fix: Priority fee copy and scoring logic #134

gzalz · 2025-07-30T04:42:28Z

Problem
Current copy_priority_fee_distribution_account and compute_score instruction handlers don't consider all edge cases considering their permissionless nature.

There is no way to distinguish a PFDA that has not been copied vs. a PFDA that does not exist (no PF sharing opt-in from validator)
PFDAs can be copied before they are finalized
Scoring can be invoked before we can guarantee a PFDA has been copied potentially causing erroneous unstake conditions for a validator

Solution

Disallow calling copy_priority_fee_distribution_account until the epoch that account tracks is over
Allow copying PFDAs that do not exist - this is the only way we can distinguish DNE from not yet copied on-chain
Add new DNE merkle_root_upload_authority enum value
Disallow re-copying a specific PFDA to remove any possibility of copying a reclaimed account

IMPORTANT
This diff has been modified to have priority fee related score defaulted to 1.0. The intended logic has been commented out until the time of priority fee scoring launch.

…grity

gzalz · 2025-07-30T04:46:11Z

programs/validator-history/src/state.rs

            epoch,
            priority_fee_commission: commission,
            priority_fee_tips,
-            merkle_root_upload_authority,


This was wrong.

great catch

for the record: this bug existed for the case where the copy_priority_fee_distribution ix is called on a validator as the first instruction before any other validator history ixs are called and there existed no entry yet for this validator

mrmizz · 2025-07-30T16:18:14Z

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

        ],
        bump,
        seeds::program = config.priority_fee_distribution_program.key(),
-        owner = config.priority_fee_distribution_program.key()


I do think this check is redundant (bc we can the seeds and seeds::program constraints). So should we update the docstring above? It's specifically talking about the owner check.

Good call-out. The big logic change here is that now we actually do want to allow a non-existent (still owned by system program) account to be passed in but it must abide by the PDA constraints. I will make sure this is reflected.

Oh even better. Did not quite catch that. Nice work.

Was copying a non-existent account possible before? I think it wouldn't have been because of the owner constraint

mrmizz · 2025-07-30T17:07:56Z

programs/validator-history/src/state.rs

    }
+
+    /// Boolean representing if fields required for scoring have been set
+    pub fn can_score(&self) -> bool {


Should we rename this method to is_finalized?

I do think we should lean into this concept of "finalizing" a validator history entry. If there is outstanding/missing data, is it okay that the steward can score without it?

The following PR will require the CopyStakeInfo instruction to run during the current epoch. While this CopyPriorityFeeDistribution is required to run after the current epoch. This means that history accounts will always be partially updated, up until being finalized.

There's a few other requirements for finalization, kind of baked into the StewardState::compute_score fn like the vote credits being updated in the current epoch and cluster history being updated

Open to putting those all in one method but we should keep scoring logic on the Steward side since validator history is supposed to be just the data store

Dropping this function as discussed offline. The process of "skipping" an epoch while scoring will look different and impact score rather than instruction success/failure.

I jive with this ^

I would actually name it something a little more verbose, has_copied_priority_fees()

In this case can_score() seems to vague, and is_finalized() is closer, but its meaning is lost.

programs/steward/src/score.rs

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

tests/tests/validator_history/test_copy_priority_fee_distribution.rs

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

programs/validator-history/src/state.rs

coachchucksol

This all seems to hit what we need!

Cannot call copy on N epoch
Cannot overwrite data once its been written
Added the DNE
Cannot score without a valid MRUA

Just a couple of comments

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

coachchucksol · 2025-07-30T21:09:06Z

programs/validator-history/src/state.rs

    }
+
+    /// Boolean representing if fields required for scoring have been set
+    pub fn can_score(&self) -> bool {


I jive with this ^

coachchucksol · 2025-07-30T21:11:03Z

programs/validator-history/src/state.rs

    }
+
+    /// Boolean representing if fields required for scoring have been set
+    pub fn can_score(&self) -> bool {


I would actually name it something a little more verbose, has_copied_priority_fees()

In this case can_score() seems to vague, and is_finalized() is closer, but its meaning is lost.

tests/tests/validator_history/test_copy_priority_fee_distribution.rs

coachchucksol

Another thought

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

programs/steward/src/score.rs

coachchucksol

A couple more things

coachchucksol · 2025-07-31T16:27:36Z

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs

+    let distribution_account = PriorityFeeDistributionAccount::try_deserialize(&mut pdfa_data)
+        .unwrap_or(PriorityFeeDistributionAccount {
+            validator_vote_account: Pubkey::default(),
+            merkle_root_upload_authority: Pubkey::default(),


nit: Set as DNE_AUTHORITY not Pubkey::default(),

coachchucksol · 2025-07-31T16:29:58Z

tests/tests/steward/test_scoring.rs

+            &[
+                MerkleRootUploadAuthority::TipRouter,
+                MerkleRootUploadAuthority::TipRouter,
+                MerkleRootUploadAuthority::TipRouter,


Any test cases with MerkleRootUploadAuthority::Other?

priority fees: additional constraints for pf copying and scoring inte…

d8f643c

…grity

gzalz changed the base branch from master to feat/add-priority-fee-support July 30, 2025 04:42

gzalz commented Jul 30, 2025

View reviewed changes

gzalz added 4 commits July 29, 2025 22:48

rm comment

db91ecc

clippy

4b3d4fc

clippy

2f262e9

lock

b986951

gzalz marked this pull request as ready for review July 30, 2025 05:05

mrmizz reviewed Jul 30, 2025

View reviewed changes

coachchucksol reviewed Jul 30, 2025

View reviewed changes

programs/steward/src/score.rs Show resolved Hide resolved

ebatsell reviewed Jul 30, 2025

View reviewed changes

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs Show resolved Hide resolved

ebatsell reviewed Jul 30, 2025

View reviewed changes

tests/tests/validator_history/test_copy_priority_fee_distribution.rs Outdated Show resolved Hide resolved

ebatsell requested changes Jul 30, 2025

View reviewed changes

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs Show resolved Hide resolved

programs/validator-history/src/state.rs Outdated Show resolved Hide resolved

gzalz added 6 commits July 30, 2025 14:26

address PR feedback - move logic to commission calculation

967d464

use max commission var in empty realizesd commission return statement

79a362e

lock

a73d244

s/0 as u64/0u64

88db39e

lock

9a53843

unused dep

0b26fbe

coachchucksol requested changes Jul 30, 2025

View reviewed changes

programs/validator-history/src/instructions/copy_priority_fee_distribution.rs Outdated Show resolved Hide resolved

ebatsell reviewed Jul 30, 2025

View reviewed changes

programs/steward/src/score.rs Outdated Show resolved Hide resolved

ebatsell reviewed Jul 30, 2025

View reviewed changes

programs/steward/src/score.rs Outdated Show resolved Hide resolved

gzalz added 3 commits July 30, 2025 16:42

additional testcases mixing unset dne and tiprouter MRUAs

8aedee0

pr feedback

6c1eec6

pr

4fa7e34

coachchucksol requested changes Jul 31, 2025

View reviewed changes

gzalz added 2 commits July 31, 2025 10:36

latest round

213022b

default pf related score components to 1 until pf score launch

c61b61e

gzalz added 4 commits July 31, 2025 12:48

clippy

2378c11

test changes for default 1.0 pf score components

7344086

relint

a3c31e2

lock

874a7cb

gzalz merged commit 0f997a2 into feat/add-priority-fee-support Jul 31, 2025
6 checks passed

gzalz deleted the gzalz/pf-scoring-fix branch July 31, 2025 21:12

fix: Priority fee copy and scoring logic #134

fix: Priority fee copy and scoring logic #134

Uh oh!

Conversation

gzalz commented Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ebatsell Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mrmizz Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

coachchucksol left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

coachchucksol left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

coachchucksol left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

gzalz commented Jul 30, 2025 •

edited

Loading

ebatsell Jul 30, 2025 •

edited

Loading

mrmizz Jul 30, 2025 •

edited

Loading