KVM: PPC: Book 3S: XICS: Implement ICS P/Q states

This patch implements P(Presented)/Q(Queued) states for ICS irqs.

When the interrupt is presented, set P. Present if P was not set.
If P is already set, don't present again, set Q.
When the interrupt is EOI'ed, move Q into P (and clear Q). If it is
set, re-present.

The asserted flag used by LSI is also incorporated into the P bit.

When the irq state is saved, P/Q bits are also saved, they need some
qemu modifications to be recognized and passed around to be restored.
KVM_XICS_PENDING bit set and saved should also indicate
KVM_XICS_PRESENTED bit set and saved. But it is possible some old
code doesn't have/recognize the P bit, so when we restore, we set P
for PENDING bit, too.

The idea and much of the code come from Ben.

Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

4 files changed, 161 insertions, 71 deletions

diff --git a/arch/powerpc/include/uapi/asm/kvm.h b/arch/powerpc/include/uapi/asm/kvm.h
index 3603b6f51b11..e3db3a50127b 100644
--- a/arch/powerpc/include/uapi/asm/kvm.h
+++ b/arch/powerpc/include/uapi/asm/kvm.h
@@ -613,5 +613,7 @@ struct kvm_get_htab_header {
 #define  KVM_XICS_LEVEL_SENSITIVE	(1ULL << 40)
 #define  KVM_XICS_MASKED		(1ULL << 41)
 #define  KVM_XICS_PENDING		(1ULL << 42)
+#define  KVM_XICS_PRESENTED		(1ULL << 43)
+#define  KVM_XICS_QUEUED		(1ULL << 44)
 
 #endif /* __LINUX_KVM_POWERPC_H */
diff --git a/arch/powerpc/kvm/book3s_hv_rm_xics.c b/arch/powerpc/kvm/book3s_hv_rm_xics.c
index 16349c9e86ba..30f82c79de5d 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_xics.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_xics.c
@@ -672,51 +672,39 @@ int kvmppc_rm_h_cppr(struct kvm_vcpu *vcpu, unsigned long cppr)
 	return check_too_hard(xics, icp);
 }
 
-int kvmppc_rm_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
+static int ics_rm_eoi(struct kvm_vcpu *vcpu, u32 irq)
 {
 	struct kvmppc_xics *xics = vcpu->kvm->arch.xics;
 	struct kvmppc_icp *icp = vcpu->arch.icp;
 	struct kvmppc_ics *ics;
 	struct ics_irq_state *state;
-	u32 irq = xirr & 0x00ffffff;
 	u16 src;
-
-	if (!xics || !xics->real_mode)
-		return H_TOO_HARD;
+	u32 pq_old, pq_new;
 
 	/*
-	 * ICP State: EOI
-	 *
-	 * Note: If EOI is incorrectly used by SW to lower the CPPR
-	 * value (ie more favored), we do not check for rejection of
-	 * a pending interrupt, this is a SW error and PAPR sepcifies
-	 * that we don't have to deal with it.
+	 * ICS EOI handling: For LSI, if P bit is still set, we need to
+	 * resend it.
 	 *
-	 * The sending of an EOI to the ICS is handled after the
-	 * CPPR update
-	 *
-	 * ICP State: Down_CPPR which we handle
-	 * in a separate function as it's shared with H_CPPR.
+	 * For MSI, we move Q bit into P (and clear Q). If it is set,
+	 * resend it.
 	 */
-	icp_rm_down_cppr(xics, icp, xirr >> 24);
 
-	/* IPIs have no EOI */
-	if (irq == XICS_IPI)
-		goto bail;
-	/*
-	 * EOI handling: If the interrupt is still asserted, we need to
-	 * resend it. We can take a lockless "peek" at the ICS state here.
-	 *
-	 * "Message" interrupts will never have "asserted" set
-	 */
 	ics = kvmppc_xics_find_ics(xics, irq, &src);
 	if (!ics)
 		goto bail;
+
 	state = &ics->irq_state[src];
 
-	/* Still asserted, resend it */
-	if (state->asserted)
-		icp_rm_deliver_irq(xics, icp, irq);
+	if (state->lsi)
+		pq_new = state->pq_state;
+	else
+		do {
+			pq_old = state->pq_state;
+			pq_new = pq_old >> 1;
+		} while (cmpxchg(&state->pq_state, pq_old, pq_new) != pq_old);
+
+	if (pq_new & PQ_PRESENTED)
+		icp_rm_deliver_irq(xics, NULL, irq);
 
 	if (!hlist_empty(&vcpu->kvm->irq_ack_notifier_list)) {
 		icp->rm_action |= XICS_RM_NOTIFY_EOI;
@@ -737,10 +725,43 @@ int kvmppc_rm_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
 			state->intr_cpu = -1;
 		}
 	}
+
  bail:
 	return check_too_hard(xics, icp);
 }
 
+int kvmppc_rm_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
+{
+	struct kvmppc_xics *xics = vcpu->kvm->arch.xics;
+	struct kvmppc_icp *icp = vcpu->arch.icp;
+	u32 irq = xirr & 0x00ffffff;
+
+	if (!xics || !xics->real_mode)
+		return H_TOO_HARD;
+
+	/*
+	 * ICP State: EOI
+	 *
+	 * Note: If EOI is incorrectly used by SW to lower the CPPR
+	 * value (ie more favored), we do not check for rejection of
+	 * a pending interrupt, this is a SW error and PAPR specifies
+	 * that we don't have to deal with it.
+	 *
+	 * The sending of an EOI to the ICS is handled after the
+	 * CPPR update
+	 *
+	 * ICP State: Down_CPPR which we handle
+	 * in a separate function as it's shared with H_CPPR.
+	 */
+	icp_rm_down_cppr(xics, icp, xirr >> 24);
+
+	/* IPIs have no EOI */
+	if (irq == XICS_IPI)
+		return check_too_hard(xics, icp);
+
+	return ics_rm_eoi(vcpu, irq);
+}
+
 unsigned long eoi_rc;
 
 static void icp_eoi(struct irq_chip *c, u32 hwirq, __be32 xirr, bool *again)
@@ -827,14 +848,33 @@ long kvmppc_deliver_irq_passthru(struct kvm_vcpu *vcpu,
 {
 	struct kvmppc_xics *xics;
 	struct kvmppc_icp *icp;
+	struct kvmppc_ics *ics;
+	struct ics_irq_state *state;
 	u32 irq;
+	u16 src;
+	u32 pq_old, pq_new;
 
 	irq = irq_map->v_hwirq;
 	xics = vcpu->kvm->arch.xics;
 	icp = vcpu->arch.icp;
 
 	kvmppc_rm_handle_irq_desc(irq_map->desc);
-	icp_rm_deliver_irq(xics, icp, irq);
+
+	ics = kvmppc_xics_find_ics(xics, irq, &src);
+	if (!ics)
+		return 2;
+
+	state = &ics->irq_state[src];
+
+	/* only MSIs register bypass producers, so it must be MSI here */
+	do {
+		pq_old = state->pq_state;
+		pq_new = ((pq_old << 1) & 3) | PQ_PRESENTED;
+	} while (cmpxchg(&state->pq_state, pq_old, pq_new) != pq_old);
+
+	/* Test P=1, Q=0, this is the only case where we present */
+	if (pq_new == PQ_PRESENTED)
+		icp_rm_deliver_irq(xics, icp, irq);
 
 	/* EOI the interrupt */
 	icp_eoi(irq_desc_get_chip(irq_map->desc), irq_map->r_hwirq, xirr,
diff --git a/arch/powerpc/kvm/book3s_xics.c b/arch/powerpc/kvm/book3s_xics.c
index cdfb4ed73601..c7620622c846 100644
--- a/arch/powerpc/kvm/book3s_xics.c
+++ b/arch/powerpc/kvm/book3s_xics.c
@@ -75,6 +75,7 @@ static int ics_deliver_irq(struct kvmppc_xics *xics, u32 irq, u32 level)
 	struct ics_irq_state *state;
 	struct kvmppc_ics *ics;
 	u16 src;
+	u32 pq_old, pq_new;
 
 	XICS_DBG("ics deliver %#x (level: %d)\n", irq, level);
 
@@ -87,25 +88,41 @@ static int ics_deliver_irq(struct kvmppc_xics *xics, u32 irq, u32 level)
 	if (!state->exists)
 		return -EINVAL;
 
+	if (level == KVM_INTERRUPT_SET_LEVEL || level == KVM_INTERRUPT_SET)
+		level = 1;
+	else if (level == KVM_INTERRUPT_UNSET)
+		level = 0;
 	/*
-	 * We set state->asserted locklessly. This should be fine as
-	 * we are the only setter, thus concurrent access is undefined
-	 * to begin with.
+	 * Take other values the same as 1, consistent with original code.
+	 * maybe WARN here?
 	 */
-	if ((level == 1 && state->lsi) || level == KVM_INTERRUPT_SET_LEVEL)
-		state->asserted = 1;
-	else if (level == 0 || level == KVM_INTERRUPT_UNSET) {
-		state->asserted = 0;
+
+	if (!state->lsi && level == 0) /* noop for MSI */
 		return 0;
-	}
+
+	do {
+		pq_old = state->pq_state;
+		if (state->lsi) {
+			if (level) {
+				if (pq_old & PQ_PRESENTED)
+					/* Setting already set LSI ... */
+					return 0;
+
+				pq_new = PQ_PRESENTED;
+			} else
+				pq_new = 0;
+		} else
+			pq_new = ((pq_old << 1) & 3) | PQ_PRESENTED;
+	} while (cmpxchg(&state->pq_state, pq_old, pq_new) != pq_old);
+
+	/* Test P=1, Q=0, this is the only case where we present */
+	if (pq_new == PQ_PRESENTED)
+		icp_deliver_irq(xics, NULL, irq);
 
 	/* Record which CPU this arrived on for passed-through interrupts */
 	if (state->host_irq)
 		state->intr_cpu = raw_smp_processor_id();
 
-	/* Attempt delivery */
-	icp_deliver_irq(xics, NULL, irq);
-
 	return 0;
 }
 
@@ -768,14 +785,51 @@ static noinline void kvmppc_h_cppr(struct kvm_vcpu *vcpu, unsigned long cppr)
 		icp_deliver_irq(xics, icp, reject);
 }
 
-static noinline int kvmppc_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
+static int ics_eoi(struct kvm_vcpu *vcpu, u32 irq)
 {
 	struct kvmppc_xics *xics = vcpu->kvm->arch.xics;
 	struct kvmppc_icp *icp = vcpu->arch.icp;
 	struct kvmppc_ics *ics;
 	struct ics_irq_state *state;
-	u32 irq = xirr & 0x00ffffff;
 	u16 src;
+	u32 pq_old, pq_new;
+
+	/*
+	 * ICS EOI handling: For LSI, if P bit is still set, we need to
+	 * resend it.
+	 *
+	 * For MSI, we move Q bit into P (and clear Q). If it is set,
+	 * resend it.
+	 */
+
+	ics = kvmppc_xics_find_ics(xics, irq, &src);
+	if (!ics) {
+		XICS_DBG("ios_eoi: IRQ 0x%06x not found !\n", irq);
+		return H_PARAMETER;
+	}
+	state = &ics->irq_state[src];
+
+	if (state->lsi)
+		pq_new = state->pq_state;
+	else
+		do {
+			pq_old = state->pq_state;
+			pq_new = pq_old >> 1;
+		} while (cmpxchg(&state->pq_state, pq_old, pq_new) != pq_old);
+
+	if (pq_new & PQ_PRESENTED)
+		icp_deliver_irq(xics, icp, irq);
+
+	kvm_notify_acked_irq(vcpu->kvm, 0, irq);
+
+	return H_SUCCESS;
+}
+
+static noinline int kvmppc_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
+{
+	struct kvmppc_xics *xics = vcpu->kvm->arch.xics;
+	struct kvmppc_icp *icp = vcpu->arch.icp;
+	u32 irq = xirr & 0x00ffffff;
 
 	XICS_DBG("h_eoi vcpu %d eoi %#lx\n", vcpu->vcpu_id, xirr);
 
@@ -798,26 +852,8 @@ static noinline int kvmppc_h_eoi(struct kvm_vcpu *vcpu, unsigned long xirr)
 	/* IPIs have no EOI */
 	if (irq == XICS_IPI)
 		return H_SUCCESS;
-	/*
-	 * EOI handling: If the interrupt is still asserted, we need to
-	 * resend it. We can take a lockless "peek" at the ICS state here.
-	 *
-	 * "Message" interrupts will never have "asserted" set
-	 */
-	ics = kvmppc_xics_find_ics(xics, irq, &src);
-	if (!ics) {
-		XICS_DBG("h_eoi: IRQ 0x%06x not found !\n", irq);
-		return H_PARAMETER;
-	}
-	state = &ics->irq_state[src];
-
-	/* Still asserted, resend it */
-	if (state->asserted)
-		icp_deliver_irq(xics, icp, irq);
-
-	kvm_notify_acked_irq(vcpu->kvm, 0, irq);
 
-	return H_SUCCESS;
+	return ics_eoi(vcpu, irq);
 }
 
 int kvmppc_xics_rm_complete(struct kvm_vcpu *vcpu, u32 hcall)
@@ -975,9 +1011,9 @@ static int xics_debug_show(struct seq_file *m, void *private)
 		for (i = 0; i < KVMPPC_XICS_IRQ_PER_ICS; i++) {
 			struct ics_irq_state *irq = &ics->irq_state[i];
 
-			seq_printf(m, "irq 0x%06x: server %#x prio %#x save prio %#x asserted %d resend %d masked pending %d\n",
+			seq_printf(m, "irq 0x%06x: server %#x prio %#x save prio %#x pq_state %d resend %d masked pending %d\n",
 				   irq->number, irq->server, irq->priority,
-				   irq->saved_priority, irq->asserted,
+				   irq->saved_priority, irq->pq_state,
 				   irq->resend, irq->masked_pending);
 
 		}
@@ -1196,10 +1232,17 @@ static int xics_get_source(struct kvmppc_xics *xics, long irq, u64 addr)
 		val |= prio << KVM_XICS_PRIORITY_SHIFT;
 		if (irqp->lsi) {
 			val |= KVM_XICS_LEVEL_SENSITIVE;
-			if (irqp->asserted)
+			if (irqp->pq_state & PQ_PRESENTED)
 				val |= KVM_XICS_PENDING;
 		} else if (irqp->masked_pending || irqp->resend)
 			val |= KVM_XICS_PENDING;
+
+		if (irqp->pq_state & PQ_PRESENTED)
+			val |= KVM_XICS_PRESENTED;
+
+		if (irqp->pq_state & PQ_QUEUED)
+			val |= KVM_XICS_QUEUED;
+
 		ret = 0;
 	}
 	arch_spin_unlock(&ics->lock);
@@ -1251,12 +1294,14 @@ static int xics_set_source(struct kvmppc_xics *xics, long irq, u64 addr)
 	irqp->resend = 0;
 	irqp->masked_pending = 0;
 	irqp->lsi = 0;
-	irqp->asserted = 0;
-	if (val & KVM_XICS_LEVEL_SENSITIVE) {
+	irqp->pq_state = 0;
+	if (val & KVM_XICS_LEVEL_SENSITIVE)
 		irqp->lsi = 1;
-		if (val & KVM_XICS_PENDING)
-			irqp->asserted = 1;
-	}
+	/* If PENDING, set P in case P is not saved because of old code */
+	if (val & KVM_XICS_PRESENTED || val & KVM_XICS_PENDING)
+		irqp->pq_state |= PQ_PRESENTED;
+	if (val & KVM_XICS_QUEUED)
+		irqp->pq_state |= PQ_QUEUED;
 	irqp->exists = 1;
 	arch_spin_unlock(&ics->lock);
 	local_irq_restore(flags);
diff --git a/arch/powerpc/kvm/book3s_xics.h b/arch/powerpc/kvm/book3s_xics.h
index 1d5fac80b706..ec5474cf70c6 100644
--- a/arch/powerpc/kvm/book3s_xics.h
+++ b/arch/powerpc/kvm/book3s_xics.h
@@ -31,16 +31,19 @@
 /* Priority value to use for disabling an interrupt */
 #define MASKED	0xff
 
+#define PQ_PRESENTED	1
+#define PQ_QUEUED	2
+
 /* State for one irq source */
 struct ics_irq_state {
 	u32 number;
 	u32 server;
+	u32 pq_state;
 	u8  priority;
 	u8  saved_priority;
 	u8  resend;
 	u8  masked_pending;
 	u8  lsi;		/* level-sensitive interrupt */
-	u8  asserted; /* Only for LSI */
 	u8  exists;
 	int intr_cpu;
 	u32 host_irq;