Merge pull request 'Every node records DMARC and TLS results for the aggregate reports' (#47) from fix/front-node-dmarc into main
ci / fork-checks (push) Successful in 1m27s
ci / build (push) Successful in 40m18s

This commit was merged in pull request #47.
This commit is contained in:
2026-09-25 01:46:19 +00:00
8 changed files with 464 additions and 55 deletions
+4 -2
View File
@@ -47,8 +47,10 @@ impl SpawnQueueManager for IpcReceivers {
// inbuxa: upstream started these only when the node's role included
// outboundMta at boot, so turning the role on later did nothing and
// turning it off left them delivering until a restart. They now run
// on every node and follow the role live (see Queue::start and the
// report scheduler). This also drains the queue channel on nodes
// on every node: the queue follows the role live (see Queue::start),
// and the report scheduler records DMARC and TLS results on every
// node, whatever its role (see reporting/scheduler.rs). This also
// drains the queue channel on nodes
// without the role, where every queued message's refresh used to sit
// in a channel nobody read until it filled and queueing blocked.
if !core.storage.registry.is_recovery_mode() {
+43 -23
View File
@@ -2,9 +2,12 @@
* SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
*
* SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
*
* Modified by Coffey Labs in 2026 for INBUXA.
*/
use super::AggregateTimestamp;
use super::shared::{MAX_WRITE_RETRIES, Revisioned, write_retry_pause};
use crate::{
core::Session,
queue::RecipientDomain,
@@ -349,29 +352,43 @@ impl DmarcReporting for Server {
let object_id = ObjectType::DmarcInternalReport.to_id();
let key = ValueClass::Registry(RegistryClass::Item { object_id, item_id });
let Some(report) = self
.store()
.get_value::<DmarcInternalReport>(ValueKey::from(key.clone()))
.await
.caused_by(trc::location!())?
else {
return Ok(());
};
// Delete report. inbuxa: only the version read here, so a record
// another node appends meanwhile is sent with it rather than lost
let mut attempt = 0;
let report = loop {
let Some(Revisioned {
revision,
value: report,
}) = self
.store()
.get_value::<Revisioned<DmarcInternalReport>>(ValueKey::from(key.clone()))
.await
.caused_by(trc::location!())?
else {
return Ok(());
};
// Delete report
let mut batch = BatchBuilder::new();
batch.clear(key).clear(RegistryClass::PrimaryKey {
object_id: object_id.into(),
index_id: Property::Domain.to_id(),
key: KeySerializer::new(report.domain.len() + U64_LEN)
.write(&report.domain)
.write(report.policy_identifier)
.finalize(),
});
self.store()
.write(batch.build_all())
.await
.caused_by(trc::location!())?;
let mut batch = BatchBuilder::new();
batch
.assert_value(key.clone(), AssertValue::Hash(revision))
.clear(key.clone())
.clear(RegistryClass::PrimaryKey {
object_id: object_id.into(),
index_id: Property::Domain.to_id(),
key: KeySerializer::new(report.domain.len() + U64_LEN)
.write(&report.domain)
.write(report.policy_identifier)
.finalize(),
});
match self.store().write(batch.build_all()).await {
Ok(_) => break report,
Err(err) if err.is_assertion_failure() && attempt < MAX_WRITE_RETRIES => {
attempt += 1;
write_retry_pause(attempt).await;
}
Err(err) => return Err(err.caused_by(trc::location!())),
}
};
let span_id = self.inner.data.span_id_gen.generate();
let event_from = report.report.date_range_begin.timestamp() as u64;
@@ -676,8 +693,11 @@ impl DmarcReporting for Server {
break;
}
Err(err) => {
if err.is_assertion_failure() && rety_count < 3 {
// inbuxa: another node appended first; try again
// after a short pause
if err.is_assertion_failure() && rety_count < MAX_WRITE_RETRIES {
rety_count += 1;
write_retry_pause(rety_count).await;
continue;
}
trc::error!(
+3
View File
@@ -2,6 +2,8 @@
* SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
*
* SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
*
* Modified by Coffey Labs in 2026 for INBUXA.
*/
use common::config::smtp::report::AggregateFrequency;
@@ -15,6 +17,7 @@ pub mod inbound;
pub mod index;
pub mod scheduler;
pub mod send;
pub mod shared; // inbuxa: reports written by every node
pub mod spf;
pub mod tls;
+11 -8
View File
@@ -20,14 +20,17 @@ impl SpawnReport for mpsc::Receiver<ReportingEvent> {
tokio::spawn(async move {
while let Some(event) = self.recv().await {
let server = inner.build_server();
// inbuxa: reports are the outbound MTA's business, as at
// boot, but the role is read per event so a change applies
// without a restart. Events that arrive while the role is
// off are dropped, as they were on a node started without it
if !matches!(event, ReportingEvent::Stop) && !server.core.network.roles.outbound_mta
{
continue;
}
// inbuxa: every node records what it received, whatever its
// role. An aggregate report covers all of a domain's mail,
// whichever node took it, and recording is a store write
// that nodes already share: the report's primary key is
// versioned, so concurrent appends from several nodes retry
// rather than overwrite. Only building and sending the
// report (the DmarcReport and TlsReport tasks) belongs to
// the outbound MTA; the task manager keeps those to nodes
// with that role. Upstream ran this only on outbound MTA
// nodes, so mail received anywhere else never reached a
// report.
match event {
ReportingEvent::Dmarc(event) => server.schedule_dmarc(event).await,
ReportingEvent::Tls(event) => server.schedule_tls(event).await,
+45
View File
@@ -0,0 +1,45 @@
/*
* SPDX-FileCopyrightText: 2026 Coffey Labs
*
* SPDX-License-Identifier: AGPL-3.0-only
*/
//! inbuxa: internal DMARC and TLS reports are shared by every node. Any node
//! that receives mail appends to them, so several nodes can write one report
//! at once, and the node that sends it may do so while another is appending.
//! Appends already guard the report's versioned primary key and retry when
//! another writer got there first; these helpers give those retries room and
//! let the sender delete exactly the report it read.
use rand::RngExt;
use std::time::Duration;
use store::{Deserialize, xxhash_rust::xxh3::xxh3_64};
/// How many times a report write that lost to another writer is retried.
/// Upstream retried three times, when only outbound MTA nodes wrote.
pub(crate) const MAX_WRITE_RETRIES: u32 = 10;
/// A short random pause, longer on each attempt, before retrying a report
/// write that lost to another node, so the writers spread out instead of
/// colliding again.
pub(crate) async fn write_retry_pause(attempt: u32) {
let ms = rand::rng().random_range(5..=25u64) * u64::from(attempt.max(1));
tokio::time::sleep(Duration::from_millis(ms)).await;
}
/// A stored value with the hash of the bytes it was read from, for
/// `AssertValue::Hash`: a write asserting it fails if anyone changed the
/// value since.
pub(crate) struct Revisioned<T> {
pub revision: u64,
pub value: T,
}
impl<T: Deserialize> Deserialize for Revisioned<T> {
fn deserialize(bytes: &[u8]) -> trc::Result<Self> {
Ok(Revisioned {
revision: xxh3_64(bytes),
value: T::deserialize(bytes)?,
})
}
}
+40 -22
View File
@@ -2,9 +2,12 @@
* SPDX-FileCopyrightText: 2020 Stalwart Labs LLC <[email protected]>
*
* SPDX-License-Identifier: AGPL-3.0-only OR LicenseRef-SEL
*
* Modified by Coffey Labs in 2026 for INBUXA.
*/
use super::AggregateTimestamp;
use super::shared::{MAX_WRITE_RETRIES, Revisioned, write_retry_pause};
use crate::{
queue::RecipientDomain,
reporting::{index::InternalReportIndex, send::MtaReportSend},
@@ -70,28 +73,40 @@ impl TlsReporting for Server {
let object_id = ObjectType::TlsInternalReport.to_id();
let key = ValueClass::Registry(RegistryClass::Item { object_id, item_id });
let Some(report) = self
.store()
.get_value::<TlsInternalReport>(ValueKey::from(key.clone()))
.await
.caused_by(trc::location!())?
else {
return Ok(());
};
// Delete report. inbuxa: only the version read here, so a result
// another node appends meanwhile is sent with it rather than lost
let mut attempt = 0;
let report = loop {
let Some(Revisioned {
revision,
value: report,
}) = self
.store()
.get_value::<Revisioned<TlsInternalReport>>(ValueKey::from(key.clone()))
.await
.caused_by(trc::location!())?
else {
return Ok(());
};
// Delete report
let mut batch = BatchBuilder::new();
batch.clear(key).clear(RegistryClass::PrimaryKey {
object_id: object_id.into(),
index_id: Property::Domain.to_id(),
key: report.domain.as_bytes().to_vec(),
});
self.core
.storage
.data
.write(batch.build_all())
.await
.caused_by(trc::location!())?;
let mut batch = BatchBuilder::new();
batch
.assert_value(key.clone(), AssertValue::Hash(revision))
.clear(key.clone())
.clear(RegistryClass::PrimaryKey {
object_id: object_id.into(),
index_id: Property::Domain.to_id(),
key: report.domain.as_bytes().to_vec(),
});
match self.core.storage.data.write(batch.build_all()).await {
Ok(_) => break report,
Err(err) if err.is_assertion_failure() && attempt < MAX_WRITE_RETRIES => {
attempt += 1;
write_retry_pause(attempt).await;
}
Err(err) => return Err(err.caused_by(trc::location!())),
}
};
let domain_name = report.domain.as_str();
let event_from = report.report.date_range_start.timestamp() as u64;
@@ -477,8 +492,11 @@ impl TlsReporting for Server {
break;
}
Err(err) => {
if err.is_assertion_failure() && rety_count < 3 {
// inbuxa: another node appended first; try again
// after a short pause
if err.is_assertion_failure() && rety_count < MAX_WRITE_RETRIES {
rety_count += 1;
write_retry_pause(rety_count).await;
continue;
}
trc::error!(