KVM: arm64: timers: Convert per-vcpu virtual offset to a global value

Having a per-vcpu virtual offset is a pain. It needs to be synchronized
on each update, and expands badly to a setup where different timers can
have different offsets, or have composite offsets (as with NV).

So let's start by replacing the use of the CNTVOFF_EL2 shadow register
(which we want to reclaim for NV anyway), and make the virtual timer
carry a pointer to a VM-wide offset.

This simplifies the code significantly. It also addresses two terrible bugs:

- The use of CNTVOFF_EL2 leads to some nice offset corruption
  when the sysreg gets reset, as reported by Joey.

- The kvm mutex is taken from a vcpu ioctl, which goes against
  the locking rules...

Reported-by: Joey Gouly <[email protected]>
Reviewed-by: Reiji Watanabe <[email protected]>
Signed-off-by: Marc Zyngier <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Tested-by: Joey Gouly <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Oliver Upton <[email protected]>

Author: Marc Zyngier

arch/arm64/include/asm/kvm_host.h

@@ -193,6 +193,9 @@ struct kvm_arch {
 	/* Interrupt controller */
 	struct vgic_dist	vgic;
 
+	/* Timers */
+	struct arch_timer_vm_data timer_data;
+
 	/* Mandated version of PSCI */
 	u32 psci_version;
 

arch/arm64/kvm/arch_timer.c

@@ -84,14 +84,10 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)
 
 static u64 timer_get_offset(struct arch_timer_context *ctxt)
 {
-	struct kvm_vcpu *vcpu = ctxt->vcpu;
+	if (ctxt->offset.vm_offset)
+		return *ctxt->offset.vm_offset;
 
-	switch(arch_timer_ctx_index(ctxt)) {
-	case TIMER_VTIMER:
-		return __vcpu_sys_reg(vcpu, CNTVOFF_EL2);
-	default:
-		return 0;
-	}
+	return 0;
 }
 
 static void timer_set_ctl(struct arch_timer_context *ctxt, u32 ctl)
@@ -128,15 +124,12 @@ static void timer_set_cval(struct arch_timer_context *ctxt, u64 cval)
 
 static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)
 {
-	struct kvm_vcpu *vcpu = ctxt->vcpu;
-
-	switch(arch_timer_ctx_index(ctxt)) {
-	case TIMER_VTIMER:
-		__vcpu_sys_reg(vcpu, CNTVOFF_EL2) = offset;
-		break;
-	default:
+	if (!ctxt->offset.vm_offset) {
 		WARN(offset, "timer %ld\n", arch_timer_ctx_index(ctxt));
+		return;
 	}
+
+	WRITE_ONCE(*ctxt->offset.vm_offset, offset);
 }
 
 u64 kvm_phys_timer_read(void)
@@ -765,36 +758,18 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
-/* Make the updates of cntvoff for all vtimer contexts atomic */
-static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)
-{
-	unsigned long i;
-	struct kvm *kvm = vcpu->kvm;
-	struct kvm_vcpu *tmp;
-
-	mutex_lock(&kvm->lock);
-	kvm_for_each_vcpu(i, tmp, kvm)
-		timer_set_offset(vcpu_vtimer(tmp), cntvoff);
-
-	/*
-	 * When called from the vcpu create path, the CPU being created is not
-	 * included in the loop above, so we just set it here as well.
-	 */
-	timer_set_offset(vcpu_vtimer(vcpu), cntvoff);
-	mutex_unlock(&kvm->lock);
-}
-
 void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
 {
 	struct arch_timer_cpu *timer = vcpu_timer(vcpu);
 	struct arch_timer_context *vtimer = vcpu_vtimer(vcpu);
 	struct arch_timer_context *ptimer = vcpu_ptimer(vcpu);
 
 	vtimer->vcpu = vcpu;
+	vtimer->offset.vm_offset = &vcpu->kvm->arch.timer_data.voffset;
 	ptimer->vcpu = vcpu;
 
 	/* Synchronize cntvoff across all vtimers of a VM. */
-	update_vtimer_cntvoff(vcpu, kvm_phys_timer_read());
+	timer_set_offset(vtimer, kvm_phys_timer_read());
 	timer_set_offset(ptimer, 0);
 
 	hrtimer_init(&timer->bg_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_HARD);
@@ -840,7 +815,7 @@ int kvm_arm_timer_set_reg(struct kvm_vcpu *vcpu, u64 regid, u64 value)
 		break;
 	case KVM_REG_ARM_TIMER_CNT:
 		timer = vcpu_vtimer(vcpu);
-		update_vtimer_cntvoff(vcpu, kvm_phys_timer_read() - value);
+		timer_set_offset(timer, kvm_phys_timer_read() - value);
 		break;
 	case KVM_REG_ARM_TIMER_CVAL:
 		timer = vcpu_vtimer(vcpu);

arch/arm64/kvm/hypercalls.c

@@ -44,7 +44,7 @@ static void kvm_ptp_get_time(struct kvm_vcpu *vcpu, u64 *val)
 	feature = smccc_get_arg1(vcpu);
 	switch (feature) {
 	case KVM_PTP_VIRT_COUNTER:
-		cycles = systime_snapshot.cycles - vcpu_read_sys_reg(vcpu, CNTVOFF_EL2);
+		cycles = systime_snapshot.cycles - vcpu->kvm->arch.timer_data.voffset;
 		break;
 	case KVM_PTP_PHYS_COUNTER:
 		cycles = systime_snapshot.cycles;

include/kvm/arm_arch_timer.h

@@ -23,6 +23,19 @@ enum kvm_arch_timer_regs {
 	TIMER_REG_CTL,
 };
 
+struct arch_timer_offset {
+	/*
+	 * If set, pointer to one of the offsets in the kvm's offset
+	 * structure. If NULL, assume a zero offset.
+	 */
+	u64	*vm_offset;
+};
+
+struct arch_timer_vm_data {
+	/* Offset applied to the virtual timer/counter */
+	u64	voffset;
+};
+
 struct arch_timer_context {
 	struct kvm_vcpu			*vcpu;
 
@@ -32,6 +45,8 @@ struct arch_timer_context {
 	/* Emulated Timer (may be unused) */
 	struct hrtimer			hrtimer;
 
+	/* Offset for this counter/timer */
+	struct arch_timer_offset	offset;
 	/*
 	 * We have multiple paths which can save/restore the timer state onto
 	 * the hardware, so we need some way of keeping track of where the