lkml.org 
[lkml]   [2017]   [Aug]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    SubjectRe: [PATCH v3 41/59] KVM: arm/arm64: GICv4: Wire mapping/unmapping of VLPIs in VFIO irq bypass
    From
    Date
    Hi Eric,

    On 30/08/17 11:20, Auger Eric wrote:
    > Hi Marc,
    >
    > On 30/08/2017 11:42, Marc Zyngier wrote:
    >> On 26/08/17 20:48, Christoffer Dall wrote:
    >>> On Mon, Jul 31, 2017 at 06:26:19PM +0100, Marc Zyngier wrote:
    >>>> Let's use the irq bypass mechanism introduced for platform device
    >>>> interrupts to intercept the virtual PCIe endpoint configuration
    >>>> and establish our LPI->VLPI mapping.
    >>>>
    >>>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
    >>>> ---
    >>>> include/kvm/arm_vgic.h | 8 ++++
    >>>> virt/kvm/arm/arm.c | 27 ++++++++----
    >>>> virt/kvm/arm/vgic/vgic-v4.c | 103 ++++++++++++++++++++++++++++++++++++++++++++
    >>>> 3 files changed, 130 insertions(+), 8 deletions(-)
    >>>>
    >>>> diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h
    >>>> index 359eeffe9857..050f78d4fb42 100644
    >>>> --- a/include/kvm/arm_vgic.h
    >>>> +++ b/include/kvm/arm_vgic.h
    >>>> @@ -367,4 +367,12 @@ int kvm_vgic_set_forwarding(struct kvm *kvm, unsigned int host_irq,
    >>>> void kvm_vgic_unset_forwarding(struct kvm *kvm, unsigned int host_irq,
    >>>> unsigned int vintid);
    >>>>
    >>>> +struct kvm_kernel_irq_routing_entry;
    >>>> +
    >>>> +int kvm_vgic_v4_set_forwarding(struct kvm *kvm, int irq,
    >>>> + struct kvm_kernel_irq_routing_entry *irq_entry);
    >>>> +
    >>>> +int kvm_vgic_v4_unset_forwarding(struct kvm *kvm, int irq,
    >>>> + struct kvm_kernel_irq_routing_entry *irq_entry);
    >>>> +
    >>>> #endif /* __KVM_ARM_VGIC_H */
    >>>> diff --git a/virt/kvm/arm/arm.c b/virt/kvm/arm/arm.c
    >>>> index ebab6c29e3be..6803ea27c47d 100644
    >>>> --- a/virt/kvm/arm/arm.c
    >>>> +++ b/virt/kvm/arm/arm.c
    >>>> @@ -1457,11 +1457,16 @@ int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *cons,
    >>>> struct kvm_kernel_irqfd *irqfd =
    >>>> container_of(cons, struct kvm_kernel_irqfd, consumer);
    >>>>
    >>>> - if (prod->type != IRQ_BYPASS_VFIO_PLATFORM)
    >>>> + switch (prod->type) {
    >>>> + case IRQ_BYPASS_VFIO_PLATFORM:
    >>>> + return kvm_vgic_set_forwarding(irqfd->kvm, prod->irq,
    >>>> + irqfd->gsi + VGIC_NR_PRIVATE_IRQS);
    >>>> + case IRQ_BYPASS_VFIO_PCI_MSI:
    >>>> + return kvm_vgic_v4_set_forwarding(irqfd->kvm, prod->irq,
    >>>> + &irqfd->irq_entry);
    >>>> + default:
    >>>> return 0;
    >>>> -
    >>>> - return kvm_vgic_set_forwarding(irqfd->kvm, prod->irq,
    >>>> - irqfd->gsi + VGIC_NR_PRIVATE_IRQS);
    >>>> + }
    >>>> }
    >>>> void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons,
    >>>> struct irq_bypass_producer *prod)
    >>>> @@ -1469,11 +1474,17 @@ void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons,
    >>>> struct kvm_kernel_irqfd *irqfd =
    >>>> container_of(cons, struct kvm_kernel_irqfd, consumer);
    >>>>
    >>>> - if (prod->type != IRQ_BYPASS_VFIO_PLATFORM)
    >>>> - return;
    >>>> + switch (prod->type) {
    >>>> + case IRQ_BYPASS_VFIO_PLATFORM:
    >>>> + kvm_vgic_unset_forwarding(irqfd->kvm, prod->irq,
    >>>> + irqfd->gsi + VGIC_NR_PRIVATE_IRQS);
    >>>> + break;
    >>>>
    >>>> - kvm_vgic_unset_forwarding(irqfd->kvm, prod->irq,
    >>>> - irqfd->gsi + VGIC_NR_PRIVATE_IRQS);
    >>>> + case IRQ_BYPASS_VFIO_PCI_MSI:
    >>>> + kvm_vgic_v4_unset_forwarding(irqfd->kvm, prod->irq,
    >>>> + &irqfd->irq_entry);
    >>>> + break;
    >>>> + }
    >>>> }
    >>>>
    >>>> void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *cons)
    >>>> diff --git a/virt/kvm/arm/vgic/vgic-v4.c b/virt/kvm/arm/vgic/vgic-v4.c
    >>>> index 207e1fda0dcd..338c86c5159f 100644
    >>>> --- a/virt/kvm/arm/vgic/vgic-v4.c
    >>>> +++ b/virt/kvm/arm/vgic/vgic-v4.c
    >>>> @@ -72,3 +72,106 @@ void vgic_v4_teardown(struct kvm *kvm)
    >>>> its_vm->nr_vpes = 0;
    >>>> its_vm->vpes = NULL;
    >>>> }
    >>>> +
    >>>> +static struct vgic_its *vgic_get_its(struct kvm *kvm,
    >>>> + struct kvm_kernel_irq_routing_entry *irq_entry)
    >>>> +{
    >>>> + struct kvm_msi msi = (struct kvm_msi) {
    >>>> + .address_lo = irq_entry->msi.address_lo,
    >>>> + .address_hi = irq_entry->msi.address_hi,
    >>>> + .data = irq_entry->msi.data,
    >>>> + .flags = irq_entry->msi.flags,
    >>>> + .devid = irq_entry->msi.devid,
    >>>> + };
    >>>> +
    >>>> + /*
    >>>> + * Get a reference on the LPI. If NULL, this is not a valid
    >>>> + * translation for any of our vITSs.
    >>>> + */
    >>>> + return vgic_msi_to_its(kvm, &msi);
    >>>> +}
    >>>> +
    >>>> +int kvm_vgic_v4_set_forwarding(struct kvm *kvm, int virq,
    >>>> + struct kvm_kernel_irq_routing_entry *irq_entry)
    >>>> +{
    >>>> + struct vgic_its *its;
    >>>> + struct vgic_irq *irq;
    >>>> + struct its_vlpi_map map;
    >>>> + int ret;
    >>>> +
    >>>> + if (!vgic_is_v4_capable(kvm))
    >>>> + return 0;
    >>>> +
    >>>> + /*
    >>>> + * Get the ITS, and escape early on error (not a valid
    >>>> + * doorbell for any of our vITSs).
    >>>> + */
    >>>> + its = vgic_get_its(kvm, irq_entry);
    >>>> + if (IS_ERR(its))
    >>>> + return 0;
    >>>> +
    >>>> + mutex_lock(&its->its_lock);
    >>>> +
    >>>> + /* Perform then actual DevID/EventID -> LPI translation. */
    >>>> + ret = vgic_its_resolve_lpi(kvm, its, irq_entry->msi.devid,
    >>>> + irq_entry->msi.data, &irq);
    >>>> + if (ret)
    >>>> + goto out;
    >>>> +
    >>>> + /*
    >>>> + * Emit the mapping request. If it fails, the ITS probably
    >>>> + * isn't v4 compatible, so let's silently bail out. Holding
    >>>> + * the ITS lock should ensure that nothing can modify the
    >>>> + * target vcpu.
    >>>> + */
    >>>> + map = (struct its_vlpi_map) {
    >>>> + .vm = &kvm->arch.vgic.its_vm,
    >>>> + .vintid = irq->intid,
    >>>> + .db_enabled = true,
    >>>> + .vpe_idx = irq->target_vcpu->vcpu_id,
    >>>> + };
    >>>> +
    >>>> + if (its_map_vlpi(virq, &map))
    >>>> + goto out;
    >>>
    >>> This seems to be able to return things like -ENOMEM, whould we really
    >>> not report this back to the caller in any way?
    >>
    >>
    >> That's a good question.
    >>
    >> If we return -ENOMEM, we'll probably end-up returning an error to
    >> userspace (as a result of the VFIO ioctl), which will in turn probably
    >> terminate the guest (I'm guessing, I haven't actually looked at what
    >> userspace does).
    >>
    >> If we don't return the error, then we have a chance to keep the guest
    >> running by sticking to software injection.
    > I have not read the whole stuff yet but userspace is not aware of this
    > negotiation. Everything happens under the hood in kernel, see
    > virt/lib/irqbypass.c __connect(): if add_producer() fails
    > prod->del_consumer() is called and we should return to the not optimized
    > injection.

    Ah, fair enough. I guess del_consumer() does nothing on PCI?

    Thanks,

    M.
    --
    Jazz is not dead. It just smells funny...

    \
     
     \ /
      Last update: 2017-08-30 12:43    [W:4.100 / U:0.196 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site