| Message ID | 1241060322-25807-1-git-send-email-sheng@linux.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Sheng Yang wrote: > Disable interrupt at interrupt handler and enable it when guest ack is for > the level triggered interrupt, to prevent reinjected interrupt. MSI/MSI-X don't > need it. > > One possible problem is multiply same vector interrupt injected between irq > handler and scheduled work handler would be merged as one for MSI/MSI-X. > But AFAIK, the drivers handle it well. > > The patch fixed the oplin card performance issue(MSI-X performance is half of > MSI/INTx). > > > schedule_work(&assigned_dev->interrupt_work); > > - disable_irq_nosync(irq); > - assigned_dev->host_irq_disabled = true; > + if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) { > + disable_irq_nosync(irq); > + assigned_dev->host_irq_disabled = true; > + } > > out: > spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); > I don't have the spin_lock_irqrestore() here. The patch applies, but with fuzz. Am I missing a patch?
On Monday 04 May 2009 16:25:55 Avi Kivity wrote: > Sheng Yang wrote: > > Disable interrupt at interrupt handler and enable it when guest ack is > > for the level triggered interrupt, to prevent reinjected interrupt. > > MSI/MSI-X don't need it. > > > > One possible problem is multiply same vector interrupt injected between > > irq handler and scheduled work handler would be merged as one for > > MSI/MSI-X. But AFAIK, the drivers handle it well. > > > > The patch fixed the oplin card performance issue(MSI-X performance is > > half of MSI/INTx). > > > > > > schedule_work(&assigned_dev->interrupt_work); > > > > - disable_irq_nosync(irq); > > - assigned_dev->host_irq_disabled = true; > > + if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) { > > + disable_irq_nosync(irq); > > + assigned_dev->host_irq_disabled = true; > > + } > > > > out: > > spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); > > I don't have the spin_lock_irqrestore() here. The patch applies, but > with fuzz. Am I missing a patch? Oh, it's Marcelo's patchset... [patch 0/4] use smp_send_reschedule in vcpu_kick / assigned dev host intx race fix
Sheng Yang wrote: > On Monday 04 May 2009 16:25:55 Avi Kivity wrote: > >> Sheng Yang wrote: >> >>> Disable interrupt at interrupt handler and enable it when guest ack is >>> for the level triggered interrupt, to prevent reinjected interrupt. >>> MSI/MSI-X don't need it. >>> >>> One possible problem is multiply same vector interrupt injected between >>> irq handler and scheduled work handler would be merged as one for >>> MSI/MSI-X. But AFAIK, the drivers handle it well. >>> >>> The patch fixed the oplin card performance issue(MSI-X performance is >>> half of MSI/INTx). >>> >>> >>> schedule_work(&assigned_dev->interrupt_work); >>> >>> - disable_irq_nosync(irq); >>> - assigned_dev->host_irq_disabled = true; >>> + if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) { >>> + disable_irq_nosync(irq); >>> + assigned_dev->host_irq_disabled = true; >>> + } >>> >>> out: >>> spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); >>> >> I don't have the spin_lock_irqrestore() here. The patch applies, but >> with fuzz. Am I missing a patch? >> > > Oh, it's Marcelo's patchset... > > [patch 0/4] use smp_send_reschedule in vcpu_kick / assigned dev host intx race > fix > Okay, I am working my queue in reverse order :) But please note dependencies on other patches for me.
On Monday 04 May 2009 16:34:00 Avi Kivity wrote: > Sheng Yang wrote: > > On Monday 04 May 2009 16:25:55 Avi Kivity wrote: > >> Sheng Yang wrote: > >>> Disable interrupt at interrupt handler and enable it when guest ack is > >>> for the level triggered interrupt, to prevent reinjected interrupt. > >>> MSI/MSI-X don't need it. > >>> > >>> One possible problem is multiply same vector interrupt injected between > >>> irq handler and scheduled work handler would be merged as one for > >>> MSI/MSI-X. But AFAIK, the drivers handle it well. > >>> > >>> The patch fixed the oplin card performance issue(MSI-X performance is > >>> half of MSI/INTx). > >>> > >>> > >>> schedule_work(&assigned_dev->interrupt_work); > >>> > >>> - disable_irq_nosync(irq); > >>> - assigned_dev->host_irq_disabled = true; > >>> + if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) { > >>> + disable_irq_nosync(irq); > >>> + assigned_dev->host_irq_disabled = true; > >>> + } > >>> > >>> out: > >>> spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); > >> > >> I don't have the spin_lock_irqrestore() here. The patch applies, but > >> with fuzz. Am I missing a patch? > > > > Oh, it's Marcelo's patchset... > > > > [patch 0/4] use smp_send_reschedule in vcpu_kick / assigned dev host intx > > race fix > > Okay, I am working my queue in reverse order :) > > But please note dependencies on other patches for me. Yes, sorry for totally forgot (mine based on his patchset)... :(
Sheng Yang wrote: > Disable interrupt at interrupt handler and enable it when guest ack is for > the level triggered interrupt, to prevent reinjected interrupt. MSI/MSI-X don't > need it. > > One possible problem is multiply same vector interrupt injected between irq > handler and scheduled work handler would be merged as one for MSI/MSI-X. > But AFAIK, the drivers handle it well. > > The patch fixed the oplin card performance issue(MSI-X performance is half of > MSI/INTx). > Applied, thanks.
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index eebc5ed..de8f205 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -120,7 +120,7 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) { struct kvm_assigned_dev_kernel *assigned_dev; struct kvm *kvm; - int irq, i; + int i; assigned_dev = container_of(work, struct kvm_assigned_dev_kernel, interrupt_work); @@ -143,20 +143,10 @@ static void kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id, guest_entries[i].vector, 1); - irq = assigned_dev->host_msix_entries[i].vector; - if (irq != 0) - enable_irq(irq); - assigned_dev->host_irq_disabled = false; } - } else { + } else kvm_set_irq(assigned_dev->kvm, assigned_dev->irq_source_id, assigned_dev->guest_irq, 1); - if (assigned_dev->irq_requested_type & - KVM_DEV_IRQ_GUEST_MSI) { - enable_irq(assigned_dev->host_irq); - assigned_dev->host_irq_disabled = false; - } - } spin_unlock_irq(&assigned_dev->assigned_dev_lock); mutex_unlock(&assigned_dev->kvm->lock); @@ -179,8 +169,10 @@ static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id) schedule_work(&assigned_dev->interrupt_work); - disable_irq_nosync(irq); - assigned_dev->host_irq_disabled = true; + if (assigned_dev->irq_requested_type & KVM_DEV_IRQ_GUEST_INTX) { + disable_irq_nosync(irq); + assigned_dev->host_irq_disabled = true; + } out: spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags); @@ -417,6 +409,7 @@ static int assigned_device_enable_guest_msi(struct kvm *kvm, { dev->guest_irq = irq->guest_irq; dev->ack_notifier.gsi = -1; + dev->host_irq_disabled = false; return 0; } #endif @@ -427,6 +420,7 @@ static int assigned_device_enable_guest_msix(struct kvm *kvm, { dev->guest_irq = irq->guest_irq; dev->ack_notifier.gsi = -1; + dev->host_irq_disabled = false; return 0; } #endif
Disable interrupt at interrupt handler and enable it when guest ack is for the level triggered interrupt, to prevent reinjected interrupt. MSI/MSI-X don't need it. One possible problem is multiply same vector interrupt injected between irq handler and scheduled work handler would be merged as one for MSI/MSI-X. But AFAIK, the drivers handle it well. The patch fixed the oplin card performance issue(MSI-X performance is half of MSI/INTx). Signed-off-by: Sheng Yang <sheng@linux.intel.com> --- virt/kvm/kvm_main.c | 22 ++++++++-------------- 1 files changed, 8 insertions(+), 14 deletions(-)