fix: reduce idle CPU wakeups and deferred-work polling

This commit is contained in:
2026-08-15 16:50:36 +02:00
parent f7677ac3f1
commit 788b662d44
10 changed files with 109 additions and 14 deletions
+23
View File
@@ -440,6 +440,29 @@ namespace Drivers::Net::Wifi {
static void ServiceAsync(); static void ServiceAsync();
static void ServiceRecovery(); static void ServiceRecovery();
bool HasDeferredWork() {
if (g_initPending.load(std::memory_order_acquire) &&
!g_initialized && Fs::Vfs::IsDriveRegistered(0)) {
return true;
}
if (!g_iwx.Mmio) return false;
if (g_iwx.WorkPending) return true;
uint64_t now = Timekeeping::GetMilliseconds();
if (g_scanDeadline != 0 && now >= g_scanDeadline) return true;
// The MLME/WPA state machine owns sub-second retransmission timers in
// addition to the overall async deadline. Service it until ServiceAsync
// observes Connected/Failed/Idle and clears this flag.
if (g_asyncConnect) return true;
if (g_iwx.State == IwxFwState::Error && g_initialized &&
!g_recoveryGaveUp &&
(g_lastRecoveryMs == 0 ||
now - g_lastRecoveryMs >= RECOVERY_BACKOFF_MS)) {
return true;
}
return false;
}
void ServiceEvents() { void ServiceEvents() {
if (!g_iwx.Mmio) return; if (!g_iwx.Mmio) return;
if (g_iwx.WorkPending) IwxProcessEvents(); if (g_iwx.WorkPending) IwxProcessEvents();
+4
View File
@@ -21,6 +21,10 @@ namespace Drivers::Net::Wifi {
// Steady-state event pump (RX ring, notifications). Idle-loop callback. // Steady-state event pump (RX ring, notifications). Idle-loop callback.
void ServiceEvents(); void ServiceEvents();
// True when firmware initialization, an RX notification, an expired async
// deadline, or a due recovery attempt needs idle-context servicing.
bool HasDeferredWork();
bool IsInitialized(); bool IsInitialized();
bool IsPresent(); bool IsPresent();
@@ -419,6 +419,14 @@ namespace Drivers::USB::Bluetooth {
// ServiceEvents — steady-state event pump (idle loop) // ServiceEvents — steady-state event pump (idle loop)
// ========================================================================= // =========================================================================
bool HasDeferredWork() {
if (g_initPending.load(std::memory_order_acquire) &&
!g_initialized && Fs::Vfs::IsDriveRegistered(0)) {
return true;
}
return g_initialized && Hci::HasPendingCommands();
}
void ServiceEvents() { void ServiceEvents() {
if (!g_initialized) return; if (!g_initialized) return;
if (Xhci::InPollContext()) return; // never nest under PollEvents if (Xhci::InPollContext()) return; // never nest under PollEvents
@@ -27,6 +27,11 @@ namespace Drivers::USB::Bluetooth {
// (PollEvents/DrainEvents/ProcessPendingCommands all self-serialize). // (PollEvents/DrainEvents/ProcessPendingCommands all self-serialize).
void ServiceEvents(); void ServiceEvents();
// True when boot-deferred initialization or queued HCI control work needs
// an idle-context service pass. USB receive events are signaled separately
// by xHCI and cause the dispatcher to service Bluetooth in the same pass.
bool HasDeferredWork();
// Query adapter state // Query adapter state
bool IsInitialized(); bool IsInitialized();
uint8_t GetSlotId(); uint8_t GetSlotId();
+5
View File
@@ -1719,6 +1719,11 @@ namespace Drivers::USB::Bluetooth::Hci {
s_active.store(false, std::memory_order_release); s_active.store(false, std::memory_order_release);
} }
bool HasPendingCommands() {
return g_pendingTail.load(std::memory_order_acquire) !=
g_pendingHead.load(std::memory_order_acquire);
}
bool WaitSecureSendResult(uint32_t timeoutMs, uint8_t* outResult, uint8_t* outStatus) { bool WaitSecureSendResult(uint32_t timeoutMs, uint8_t* outResult, uint8_t* outStatus) {
uint64_t start = Timekeeping::GetMilliseconds(); uint64_t start = Timekeeping::GetMilliseconds();
while (Timekeeping::GetMilliseconds() - start < timeoutMs) { while (Timekeeping::GetMilliseconds() - start < timeoutMs) {
+1
View File
@@ -331,6 +331,7 @@ namespace Drivers::USB::Bluetooth::Hci {
// real confirmed transfers. Call from top-level (e.g. the connect loop), // real confirmed transfers. Call from top-level (e.g. the connect loop),
// NOT from an event handler -- event handlers only enqueue. // NOT from an event handler -- event handlers only enqueue.
void ProcessPendingCommands(); void ProcessPendingCommands();
bool HasPendingCommands();
// ACL TX flow control: outstanding (un-acked) ACL packets, and the // ACL TX flow control: outstanding (un-acked) ACL packets, and the
// controller's ACL buffer count (Number-Of-Completed-Packets credits). The // controller's ACL buffer count (Number-Of-Completed-Packets credits). The
+4
View File
@@ -218,6 +218,10 @@ namespace Smp {
for (;;) { for (;;) {
// Pick up thermal-governor frequency changes decided by the BSP. // Pick up thermal-governor frequency changes decided by the BSP.
Hal::CpuPower::ApplyPolicyIfChanged(); Hal::CpuPower::ApplyPolicyIfChanged();
// Runnable work sends this AP a reschedule IPI. Keep its periodic
// scheduler tick masked for the entire idle-context pass so long
// firmware waits do not keep generating useless timer interrupts.
Timekeeping::ApicTimerEnterApIdle();
// Any idle core may run bounded USB/NIC bottom halves. Preserve // Any idle core may run bounded USB/NIC bottom halves. Preserve
// the AP's ACPI/MWAIT idle selection after servicing them. // the AP's ACPI/MWAIT idle selection after servicing them.
Timekeeping::ServiceDeferredWork(); Timekeeping::ServiceDeferredWork();
+9
View File
@@ -1284,6 +1284,15 @@ namespace Sched {
uint8_t* oldFpu = (oldSlot >= 0) ? processTable[oldSlot].fpuState : nullptr; uint8_t* oldFpu = (oldSlot >= 0) ? processTable[oldSlot].fpuState : nullptr;
uint8_t* newFpu = processTable[next].fpuState; uint8_t* newFpu = processTable[next].fpuState;
if (oldSlot < 0) {
// AP idle loops mask their local periodic timer. Rearm it before
// dispatching user work so preemption resumes with the process.
// Also pick up a thermal-governor policy epoch that may have
// changed while this CPU remained asleep without timer ticks.
Hal::CpuPower::ApplyPolicyIfChanged();
Timekeeping::ApicTimerLeaveApIdle();
}
LoadUserFsBase(cpu, processTable[next].fsBase); LoadUserFsBase(cpu, processTable[next].fsBase);
// DO NOT release schedLock here! It is held across the context // DO NOT release schedLock here! It is held across the context
+38 -7
View File
@@ -41,8 +41,8 @@ namespace Timekeeping {
static constexpr uint32_t DIVIDE_BY_16 = 0x03; static constexpr uint32_t DIVIDE_BY_16 = 0x03;
// The BSP keeps a 1 ms tick for timekeeping and sleep deadlines. // The BSP keeps a 1 ms tick for timekeeping and sleep deadlines.
// APs use a coarser 10 ms scheduler tick to avoid waking idle cores // Running APs use a 10 ms scheduler tick. Idle APs mask it entirely and
// 1000 times per second with no useful work to do. // rely on reschedule IPIs, avoiding periodic package wakeups.
static constexpr uint32_t BSP_TICK_INTERVAL_MS = 1; static constexpr uint32_t BSP_TICK_INTERVAL_MS = 1;
static constexpr uint32_t BSP_TIMER_HZ = 1000 / BSP_TICK_INTERVAL_MS; static constexpr uint32_t BSP_TIMER_HZ = 1000 / BSP_TICK_INTERVAL_MS;
static constexpr uint32_t AP_TICK_INTERVAL_MS = 10; static constexpr uint32_t AP_TICK_INTERVAL_MS = 10;
@@ -220,8 +220,28 @@ namespace Timekeeping {
// identical. This avoids PIT contention during AP boot. // identical. This avoids PIT contention during AP boot.
if (g_ticksPerMs == 0) return; if (g_ticksPerMs == 0) return;
// Configure a coarser periodic timer on APs. The scheduler still gets // Configure the 10 ms scheduler timer for running APs. Their idle loop
// a 10 ms time slice, but idle APs stop taking 1000 timer interrupts/sec. // masks it after initialization and rearms it when dispatching work.
ProgramTimer(true, AP_TICK_INTERVAL_MS);
}
void ApicTimerEnterApIdle() {
auto* cpu = Smp::GetCurrentCpuData();
if (cpu == nullptr || cpu->cpuIndex == 0 || g_ticksPerMs == 0) return;
uint32_t lvt = Hal::LocalApic::ReadRegister(Hal::LocalApic::REG_TIMER_LVT);
if ((lvt & LVT_MASKED) == 0) {
Hal::LocalApic::WriteRegister(Hal::LocalApic::REG_TIMER_LVT,
lvt | LVT_MASKED);
}
}
void ApicTimerLeaveApIdle() {
auto* cpu = Smp::GetCurrentCpuData();
if (cpu == nullptr || cpu->cpuIndex == 0 || g_ticksPerMs == 0) return;
// Reprogram the initial count as well as unmasking. A deep idle state
// may have stopped the local timer at an arbitrary point in its period.
ProgramTimer(true, AP_TICK_INTERVAL_MS); ProgramTimer(true, AP_TICK_INTERVAL_MS);
} }
@@ -239,19 +259,24 @@ namespace Timekeeping {
bool wasReserved = cpu->reservedForKernelWork; bool wasReserved = cpu->reservedForKernelWork;
cpu->reservedForKernelWork = true; cpu->reservedForKernelWork = true;
// Drain USB hot-plug deferred work from any idle core, not just the BSP. // Drain USB work only when the MSI path has actually queued something.
if (Drivers::USB::Xhci::HasDeferredWork()) { // Bluetooth shares this controller, so service its protocol queues in
// the same pass after xHCI has delivered completion callbacks.
bool usbWork = Drivers::USB::Xhci::HasDeferredWork();
if (usbWork) {
Drivers::USB::Xhci::ProcessDeferredWork(); Drivers::USB::Xhci::ProcessDeferredWork();
} }
// NIC hard IRQs only acknowledge/mask and queue RX work. Dispatching // NIC hard IRQs only acknowledge/mask and queue RX work. Dispatching
// Ethernet/TCP/UDP here keeps process-context IPC mutexes out of IRQs. // Ethernet/TCP/UDP here keeps process-context IPC mutexes out of IRQs.
if (Drivers::Net::E1000::HasDeferredWork())
Drivers::Net::E1000::ProcessDeferredWork(); Drivers::Net::E1000::ProcessDeferredWork();
if (Drivers::Net::E1000E::HasDeferredWork())
Drivers::Net::E1000E::ProcessDeferredWork(); Drivers::Net::E1000E::ProcessDeferredWork();
// HDA completion IRQs only acknowledge/mask. Resampling and DMA-ring // HDA completion IRQs only acknowledge/mask. Resampling and DMA-ring
// refill are far too expensive for hard interrupt context. // refill are far too expensive for hard interrupt context.
if (cpu->cpuIndex == 0) { if (cpu->cpuIndex == 0 && Drivers::Audio::IntelHda::HasDeferredWork()) {
Drivers::Audio::IntelHda::ProcessDeferredWork(); Drivers::Audio::IntelHda::ProcessDeferredWork();
} }
@@ -260,6 +285,9 @@ namespace Timekeeping {
// seconds and used to stall kmain before the first process spawned. // seconds and used to stall kmain before the first process spawned.
// Cheap no-op unless an adapter is waiting; self-claiming, and safe // Cheap no-op unless an adapter is waiting; self-claiming, and safe
// to preempt (the scheduler saves/resumes the idle context). // to preempt (the scheduler saves/resumes the idle context).
bool bluetoothWork = usbWork ||
Drivers::USB::Bluetooth::HasDeferredWork();
if (bluetoothWork)
Drivers::USB::Bluetooth::ServiceDeferredInit(); Drivers::USB::Bluetooth::ServiceDeferredInit();
// Service Bluetooth inbound traffic (the headset's SDP/AVRCP queries, // Service Bluetooth inbound traffic (the headset's SDP/AVRCP queries,
@@ -269,13 +297,16 @@ namespace Timekeeping {
// writes got silence (observed: Bose re-dialing SDP during playback, // writes got silence (observed: Bose re-dialing SDP during playback,
// queries never answered). Self-serializing and a cheap no-op when // queries never answered). Self-serializing and a cheap no-op when
// the adapter is down. // the adapter is down.
if (bluetoothWork)
Drivers::USB::Bluetooth::ServiceEvents(); Drivers::USB::Bluetooth::ServiceEvents();
// Wi-Fi mirrors the Bluetooth split: the firmware load needs the // Wi-Fi mirrors the Bluetooth split: the firmware load needs the
// ramdisk, and the RX/notification ring must be drained outside hard // ramdisk, and the RX/notification ring must be drained outside hard
// interrupt context (the MSI handler only latches a flag). // interrupt context (the MSI handler only latches a flag).
if (Drivers::Net::Wifi::HasDeferredWork()) {
Drivers::Net::Wifi::ServiceDeferredInit(); Drivers::Net::Wifi::ServiceDeferredInit();
Drivers::Net::Wifi::ServiceEvents(); Drivers::Net::Wifi::ServiceEvents();
}
// Thermal policy records transitions during BSP maintenance; print // Thermal policy records transitions during BSP maintenance; print
// them from this explicitly non-interrupt idle path. // them from this explicitly non-interrupt idle path.
+6 -1
View File
@@ -11,9 +11,14 @@ namespace Timekeeping {
// Initialize the APIC timer: calibrate against PIT, start periodic interrupts // Initialize the APIC timer: calibrate against PIT, start periodic interrupts
void ApicTimerInitialize(); void ApicTimerInitialize();
// Initialize the APIC timer on an AP (calibrate + start, no IRQ handler registration) // Initialize the scheduler timer on an AP using the BSP calibration.
void ApicTimerInitializeAP(); void ApicTimerInitializeAP();
// Idle APs are woken for runnable work by the reschedule IPI, so their
// periodic scheduler timer can remain masked until a process is dispatched.
void ApicTimerEnterApIdle();
void ApicTimerLeaveApIdle();
// Reinitialize the APIC timer after S3 resume using the previously // Reinitialize the APIC timer after S3 resume using the previously
// calibrated tick rate. Skips PIT calibration and IRQ registration // calibrated tick rate. Skips PIT calibration and IRQ registration
// (both survive in RAM). Only reprograms the timer hardware registers. // (both survive in RAM). Only reprograms the timer hardware registers.