On Saturday 28 February 2009 01:54:31 Marcelo Tosatti wrote:
> On Fri, Feb 27, 2009 at 12:17:01PM +0800, Yang, Sheng wrote:
> > On Friday 27 February 2009 07:50:54 Marcelo Tosatti wrote:
> > > Can someone with HW test this please?
> >
> > Good catch! The patch works fine on my side.
> >
> > Can it be a per-device lock? One big lock for all assigned device seems
> > restrict scalability.
>
> Since all state is per-device, yes.
>
> Can you please review, test and ack the patch below?
I just checked dmesg and got this...
[ 1105.343824] ------------[ cut here ]------------
[ 1105.347814] WARNING: at kernel/smp.c:226
smp_call_function_single+0x41/0x10b()
[ 1105.347814] Hardware name: To Be Filled By O.E.M.
[ 1105.347814] Modules linked in: kvm_intel kvm bridge stp llc i2c_dev
i2c_core e1000 e1000e ehci_hcd ohci_hcd uhci_hcd
[ 1105.347814] Pid: 9, comm: events/0 Tainted: G W 2.6.29-rc4-00001-
gd5b5623 #20
[ 1105.347814] Call Trace:
[ 1105.347814] [<ffffffff80239ddc>] warn_slowpath+0xd3/0xf2
[ 1105.347814] [<ffffffff8022ec6d>] ? __enqueue_entity+0x74/0x76
[ 1105.347814] [<ffffffff8022ed1c>] ? enqueue_entity+0xad/0xb6
[ 1105.347814] [<ffffffff80232771>] ? try_to_wake_up+0x1ff/0x211
[ 1105.347814] [<ffffffff80260061>] smp_call_function_single+0x41/0x10b
[ 1105.347814] [<ffffffffa0089bc5>] kvm_vcpu_kick+0x74/0x7c [kvm]
[ 1105.347814] [<ffffffffa009c559>] kvm_apic_set_irq+0x70/0x77 [kvm]
[ 1105.347814] [<ffffffffa009dfa8>] kvm_set_msi+0xe8/0x10d [kvm]
[ 1105.347814] [<ffffffffa00869b6>] ?
kvm_assigned_dev_interrupt_work_handler+0x30/0xfd [kvm]
[ 1105.347814] [<ffffffffa009e0a4>] kvm_set_irq+0x6f/0xb3 [kvm]
[ 1105.347814] [<ffffffffa0086a02>]
kvm_assigned_dev_interrupt_work_handler+0x7c/0xfd [kvm]
[ 1105.347814] [<ffffffffa0086986>] ?
kvm_assigned_dev_interrupt_work_handler+0x0/0xfd [kvm]
[ 1105.347814] [<ffffffff802497ac>] run_workqueue+0xf5/0x1fd
[ 1105.347814] [<ffffffff80249756>] ? run_workqueue+0x9f/0x1fd
[ 1105.347814] [<ffffffff8024a49e>] worker_thread+0xdb/0xe8
[ 1105.347814] [<ffffffff8024d342>] ? autoremove_wake_function+0x0/0x38
[ 1105.347814] [<ffffffff8024a3c3>] ? worker_thread+0x0/0xe8
[ 1105.347814] [<ffffffff8024d231>] kthread+0x49/0x78
[ 1105.347814] [<ffffffff8020c53a>] child_rip+0xa/0x20
[ 1105.347814] [<ffffffff8020bf3c>] ? restore_args+0x0/0x30
[ 1105.347814] [<ffffffff802342c2>] ? finish_task_switch+0x0/0xf3
[ 1105.347814] [<ffffffff8024d1e8>] ? kthread+0x0/0x78
[ 1105.347814] [<ffffffff8020c530>] ? child_rip+0x0/0x20
[ 1105.347814] ---[ end trace 3b3fe301343db608 ]---
--
regards
Yang, Sheng
> Thanks.
>
> > > -----
> > >
> > > kvm_assigned_dev_ack_irq is vulnerable to a race condition with the
> > > interrupt handler function. It does:
> > >
> > > if (dev->host_irq_disabled) {
> > > enable_irq(dev->host_irq);
> > > dev->host_irq_disabled = false;
> > > }
> > >
> > > If an interrupt triggers before the host->dev_irq_disabled assignment,
> > > it will disable the interrupt and set dev->host_irq_disabled to true.
> > >
> > > On return to kvm_assigned_dev_ack_irq, dev->host_irq_disabled is set to
> > > false, and the next kvm_assigned_dev_ack_irq call will fail to reenable
> > > it.
> > >
> > > Other than that, having the interrupt handler and work handlers run in
> > > parallel sounds like asking for trouble (could not spot any obvious
> > > problem, but better not have to, its fragile).
> >
> > Well, my original purpose is a FIFO between interrupt handler and
> > work(for MSI-X), but seems too complex... And I also don't see any
> > problem for now...
> >
> > --
> > regards
> > Yang, Sheng
>
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 3832243..16bf60b 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -349,6 +349,7 @@ struct kvm_assigned_dev_kernel {
> int flags;
> struct pci_dev *dev;
> struct kvm *kvm;
> + spinlock_t assigned_dev_lock;
> };
>
> struct kvm_irq_mask_notifier {
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index 4d2be16..b7d060f 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -41,6 +41,7 @@
> #include <linux/pagemap.h>
> #include <linux/mman.h>
> #include <linux/swap.h>
> +#include <linux/spinlock.h>
>
> #include <asm/processor.h>
> #include <asm/io.h>
> @@ -132,6 +133,7 @@ static void
> kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) *
> finer-grained lock, update this
> */
> mutex_lock(&kvm->lock);
> + spin_lock_irq(&assigned_dev->assigned_dev_lock);
> if (assigned_dev->irq_requested_type & KVM_ASSIGNED_DEV_MSIX) {
> struct kvm_guest_msix_entry *guest_entries =
> assigned_dev->guest_msix_entries;
> @@ -158,18 +160,21 @@ static void
> kvm_assigned_dev_interrupt_work_handler(struct work_struct *work) }
> }
>
> + spin_unlock_irq(&assigned_dev->assigned_dev_lock);
> mutex_unlock(&assigned_dev->kvm->lock);
> }
>
> static irqreturn_t kvm_assigned_dev_intr(int irq, void *dev_id)
> {
> + unsigned long flags;
> struct kvm_assigned_dev_kernel *assigned_dev =
> (struct kvm_assigned_dev_kernel *) dev_id;
>
> + spin_lock_irqsave(&assigned_dev->assigned_dev_lock, flags);
> if (assigned_dev->irq_requested_type == KVM_ASSIGNED_DEV_MSIX) {
> int index = find_index_from_host_irq(assigned_dev, irq);
> if (index < 0)
> - return IRQ_HANDLED;
> + goto out;
> assigned_dev->guest_msix_entries[index].flags |=
> KVM_ASSIGNED_MSIX_PENDING;
> }
> @@ -179,6 +184,8 @@ static irqreturn_t kvm_assigned_dev_intr(int irq, void
> *dev_id) disable_irq_nosync(irq);
> assigned_dev->host_irq_disabled = true;
>
> +out:
> + spin_unlock_irqrestore(&assigned_dev->assigned_dev_lock, flags);
> return IRQ_HANDLED;
> }
>
> @@ -186,6 +193,7 @@ static irqreturn_t kvm_assigned_dev_intr(int irq, void
> *dev_id) static void kvm_assigned_dev_ack_irq(struct kvm_irq_ack_notifier
> *kian) {
> struct kvm_assigned_dev_kernel *dev;
> + unsigned long flags;
>
> if (kian->gsi == -1)
> return;
> @@ -198,10 +206,12 @@ static void kvm_assigned_dev_ack_irq(struct
> kvm_irq_ack_notifier *kian) /* The guest irq may be shared so this ack may
> be
> * from another device.
> */
> + spin_lock_irqsave(&dev->assigned_dev_lock, flags);
> if (dev->host_irq_disabled) {
> enable_irq(dev->host_irq);
> dev->host_irq_disabled = false;
> }
> + spin_unlock_irqrestore(&dev->assigned_dev_lock, flags);
> }
>
> /* The function implicit hold kvm->lock mutex due to cancel_work_sync() */
> @@ -604,6 +614,7 @@ static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
> match->dev = dev;
> match->irq_source_id = -1;
> match->kvm = kvm;
> + spin_lock_init(&match->assigned_dev_lock);
>
> list_add(&match->list, &kvm->arch.assigned_dev_head);
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
