Re: [PATCH 15/17] KVM: PPC: Support irq routing and irqfd for in-kernel MPIC

[Scott Wood <scottwood@freescale.com>]

On 04/19/2013 09:06:26 AM, Alexander Graf wrote:
> diff --git a/Documentation/virtual/kvm/devices/mpic.txt  
> b/Documentation/virtual/kvm/devices/mpic.txt
> index ce98e32..dadc1e0 100644
> --- a/Documentation/virtual/kvm/devices/mpic.txt
> +++ b/Documentation/virtual/kvm/devices/mpic.txt
> @@ -35,3 +35,14 @@ Groups:
> 
>      "attr" is the IRQ number.  IRQ numbers for standard sources are  
> the
>      byte offset of the relevant IVPR from EIVPR0, divided by 32.
> +
> +IRQ Routing:
> +
> +  The MPIC emulation supports IRQ routing. Only a single MPIC device  
> can
> +  be instantiated. Once that device has been created, it's available  
> as
> +  irqchip id 0.
> +

> +  This irqchip 0 has 256 interrupt pins. These pins reflect the SRC  
> pins
> +  on the MPIC controller.

This irqchip 0 has 256 interrupt pins, which expose the interrupts in  
the main array of interrupt sources (a.k.a. "SRC" interrupts).  The  
numbering is the same as the MPIC device tree binding -- based on the  
register offset from the beginning of the sources array, without regard  
to any subdivisions in chip documentation such as "internal" or  
"external" interrupts.  Default routes are established for these pins,  
with the GSI being equal to the pin number.

> +  Access to on-SRC registers is not implemented through IRQ routing  
> mechanisms.

s/on-SRC registers/non-SRC interrupts/

> diff --git a/arch/powerpc/kvm/mpic.c b/arch/powerpc/kvm/mpic.c
> index 10bc08a..d137df8 100644
> --- a/arch/powerpc/kvm/mpic.c
> +++ b/arch/powerpc/kvm/mpic.c
> @@ -1029,6 +1029,7 @@ static int openpic_cpu_write_internal(void  
> *opaque, gpa_t addr,
>  	struct irq_source *src;
>  	struct irq_dest *dst;
>  	int s_IRQ, n_IRQ;
> +	int notify_eoi = -1;
> 
>  	pr_debug("%s: cpu %d addr %#llx <= 0x%08x\n", __func__, idx,
>  		addr, val);
> @@ -1087,6 +1088,8 @@ static int openpic_cpu_write_internal(void  
> *opaque, gpa_t addr,
>  		}
> 
>  		IRQ_resetbit(&dst->servicing, s_IRQ);
> +		/* Notify listeners that the IRQ is over */
> +		notify_eoi = s_IRQ;
>  		/* Set up next servicing IRQ */
>  		s_IRQ = IRQ_get_next(opp, &dst->servicing);
>  		/* Check queued interrupts. */
> @@ -1104,6 +1107,12 @@ static int openpic_cpu_write_internal(void  
> *opaque, gpa_t addr,
>  		break;
>  	}
> 
> +	if (notify_eoi != -1) {
> +		spin_unlock_irq(&opp->lock);
> +		kvm_notify_acked_irq(opp->kvm, 0, notify_eoi);
> +		spin_lock_irq(&opp->lock);
> +	}

I'd rather not have the "_irq" here, which could break if we enter this  
patch via an "_irqsave" (I realize there currently is no such path that  
reaches EOI emulation).

Will we ever set notify_eoi when addr != EOI?  I'm wondering why it was  
moved out of the switch statement, instead of being put at the end of  
the case EOI: code.

> +/*
> + * Return value:
> + *  < 0   Interrupt was ignored (masked or not delivered for other  
> reasons)
> + *  = 0   Interrupt was coalesced (previous irq is still pending)
> + *  > 0   Number of CPUs interrupt was delivered to
> + */
> +static int mpic_set_irq(struct kvm_kernel_irq_routing_entry *e,
> +			struct kvm *kvm, int irq_source_id, int level,
> +			bool line_status)
> +{
> +	u32 irq = e->irqchip.pin;
> +	struct openpic *opp = kvm->arch.mpic;
> +
> +	spin_lock_irq(&opp->lock);
> +	openpic_set_irq(opp, irq, level);
> +	spin_unlock_irq(&opp->lock);

Use irqsave here and in kvm_set_msi.  The latter can already be called  
with interrupts disabled, and we may want to do the same for non-MSIs  
once we start assigning non-PCI devices (where there's no longer the  
excuse of "if you want it to be fast, use MSIs").

-Scott