On 2012-08-06 19:05, Peter Maydell wrote:
> Move the init of the irqchip_inject_ioctl field of KVMState out of
> kvm_irqchip_create() and into kvm_init(), so that kvm_set_irq()
> can be used even when no irqchip is created (for architectures
> that support async interrupt notification even without an in
> kernel irqchip).
>
> Signed-off-by: Peter Maydell <peter.maydell@xxxxxxxxxx>
> ---
> We can't just set irqchip_inject_ioctl in target-*/kvm.c because
> the KVMState struct layout is private to kvm-all.c. Moving the
> default initialisation of this field seemed the simplest approach.
> It's safe because the use in kvm_set_irq() is guarded by a check of
> kvm_async_interrupts_enabled().
>
> The other approach would be to have a helper function for setting
> the field, but that seems overkill when KVM_IRQ_LINE is the standard
> default value. (KVM_IRQ_LINE_STATUS seems to be undocumented,
> incidentally. I am going to assume it's another x86ism until somebody
> does document it :-))
>
> kvm-all.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/kvm-all.c b/kvm-all.c
> index 6def6c9..9a1f001 100644
> --- a/kvm-all.c
> +++ b/kvm-all.c
> @@ -1200,7 +1200,6 @@ static int kvm_irqchip_create(KVMState *s)
> return ret;
> }
>
> - s->irqchip_inject_ioctl = KVM_IRQ_LINE;
> if (kvm_check_extension(s, KVM_CAP_IRQ_INJECT_STATUS)) {
> s->irqchip_inject_ioctl = KVM_IRQ_LINE_STATUS;
> }
Either you move both or none.
KVM_IRQ_LINE is old-style, deprecated, KVM_IRQ_LINE_STATUS (i.e
injection with feedback to allow lost-tick compensation) is the current
standard that other archs should pick up.
Jan
Attachment:
signature.asc
Description: OpenPGP digital signature
