* Ensure relay chain block number strictly increases
This is a safeguard for when async backing is enabled on the relay chain and multiple parachain
blocks per relay chain block are legal. We will need to later change this to support then multiple
parachain blocks per relay chain block.
* Make the check configurable
* Fix compilation
* Update pallets/parachain-system/src/lib.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update pallets/parachain-system/src/lib.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Fix test
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* polkadot-collator: Switch to wasm only
This switches the polkadot-collator to run everything in wasm only mode. While we should not that
yet with the relay chain, because it can happen that we run out of memory (very unlikely). On the
relay chain that would be bad, because we only have at max 2 sessions to bring everything back, for
Parachains that isn't such a problem as they would only stall and we could roll out a release that
fixes it. Besides that, Parachain validation on the relay chain happens in Wasm already all the time
and there is the memory usage even higher then on block import.
* cargo fmt
* remove unused var
Co-authored-by: Squirrel <gilescope@gmail.com>
This is a crate that shouldn't be used outside of the Substrate Node. Here the correct crate is the
`polkadot-core-primitives` to get the `Balance` type.
Before this fix the relay chain node name was always a random generated, because we did not had
implemented all the required methods of the `CliConfiguration` trait.
* Return compressed PoV from collation
This enables us to also print the size of the compressed PoV.
* FMT
* Update Polkadot & Substrate
* Fix compilation
* Fixes
* FMT
* pov-recovery: Enable pov-recovery as well on full nodes
Pov recovery was before only enabled for collators. The reason behind this was prevention of spam of
the relay chain for block recovery. However, recent events has shown that this wasn't a good idea.
So, this pr enables pov-recover for normal full nodes as well, but with a much higher delay before
trying to recover a block. This means that full nodes will wait in minimum 2.5 minutes and in
maximum 5 minutes before recovering a block. This should give collators in "normal mode" enough time
to recover a block (they wait in maximum 6 seconds after they have seen a new candidate in the relay
chain) before recovering a block. So, we should hopefully not spam the relay chain.
* FMT
* Fixes
* Fix documentation
* xcmp-queue: Fix handling of encoded blobs
With #701 we tried to fix some infinite loop related to encoded blobs, however that lead actually to
not being able to process encoded blobs at all. The reason for this is that `decode_all` doesn't
consume the given input. The point of this function is that it returns an error if the data couldn't
be decoded or there is still data left. However, this means that the check
`remaining_fragments.len() < last_remaining_fragments.len()` would always fail.
We remove the while loop, because we decode the entire fragment anyway or it fails. Aka, we don't
need to loop here. Next we remove the broken check and we directly reset the
`remaining_fragments` (because `decode_all` doesn't consume anything).
* Restore correct behavior
We need to use a while loop, because there can be multiple `Vec<u8>`s. We also need to use `decode`,
because `decode_all` would otherwise return an error if the input is not empty afterwards.
* Remove unused import
* sketch downward messages
* bring in attempt to mock mqc-head from moonbeam
* just patch individual crates
* fing comma
* add some logs
* Holy shit, we actually imported a block!
* Actually mock the message queue chain
* use relay parent number for `sent_at`
* finish moving MQC to primitives
* more complete mock and better config type
* change name
* fix export
* better map types
* fix dependencies after rebase
* try-rejigging branches because this is an override
* try to re-jig for hrmp mcqs
* fix branches
* actually fix branches better
* even better
* Removestray log lines
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* Nicer handling of default `ParachainSystem` name
* better docs
* Default MockXcm for people who only who don't care to mock xcm.
* cargo fmt
* trailing commas
* Apply suggestions from code review
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* use the variable for hrmp to
* fix deref
* deduplicate MessageQueueChain
* better docs for MessageQueueChain
* Use `Vec<u8>` instead of `&'static [u8]`
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* cargo fmt
* associated changes for using Vec<u8>
* Unused import
* Fix compilation
Co-authored-by: Joshy Orndorff <admin@joshyorndorff.com>
Co-authored-by: Joshy Orndorff <JoshOrndorff@users.noreply.github.com>
* Adds support for returning a custom header from `validate_block`
This adds support for returning a custom header from `validate_block`. Before this, we always
returned the header of the block that was validated (and still do it by default). However, after
this pr it is possible to set a custom header or better custom head data that will be returned
instead from `validate_block`. This can be for example when a chain wants to fork.
* FMT
This was never planned to be a permanent feature. This was mainly added for testing purposes, but
now was copied by everybody. The users should be more specific about the para id and set this
properly in the chain spec.
* Add `relay_parent` to block announce data
This is required to make it more future proof when contextual execution is coming. While this
changes the `BlockAnnounceData` type, it is written in a way to make it backwards compatible.
* FMT
* Adds support for checking the timestamp inherent while validating a block
This adds support for checking the timestamp inherent while validating a
block. This will use the relay chain slot number * relay chain slot
duration to calculate a timestamp. This timestamp is used to check the
timestamp in the timestamp inherent.
* Update polkadot-parachains/rococo-runtime/src/lib.rs
Co-authored-by: Shawn Tabrizi <shawntabrizi@gmail.com>
* Update polkadot-parachains/statemine-runtime/src/lib.rs
Co-authored-by: Shawn Tabrizi <shawntabrizi@gmail.com>
* Update primitives/timestamp/src/lib.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Fix warnings
Co-authored-by: Shawn Tabrizi <shawntabrizi@gmail.com>
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
This is basically a hack that prevents the inclusion of `:code` in the
storage proof. A proper fix requires some changes to the trie crate to
not include nodes that aren't read.
It doesn't make sense to log them as a warning, especially when it could
happen that a parachain id could be reused over the lifetime of a relay
chain.
* Start with a failing integration test & some refactorings
* More work
* Make it "work"
* Add `NullConsensus` for the test
* More refactorings
* Move stuff over to its own crate
* Refactorings
* Integrate it into `service` and make the test working
* Docs and some exit condition
* Use the real import queue
* Fix tests
* Update client/pov-recovery/src/active_candidate_recovery.rs
Co-authored-by: Bernhard Schuster <bernhard@ahoi.io>
* Fetch slot duration from the relay chain
* Docs
* Fixes
Co-authored-by: Bernhard Schuster <bernhard@ahoi.io>
* Introduce `CollectCollationInfo` runtime api
Instead of using well known keys to communicate information about a
collation between the runtime and the collator, we now use a runtime api
for this.
* Fixes bug
* Apply suggestions from code review
Co-authored-by: Sergei Shulepov <sergei@parity.io>
* Doc update
Co-authored-by: Sergei Shulepov <sergei@parity.io>
Before when loading a chain spec from a file it was always using the
rococo parachains runtime genesis config. This leaded to problems when
trying to convert a shell chain spec json file to a raw chain spec.
* Update polkadot
* Migrate all uses of MQC heads to merkle proofs
* Mass rename `relay_parent_storage_root`
* Restore parachain-system tests
* Update polkadot and libp2p swarm for testing
* Collapse match into an if let
Co-authored-by: Bastian Köcher <bkchr@users.noreply.github.com>
* Start with something
* Update Substrate & Polkadot
* Start to make it compile
* Make it compile
* Begin with something
* Yep
* I'm a hacker
* Bring back the builder
* Make it work in some way
* Compile
* Parachains use their own "slot"
* Adds cumulus-pallet-aura
* Wrap AuRa import queue to disable equivocation checking by default
* Pass slot duration
* Check the seal when validating a block
* Adds missing file
* Try to make the seal working
* Fix it
* Some fixes
* Bring in the latest features to cleanup the code
* Update and make it compile
* Improve the import
* Start fixing
* More work
* Fix fix fix
* Make everything compile
* Small cleanups
* Rename and more docs
* Docs
* Fixes fixes fixes
* Update rococo-parachains/src/chain_spec.rs
* Update client/consensus/aura/src/lib.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update client/consensus/aura/src/lib.rs
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update primitives/parachain-inherent/Cargo.toml
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Update primitives/parachain-inherent/Cargo.toml
* Update primitives/parachain-inherent/Cargo.toml
* Update primitives/parachain-inherent/Cargo.toml
Co-authored-by: Sergei Shulepov <sergei@parity.io>
Co-authored-by: André Silva <123550+andresilva@users.noreply.github.com>
* Remove pov hash comparison when waiting on a block to announce
Internally we get notified on the channel for our candidate anyway.
Besides that polkadot will compress the pov which leads to a different
pov hash and thus, would lead to a failing check on the `pov_hash`.
* Fix
* Fix tests
* Approve block announcements of backed blocks
If we receive a block announcement without a statement attached that
matches the latest backed block, it is valid and we need to approve the
block announcement to download the block.
* Fix tests
* Approve block announcement if it comes from the best known block
* Fetch backed block only when required
Currently validation data is shared by using a well known key between
the parachain system pallet and the validate block implementation. This
pr changes this by passing the parachain system directly to the validate
block implementation to make use of it. Besides that, we also store the
validation params in some thread local variable to make it inspectable
by parachain system. This moves the validation of validation data and
validation params to the parachain system pallet directly, instead of
having this hidden inside the validate block implementation.
Fixes: https://github.com/paritytech/cumulus/issues/217
* Move consensus to consensus-common crate
* Move the parachain consensus out of the collator
* Add first relay chain consensus stuff
* Remove some warnings
* Fix more stuff
* Fix collator test
* Change `ParachainConsensus` to take a mutable self
* Make everything compile
* Feedback
