Re: [PATCH 12/14] KVM: arm/arm64: arch_timer: Assign the phys timer on VHE systems

[Christoffer Dall <christoffer.dall@arm.com>]

On Mon, Feb 18, 2019 at 03:10:49PM +0000, André Przywara wrote:
> On Thu, 24 Jan 2019 15:00:30 +0100
> Christoffer Dall <christoffer.dall@arm.com> wrote:
> 
> Hi,
> 
> > VHE systems don't have to emulate the physical timer, we can simply
> > assigne the EL1 physical timer directly to the VM as the host always
> > uses the EL2 timers.
> > 
> > In order to minimize the amount of cruft, AArch32 gets definitions for
> > the physical timer too, but is should be generally unused on this
> > architecture.
> > 
> > Co-written with Marc Zyngier <marc.zyngier@arm.com>
> > 
> > Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
> > Signed-off-by: Christoffer Dall <christoffer.dall@arm.com>
> > ---
> >  arch/arm/include/asm/kvm_hyp.h |   4 +
> >  include/kvm/arm_arch_timer.h   |   6 +
> >  virt/kvm/arm/arch_timer.c      | 206 ++++++++++++++++++++++++++-------
> >  3 files changed, 171 insertions(+), 45 deletions(-)
> > 
> > diff --git a/arch/arm/include/asm/kvm_hyp.h b/arch/arm/include/asm/kvm_hyp.h
> > index e93a0cac9add..87bcd18df8d5 100644
> > --- a/arch/arm/include/asm/kvm_hyp.h
> > +++ b/arch/arm/include/asm/kvm_hyp.h
> > @@ -40,6 +40,7 @@
> >  #define TTBR1		__ACCESS_CP15_64(1, c2)
> >  #define VTTBR		__ACCESS_CP15_64(6, c2)
> >  #define PAR		__ACCESS_CP15_64(0, c7)
> > +#define CNTP_CVAL	__ACCESS_CP15_64(2, c14)
> >  #define CNTV_CVAL	__ACCESS_CP15_64(3, c14)
> >  #define CNTVOFF		__ACCESS_CP15_64(4, c14)
> >  
> > @@ -85,6 +86,7 @@
> >  #define TID_PRIV	__ACCESS_CP15(c13, 0, c0, 4)
> >  #define HTPIDR		__ACCESS_CP15(c13, 4, c0, 2)
> >  #define CNTKCTL		__ACCESS_CP15(c14, 0, c1, 0)
> > +#define CNTP_CTL	__ACCESS_CP15(c14, 0, c2, 1)
> >  #define CNTV_CTL	__ACCESS_CP15(c14, 0, c3, 1)
> >  #define CNTHCTL		__ACCESS_CP15(c14, 4, c1, 0)
> >  
> > @@ -94,6 +96,8 @@
> >  #define read_sysreg_el0(r)		read_sysreg(r##_el0)
> >  #define write_sysreg_el0(v, r)		write_sysreg(v, r##_el0)
> >  
> > +#define cntp_ctl_el0			CNTP_CTL
> > +#define cntp_cval_el0			CNTP_CVAL
> >  #define cntv_ctl_el0			CNTV_CTL
> >  #define cntv_cval_el0			CNTV_CVAL
> >  #define cntvoff_el2			CNTVOFF
> > diff --git a/include/kvm/arm_arch_timer.h b/include/kvm/arm_arch_timer.h
> > index d40fe57a2d0d..722e0481f310 100644
> > --- a/include/kvm/arm_arch_timer.h
> > +++ b/include/kvm/arm_arch_timer.h
> > @@ -50,6 +50,10 @@ struct arch_timer_context {
> >  
> >  	/* Emulated Timer (may be unused) */
> >  	struct hrtimer			hrtimer;
> > +
> > +	/* Duplicated state from arch_timer.c for convenience */
> > +	u32				host_timer_irq;
> > +	u32				host_timer_irq_flags;
> >  };
> >  
> >  enum loaded_timer_state {
> > @@ -107,6 +111,8 @@ bool kvm_arch_timer_get_input_level(int vintid);
> >  #define vcpu_vtimer(v)	(&(v)->arch.timer_cpu.timers[TIMER_VTIMER])
> >  #define vcpu_ptimer(v)	(&(v)->arch.timer_cpu.timers[TIMER_PTIMER])
> >  
> > +#define arch_timer_ctx_index(ctx)	((ctx) - vcpu_timer((ctx)->vcpu)->timers)
> > +
> >  u64 kvm_arm_timer_read_sysreg(struct kvm_vcpu *vcpu,
> >  			      enum kvm_arch_timers tmr,
> >  			      enum kvm_arch_timer_regs treg);
> > diff --git a/virt/kvm/arm/arch_timer.c b/virt/kvm/arm/arch_timer.c
> > index 8b0eca5fbad1..eed8f48fbf9b 100644
> > --- a/virt/kvm/arm/arch_timer.c
> > +++ b/virt/kvm/arm/arch_timer.c
> > @@ -35,7 +35,9 @@
> >  
> >  static struct timecounter *timecounter;
> >  static unsigned int host_vtimer_irq;
> > +static unsigned int host_ptimer_irq;
> >  static u32 host_vtimer_irq_flags;
> > +static u32 host_ptimer_irq_flags;
> >  
> >  static DEFINE_STATIC_KEY_FALSE(has_gic_active_state);
> >  
> > @@ -86,20 +88,24 @@ static void soft_timer_cancel(struct hrtimer *hrt)
> >  static irqreturn_t kvm_arch_timer_handler(int irq, void *dev_id)
> >  {
> >  	struct kvm_vcpu *vcpu = *(struct kvm_vcpu **)dev_id;
> > -	struct arch_timer_context *vtimer;
> > +	struct arch_timer_context *ctx;
> >  
> >  	/*
> >  	 * We may see a timer interrupt after vcpu_put() has been called which
> >  	 * sets the CPU's vcpu pointer to NULL, because even though the timer
> > -	 * has been disabled in vtimer_save_state(), the hardware interrupt
> > +	 * has been disabled in timer_save_state(), the hardware interrupt
> >  	 * signal may not have been retired from the interrupt controller yet.
> >  	 */
> >  	if (!vcpu)
> >  		return IRQ_HANDLED;
> >  
> > -	vtimer = vcpu_vtimer(vcpu);
> > -	if (kvm_timer_should_fire(vtimer))
> > -		kvm_timer_update_irq(vcpu, true, vtimer);
> > +	if (irq == host_vtimer_irq)
> > +		ctx = vcpu_vtimer(vcpu);
> > +	else
> > +		ctx = vcpu_ptimer(vcpu);
> > +
> > +	if (kvm_timer_should_fire(ctx))
> > +		kvm_timer_update_irq(vcpu, true, ctx);
> >  
> >  	if (userspace_irqchip(vcpu->kvm) &&
> >  	    !static_branch_unlikely(&has_gic_active_state))
> > @@ -208,13 +214,25 @@ static enum hrtimer_restart kvm_phys_timer_expire(struct hrtimer *hrt)
> >  static bool kvm_timer_should_fire(struct arch_timer_context *timer_ctx)
> >  {
> >  	struct arch_timer_cpu *timer = vcpu_timer(timer_ctx->vcpu);
> > +	enum kvm_arch_timers index = arch_timer_ctx_index(timer_ctx);
> >  	u64 cval, now;
> >  
> >  	if (timer->loaded == TIMER_EL1_LOADED) {
> > -		u32 cnt_ctl;
> > +		u32 cnt_ctl = 0;
> > +
> > +		switch (index) {
> > +		case TIMER_VTIMER:
> > +			cnt_ctl = read_sysreg_el0(cntv_ctl);
> > +			break;
> > +		case TIMER_PTIMER:
> > +			cnt_ctl = read_sysreg_el0(cntp_ctl);
> > +			break;
> > +		case NR_KVM_TIMERS:
> > +			/* GCC is braindead */
> > +			cnt_ctl = 0;
> > +			break;
> > +		}
> >  
> > -		/* Only the virtual timer can be loaded so far */
> > -		cnt_ctl = read_sysreg_el0(cntv_ctl);
> >  		return  (cnt_ctl & ARCH_TIMER_CTRL_ENABLE) &&
> >  		        (cnt_ctl & ARCH_TIMER_CTRL_IT_STAT) &&
> >  		       !(cnt_ctl & ARCH_TIMER_CTRL_IT_MASK);
> > @@ -310,7 +328,7 @@ static void kvm_timer_update_state(struct kvm_vcpu *vcpu)
> >  		return;
> >  
> >  	/*
> > -	 * The vtimer virtual interrupt is a 'mapped' interrupt, meaning part
> > +	 * If the timer virtual interrupt is a 'mapped' interrupt, part
> >  	 * of its lifecycle is offloaded to the hardware, and we therefore may
> >  	 * not have lowered the irq.level value before having to signal a new
> >  	 * interrupt, but have to signal an interrupt every time the level is
> > @@ -319,31 +337,55 @@ static void kvm_timer_update_state(struct kvm_vcpu *vcpu)
> >  	level = kvm_timer_should_fire(vtimer);
> >  	kvm_timer_update_irq(vcpu, level, vtimer);
> >  
> > +	if (has_vhe()) {
> 
> Not really critical, but wouldn't it be cleaner to use "if
> (host_ptimer_irq > 0)" here to check if we map the phys timer as well?
> This is at least how we originally derive the decision to initialise
> everything in kvm_timer_hyp_init() below. Applies to the other instances
> of "if (has_vhe())" as well.
> But I guess we use has_vhe() because it's a static key? In this case
> would it be worth to define some cosmetic wrapper, to improve readability?
> 
> > +		level = kvm_timer_should_fire(ptimer);
> > +		kvm_timer_update_irq(vcpu, level, ptimer);
> > +
> > +		return;
> > +	}
> > +
> >  	phys_timer_emulate(vcpu);
> >  
> >  	if (kvm_timer_should_fire(ptimer) != ptimer->irq.level)
> >  		kvm_timer_update_irq(vcpu, !ptimer->irq.level, ptimer);
> >  }
> >  
> > -static void vtimer_save_state(struct kvm_vcpu *vcpu)
> > +static void timer_save_state(struct arch_timer_context *ctx)
> >  {
> > -	struct arch_timer_cpu *timer = vcpu_timer(vcpu);
> > -	struct arch_timer_context *vtimer = vcpu_vtimer(vcpu);
> > +	struct arch_timer_cpu *timer = vcpu_timer(ctx->vcpu);
> > +	enum kvm_arch_timers index = arch_timer_ctx_index(ctx);
> >  	unsigned long flags;
> >  
> > +	if (!timer->enabled)
> > +		return;
> > +
> >  	local_irq_save(flags);
> >  
> >  	if (timer->loaded == TIMER_NOT_LOADED)
> >  		goto out;
> >  
> > -	if (timer->enabled) {
> > -		vtimer->cnt_ctl = read_sysreg_el0(cntv_ctl);
> > -		vtimer->cnt_cval = read_sysreg_el0(cntv_cval);
> > -	}
> > +	switch (index) {
> > +	case TIMER_VTIMER:
> > +		ctx->cnt_ctl = read_sysreg_el0(cntv_ctl);
> > +		ctx->cnt_cval = read_sysreg_el0(cntv_cval);
> >  
> > -	/* Disable the virtual timer */
> > -	write_sysreg_el0(0, cntv_ctl);
> > -	isb();
> > +		/* Disable the timer */
> > +		write_sysreg_el0(0, cntv_ctl);
> > +		isb();
> > +
> > +		break;
> > +	case TIMER_PTIMER:
> > +		ctx->cnt_ctl = read_sysreg_el0(cntp_ctl);
> > +		ctx->cnt_cval = read_sysreg_el0(cntp_cval);
> > +
> > +		/* Disable the timer */
> > +		write_sysreg_el0(0, cntp_ctl);
> > +		isb();
> > +
> > +		break;
> > +	case NR_KVM_TIMERS:
> > +		break; /* GCC is braindead */
> > +	}
> >  
> >  	timer->loaded = TIMER_NOT_LOADED;
> >  out:
> > @@ -382,21 +424,33 @@ static void kvm_timer_unblocking(struct kvm_vcpu *vcpu)
> >  	soft_timer_cancel(&timer->bg_timer);
> >  }
> >  
> > -static void vtimer_restore_state(struct kvm_vcpu *vcpu)
> > +static void timer_restore_state(struct arch_timer_context *ctx)
> >  {
> > -	struct arch_timer_cpu *timer = vcpu_timer(vcpu);
> > -	struct arch_timer_context *vtimer = vcpu_vtimer(vcpu);
> > +	struct arch_timer_cpu *timer = vcpu_timer(ctx->vcpu);
> > +	enum kvm_arch_timers index = arch_timer_ctx_index(ctx);
> >  	unsigned long flags;
> >  
> > +	if (!timer->enabled)
> > +		return;
> > +
> >  	local_irq_save(flags);
> >  
> >  	if (timer->loaded == TIMER_EL1_LOADED)
> >  		goto out;
> >  
> > -	if (timer->enabled) {
> > -		write_sysreg_el0(vtimer->cnt_cval, cntv_cval);
> > +	switch (index) {
> > +	case TIMER_VTIMER:
> > +		write_sysreg_el0(ctx->cnt_cval, cntv_cval);
> > +		isb();
> > +		write_sysreg_el0(ctx->cnt_ctl, cntv_ctl);
> > +		break;
> > +	case TIMER_PTIMER:
> > +		write_sysreg_el0(ctx->cnt_cval, cntp_cval);
> >  		isb();
> > -		write_sysreg_el0(vtimer->cnt_ctl, cntv_ctl);
> > +		write_sysreg_el0(ctx->cnt_ctl, cntp_ctl);
> > +		break;
> > +	case NR_KVM_TIMERS:
> > +		break; /* GCC is braindead */
> >  	}
> >  
> >  	timer->loaded = TIMER_EL1_LOADED;
> > @@ -419,23 +473,23 @@ static void set_cntvoff(u64 cntvoff)
> >  	kvm_call_hyp(__kvm_timer_set_cntvoff, low, high);
> >  }
> >  
> > -static inline void set_vtimer_irq_phys_active(struct kvm_vcpu *vcpu, bool active)
> > +static inline void set_timer_irq_phys_active(struct arch_timer_context *ctx, bool active)
> >  {
> >  	int r;
> > -	r = irq_set_irqchip_state(host_vtimer_irq, IRQCHIP_STATE_ACTIVE, active);
> > +	r = irq_set_irqchip_state(ctx->host_timer_irq, IRQCHIP_STATE_ACTIVE, active);
> >  	WARN_ON(r);
> >  }
> >  
> > -static void kvm_timer_vcpu_load_gic(struct kvm_vcpu *vcpu)
> > +static void kvm_timer_vcpu_load_gic(struct arch_timer_context *ctx)
> >  {
> > -	struct arch_timer_context *vtimer = vcpu_vtimer(vcpu);
> > +	struct kvm_vcpu *vcpu = ctx->vcpu;
> >  	bool phys_active;
> >  
> >  	if (irqchip_in_kernel(vcpu->kvm))
> > -		phys_active = kvm_vgic_map_is_active(vcpu, vtimer->irq.irq);
> > +		phys_active = kvm_vgic_map_is_active(vcpu, ctx->irq.irq);
> >  	else
> > -		phys_active = vtimer->irq.level;
> > -	set_vtimer_irq_phys_active(vcpu, phys_active);
> > +		phys_active = ctx->irq.level;
> > +	set_timer_irq_phys_active(ctx, phys_active);
> >  }
> >  
> >  static void kvm_timer_vcpu_load_nogic(struct kvm_vcpu *vcpu)
> > @@ -467,14 +521,22 @@ void kvm_timer_vcpu_load(struct kvm_vcpu *vcpu)
> >  	if (unlikely(!timer->enabled))
> >  		return;
> >  
> > -	if (static_branch_likely(&has_gic_active_state))
> > -		kvm_timer_vcpu_load_gic(vcpu);
> > -	else
> > +	if (static_branch_likely(&has_gic_active_state)) {
> > +		kvm_timer_vcpu_load_gic(vtimer);
> > +		if (has_vhe())
> > +			kvm_timer_vcpu_load_gic(ptimer);
> > +	} else {
> >  		kvm_timer_vcpu_load_nogic(vcpu);
> > +	}
> >  
> >  	set_cntvoff(vtimer->cntvoff);
> >  
> > -	vtimer_restore_state(vcpu);
> > +	timer_restore_state(vtimer);
> > +
> > +	if (has_vhe()) {
> > +		timer_restore_state(ptimer);
> > +		return;
> > +	}
> >  
> >  	/* Set the background timer for the physical timer emulation. */
> >  	phys_timer_emulate(vcpu);
> > @@ -506,12 +568,17 @@ bool kvm_timer_should_notify_user(struct kvm_vcpu *vcpu)
> >  void kvm_timer_vcpu_put(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);
> >  
> >  	if (unlikely(!timer->enabled))
> >  		return;
> >  
> > -	vtimer_save_state(vcpu);
> > +	timer_save_state(vtimer);
> > +	if (has_vhe()) {
> > +		timer_save_state(ptimer);
> > +		return;
> > +	}
> >  
> >  	/*
> >  	 * Cancel the physical timer emulation, because the only case where we
> > @@ -534,8 +601,7 @@ void kvm_timer_vcpu_put(struct kvm_vcpu *vcpu)
> >  	 * counter of non-VHE case. For VHE, the virtual counter uses a fixed
> >  	 * virtual offset of zero, so no need to zero CNTVOFF_EL2 register.
> >  	 */
> > -	if (!has_vhe())
> > -		set_cntvoff(0);
> > +	set_cntvoff(0);
> >  }
> >  
> >  /*
> > @@ -550,7 +616,7 @@ static void unmask_vtimer_irq_user(struct kvm_vcpu *vcpu)
> >  	if (!kvm_timer_should_fire(vtimer)) {
> >  		kvm_timer_update_irq(vcpu, false, vtimer);
> >  		if (static_branch_likely(&has_gic_active_state))
> > -			set_vtimer_irq_phys_active(vcpu, false);
> > +			set_timer_irq_phys_active(vtimer, false);
> >  		else
> >  			enable_percpu_irq(host_vtimer_irq, host_vtimer_irq_flags);
> >  	}
> > @@ -625,7 +691,12 @@ void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
> >  	ptimer->hrtimer.function = kvm_phys_timer_expire;
> >  
> >  	vtimer->irq.irq = default_vtimer_irq.irq;
> > +	vtimer->host_timer_irq = host_vtimer_irq;
> > +	vtimer->host_timer_irq_flags = host_vtimer_irq_flags;
> > +
> >  	ptimer->irq.irq = default_ptimer_irq.irq;
> > +	ptimer->host_timer_irq = host_ptimer_irq;
> > +	ptimer->host_timer_irq_flags = host_ptimer_irq_flags;
> >  
> >  	vtimer->vcpu = vcpu;
> >  	ptimer->vcpu = vcpu;
> > @@ -634,6 +705,7 @@ void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
> >  static void kvm_timer_init_interrupt(void *info)
> >  {
> >  	enable_percpu_irq(host_vtimer_irq, host_vtimer_irq_flags);
> > +	enable_percpu_irq(host_ptimer_irq, host_ptimer_irq_flags);
> 
> Shouldn't we only enable host_ptimer_irq() if we actually pass this on
> here? So either guarded by has_vhe() or (host_ptimer_irq > 0)?
> Otherwise we would enable IRQ 0?
> 

I think the right fix is to raise an error if a VHE system doesn't give
us a valid physical irq number during init, and then leave the checks
for has_vhe() here.

Thanks,

    Christoffer
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm