Re: [PATCH 3/8] KVM: Handle device assignment to guests

[Avi Kivity <avi@qumranet.com>]

Han, Weidong wrote:
> Avi Kivity wrote:
>   
>>> +static void kvm_pci_pt_work_fn(struct work_struct *work) +{
>>> +	struct kvm_pci_pt_dev_list *match;
>>> +	struct kvm_pci_pt_work *int_work;
>>> +	int source;
>>> +	unsigned long flags;
>>> +	int guest_irq;
>>> +	int host_irq;
>>> +
>>> +	int_work = container_of(work, struct kvm_pci_pt_work, work); +
>>> +	source = int_work->source ? KVM_PT_SOURCE_IRQ_ACK :
>>> KVM_PT_SOURCE_IRQ; + +	/* This is taken to safely inject irq
>>>       
> inside
>   
>>> the guest. When +	 * the interrupt injection (or the ioapic code)
>>> uses a +	 * finer-grained lock, update this
>>> +	 */
>>> +	mutex_lock(&int_work->kvm->lock);
>>> +	read_lock_irqsave(&kvm_pci_pt_lock, flags);
>>> +	match =
>>>       
> kvm_find_pci_pt_dev(&int_work->kvm->arch.pci_pt_dev_head,
>   
>>> NULL, +				    int_work->irq, source);
>>> +	if (!match) {
>>> +		printk(KERN_ERR "%s: no matching device assigned to
>>>       
> guest "
>   
>>> +		       "found for irq %d, source = %d!\n",
>>> +		       __func__, int_work->irq, int_work->source);
>>> +		read_unlock_irqrestore(&kvm_pci_pt_lock, flags); +
>>>       
> goto out;
>   
>>> +	}
>>> +	guest_irq = match->pt_dev.guest.irq;
>>> +	host_irq = match->pt_dev.host.irq;
>>> +	read_unlock_irqrestore(&kvm_pci_pt_lock, flags);
>>> +
>>> +	if (source == KVM_PT_SOURCE_IRQ)
>>> +		kvm_set_irq(int_work->kvm, guest_irq, 1);
>>> +	else {
>>> +		kvm_set_irq(int_work->kvm, int_work->irq, 0);
>>> +		enable_irq(host_irq);
>>> +	}
>>> +out:
>>> +	mutex_unlock(&int_work->kvm->lock);
>>> +	kvm_put_kvm(int_work->kvm);
>>> +}
>>>
>>> +
>>> +/* FIXME: Implement the OR logic needed to make shared interrupts
>>> on + * this line behave properly + */
>>>
>>>       
>> Isn't this a showstopper?  There is no easy way for a user to avoid
>> sharing, especially as we have only three pci irqs at present.
>>
>>     
>
> Currently it's not easy to avoid sharing. I think we can support MSI for
> assgined device to solve sharing problem. 
>   

MSI is definitely the right direction, but we also need to support the 
OR logic for guests that do not support MSI.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.