* Fixes `storage_hash` caching issue and enables better caching for Cumulus
There was a caching issue with `storage_hash` that resulted in not
reverting cached storage hashes when required. In Cumulus this resulted
in nodes failing to import new blocks after a runtime upgrade, because
they were using the old runtime version.
Besides that, this pr optimizes for the Cumulus use case. In particular
that we always import blocks first as non-best blocks and enact them
later. In current version of the caching that would mean we would always
throw away the complete cache of the latest imported block. Now, we
always update the cache for the first block of a new block height. This
enables us to use the cache if this block will enacted as best block
later. If there is a fork and that is enacted as best, we revert all the
changes to the cache.
* Apply suggestions from code review
Co-authored-by: Arkadiy Paronyan <arkady.paronyan@gmail.com>
* Indentation
* Update client/db/src/storage_cache.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
Co-authored-by: Arkadiy Paronyan <arkady.paronyan@gmail.com>
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* grandpa: always store justification for best finalized block
* grandpa-warp-sync: add latest justification when finished proving
* grandpa-warp-sync: change logic for sending best justification when finished
* grandpa: test storing best justification
* grandpa: reorder variants in WarpSyncFinished
* Make grandpa work
* Introduce `SharedData`
* Add test and fix bugs
* Switch to `SharedData`
* Make grandpa tests working
* More Babe work
* Make it async
* Fix fix
* Use `async_trait` in sc-consensus-slots
This makes the code a little bit easier to read and also expresses that
there can always only be one call at a time to `on_slot`.
* Make grandpa tests compile
* More Babe tests work
* Fix network test
* Start fixing service test
* Finish service-test
* Fix sc-consensus-aura
* Fix fix fix
* More fixes
* Make everything compile *yeah*
* Fix build when we have Rust 1.51
* Update client/consensus/common/src/shared_data.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/common/src/shared_data.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/common/src/shared_data.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/common/src/shared_data.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/common/src/shared_data.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/babe/src/tests.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/babe/src/tests.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Fix warning
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Use `async_trait` in sc-consensus-slots
This makes the code a little bit easier to read and also expresses that
there can always only be one call at a time to `on_slot`.
* slots: remove mutex around BlockImport in SlotWorker
Co-authored-by: André Silva <andrerfosilva@gmail.com>
* Improve logging in network gossip
This adds some more information to the logging output to get a better
understanding when something fails.
* Update client/network-gossip/src/state_machine.rs
Co-authored-by: Peter Goodspeed-Niklaus <coriolinus@users.noreply.github.com>
Co-authored-by: Peter Goodspeed-Niklaus <coriolinus@users.noreply.github.com>
* primitives/runtime: initial changes on supporting multiple Justifications
* primitives/runtime: make Justifications strongly typed
* Encode/decode Justifications
* primitives/runtime: add Justification type
* backend: apply_finality and finalize_block takes a single Justification
* manual-seal: create engine id and let rpc take encoded justification
* backend: skeleton functions for appending justifications
* backend: initial implementation append_justification
Initial implementation of append_justification on the Backend trait, and also remove unused skeleton
functions for append_justificaton on Finaziler trait.
k
* backend: guard against duplicate consensus engine id
* client/db: add check for block finality
* client/api: add append_justification to in_mem db
* client/light: add no-op append_justification
* network: fix decode call for Justification
* network: only send a single Justification in BlockData
* network: minor comment update
* protocol: update field names to distinguish single justification
* client: further field renames to plural
* client: update function names to plural justifications
* client/db: upgrade existing database for new format
* network: remove dependency on grandpa crate
* db: fix check for finalized block
* grandpa: check for multiple grandpa justifications hwne importing
* backend: update Finalizer trait to take multiple Justifications
* db: remove debugging statements in migration code
* manual-seal: update note about engine id
* db: fix check for finalized block
* client: update variable name to reflect it is now plural
* grandpa: fix incorrect empty Justications in test
* primitives: make Justifications opaque to avoid being empty
* network: fix detecting empty Justification
* runtime: doc strings for Justifications functions
* runtime: add into_justifications
* primitives: check for duplicates in when adding to Justifications
* network/test: use real grandpa engine id in test
* client: fix reviewer comments
* primitives: rename Justifications::push to append
* backend: revert changes to Finalizer trait
* backend: revert mark_finalized
* backend: revert changes to finalize_block
* backend: revert finalized_blocks
* db: add a quick early return for performance
* client: minor reviewer comments
* service/test: use local ConsensusEngineId
* network: add link to issue for sending multiple Justifications
* Apply suggestions from code review
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
* Apply suggestions from code review
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
* network: tweaks to review suggestions
* network: revert change to BlockData for backwards compatibility
* Apply suggestion from code review
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
* Apply suggestions from code review
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* primitives: update doc comment for Justifications
* client/db/upgrade: avoid grandpa crate dependency
* consensus: revert to single Justification for import_justification
* primitives: improve justifications docs
* style cleanups
* use and_then
* client: rename JUSTIFICATIONS db column
* network: revert to using FRNK in network-test
Co-authored-by: Pierre Krieger <pierre.krieger1708@gmail.com>
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
Co-authored-by: André Silva <andrerfosilva@gmail.com>
We introduced banning of peers who spam us with the same request (more
than 2 times). However, we missed that it is completely legal to send
the same request multiple times as long as we did not provide any
answer. An example for that is the justification request. This request
is send multiple times until we could fetch the justification from one
of our peers. So, the solution to this problem is to tag requests as
fulfilled and to start counting these fulfilled requests. If the number
is higher than what we allow, the peer should be banned.
The timestamp inherent type was up to now just a simple `u64`. This
worked, but doesn't give you that much guarantees at compile time about
the type. This pr changes that by converting this type to a unit type
wrapper, similar to what we have done for `Slot`.
This is required for some future pr that touches quite a lot of the
inherents stuff :)
Besides this unit wrapper type, this pr also moves the `OnTimestampSet`
trait to `frame_support::traits`.
* Add MMR custom RPC.
* Change RuntimeApi to avoid hardcoding leaf type.
* Properly implement the new RuntimeAPI and wire up RPC.
* Extract Offchain DB as separate execution extension.
* Enable offchain DB access for offchain calls.
* Fix offchain_election tests.
* Skip block initialisation for proof generation.
* Fix integration test setup.
* Fix offchain tests. Not sure how I missed them earlier 🤷.
* Fix long line.
* One more test missing.
* Update mock for multi-phase.
* Address review grumbbles.
* Address review grumbles.
* Fix line width of a comment
* Make changes
* Add serialize/deserialize, copy babe epoch config defaults from node runtime
* Fix line widths and turn default features off for serde
* Remove ser/deser from Epoch, fix node-cli
* Apply suggestions
* Add comment to BABE_GENESIS_EPOCH_CONFIG in bin
* Apply suggestions
* Add a sketchy migration function
* Add a migration test
* Check for PendingEpochConfigChange as well
* Make epoch_config in node-cli
* Move updating EpochConfig out of the if
* Fix executor tests
* Calculate weight for add_epoch_configurations
* Fix babe test
* Apply suggestions from code review
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Add more asserts to tests, remove unused changes to primitives/slots
* Allow setting the migration pallet prefix
* Rename to BabePalletPrefix
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Introduce new concept of "slot portion for proposing"
Currently when building a block we actually give the proposer all of the
time in the slot, while this is wrong. The slot is actually split in at
least two phases proposing and propagation or in the polkadot case into
three phases validating pov's, proposing and propagation. As we don't
want to bring that much polkadot concepts into Substrate, we only
support splitting the slot into proposing and propagation. The portion
can now be passed as parameter to AuRa and BABE to configure this value.
However, this slot portion for propagation doesn't mean that the
proposer can not go over this limit. When we miss slots we still apply
the lenience factor to increase the proposing time, so that we have
enough time to build a heavy block.
Besides all what was said above, this is especially required for
parachains. Parachains have a much more constraint proposing window.
Currently the slot duration is at minimum 12 seconds, but we only have
around 500ms for proposing. So, this slot portion for proposing is
really required to make it working without hacks.
* Offgit feedback
* Cast cast cast
* Decrease the peer reputation on invalid block requests
This pr changes the block request handler to decrease the reputation of
peers when they send the same request multiple times or they send us an
invalid block request.
* Review feedback
* Change log target
* Remove unused code
* AuRa improvements
Hot and fresh AuRa improvements. This pr does the following:
- Move code belonging to the import queue etc to import_queue.rs
- Introduce `ImportQueueParams` and `StartAuraParams` structs to make
it more easier to understand what parameters we pass to AuRa.
- Introduce `CheckForEquivocation` to tell AuRa if it should check for
equivocation on block import. This is required for parachains, because
they are allowed to equivocate when they build two blocks for the same
slot, but for different relay chain parents.
* Update client/consensus/aura/src/import_queue.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Fix compilation
* AAA
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
We were using the wrong syntax and that will be dropped with Rust 2021.
The compiler already starts to hint the wrong syntax with warnings. So,
we fix this here.
* Move AuRa digest from client to primitives
This makes the digest stuff usable from inside the runtime ;)
* Update primitives/runtime/src/generic/digest.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Review feedback
* Make BABE use the new functionality
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
Arkadiy Paronyan