feat(causal): timer-heap virtual time — deadlines chase injected delay (RFC 007)

Injected delays dilate virtual time for the workload, but timer deadlines
stayed wall-anchored: a sleep or receive-timeout fired early in virtual
terms, so timeout/retry behaviour sped up relative to the dilated world
(v1 known gap #1).

Every heap Entry now carries delay_stamp — the global delay ledger at
(re-)queue time, cfg-gated on smarm-causal. pop_due converts any debt
accrued since the stamp to wall time (tsc_hz) and shifts the effective
deadline; a not-yet-due entry is re-queued at the shifted deadline with a
fresh stamp, so it keeps chasing delay injected while it waits. seq is
preserved across re-queues, keeping send_after cancellation identity
intact (cancelled entries are discarded before any shift). Zero debt is
byte-identical to the old path; peek_deadline may under-report, costing
one spurious scheduler wake per injected chunk (documented).

This also makes the park-gated resume credit *correct* rather than
forgiving for sleepers: a sleeping actor now physically pays its debt by
sleeping longer, so the on_resume fast-forward reflects real payment
(sleeping_actor_pays_injected_delay pins this end-to-end through the
runtime).

New test hooks: inject_delay_cycles_for_test (deterministic ledger
driver, eagerly TSC-calibrating so conversion never stalls a scheduler
loop) and cycles_to_duration. The ledger is process-global, so the
delta-sensitive causal tests now serialize on a shared test mutex — they
were racy under the parallel test harness before this, in principle.

src/causal.rs

@@ -397,6 +397,20 @@ mod inner {
         end();
     }
 
+    /// Test support: grow the global delay ledger directly, as if target-site
+    /// samples had attributed `cycles` — deterministic driver for the timer
+    /// virtual-time tests. Calibrates the TSC eagerly so conversion later
+    /// never stalls a scheduler loop.
+    pub fn inject_delay_cycles_for_test(cycles: u64) {
+        let _ = tsc_hz();
+        GLOBAL_DELAY.fetch_add(cycles, Ordering::Relaxed);
+    }
+
+    /// Convert ledger cycles to wall time at the measured TSC rate.
+    pub fn cycles_to_duration(cycles: u64) -> Duration {
+        Duration::from_secs_f64(cycles as f64 / tsc_hz())
+    }
+
     /// Controller parameters: which speedups to try per site, and the
     /// experiment/cooldown windows.
     pub struct ExperimentPlan {

src/timer.rs

@@ -102,6 +102,12 @@ pub struct Entry {
     seq: u64,
     pub pid: Pid,
     pub reason: Reason,
+    /// RFC 007 virtual time: the global delay ledger reading when this entry
+    /// was (re-)queued. `pop_due` shifts the effective deadline by any delay
+    /// injected since, so timers dilate together with the causally-delayed
+    /// workload instead of firing early in virtual terms.
+    #[cfg(feature = "smarm-causal")]
+    delay_stamp: u64,
 }
 
 impl PartialEq for Entry {
@@ -166,7 +172,14 @@ impl Timers {
         let seq = self.next_seq;
         self.next_seq = self.next_seq.wrapping_add(1);
         self.armed.insert(seq);
-        self.heap.push(Reverse(Entry { deadline, seq, pid, reason: Reason::Send { fire } }));
+        self.heap.push(Reverse(Entry {
+            deadline,
+            seq,
+            pid,
+            reason: Reason::Send { fire },
+            #[cfg(feature = "smarm-causal")]
+            delay_stamp: crate::causal::global_delay_cycles(),
+        }));
         TimerId(seq)
     }
 
@@ -183,7 +196,14 @@ impl Timers {
     pub fn insert(&mut self, deadline: Instant, pid: Pid, reason: Reason) {
         let seq = self.next_seq;
         self.next_seq = self.next_seq.wrapping_add(1);
-        self.heap.push(Reverse(Entry { deadline, seq, pid, reason }));
+        self.heap.push(Reverse(Entry {
+            deadline,
+            seq,
+            pid,
+            reason,
+            #[cfg(feature = "smarm-causal")]
+            delay_stamp: crate::causal::global_delay_cycles(),
+        }));
     }
 
     pub fn is_empty(&self) -> bool {
@@ -210,23 +230,59 @@ impl Timers {
     /// one is silently dropped here (its `seq` was already removed from
     /// `armed` by [`cancel`](Self::cancel)). Returning it removes it from
     /// `armed`, so a later `cancel` of a fired timer reports `false`.
+    ///
+    /// RFC 007 virtual time (feature `smarm-causal`): before an entry fires,
+    /// any global delay injected since it was (re-)queued is added to its
+    /// deadline; an entry whose *effective* deadline hasn't passed is pushed
+    /// back with the shifted deadline and a fresh stamp, so it keeps chasing
+    /// delay injected while it waits. Consequences, both benign:
+    /// [`peek_deadline`](Self::peek_deadline) may under-report (raw deadline
+    /// earlier than effective), costing at most one spurious scheduler wake
+    /// per injected chunk; and a shift never converts wall time — with zero
+    /// debt the path is byte-identical to the featureless one.
     pub fn pop_due(&mut self, now: Instant) -> Vec<Entry> {
         let mut out = Vec::new();
+        #[cfg(feature = "smarm-causal")]
+        let global = crate::causal::global_delay_cycles();
         while let Some(r) = self.heap.peek() {
-            if r.0.deadline <= now {
-                let entry = match self.heap.pop() {
+            if r.0.deadline > now {
+                break;
+            }
+            #[allow(unused_mut)]
+            let mut entry = match self.heap.pop() {
                 Some(e) => e.0,
                 None => panic!("smarm: timer heap pop after peek returned None (core corrupt)"),
             };
-                if matches!(entry.reason, Reason::Send { .. }) && !self.armed.remove(&entry.seq) {
+            if matches!(entry.reason, Reason::Send { .. }) && !self.armed.contains(&entry.seq) {
                 // Cancelled before it came due: discard, do not deliver.
+                // (Checked before any shift so a cancelled entry is never
+                // re-queued just to be discarded later.)
                 continue;
             }
-                out.push(entry);
-            } else {
-                break;
+            #[cfg(feature = "smarm-causal")]
+            {
+                let debt = global.saturating_sub(entry.delay_stamp);
+                if debt > 0 {
+                    let shifted = entry
+                        .deadline
+                        .checked_add(crate::causal::cycles_to_duration(debt))
+                        .unwrap_or(entry.deadline);
+                    if shifted > now {
+                        // Not due in virtual time: re-queue at the shifted
+                        // deadline, stamped, keeping `seq` (and thus `Send`
+                        // cancellation identity) intact.
+                        entry.deadline = shifted;
+                        entry.delay_stamp = global;
+                        self.heap.push(Reverse(entry));
+                        continue;
                     }
                 }
+            }
+            if matches!(entry.reason, Reason::Send { .. }) {
+                self.armed.remove(&entry.seq);
+            }
+            out.push(entry);
+        }
         out
     }
 }

tests/causal.rs

@@ -73,6 +73,7 @@ fn site_guard_nesting_restores() {
 /// catches up to the global one) instead of running delay-free.
 #[test]
 fn virtual_speedup_ledger() {
+    let _s = serial();
     let global_out = Arc::new(AtomicU64::new(0));
     let absorbed_out = Arc::new(AtomicU64::new(0));
     let g_out = global_out.clone();
@@ -137,6 +138,7 @@ fn virtual_speedup_ledger() {
 /// resume made all bystanders delay-immune, so experiments moved nothing.)
 #[test]
 fn runnable_bystander_pays_delay() {
+    let _s = serial();
     let baseline = Arc::new(AtomicU64::new(0));
     let during = Arc::new(AtomicU64::new(0));
     let b_out = baseline.clone();
@@ -218,6 +220,7 @@ fn runnable_bystander_pays_delay() {
 /// baseline Coz relies on, and it doubles as the overhead sanity check.
 #[test]
 fn zero_speedup_injects_nothing() {
+    let _s = serial();
     smarm::init(smarm::Config::exact(2)).run(|| {
         let stop = Arc::new(AtomicBool::new(false));
         let stop2 = stop.clone();
@@ -299,6 +302,7 @@ fn impact_from_synthetic_results() {
 /// section with experiment / throughput-point lines.
 #[test]
 fn controller_produces_report() {
+    let _s = serial();
     smarm::init(smarm::Config::exact(2)).run(|| {
         let stop = Arc::new(AtomicBool::new(false));
         let stop2 = stop.clone();
@@ -341,6 +345,7 @@ fn controller_produces_report() {
 /// (+83.5% where theory says +100%).
 #[test]
 fn site_boundaries_flush_tail() {
+    let _s = serial();
     fn tsc() -> u64 {
         // SAFETY: rdtsc has no preconditions on x86_64.
         unsafe { core::arch::x86_64::_rdtsc() }
@@ -388,3 +393,152 @@ fn site_boundaries_flush_tail() {
         "boundary flush attributed {injected} of {spent} spent cycles (want ~50%)"
     );
 }
+
+// ---------------------------------------------------------------------------
+// RFC 007 timer virtual time: injected delay dilates virtual time for
+// everyone, so a pending timer's *effective* deadline is its raw deadline
+// plus whatever delay was injected while it was armed. Without this, a
+// sleep or receive-timeout fires "early" in virtual terms and timeout-driven
+// behaviour speeds up relative to the dilated world.
+//
+// These tests (and every ledger-delta test above) share the process-global
+// delay ledger, so they serialize on `serial()`.
+// ---------------------------------------------------------------------------
+
+fn serial() -> std::sync::MutexGuard<'static, ()> {
+    static M: std::sync::Mutex<()> = std::sync::Mutex::new(());
+    M.lock().unwrap_or_else(std::sync::PoisonError::into_inner)
+}
+
+/// A pending timer must not fire at its raw deadline if delay was injected
+/// after it was armed; it fires once wall time covers raw + injected.
+#[test]
+fn timer_deadline_shifts_with_injected_delay() {
+    let _s = serial();
+    use smarm::pid::Pid;
+    use smarm::timer::Timers;
+    use std::time::Instant;
+
+    let hz = smarm::causal::tsc_hz();
+    let mut t = Timers::new();
+    let now = Instant::now();
+    t.insert_sleep(now + Duration::from_millis(50), Pid::new(0, 0), 1);
+
+    // 100ms of debt lands while the timer is pending.
+    smarm::causal::inject_delay_cycles_for_test((hz * 0.100) as u64);
+
+    // Raw deadline passed, effective deadline not: nothing fires, entry kept.
+    assert!(t.pop_due(now + Duration::from_millis(60)).is_empty());
+    assert!(!t.is_empty(), "shifted entry must be re-queued, not dropped");
+
+    // Past raw + injected (with margin) it must fire. Chase in case a
+    // parallel test injected more debt meanwhile.
+    let mut fire_at = now + Duration::from_millis(170);
+    let mut fired = false;
+    for _ in 0..40 {
+        if !t.pop_due(fire_at).is_empty() {
+            fired = true;
+            break;
+        }
+        fire_at += Duration::from_millis(50);
+    }
+    assert!(fired, "timer never fired after covering injected delay");
+}
+
+/// With no debt accrued since insertion, behaviour is byte-identical to
+/// before: due entries fire at their raw deadline.
+#[test]
+fn timer_zero_debt_pops_at_raw_deadline() {
+    let _s = serial();
+    use smarm::pid::Pid;
+    use smarm::timer::Timers;
+    use std::time::Instant;
+
+    let mut t = Timers::new();
+    let now = Instant::now();
+    t.insert_sleep(now - Duration::from_millis(1), Pid::new(0, 0), 1);
+    assert_eq!(t.pop_due(now).len(), 1);
+    assert!(t.is_empty());
+}
+
+/// Re-queueing a shifted entry must preserve its identity: a `send_after`
+/// cancelled *after* being shifted past its raw deadline must still cancel
+/// (return true) and must never deliver.
+#[test]
+fn send_after_cancel_survives_reheap() {
+    let _s = serial();
+    use smarm::pid::Pid;
+    use smarm::timer::Timers;
+    use std::sync::atomic::AtomicBool;
+    use std::time::Instant;
+
+    let hz = smarm::causal::tsc_hz();
+    let mut t = Timers::new();
+    let now = Instant::now();
+    let delivered = Arc::new(AtomicBool::new(false));
+    let d2 = delivered.clone();
+    let id = t.insert_send(
+        now + Duration::from_millis(20),
+        Pid::new(0, 0),
+        Box::new(move || d2.store(true, Ordering::Relaxed)),
+    );
+
+    smarm::causal::inject_delay_cycles_for_test((hz * 0.100) as u64);
+
+    // Raw deadline passed: shifted, not fired.
+    assert!(t.pop_due(now + Duration::from_millis(30)).is_empty());
+    // Cancellation still works on the re-queued entry.
+    assert!(t.cancel(id), "cancel lost track of a re-queued send_after");
+    // Even far past the effective deadline nothing is delivered.
+    for e in t.pop_due(now + Duration::from_secs(3600)) {
+        if let smarm::timer::Reason::Send { fire } = e.reason {
+            fire();
+        }
+    }
+    assert!(
+        !delivered.load(Ordering::Relaxed),
+        "cancelled send_after delivered after re-heap"
+    );
+}
+
+/// End-to-end through the runtime: an actor mid-`sleep` pays delay injected
+/// while it is parked by sleeping *longer* — which is exactly what makes the
+/// park-gated resume-credit fast-forward correct rather than forgiving.
+#[test]
+fn sleeping_actor_pays_injected_delay() {
+    let _s = serial();
+    use std::time::Instant;
+
+    let slept = Arc::new(AtomicU64::new(0));
+    let s_out = slept.clone();
+
+    smarm::init(smarm::Config::exact(1)).run(move || {
+        let hz = smarm::causal::tsc_hz();
+        let s2 = s_out.clone();
+        let sleeper = smarm::spawn(move || {
+            let t0 = Instant::now();
+            smarm::sleep(Duration::from_millis(50));
+            s2.store(t0.elapsed().as_millis() as u64, Ordering::Relaxed);
+        });
+
+        // Let the sleeper arm its timer, then inject 150ms of debt from a
+        // plain OS thread — no slot, so nothing spin-absorbs it here and the
+        // only path by which it can slow the sleeper is the timer shift.
+        smarm::sleep(Duration::from_millis(10));
+        std::thread::spawn(move || {
+            smarm::causal::inject_delay_cycles_for_test((hz * 0.150) as u64);
+        })
+        .join()
+        .unwrap();
+
+        sleeper.join().unwrap();
+    });
+
+    let ms = slept.load(Ordering::Relaxed);
+    // Raw sleep 50ms + ~150ms debt injected mid-sleep => ~200ms effective.
+    // Loose lower bound; the pre-fix behaviour is ~50ms.
+    assert!(
+        ms >= 150,
+        "sleeper paid nothing: slept {ms}ms, expected >= 150ms (raw 50 + injected 150)"
+    );
+}