-
Notifications
You must be signed in to change notification settings - Fork 220
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat!: network specific domain hashers #5980
feat!: network specific domain hashers #5980
Conversation
a648abf
to
a0d3b96
Compare
b931592
to
f503b87
Compare
@@ -41,8 +42,9 @@ where D: Default | |||
{ | |||
#[allow(clippy::new_ret_no_self)] | |||
pub fn new(label: &'static str) -> ConsensusHasher<D> { | |||
let network = *CURRENT_NETWORK.lock().unwrap(); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would there be any critical consequences if this were to panic?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it would result in an unusable service. Which is actually a pretty good scenario. I'd rather the service crash than do something like mangle or produce bad transactions.
f503b87
to
a44f059
Compare
We need to regenerate the faucets asap, but we won't do it in this PR. So we're temporarily disabling them.
a44f059
to
ca9c9e9
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
utACK
The changes look good, but on further thought, is it straightforward to write a sanity check that hashers for different networks are independent? That is, if you hash the same input data against two different networks, the output should at least be distinct. |
Straight forward enough it shouldn't be skipped. I've written the test now and will push it once I see the CI results and what else needs changing. |
Sounds good, thanks! I'd be surprised if the test uncovered any problems, but at least it's there in case of any future regressions. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I am going to merge this as is, the test I will mark as broken.
I tested this locally and it works.
Description --- Refactors the handling of network-based hashing operations introduced in #5980 to better handle read operations. Adds a sanity test for hash independence. Supersedes #6014. Closes #6003. Motivation and Context --- Recent work in #5980 binds the current network into consensus hashing. It uses a `Mutex`, which has two subtle issues. First, it allows the network to be set multiple times, which should not occur. Second, it locks on reads, which is unnecessary and inefficient. This PR updates how the current network is handled. It adds `Network::get_current_or_default` that will return either the current network (if it has been set) or the default network (if it has not). It adds `Network:set_current` that attempts to set the network; if it has been set before, this operation will fail. Note that calling `Network::get_current_or_default` does _not_ set the network for this purpose. It modifies the API for consensus hashing to add a wrapper that uses the current network. This wraps functionality allowing for specification of a network, which is useful for testing. On top of this new API, a new test is added that checks for distinct hashes using the same input but different networks. This is not comprehensive, but will detect obvious issues. How Has This Been Tested? --- Existing tests pass. A new test passes. What process can a PR reviewer use to test or verify this change? --- Check that the tests do what they claim. Check that the updates to consensus hashing properly introduce the expected wrapping functionality. Check that the updated network API does what it is supposed to.
Description
Add a lazy static value that allows us to access what the current network is globally, so we can use the network byte to create network & domain-specific hashers.
Motivation and Context
This prevents the spending of duplicated UTXO's across networks as the hashes would compute differently based on the network byte.
Closes: #5652
How Has This Been Tested?
CI
Breaking Changes
This will invalidate all previous hashes on the network, and require both a network reset, with full data directory deletion.