Re: [PATCH] KVM: Implement support for the RH bit

[Sasha Levin <levinsasha928@gmail.com>]

On Thu, 2011-09-01 at 20:45 +0300, Gleb Natapov wrote:
> On Thu, Sep 01, 2011 at 07:05:35PM +0300, Sasha Levin wrote:
> > The RH bit exists in the message address register (lower 32 bits of
> > the address).
> > 
> > The bit indicates whether the message should go to the processor which was
> > indicated in the destination ID bits, or whether it should go to the
> > processor running at the lowest priority.
> > 
> > Cc: Avi Kivity <avi@redhat.com>
> > Cc: Marcelo Tosatti <mtosatti@redhat.com>
> > Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
> > ---
> >  arch/x86/kvm/lapic.c      |   17 +++++++++++++++--
> >  include/linux/kvm_types.h |    1 +
> >  virt/kvm/ioapic.h         |    2 +-
> >  virt/kvm/irq_comm.c       |   14 +++++++++++---
> >  4 files changed, 28 insertions(+), 6 deletions(-)
> > 
> > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > index 57dcbd4..3347f9f 100644
> > --- a/arch/x86/kvm/lapic.c
> > +++ b/arch/x86/kvm/lapic.c
> > @@ -33,6 +33,7 @@
> >  #include <asm/page.h>
> >  #include <asm/current.h>
> >  #include <asm/apicdef.h>
> > +#include <asm/msidef.h>
> >  #include <linux/atomic.h>
> >  #include "kvm_cache_regs.h"
> >  #include "irq.h"
> > @@ -325,7 +326,7 @@ int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
> >  }
> >  
> >  int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
> > -			   int short_hand, int dest, int dest_mode)
> > +			   int short_hand, int dest, int dest_mode, int rh)
> >  {
> >  	int result = 0;
> >  	struct kvm_lapic *target = vcpu->arch.apic;
> > @@ -335,9 +336,21 @@ int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
> >  		   target, source, dest, dest_mode, short_hand);
> >  
> >  	ASSERT(target);
> > +
> > +	/* If RH bit is not set, use addressing mode from ICR */
> > +	if (rh == 0) {
> > +		u32 icr_low = apic_get_reg(target, APIC_ICR);
> > +
> > +		dest_mode = icr_low & APIC_DEST_MASK;
> 
> Why is that? This function is called for messages generated by
> ioapic/ipi/msi. In all those cases destination mode does not depend on
> the content of icr_low register in the destination processor. This
> function gets correct dest_mode as a parameter.
> 

According to the spec:

"If RH is 0, then the DM bit is ignored and the message is sent ahead
independent of whether the physical or logical destination mode is used"

As far as I understand it, dest_mode parameter is invalid if the RH bit
is set.

> > +
> > +		/* dest apic can't be 0xff if targeting specific processor */
> > +		BUG_ON(dest_mode == MSI_ADDR_DEST_MODE_PHYSICAL &&
> > +			(u8)dest == 0xff);
> Can't this be triggered by a guest?

I wrote this because of this line:

"When RH is 1 and the physical destination mode is used, the Destination
ID field must not be set to 0xFF"

I did get this check in the wrong location, it was supposed to be in the
'else' case.

> 
> > +	}
> > +
> >  	switch (short_hand) {
> >  	case APIC_DEST_NOSHORT:
> > -		if (dest_mode == 0)
> > +		if (dest_mode == MSI_ADDR_DEST_MODE_PHYSICAL)
> This is apic code, why MSI defines?
> 

You're right - I'll fix that.

> >  			/* Physical mode. */
> >  			result = kvm_apic_match_physical_addr(target, dest);
> >  		else
> > diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h
> > index fa7cc72..55d4d9a 100644
> > --- a/include/linux/kvm_types.h
> > +++ b/include/linux/kvm_types.h
> > @@ -65,6 +65,7 @@ struct kvm_lapic_irq {
> >  	u32 trig_mode;
> >  	u32 shorthand;
> >  	u32 dest_id;
> > +	u32 redir_hint;
> >  };
> >  
> >  struct gfn_to_hva_cache {
> > diff --git a/virt/kvm/ioapic.h b/virt/kvm/ioapic.h
> > index 0b190c3..fd3bb64 100644
> > --- a/virt/kvm/ioapic.h
> > +++ b/virt/kvm/ioapic.h
> > @@ -68,7 +68,7 @@ static inline struct kvm_ioapic *ioapic_irqchip(struct kvm *kvm)
> >  }
> >  
> >  int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
> > -		int short_hand, int dest, int dest_mode);
> > +		   int short_hand, int dest, int dest_mode, int rh);
> >  int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2);
> >  void kvm_ioapic_update_eoi(struct kvm *kvm, int vector, int trigger_mode);
> >  int kvm_ioapic_init(struct kvm *kvm);
> > diff --git a/virt/kvm/irq_comm.c b/virt/kvm/irq_comm.c
> > index 9f614b4..e53fc83 100644
> > --- a/virt/kvm/irq_comm.c
> > +++ b/virt/kvm/irq_comm.c
> > @@ -68,6 +68,11 @@ static int kvm_set_ioapic_irq(struct kvm_kernel_irq_routing_entry *e,
> >  	return kvm_ioapic_set_irq(ioapic, e->irqchip.pin, level);
> >  }
> >  
> > +static inline bool kvm_is_rh_lowest_prio(struct kvm_lapic_irq *irq)
> > +{
> > +	return irq->redir_hint == MSI_ADDR_REDIRECTION_LOWPRI;
> > +}
> > +
> >  inline static bool kvm_is_dm_lowest_prio(struct kvm_lapic_irq *irq)
> >  {
> >  #ifdef CONFIG_IA64
> > @@ -93,10 +98,12 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct kvm_lapic *src,
> >  			continue;
> >  
> >  		if (!kvm_apic_match_dest(vcpu, src, irq->shorthand,
> > -					irq->dest_id, irq->dest_mode))
> > +					irq->dest_id, irq->dest_mode,
> > +					irq->redir_hint))
> >  			continue;
> >  
> > -		if (!kvm_is_dm_lowest_prio(irq)) {
> > +		if (!kvm_is_dm_lowest_prio(irq) &&
> > +		    !kvm_is_rh_lowest_prio(irq)) {
> >  			if (r < 0)
> >  				r = 0;
> >  			r += kvm_apic_set_irq(vcpu, irq);
> > @@ -128,13 +135,14 @@ int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
> >  			MSI_ADDR_DEST_ID_MASK) >> MSI_ADDR_DEST_ID_SHIFT;
> >  	irq.vector = (e->msi.data &
> >  			MSI_DATA_VECTOR_MASK) >> MSI_DATA_VECTOR_SHIFT;
> > +
> > +	irq.redir_hint = (1 << MSI_ADDR_REDIRECTION_SHIFT) & e->msi.address_lo;
> >  	irq.dest_mode = (1 << MSI_ADDR_DEST_MODE_SHIFT) & e->msi.address_lo;
> >  	irq.trig_mode = (1 << MSI_DATA_TRIGGER_SHIFT) & e->msi.data;
> >  	irq.delivery_mode = e->msi.data & 0x700;
> >  	irq.level = 1;
> >  	irq.shorthand = 0;
> >  
> > -	/* TODO Deal with RH bit of MSI message address */
> >  	return kvm_irq_delivery_to_apic(kvm, NULL, &irq);
> >  }
> >  
> Reading the spec it looks like it will be enough to set irq.delivery_mode
> to lowest priority if RH bit is set and destination mode is logical. The
> change will be contained in kvm_set_msi() function. Do I miss something?

That was my plan initially, the problem I stumbled on was that quote
that I've pasted before:

"If RH is 0, then the DM bit is ignored and the message is sent ahead
independent of whether the physical or logical destination mode is used"

Which means that I can't use the DM bit from the address register, and
have to find the destination processor based on actual addressing model.

-- 

Sasha.