TD, P>
{
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
// Check if we produced any cosigns we were supposed to
let mut pending_notable_cosign = false;
loop {
let mut txn = self.db.txn();
// Fetch the next cosign this tributary should handle
let Some(cosign) = PendingCosigns::try_recv(&mut txn, self.set.set) else { break };
pending_notable_cosign = cosign.notable;
// If we (Serai) haven't cosigned this block, break as this is still pending
let latest = match Cosigning::<CD>::latest_cosigned_block_number(&txn) {
Ok(latest) => latest,
Err(Faulted) => {
log::error!("cosigning faulted");
Err("cosigning faulted")?
}
};
if latest < cosign.block_number {
break;
}
// Because we've cosigned it, provide the TX for that
{
let mut txn = self.tributary_db.txn();
CosignIntents::provide(&mut txn, self.set.set, &cosign);
txn.commit();
}
provide_transaction(
self.set.set,
&self.tributary,
Transaction::Cosigned { substrate_block_hash: cosign.block_hash },
)
.await;
// Clear pending_notable_cosign since this cosign isn't pending
pending_notable_cosign = false;
// Commit the txn to clear this from PendingCosigns
txn.commit();
made_progress = true;
}
// If we don't have any notable cosigns pending, provide the next set of cosign intents
if !pending_notable_cosign {
let mut txn = self.db.txn();
// intended_cosigns will only yield up to and including the next notable cosign
for cosign in Cosigning::<CD>::intended_cosigns(&mut txn, self.set.set) {
// Flag this cosign as pending
PendingCosigns::send(&mut txn, self.set.set, &cosign);
// Provide the transaction to queue it for work
provide_transaction(
self.set.set,
&self.tributary,
Transaction::Cosign { substrate_block_hash: cosign.block_hash },
)
.await;
}
txn.commit();
made_progress = true;
}
Ok(made_progress)
}
}
}
/// Adds all of the transactions sent via `TributaryTransactions`.
pub(crate) struct AddTributaryTransactionsTask<CD: DbTrait, TD: DbTrait, P: P2p> {
db: CD,
tributary_db: TD,
tributary: Tributary<TD, Transaction, P>,
set: ValidatorSet,
key: Zeroizing<<Ristretto as Ciphersuite>::F>,
}
impl<CD: DbTrait, TD: DbTrait, P: P2p> ContinuallyRan for AddTributaryTransactionsTask<CD, TD, P> {
type Error = DoesNotError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
loop {
let mut txn = self.db.txn();
let Some(mut tx) = TributaryTransactions::try_recv(&mut txn, self.set) else { break };
let kind = tx.kind();
match kind {
TransactionKind::Provided(_) => provide_transaction(self.set, &self.tributary, tx).await,
TransactionKind::Unsigned | TransactionKind::Signed(_, _) => {
// If this is a signed transaction, sign it
if matches!(kind, TransactionKind::Signed(_, _)) {
tx.sign(&mut OsRng, self.tributary.genesis(), &self.key);
}
// Actually add the transaction
// TODO: If this is a preprocess, make sure the topic has been recognized
let res = self.tributary.add_transaction(tx.clone()).await;
match &res {
// Fresh publication, already published
Ok(true | false) => {}
Err(
TransactionError::TooLargeTransaction |
TransactionError::InvalidSigner |
TransactionError::InvalidNonce |
TransactionError::InvalidSignature |
TransactionError::InvalidContent,
) => {
panic!("created an invalid transaction, tx: {tx:?}, err: {res:?}");
}
// We've published too many transactions recently
// Drop this txn to try to publish it again later on a future iteration
Err(TransactionError::TooManyInMempool) => {
drop(txn);
break;
}
// This isn't a Provided transaction so this should never be hit
Err(TransactionError::ProvidedAddedToMempool) => unreachable!(),
}
}
}
made_progress = true;
txn.commit();
}
Ok(made_progress)
}
}
}
/// Takes the messages from ScanTributaryTask and publishes them to the message-queue.
pub(crate) struct TributaryProcessorMessagesTask<TD: DbTrait> {
tributary_db: TD,
set: ValidatorSet,
message_queue: Arc<MessageQueue>,
}
impl<TD: DbTrait> ContinuallyRan for TributaryProcessorMessagesTask<TD> {
type Error = String; // TODO
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
loop {
let mut txn = self.tributary_db.txn();
let Some(msg) = ProcessorMessages::try_recv(&mut txn, self.set) else { break };
let metadata = Metadata {
from: Service::Coordinator,
to: Service::Processor(self.set.network),
intent: msg.intent(),
};
let msg = borsh::to_vec(&msg).unwrap();
self.message_queue.queue(metadata, msg).await?;
txn.commit();
made_progress = true;
}
Ok(made_progress)
}
}
}
/// Checks for the notification to sign a slash report and does so if present.
pub(crate) struct SignSlashReportTask<CD: DbTrait, TD: DbTrait, P: P2p> {
db: CD,
tributary_db: TD,
tributary: Tributary<TD, Transaction, P>,
set: NewSetInformation,
key: Zeroizing<<Ristretto as Ciphersuite>::F>,
}
impl<CD: DbTrait, TD: DbTrait, P: P2p> ContinuallyRan for SignSlashReportTask<CD, TD, P> {
type Error = DoesNotError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut txn = self.db.txn();
let Some(()) = SignSlashReport::try_recv(&mut txn, self.set.set) else { return Ok(false) };
// Fetch the slash report for this Tributary
let mut tx =
serai_coordinator_tributary::slash_report_transaction(&self.tributary_db, &self.set);
tx.sign(&mut OsRng, self.tributary.genesis(), &self.key);
let res = self.tributary.add_transaction(tx.clone()).await;
match &res {
// Fresh publication, already published
Ok(true | false) => {}
Err(
TransactionError::TooLargeTransaction |
TransactionError::InvalidSigner |
TransactionError::InvalidNonce |
TransactionError::InvalidSignature |
TransactionError::InvalidContent,
) => {
panic!("created an invalid SlashReport transaction, tx: {tx:?}, err: {res:?}");
}
// We've published too many transactions recently
// Drop this txn to try to publish it again later on a future iteration
Err(TransactionError::TooManyInMempool) => {
drop(txn);
return Ok(false);
}
// This isn't a Provided transaction so this should never be hit
Err(TransactionError::ProvidedAddedToMempool) => unreachable!(),
}
txn.commit();
Ok(true)
}
}
}
/// Run the scan task whenever the Tributary adds a new block.
async fn scan_on_new_block<CD: DbTrait, TD: DbTrait, P: P2p>(
db: CD,
set: ValidatorSet,
tributary: Tributary<TD, Transaction, P>,
scan_tributary_task: TaskHandle,
tasks_to_keep_alive: Vec<TaskHandle>,
) {
loop {
// Break once this Tributary is retired
if crate::RetiredTributary::get(&db, set.network).map(|session| session.0) >=
Some(set.session.0)
{
drop(tasks_to_keep_alive);
break;
}
// Have the tributary scanner run as soon as there's a new block
match tributary.next_block_notification().await.await {
Ok(()) => scan_tributary_task.run_now(),
// unreachable since this owns the tributary object and doesn't drop it
Err(_) => panic!("tributary was dropped causing notification to error"),
}
}
}
/// Spawn a Tributary.
///
/// This will:
/// - Spawn the Tributary
/// - Inform the P2P network of the Tributary
/// - Spawn the ScanTributaryTask
/// - Spawn the ProvideCosignCosignedTransactionsTask
/// - Spawn the TributaryProcessorMessagesTask
/// - Spawn the SignSlashReportTask
/// - Iterate the scan task whenever a new block occurs (not just on the standard interval)
pub(crate) async fn spawn_tributary<P: P2p>(
db: Db,
message_queue: Arc<MessageQueue>,
p2p: P,
p2p_add_tributary: &mpsc::UnboundedSender<(ValidatorSet, Tributary<Db, Transaction, P>)>,
set: NewSetInformation,
serai_key: Zeroizing<<Ristretto as Ciphersuite>::F>,
) {
// Don't spawn retired Tributaries
if crate::db::RetiredTributary::get(&db, set.set.network).map(|session| session.0) >=
Some(set.set.session.0)
{
return;
}
let genesis = <[u8; 32]>::from(Blake2s::<U32>::digest((set.serai_block, set.set).encode()));
// Since the Serai block will be finalized, then cosigned, before we handle this, this time will
// be a couple of minutes stale. While the Tributary will still function with a start time in the
// past, the Tributary will immediately incur round timeouts. We reduce these by adding a
// constant delay of a couple of minutes.
const TRIBUTARY_START_TIME_DELAY: u64 = 120;
let start_time = set.declaration_time + TRIBUTARY_START_TIME_DELAY;
let mut tributary_validators = Vec::with_capacity(set.validators.len());
for (validator, weight) in set.validators.iter().copied() {
let validator_key = <Ristretto as Ciphersuite>::read_G(&mut validator.0.as_slice())
.expect("Serai validator had an invalid public key");
let weight = u64::from(weight);
tributary_validators.push((validator_key, weight));
}
// Spawn the Tributary
let tributary_db = crate::db::tributary_db(set.set);
let tributary = Tributary::new(
tributary_db.clone(),
genesis,
start_time,
serai_key.clone(),
tributary_validators,
p2p,
)
.await
.unwrap();
let reader = tributary.reader();
// Inform the P2P network
p2p_add_tributary
.send((set.set, tributary.clone()))
.expect("p2p's add_tributary channel was closed?");
// Spawn the task to provide Cosign/Cosigned transactions onto the Tributary
let (provide_cosign_cosigned_transactions_task_def, provide_cosign_cosigned_transactions_task) =
Task::new();
tokio::spawn(
(ProvideCosignCosignedTransactionsTask {
db: db.clone(),
tributary_db: tributary_db.clone(),
set: set.clone(),
tributary: tributary.clone(),
})
.continually_run(provide_cosign_cosigned_transactions_task_def, vec![]),
);
// Spawn the task to send all messages from the Tributary scanner to the message-queue
let (scan_tributary_messages_task_def, scan_tributary_messages_task) = Task::new();
tokio::spawn(
(TributaryProcessorMessagesTask {
tributary_db: tributary_db.clone(),
set: set.set,
message_queue,
})
.continually_run(scan_tributary_messages_task_def, vec![]),
);
// Spawn the scan task
let (scan_tributary_task_def, scan_tributary_task) = Task::new();
tokio::spawn(
ScanTributaryTask::<_, P>::new(tributary_db.clone(), &set, reader)
// This is the only handle for this TributaryProcessorMessagesTask, so when this task is
// dropped, it will be too
.continually_run(scan_tributary_task_def, vec![scan_tributary_messages_task]),
);
// Spawn the sign slash report task
let (sign_slash_report_task_def, sign_slash_report_task) = Task::new();
tokio::spawn(
(SignSlashReportTask {
db: db.clone(),
tributary_db: tributary_db.clone(),
tributary: tributary.clone(),
set: set.clone(),
key: serai_key.clone(),
})
.continually_run(sign_slash_report_task_def, vec![]),
);
// Spawn the add transactions task
let (add_tributary_transactions_task_def, add_tributary_transactions_task) = Task::new();
tokio::spawn(
(AddTributaryTransactionsTask {
db: db.clone(),
tributary_db,
tributary: tributary.clone(),
set: set.set,
key: serai_key,
})
.continually_run(add_tributary_transactions_task_def, vec![]),
);
// Whenever a new block occurs, immediately run the scan task
// This function also preserves the ProvideCosignCosignedTransactionsTask handle until the
// Tributary is retired, ensuring it isn't dropped prematurely and that the task don't run ad
// infinitum
tokio::spawn(scan_on_new_block(
db,
set.set,
tributary,
scan_tributary_task,
vec![
provide_cosign_cosigned_transactions_task,
sign_slash_report_task,
add_tributary_transactions_task,
],
));
}

View File

@@ -1,197 +0,0 @@
use std::collections::HashMap;
use scale::Encode;
use borsh::{BorshSerialize, BorshDeserialize};
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use frost::Participant;
use serai_client::validator_sets::primitives::{KeyPair, ValidatorSet};
use processor_messages::coordinator::SubstrateSignableId;
pub use serai_db::*;
use tributary::ReadWrite;
use crate::tributary::{Label, Transaction};
#[derive(Clone, Copy, PartialEq, Eq, Debug, Encode, BorshSerialize, BorshDeserialize)]
pub enum Topic {
Dkg,
DkgConfirmation,
SubstrateSign(SubstrateSignableId),
Sign([u8; 32]),
}
// A struct to refer to a piece of data all validators will presumably provide a value for.
#[derive(Clone, Copy, PartialEq, Eq, Debug, Encode)]
pub struct DataSpecification {
pub topic: Topic,
pub label: Label,
pub attempt: u32,
}
pub enum DataSet {
Participating(HashMap<Participant, Vec<u8>>),
NotParticipating,
}
pub enum Accumulation {
Ready(DataSet),
NotReady,
}
// TODO: Move from genesis to set for indexing
create_db!(
Tributary {
SeraiBlockNumber: (hash: [u8; 32]) -> u64,
SeraiDkgCompleted: (spec: ValidatorSet) -> [u8; 32],
TributaryBlockNumber: (block: [u8; 32]) -> u32,
LastHandledBlock: (genesis: [u8; 32]) -> [u8; 32],
// TODO: Revisit the point of this
FatalSlashes: (genesis: [u8; 32]) -> Vec<[u8; 32]>,
RemovedAsOfDkgAttempt: (genesis: [u8; 32], attempt: u32) -> Vec<[u8; 32]>,
OfflineDuringDkg: (genesis: [u8; 32]) -> Vec<[u8; 32]>,
// TODO: Combine these two
FatallySlashed: (genesis: [u8; 32], account: [u8; 32]) -> (),
SlashPoints: (genesis: [u8; 32], account: [u8; 32]) -> u32,
VotedToRemove: (genesis: [u8; 32], voter: [u8; 32], to_remove: [u8; 32]) -> (),
VotesToRemove: (genesis: [u8; 32], to_remove: [u8; 32]) -> u16,
AttemptDb: (genesis: [u8; 32], topic: &Topic) -> u32,
ReattemptDb: (genesis: [u8; 32], block: u32) -> Vec<Topic>,
DataReceived: (genesis: [u8; 32], data_spec: &DataSpecification) -> u16,
DataDb: (genesis: [u8; 32], data_spec: &DataSpecification, signer_bytes: &[u8; 32]) -> Vec<u8>,
DkgShare: (genesis: [u8; 32], from: u16, to: u16) -> Vec<u8>,
ConfirmationNonces: (genesis: [u8; 32], attempt: u32) -> HashMap<Participant, Vec<u8>>,
DkgKeyPair: (genesis: [u8; 32], attempt: u32) -> KeyPair,
KeyToDkgAttempt: (key: [u8; 32]) -> u32,
DkgLocallyCompleted: (genesis: [u8; 32]) -> (),
PlanIds: (genesis: &[u8], block: u64) -> Vec<[u8; 32]>,
SignedTransactionDb: (order: &[u8], nonce: u32) -> Vec<u8>,
SlashReports: (genesis: [u8; 32], signer: [u8; 32]) -> Vec<u32>,
SlashReported: (genesis: [u8; 32]) -> u16,
SlashReportCutOff: (genesis: [u8; 32]) -> u64,
SlashReport: (set: ValidatorSet) -> Vec<([u8; 32], u32)>,
}
);
impl FatalSlashes {
pub fn get_as_keys(getter: &impl Get, genesis: [u8; 32]) -> Vec<<Ristretto as Ciphersuite>::G> {
FatalSlashes::get(getter, genesis)
.unwrap_or(vec![])
.iter()
.map(|key| <Ristretto as Ciphersuite>::G::from_bytes(key).unwrap())
.collect::<Vec<_>>()
}
}
impl FatallySlashed {
pub fn set_fatally_slashed(txn: &mut impl DbTxn, genesis: [u8; 32], account: [u8; 32]) {
Self::set(txn, genesis, account, &());
let mut existing = FatalSlashes::get(txn, genesis).unwrap_or_default();
// Don't append if we already have it, which can occur upon multiple faults
if existing.iter().any(|existing| existing == &account) {
return;
}
existing.push(account);
FatalSlashes::set(txn, genesis, &existing);
}
}
impl AttemptDb {
pub fn recognize_topic(txn: &mut impl DbTxn, genesis: [u8; 32], topic: Topic) {
Self::set(txn, genesis, &topic, &0u32);
}
pub fn start_next_attempt(txn: &mut impl DbTxn, genesis: [u8; 32], topic: Topic) -> u32 {
let next =
Self::attempt(txn, genesis, topic).expect("starting next attempt for unknown topic") + 1;
Self::set(txn, genesis, &topic, &next);
next
}
pub fn attempt(getter: &impl Get, genesis: [u8; 32], topic: Topic) -> Option<u32> {
let attempt = Self::get(getter, genesis, &topic);
// Don't require explicit recognition of the Dkg topic as it starts when the chain does
// Don't require explicit recognition of the SlashReport topic as it isn't a DoS risk and it
// should always happen (eventually)
if attempt.is_none() &&
((topic == Topic::Dkg) ||
(topic == Topic::DkgConfirmation) ||
(topic == Topic::SubstrateSign(SubstrateSignableId::SlashReport)))
{
return Some(0);
}
attempt
}
}
impl ReattemptDb {
pub fn schedule_reattempt(
txn: &mut impl DbTxn,
genesis: [u8; 32],
current_block_number: u32,
topic: Topic,
) {
// 5 minutes
#[cfg(not(feature = "longer-reattempts"))]
const BASE_REATTEMPT_DELAY: u32 = (5 * 60 * 1000) / tributary::tendermint::TARGET_BLOCK_TIME;
// 10 minutes, intended for latent environments like the GitHub CI
#[cfg(feature = "longer-reattempts")]
const BASE_REATTEMPT_DELAY: u32 = (10 * 60 * 1000) / tributary::tendermint::TARGET_BLOCK_TIME;
// 5 minutes for attempts 0 ..= 2, 10 minutes for attempts 3 ..= 5, 15 minutes for attempts > 5
// Assumes no event will take longer than 15 minutes, yet grows the time in case there are
// network bandwidth issues
let mut reattempt_delay = BASE_REATTEMPT_DELAY *
((AttemptDb::attempt(txn, genesis, topic)
.expect("scheduling re-attempt for unknown topic") /
3) +
1)
.min(3);
// Allow more time for DKGs since they have an extra round and much more data
if matches!(topic, Topic::Dkg) {
reattempt_delay *= 4;
}
let upon_block = current_block_number + reattempt_delay;
let mut reattempts = Self::get(txn, genesis, upon_block).unwrap_or(vec![]);
reattempts.push(topic);
Self::set(txn, genesis, upon_block, &reattempts);
}
pub fn take(txn: &mut impl DbTxn, genesis: [u8; 32], block_number: u32) -> Vec<Topic> {
let res = Self::get(txn, genesis, block_number).unwrap_or(vec![]);
if !res.is_empty() {
Self::del(txn, genesis, block_number);
}
res
}
}
impl SignedTransactionDb {
pub fn take_signed_transaction(
txn: &mut impl DbTxn,
order: &[u8],
nonce: u32,
) -> Option<Transaction> {
let res = SignedTransactionDb::get(txn, order, nonce)
.map(|bytes| Transaction::read(&mut bytes.as_slice()).unwrap());
if res.is_some() {
Self::del(txn, order, nonce);
}
res
}
}

View File

@@ -1,776 +0,0 @@
use core::ops::Deref;
use std::collections::HashMap;
use zeroize::Zeroizing;
use rand_core::OsRng;
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use frost::dkg::Participant;
use scale::{Encode, Decode};
use serai_client::{Signature, validator_sets::primitives::KeyPair};
use tributary::{Signed, TransactionKind, TransactionTrait};
use processor_messages::{
key_gen::{self, KeyGenId},
coordinator::{self, SubstrateSignableId, SubstrateSignId},
sign::{self, SignId},
};
use serai_db::*;
use crate::{
processors::Processors,
tributary::{
*,
signing_protocol::DkgConfirmer,
scanner::{
RecognizedIdType, RIDTrait, PublishSeraiTransaction, PTTTrait, TributaryBlockHandler,
},
},
P2p,
};
pub fn dkg_confirmation_nonces(
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
spec: &TributarySpec,
txn: &mut impl DbTxn,
attempt: u32,
) -> [u8; 64] {
DkgConfirmer::new(key, spec, txn, attempt)
.expect("getting DKG confirmation nonces for unknown attempt")
.preprocess()
}
pub fn generated_key_pair<D: Db>(
txn: &mut D::Transaction<'_>,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
spec: &TributarySpec,
key_pair: &KeyPair,
attempt: u32,
) -> Result<[u8; 32], Participant> {
DkgKeyPair::set(txn, spec.genesis(), attempt, key_pair);
KeyToDkgAttempt::set(txn, key_pair.0 .0, &attempt);
let preprocesses = ConfirmationNonces::get(txn, spec.genesis(), attempt).unwrap();
DkgConfirmer::new(key, spec, txn, attempt)
.expect("claiming to have generated a key pair for an unrecognized attempt")
.share(preprocesses, key_pair)
}
fn unflatten(
spec: &TributarySpec,
removed: &[<Ristretto as Ciphersuite>::G],
data: &mut HashMap<Participant, Vec<u8>>,
) {
for (validator, _) in spec.validators() {
let Some(range) = spec.i(removed, validator) else { continue };
let Some(all_segments) = data.remove(&range.start) else {
continue;
};
let mut data_vec = Vec::<_>::decode(&mut all_segments.as_slice()).unwrap();
for i in u16::from(range.start) .. u16::from(range.end) {
let i = Participant::new(i).unwrap();
data.insert(i, data_vec.remove(0));
}
}
}
impl<
D: Db,
T: DbTxn,
Pro: Processors,
PST: PublishSeraiTransaction,
PTT: PTTTrait,
RID: RIDTrait,
P: P2p,
> TributaryBlockHandler<'_, D, T, Pro, PST, PTT, RID, P>
{
fn accumulate(
&mut self,
removed: &[<Ristretto as Ciphersuite>::G],
data_spec: &DataSpecification,
signer: <Ristretto as Ciphersuite>::G,
data: &Vec<u8>,
) -> Accumulation {
log::debug!("accumulating entry for {:?} attempt #{}", &data_spec.topic, &data_spec.attempt);
let genesis = self.spec.genesis();
if DataDb::get(self.txn, genesis, data_spec, &signer.to_bytes()).is_some() {
panic!("accumulating data for a participant multiple times");
}
let signer_shares = {
let Some(signer_i) = self.spec.i(removed, signer) else {
log::warn!("accumulating data from {} who was removed", hex::encode(signer.to_bytes()));
return Accumulation::NotReady;
};
u16::from(signer_i.end) - u16::from(signer_i.start)
};
let prior_received = DataReceived::get(self.txn, genesis, data_spec).unwrap_or_default();
let now_received = prior_received + signer_shares;
DataReceived::set(self.txn, genesis, data_spec, &now_received);
DataDb::set(self.txn, genesis, data_spec, &signer.to_bytes(), data);
let received_range = (prior_received + 1) ..= now_received;
// If 2/3rds of the network participated in this preprocess, queue it for an automatic
// re-attempt
// DkgConfirmation doesn't have a re-attempt as it's just an extension for Dkg
if (data_spec.label == Label::Preprocess) &&
received_range.contains(&self.spec.t()) &&
(data_spec.topic != Topic::DkgConfirmation)
{
// Double check the attempt on this entry, as we don't want to schedule a re-attempt if this
// is an old entry
// This is an assert, not part of the if check, as old data shouldn't be here in the first
// place
assert_eq!(AttemptDb::attempt(self.txn, genesis, data_spec.topic), Some(data_spec.attempt));
ReattemptDb::schedule_reattempt(self.txn, genesis, self.block_number, data_spec.topic);
}
// If we have all the needed commitments/preprocesses/shares, tell the processor
let needs_everyone =
(data_spec.topic == Topic::Dkg) || (data_spec.topic == Topic::DkgConfirmation);
let needed = if needs_everyone { self.spec.n(removed) } else { self.spec.t() };
if received_range.contains(&needed) {
log::debug!(
"accumulation for entry {:?} attempt #{} is ready",
&data_spec.topic,
&data_spec.attempt
);
let mut data = HashMap::new();
for validator in self.spec.validators().iter().map(|validator| validator.0) {
let Some(i) = self.spec.i(removed, validator) else { continue };
data.insert(
i.start,
if let Some(data) = DataDb::get(self.txn, genesis, data_spec, &validator.to_bytes()) {
data
} else {
continue;
},
);
}
assert_eq!(data.len(), usize::from(needed));
// Remove our own piece of data, if we were involved
if let Some(i) = self.spec.i(removed, Ristretto::generator() * self.our_key.deref()) {
if data.remove(&i.start).is_some() {
return Accumulation::Ready(DataSet::Participating(data));
}
}
return Accumulation::Ready(DataSet::NotParticipating);
}
Accumulation::NotReady
}
fn handle_data(
&mut self,
removed: &[<Ristretto as Ciphersuite>::G],
data_spec: &DataSpecification,
bytes: &Vec<u8>,
signed: &Signed,
) -> Accumulation {
let genesis = self.spec.genesis();
let Some(curr_attempt) = AttemptDb::attempt(self.txn, genesis, data_spec.topic) else {
// Premature publication of a valid ID/publication of an invalid ID
self.fatal_slash(signed.signer.to_bytes(), "published data for ID without an attempt");
return Accumulation::NotReady;
};
// If they've already published a TX for this attempt, slash
// This shouldn't be reachable since nonces were made inserted by the coordinator, yet it's a
// cheap check to leave in for safety
if DataDb::get(self.txn, genesis, data_spec, &signed.signer.to_bytes()).is_some() {
self.fatal_slash(signed.signer.to_bytes(), "published data multiple times");
return Accumulation::NotReady;
}
// If the attempt is lesser than the blockchain's, return
if data_spec.attempt < curr_attempt {
log::debug!(
"dated attempt published onto tributary for topic {:?} (used attempt {}, current {})",
data_spec.topic,
data_spec.attempt,
curr_attempt
);
return Accumulation::NotReady;
}
// If the attempt is greater, this is a premature publication, full slash
if data_spec.attempt > curr_attempt {
self.fatal_slash(
signed.signer.to_bytes(),
"published data with an attempt which hasn't started",
);
return Accumulation::NotReady;
}
// TODO: We can also full slash if shares before all commitments, or share before the
// necessary preprocesses
// TODO: If this is shares, we need to check they are part of the selected signing set
// Accumulate this data
self.accumulate(removed, data_spec, signed.signer, bytes)
}
fn check_sign_data_len(
&mut self,
removed: &[<Ristretto as Ciphersuite>::G],
signer: <Ristretto as Ciphersuite>::G,
len: usize,
) -> Result<(), ()> {
let Some(signer_i) = self.spec.i(removed, signer) else {
// TODO: Ensure processor doesn't so participate/check how it handles removals for being
// offline
self.fatal_slash(signer.to_bytes(), "signer participated despite being removed");
Err(())?
};
if len != usize::from(u16::from(signer_i.end) - u16::from(signer_i.start)) {
self.fatal_slash(
signer.to_bytes(),
"signer published a distinct amount of sign data than they had shares",
);
Err(())?;
}
Ok(())
}
// TODO: Don't call fatal_slash in here, return the party to fatal_slash to ensure no further
// execution occurs
pub(crate) async fn handle_application_tx(&mut self, tx: Transaction) {
let genesis = self.spec.genesis();
// Don't handle transactions from fatally slashed participants
// This prevents removed participants from sabotaging the removal signing sessions and so on
// TODO: Because fatally slashed participants can still publish onto the blockchain, they have
// a notable DoS ability
if let TransactionKind::Signed(_, signed) = tx.kind() {
if FatallySlashed::get(self.txn, genesis, signed.signer.to_bytes()).is_some() {
return;
}
}
match tx {
Transaction::RemoveParticipantDueToDkg { participant, signed } => {
if self.spec.i(&[], participant).is_none() {
self.fatal_slash(
participant.to_bytes(),
"RemoveParticipantDueToDkg vote for non-validator",
);
return;
}
let participant = participant.to_bytes();
let signer = signed.signer.to_bytes();
assert!(
VotedToRemove::get(self.txn, genesis, signer, participant).is_none(),
"VotedToRemove multiple times despite a single nonce being allocated",
);
VotedToRemove::set(self.txn, genesis, signer, participant, &());
let prior_votes = VotesToRemove::get(self.txn, genesis, participant).unwrap_or(0);
let signer_votes =
self.spec.i(&[], signed.signer).expect("signer wasn't a validator for this network?");
let new_votes = prior_votes + u16::from(signer_votes.end) - u16::from(signer_votes.start);
VotesToRemove::set(self.txn, genesis, participant, &new_votes);
if ((prior_votes + 1) ..= new_votes).contains(&self.spec.t()) {
self.fatal_slash(participant, "RemoveParticipantDueToDkg vote")
}
}
Transaction::DkgCommitments { attempt, commitments, signed } => {
let Some(removed) = removed_as_of_dkg_attempt(self.txn, genesis, attempt) else {
self.fatal_slash(signed.signer.to_bytes(), "DkgCommitments with an unrecognized attempt");
return;
};
let Ok(()) = self.check_sign_data_len(&removed, signed.signer, commitments.len()) else {
return;
};
let data_spec = DataSpecification { topic: Topic::Dkg, label: Label::Preprocess, attempt };
match self.handle_data(&removed, &data_spec, &commitments.encode(), &signed) {
Accumulation::Ready(DataSet::Participating(mut commitments)) => {
log::info!("got all DkgCommitments for {}", hex::encode(genesis));
unflatten(self.spec, &removed, &mut commitments);
self
.processors
.send(
self.spec.set().network,
key_gen::CoordinatorMessage::Commitments {
id: KeyGenId { session: self.spec.set().session, attempt },
commitments,
},
)
.await;
}
Accumulation::Ready(DataSet::NotParticipating) => {
assert!(
removed.contains(&(Ristretto::generator() * self.our_key.deref())),
"NotParticipating in a DkgCommitments we weren't removed for"
);
}
Accumulation::NotReady => {}
}
}
Transaction::DkgShares { attempt, mut shares, confirmation_nonces, signed } => {
let Some(removed) = removed_as_of_dkg_attempt(self.txn, genesis, attempt) else {
self.fatal_slash(signed.signer.to_bytes(), "DkgShares with an unrecognized attempt");
return;
};
let not_participating = removed.contains(&(Ristretto::generator() * self.our_key.deref()));
let Ok(()) = self.check_sign_data_len(&removed, signed.signer, shares.len()) else {
return;
};
let Some(sender_i) = self.spec.i(&removed, signed.signer) else {
self.fatal_slash(
signed.signer.to_bytes(),
"DkgShares for a DKG they aren't participating in",
);
return;
};
let sender_is_len = u16::from(sender_i.end) - u16::from(sender_i.start);
for shares in &shares {
if shares.len() != (usize::from(self.spec.n(&removed) - sender_is_len)) {
self.fatal_slash(signed.signer.to_bytes(), "invalid amount of DKG shares");
return;
}
}
// Save each share as needed for blame
for (from_offset, shares) in shares.iter().enumerate() {
let from =
Participant::new(u16::from(sender_i.start) + u16::try_from(from_offset).unwrap())
.unwrap();
for (to_offset, share) in shares.iter().enumerate() {
// 0-indexed (the enumeration) to 1-indexed (Participant)
let mut to = u16::try_from(to_offset).unwrap() + 1;
// Adjust for the omission of the sender's own shares
if to >= u16::from(sender_i.start) {
to += u16::from(sender_i.end) - u16::from(sender_i.start);
}
let to = Participant::new(to).unwrap();
DkgShare::set(self.txn, genesis, from.into(), to.into(), share);
}
}
// Filter down to only our share's bytes for handle
let our_shares = if let Some(our_i) =
self.spec.i(&removed, Ristretto::generator() * self.our_key.deref())
{
if sender_i == our_i {
vec![]
} else {
// 1-indexed to 0-indexed
let mut our_i_pos = u16::from(our_i.start) - 1;
// Handle the omission of the sender's own data
if u16::from(our_i.start) > u16::from(sender_i.start) {
our_i_pos -= sender_is_len;
}
let our_i_pos = usize::from(our_i_pos);
shares
.iter_mut()
.map(|shares| {
shares
.drain(
our_i_pos ..
(our_i_pos + usize::from(u16::from(our_i.end) - u16::from(our_i.start))),
)
.collect::<Vec<_>>()
})
.collect()
}
} else {
assert!(
not_participating,
"we didn't have an i while handling DkgShares we weren't removed for"
);
// Since we're not participating, simply save vec![] for our shares
vec![]
};
// Drop shares as it's presumably been mutated into invalidity
drop(shares);
let data_spec = DataSpecification { topic: Topic::Dkg, label: Label::Share, attempt };
let encoded_data = (confirmation_nonces.to_vec(), our_shares.encode()).encode();
match self.handle_data(&removed, &data_spec, &encoded_data, &signed) {
Accumulation::Ready(DataSet::Participating(confirmation_nonces_and_shares)) => {
log::info!("got all DkgShares for {}", hex::encode(genesis));
let mut confirmation_nonces = HashMap::new();
let mut shares = HashMap::new();
for (participant, confirmation_nonces_and_shares) in confirmation_nonces_and_shares {
let (these_confirmation_nonces, these_shares) =
<(Vec<u8>, Vec<u8>)>::decode(&mut confirmation_nonces_and_shares.as_slice())
.unwrap();
confirmation_nonces.insert(participant, these_confirmation_nonces);
shares.insert(participant, these_shares);
}
ConfirmationNonces::set(self.txn, genesis, attempt, &confirmation_nonces);
// shares is a HashMap<Participant, Vec<Vec<Vec<u8>>>>, with the values representing:
// - Each of the sender's shares
// - Each of the our shares
// - Each share
// We need a Vec<HashMap<Participant, Vec<u8>>>, with the outer being each of ours
let mut expanded_shares = vec![];
for (sender_start_i, shares) in shares {
let shares: Vec<Vec<Vec<u8>>> = Vec::<_>::decode(&mut shares.as_slice()).unwrap();
for (sender_i_offset, our_shares) in shares.into_iter().enumerate() {
for (our_share_i, our_share) in our_shares.into_iter().enumerate() {
if expanded_shares.len() <= our_share_i {
expanded_shares.push(HashMap::new());
}
expanded_shares[our_share_i].insert(
Participant::new(
u16::from(sender_start_i) + u16::try_from(sender_i_offset).unwrap(),
)
.unwrap(),
our_share,
);
}
}
}
self
.processors
.send(
self.spec.set().network,
key_gen::CoordinatorMessage::Shares {
id: KeyGenId { session: self.spec.set().session, attempt },
shares: expanded_shares,
},
)
.await;
}
Accumulation::Ready(DataSet::NotParticipating) => {
assert!(not_participating, "NotParticipating in a DkgShares we weren't removed for");
}
Accumulation::NotReady => {}
}
}
Transaction::InvalidDkgShare { attempt, accuser, faulty, blame, signed } => {
let Some(removed) = removed_as_of_dkg_attempt(self.txn, genesis, attempt) else {
self
.fatal_slash(signed.signer.to_bytes(), "InvalidDkgShare with an unrecognized attempt");
return;
};
let Some(range) = self.spec.i(&removed, signed.signer) else {
self.fatal_slash(
signed.signer.to_bytes(),
"InvalidDkgShare for a DKG they aren't participating in",
);
return;
};
if !range.contains(&accuser) {
self.fatal_slash(
signed.signer.to_bytes(),
"accused with a Participant index which wasn't theirs",
);
return;
}
if range.contains(&faulty) {
self.fatal_slash(signed.signer.to_bytes(), "accused self of having an InvalidDkgShare");
return;
}
let Some(share) = DkgShare::get(self.txn, genesis, accuser.into(), faulty.into()) else {
self.fatal_slash(
signed.signer.to_bytes(),
"InvalidDkgShare had a non-existent faulty participant",
);
return;
};
self
.processors
.send(
self.spec.set().network,
key_gen::CoordinatorMessage::VerifyBlame {
id: KeyGenId { session: self.spec.set().session, attempt },
accuser,
accused: faulty,
share,
blame,
},
)
.await;
}
Transaction::DkgConfirmed { attempt, confirmation_share, signed } => {
let Some(removed) = removed_as_of_dkg_attempt(self.txn, genesis, attempt) else {
self.fatal_slash(signed.signer.to_bytes(), "DkgConfirmed with an unrecognized attempt");
return;
};
let data_spec =
DataSpecification { topic: Topic::DkgConfirmation, label: Label::Share, attempt };
match self.handle_data(&removed, &data_spec, &confirmation_share.to_vec(), &signed) {
Accumulation::Ready(DataSet::Participating(shares)) => {
log::info!("got all DkgConfirmed for {}", hex::encode(genesis));
let Some(removed) = removed_as_of_dkg_attempt(self.txn, genesis, attempt) else {
panic!(
"DkgConfirmed for everyone yet didn't have the removed parties for this attempt",
);
};
let preprocesses = ConfirmationNonces::get(self.txn, genesis, attempt).unwrap();
// TODO: This can technically happen under very very very specific timing as the txn
// put happens before DkgConfirmed, yet the txn commit isn't guaranteed to
let key_pair = DkgKeyPair::get(self.txn, genesis, attempt).expect(
"in DkgConfirmed handling, which happens after everyone \
(including us) fires DkgConfirmed, yet no confirming key pair",
);
let mut confirmer = DkgConfirmer::new(self.our_key, self.spec, self.txn, attempt)
.expect("confirming DKG for unrecognized attempt");
let sig = match confirmer.complete(preprocesses, &key_pair, shares) {
Ok(sig) => sig,
Err(p) => {
let mut tx = Transaction::RemoveParticipantDueToDkg {
participant: self.spec.reverse_lookup_i(&removed, p).unwrap(),
signed: Transaction::empty_signed(),
};
tx.sign(&mut OsRng, genesis, self.our_key);
self.publish_tributary_tx.publish_tributary_tx(tx).await;
return;
}
};
DkgLocallyCompleted::set(self.txn, genesis, &());
self
.publish_serai_tx
.publish_set_keys(
self.db,
self.spec.set(),
removed.into_iter().map(|key| key.to_bytes().into()).collect(),
key_pair,
Signature(sig),
)
.await;
}
Accumulation::Ready(DataSet::NotParticipating) => {
panic!("wasn't a participant in DKG confirmination shares")
}
Accumulation::NotReady => {}
}
}
Transaction::CosignSubstrateBlock(hash) => {
AttemptDb::recognize_topic(
self.txn,
genesis,
Topic::SubstrateSign(SubstrateSignableId::CosigningSubstrateBlock(hash)),
);
let block_number = SeraiBlockNumber::get(self.txn, hash)
.expect("CosignSubstrateBlock yet didn't save Serai block number");
let msg = coordinator::CoordinatorMessage::CosignSubstrateBlock {
id: SubstrateSignId {
session: self.spec.set().session,
id: SubstrateSignableId::CosigningSubstrateBlock(hash),
attempt: 0,
},
block_number,
};
self.processors.send(self.spec.set().network, msg).await;
}
Transaction::Batch { block: _, batch } => {
// Because this Batch has achieved synchrony, its batch ID should be authorized
AttemptDb::recognize_topic(
self.txn,
genesis,
Topic::SubstrateSign(SubstrateSignableId::Batch(batch)),
);
self
.recognized_id
.recognized_id(
self.spec.set(),
genesis,
RecognizedIdType::Batch,
batch.to_le_bytes().to_vec(),
)
.await;
}
Transaction::SubstrateBlock(block) => {
let plan_ids = PlanIds::get(self.txn, &genesis, block).expect(
"synced a tributary block finalizing a substrate block in a provided transaction \
despite us not providing that transaction",
);
for id in plan_ids {
AttemptDb::recognize_topic(self.txn, genesis, Topic::Sign(id));
self
.recognized_id
.recognized_id(self.spec.set(), genesis, RecognizedIdType::Plan, id.to_vec())
.await;
}
}
Transaction::SubstrateSign(data) => {
// Provided transactions ensure synchrony on any signing protocol, and we won't start
// signing with threshold keys before we've confirmed them on-chain
let Some(removed) =
crate::tributary::removed_as_of_set_keys(self.txn, self.spec.set(), genesis)
else {
self.fatal_slash(
data.signed.signer.to_bytes(),
"signing despite not having set keys on substrate",
);
return;
};
let signer = data.signed.signer;
let Ok(()) = self.check_sign_data_len(&removed, signer, data.data.len()) else {
return;
};
let expected_len = match data.label {
Label::Preprocess => 64,
Label::Share => 32,
};
for data in &data.data {
if data.len() != expected_len {
self.fatal_slash(
signer.to_bytes(),
"unexpected length data for substrate signing protocol",
);
return;
}
}
let data_spec = DataSpecification {
topic: Topic::SubstrateSign(data.plan),
label: data.label,
attempt: data.attempt,
};
let Accumulation::Ready(DataSet::Participating(mut results)) =
self.handle_data(&removed, &data_spec, &data.data.encode(), &data.signed)
else {
return;
};
unflatten(self.spec, &removed, &mut results);
let id = SubstrateSignId {
session: self.spec.set().session,
id: data.plan,
attempt: data.attempt,
};
let msg = match data.label {
Label::Preprocess => coordinator::CoordinatorMessage::SubstratePreprocesses {
id,
preprocesses: results.into_iter().map(|(v, p)| (v, p.try_into().unwrap())).collect(),
},
Label::Share => coordinator::CoordinatorMessage::SubstrateShares {
id,
shares: results.into_iter().map(|(v, p)| (v, p.try_into().unwrap())).collect(),
},
};
self.processors.send(self.spec.set().network, msg).await;
}
Transaction::Sign(data) => {
let Some(removed) =
crate::tributary::removed_as_of_set_keys(self.txn, self.spec.set(), genesis)
else {
self.fatal_slash(
data.signed.signer.to_bytes(),
"signing despite not having set keys on substrate",
);
return;
};
let Ok(()) = self.check_sign_data_len(&removed, data.signed.signer, data.data.len()) else {
return;
};
let data_spec = DataSpecification {
topic: Topic::Sign(data.plan),
label: data.label,
attempt: data.attempt,
};
if let Accumulation::Ready(DataSet::Participating(mut results)) =
self.handle_data(&removed, &data_spec, &data.data.encode(), &data.signed)
{
unflatten(self.spec, &removed, &mut results);
let id =
SignId { session: self.spec.set().session, id: data.plan, attempt: data.attempt };
self
.processors
.send(
self.spec.set().network,
match data.label {
Label::Preprocess => {
sign::CoordinatorMessage::Preprocesses { id, preprocesses: results }
}
Label::Share => sign::CoordinatorMessage::Shares { id, shares: results },
},
)
.await;
}
}
Transaction::SignCompleted { plan, tx_hash, first_signer, signature: _ } => {
log::info!(
"on-chain SignCompleted claims {} completes {}",
hex::encode(&tx_hash),
hex::encode(plan)
);
if AttemptDb::attempt(self.txn, genesis, Topic::Sign(plan)).is_none() {
self.fatal_slash(first_signer.to_bytes(), "claimed an unrecognized plan was completed");
return;
};
// TODO: Confirm this signer hasn't prior published a completion
let msg = sign::CoordinatorMessage::Completed {
session: self.spec.set().session,
id: plan,
tx: tx_hash,
};
self.processors.send(self.spec.set().network, msg).await;
}
Transaction::SlashReport(points, signed) => {
// Uses &[] as we only need the length which is independent to who else was removed
let signer_range = self.spec.i(&[], signed.signer).unwrap();
let signer_len = u16::from(signer_range.end) - u16::from(signer_range.start);
if points.len() != (self.spec.validators().len() - 1) {
self.fatal_slash(
signed.signer.to_bytes(),
"submitted a distinct amount of slash points to participants",
);
return;
}
if SlashReports::get(self.txn, genesis, signed.signer.to_bytes()).is_some() {
self.fatal_slash(signed.signer.to_bytes(), "submitted multiple slash points");
return;
}
SlashReports::set(self.txn, genesis, signed.signer.to_bytes(), &points);
let prior_reported = SlashReported::get(self.txn, genesis).unwrap_or(0);
let now_reported = prior_reported + signer_len;
SlashReported::set(self.txn, genesis, &now_reported);
if (prior_reported < self.spec.t()) && (now_reported >= self.spec.t()) {
SlashReportCutOff::set(
self.txn,
genesis,
// 30 minutes into the future
&(u64::from(self.block_number) +
((30 * 60 * 1000) / u64::from(tributary::tendermint::TARGET_BLOCK_TIME))),
);
}
}
}
}
}

View File

@@ -1,100 +0,0 @@
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use serai_client::validator_sets::primitives::ValidatorSet;
use tributary::{
ReadWrite,
transaction::{TransactionError, TransactionKind, Transaction as TransactionTrait},
Tributary,
};
mod db;
pub use db::*;
mod spec;
pub use spec::TributarySpec;
mod transaction;
pub use transaction::{Label, SignData, Transaction};
mod signing_protocol;
mod handle;
pub use handle::*;
pub mod scanner;
pub fn removed_as_of_dkg_attempt(
getter: &impl Get,
genesis: [u8; 32],
attempt: u32,
) -> Option<Vec<<Ristretto as Ciphersuite>::G>> {
if attempt == 0 {
Some(vec![])
} else {
RemovedAsOfDkgAttempt::get(getter, genesis, attempt).map(|keys| {
keys.iter().map(|key| <Ristretto as Ciphersuite>::G::from_bytes(key).unwrap()).collect()
})
}
}
pub fn removed_as_of_set_keys(
getter: &impl Get,
set: ValidatorSet,
genesis: [u8; 32],
) -> Option<Vec<<Ristretto as Ciphersuite>::G>> {
// SeraiDkgCompleted has the key placed on-chain.
// This key can be uniquely mapped to an attempt so long as one participant was honest, which we
// assume as a presumably honest participant.
// Resolve from generated key to attempt to fatally slashed as of attempt.
// This expect will trigger if this is prematurely called and Substrate has tracked the keys yet
// we haven't locally synced and handled the Tributary
// All callers of this, at the time of writing, ensure the Tributary has sufficiently synced
// making the panic with context more desirable than the None
let attempt = KeyToDkgAttempt::get(getter, SeraiDkgCompleted::get(getter, set)?)
.expect("key completed on-chain didn't have an attempt related");
removed_as_of_dkg_attempt(getter, genesis, attempt)
}
pub async fn publish_signed_transaction<D: Db, P: crate::P2p>(
txn: &mut D::Transaction<'_>,
tributary: &Tributary<D, Transaction, P>,
tx: Transaction,
) {
log::debug!("publishing transaction {}", hex::encode(tx.hash()));
let (order, signer) = if let TransactionKind::Signed(order, signed) = tx.kind() {
let signer = signed.signer;
// Safe as we should deterministically create transactions, meaning if this is already on-disk,
// it's what we're saving now
SignedTransactionDb::set(txn, &order, signed.nonce, &tx.serialize());
(order, signer)
} else {
panic!("non-signed transaction passed to publish_signed_transaction");
};
// If we're trying to publish 5, when the last transaction published was 3, this will delay
// publication until the point in time we publish 4
while let Some(tx) = SignedTransactionDb::take_signed_transaction(
txn,
&order,
tributary
.next_nonce(&signer, &order)
.await
.expect("we don't have a nonce, meaning we aren't a participant on this tributary"),
) {
// We need to return a proper error here to enable that, due to a race condition around
// multiple publications
match tributary.add_transaction(tx.clone()).await {
Ok(_) => {}
// Some asynchonicity if InvalidNonce, assumed safe to deterministic nonces
Err(TransactionError::InvalidNonce) => {
log::warn!("publishing TX {tx:?} returned InvalidNonce. was it already added?")
}
Err(e) => panic!("created an invalid transaction: {e:?}"),
}
}
}

View File

@@ -1,804 +0,0 @@
use core::{marker::PhantomData, ops::Deref, future::Future, time::Duration};
use std::{sync::Arc, collections::HashSet};
use zeroize::Zeroizing;
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use tokio::sync::broadcast;
use scale::{Encode, Decode};
use serai_client::{
primitives::{SeraiAddress, Signature},
validator_sets::primitives::{KeyPair, ValidatorSet},
Serai,
};
use serai_db::DbTxn;
use processor_messages::coordinator::{SubstrateSignId, SubstrateSignableId};
use tributary::{
TransactionKind, Transaction as TributaryTransaction, TransactionError, Block, TributaryReader,
tendermint::{
tx::{TendermintTx, Evidence, decode_signed_message},
TendermintNetwork,
},
};
use crate::{Db, processors::Processors, substrate::BatchInstructionsHashDb, tributary::*, P2p};
#[derive(Clone, Copy, PartialEq, Eq, Debug, Encode, Decode)]
pub enum RecognizedIdType {
Batch,
Plan,
}
#[async_trait::async_trait]
pub trait RIDTrait {
async fn recognized_id(
&self,
set: ValidatorSet,
genesis: [u8; 32],
kind: RecognizedIdType,
id: Vec<u8>,
);
}
#[async_trait::async_trait]
impl<
FRid: Send + Future<Output = ()>,
F: Sync + Fn(ValidatorSet, [u8; 32], RecognizedIdType, Vec<u8>) -> FRid,
> RIDTrait for F
{
async fn recognized_id(
&self,
set: ValidatorSet,
genesis: [u8; 32],
kind: RecognizedIdType,
id: Vec<u8>,
) {
(self)(set, genesis, kind, id).await
}
}
#[async_trait::async_trait]
pub trait PublishSeraiTransaction {
async fn publish_set_keys(
&self,
db: &(impl Sync + Get),
set: ValidatorSet,
removed: Vec<SeraiAddress>,
key_pair: KeyPair,
signature: Signature,
);
}
mod impl_pst_for_serai {
use super::*;
use serai_client::SeraiValidatorSets;
// Uses a macro because Rust can't resolve the lifetimes/generics around the check function
// check is expected to return true if the effect has already occurred
// The generated publish function will return true if *we* published the transaction
macro_rules! common_pst {
($Meta: ty, $check: ident) => {
async fn publish(
serai: &Serai,
db: &impl Get,
set: ValidatorSet,
tx: serai_client::Transaction,
meta: $Meta,
) -> bool {
loop {
match serai.publish(&tx).await {
Ok(_) => return true,
// This is assumed to be some ephemeral error due to the assumed fault-free
// creation
// TODO2: Differentiate connection errors from invariants
Err(e) => {
// The following block is irrelevant, and can/likely will fail, if we're publishing
// a TX for an old session
// If we're on a newer session, move on
if crate::RetiredTributaryDb::get(db, set).is_some() {
log::warn!("trying to publish a TX relevant to set {set:?} which isn't the latest");
return false;
}
if let Ok(serai) = serai.as_of_latest_finalized_block().await {
let serai = serai.validator_sets();
// Check if someone else published the TX in question
if $check(serai, set, meta).await {
return false;
}
}
log::error!("couldn't connect to Serai node to publish TX: {e:?}");
tokio::time::sleep(core::time::Duration::from_secs(5)).await;
}
}
}
}
};
}
#[async_trait::async_trait]
impl PublishSeraiTransaction for Serai {
async fn publish_set_keys(
&self,
db: &(impl Sync + Get),
set: ValidatorSet,
removed: Vec<SeraiAddress>,
key_pair: KeyPair,
signature: Signature,
) {
// TODO: BoundedVec as an arg to avoid this expect
let tx = SeraiValidatorSets::set_keys(
set.network,
removed.try_into().expect("removing more than allowed"),
key_pair,
signature,
);
async fn check(serai: SeraiValidatorSets<'_>, set: ValidatorSet, (): ()) -> bool {
if matches!(serai.keys(set).await, Ok(Some(_))) {
log::info!("another coordinator set key pair for {:?}", set);
return true;
}
false
}
common_pst!((), check);
if publish(self, db, set, tx, ()).await {
log::info!("published set keys for {set:?}");
}
}
}
}
#[async_trait::async_trait]
pub trait PTTTrait {
async fn publish_tributary_tx(&self, tx: Transaction);
}
#[async_trait::async_trait]
impl<FPtt: Send + Future<Output = ()>, F: Sync + Fn(Transaction) -> FPtt> PTTTrait for F {
async fn publish_tributary_tx(&self, tx: Transaction) {
(self)(tx).await
}
}
pub struct TributaryBlockHandler<
'a,
D: Db,
T: DbTxn,
Pro: Processors,
PST: PublishSeraiTransaction,
PTT: PTTTrait,
RID: RIDTrait,
P: P2p,
> {
pub db: &'a D,
pub txn: &'a mut T,
pub our_key: &'a Zeroizing<<Ristretto as Ciphersuite>::F>,
pub recognized_id: &'a RID,
pub processors: &'a Pro,
pub publish_serai_tx: &'a PST,
pub publish_tributary_tx: &'a PTT,
pub spec: &'a TributarySpec,
block: Block<Transaction>,
pub block_number: u32,
_p2p: PhantomData<P>,
}
impl<
D: Db,
T: DbTxn,
Pro: Processors,
PST: PublishSeraiTransaction,
PTT: PTTTrait,
RID: RIDTrait,
P: P2p,
> TributaryBlockHandler<'_, D, T, Pro, PST, PTT, RID, P>
{
pub fn fatal_slash(&mut self, slashing: [u8; 32], reason: &str) {
let genesis = self.spec.genesis();
log::warn!("fatally slashing {}. reason: {}", hex::encode(slashing), reason);
FatallySlashed::set_fatally_slashed(self.txn, genesis, slashing);
// TODO: disconnect the node from network/ban from further participation in all Tributaries
}
// TODO: Once Substrate confirms a key, we need to rotate our validator set OR form a second
// Tributary post-DKG
// https://github.com/serai-dex/serai/issues/426
async fn handle(mut self) {
log::info!("found block for Tributary {:?}", self.spec.set());
let transactions = self.block.transactions.clone();
for tx in transactions {
match tx {
TributaryTransaction::Tendermint(TendermintTx::SlashEvidence(ev)) => {
// Since the evidence is on the chain, it should already have been validated
// We can just punish the signer
let data = match ev {
Evidence::ConflictingMessages(first, second) => (first, Some(second)),
Evidence::InvalidPrecommit(first) | Evidence::InvalidValidRound(first) => (first, None),
};
let msgs = (
decode_signed_message::<TendermintNetwork<D, Transaction, P>>(&data.0).unwrap(),
if data.1.is_some() {
Some(
decode_signed_message::<TendermintNetwork<D, Transaction, P>>(&data.1.unwrap())
.unwrap(),
)
} else {
None
},
);
// Since anything with evidence is fundamentally faulty behavior, not just temporal
// errors, mark the node as fatally slashed
self.fatal_slash(msgs.0.msg.sender, &format!("invalid tendermint messages: {msgs:?}"));
}
TributaryTransaction::Application(tx) => {
self.handle_application_tx(tx).await;
}
}
}
let genesis = self.spec.genesis();
let current_fatal_slashes = FatalSlashes::get_as_keys(self.txn, genesis);
// Calculate the shares still present, spinning if not enough are
// still_present_shares is used by a below branch, yet it's a natural byproduct of checking if
// we should spin, hence storing it in a variable here
let still_present_shares = {
// Start with the original n value
let mut present_shares = self.spec.n(&[]);
// Remove everyone fatally slashed
for removed in &current_fatal_slashes {
let original_i_for_removed =
self.spec.i(&[], *removed).expect("removed party was never present");
let removed_shares =
u16::from(original_i_for_removed.end) - u16::from(original_i_for_removed.start);
present_shares -= removed_shares;
}
// Spin if the present shares don't satisfy the required threshold
if present_shares < self.spec.t() {
loop {
log::error!(
"fatally slashed so many participants for {:?} we no longer meet the threshold",
self.spec.set()
);
tokio::time::sleep(core::time::Duration::from_secs(60)).await;
}
}
present_shares
};
for topic in ReattemptDb::take(self.txn, genesis, self.block_number) {
let attempt = AttemptDb::start_next_attempt(self.txn, genesis, topic);
log::info!("re-attempting {topic:?} with attempt {attempt}");
// Slash people who failed to participate as expected in the prior attempt
{
let prior_attempt = attempt - 1;
let (removed, expected_participants) = match topic {
Topic::Dkg => {
// Every validator who wasn't removed is expected to have participated
let removed =
crate::tributary::removed_as_of_dkg_attempt(self.txn, genesis, prior_attempt)
.expect("prior attempt didn't have its removed saved to disk");
let removed_set = removed.iter().copied().collect::<HashSet<_>>();
(
removed,
self
.spec
.validators()
.into_iter()
.filter_map(|(validator, _)| {
Some(validator).filter(|validator| !removed_set.contains(validator))
})
.collect(),
)
}
Topic::DkgConfirmation => {
panic!("TODO: re-attempting DkgConfirmation when we should be re-attempting the Dkg")
}
Topic::SubstrateSign(_) | Topic::Sign(_) => {
let removed =
crate::tributary::removed_as_of_set_keys(self.txn, self.spec.set(), genesis)
.expect("SubstrateSign/Sign yet have yet to set keys");
// TODO: If 67% sent preprocesses, this should be them. Else, this should be vec![]
let expected_participants = vec![];
(removed, expected_participants)
}
};
let (expected_topic, expected_label) = match topic {
Topic::Dkg => {
let n = self.spec.n(&removed);
// If we got all the DKG shares, we should be on DKG confirmation
let share_spec =
DataSpecification { topic: Topic::Dkg, label: Label::Share, attempt: prior_attempt };
if DataReceived::get(self.txn, genesis, &share_spec).unwrap_or(0) == n {
// Label::Share since there is no Label::Preprocess for DkgConfirmation since the
// preprocess is part of Topic::Dkg Label::Share
(Topic::DkgConfirmation, Label::Share)
} else {
let preprocess_spec = DataSpecification {
topic: Topic::Dkg,
label: Label::Preprocess,
attempt: prior_attempt,
};
// If we got all the DKG preprocesses, DKG shares
if DataReceived::get(self.txn, genesis, &preprocess_spec).unwrap_or(0) == n {
// Label::Share since there is no Label::Preprocess for DkgConfirmation since the
// preprocess is part of Topic::Dkg Label::Share
(Topic::Dkg, Label::Share)
} else {
(Topic::Dkg, Label::Preprocess)
}
}
}
Topic::DkgConfirmation => unreachable!(),
// If we got enough participants to move forward, then we expect shares from them all
Topic::SubstrateSign(_) | Topic::Sign(_) => (topic, Label::Share),
};
let mut did_not_participate = vec![];
for expected_participant in expected_participants {
if DataDb::get(
self.txn,
genesis,
&DataSpecification {
topic: expected_topic,
label: expected_label,
attempt: prior_attempt,
},
&expected_participant.to_bytes(),
)
.is_none()
{
did_not_participate.push(expected_participant);
}
}
// If a supermajority didn't participate as expected, the protocol was likely aborted due
// to detection of a completion or some larger networking error
// Accordingly, clear did_not_participate
// TODO
// If during the DKG, explicitly mark these people as having been offline
// TODO: If they were offline sufficiently long ago, don't strike them off
if topic == Topic::Dkg {
let mut existing = OfflineDuringDkg::get(self.txn, genesis).unwrap_or(vec![]);
for did_not_participate in did_not_participate {
existing.push(did_not_participate.to_bytes());
}
OfflineDuringDkg::set(self.txn, genesis, &existing);
}
// Slash everyone who didn't participate as expected
// This may be overzealous as if a minority detects a completion, they'll abort yet the
// supermajority will cause the above allowance to not trigger, causing an honest minority
// to be slashed
// At the end of the protocol, the accumulated slashes are reduced by the amount obtained
// by the worst-performing member of the supermajority, and this is expected to
// sufficiently compensate for slashes which occur under normal operation
// TODO
}
/*
All of these have the same common flow:
1) Check if this re-attempt is actually needed
2) If so, dispatch whatever events as needed
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).
*/
match topic {
Topic::Dkg => {
let mut removed = current_fatal_slashes.clone();
let t = self.spec.t();
{
let mut present_shares = still_present_shares;
// Load the parties marked as offline across the various attempts
let mut offline = OfflineDuringDkg::get(self.txn, genesis)
.unwrap_or(vec![])
.iter()
.map(|key| <Ristretto as Ciphersuite>::G::from_bytes(key).unwrap())
.collect::<Vec<_>>();
// Pop from the list to prioritize the removal of those recently offline
while let Some(offline) = offline.pop() {
// Make sure they weren't removed already (such as due to being fatally slashed)
// This also may trigger if they were offline across multiple attempts
if removed.contains(&offline) {
continue;
}
// If we can remove them and still meet the threshold, do so
let original_i_for_offline =
self.spec.i(&[], offline).expect("offline was never present?");
let offline_shares =
u16::from(original_i_for_offline.end) - u16::from(original_i_for_offline.start);
if (present_shares - offline_shares) >= t {
present_shares -= offline_shares;
removed.push(offline);
}
// If we've removed as many people as we can, break
if present_shares == t {
break;
}
}
}
RemovedAsOfDkgAttempt::set(
self.txn,
genesis,
attempt,
&removed.iter().map(<Ristretto as Ciphersuite>::G::to_bytes).collect(),
);
if DkgLocallyCompleted::get(self.txn, genesis).is_none() {
let Some(our_i) = self.spec.i(&removed, Ristretto::generator() * self.our_key.deref())
else {
continue;
};
// Since it wasn't completed, instruct the processor to start the next attempt
let id =
processor_messages::key_gen::KeyGenId { session: self.spec.set().session, attempt };
let params =
frost::ThresholdParams::new(t, self.spec.n(&removed), our_i.start).unwrap();
let shares = u16::from(our_i.end) - u16::from(our_i.start);
self
.processors
.send(
self.spec.set().network,
processor_messages::key_gen::CoordinatorMessage::GenerateKey { id, params, shares },
)
.await;
}
}
Topic::DkgConfirmation => unreachable!(),
Topic::SubstrateSign(inner_id) => {
let id = processor_messages::coordinator::SubstrateSignId {
session: self.spec.set().session,
id: inner_id,
attempt,
};
match inner_id {
SubstrateSignableId::CosigningSubstrateBlock(block) => {
let block_number = SeraiBlockNumber::get(self.txn, block)
.expect("couldn't get the block number for prior attempted cosign");
// Check if the cosigner has a signature from our set for this block/a newer one
let latest_cosign =
crate::cosign_evaluator::LatestCosign::get(self.txn, self.spec.set().network)
.map_or(0, |cosign| cosign.block_number);
if latest_cosign < block_number {
// Instruct the processor to start the next attempt
self
.processors
.send(
self.spec.set().network,
processor_messages::coordinator::CoordinatorMessage::CosignSubstrateBlock {
id,
block_number,
},
)
.await;
}
}
SubstrateSignableId::Batch(batch) => {
// If the Batch hasn't appeared on-chain...
if BatchInstructionsHashDb::get(self.txn, self.spec.set().network, batch).is_none() {
// Instruct the processor to start the next attempt
// The processor won't continue if it's already signed a Batch
// Prior checking if the Batch is on-chain just may reduce the non-participating
// 33% from publishing their re-attempt messages
self
.processors
.send(
self.spec.set().network,
processor_messages::coordinator::CoordinatorMessage::BatchReattempt { id },
)
.await;
}
}
SubstrateSignableId::SlashReport => {
// If this Tributary hasn't been retired...
// (published SlashReport/took too long to do so)
if crate::RetiredTributaryDb::get(self.txn, self.spec.set()).is_none() {
let report = SlashReport::get(self.txn, self.spec.set())
.expect("re-attempting signing a SlashReport we don't have?");
self
.processors
.send(
self.spec.set().network,
processor_messages::coordinator::CoordinatorMessage::SignSlashReport {
id,
report,
},
)
.await;
}
}
}
}
Topic::Sign(id) => {
// Instruct the processor to start the next attempt
// If it has already noted a completion, it won't send a preprocess and will simply drop
// the re-attempt message
self
.processors
.send(
self.spec.set().network,
processor_messages::sign::CoordinatorMessage::Reattempt {
id: processor_messages::sign::SignId {
session: self.spec.set().session,
id,
attempt,
},
},
)
.await;
}
}
}
if Some(u64::from(self.block_number)) == SlashReportCutOff::get(self.txn, genesis) {
// Grab every slash report
let mut all_reports = vec![];
for (i, (validator, _)) in self.spec.validators().into_iter().enumerate() {
let Some(mut report) = SlashReports::get(self.txn, genesis, validator.to_bytes()) else {
continue;
};
// Assign them 0 points for themselves
report.insert(i, 0);
// Uses &[] as we only need the length which is independent to who else was removed
let signer_i = self.spec.i(&[], validator).unwrap();
let signer_len = u16::from(signer_i.end) - u16::from(signer_i.start);
// Push `n` copies, one for each of their shares
for _ in 0 .. signer_len {
all_reports.push(report.clone());
}
}
// For each participant, grab their median
let mut medians = vec![];
for p in 0 .. self.spec.validators().len() {
let mut median_calc = vec![];
for report in &all_reports {
median_calc.push(report[p]);
}
median_calc.sort_unstable();
medians.push(median_calc[median_calc.len() / 2]);
}
// Grab the points of the last party within the best-performing threshold
// This is done by first expanding the point values by the amount of shares
let mut sorted_medians = vec![];
for (i, (_, shares)) in self.spec.validators().into_iter().enumerate() {
for _ in 0 .. shares {
sorted_medians.push(medians[i]);
}
}
// Then performing the sort
sorted_medians.sort_unstable();
let worst_points_by_party_within_threshold = sorted_medians[usize::from(self.spec.t()) - 1];
// Reduce everyone's points by this value
for median in &mut medians {
*median = median.saturating_sub(worst_points_by_party_within_threshold);
}
// The threshold now has the proper incentive to report this as they no longer suffer
// negative effects
//
// Additionally, if all validators had degraded performance, they don't all get penalized for
// what's likely outside their control (as it occurred universally)
// Mark everyone fatally slashed with u32::MAX
for (i, (validator, _)) in self.spec.validators().into_iter().enumerate() {
if FatallySlashed::get(self.txn, genesis, validator.to_bytes()).is_some() {
medians[i] = u32::MAX;
}
}
let mut report = vec![];
for (i, (validator, _)) in self.spec.validators().into_iter().enumerate() {
if medians[i] != 0 {
report.push((validator.to_bytes(), medians[i]));
}
}
// This does lock in the report, meaning further slash point accumulations won't be reported
// They still have value to be locally tracked due to local decisions made based off
// accumulated slash reports
SlashReport::set(self.txn, self.spec.set(), &report);
// Start a signing protocol for this
self
.processors
.send(
self.spec.set().network,
processor_messages::coordinator::CoordinatorMessage::SignSlashReport {
id: SubstrateSignId {
session: self.spec.set().session,
id: SubstrateSignableId::SlashReport,
attempt: 0,
},
report,
},
)
.await;
}
}
}
#[allow(clippy::too_many_arguments)]
pub(crate) async fn handle_new_blocks<
D: Db,
Pro: Processors,
PST: PublishSeraiTransaction,
PTT: PTTTrait,
RID: RIDTrait,
P: P2p,
>(
db: &mut D,
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
recognized_id: &RID,
processors: &Pro,
publish_serai_tx: &PST,
publish_tributary_tx: &PTT,
spec: &TributarySpec,
tributary: &TributaryReader<D, Transaction>,
) {
let genesis = tributary.genesis();
let mut last_block = LastHandledBlock::get(db, genesis).unwrap_or(genesis);
let mut block_number = TributaryBlockNumber::get(db, last_block).unwrap_or(0);
while let Some(next) = tributary.block_after(&last_block) {
let block = tributary.block(&next).unwrap();
block_number += 1;
// Make sure we have all of the provided transactions for this block
for tx in &block.transactions {
// Provided TXs will appear first in the Block, so we can break after we hit a non-Provided
let TransactionKind::Provided(order) = tx.kind() else {
break;
};
// make sure we have all the provided txs in this block locally
if !tributary.locally_provided_txs_in_block(&block.hash(), order) {
return;
}
}
let mut db_clone = db.clone();
let mut txn = db_clone.txn();
TributaryBlockNumber::set(&mut txn, next, &block_number);
(TributaryBlockHandler {
db,
txn: &mut txn,
spec,
our_key: key,
recognized_id,
processors,
publish_serai_tx,
publish_tributary_tx,
block,
block_number,
_p2p: PhantomData::<P>,
})
.handle()
.await;
last_block = next;
LastHandledBlock::set(&mut txn, genesis, &next);
txn.commit();
}
}
pub(crate) async fn scan_tributaries_task<
D: Db,
Pro: Processors,
P: P2p,
RID: 'static + Send + Sync + Clone + RIDTrait,
>(
raw_db: D,
key: Zeroizing<<Ristretto as Ciphersuite>::F>,
recognized_id: RID,
processors: Pro,
serai: Arc<Serai>,
mut tributary_event: broadcast::Receiver<crate::TributaryEvent<D, P>>,
) {
log::info!("scanning tributaries");
loop {
match tributary_event.recv().await {
Ok(crate::TributaryEvent::NewTributary(crate::ActiveTributary { spec, tributary })) => {
// For each Tributary, spawn a dedicated scanner task
tokio::spawn({
let raw_db = raw_db.clone();
let key = key.clone();
let recognized_id = recognized_id.clone();
let processors = processors.clone();
let serai = serai.clone();
async move {
let spec = &spec;
let reader = tributary.reader();
let mut tributary_db = raw_db.clone();
loop {
// Check if the set was retired, and if so, don't further operate
if crate::db::RetiredTributaryDb::get(&raw_db, spec.set()).is_some() {
break;
}
// Obtain the next block notification now to prevent obtaining it immediately after
// the next block occurs
let next_block_notification = tributary.next_block_notification().await;
handle_new_blocks::<_, _, _, _, _, P>(
&mut tributary_db,
&key,
&recognized_id,
&processors,
&*serai,
&|tx: Transaction| {
let tributary = tributary.clone();
async move {
match tributary.add_transaction(tx.clone()).await {
Ok(_) => {}
// Can happen as this occurs on a distinct DB TXN
Err(TransactionError::InvalidNonce) => {
log::warn!(
"publishing TX {tx:?} returned InvalidNonce. was it already added?"
)
}
Err(e) => panic!("created an invalid transaction: {e:?}"),
}
}
},
spec,
&reader,
)
.await;
// Run either when the notification fires, or every interval of block_time
let _ = tokio::time::timeout(
Duration::from_secs(tributary::Tributary::<D, Transaction, P>::block_time().into()),
next_block_notification,
)
.await;
}
}
});
}
// The above loop simply checks the DB every few seconds, voiding the need for this event
Ok(crate::TributaryEvent::TributaryRetired(_)) => {}
Err(broadcast::error::RecvError::Lagged(_)) => {
panic!("scan_tributaries lagged to handle tributary_event")
}
Err(broadcast::error::RecvError::Closed) => panic!("tributary_event sender closed"),
}
}
}

View File

@@ -1,331 +0,0 @@
/*
A MuSig-based signing protocol executed with the validators' keys.
This is used for confirming the results of a DKG on-chain, an operation requiring all validators
which aren't specified as removed while still satisfying a supermajority.
Since we're using the validator's keys, as needed for their being the root of trust, the
coordinator must perform the signing. This is distinct from all other group-signing operations,
as they're all done by the processor.
The MuSig-aggregation achieves on-chain efficiency and enables a more secure design pattern.
While we could individually tack votes, that'd require logic to prevent voting multiple times and
tracking the accumulated votes. MuSig-aggregation simply requires checking the list is sorted and
the list's weight exceeds the threshold.
Instead of maintaining state in memory, a combination of the DB and re-execution are used. This
is deemed acceptable re: performance as:
1) This is only done prior to a DKG being confirmed on Substrate and is assumed infrequent.
2) This is an O(n) algorithm.
3) The size of the validator set is bounded by MAX_KEY_SHARES_PER_SET.
Accordingly, this should be tolerable.
As for safety, it is explicitly unsafe to reuse nonces across signing sessions. This raises
concerns regarding our re-execution which is dependent on fixed nonces. Safety is derived from
the nonces being context-bound under a BFT protocol. The flow is as follows:
1) Decide the nonce.
2) Publish the nonces' commitments, receiving everyone elses *and potentially the message to be
signed*.
3) Sign and publish the signature share.
In order for nonce re-use to occur, the received nonce commitments (or the message to be signed)
would have to be distinct and sign would have to be called again.
Before we act on any received messages, they're ordered and finalized by a BFT algorithm. The
only way to operate on distinct received messages would be if:
1) A logical flaw exists, letting new messages over write prior messages
2) A reorganization occurred from chain A to chain B, and with it, different messages
Reorganizations are not supported, as BFT is assumed by the presence of a BFT algorithm. While
a significant amount of processes may be byzantine, leading to BFT being broken, that still will
not trigger a reorganization. The only way to move to a distinct chain, with distinct messages,
would be by rebuilding the local process (this time following chain B). Upon any complete
rebuild, we'd re-decide nonces, achieving safety. This does set a bound preventing partial
rebuilds which is accepted.
Additionally, to ensure a rebuilt service isn't flagged as malicious, we have to check the
commitments generated from the decided nonces are in fact its commitments on-chain (TODO).
TODO: We also need to review how we're handling Processor preprocesses and likely implement the
same on-chain-preprocess-matches-presumed-preprocess check before publishing shares.
*/
use core::ops::Deref;
use std::collections::HashMap;
use zeroize::{Zeroize, Zeroizing};
use rand_core::OsRng;
use blake2::{Digest, Blake2s256};
use ciphersuite::{
group::{ff::PrimeField, GroupEncoding},
Ciphersuite, Ristretto,
};
use frost::{
FrostError,
dkg::{Participant, musig::musig},
ThresholdKeys,
sign::*,
};
use frost_schnorrkel::Schnorrkel;
use scale::Encode;
use serai_client::{
Public,
validator_sets::primitives::{KeyPair, musig_context, set_keys_message},
};
use serai_db::*;
use crate::tributary::TributarySpec;
create_db!(
SigningProtocolDb {
CachedPreprocesses: (context: &impl Encode) -> [u8; 32]
}
);
struct SigningProtocol<'a, T: DbTxn, C: Encode> {
pub(crate) key: &'a Zeroizing<<Ristretto as Ciphersuite>::F>,
pub(crate) spec: &'a TributarySpec,
pub(crate) txn: &'a mut T,
pub(crate) context: C,
}
impl<T: DbTxn, C: Encode> SigningProtocol<'_, T, C> {
fn preprocess_internal(
&mut self,
participants: &[<Ristretto as Ciphersuite>::G],
) -> (AlgorithmSignMachine<Ristretto, Schnorrkel>, [u8; 64]) {
// Encrypt the cached preprocess as recovery of it will enable recovering the private key
// While the DB isn't expected to be arbitrarily readable, it isn't a proper secret store and
// shouldn't be trusted as one
let mut encryption_key = {
let mut encryption_key_preimage =
Zeroizing::new(b"Cached Preprocess Encryption Key".to_vec());
encryption_key_preimage.extend(self.context.encode());
let repr = Zeroizing::new(self.key.to_repr());
encryption_key_preimage.extend(repr.deref());
Blake2s256::digest(&encryption_key_preimage)
};
let encryption_key_slice: &mut [u8] = encryption_key.as_mut();
let algorithm = Schnorrkel::new(b"substrate");
let keys: ThresholdKeys<Ristretto> =
musig(&musig_context(self.spec.set()), self.key, participants)
.expect("signing for a set we aren't in/validator present multiple times")
.into();
if CachedPreprocesses::get(self.txn, &self.context).is_none() {
let (machine, _) =
AlgorithmMachine::new(algorithm.clone(), keys.clone()).preprocess(&mut OsRng);
let mut cache = machine.cache();
assert_eq!(cache.0.len(), 32);
#[allow(clippy::needless_range_loop)]
for b in 0 .. 32 {
cache.0[b] ^= encryption_key_slice[b];
}
CachedPreprocesses::set(self.txn, &self.context, &cache.0);
}
let cached = CachedPreprocesses::get(self.txn, &self.context).unwrap();
let mut cached: Zeroizing<[u8; 32]> = Zeroizing::new(cached);
#[allow(clippy::needless_range_loop)]
for b in 0 .. 32 {
cached[b] ^= encryption_key_slice[b];
}
encryption_key_slice.zeroize();
let (machine, preprocess) =
AlgorithmSignMachine::from_cache(algorithm, keys, CachedPreprocess(cached));
(machine, preprocess.serialize().try_into().unwrap())
}
fn share_internal(
&mut self,
participants: &[<Ristretto as Ciphersuite>::G],
mut serialized_preprocesses: HashMap<Participant, Vec<u8>>,
msg: &[u8],
) -> Result<(AlgorithmSignatureMachine<Ristretto, Schnorrkel>, [u8; 32]), Participant> {
let machine = self.preprocess_internal(participants).0;
let mut participants = serialized_preprocesses.keys().copied().collect::<Vec<_>>();
participants.sort();
let mut preprocesses = HashMap::new();
for participant in participants {
preprocesses.insert(
participant,
machine
.read_preprocess(&mut serialized_preprocesses.remove(&participant).unwrap().as_slice())
.map_err(|_| participant)?,
);
}
let (machine, share) = machine.sign(preprocesses, msg).map_err(|e| match e {
FrostError::InternalError(e) => unreachable!("FrostError::InternalError {e}"),
FrostError::InvalidParticipant(_, _) |
FrostError::InvalidSigningSet(_) |
FrostError::InvalidParticipantQuantity(_, _) |
FrostError::DuplicatedParticipant(_) |
FrostError::MissingParticipant(_) => unreachable!("{e:?}"),
FrostError::InvalidPreprocess(p) | FrostError::InvalidShare(p) => p,
})?;
Ok((machine, share.serialize().try_into().unwrap()))
}
fn complete_internal(
machine: AlgorithmSignatureMachine<Ristretto, Schnorrkel>,
shares: HashMap<Participant, Vec<u8>>,
) -> Result<[u8; 64], Participant> {
let shares = shares
.into_iter()
.map(|(p, share)| {
machine.read_share(&mut share.as_slice()).map(|share| (p, share)).map_err(|_| p)
})
.collect::<Result<HashMap<_, _>, _>>()?;
let signature = machine.complete(shares).map_err(|e| match e {
FrostError::InternalError(e) => unreachable!("FrostError::InternalError {e}"),
FrostError::InvalidParticipant(_, _) |
FrostError::InvalidSigningSet(_) |
FrostError::InvalidParticipantQuantity(_, _) |
FrostError::DuplicatedParticipant(_) |
FrostError::MissingParticipant(_) => unreachable!("{e:?}"),
FrostError::InvalidPreprocess(p) | FrostError::InvalidShare(p) => p,
})?;
Ok(signature.to_bytes())
}
}
// Get the keys of the participants, noted by their threshold is, and return a new map indexed by
// the MuSig is.
fn threshold_i_map_to_keys_and_musig_i_map(
spec: &TributarySpec,
removed: &[<Ristretto as Ciphersuite>::G],
our_key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
mut map: HashMap<Participant, Vec<u8>>,
) -> (Vec<<Ristretto as Ciphersuite>::G>, HashMap<Participant, Vec<u8>>) {
// Insert our own index so calculations aren't offset
let our_threshold_i = spec
.i(removed, <Ristretto as Ciphersuite>::generator() * our_key.deref())
.expect("MuSig t-of-n signing a for a protocol we were removed from")
.start;
assert!(map.insert(our_threshold_i, vec![]).is_none());
let spec_validators = spec.validators();
let key_from_threshold_i = |threshold_i| {
for (key, _) in &spec_validators {
if threshold_i == spec.i(removed, *key).expect("MuSig t-of-n participant was removed").start {
return *key;
}
}
panic!("requested info for threshold i which doesn't exist")
};
let mut sorted = vec![];
let mut threshold_is = map.keys().copied().collect::<Vec<_>>();
threshold_is.sort();
for threshold_i in threshold_is {
sorted.push((key_from_threshold_i(threshold_i), map.remove(&threshold_i).unwrap()));
}
// Now that signers are sorted, with their shares, create a map with the is needed for MuSig
let mut participants = vec![];
let mut map = HashMap::new();
for (raw_i, (key, share)) in sorted.into_iter().enumerate() {
let musig_i = u16::try_from(raw_i).unwrap() + 1;
participants.push(key);
map.insert(Participant::new(musig_i).unwrap(), share);
}
map.remove(&our_threshold_i).unwrap();
(participants, map)
}
type DkgConfirmerSigningProtocol<'a, T> = SigningProtocol<'a, T, (&'static [u8; 12], u32)>;
pub(crate) struct DkgConfirmer<'a, T: DbTxn> {
key: &'a Zeroizing<<Ristretto as Ciphersuite>::F>,
spec: &'a TributarySpec,
removed: Vec<<Ristretto as Ciphersuite>::G>,
txn: &'a mut T,
attempt: u32,
}
impl<T: DbTxn> DkgConfirmer<'_, T> {
pub(crate) fn new<'a>(
key: &'a Zeroizing<<Ristretto as Ciphersuite>::F>,
spec: &'a TributarySpec,
txn: &'a mut T,
attempt: u32,
) -> Option<DkgConfirmer<'a, T>> {
// This relies on how confirmations are inlined into the DKG protocol and they accordingly
// share attempts
let removed = crate::tributary::removed_as_of_dkg_attempt(txn, spec.genesis(), attempt)?;
Some(DkgConfirmer { key, spec, removed, txn, attempt })
}
fn signing_protocol(&mut self) -> DkgConfirmerSigningProtocol<'_, T> {
let context = (b"DkgConfirmer", self.attempt);
SigningProtocol { key: self.key, spec: self.spec, txn: self.txn, context }
}
fn preprocess_internal(&mut self) -> (AlgorithmSignMachine<Ristretto, Schnorrkel>, [u8; 64]) {
let participants = self.spec.validators().iter().map(|val| val.0).collect::<Vec<_>>();
self.signing_protocol().preprocess_internal(&participants)
}
// Get the preprocess for this confirmation.
pub(crate) fn preprocess(&mut self) -> [u8; 64] {
self.preprocess_internal().1
}
fn share_internal(
&mut self,
preprocesses: HashMap<Participant, Vec<u8>>,
key_pair: &KeyPair,
) -> Result<(AlgorithmSignatureMachine<Ristretto, Schnorrkel>, [u8; 32]), Participant> {
let participants = self.spec.validators().iter().map(|val| val.0).collect::<Vec<_>>();
let preprocesses =
threshold_i_map_to_keys_and_musig_i_map(self.spec, &self.removed, self.key, preprocesses).1;
let msg = set_keys_message(
&self.spec.set(),
&self.removed.iter().map(|key| Public(key.to_bytes())).collect::<Vec<_>>(),
key_pair,
);
self.signing_protocol().share_internal(&participants, preprocesses, &msg)
}
// Get the share for this confirmation, if the preprocesses are valid.
pub(crate) fn share(
&mut self,
preprocesses: HashMap<Participant, Vec<u8>>,
key_pair: &KeyPair,
) -> Result<[u8; 32], Participant> {
self.share_internal(preprocesses, key_pair).map(|(_, share)| share)
}
pub(crate) fn complete(
&mut self,
preprocesses: HashMap<Participant, Vec<u8>>,
key_pair: &KeyPair,
shares: HashMap<Participant, Vec<u8>>,
) -> Result<[u8; 64], Participant> {
let shares =
threshold_i_map_to_keys_and_musig_i_map(self.spec, &self.removed, self.key, shares).1;
let machine = self
.share_internal(preprocesses, key_pair)
.expect("trying to complete a machine which failed to preprocess")
.0;
DkgConfirmerSigningProtocol::<'_, T>::complete_internal(machine, shares)
}
}

View File

@@ -1,156 +0,0 @@
use core::{ops::Range, fmt::Debug};
use std::{io, collections::HashMap};
use transcript::{Transcript, RecommendedTranscript};
use ciphersuite::{group::GroupEncoding, Ciphersuite, Ristretto};
use frost::Participant;
use scale::Encode;
use borsh::{BorshSerialize, BorshDeserialize};
use serai_client::{primitives::PublicKey, validator_sets::primitives::ValidatorSet};
fn borsh_serialize_validators<W: io::Write>(
validators: &Vec<(<Ristretto as Ciphersuite>::G, u16)>,
writer: &mut W,
) -> Result<(), io::Error> {
let len = u16::try_from(validators.len()).unwrap();
BorshSerialize::serialize(&len, writer)?;
for validator in validators {
BorshSerialize::serialize(&validator.0.to_bytes(), writer)?;
BorshSerialize::serialize(&validator.1, writer)?;
}
Ok(())
}
fn borsh_deserialize_validators<R: io::Read>(
reader: &mut R,
) -> Result<Vec<(<Ristretto as Ciphersuite>::G, u16)>, io::Error> {
let len: u16 = BorshDeserialize::deserialize_reader(reader)?;
let mut res = vec![];
for _ in 0 .. len {
let compressed: [u8; 32] = BorshDeserialize::deserialize_reader(reader)?;
let point = Option::from(<Ristretto as Ciphersuite>::G::from_bytes(&compressed))
.ok_or_else(|| io::Error::other("invalid point for validator"))?;
let weight: u16 = BorshDeserialize::deserialize_reader(reader)?;
res.push((point, weight));
}
Ok(res)
}
#[derive(Clone, PartialEq, Eq, Debug, BorshSerialize, BorshDeserialize)]
pub struct TributarySpec {
serai_block: [u8; 32],
start_time: u64,
set: ValidatorSet,
#[borsh(
serialize_with = "borsh_serialize_validators",
deserialize_with = "borsh_deserialize_validators"
)]
validators: Vec<(<Ristretto as Ciphersuite>::G, u16)>,
}
impl TributarySpec {
pub fn new(
serai_block: [u8; 32],
start_time: u64,
set: ValidatorSet,
set_participants: Vec<(PublicKey, u16)>,
) -> TributarySpec {
let mut validators = vec![];
for (participant, shares) in set_participants {
let participant = <Ristretto as Ciphersuite>::read_G::<&[u8]>(&mut participant.0.as_ref())
.expect("invalid key registered as participant");
validators.push((participant, shares));
}
Self { serai_block, start_time, set, validators }
}
pub fn set(&self) -> ValidatorSet {
self.set
}
pub fn genesis(&self) -> [u8; 32] {
// Calculate the genesis for this Tributary
let mut genesis = RecommendedTranscript::new(b"Serai Tributary Genesis");
// This locks it to a specific Serai chain
genesis.append_message(b"serai_block", self.serai_block);
genesis.append_message(b"session", self.set.session.0.to_le_bytes());
genesis.append_message(b"network", self.set.network.encode());
let genesis = genesis.challenge(b"genesis");
let genesis_ref: &[u8] = genesis.as_ref();
genesis_ref[.. 32].try_into().unwrap()
}
pub fn start_time(&self) -> u64 {
self.start_time
}
pub fn n(&self, removed_validators: &[<Ristretto as Ciphersuite>::G]) -> u16 {
self
.validators
.iter()
.map(|(validator, weight)| if removed_validators.contains(validator) { 0 } else { *weight })
.sum()
}
pub fn t(&self) -> u16 {
// t doesn't change with regards to the amount of removed validators
((2 * self.n(&[])) / 3) + 1
}
pub fn i(
&self,
removed_validators: &[<Ristretto as Ciphersuite>::G],
key: <Ristretto as Ciphersuite>::G,
) -> Option<Range<Participant>> {
let mut all_is = HashMap::new();
let mut i = 1;
for (validator, weight) in &self.validators {
all_is.insert(
*validator,
Range { start: Participant::new(i).unwrap(), end: Participant::new(i + weight).unwrap() },
);
i += weight;
}
let original_i = all_is.get(&key)?.clone();
let mut result_i = original_i.clone();
for removed_validator in removed_validators {
let removed_i = all_is
.get(removed_validator)
.expect("removed validator wasn't present in set to begin with");
// If the queried key was removed, return None
if &original_i == removed_i {
return None;
}
// If the removed was before the queried, shift the queried down accordingly
if removed_i.start < original_i.start {
let removed_shares = u16::from(removed_i.end) - u16::from(removed_i.start);
result_i.start = Participant::new(u16::from(original_i.start) - removed_shares).unwrap();
result_i.end = Participant::new(u16::from(original_i.end) - removed_shares).unwrap();
}
}
Some(result_i)
}
pub fn reverse_lookup_i(
&self,
removed_validators: &[<Ristretto as Ciphersuite>::G],
i: Participant,
) -> Option<<Ristretto as Ciphersuite>::G> {
for (validator, _) in &self.validators {
if self.i(removed_validators, *validator).map_or(false, |range| range.contains(&i)) {
return Some(*validator);
}
}
None
}
pub fn validators(&self) -> Vec<(<Ristretto as Ciphersuite>::G, u64)> {
self.validators.iter().map(|(validator, weight)| (*validator, u64::from(*weight))).collect()
}
}

View File

@@ -1,715 +0,0 @@
use core::{ops::Deref, fmt::Debug};
use std::io;
use zeroize::Zeroizing;
use rand_core::{RngCore, CryptoRng};
use blake2::{Digest, Blake2s256};
use transcript::{Transcript, RecommendedTranscript};
use ciphersuite::{
group::{ff::Field, GroupEncoding},
Ciphersuite, Ristretto,
};
use schnorr::SchnorrSignature;
use frost::Participant;
use scale::{Encode, Decode};
use processor_messages::coordinator::SubstrateSignableId;
use tributary::{
TRANSACTION_SIZE_LIMIT, ReadWrite,
transaction::{Signed, TransactionError, TransactionKind, Transaction as TransactionTrait},
};
#[derive(Clone, Copy, PartialEq, Eq, Debug, Encode)]
pub enum Label {
Preprocess,
Share,
}
impl Label {
// TODO: Should nonces be u8 thanks to our use of topics?
pub fn nonce(&self) -> u32 {
match self {
Label::Preprocess => 0,
Label::Share => 1,
}
}
}
#[derive(Clone, PartialEq, Eq)]
pub struct SignData<Id: Clone + PartialEq + Eq + Debug + Encode + Decode> {
pub plan: Id,
pub attempt: u32,
pub label: Label,
pub data: Vec<Vec<u8>>,
pub signed: Signed,
}
impl<Id: Clone + PartialEq + Eq + Debug + Encode + Decode> Debug for SignData<Id> {
fn fmt(&self, fmt: &mut core::fmt::Formatter<'_>) -> Result<(), core::fmt::Error> {
fmt
.debug_struct("SignData")
.field("id", &hex::encode(self.plan.encode()))
.field("attempt", &self.attempt)
.field("label", &self.label)
.field("signer", &hex::encode(self.signed.signer.to_bytes()))
.finish_non_exhaustive()
}
}
impl<Id: Clone + PartialEq + Eq + Debug + Encode + Decode> SignData<Id> {
pub(crate) fn read<R: io::Read>(reader: &mut R) -> io::Result<Self> {
let plan = Id::decode(&mut scale::IoReader(&mut *reader))
.map_err(|_| io::Error::other("invalid plan in SignData"))?;
let mut attempt = [0; 4];
reader.read_exact(&mut attempt)?;
let attempt = u32::from_le_bytes(attempt);
let mut label = [0; 1];
reader.read_exact(&mut label)?;
let label = match label[0] {
0 => Label::Preprocess,
1 => Label::Share,
_ => Err(io::Error::other("invalid label in SignData"))?,
};
let data = {
let mut data_pieces = [0];
reader.read_exact(&mut data_pieces)?;
if data_pieces[0] == 0 {
Err(io::Error::other("zero pieces of data in SignData"))?;
}
let mut all_data = vec![];
for _ in 0 .. data_pieces[0] {
let mut data_len = [0; 2];
reader.read_exact(&mut data_len)?;
let mut data = vec![0; usize::from(u16::from_le_bytes(data_len))];
reader.read_exact(&mut data)?;
all_data.push(data);
}
all_data
};
let signed = Signed::read_without_nonce(reader, label.nonce())?;
Ok(SignData { plan, attempt, label, data, signed })
}
pub(crate) fn write<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
writer.write_all(&self.plan.encode())?;
writer.write_all(&self.attempt.to_le_bytes())?;
writer.write_all(&[match self.label {
Label::Preprocess => 0,
Label::Share => 1,
}])?;
writer.write_all(&[u8::try_from(self.data.len()).unwrap()])?;
for data in &self.data {
if data.len() > u16::MAX.into() {
// Currently, the largest individual preprocess is a Monero transaction
// It provides 4 commitments per input (128 bytes), a 64-byte proof for them, along with a
// key image and proof (96 bytes)
// Even with all of that, we could support 227 inputs in a single TX
// Monero is limited to ~120 inputs per TX
//
// Bitcoin has a much higher input count of 520, yet it only uses 64 bytes per preprocess
Err(io::Error::other("signing data exceeded 65535 bytes"))?;
}
writer.write_all(&u16::try_from(data.len()).unwrap().to_le_bytes())?;
writer.write_all(data)?;
}
self.signed.write_without_nonce(writer)
}
}
#[derive(Clone, PartialEq, Eq)]
pub enum Transaction {
RemoveParticipantDueToDkg {
participant: <Ristretto as Ciphersuite>::G,
signed: Signed,
},
DkgCommitments {
attempt: u32,
commitments: Vec<Vec<u8>>,
signed: Signed,
},
DkgShares {
attempt: u32,
// Sending Participant, Receiving Participant, Share
shares: Vec<Vec<Vec<u8>>>,
confirmation_nonces: [u8; 64],
signed: Signed,
},
InvalidDkgShare {
attempt: u32,
accuser: Participant,
faulty: Participant,
blame: Option<Vec<u8>>,
signed: Signed,
},
DkgConfirmed {
attempt: u32,
confirmation_share: [u8; 32],
signed: Signed,
},
// Co-sign a Substrate block.
CosignSubstrateBlock([u8; 32]),
// When we have synchrony on a batch, we can allow signing it
// TODO (never?): This is less efficient compared to an ExternalBlock provided transaction,
// which would be binding over the block hash and automatically achieve synchrony on all
// relevant batches. ExternalBlock was removed for this due to complexity around the pipeline
// with the current processor, yet it would still be an improvement.
Batch {
block: [u8; 32],
batch: u32,
},
// When a Serai block is finalized, with the contained batches, we can allow the associated plan
// IDs
SubstrateBlock(u64),
SubstrateSign(SignData<SubstrateSignableId>),
Sign(SignData<[u8; 32]>),
// This is defined as an Unsigned transaction in order to de-duplicate SignCompleted amongst
// reporters (who should all report the same thing)
// We do still track the signer in order to prevent a single signer from publishing arbitrarily
// many TXs without penalty
// Here, they're denoted as the first_signer, as only the signer of the first TX to be included
// with this pairing will be remembered on-chain
SignCompleted {
plan: [u8; 32],
tx_hash: Vec<u8>,
first_signer: <Ristretto as Ciphersuite>::G,
signature: SchnorrSignature<Ristretto>,
},
SlashReport(Vec<u32>, Signed),
}
impl Debug for Transaction {
fn fmt(&self, fmt: &mut core::fmt::Formatter<'_>) -> Result<(), core::fmt::Error> {
match self {
Transaction::RemoveParticipantDueToDkg { participant, signed } => fmt
.debug_struct("Transaction::RemoveParticipantDueToDkg")
.field("participant", &hex::encode(participant.to_bytes()))
.field("signer", &hex::encode(signed.signer.to_bytes()))
.finish_non_exhaustive(),
Transaction::DkgCommitments { attempt, commitments: _, signed } => fmt
.debug_struct("Transaction::DkgCommitments")
.field("attempt", attempt)
.field("signer", &hex::encode(signed.signer.to_bytes()))
.finish_non_exhaustive(),
Transaction::DkgShares { attempt, signed, .. } => fmt
.debug_struct("Transaction::DkgShares")
.field("attempt", attempt)
.field("signer", &hex::encode(signed.signer.to_bytes()))
.finish_non_exhaustive(),
Transaction::InvalidDkgShare { attempt, accuser, faulty, .. } => fmt
.debug_struct("Transaction::InvalidDkgShare")
.field("attempt", attempt)
.field("accuser", accuser)
.field("faulty", faulty)
.finish_non_exhaustive(),
Transaction::DkgConfirmed { attempt, confirmation_share: _, signed } => fmt
.debug_struct("Transaction::DkgConfirmed")
.field("attempt", attempt)
.field("signer", &hex::encode(signed.signer.to_bytes()))
.finish_non_exhaustive(),
Transaction::CosignSubstrateBlock(block) => fmt
.debug_struct("Transaction::CosignSubstrateBlock")
.field("block", &hex::encode(block))
.finish(),
Transaction::Batch { block, batch } => fmt
.debug_struct("Transaction::Batch")
.field("block", &hex::encode(block))
.field("batch", &batch)
.finish(),
Transaction::SubstrateBlock(block) => {
fmt.debug_struct("Transaction::SubstrateBlock").field("block", block).finish()
}
Transaction::SubstrateSign(sign_data) => {
fmt.debug_struct("Transaction::SubstrateSign").field("sign_data", sign_data).finish()
}
Transaction::Sign(sign_data) => {
fmt.debug_struct("Transaction::Sign").field("sign_data", sign_data).finish()
}
Transaction::SignCompleted { plan, tx_hash, .. } => fmt
.debug_struct("Transaction::SignCompleted")
.field("plan", &hex::encode(plan))
.field("tx_hash", &hex::encode(tx_hash))
.finish_non_exhaustive(),
Transaction::SlashReport(points, signed) => fmt
.debug_struct("Transaction::SignCompleted")
.field("points", points)
.field("signed", signed)
.finish(),
}
}
}
impl ReadWrite for Transaction {
fn read<R: io::Read>(reader: &mut R) -> io::Result<Self> {
let mut kind = [0];
reader.read_exact(&mut kind)?;
match kind[0] {
0 => Ok(Transaction::RemoveParticipantDueToDkg {
participant: Ristretto::read_G(reader)?,
signed: Signed::read_without_nonce(reader, 0)?,
}),
1 => {
let mut attempt = [0; 4];
reader.read_exact(&mut attempt)?;
let attempt = u32::from_le_bytes(attempt);
let commitments = {
let mut commitments_len = [0; 1];
reader.read_exact(&mut commitments_len)?;
let commitments_len = usize::from(commitments_len[0]);
if commitments_len == 0 {
Err(io::Error::other("zero commitments in DkgCommitments"))?;
}
let mut each_commitments_len = [0; 2];
reader.read_exact(&mut each_commitments_len)?;
let each_commitments_len = usize::from(u16::from_le_bytes(each_commitments_len));
if (commitments_len * each_commitments_len) > TRANSACTION_SIZE_LIMIT {
Err(io::Error::other(
"commitments present in transaction exceeded transaction size limit",
))?;
}
let mut commitments = vec![vec![]; commitments_len];
for commitments in &mut commitments {
*commitments = vec![0; each_commitments_len];
reader.read_exact(commitments)?;
}
commitments
};
let signed = Signed::read_without_nonce(reader, 0)?;
Ok(Transaction::DkgCommitments { attempt, commitments, signed })
}
2 => {
let mut attempt = [0; 4];
reader.read_exact(&mut attempt)?;
let attempt = u32::from_le_bytes(attempt);
let shares = {
let mut share_quantity = [0; 1];
reader.read_exact(&mut share_quantity)?;
let mut key_share_quantity = [0; 1];
reader.read_exact(&mut key_share_quantity)?;
let mut share_len = [0; 2];
reader.read_exact(&mut share_len)?;
let share_len = usize::from(u16::from_le_bytes(share_len));
let mut all_shares = vec![];
for _ in 0 .. share_quantity[0] {
let mut shares = vec![];
for _ in 0 .. key_share_quantity[0] {
let mut share = vec![0; share_len];
reader.read_exact(&mut share)?;
shares.push(share);
}
all_shares.push(shares);
}
all_shares
};
let mut confirmation_nonces = [0; 64];
reader.read_exact(&mut confirmation_nonces)?;
let signed = Signed::read_without_nonce(reader, 1)?;
Ok(Transaction::DkgShares { attempt, shares, confirmation_nonces, signed })
}
3 => {
let mut attempt = [0; 4];
reader.read_exact(&mut attempt)?;
let attempt = u32::from_le_bytes(attempt);
let mut accuser = [0; 2];
reader.read_exact(&mut accuser)?;
let accuser = Participant::new(u16::from_le_bytes(accuser))
.ok_or_else(|| io::Error::other("invalid participant in InvalidDkgShare"))?;
let mut faulty = [0; 2];
reader.read_exact(&mut faulty)?;
let faulty = Participant::new(u16::from_le_bytes(faulty))
.ok_or_else(|| io::Error::other("invalid participant in InvalidDkgShare"))?;
let mut blame_len = [0; 2];
reader.read_exact(&mut blame_len)?;
let mut blame = vec![0; u16::from_le_bytes(blame_len).into()];
reader.read_exact(&mut blame)?;
// This shares a nonce with DkgConfirmed as only one is expected
let signed = Signed::read_without_nonce(reader, 2)?;
Ok(Transaction::InvalidDkgShare {
attempt,
accuser,
faulty,
blame: Some(blame).filter(|blame| !blame.is_empty()),
signed,
})
}
4 => {
let mut attempt = [0; 4];
reader.read_exact(&mut attempt)?;
let attempt = u32::from_le_bytes(attempt);
let mut confirmation_share = [0; 32];
reader.read_exact(&mut confirmation_share)?;
let signed = Signed::read_without_nonce(reader, 2)?;
Ok(Transaction::DkgConfirmed { attempt, confirmation_share, signed })
}
5 => {
let mut block = [0; 32];
reader.read_exact(&mut block)?;
Ok(Transaction::CosignSubstrateBlock(block))
}
6 => {
let mut block = [0; 32];
reader.read_exact(&mut block)?;
let mut batch = [0; 4];
reader.read_exact(&mut batch)?;
Ok(Transaction::Batch { block, batch: u32::from_le_bytes(batch) })
}
7 => {
let mut block = [0; 8];
reader.read_exact(&mut block)?;
Ok(Transaction::SubstrateBlock(u64::from_le_bytes(block)))
}
8 => SignData::read(reader).map(Transaction::SubstrateSign),
9 => SignData::read(reader).map(Transaction::Sign),
10 => {
let mut plan = [0; 32];
reader.read_exact(&mut plan)?;
let mut tx_hash_len = [0];
reader.read_exact(&mut tx_hash_len)?;
let mut tx_hash = vec![0; usize::from(tx_hash_len[0])];
reader.read_exact(&mut tx_hash)?;
let first_signer = Ristretto::read_G(reader)?;
let signature = SchnorrSignature::<Ristretto>::read(reader)?;
Ok(Transaction::SignCompleted { plan, tx_hash, first_signer, signature })
}
11 => {
let mut len = [0];
reader.read_exact(&mut len)?;
let len = len[0];
// If the set has as many validators as MAX_KEY_SHARES_PER_SET, then the amount of distinct
// validators (the amount of validators reported on) will be at most
// `MAX_KEY_SHARES_PER_SET - 1`
if u32::from(len) > (serai_client::validator_sets::primitives::MAX_KEY_SHARES_PER_SET - 1) {
Err(io::Error::other("more points reported than allowed validator"))?;
}
let mut points = vec![0u32; len.into()];
for points in &mut points {
let mut these_points = [0; 4];
reader.read_exact(&mut these_points)?;
*points = u32::from_le_bytes(these_points);
}
Ok(Transaction::SlashReport(points, Signed::read_without_nonce(reader, 0)?))
}
_ => Err(io::Error::other("invalid transaction type")),
}
}
fn write<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
match self {
Transaction::RemoveParticipantDueToDkg { participant, signed } => {
writer.write_all(&[0])?;
writer.write_all(&participant.to_bytes())?;
signed.write_without_nonce(writer)
}
Transaction::DkgCommitments { attempt, commitments, signed } => {
writer.write_all(&[1])?;
writer.write_all(&attempt.to_le_bytes())?;
if commitments.is_empty() {
Err(io::Error::other("zero commitments in DkgCommitments"))?
}
writer.write_all(&[u8::try_from(commitments.len()).unwrap()])?;
for commitments_i in commitments {
if commitments_i.len() != commitments[0].len() {
Err(io::Error::other("commitments of differing sizes in DkgCommitments"))?
}
}
writer.write_all(&u16::try_from(commitments[0].len()).unwrap().to_le_bytes())?;
for commitments in commitments {
writer.write_all(commitments)?;
}
signed.write_without_nonce(writer)
}
Transaction::DkgShares { attempt, shares, confirmation_nonces, signed } => {
writer.write_all(&[2])?;
writer.write_all(&attempt.to_le_bytes())?;
// `shares` is a Vec which is supposed to map to a HashMap<Participant, Vec<u8>>. Since we
// bound participants to 150, this conversion is safe if a valid in-memory transaction.
writer.write_all(&[u8::try_from(shares.len()).unwrap()])?;
// This assumes at least one share is being sent to another party
writer.write_all(&[u8::try_from(shares[0].len()).unwrap()])?;
let share_len = shares[0][0].len();
// For BLS12-381 G2, this would be:
// - A 32-byte share
// - A 96-byte ephemeral key
// - A 128-byte signature
// Hence why this has to be u16
writer.write_all(&u16::try_from(share_len).unwrap().to_le_bytes())?;
for these_shares in shares {
assert_eq!(these_shares.len(), shares[0].len(), "amount of sent shares was variable");
for share in these_shares {
assert_eq!(share.len(), share_len, "sent shares were of variable length");
writer.write_all(share)?;
}
}
writer.write_all(confirmation_nonces)?;
signed.write_without_nonce(writer)
}
Transaction::InvalidDkgShare { attempt, accuser, faulty, blame, signed } => {
writer.write_all(&[3])?;
writer.write_all(&attempt.to_le_bytes())?;
writer.write_all(&u16::from(*accuser).to_le_bytes())?;
writer.write_all(&u16::from(*faulty).to_le_bytes())?;
// Flattens Some(vec![]) to None on the expectation no actual blame will be 0-length
assert!(blame.as_ref().map_or(1, Vec::len) != 0);
let blame_len =
u16::try_from(blame.as_ref().unwrap_or(&vec![]).len()).expect("blame exceeded 64 KB");
writer.write_all(&blame_len.to_le_bytes())?;
writer.write_all(blame.as_ref().unwrap_or(&vec![]))?;
signed.write_without_nonce(writer)
}
Transaction::DkgConfirmed { attempt, confirmation_share, signed } => {
writer.write_all(&[4])?;
writer.write_all(&attempt.to_le_bytes())?;
writer.write_all(confirmation_share)?;
signed.write_without_nonce(writer)
}
Transaction::CosignSubstrateBlock(block) => {
writer.write_all(&[5])?;
writer.write_all(block)
}
Transaction::Batch { block, batch } => {
writer.write_all(&[6])?;
writer.write_all(block)?;
writer.write_all(&batch.to_le_bytes())
}
Transaction::SubstrateBlock(block) => {
writer.write_all(&[7])?;
writer.write_all(&block.to_le_bytes())
}
Transaction::SubstrateSign(data) => {
writer.write_all(&[8])?;
data.write(writer)
}
Transaction::Sign(data) => {
writer.write_all(&[9])?;
data.write(writer)
}
Transaction::SignCompleted { plan, tx_hash, first_signer, signature } => {
writer.write_all(&[10])?;
writer.write_all(plan)?;
writer
.write_all(&[u8::try_from(tx_hash.len()).expect("tx hash length exceed 255 bytes")])?;
writer.write_all(tx_hash)?;
writer.write_all(&first_signer.to_bytes())?;
signature.write(writer)
}
Transaction::SlashReport(points, signed) => {
writer.write_all(&[11])?;
writer.write_all(&[u8::try_from(points.len()).unwrap()])?;
for points in points {
writer.write_all(&points.to_le_bytes())?;
}
signed.write_without_nonce(writer)
}
}
}
}
impl TransactionTrait for Transaction {
fn kind(&self) -> TransactionKind<'_> {
match self {
Transaction::RemoveParticipantDueToDkg { participant, signed } => {
TransactionKind::Signed((b"remove", participant.to_bytes()).encode(), signed)
}
Transaction::DkgCommitments { attempt, commitments: _, signed } |
Transaction::DkgShares { attempt, signed, .. } |
Transaction::InvalidDkgShare { attempt, signed, .. } |
Transaction::DkgConfirmed { attempt, signed, .. } => {
TransactionKind::Signed((b"dkg", attempt).encode(), signed)
}
Transaction::CosignSubstrateBlock(_) => TransactionKind::Provided("cosign"),
Transaction::Batch { .. } => TransactionKind::Provided("batch"),
Transaction::SubstrateBlock(_) => TransactionKind::Provided("serai"),
Transaction::SubstrateSign(data) => {
TransactionKind::Signed((b"substrate", data.plan, data.attempt).encode(), &data.signed)
}
Transaction::Sign(data) => {
TransactionKind::Signed((b"sign", data.plan, data.attempt).encode(), &data.signed)
}
Transaction::SignCompleted { .. } => TransactionKind::Unsigned,
Transaction::SlashReport(_, signed) => {
TransactionKind::Signed(b"slash_report".to_vec(), signed)
}
}
}
fn hash(&self) -> [u8; 32] {
let mut tx = self.serialize();
if let TransactionKind::Signed(_, signed) = self.kind() {
// Make sure the part we're cutting off is the signature
assert_eq!(tx.drain((tx.len() - 64) ..).collect::<Vec<_>>(), signed.signature.serialize());
}
Blake2s256::digest([b"Coordinator Tributary Transaction".as_slice(), &tx].concat()).into()
}
fn verify(&self) -> Result<(), TransactionError> {
// TODO: Check SubstrateSign's lengths here
if let Transaction::SignCompleted { first_signer, signature, .. } = self {
if !signature.verify(*first_signer, self.sign_completed_challenge()) {
Err(TransactionError::InvalidContent)?;
}
}
Ok(())
}
}
impl Transaction {
// Used to initially construct transactions so we can then get sig hashes and perform signing
pub fn empty_signed() -> Signed {
Signed {
signer: Ristretto::generator(),
nonce: 0,
signature: SchnorrSignature::<Ristretto> {
R: Ristretto::generator(),
s: <Ristretto as Ciphersuite>::F::ZERO,
},
}
}
// Sign a transaction
pub fn sign<R: RngCore + CryptoRng>(
&mut self,
rng: &mut R,
genesis: [u8; 32],
key: &Zeroizing<<Ristretto as Ciphersuite>::F>,
) {
fn signed(tx: &mut Transaction) -> (u32, &mut Signed) {
#[allow(clippy::match_same_arms)] // Doesn't make semantic sense here
let nonce = match tx {
Transaction::RemoveParticipantDueToDkg { .. } => 0,
Transaction::DkgCommitments { .. } => 0,
Transaction::DkgShares { .. } => 1,
Transaction::InvalidDkgShare { .. } | Transaction::DkgConfirmed { .. } => 2,
Transaction::CosignSubstrateBlock(_) => panic!("signing CosignSubstrateBlock"),
Transaction::Batch { .. } => panic!("signing Batch"),
Transaction::SubstrateBlock(_) => panic!("signing SubstrateBlock"),
Transaction::SubstrateSign(data) => data.label.nonce(),
Transaction::Sign(data) => data.label.nonce(),
Transaction::SignCompleted { .. } => panic!("signing SignCompleted"),
Transaction::SlashReport(_, _) => 0,
};
(
nonce,
#[allow(clippy::match_same_arms)]
match tx {
Transaction::RemoveParticipantDueToDkg { ref mut signed, .. } |
Transaction::DkgCommitments { ref mut signed, .. } |
Transaction::DkgShares { ref mut signed, .. } |
Transaction::InvalidDkgShare { ref mut signed, .. } |
Transaction::DkgConfirmed { ref mut signed, .. } => signed,
Transaction::CosignSubstrateBlock(_) => panic!("signing CosignSubstrateBlock"),
Transaction::Batch { .. } => panic!("signing Batch"),
Transaction::SubstrateBlock(_) => panic!("signing SubstrateBlock"),
Transaction::SubstrateSign(ref mut data) => &mut data.signed,
Transaction::Sign(ref mut data) => &mut data.signed,
Transaction::SignCompleted { .. } => panic!("signing SignCompleted"),
Transaction::SlashReport(_, ref mut signed) => signed,
},
)
}
let (nonce, signed_ref) = signed(self);
signed_ref.signer = Ristretto::generator() * key.deref();
signed_ref.nonce = nonce;
let sig_nonce = Zeroizing::new(<Ristretto as Ciphersuite>::F::random(rng));
signed(self).1.signature.R = <Ristretto as Ciphersuite>::generator() * sig_nonce.deref();
let sig_hash = self.sig_hash(genesis);
signed(self).1.signature = SchnorrSignature::<Ristretto>::sign(key, sig_nonce, sig_hash);
}
pub fn sign_completed_challenge(&self) -> <Ristretto as Ciphersuite>::F {
if let Transaction::SignCompleted { plan, tx_hash, first_signer, signature } = self {
let mut transcript =
RecommendedTranscript::new(b"Coordinator Tributary Transaction SignCompleted");
transcript.append_message(b"plan", plan);
transcript.append_message(b"tx_hash", tx_hash);
transcript.append_message(b"signer", first_signer.to_bytes());
transcript.append_message(b"nonce", signature.R.to_bytes());
Ristretto::hash_to_F(b"SignCompleted signature", &transcript.challenge(b"challenge"))
} else {
panic!("sign_completed_challenge called on transaction which wasn't SignCompleted")
}
}
}

View File

@@ -0,0 +1,37 @@
[package]
name = "serai-coordinator-substrate"
version = "0.1.0"
description = "Serai Coordinator's Substrate Scanner"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/coordinator/substrate"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
keywords = []
edition = "2021"
publish = false
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
bitvec = { version = "1", default-features = false, features = ["std"] }
scale = { package = "parity-scale-codec", version = "3", default-features = false, features = ["std", "derive", "bit-vec"] }
borsh = { version = "1", default-features = false, features = ["std", "derive", "de_strict_order"] }
serai-client = { path = "../../substrate/client", version = "0.1", default-features = false, features = ["serai", "borsh"] }
log = { version = "0.4", default-features = false, features = ["std"] }
futures = { version = "0.3", default-features = false, features = ["std"] }
tokio = { version = "1", default-features = false }
serai-db = { path = "../../common/db", version = "0.1.1" }
serai-task = { path = "../../common/task", version = "0.1" }
serai-cosign = { path = "../cosign", version = "0.1" }
messages = { package = "serai-processor-messages", version = "0.1", path = "../../processor/messages" }

View File

@@ -0,0 +1,15 @@
AGPL-3.0-only license
Copyright (c) 2023-2024 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.

View File

@@ -0,0 +1,20 @@
# Serai Coordinator Substrate
This crate manages the Serai coordinators's interactions with Serai's Substrate blockchain.
Two event streams are defined:
- Canonical events, which must be handled by every validator, regardless of the sets they're present
in. These are represented by `serai_processor_messages::substrate::CoordinatorMessage`.
- Ephemeral events, which only need to be handled by the validators present within the sets they
relate to. These are represented by two channels, `NewSet` and `SignSlashReport`.
The canonical event stream is available without provision of a validator's public key. The ephemeral
event stream requires provision of a validator's public key. Both are ordered within themselves, yet
there are no ordering guarantees across the two.
Additionally, a collection of tasks are defined to publish data onto Serai:
- `SetKeysTask`, which sets the keys generated via DKGs onto Serai.
- `PublishBatchTask`, which publishes `Batch`s onto Serai.
- `PublishSlashReportTask`, which publishes `SlashReport`s onto Serai.

View File

@@ -0,0 +1,224 @@
use core::future::Future;
use std::sync::Arc;
use futures::stream::{StreamExt, FuturesOrdered};
use serai_client::Serai;
use messages::substrate::{InInstructionResult, ExecutedBatch, CoordinatorMessage};
use serai_db::*;
use serai_task::ContinuallyRan;
use serai_cosign::Cosigning;
create_db!(
CoordinatorSubstrateCanonical {
NextBlock: () -> u64,
}
);
/// The event stream for canonical events.
pub struct CanonicalEventStream<D: Db> {
db: D,
serai: Arc<Serai>,
}
impl<D: Db> CanonicalEventStream<D> {
/// Create a new canonical event stream.
///
/// Only one of these may exist over the provided database.
pub fn new(db: D, serai: Arc<Serai>) -> Self {
Self { db, serai }
}
}
impl<D: Db> ContinuallyRan for CanonicalEventStream<D> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let next_block = NextBlock::get(&self.db).unwrap_or(0);
let latest_finalized_block =
Cosigning::<D>::latest_cosigned_block_number(&self.db).map_err(|e| format!("{e:?}"))?;
// These are all the events which generate canonical messages
struct CanonicalEvents {
time: u64,
key_gen_events: Vec<serai_client::validator_sets::ValidatorSetsEvent>,
set_retired_events: Vec<serai_client::validator_sets::ValidatorSetsEvent>,
batch_events: Vec<serai_client::in_instructions::InInstructionsEvent>,
burn_events: Vec<serai_client::coins::CoinsEvent>,
}
// For a cosigned block, fetch all relevant events
let scan = {
let db = self.db.clone();
let serai = &self.serai;
move |block_number| {
let block_hash = Cosigning::<D>::cosigned_block(&db, block_number);
async move {
let block_hash = match block_hash {
Ok(Some(block_hash)) => block_hash,
Ok(None) => {
panic!("iterating to latest cosigned block but couldn't get cosigned block")
}
Err(serai_cosign::Faulted) => return Err("cosigning process faulted".to_string()),
};
let temporal_serai = serai.as_of(block_hash);
let temporal_serai_validators = temporal_serai.validator_sets();
let temporal_serai_instructions = temporal_serai.in_instructions();
let temporal_serai_coins = temporal_serai.coins();
let (block, key_gen_events, set_retired_events, batch_events, burn_events) =
tokio::try_join!(
serai.block(block_hash),
temporal_serai_validators.key_gen_events(),
temporal_serai_validators.set_retired_events(),
temporal_serai_instructions.batch_events(),
temporal_serai_coins.burn_with_instruction_events(),
)
.map_err(|e| format!("{e:?}"))?;
let Some(block) = block else {
Err(format!("Serai node didn't have cosigned block #{block_number}"))?
};
let time = if block_number == 0 {
block.time().unwrap_or(0)
} else {
// Serai's block time is in milliseconds
block
.time()
.ok_or_else(|| "non-genesis Serai block didn't have a time".to_string())? /
1000
};
Ok((
block_number,
CanonicalEvents {
time,
key_gen_events,
set_retired_events,
batch_events,
burn_events,
},
))
}
}
};
// Sync the next set of upcoming blocks all at once to minimize latency
const BLOCKS_TO_SYNC_AT_ONCE: u64 = 10;
// FuturesOrdered can be bad practice due to potentially causing tiemouts if it isn't
// sufficiently polled. Considering our processing loop is minimal and it does poll this,
// it's fine.
let mut set = FuturesOrdered::new();
for block_number in
next_block ..= latest_finalized_block.min(next_block + BLOCKS_TO_SYNC_AT_ONCE)
{
set.push_back(scan(block_number));
}
for block_number in next_block ..= latest_finalized_block {
// Get the next block in our queue
let (popped_block_number, block) = set.next().await.unwrap()?;
assert_eq!(block_number, popped_block_number);
// Re-populate the queue
if (block_number + BLOCKS_TO_SYNC_AT_ONCE) <= latest_finalized_block {
set.push_back(scan(block_number + BLOCKS_TO_SYNC_AT_ONCE));
}
let mut txn = self.db.txn();
for key_gen in block.key_gen_events {
let serai_client::validator_sets::ValidatorSetsEvent::KeyGen { set, key_pair } = &key_gen
else {
panic!("KeyGen event wasn't a KeyGen event: {key_gen:?}");
};
crate::Canonical::send(
&mut txn,
set.network,
&CoordinatorMessage::SetKeys {
serai_time: block.time,
session: set.session,
key_pair: key_pair.clone(),
},
);
}
for set_retired in block.set_retired_events {
let serai_client::validator_sets::ValidatorSetsEvent::SetRetired { set } = &set_retired
else {
panic!("SetRetired event wasn't a SetRetired event: {set_retired:?}");
};
crate::Canonical::send(
&mut txn,
set.network,
&CoordinatorMessage::SlashesReported { session: set.session },
);
}
for network in serai_client::primitives::NETWORKS {
let mut batch = None;
for this_batch in &block.batch_events {
let serai_client::in_instructions::InInstructionsEvent::Batch {
network: batch_network,
publishing_session,
id,
external_network_block_hash,
in_instructions_hash,
in_instruction_results,
} = this_batch
else {
panic!("Batch event wasn't a Batch event: {this_batch:?}");
};
if network == *batch_network {
if batch.is_some() {
Err("Serai block had multiple batches for the same network".to_string())?;
}
batch = Some(ExecutedBatch {
id: *id,
publisher: *publishing_session,
external_network_block_hash: *external_network_block_hash,
in_instructions_hash: *in_instructions_hash,
in_instruction_results: in_instruction_results
.iter()
.map(|bit| {
if *bit {
InInstructionResult::Succeeded
} else {
InInstructionResult::Failed
}
})
.collect(),
});
}
}
let mut burns = vec![];
for burn in &block.burn_events {
let serai_client::coins::CoinsEvent::BurnWithInstruction { from: _, instruction } =
&burn
else {
panic!("Burn event wasn't a Burn.in event: {burn:?}");
};
if instruction.balance.coin.network() == network {
burns.push(instruction.clone());
}
}
crate::Canonical::send(
&mut txn,
network,
&CoordinatorMessage::Block { serai_block_number: block_number, batch, burns },
);
}
txn.commit();
}
Ok(next_block <= latest_finalized_block)
}
}
}

View File

@@ -0,0 +1,249 @@
use core::future::Future;
use std::sync::Arc;
use futures::stream::{StreamExt, FuturesOrdered};
use serai_client::{
primitives::{PublicKey, NetworkId, EmbeddedEllipticCurve},
validator_sets::primitives::MAX_KEY_SHARES_PER_SET,
Serai,
};
use serai_db::*;
use serai_task::ContinuallyRan;
use serai_cosign::Cosigning;
use crate::NewSetInformation;
create_db!(
CoordinatorSubstrateEphemeral {
NextBlock: () -> u64,
}
);
/// The event stream for ephemeral events.
pub struct EphemeralEventStream<D: Db> {
db: D,
serai: Arc<Serai>,
validator: PublicKey,
}
impl<D: Db> EphemeralEventStream<D> {
/// Create a new ephemeral event stream.
///
/// Only one of these may exist over the provided database.
pub fn new(db: D, serai: Arc<Serai>, validator: PublicKey) -> Self {
Self { db, serai, validator }
}
}
impl<D: Db> ContinuallyRan for EphemeralEventStream<D> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let next_block = NextBlock::get(&self.db).unwrap_or(0);
let latest_finalized_block =
Cosigning::<D>::latest_cosigned_block_number(&self.db).map_err(|e| format!("{e:?}"))?;
// These are all the events which generate canonical messages
struct EphemeralEvents {
block_hash: [u8; 32],
time: u64,
new_set_events: Vec<serai_client::validator_sets::ValidatorSetsEvent>,
accepted_handover_events: Vec<serai_client::validator_sets::ValidatorSetsEvent>,
}
// For a cosigned block, fetch all relevant events
let scan = {
let db = self.db.clone();
let serai = &self.serai;
move |block_number| {
let block_hash = Cosigning::<D>::cosigned_block(&db, block_number);
async move {
let block_hash = match block_hash {
Ok(Some(block_hash)) => block_hash,
Ok(None) => {
panic!("iterating to latest cosigned block but couldn't get cosigned block")
}
Err(serai_cosign::Faulted) => return Err("cosigning process faulted".to_string()),
};
let temporal_serai = serai.as_of(block_hash);
let temporal_serai_validators = temporal_serai.validator_sets();
let (block, new_set_events, accepted_handover_events) = tokio::try_join!(
serai.block(block_hash),
temporal_serai_validators.new_set_events(),
temporal_serai_validators.accepted_handover_events(),
)
.map_err(|e| format!("{e:?}"))?;
let Some(block) = block else {
Err(format!("Serai node didn't have cosigned block #{block_number}"))?
};
let time = if block_number == 0 {
block.time().unwrap_or(0)
} else {
// Serai's block time is in milliseconds
block
.time()
.ok_or_else(|| "non-genesis Serai block didn't have a time".to_string())? /
1000
};
Ok((
block_number,
EphemeralEvents { block_hash, time, new_set_events, accepted_handover_events },
))
}
}
};
// Sync the next set of upcoming blocks all at once to minimize latency
const BLOCKS_TO_SYNC_AT_ONCE: u64 = 50;
// FuturesOrdered can be bad practice due to potentially causing tiemouts if it isn't
// sufficiently polled. Our processing loop isn't minimal, itself making multiple requests,
// but the loop body should only be executed a few times a week. It's better to get through
// most blocks with this optimization, and have timeouts a few times a week, than not have
// this at all.
let mut set = FuturesOrdered::new();
for block_number in
next_block ..= latest_finalized_block.min(next_block + BLOCKS_TO_SYNC_AT_ONCE)
{
set.push_back(scan(block_number));
}
for block_number in next_block ..= latest_finalized_block {
// Get the next block in our queue
let (popped_block_number, block) = set.next().await.unwrap()?;
assert_eq!(block_number, popped_block_number);
// Re-populate the queue
if (block_number + BLOCKS_TO_SYNC_AT_ONCE) <= latest_finalized_block {
set.push_back(scan(block_number + BLOCKS_TO_SYNC_AT_ONCE));
}
let mut txn = self.db.txn();
for new_set in block.new_set_events {
let serai_client::validator_sets::ValidatorSetsEvent::NewSet { set } = &new_set else {
panic!("NewSet event wasn't a NewSet event: {new_set:?}");
};
// We only coordinate over external networks
if set.network == NetworkId::Serai {
continue;
}
let serai = self.serai.as_of(block.block_hash);
let serai = serai.validator_sets();
let Some(validators) =
serai.participants(set.network).await.map_err(|e| format!("{e:?}"))?
else {
Err(format!(
"block #{block_number} declared a new set but didn't have the participants"
))?
};
let in_set = validators.iter().any(|(validator, _)| *validator == self.validator);
if in_set {
if u16::try_from(validators.len()).is_err() {
Err("more than u16::MAX validators sent")?;
}
let Ok(validators) = validators
.into_iter()
.map(|(validator, weight)| u16::try_from(weight).map(|weight| (validator, weight)))
.collect::<Result<Vec<_>, _>>()
else {
Err("validator's weight exceeded u16::MAX".to_string())?
};
// Do the summation in u32 so we don't risk a u16 overflow
let total_weight = validators.iter().map(|(_, weight)| u32::from(*weight)).sum::<u32>();
if total_weight > u32::from(MAX_KEY_SHARES_PER_SET) {
Err(format!(
"{set:?} has {total_weight} key shares when the max is {MAX_KEY_SHARES_PER_SET}"
))?;
}
let total_weight = u16::try_from(total_weight).unwrap();
// Fetch all of the validators' embedded elliptic curve keys
let mut embedded_elliptic_curve_keys = FuturesOrdered::new();
for (validator, _) in &validators {
let validator = *validator;
// try_join doesn't return a future so we need to wrap it in this additional async
// block
embedded_elliptic_curve_keys.push_back(async move {
tokio::try_join!(
// One future to fetch the substrate embedded key
serai
.embedded_elliptic_curve_key(validator, EmbeddedEllipticCurve::Embedwards25519),
// One future to fetch the external embedded key, if there is a distinct curve
async {
// `embedded_elliptic_curves` is documented to have the second entry be the
// network-specific curve (if it exists and is distinct from Embedwards25519)
if let Some(curve) = set.network.embedded_elliptic_curves().get(1) {
serai.embedded_elliptic_curve_key(validator, *curve).await.map(Some)
} else {
Ok(None)
}
}
)
.map(|(substrate_embedded_key, external_embedded_key)| {
(validator, substrate_embedded_key, external_embedded_key)
})
});
}
let mut evrf_public_keys = Vec::with_capacity(usize::from(total_weight));
for (validator, weight) in &validators {
let (future_validator, substrate_embedded_key, external_embedded_key) =
embedded_elliptic_curve_keys.next().await.unwrap().map_err(|e| format!("{e:?}"))?;
assert_eq!(*validator, future_validator);
let external_embedded_key =
external_embedded_key.unwrap_or(substrate_embedded_key.clone());
match (substrate_embedded_key, external_embedded_key) {
(Some(substrate_embedded_key), Some(external_embedded_key)) => {
let substrate_embedded_key = <[u8; 32]>::try_from(substrate_embedded_key)
.map_err(|_| "Embedwards25519 key wasn't 32 bytes".to_string())?;
for _ in 0 .. *weight {
evrf_public_keys.push((substrate_embedded_key, external_embedded_key.clone()));
}
}
_ => Err("NewSet with validator missing an embedded key".to_string())?,
}
}
crate::NewSet::send(
&mut txn,
&NewSetInformation {
set: *set,
serai_block: block.block_hash,
declaration_time: block.time,
// TODO: Why do we have this as an explicit field here?
// Shouldn't thiis be inlined into the Processor's key gen code, where it's used?
threshold: ((total_weight * 2) / 3) + 1,
validators,
evrf_public_keys,
},
);
}
}
for accepted_handover in block.accepted_handover_events {
let serai_client::validator_sets::ValidatorSetsEvent::AcceptedHandover { set } =
&accepted_handover
else {
panic!("AcceptedHandover event wasn't a AcceptedHandover event: {accepted_handover:?}");
};
crate::SignSlashReport::send(&mut txn, *set);
}
txn.commit();
}
Ok(next_block <= latest_finalized_block)
}
}
}

View File

@@ -0,0 +1,228 @@
#![cfg_attr(docsrs, feature(doc_auto_cfg))]
#![doc = include_str!("../README.md")]
#![deny(missing_docs)]
use scale::{Encode, Decode};
use borsh::{io, BorshSerialize, BorshDeserialize};
use serai_client::{
primitives::{NetworkId, PublicKey, Signature, SeraiAddress},
validator_sets::primitives::{Session, ValidatorSet, KeyPair},
in_instructions::primitives::SignedBatch,
Transaction,
};
use serai_db::*;
mod canonical;
pub use canonical::CanonicalEventStream;
mod ephemeral;
pub use ephemeral::EphemeralEventStream;
mod set_keys;
pub use set_keys::SetKeysTask;
mod publish_batch;
pub use publish_batch::PublishBatchTask;
mod publish_slash_report;
pub use publish_slash_report::PublishSlashReportTask;
fn borsh_serialize_validators<W: io::Write>(
validators: &Vec<(PublicKey, u16)>,
writer: &mut W,
) -> Result<(), io::Error> {
// This doesn't use `encode_to` as `encode_to` panics if the writer returns an error
writer.write_all(&validators.encode())
}
fn borsh_deserialize_validators<R: io::Read>(
reader: &mut R,
) -> Result<Vec<(PublicKey, u16)>, io::Error> {
Decode::decode(&mut scale::IoReader(reader)).map_err(io::Error::other)
}
/// The information for a new set.
#[derive(Clone, Debug, BorshSerialize, BorshDeserialize)]
pub struct NewSetInformation {
/// The set.
pub set: ValidatorSet,
/// The Serai block which declared it.
pub serai_block: [u8; 32],
/// The time of the block which declared it, in seconds.
pub declaration_time: u64,
/// The threshold to use.
pub threshold: u16,
/// The validators, with the amount of key shares they have.
#[borsh(
serialize_with = "borsh_serialize_validators",
deserialize_with = "borsh_deserialize_validators"
)]
pub validators: Vec<(PublicKey, u16)>,
/// The eVRF public keys.
pub evrf_public_keys: Vec<([u8; 32], Vec<u8>)>,
}
mod _public_db {
use super::*;
db_channel!(
CoordinatorSubstrate {
// Canonical messages to send to the processor
Canonical: (network: NetworkId) -> messages::substrate::CoordinatorMessage,
// Relevant new set, from an ephemeral event stream
NewSet: () -> NewSetInformation,
// Potentially relevant sign slash report, from an ephemeral event stream
SignSlashReport: (set: ValidatorSet) -> (),
// Signed batches to publish onto the Serai network
SignedBatches: (network: NetworkId) -> SignedBatch,
}
);
create_db!(
CoordinatorSubstrate {
// Keys to set on the Serai network
Keys: (network: NetworkId) -> (Session, Vec<u8>),
// Slash reports to publish onto the Serai network
SlashReports: (network: NetworkId) -> (Session, Vec<u8>),
}
);
}
/// The canonical event stream.
pub struct Canonical;
impl Canonical {
pub(crate) fn send(
txn: &mut impl DbTxn,
network: NetworkId,
msg: &messages::substrate::CoordinatorMessage,
) {
_public_db::Canonical::send(txn, network, msg);
}
/// Try to receive a canonical event, returning `None` if there is none to receive.
pub fn try_recv(
txn: &mut impl DbTxn,
network: NetworkId,
) -> Option<messages::substrate::CoordinatorMessage> {
_public_db::Canonical::try_recv(txn, network)
}
}
/// The channel for new set events emitted by an ephemeral event stream.
pub struct NewSet;
impl NewSet {
pub(crate) fn send(txn: &mut impl DbTxn, msg: &NewSetInformation) {
_public_db::NewSet::send(txn, msg);
}
/// Try to receive a new set's information, returning `None` if there is none to receive.
pub fn try_recv(txn: &mut impl DbTxn) -> Option<NewSetInformation> {
_public_db::NewSet::try_recv(txn)
}
}
/// The channel for notifications to sign a slash report, as emitted by an ephemeral event stream.
///
/// These notifications MAY be for irrelevant validator sets. The only guarantee is the
/// notifications for all relevant validator sets will be included.
pub struct SignSlashReport;
impl SignSlashReport {
pub(crate) fn send(txn: &mut impl DbTxn, set: ValidatorSet) {
_public_db::SignSlashReport::send(txn, set, &());
}
/// Try to receive a notification to sign a slash report, returning `None` if there is none to
/// receive.
pub fn try_recv(txn: &mut impl DbTxn, set: ValidatorSet) -> Option<()> {
_public_db::SignSlashReport::try_recv(txn, set)
}
}
/// The keys to set on Serai.
pub struct Keys;
impl Keys {
/// Set the keys to report for a validator set.
///
/// This only saves the most recent keys as only a single session is eligible to have its keys
/// reported at once.
pub fn set(
txn: &mut impl DbTxn,
set: ValidatorSet,
key_pair: KeyPair,
signature_participants: bitvec::vec::BitVec<u8, bitvec::order::Lsb0>,
signature: Signature,
) {
// If we have a more recent pair of keys, don't write this historic one
if let Some((existing_session, _)) = _public_db::Keys::get(txn, set.network) {
if existing_session.0 >= set.session.0 {
return;
}
}
let tx = serai_client::validator_sets::SeraiValidatorSets::set_keys(
set.network,
key_pair,
signature_participants,
signature,
);
_public_db::Keys::set(txn, set.network, &(set.session, tx.encode()));
}
pub(crate) fn take(txn: &mut impl DbTxn, network: NetworkId) -> Option<(Session, Transaction)> {
let (session, tx) = _public_db::Keys::take(txn, network)?;
Some((session, <_>::decode(&mut tx.as_slice()).unwrap()))
}
}
/// The signed batches to publish onto Serai.
pub struct SignedBatches;
impl SignedBatches {
/// Send a `SignedBatch` to publish onto Serai.
///
/// These will be published sequentially. Out-of-order sending risks hanging the task.
pub fn send(txn: &mut impl DbTxn, batch: &SignedBatch) {
_public_db::SignedBatches::send(txn, batch.batch.network, batch);
}
pub(crate) fn try_recv(txn: &mut impl DbTxn, network: NetworkId) -> Option<SignedBatch> {
_public_db::SignedBatches::try_recv(txn, network)
}
}
/// The slash report was invalid.
#[derive(Debug)]
pub struct InvalidSlashReport;
/// The slash reports to publish onto Serai.
pub struct SlashReports;
impl SlashReports {
/// Set the slashes to report for a validator set.
///
/// This only saves the most recent slashes as only a single session is eligible to have its
/// slashes reported at once.
///
/// Returns Err if the slashes are invalid. Returns Ok if the slashes weren't detected as
/// invalid. Slashes may be considered invalid by the Serai blockchain later even if not detected
/// as invalid here.
pub fn set(
txn: &mut impl DbTxn,
set: ValidatorSet,
slashes: Vec<(SeraiAddress, u32)>,
signature: Signature,
) -> Result<(), InvalidSlashReport> {
// If we have a more recent slash report, don't write this historic one
if let Some((existing_session, _)) = _public_db::SlashReports::get(txn, set.network) {
if existing_session.0 >= set.session.0 {
return Ok(());
}
}
let tx = serai_client::validator_sets::SeraiValidatorSets::report_slashes(
set.network,
slashes.try_into().map_err(|_| InvalidSlashReport)?,
signature,
);
_public_db::SlashReports::set(txn, set.network, &(set.session, tx.encode()));
Ok(())
}
pub(crate) fn take(txn: &mut impl DbTxn, network: NetworkId) -> Option<(Session, Transaction)> {
let (session, tx) = _public_db::SlashReports::take(txn, network)?;
Some((session, <_>::decode(&mut tx.as_slice()).unwrap()))
}
}

View File

@@ -0,0 +1,66 @@
use core::future::Future;
use std::sync::Arc;
use serai_db::{DbTxn, Db};
use serai_client::{primitives::NetworkId, SeraiError, Serai};
use serai_task::ContinuallyRan;
use crate::SignedBatches;
/// Publish `SignedBatch`s from `SignedBatches` onto Serai.
pub struct PublishBatchTask<D: Db> {
db: D,
serai: Arc<Serai>,
network: NetworkId,
}
impl<D: Db> PublishBatchTask<D> {
/// Create a task to publish `SignedBatch`s onto Serai.
///
/// Returns None if `network == NetworkId::Serai`.
// TODO: ExternalNetworkId
pub fn new(db: D, serai: Arc<Serai>, network: NetworkId) -> Option<Self> {
if network == NetworkId::Serai {
None?
};
Some(Self { db, serai, network })
}
}
impl<D: Db> ContinuallyRan for PublishBatchTask<D> {
type Error = SeraiError;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
loop {
let mut txn = self.db.txn();
let Some(batch) = SignedBatches::try_recv(&mut txn, self.network) else {
// No batch to publish at this time
break;
};
// Publish this Batch if it hasn't already been published
let serai = self.serai.as_of_latest_finalized_block().await?;
let last_batch = serai.in_instructions().last_batch_for_network(self.network).await?;
if last_batch < Some(batch.batch.id) {
// This stream of Batches *should* be sequential within the larger context of the Serai
// coordinator. In this library, we use a more relaxed definition and don't assert
// sequence. This does risk hanging the task, if Batch #n+1 is sent before Batch #n, but
// that is a documented fault of the `SignedBatches` API.
self
.serai
.publish(&serai_client::in_instructions::SeraiInInstructions::execute_batch(batch))
.await?;
}
txn.commit();
made_progress = true;
}
Ok(made_progress)
}
}
}

View File

@@ -0,0 +1,89 @@
use core::future::Future;
use std::sync::Arc;
use serai_db::{DbTxn, Db};
use serai_client::{primitives::NetworkId, validator_sets::primitives::Session, Serai};
use serai_task::ContinuallyRan;
use crate::SlashReports;
/// Publish slash reports from `SlashReports` onto Serai.
pub struct PublishSlashReportTask<D: Db> {
db: D,
serai: Arc<Serai>,
}
impl<D: Db> PublishSlashReportTask<D> {
/// Create a task to publish slash reports onto Serai.
pub fn new(db: D, serai: Arc<Serai>) -> Self {
Self { db, serai }
}
}
impl<D: Db> ContinuallyRan for PublishSlashReportTask<D> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
for network in serai_client::primitives::NETWORKS {
if network == NetworkId::Serai {
continue;
};
let mut txn = self.db.txn();
let Some((session, slash_report)) = SlashReports::take(&mut txn, network) else {
// No slash report to publish
continue;
};
let serai =
self.serai.as_of_latest_finalized_block().await.map_err(|e| format!("{e:?}"))?;
let serai = serai.validator_sets();
let session_after_slash_report = Session(session.0 + 1);
let current_session = serai.session(network).await.map_err(|e| format!("{e:?}"))?;
let current_session = current_session.map(|session| session.0);
// Only attempt to publish the slash report for session #n while session #n+1 is still
// active
let session_after_slash_report_retired =
current_session > Some(session_after_slash_report.0);
if session_after_slash_report_retired {
// Commit the txn to drain this slash report from the database and not try it again later
txn.commit();
continue;
}
if Some(session_after_slash_report.0) != current_session {
// We already checked the current session wasn't greater, and they're not equal
assert!(current_session < Some(session_after_slash_report.0));
// This would mean the Serai node is resyncing and is behind where it prior was
Err("have a slash report for a session Serai has yet to retire".to_string())?;
}
// If this session which should publish a slash report already has, move on
let key_pending_slash_report =
serai.key_pending_slash_report(network).await.map_err(|e| format!("{e:?}"))?;
if key_pending_slash_report.is_none() {
txn.commit();
continue;
};
match self.serai.publish(&slash_report).await {
Ok(()) => {
txn.commit();
made_progress = true;
}
// This could be specific to this TX (such as an already in mempool error) and it may be
// worthwhile to continue iteration with the other pending slash reports. We assume this
// error ephemeral and that the latency incurred for this ephemeral error to resolve is
// miniscule compared to the window available to publish the slash report. That makes
// this a non-issue.
Err(e) => Err(format!("couldn't publish slash report transaction: {e:?}"))?,
}
}
Ok(made_progress)
}
}
}

View File

@@ -0,0 +1,88 @@
use core::future::Future;
use std::sync::Arc;
use serai_db::{DbTxn, Db};
use serai_client::{primitives::NetworkId, validator_sets::primitives::ValidatorSet, Serai};
use serai_task::ContinuallyRan;
use crate::Keys;
/// Set keys from `Keys` on Serai.
pub struct SetKeysTask<D: Db> {
db: D,
serai: Arc<Serai>,
}
impl<D: Db> SetKeysTask<D> {
/// Create a task to publish slash reports onto Serai.
pub fn new(db: D, serai: Arc<Serai>) -> Self {
Self { db, serai }
}
}
impl<D: Db> ContinuallyRan for SetKeysTask<D> {
type Error = String;
fn run_iteration(&mut self) -> impl Send + Future<Output = Result<bool, Self::Error>> {
async move {
let mut made_progress = false;
for network in serai_client::primitives::NETWORKS {
if network == NetworkId::Serai {
continue;
};
let mut txn = self.db.txn();
let Some((session, keys)) = Keys::take(&mut txn, network) else {
// No keys to set
continue;
};
let serai =
self.serai.as_of_latest_finalized_block().await.map_err(|e| format!("{e:?}"))?;
let serai = serai.validator_sets();
let current_session = serai.session(network).await.map_err(|e| format!("{e:?}"))?;
let current_session = current_session.map(|session| session.0);
// Only attempt to set these keys if this isn't a retired session
if Some(session.0) < current_session {
// Commit the txn to take these keys from the database and not try it again later
txn.commit();
continue;
}
if Some(session.0) != current_session {
// We already checked the current session wasn't greater, and they're not equal
assert!(current_session < Some(session.0));
// This would mean the Serai node is resyncing and is behind where it prior was
Err("have a keys for a session Serai has yet to start".to_string())?;
}
// If this session already has had its keys set, move on
if serai
.keys(ValidatorSet { network, session })
.await
.map_err(|e| format!("{e:?}"))?
.is_some()
{
txn.commit();
continue;
};
match self.serai.publish(&keys).await {
Ok(()) => {
txn.commit();
made_progress = true;
}
// This could be specific to this TX (such as an already in mempool error) and it may be
// worthwhile to continue iteration with the other pending slash reports. We assume this
// error ephemeral and that the latency incurred for this ephemeral error to resolve is
// miniscule compared to the window reasonable to set the keys. That makes this a
// non-issue.
Err(e) => Err(format!("couldn't publish set keys transaction: {e:?}"))?,
}
}
Ok(made_progress)
}
}
}

View File

@@ -0,0 +1,49 @@
[package]
name = "tributary-sdk"
version = "0.1.0"
description = "A micro-blockchain to provide consensus and ordering to P2P communication"
license = "AGPL-3.0-only"
repository = "https://github.com/serai-dex/serai/tree/develop/coordinator/tributary-sdk"
authors = ["Luke Parker <lukeparker5132@gmail.com>"]
edition = "2021"
rust-version = "1.81"
[package.metadata.docs.rs]
all-features = true
rustdoc-args = ["--cfg", "docsrs"]
[lints]
workspace = true
[dependencies]
thiserror = { version = "2", default-features = false, features = ["std"] }
subtle = { version = "^2", default-features = false, features = ["std"] }
zeroize = { version = "^1.5", default-features = false, features = ["std"] }
rand = { version = "0.8", default-features = false, features = ["std"] }
rand_chacha = { version = "0.3", default-features = false, features = ["std"] }
blake2 = { version = "0.10", default-features = false, features = ["std"] }
transcript = { package = "flexible-transcript", path = "../../crypto/transcript", version = "0.3", default-features = false, features = ["std", "recommended"] }
ciphersuite = { package = "ciphersuite", path = "../../crypto/ciphersuite", version = "0.4", default-features = false, features = ["std", "ristretto"] }
schnorr = { package = "schnorr-signatures", path = "../../crypto/schnorr", version = "0.5", default-features = false, features = ["std"] }
hex = { version = "0.4", default-features = false, features = ["std"] }
log = { version = "0.4", default-features = false, features = ["std"] }
serai-db = { path = "../../common/db", version = "0.1" }
scale = { package = "parity-scale-codec", version = "3", default-features = false, features = ["std", "derive"] }
futures-util = { version = "0.3", default-features = false, features = ["std", "sink", "channel"] }
futures-channel = { version = "0.3", default-features = false, features = ["std", "sink"] }
tendermint = { package = "tendermint-machine", path = "./tendermint", version = "0.2" }
tokio = { version = "1", default-features = false, features = ["sync", "time", "rt"] }
[dev-dependencies]
tokio = { version = "1", features = ["macros"] }
[features]
tests = []

View File

@@ -0,0 +1,15 @@
AGPL-3.0-only license
Copyright (c) 2023 Luke Parker
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License Version 3 as
published by the Free Software Foundation.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.

View File

@@ -0,0 +1,3 @@
# Tributary
A verifiable, ordered broadcast layer implemented as a BFT micro-blockchain.

View File

@@ -135,7 +135,7 @@ impl<T: TransactionTrait> Block<T> {
// Check TXs are sorted by nonce.
let nonce = |tx: &Transaction<T>| {
if let TransactionKind::Signed(_, Signed { nonce, .. }) = tx.kind() {
*nonce
nonce
} else {
0
}

View File

@@ -323,7 +323,7 @@ impl<D: Db, T: TransactionTrait> Blockchain<D, T> {
}
TransactionKind::Signed(order, Signed { signer, nonce, .. }) => {
let next_nonce = nonce + 1;
txn.put(Self::next_nonce_key(&self.genesis, signer, &order), next_nonce.to_le_bytes());
txn.put(Self::next_nonce_key(&self.genesis, &signer, &order), next_nonce.to_le_bytes());
self.mempool.remove(&tx.hash());
}
}

View File

@@ -0,0 +1,388 @@
use core::{marker::PhantomData, fmt::Debug, future::Future};
use std::{sync::Arc, io};
use zeroize::Zeroizing;
use ciphersuite::{Ciphersuite, Ristretto};
use scale::Decode;
use futures_channel::mpsc::UnboundedReceiver;
use futures_util::{StreamExt, SinkExt};
use ::tendermint::{
ext::{BlockNumber, Commit, Block as BlockTrait, Network},
SignedMessageFor, SyncedBlock, SyncedBlockSender, SyncedBlockResultReceiver, MessageSender,
TendermintMachine, TendermintHandle,
};
pub use ::tendermint::Evidence;
use serai_db::Db;
use tokio::sync::RwLock;
mod merkle;
pub(crate) use merkle::*;
pub mod transaction;
pub use transaction::{TransactionError, Signed, TransactionKind, Transaction as TransactionTrait};
use crate::tendermint::tx::TendermintTx;
mod provided;
pub(crate) use provided::*;
pub use provided::ProvidedError;
mod block;
pub use block::*;
mod blockchain;
pub(crate) use blockchain::*;
mod mempool;
pub(crate) use mempool::*;
pub mod tendermint;
pub(crate) use crate::tendermint::*;
#[cfg(any(test, feature = "tests"))]
pub mod tests;
/// Size limit for an individual transaction.
// This needs to be big enough to participate in a 101-of-150 eVRF DKG with each element taking
// `MAX_KEY_LEN`. This also needs to be big enough to pariticpate in signing 520 Bitcoin inputs
// with 49 key shares, and signing 120 Monero inputs with 49 key shares.
// TODO: Add a test for these properties
pub const TRANSACTION_SIZE_LIMIT: usize = 2_000_000;
/// Amount of transactions a single account may have in the mempool.
pub const ACCOUNT_MEMPOOL_LIMIT: u32 = 50;
/// Block size limit.
// This targets a growth limit of roughly 30 GB a day, under load, in order to prevent a malicious
// participant from flooding disks and causing out of space errors in order processes.
pub const BLOCK_SIZE_LIMIT: usize = 2_001_000;
pub(crate) const TENDERMINT_MESSAGE: u8 = 0;
pub(crate) const TRANSACTION_MESSAGE: u8 = 1;
#[allow(clippy::large_enum_variant)]
#[derive(Clone, PartialEq, Eq, Debug)]
pub enum Transaction<T: TransactionTrait> {
Tendermint(TendermintTx),
Application(T),
}
impl<T: TransactionTrait> ReadWrite for Transaction<T> {
fn read<R: io::Read>(reader: &mut R) -> io::Result<Self> {
let mut kind = [0];
reader.read_exact(&mut kind)?;
match kind[0] {
0 => {
let tx = TendermintTx::read(reader)?;
Ok(Transaction::Tendermint(tx))
}
1 => {
let tx = T::read(reader)?;
Ok(Transaction::Application(tx))
}
_ => Err(io::Error::other("invalid transaction type")),
}
}
fn write<W: io::Write>(&self, writer: &mut W) -> io::Result<()> {
match self {
Transaction::Tendermint(tx) => {
writer.write_all(&[0])?;
tx.write(writer)
}
Transaction::Application(tx) => {
writer.write_all(&[1])?;
tx.write(writer)
}
}
}
}
impl<T: TransactionTrait> Transaction<T> {
pub fn hash(&self) -> [u8; 32] {
match self {
Transaction::Tendermint(tx) => tx.hash(),
Transaction::Application(tx) => tx.hash(),
}
}
pub fn kind(&self) -> TransactionKind {
match self {
Transaction::Tendermint(tx) => tx.kind(),
Transaction::Application(tx) => tx.kind(),
}
}
}
/// An item which can be read and written.
pub trait ReadWrite: Sized {
fn read<R: io::Read>(reader: &mut R) -> io::Result<Self>;
fn write<W: io::Write>(&self, writer: &mut W) -> io::Result<()>;
fn serialize(&self) -> Vec<u8> {
// BlockHeader is 64 bytes and likely the smallest item in this system
let mut buf = Vec::with_capacity(64);
self.write(&mut buf).unwrap();
buf
}
}
pub trait P2p: 'static + Send + Sync + Clone {
/// Broadcast a message to all other members of the Tributary with the specified genesis.
///
/// The Tributary will re-broadcast consensus messages on a fixed interval to ensure they aren't
/// prematurely dropped from the P2P layer. THe P2P layer SHOULD perform content-based
/// deduplication to ensure a sane amount of load.
fn broadcast(&self, genesis: [u8; 32], msg: Vec<u8>) -> impl Send + Future<Output = ()>;
}
impl<P: P2p> P2p for Arc<P> {
fn broadcast(&self, genesis: [u8; 32], msg: Vec<u8>) -> impl Send + Future<Output = ()> {
P::broadcast(self, genesis, msg)
}
}
#[derive(Clone)]
pub struct Tributary<D: Db, T: TransactionTrait, P: P2p> {
db: D,
genesis: [u8; 32],
network: TendermintNetwork<D, T, P>,
synced_block: Arc<RwLock<SyncedBlockSender<TendermintNetwork<D, T, P>>>>,
synced_block_result: Arc<RwLock<SyncedBlockResultReceiver>>,
messages: Arc<RwLock<MessageSender<TendermintNetwork<D, T, P>>>>,
}
impl<D: Db, T: TransactionTrait, P: P2p> Tributary<D, T, P> {
pub async fn new(
db: D,
genesis: [u8; 32],
start_time: u64,
key: Zeroizing<<Ristretto as Ciphersuite>::F>,
validators: Vec<(<Ristretto as Ciphersuite>::G, u64)>,
p2p: P,
) -> Option<Self> {
log::info!("new Tributary with genesis {}", hex::encode(genesis));
let validators_vec = validators.iter().map(|validator| validator.0).collect::<Vec<_>>();
let signer = Arc::new(Signer::new(genesis, key));
let validators = Arc::new(Validators::new(genesis, validators)?);
let mut blockchain = Blockchain::new(db.clone(), genesis, &validators_vec);
let block_number = BlockNumber(blockchain.block_number());
let start_time = if let Some(commit) = blockchain.commit(&blockchain.tip()) {
Commit::<Validators>::decode(&mut commit.as_ref()).unwrap().end_time
} else {
start_time
};
let proposal = TendermintBlock(
blockchain.build_block::<TendermintNetwork<D, T, P>>(&validators).serialize(),
);
let blockchain = Arc::new(RwLock::new(blockchain));
let network = TendermintNetwork { genesis, signer, validators, blockchain, p2p };
let TendermintHandle { synced_block, synced_block_result, messages, machine } =
TendermintMachine::new(
db.clone(),
network.clone(),
genesis,
block_number,
start_time,
proposal,
)
.await;
tokio::spawn(machine.run());
Some(Self {
db,
genesis,
network,
synced_block: Arc::new(RwLock::new(synced_block)),
synced_block_result: Arc::new(RwLock::new(synced_block_result)),
messages: Arc::new(RwLock::new(messages)),
})
}
pub fn block_time() -> u32 {
TendermintNetwork::<D, T, P>::block_time()
}
pub fn genesis(&self) -> [u8; 32] {
self.genesis
}
pub async fn block_number(&self) -> u64 {
self.network.blockchain.read().await.block_number()
}
pub async fn tip(&self) -> [u8; 32] {
self.network.blockchain.read().await.tip()
}
pub fn reader(&self) -> TributaryReader<D, T> {
TributaryReader(self.db.clone(), self.genesis, PhantomData)
}
pub async fn provide_transaction(&self, tx: T) -> Result<(), ProvidedError> {
self.network.blockchain.write().await.provide_transaction(tx)
}
pub async fn next_nonce(
&self,
signer: &<Ristretto as Ciphersuite>::G,
order: &[u8],
) -> Option<u32> {
self.network.blockchain.read().await.next_nonce(signer, order)
}
// Returns Ok(true) if new, Ok(false) if an already present unsigned, or the error.
// Safe to be &self since the only meaningful usage of self is self.network.blockchain which
// successfully acquires its own write lock
pub async fn add_transaction(&self, tx: T) -> Result<bool, TransactionError> {
let tx = Transaction::Application(tx);
let mut to_broadcast = vec![TRANSACTION_MESSAGE];
tx.write(&mut to_broadcast).unwrap();
let res = self.network.blockchain.write().await.add_transaction::<TendermintNetwork<D, T, P>>(
true,
tx,
&self.network.signature_scheme(),
);
if res == Ok(true) {
self.network.p2p.broadcast(self.genesis, to_broadcast).await;
}
res
}
async fn sync_block_internal(
&self,
block: Block<T>,
commit: Vec<u8>,
result: &mut UnboundedReceiver<bool>,
) -> bool {
let (tip, block_number) = {
let blockchain = self.network.blockchain.read().await;
(blockchain.tip(), blockchain.block_number())
};
if block.header.parent != tip {
log::debug!("told to sync a block whose parent wasn't our tip");
return false;
}
let block = TendermintBlock(block.serialize());
let mut commit_ref = commit.as_ref();
let Ok(commit) = Commit::<Arc<Validators>>::decode(&mut commit_ref) else {
log::error!("sent an invalidly serialized commit");
return false;
};
// Storage DoS vector. We *could* truncate to solely the relevant portion, trying to save this,
// yet then we'd have to test the truncation was performed correctly.
if !commit_ref.is_empty() {
log::error!("sent an commit with additional data after it");
return false;
}
if !self.network.verify_commit(block.id(), &commit) {
log::error!("sent an invalid commit");
return false;
}
let number = BlockNumber(block_number + 1);
self.synced_block.write().await.send(SyncedBlock { number, block, commit }).await.unwrap();
result.next().await.unwrap()
}
// Sync a block.
// TODO: Since we have a static validator set, we should only need the tail commit?
pub async fn sync_block(&self, block: Block<T>, commit: Vec<u8>) -> bool {
let mut result = self.synced_block_result.write().await;
self.sync_block_internal(block, commit, &mut result).await
}
// Return true if the message should be rebroadcasted.
pub async fn handle_message(&self, msg: &[u8]) -> bool {
match msg.first() {
Some(&TRANSACTION_MESSAGE) => {
let Ok(tx) = Transaction::read::<&[u8]>(&mut &msg[1 ..]) else {
log::error!("received invalid transaction message");
return false;
};
// TODO: Sync mempools with fellow peers
// Can we just rebroadcast transactions not included for at least two blocks?
let res =
self.network.blockchain.write().await.add_transaction::<TendermintNetwork<D, T, P>>(
false,
tx,
&self.network.signature_scheme(),
);
log::debug!("received transaction message. valid new transaction: {res:?}");
res == Ok(true)
}
Some(&TENDERMINT_MESSAGE) => {
let Ok(msg) =
SignedMessageFor::<TendermintNetwork<D, T, P>>::decode::<&[u8]>(&mut &msg[1 ..])
else {
log::error!("received invalid tendermint message");
return false;
};
self.messages.write().await.send(msg).await.unwrap();
false
}
_ => false,
}
}
/// Get a Future which will resolve once the next block has been added.
pub async fn next_block_notification(
&self,
) -> impl Send + Sync + core::future::Future<Output = Result<(), impl Send + Sync>> {
let (tx, rx) = tokio::sync::oneshot::channel();
self.network.blockchain.write().await.next_block_notifications.push_back(tx);
rx
}
}
#[derive(Clone)]
pub struct TributaryReader<D: Db, T: TransactionTrait>(D, [u8; 32], PhantomData<T>);
impl<D: Db, T: TransactionTrait> TributaryReader<D, T> {
pub fn genesis(&self) -> [u8; 32] {
self.1
}
// Since these values are static once set, they can be safely read from the database without lock
// acquisition
pub fn block(&self, hash: &[u8; 32]) -> Option<Block<T>> {
Blockchain::<D, T>::block_from_db(&self.0, self.1, hash)
}
pub fn commit(&self, hash: &[u8; 32]) -> Option<Vec<u8>> {
Blockchain::<D, T>::commit_from_db(&self.0, self.1, hash)
}
pub fn parsed_commit(&self, hash: &[u8; 32]) -> Option<Commit<Validators>> {
self.commit(hash).map(|commit| Commit::<Validators>::decode(&mut commit.as_ref()).unwrap())
}
pub fn block_after(&self, hash: &[u8; 32]) -> Option<[u8; 32]> {
Blockchain::<D, T>::block_after(&self.0, self.1, hash)
}
pub fn time_of_block(&self, hash: &[u8; 32]) -> Option<u64> {
self
.commit(hash)
.map(|commit| Commit::<Validators>::decode(&mut commit.as_ref()).unwrap().end_time)
}
pub fn locally_provided_txs_in_block(&self, hash: &[u8; 32], order: &str) -> bool {
Blockchain::<D, T>::locally_provided_txs_in_block(&self.0, &self.1, hash, order)
}
// This isn't static, yet can be read with only minor discrepancy risks
pub fn tip(&self) -> [u8; 32] {
Blockchain::<D, T>::tip_from_db(&self.0, self.1)
}
}

View File

@@ -81,11 +81,11 @@ impl<D: Db, T: TransactionTrait> Mempool<D, T> {
}
Transaction::Application(tx) => match tx.kind() {
TransactionKind::Signed(order, Signed { signer, nonce, .. }) => {
let amount = *res.txs_per_signer.get(signer).unwrap_or(&0) + 1;
res.txs_per_signer.insert(*signer, amount);
let amount = *res.txs_per_signer.get(&signer).unwrap_or(&0) + 1;
res.txs_per_signer.insert(signer, amount);
if let Some(prior_nonce) =
res.last_nonce_in_mempool.insert((*signer, order.clone()), *nonce)
res.last_nonce_in_mempool.insert((signer, order.clone()), nonce)
{
assert_eq!(prior_nonce, nonce - 1);
}
@@ -133,14 +133,14 @@ impl<D: Db, T: TransactionTrait> Mempool<D, T> {
match app_tx.kind() {
TransactionKind::Signed(order, Signed { signer, .. }) => {
// Get the nonce from the blockchain
let Some(blockchain_next_nonce) = blockchain_next_nonce(*signer, order.clone()) else {
let Some(blockchain_next_nonce) = blockchain_next_nonce(signer, order.clone()) else {
// Not a participant
Err(TransactionError::InvalidSigner)?
};
let mut next_nonce = blockchain_next_nonce;
if let Some(mempool_last_nonce) =
self.last_nonce_in_mempool.get(&(*signer, order.clone()))
self.last_nonce_in_mempool.get(&(signer, order.clone()))
{
assert!(*mempool_last_nonce >= blockchain_next_nonce);
next_nonce = *mempool_last_nonce + 1;
@@ -148,14 +148,14 @@ impl<D: Db, T: TransactionTrait> Mempool<D, T> {
// If we have too many transactions from this sender, don't add this yet UNLESS we are
// this sender
let amount_in_pool = *self.txs_per_signer.get(signer).unwrap_or(&0) + 1;
let amount_in_pool = *self.txs_per_signer.get(&signer).unwrap_or(&0) + 1;
if !internal && (amount_in_pool > ACCOUNT_MEMPOOL_LIMIT) {
Err(TransactionError::TooManyInMempool)?;
}
verify_transaction(app_tx, self.genesis, &mut |_, _| Some(next_nonce))?;
self.last_nonce_in_mempool.insert((*signer, order.clone()), next_nonce);
self.txs_per_signer.insert(*signer, amount_in_pool);
self.last_nonce_in_mempool.insert((signer, order.clone()), next_nonce);
self.txs_per_signer.insert(signer, amount_in_pool);
}
TransactionKind::Unsigned => {
// check we have the tx in the pool/chain
@@ -205,7 +205,7 @@ impl<D: Db, T: TransactionTrait> Mempool<D, T> {
// Sort signed by nonce
let nonce = |tx: &Transaction<T>| {
if let TransactionKind::Signed(_, Signed { nonce, .. }) = tx.kind() {
*nonce
nonce
} else {
unreachable!()
}
@@ -242,11 +242,11 @@ impl<D: Db, T: TransactionTrait> Mempool<D, T> {
if let Some(tx) = self.txs.remove(tx) {
if let TransactionKind::Signed(order, Signed { signer, nonce, .. }) = tx.kind() {
let amount = *self.txs_per_signer.get(signer).unwrap() - 1;
self.txs_per_signer.insert(*signer, amount);
let amount = *self.txs_per_signer.get(&signer).unwrap() - 1;
self.txs_per_signer.insert(signer, amount);
if self.last_nonce_in_mempool.get(&(*signer, order.clone())) == Some(nonce) {
self.last_nonce_in_mempool.remove(&(*signer, order));
if self.last_nonce_in_mempool.get(&(signer, order.clone())) == Some(&nonce) {
self.last_nonce_in_mempool.remove(&(signer, order));
}
}
}

View File

@@ -1,8 +1,6 @@
use core::ops::Deref;
use core::{ops::Deref, future::Future};
use std::{sync::Arc, collections::HashMap};
use async_trait::async_trait;
use subtle::ConstantTimeEq;
use zeroize::{Zeroize, Zeroizing};
@@ -74,50 +72,52 @@ impl Signer {
}
}
#[async_trait]
impl SignerTrait for Signer {
type ValidatorId = [u8; 32];
type Signature = [u8; 64];
/// Returns the validator's current ID. Returns None if they aren't a current validator.
async fn validator_id(&self) -> Option<Self::ValidatorId> {
Some((Ristretto::generator() * self.key.deref()).to_bytes())
fn validator_id(&self) -> impl Send + Future<Output = Option<Self::ValidatorId>> {
async move { Some((Ristretto::generator() * self.key.deref()).to_bytes()) }
}
/// Sign a signature with the current validator's private key.
async fn sign(&self, msg: &[u8]) -> Self::Signature {
let mut nonce = Zeroizing::new(RecommendedTranscript::new(b"Tributary Chain Tendermint Nonce"));
nonce.append_message(b"genesis", self.genesis);
nonce.append_message(b"key", Zeroizing::new(self.key.deref().to_repr()).as_ref());
nonce.append_message(b"message", msg);
let mut nonce = nonce.challenge(b"nonce");
fn sign(&self, msg: &[u8]) -> impl Send + Future<Output = Self::Signature> {
async move {
let mut nonce =
Zeroizing::new(RecommendedTranscript::new(b"Tributary Chain Tendermint Nonce"));
nonce.append_message(b"genesis", self.genesis);
nonce.append_message(b"key", Zeroizing::new(self.key.deref().to_repr()).as_ref());
nonce.append_message(b"message", msg);
let mut nonce = nonce.challenge(b"nonce");
let mut nonce_arr = [0; 64];
nonce_arr.copy_from_slice(nonce.as_ref());
let mut nonce_arr = [0; 64];
nonce_arr.copy_from_slice(nonce.as_ref());
let nonce_ref: &mut [u8] = nonce.as_mut();
nonce_ref.zeroize();
let nonce_ref: &[u8] = nonce.as_ref();
assert_eq!(nonce_ref, [0; 64].as_ref());
let nonce_ref: &mut [u8] = nonce.as_mut();
nonce_ref.zeroize();
let nonce_ref: &[u8] = nonce.as_ref();
assert_eq!(nonce_ref, [0; 64].as_ref());
let nonce =
Zeroizing::new(<Ristretto as Ciphersuite>::F::from_bytes_mod_order_wide(&nonce_arr));
nonce_arr.zeroize();
let nonce =
Zeroizing::new(<Ristretto as Ciphersuite>::F::from_bytes_mod_order_wide(&nonce_arr));
nonce_arr.zeroize();
assert!(!bool::from(nonce.ct_eq(&<Ristretto as Ciphersuite>::F::ZERO)));
assert!(!bool::from(nonce.ct_eq(&<Ristretto as Ciphersuite>::F::ZERO)));
let challenge = challenge(
self.genesis,
(Ristretto::generator() * self.key.deref()).to_bytes(),
(Ristretto::generator() * nonce.deref()).to_bytes().as_ref(),
msg,
);
let challenge = challenge(
self.genesis,
(Ristretto::generator() * self.key.deref()).to_bytes(),
(Ristretto::generator() * nonce.deref()).to_bytes().as_ref(),
msg,
);
let sig = SchnorrSignature::<Ristretto>::sign(&self.key, nonce, challenge).serialize();
let sig = SchnorrSignature::<Ristretto>::sign(&self.key, nonce, challenge).serialize();
let mut res = [0; 64];
res.copy_from_slice(&sig);
res
let mut res = [0; 64];
res.copy_from_slice(&sig);
res
}
}
}
@@ -274,7 +274,6 @@ pub const BLOCK_PROCESSING_TIME: u32 = 999;
pub const LATENCY_TIME: u32 = 1667;
pub const TARGET_BLOCK_TIME: u32 = BLOCK_PROCESSING_TIME + (3 * LATENCY_TIME);
#[async_trait]
impl<D: Db, T: TransactionTrait, P: P2p> Network for TendermintNetwork<D, T, P> {
type Db = D;
@@ -300,111 +299,126 @@ impl<D: Db, T: TransactionTrait, P: P2p> Network for TendermintNetwork<D, T, P>
self.validators.clone()
}
async fn broadcast(&mut self, msg: SignedMessageFor<Self>) {
let mut to_broadcast = vec![TENDERMINT_MESSAGE];
to_broadcast.extend(msg.encode());
self.p2p.broadcast(self.genesis, to_broadcast).await
}
async fn slash(&mut self, validator: Self::ValidatorId, slash_event: SlashEvent) {
log::error!(
"validator {} triggered a slash event on tributary {} (with evidence: {})",
hex::encode(validator),
hex::encode(self.genesis),
matches!(slash_event, SlashEvent::WithEvidence(_)),
);
let signer = self.signer();
let Some(tx) = (match slash_event {
SlashEvent::WithEvidence(evidence) => {
// create an unsigned evidence tx
Some(TendermintTx::SlashEvidence(evidence))
}
SlashEvent::Id(_reason, _block, _round) => {
// TODO: Increase locally observed slash points
None
}
}) else {
return;
};
// add tx to blockchain and broadcast to peers
let mut to_broadcast = vec![TRANSACTION_MESSAGE];
tx.write(&mut to_broadcast).unwrap();
if self.blockchain.write().await.add_transaction::<Self>(
true,
Transaction::Tendermint(tx),
&self.signature_scheme(),
) == Ok(true)
{
self.p2p.broadcast(signer.genesis, to_broadcast).await;
fn broadcast(&mut self, msg: SignedMessageFor<Self>) -> impl Send + Future<Output = ()> {
async move {
let mut to_broadcast = vec![TENDERMINT_MESSAGE];
to_broadcast.extend(msg.encode());
self.p2p.broadcast(self.genesis, to_broadcast).await
}
}
async fn validate(&self, block: &Self::Block) -> Result<(), TendermintBlockError> {
let block =
Block::read::<&[u8]>(&mut block.0.as_ref()).map_err(|_| TendermintBlockError::Fatal)?;
self
.blockchain
.read()
.await
.verify_block::<Self>(&block, &self.signature_scheme(), false)
.map_err(|e| match e {
BlockError::NonLocalProvided(_) => TendermintBlockError::Temporal,
_ => {
log::warn!("Tributary Tendermint validate returning BlockError::Fatal due to {e:?}");
TendermintBlockError::Fatal
fn slash(
&mut self,
validator: Self::ValidatorId,
slash_event: SlashEvent,
) -> impl Send + Future<Output = ()> {
async move {
log::error!(
"validator {} triggered a slash event on tributary {} (with evidence: {})",
hex::encode(validator),
hex::encode(self.genesis),
matches!(slash_event, SlashEvent::WithEvidence(_)),
);
let signer = self.signer();
let Some(tx) = (match slash_event {
SlashEvent::WithEvidence(evidence) => {
// create an unsigned evidence tx
Some(TendermintTx::SlashEvidence(evidence))
}
})
SlashEvent::Id(_reason, _block, _round) => {
// TODO: Increase locally observed slash points
None
}
}) else {
return;
};
// add tx to blockchain and broadcast to peers
let mut to_broadcast = vec![TRANSACTION_MESSAGE];
tx.write(&mut to_broadcast).unwrap();
if self.blockchain.write().await.add_transaction::<Self>(
true,
Transaction::Tendermint(tx),
&self.signature_scheme(),
) == Ok(true)
{
self.p2p.broadcast(signer.genesis, to_broadcast).await;
}
}
}
async fn add_block(
fn validate(
&self,
block: &Self::Block,
) -> impl Send + Future<Output = Result<(), TendermintBlockError>> {
async move {
let block =
Block::read::<&[u8]>(&mut block.0.as_ref()).map_err(|_| TendermintBlockError::Fatal)?;
self
.blockchain
.read()
.await
.verify_block::<Self>(&block, &self.signature_scheme(), false)
.map_err(|e| match e {
BlockError::NonLocalProvided(_) => TendermintBlockError::Temporal,
_ => {
log::warn!("Tributary Tendermint validate returning BlockError::Fatal due to {e:?}");
TendermintBlockError::Fatal
}
})
}
}
fn add_block(
&mut self,
serialized_block: Self::Block,
commit: Commit<Self::SignatureScheme>,
) -> Option<Self::Block> {
let invalid_block = || {
// There's a fatal flaw in the code, it's behind a hard fork, or the validators turned
// malicious
// All justify a halt to then achieve social consensus from
// TODO: Under multiple validator sets, a small validator set turning malicious knocks
// off the entire network. That's an unacceptable DoS.
panic!("validators added invalid block to tributary {}", hex::encode(self.genesis));
};
) -> impl Send + Future<Output = Option<Self::Block>> {
async move {
let invalid_block = || {
// There's a fatal flaw in the code, it's behind a hard fork, or the validators turned
// malicious
// All justify a halt to then achieve social consensus from
// TODO: Under multiple validator sets, a small validator set turning malicious knocks
// off the entire network. That's an unacceptable DoS.
panic!("validators added invalid block to tributary {}", hex::encode(self.genesis));
};
// Tendermint should only produce valid commits
assert!(self.verify_commit(serialized_block.id(), &commit));
// Tendermint should only produce valid commits
assert!(self.verify_commit(serialized_block.id(), &commit));
let Ok(block) = Block::read::<&[u8]>(&mut serialized_block.0.as_ref()) else {
return invalid_block();
};
let Ok(block) = Block::read::<&[u8]>(&mut serialized_block.0.as_ref()) else {
return invalid_block();
};
let encoded_commit = commit.encode();
loop {
let block_res = self.blockchain.write().await.add_block::<Self>(
&block,
encoded_commit.clone(),
&self.signature_scheme(),
);
match block_res {
Ok(()) => {
// If we successfully added this block, break
break;
let encoded_commit = commit.encode();
loop {
let block_res = self.blockchain.write().await.add_block::<Self>(
&block,
encoded_commit.clone(),
&self.signature_scheme(),
);
match block_res {
Ok(()) => {
// If we successfully added this block, break
break;
}
Err(BlockError::NonLocalProvided(hash)) => {
log::error!(
"missing provided transaction {} which other validators on tributary {} had",
hex::encode(hash),
hex::encode(self.genesis)
);
tokio::time::sleep(core::time::Duration::from_secs(5)).await;
}
_ => return invalid_block(),
}
Err(BlockError::NonLocalProvided(hash)) => {
log::error!(
"missing provided transaction {} which other validators on tributary {} had",
hex::encode(hash),
hex::encode(self.genesis)
);
tokio::time::sleep(core::time::Duration::from_secs(5)).await;
}
_ => return invalid_block(),
}
}
Some(TendermintBlock(
self.blockchain.write().await.build_block::<Self>(&self.signature_scheme()).serialize(),
))
Some(TendermintBlock(
self.blockchain.write().await.build_block::<Self>(&self.signature_scheme()).serialize(),
))
}
}
}

View File

@@ -39,7 +39,7 @@ impl ReadWrite for TendermintTx {
}
impl Transaction for TendermintTx {
fn kind(&self) -> TransactionKind<'_> {
fn kind(&self) -> TransactionKind {
// There's an assert elsewhere in the codebase expecting this behavior
// If we do want to add Provided/Signed TendermintTxs, review the implications carefully
TransactionKind::Unsigned

View File

@@ -60,8 +60,8 @@ impl ReadWrite for NonceTransaction {
}
impl TransactionTrait for NonceTransaction {
fn kind(&self) -> TransactionKind<'_> {
TransactionKind::Signed(vec![], &self.2)
fn kind(&self) -> TransactionKind {
TransactionKind::Signed(vec![], self.2.clone())
}
fn hash(&self) -> [u8; 32] {

View File

@@ -425,7 +425,7 @@ async fn block_tx_ordering() {
}
impl TransactionTrait for SignedTx {
fn kind(&self) -> TransactionKind<'_> {
fn kind(&self) -> TransactionKind {
match self {
SignedTx::Signed(signed) => signed.kind(),
SignedTx::Provided(pro) => pro.kind(),

Some files were not shown because too many files have changed in this diff Show More