Re: [PATCH 01/10] KVM: Add a route layer to convert MSI message to GSI

[Avi Kivity <avi@redhat.com>]

Sheng Yang wrote:
> Avi's purpose, to use single kvm_set_irq() to deal with all interrupt, including
> MSI. So here is it.
>
> struct gsi_msg is a mapping from a special gsi(with KVM_GSI_MSG_MASK) to
> MSI/MSI-X message address/data.
>
> Now we support up to 256 gsi_msg mapping, and gsi_msg is allocated by kernel and
> provide two ioctls to userspace, which is more flexiable.
>
>   

> +#define KVM_REQUEST_GSI_MSG _IOWR(KVMIO, 0x71, struct kvm_assigned_gsi_msg)
> +#define KVM_FREE_GSI_MSG _IOR(KVMIO, 0x72, struct kvm_assigned_gsi_msg)
>   

We will also need a pair for PIC and a pair for IOAPIC routing.

How about a single ioctl to set the entire routing table?  It would take 
an array of structures:

struct {
    __u32 gsi;
    __u32 type;
    __u32 flags;
    __u32 reserved;
    union {
         struct {
             __u32 irq;
         } pic;
         struct {
            __u32 ioapic; // can have >1 ioapic
            __u32 inti;
         } ioapic;
         struct {
            ...
         } msi;
         __u32 reserved[8];
    };
};

This way we can solve the HPET irq0/inti2 mess, and also have a simpler 
way of setting MSI.  All the mess in one ioctl.

> +
> +/* Call with kvm->gsi_msg_lock hold */
> +struct kvm_gsi_msg *kvm_find_gsi_msg(struct kvm *kvm, u32 gsi)
> +{
> +	struct kvm_gsi_msg *gsi_msg;
> +	struct hlist_node *n;
> +
> +	if (!(gsi & KVM_GSI_MSG_MASK))
> +		return NULL;
> +	hlist_for_each_entry(gsi_msg, n, &kvm->gsi_msg_list, link)
> +		if (gsi_msg->gsi == gsi)
> +			goto out;
> +	gsi_msg = NULL;
> +out:
> +	return gsi_msg;
> +}
>   

Linear search is a bit sad but fine for now.  Later we can add an array 
indexed by gsi.


-- 
error compiling committee.c: too many arguments to function