kvm: x86: add support for KVM_CAP_SPLIT_IRQCHIP

This patch adds support for split IRQ chip mode. When
KVM_CAP_SPLIT_IRQCHIP is enabled:

    1.) The PIC, PIT, and IOAPIC are implemented in userspace while
    the LAPIC is implemented by KVM.

    2.) The software IOAPIC delivers interrupts to the KVM LAPIC via
    kvm_set_irq. Interrupt delivery is configured via the MSI routing
    table, for which routes are reserved in target-i386/kvm.c then
    configured in hw/intc/ioapic.c

    3.) KVM delivers IOAPIC EOIs via a new exit KVM_EXIT_IOAPIC_EOI,
    which is handled in target-i386/kvm.c and relayed to the software
    IOAPIC via ioapic_eoi_broadcast.
Signed-off-by: NMatt Gingell <gingell@google.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

hw/i386/pc.c

@@ -65,6 +65,7 @@
 #include "hw/mem/pc-dimm.h"
 #include "qapi/visitor.h"
 #include "qapi-visit.h"
+#include "qom/cpu.h"
 
 /* debug PC/ISA interrupts */
 //#define DEBUG_IRQ
@@ -1517,7 +1518,7 @@ void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
     qemu_register_boot_set(pc_boot_set, *rtc_state);
 
     if (!xen_enabled()) {
-        if (kvm_irqchip_in_kernel()) {
+        if (kvm_pit_in_kernel()) {
             pit = kvm_pit_init(isa_bus, 0x40);
         } else {
             pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
@@ -1592,7 +1593,7 @@ void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
     SysBusDevice *d;
     unsigned int i;
 
-    if (kvm_irqchip_in_kernel()) {
+    if (kvm_ioapic_in_kernel()) {
         dev = qdev_create(NULL, "kvm-ioapic");
     } else {
         dev = qdev_create(NULL, "ioapic");

hw/i386/pc_piix.c

@@ -182,7 +182,7 @@ static void pc_init1(MachineState *machine,
     }
 
     gsi_state = g_malloc0(sizeof(*gsi_state));
-    if (kvm_irqchip_in_kernel()) {
+    if (kvm_ioapic_in_kernel()) {
         kvm_pc_setup_irq_routing(pci_enabled);
         gsi = qemu_allocate_irqs(kvm_pc_gsi_handler, gsi_state,
                                  GSI_NUM_PINS);
@@ -206,7 +206,7 @@ static void pc_init1(MachineState *machine,
     }
     isa_bus_irqs(isa_bus, gsi);
 
-    if (kvm_irqchip_in_kernel()) {
+    if (kvm_pic_in_kernel()) {
         i8259 = kvm_i8259_init(isa_bus);
     } else if (xen_enabled()) {
         i8259 = xen_interrupt_controller_init();

hw/intc/ioapic.c

@@ -25,6 +25,8 @@
 #include "hw/i386/pc.h"
 #include "hw/i386/ioapic.h"
 #include "hw/i386/ioapic_internal.h"
+#include "include/hw/pci/msi.h"
+#include "sysemu/kvm.h"
 
 //#define DEBUG_IOAPIC
 
@@ -35,6 +37,10 @@
 #define DPRINTF(fmt, ...)
 #endif
 
+#define APIC_DELIVERY_MODE_SHIFT 8
+#define APIC_POLARITY_SHIFT 14
+#define APIC_TRIG_MODE_SHIFT 15
+
 static IOAPICCommonState *ioapics[MAX_IOAPICS];
 
 /* global variable from ioapic_common.c */
@@ -54,6 +60,8 @@ static void ioapic_service(IOAPICCommonState *s)
     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
         mask = 1 << i;
         if (s->irr & mask) {
+            int coalesce = 0;
+
             entry = s->ioredtbl[i];
             if (!(entry & IOAPIC_LVT_MASKED)) {
                 trig_mode = ((entry >> IOAPIC_LVT_TRIGGER_MODE_SHIFT) & 1);
@@ -64,6 +72,7 @@ static void ioapic_service(IOAPICCommonState *s)
                 if (trig_mode == IOAPIC_TRIGGER_EDGE) {
                     s->irr &= ~mask;
                 } else {
+                    coalesce = s->ioredtbl[i] & IOAPIC_LVT_REMOTE_IRR;
                     s->ioredtbl[i] |= IOAPIC_LVT_REMOTE_IRR;
                 }
                 if (delivery_mode == IOAPIC_DM_EXTINT) {
@@ -71,8 +80,23 @@ static void ioapic_service(IOAPICCommonState *s)
                 } else {
                     vector = entry & IOAPIC_VECTOR_MASK;
                 }
-                apic_deliver_irq(dest, dest_mode, delivery_mode,
-                                 vector, trig_mode);
+#ifdef CONFIG_KVM
+                if (kvm_irqchip_is_split()) {
+                    if (trig_mode == IOAPIC_TRIGGER_EDGE) {
+                        kvm_set_irq(kvm_state, i, 1);
+                        kvm_set_irq(kvm_state, i, 0);
+                    } else {
+                        if (!coalesce) {
+                            kvm_set_irq(kvm_state, i, 1);
+                        }
+                    }
+                    continue;
+                }
+#else
+                (void)coalesce;
+#endif
+                apic_deliver_irq(dest, dest_mode, delivery_mode, vector,
+                                 trig_mode);
             }
         }
     }
@@ -116,6 +140,44 @@ static void ioapic_set_irq(void *opaque, int vector, int level)
     }
 }
 
+static void ioapic_update_kvm_routes(IOAPICCommonState *s)
+{
+#ifdef CONFIG_KVM
+    int i;
+
+    if (kvm_irqchip_is_split()) {
+        for (i = 0; i < IOAPIC_NUM_PINS; i++) {
+            uint64_t entry = s->ioredtbl[i];
+            uint8_t trig_mode;
+            uint8_t delivery_mode;
+            uint8_t dest;
+            uint8_t dest_mode;
+            uint64_t pin_polarity;
+            MSIMessage msg;
+
+            trig_mode = ((entry >> IOAPIC_LVT_TRIGGER_MODE_SHIFT) & 1);
+            dest = entry >> IOAPIC_LVT_DEST_SHIFT;
+            dest_mode = (entry >> IOAPIC_LVT_DEST_MODE_SHIFT) & 1;
+            pin_polarity = (entry >> IOAPIC_LVT_TRIGGER_MODE_SHIFT) & 1;
+            delivery_mode =
+                (entry >> IOAPIC_LVT_DELIV_MODE_SHIFT) & IOAPIC_DM_MASK;
+
+            msg.address = APIC_DEFAULT_ADDRESS;
+            msg.address |= dest_mode << 2;
+            msg.address |= dest << 12;
+
+            msg.data = entry & IOAPIC_VECTOR_MASK;
+            msg.data |= delivery_mode << APIC_DELIVERY_MODE_SHIFT;
+            msg.data |= pin_polarity << APIC_POLARITY_SHIFT;
+            msg.data |= trig_mode << APIC_TRIG_MODE_SHIFT;
+
+            kvm_irqchip_update_msi_route(kvm_state, i, msg, NULL);
+        }
+        kvm_irqchip_commit_routes(kvm_state);
+    }
+#endif
+}
+
 void ioapic_eoi_broadcast(int vector)
 {
     IOAPICCommonState *s;
@@ -229,6 +291,8 @@ ioapic_mem_write(void *opaque, hwaddr addr, uint64_t val,
         }
         break;
     }
+
+    ioapic_update_kvm_routes(s);
 }
 
 static const MemoryRegionOps ioapic_io_ops = {

include/hw/i386/pc.h

@@ -20,6 +20,19 @@
 
 #define HPET_INTCAP "hpet-intcap"
 
+#ifdef CONFIG_KVM
+#define kvm_pit_in_kernel() \
+    (kvm_irqchip_in_kernel() && !kvm_irqchip_is_split())
+#define kvm_pic_in_kernel()  \
+    (kvm_irqchip_in_kernel() && !kvm_irqchip_is_split())
+#define kvm_ioapic_in_kernel() \
+    (kvm_irqchip_in_kernel() && !kvm_irqchip_is_split())
+#else
+#define kvm_pit_in_kernel()      0
+#define kvm_pic_in_kernel()      0
+#define kvm_ioapic_in_kernel()   0
+#endif
+
 /**
  * PCMachineState:
  * @acpi_dev: link to ACPI PM device that performs ACPI hotplug handling

include/sysemu/kvm.h

@@ -174,6 +174,7 @@ extern bool kvm_ioeventfd_any_length_allowed;
 #else
 #define kvm_enabled()           (0)
 #define kvm_irqchip_in_kernel() (false)
+#define kvm_irqchip_is_split() (false)
 #define kvm_async_interrupts_enabled() (false)
 #define kvm_halt_in_kernel() (false)
 #define kvm_eventfds_enabled() (false)
@@ -317,6 +318,8 @@ MemTxAttrs kvm_arch_post_run(CPUState *cpu, struct kvm_run *run);
 
 int kvm_arch_handle_exit(CPUState *cpu, struct kvm_run *run);
 
+int kvm_arch_handle_ioapic_eoi(CPUState *cpu, struct kvm_run *run);
+
 int kvm_arch_process_async_events(CPUState *cpu);
 
 int kvm_arch_get_registers(CPUState *cpu);
@@ -484,6 +487,7 @@ void kvm_init_irq_routing(KVMState *s);
 /**
  * kvm_arch_irqchip_create:
  * @KVMState: The KVMState pointer
+ * @MachineState: The MachineState pointer
  *
  * Allow architectures to create an in-kernel irq chip themselves.
  *
@@ -491,7 +495,7 @@ void kvm_init_irq_routing(KVMState *s);
  *            0: irq chip was not created
  *          > 0: irq chip was created
  */
-int kvm_arch_irqchip_create(KVMState *s);
+int kvm_arch_irqchip_create(MachineState *ms, KVMState *s);
 
 /**
  * kvm_set_one_reg - set a register value in KVM via KVM_SET_ONE_REG ioctl

kvm-all.c

@@ -99,6 +99,7 @@ struct KVMState
 
 KVMState *kvm_state;
 bool kvm_kernel_irqchip;
+bool kvm_split_irqchip;
 bool kvm_async_interrupts_allowed;
 bool kvm_halt_in_kernel_allowed;
 bool kvm_eventfds_allowed;
@@ -1430,9 +1431,14 @@ static void kvm_irqchip_create(MachineState *machine, KVMState *s)
 
     /* First probe and see if there's a arch-specific hook to create the
      * in-kernel irqchip for us */
-    ret = kvm_arch_irqchip_create(s);
+    ret = kvm_arch_irqchip_create(machine, s);
     if (ret == 0) {
-        ret = kvm_vm_ioctl(s, KVM_CREATE_IRQCHIP);
+        if (machine_kernel_irqchip_split(machine)) {
+            perror("Split IRQ chip mode not supported.");
+            exit(1);
+        } else {
+            ret = kvm_vm_ioctl(s, KVM_CREATE_IRQCHIP);
+        }
     }
     if (ret < 0) {
         fprintf(stderr, "Create kernel irqchip failed: %s\n", strerror(-ret));

stubs/kvm.c

 #include "qemu-common.h"
 #include "sysemu/kvm.h"
 
-int kvm_arch_irqchip_create(KVMState *s)
+int kvm_arch_irqchip_create(MachineState *ms, KVMState *s)
 {
     return 0;
 }

target-arm/kvm.c

@@ -25,6 +25,7 @@
 #include "internals.h"
 #include "hw/arm/arm.h"
 #include "exec/memattrs.h"
+#include "hw/boards.h"
 
 const KVMCapabilityInfo kvm_arch_required_capabilities[] = {
     KVM_CAP_LAST_INFO
@@ -578,8 +579,13 @@ void kvm_arch_init_irq_routing(KVMState *s)
 {
 }
 
-int kvm_arch_irqchip_create(KVMState *s)
+int kvm_arch_irqchip_create(MachineState *ms, KVMState *s)
 {
+     if (machine_kernel_irqchip_split(ms)) {
+         perror("-machine kernel_irqchip=split is not supported on ARM.");
+         exit(1);
+    }
+
     /* If we can create the VGIC using the newer device control API, we
      * let the device do this when it initializes itself, otherwise we
      * fall back to the old API */

target-i386/cpu.c

@@ -2743,7 +2743,7 @@ static void x86_cpu_apic_create(X86CPU *cpu, Error **errp)
     APICCommonState *apic;
     const char *apic_type = "apic";
 
-    if (kvm_irqchip_in_kernel()) {
+    if (kvm_apic_in_kernel()) {
         apic_type = "kvm-apic";
     } else if (xen_enabled()) {
         apic_type = "xen-apic";

target-i386/kvm.c

@@ -39,6 +39,7 @@
 #include "exec/ioport.h"
 #include "standard-headers/asm-x86/hyperv.h"
 #include "hw/pci/pci.h"
+#include "hw/pci/msi.h"
 #include "migration/migration.h"
 #include "exec/memattrs.h"
 
@@ -2597,7 +2598,7 @@ void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
         }
     }
 
-    if (!kvm_irqchip_in_kernel()) {
+    if (!kvm_pic_in_kernel()) {
         qemu_mutex_lock_iothread();
     }
 
@@ -2615,7 +2616,7 @@ void kvm_arch_pre_run(CPUState *cpu, struct kvm_run *run)
         }
     }
 
-    if (!kvm_irqchip_in_kernel()) {
+    if (!kvm_pic_in_kernel()) {
         /* Try to inject an interrupt if the guest can accept it */
         if (run->ready_for_interrupt_injection &&
             (cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
@@ -3017,6 +3018,10 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
     case KVM_EXIT_HYPERV:
         ret = kvm_hv_handle_exit(cpu, &run->hyperv);
         break;
+    case KVM_EXIT_IOAPIC_EOI:
+        ioapic_eoi_broadcast(run->eoi.vector);
+        ret = 0;
+        break;
     default:
         fprintf(stderr, "KVM: unknown exit reason %d\n", run->exit_reason);
         ret = -1;
@@ -3051,6 +3056,39 @@ void kvm_arch_init_irq_routing(KVMState *s)
      */
     kvm_msi_via_irqfd_allowed = true;
     kvm_gsi_routing_allowed = true;
+
+    if (kvm_irqchip_is_split()) {
+        int i;
+
+        /* If the ioapic is in QEMU and the lapics are in KVM, reserve
+           MSI routes for signaling interrupts to the local apics. */
+        for (i = 0; i < IOAPIC_NUM_PINS; i++) {
+            struct MSIMessage msg = { 0x0, 0x0 };
+            if (kvm_irqchip_add_msi_route(s, msg, NULL) < 0) {
+                error_report("Could not enable split IRQ mode.");
+                exit(1);
+            }
+        }
+    }
+}
+
+int kvm_arch_irqchip_create(MachineState *ms, KVMState *s)
+{
+    int ret;
+    if (machine_kernel_irqchip_split(ms)) {
+        ret = kvm_vm_enable_cap(s, KVM_CAP_SPLIT_IRQCHIP, 0, 24);
+        if (ret) {
+            error_report("Could not enable split irqchip mode: %s\n",
+                         strerror(-ret));
+            exit(1);
+        } else {
+            DPRINTF("Enabled KVM_CAP_SPLIT_IRQCHIP\n");
+            kvm_split_irqchip = true;
+            return 1;
+        }
+    } else {
+        return 0;
+    }
 }
 
 /* Classic KVM device assignment interface. Will remain x86 only. */

target-i386/kvm_i386.h

@@ -13,6 +13,8 @@
 
 #include "sysemu/kvm.h"
 
+#define kvm_apic_in_kernel() (kvm_irqchip_in_kernel())
+
 bool kvm_allows_irq0_override(void);
 bool kvm_has_smm(void);
 void kvm_synchronize_all_tsc(void);