hw/xen: Support HVM_PARAM_CALLBACK_TYPE_GSI callback

The GSI callback (and later PCI_INTX) is a level triggered interrupt. It
is asserted when an event channel is delivered to vCPU0, and is supposed
to be cleared when the vcpu_info->evtchn_upcall_pending field for vCPU0
is cleared again.

Thankfully, Xen does *not* assert the GSI if the guest sets its own
evtchn_upcall_pending field; we only need to assert the GSI when we
have delivered an event for ourselves. So that's the easy part, kind of.

There's a slight complexity in that we need to hold the BQL before we
can call qemu_set_irq(), and we definitely can't do that while holding
our own port_lock (because we'll need to take that from the qemu-side
functions that the PV backend drivers will call). So if we end up
wanting to set the IRQ in a context where we *don't* already hold the
BQL, defer to a BH.

However, we *do* need to poll for the evtchn_upcall_pending flag being
cleared. In an ideal world we would poll that when the EOI happens on
the PIC/IOAPIC. That's how it works in the kernel with the VFIO eventfd
pairs — one is used to trigger the interrupt, and the other works in the
other direction to 'resample' on EOI, and trigger the first eventfd
again if the line is still active.

However, QEMU doesn't seem to do that. Even VFIO level interrupts seem
to be supported by temporarily unmapping the device's BARs from the
guest when an interrupt happens, then trapping *all* MMIO to the device
and sending the 'resample' event on *every* MMIO access until the IRQ
is cleared! Maybe in future we'll plumb the 'resample' concept through
QEMU's irq framework but for now we'll do what Xen itself does: just
check the flag on every vmexit if the upcall GSI is known to be
asserted.

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>

diff --git a/hw/i386/kvm/xen_evtchn.c b/hw/i386/kvm/xen_evtchn.c
index a73db5d2bcd1fb58e3e5d06f81d205f65d494512..e2ecee9a6f8e4279f6bc80ed59d9057bf2b7b9c1 100644 (file)
--- a/hw/i386/kvm/xen_evtchn.c
+++ b/hw/i386/kvm/xen_evtchn.c
@@ -26,6 +26,8 @@
 
 #include "hw/sysbus.h"
 #include "hw/xen/xen.h"
+#include "hw/i386/x86.h"
+#include "hw/irq.h"
 
 #include "xen_evtchn.h"
 #include "xen_overlay.h"
@@ -99,9 +101,12 @@ struct XenEvtchnState {
     uint64_t callback_param;
     bool evtchn_in_kernel;
 
+    QEMUBH *gsi_bh;
+
     QemuMutex port_lock;
     uint32_t nr_ports;
     XenEvtchnPort port_table[EVTCHN_2L_NR_CHANNELS];
+    qemu_irq gsis[GSI_NUM_PINS];
 };
 
 struct XenEvtchnState *xen_evtchn_singleton;
@@ -166,13 +171,42 @@ static const TypeInfo xen_evtchn_info = {
     .class_init    = xen_evtchn_class_init,
 };
 
+static void gsi_assert_bh(void *opaque)
+{
+    struct vcpu_info *vi = kvm_xen_get_vcpu_info_hva(0);
+    if (vi) {
+        xen_evtchn_set_callback_level(!!vi->evtchn_upcall_pending);
+    }
+}
+
 void xen_evtchn_create(void)
 {
     XenEvtchnState *s = XEN_EVTCHN(sysbus_create_simple(TYPE_XEN_EVTCHN,
                                                         -1, NULL));
+    int i;
+
     xen_evtchn_singleton = s;
 
     qemu_mutex_init(&s->port_lock);
+    s->gsi_bh = aio_bh_new(qemu_get_aio_context(), gsi_assert_bh, s);
+
+    for (i = 0; i < GSI_NUM_PINS; i++) {
+        sysbus_init_irq(SYS_BUS_DEVICE(s), &s->gsis[i]);
+    }
+}
+
+void xen_evtchn_connect_gsis(qemu_irq *system_gsis)
+{
+    XenEvtchnState *s = xen_evtchn_singleton;
+    int i;
+
+    if (!s) {
+        return;
+    }
+
+    for (i = 0; i < GSI_NUM_PINS; i++) {
+        sysbus_connect_irq(SYS_BUS_DEVICE(s), i, system_gsis[i]);
+    }
 }
 
 static void xen_evtchn_register_types(void)
@@ -182,6 +216,64 @@ static void xen_evtchn_register_types(void)
 
 type_init(xen_evtchn_register_types)
 
+void xen_evtchn_set_callback_level(int level)
+{
+    XenEvtchnState *s = xen_evtchn_singleton;
+    uint32_t param;
+
+    if (!s) {
+        return;
+    }
+
+    /*
+     * We get to this function in a number of ways:
+     *
+     *  • From I/O context, via PV backend drivers sending a notification to
+     *    the guest.
+     *
+     *  • From guest vCPU context, via loopback interdomain event channels
+     *    (or theoretically even IPIs but guests don't use those with GSI
+     *    delivery because that's pointless. We don't want a malicious guest
+     *    to be able to trigger a deadlock though, so we can't rule it out.)
+     *
+     *  • From guest vCPU context when the HVM_PARAM_CALLBACK_IRQ is being
+     *    configured.
+     *
+     *  • From guest vCPU context in the KVM exit handler, if the upcall
+     *    pending flag has been cleared and the GSI needs to be deasserted.
+     *
+     *  • Maybe in future, in an interrupt ack/eoi notifier when the GSI has
+     *    been acked in the irqchip.
+     *
+     * Whichever context we come from if we aren't already holding the BQL
+     * then e can't take it now, as we may already hold s->port_lock. So
+     * trigger the BH to set the IRQ for us instead of doing it immediately.
+     *
+     * In the HVM_PARAM_CALLBACK_IRQ and KVM exit handler cases, the caller
+     * will deliberately take the BQL because they want the change to take
+     * effect immediately. That just leaves interdomain loopback as the case
+     * which uses the BH.
+     */
+    if (!qemu_mutex_iothread_locked()) {
+        qemu_bh_schedule(s->gsi_bh);
+        return;
+    }
+
+    param = (uint32_t)s->callback_param;
+
+    switch (s->callback_param >> CALLBACK_VIA_TYPE_SHIFT) {
+    case HVM_PARAM_CALLBACK_TYPE_GSI:
+        if (param < GSI_NUM_PINS) {
+            qemu_set_irq(s->gsis[param], level);
+            if (level) {
+                /* Ensure the vCPU polls for deassertion */
+                kvm_xen_set_callback_asserted();
+            }
+        }
+        break;
+    }
+}
+
 int xen_evtchn_set_callback_param(uint64_t param)
 {
     XenEvtchnState *s = xen_evtchn_singleton;
@@ -207,6 +299,11 @@ int xen_evtchn_set_callback_param(uint64_t param)
         }
         break;
     }
+
+    case HVM_PARAM_CALLBACK_TYPE_GSI:
+        ret = 0;
+        break;
+
     default:
         ret = -ENOSYS;
         break;

diff --git a/hw/i386/kvm/xen_evtchn.h b/hw/i386/kvm/xen_evtchn.h
index 670f8b3f7d79e49d5ded5e530d5c2822ad03e408..1f9ffc3f946839e302df5680f858cf6081dbc9c2 100644 (file)
--- a/hw/i386/kvm/xen_evtchn.h
+++ b/hw/i386/kvm/xen_evtchn.h
@@ -12,9 +12,13 @@
 #ifndef QEMU_XEN_EVTCHN_H
 #define QEMU_XEN_EVTCHN_H
 
+#include "hw/sysbus.h"
+
 void xen_evtchn_create(void);
 int xen_evtchn_soft_reset(void);
 int xen_evtchn_set_callback_param(uint64_t param);
+void xen_evtchn_connect_gsis(qemu_irq *system_gsis);
+void xen_evtchn_set_callback_level(int level);
 
 void hmp_xen_event_inject(Monitor *mon, const QDict *qdict);
 void hmp_xen_event_list(Monitor *mon, const QDict *qdict);

diff --git a/hw/i386/pc.c b/hw/i386/pc.c
index 8f668a51385e10106261f9f8e94a2ccef154af73..61a90c9e5ba2cc10429028284f7aaec2d004ad7b 100644 (file)
--- a/hw/i386/pc.c
+++ b/hw/i386/pc.c
@@ -1308,6 +1308,12 @@ void pc_basic_device_init(struct PCMachineState *pcms,
     }
     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
 
+#ifdef CONFIG_XEN_EMU
+    if (xen_mode == XEN_EMULATE) {
+        xen_evtchn_connect_gsis(gsi);
+    }
+#endif
+
     qemu_register_boot_set(pc_boot_set, *rtc_state);
 
     if (!xen_enabled() &&

diff --git a/include/sysemu/kvm_xen.h b/include/sysemu/kvm_xen.h
index b2bcacd761bfada0433a82d61d09310631c8b66e..a32ee58852884898e2cfe50c12557becb96bbc91 100644 (file)
--- a/include/sysemu/kvm_xen.h
+++ b/include/sysemu/kvm_xen.h
@@ -22,6 +22,7 @@
 uint32_t kvm_xen_get_caps(void);
 void *kvm_xen_get_vcpu_info_hva(uint32_t vcpu_id);
 void kvm_xen_inject_vcpu_callback_vector(uint32_t vcpu_id, int type);
+void kvm_xen_set_callback_asserted(void);
 int kvm_xen_set_vcpu_virq(uint32_t vcpu_id, uint16_t virq, uint16_t port);
 
 #define kvm_xen_has_cap(cap) (!!(kvm_xen_get_caps() &           \

diff --git a/target/i386/cpu.h b/target/i386/cpu.h
index dba8732fc6ff59b0affefe3be4af3afbc6e232d9..e8718c31e5de737003b3a5d78f76ad4cec68a998 100644 (file)
--- a/target/i386/cpu.h
+++ b/target/i386/cpu.h
@@ -1797,6 +1797,7 @@ typedef struct CPUArchState {
     uint64_t xen_vcpu_time_info_gpa;
     uint64_t xen_vcpu_runstate_gpa;
     uint8_t xen_vcpu_callback_vector;
+    bool xen_callback_asserted;
     uint16_t xen_virq[XEN_NR_VIRQS];
     uint64_t xen_singleshot_timer_ns;
 #endif

diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
index fa08cb65744c846965e05043673746e84fd19be8..51ddf4bfa2f1fbe5b62e9d26cafc441a42007eec 100644 (file)
--- a/target/i386/kvm/kvm.c
+++ b/target/i386/kvm/kvm.c
@@ -5415,6 +5415,19 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
     char str[256];
     KVMState *state;
 
+#ifdef CONFIG_XEN_EMU
+    /*
+     * If the callback is asserted as a GSI (or PCI INTx) then check if
+     * vcpu_info->evtchn_upcall_pending has been cleared, and deassert
+     * the callback IRQ if so. Ideally we could hook into the PIC/IOAPIC
+     * EOI and only resample then, exactly how the VFIO eventfd pairs
+     * are designed to work for level triggered interrupts.
+     */
+    if (cpu->env.xen_callback_asserted) {
+        kvm_xen_maybe_deassert_callback(cs);
+    }
+#endif
+
     switch (run->exit_reason) {
     case KVM_EXIT_HLT:
         DPRINTF("handle_hlt\n");

diff --git a/target/i386/kvm/xen-emu.c b/target/i386/kvm/xen-emu.c
index bc7426b90f9d04827948604709d65363ad461eb6..27e0555baf0c97bae461d36a926405b7e1f2198c 100644 (file)
--- a/target/i386/kvm/xen-emu.c
+++ b/target/i386/kvm/xen-emu.c
@@ -311,6 +311,31 @@ void *kvm_xen_get_vcpu_info_hva(uint32_t vcpu_id)
     return X86_CPU(cs)->env.xen_vcpu_info_hva;
 }
 
+void kvm_xen_maybe_deassert_callback(CPUState *cs)
+{
+    CPUX86State *env = &X86_CPU(cs)->env;
+    struct vcpu_info *vi = env->xen_vcpu_info_hva;
+    if (!vi) {
+            return;
+    }
+
+    /* If the evtchn_upcall_pending flag is cleared, turn the GSI off. */
+    if (!vi->evtchn_upcall_pending) {
+        qemu_mutex_lock_iothread();
+        xen_evtchn_set_callback_level(0);
+        qemu_mutex_unlock_iothread();
+    }
+}
+
+void kvm_xen_set_callback_asserted(void)
+{
+    CPUState *cs = qemu_get_cpu(0);
+
+    if (cs) {
+        X86_CPU(cs)->env.xen_callback_asserted = true;
+    }
+}
+
 void kvm_xen_inject_vcpu_callback_vector(uint32_t vcpu_id, int type)
 {
     CPUState *cs = qemu_get_cpu(vcpu_id);
@@ -343,6 +368,13 @@ void kvm_xen_inject_vcpu_callback_vector(uint32_t vcpu_id, int type)
          */
         qemu_cpu_kick(cs);
         break;
+
+    case HVM_PARAM_CALLBACK_TYPE_GSI:
+    case HVM_PARAM_CALLBACK_TYPE_PCI_INTX:
+        if (vcpu_id == 0) {
+            xen_evtchn_set_callback_level(1);
+        }
+        break;
     }
 }
 

diff --git a/target/i386/kvm/xen-emu.h b/target/i386/kvm/xen-emu.h
index 452605699abb00746f0595cae5f5b6dc4b2fa390..fe85e0b19580621b78a74e0e95e70da2a39bdae3 100644 (file)
--- a/target/i386/kvm/xen-emu.h
+++ b/target/i386/kvm/xen-emu.h
@@ -28,5 +28,6 @@ int kvm_xen_init_vcpu(CPUState *cs);
 int kvm_xen_handle_exit(X86CPU *cpu, struct kvm_xen_exit *exit);
 int kvm_put_xen_state(CPUState *cs);
 int kvm_get_xen_state(CPUState *cs);
+void kvm_xen_maybe_deassert_callback(CPUState *cs);
 
 #endif /* QEMU_I386_KVM_XEN_EMU_H */