[Master] Fix low hanging issues by esemeniuc · Pull Request #1116 · jito-foundation/jito-solana

esemeniuc · 2026-01-06T21:05:49Z

Bugs:

Remove the duplicate checkout in .github/workflows/cargo.yml.
Avoid storing empty bam url via admin-rpc
Running agave-validator --ledger /solana/ledger set-bam-config will stop bam (should have --bam-url to empty to stop bam), see stoppage at https://jitolabs.grafana.net/goto/ef9gr7lzl9dz4e?orgId=1
Running agave-validator --ledger /solana/ledger set-bam-config --bam-url "" will set url to empty string (not disable! we'd keep trying to connect to "")

Perf:

Rework BAM prevalidation/verify to reuse buffers, pass batch indexes instead of cloning in bam_receive_and_buffer.rs

Correctness:

bam_receive_and_buffer.rs: Blacklisted account rejection reports a TransactionError with DeserializationErrorReason::SanitizeError as the reason. That enum value doesn’t match TransactionErrorReason (it maps to INSUFFICIENT_FUNDS_FOR_FEE), so BAM will see the wrong error code.

esemeniuc · 2026-01-07T02:22:25Z

.github/workflows/cargo.yml


      - uses: actions/checkout@v6
        with:
-          submodules: 'recursive'


esemeniuc · 2026-01-07T02:22:42Z

core/src/banking_stage/transaction_scheduler/bam_receive_and_buffer.rs

        let mut last_metrics_report = Instant::now();
        let mut metrics = BamReceiveAndBufferMetrics::default();
        let mut stats = ReceivingStats::default();
+        let mut prevalidated = Vec::new();


reuse across the containers instead of alloc'ing in the while loop

esemeniuc · 2026-01-07T02:23:17Z

core/src/banking_stage/transaction_scheduler/bam_receive_and_buffer.rs


                Ok((
-                    atomic_txn_batch.clone(),
+                    batch_index,


don't clone

esemeniuc · 2026-01-07T02:25:39Z

core/src/banking_stage/consumer.rs

-                .zip(batch.sanitized_transactions().iter()),
+            processed_transactions.into_iter(),
        );
+        self.seq_not_conflict_batch_reusables.set(reusables);


don't lose the reusables, otherwise we keep the empty containers

esemeniuc · 2026-01-07T02:30:44Z

core/src/banking_stage/consumer.rs

-            processed_transactions
-                .into_iter()
-                .zip(batch.sanitized_transactions().iter()),
+            processed_transactions.into_iter(),


already zipped above

esemeniuc · 2026-01-07T02:32:57Z

core/src/banking_stage/transaction_scheduler/bam_receive_and_buffer.rs

-type PrevalidationResult = Result<(AtomicTxnBatch, bool, u32, u64), (Reason, u32)>;
-type PrevalidationOutput = (Vec<PrevalidationResult>, ReceivingStats);
-
+type PrevalidationResult = Result<(usize, bool, u32, u64), (Reason, u32)>;


avoid clones

esemeniuc · 2026-01-07T02:36:22Z

core/src/banking_stage/consumer.rs


-        for (transaction, transaction_info) in transactions {
-            let account_keys = transaction_info.account_keys();
-            let write_account_locks = account_keys


old code does 3 iterations over all account keys for the txn

esemeniuc · 2026-01-07T02:37:20Z

core/src/banking_stage/consumer.rs

-                .iter()
-                .enumerate()
-                .filter_map(|(index, key)| (!transaction_info.is_writable(index)).then_some(key));
-            let has_contention = write_account_locks.clone().any(|key| {


this is a full iteration over all account keys

esemeniuc · 2026-01-07T02:37:53Z

core/src/banking_stage/consumer.rs

+            transaction_read_locks.clear();
+            let mut has_contention = false;
+            for (key, writable) in
+                TransactionAccountLocksIterator::new(transaction_info).accounts_with_is_writable()


single iteration over all account keys

esemeniuc · 2026-01-07T02:41:29Z

core/src/banking_stage/consume_worker.rs

                processed_results: vec![
                    TransactionResult::NotCommitted(
-                        NotCommittedReason::PohTimeout,
+                        NotCommittedReason::PohTimeout, // Note: ChannelFull, ChannelDisconnected, MaxHeightReached are misreported as a PohTimeout


should fix this

esemeniuc · 2026-01-07T03:43:45Z

validator/src/commands/bam/mod.rs

 fn normalize_bam_url(url_str: &str) -> Result<String, BamUrlError> {
-    // If empty, return empty string to disable BAM
-    if url_str.trim().is_empty() {
-        return Ok(String::new());


this is dangerous

esemeniuc · 2026-01-07T03:49:19Z

validator/src/commands/bam/mod.rs

 /// ```
 pub fn extract_bam_url(matches: &ArgMatches) -> Result<Option<String>, BamUrlError> {
    match matches.value_of("bam_url") {
-        Some(url) => normalize_bam_url(url).map(Some),


since above always returns Ok(""), we will return Ok(Some("")) here

esemeniuc · 2026-01-07T03:50:22Z

validator/src/admin_rpc_service.rs

+            );
        }

        *meta.bam_url.lock().unwrap() = bam_url;


this would cause us to store new empty url here

esemeniuc · 2026-01-07T14:43:39Z

core/src/banking_stage/transaction_scheduler/bam_receive_and_buffer.rs

@@ -807,7 +814,7 @@ impl ReceiveAndBuffer for BamReceiveAndBuffer {

                // If BAM is not enabled, drain the channel


this would multiply num_dropped_without_parsing for each packet dropped

esemeniuc · 2026-01-07T14:59:37Z

validator/src/commands/bam/mod.rs

 }

 pub fn execute(subcommand_matches: &ArgMatches, ledger_path: &Path) -> crate::commands::Result<()> {
+    if !subcommand_matches.is_present("bam_url") {


don't delete bam config if --bam-url not passed to agave-validator set-bam-config

esemeniuc · 2026-01-07T16:31:10Z

core/src/banking_stage/transaction_scheduler/bam_receive_and_buffer.rs

                        jito_protos::proto::bam_types::TransactionError {
                            index: index as u32,
-                            reason: DeserializationErrorReason::SanitizeError as i32,
+                            reason: TransactionErrorReason::SanitizeFailure as i32,


blacklisted account rejection reports a TransactionError with DeserializationErrorReason::SanitizeError as the reason. That enum doesn’t match TransactionErrorReason (it maps to INSUFFICIENT_FUNDS_FOR_FEE), so BAM will see the wrong error code

esemeniuc force-pushed the eric/master-tweak branch from 431b459 to 76bef9f Compare January 6, 2026 21:30

esemeniuc requested a review from dgski January 6, 2026 21:49

esemeniuc changed the title ~~add~~ Reduce allocs Jan 6, 2026

esemeniuc commented Jan 7, 2026

View reviewed changes

.github/workflows/cargo.yml

- uses: actions/checkout@v6

with:

submodules: 'recursive'

Copy link
Copy Markdown

Collaborator Author

esemeniuc Jan 7, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

duplicate?

esemeniuc commented Jan 7, 2026

View reviewed changes

esemeniuc force-pushed the eric/master-tweak branch from 37401ad to b944569 Compare January 7, 2026 02:45

esemeniuc changed the title ~~Reduce allocs~~ [Master] Fix low hanging issues Jan 7, 2026

esemeniuc mentioned this pull request Jan 7, 2026

master: Add Opt-In BAM Compability #1107

Merged

esemeniuc commented Jan 7, 2026

View reviewed changes

esemeniuc marked this pull request as ready for review January 7, 2026 03:51

esemeniuc force-pushed the eric/master-tweak branch 2 times, most recently from 8b07d91 to 2453df2 Compare January 7, 2026 14:15

esemeniuc commented Jan 7, 2026

View reviewed changes

dgski approved these changes Jan 9, 2026

View reviewed changes

squash

fd2be04

esemeniuc force-pushed the eric/master-tweak branch from b144abf to fd2be04 Compare January 9, 2026 17:21

esemeniuc mentioned this pull request Jan 9, 2026

[v3.1] Backport tweaks #1120

Merged

add

d0640a7

esemeniuc merged commit 66b7160 into master_bam Jan 9, 2026
1 of 3 checks passed

esemeniuc deleted the eric/master-tweak branch January 9, 2026 19:08

		@@ -807,7 +814,7 @@ impl ReceiveAndBuffer for BamReceiveAndBuffer {

		// If BAM is not enabled, drain the channel

Conversation

esemeniuc commented Jan 6, 2026 • 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!

esemeniuc Jan 7, 2026 • 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!

esemeniuc Jan 7, 2026 • 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!

esemeniuc Jan 7, 2026 • 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!

esemeniuc Jan 7, 2026 • 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!

esemeniuc Jan 7, 2026 • 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!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

esemeniuc commented Jan 6, 2026 •

edited

Loading

esemeniuc Jan 7, 2026 •

edited

Loading

esemeniuc Jan 7, 2026 •

edited

Loading

esemeniuc Jan 7, 2026 •

edited

Loading

esemeniuc Jan 7, 2026 •

edited

Loading

esemeniuc Jan 7, 2026 •

edited

Loading