[v4,46/47] ivshmem: use kvm irqfd for msi notifications

From: Marc-André Lureau <marcandre.lureau@redhat.com>

Use irqfd for improving context switch when notifying the guest.
If the host doesn't support kvm irqfd, regular msi notifications are
still supported.

Note: the ivshmem implementation doesn't allow switching between MSI and
IO interrupts, this patch doesn't either.

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
---
 hw/misc/ivshmem.c | 175 ++++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 169 insertions(+), 6 deletions(-)

On 24.09.2015 13:37, marcandre.lureau@redhat.com wrote:
> From: Marc-André Lureau <marcandre.lureau@redhat.com>
> 
> Use irqfd for improving context switch when notifying the guest.
> If the host doesn't support kvm irqfd, regular msi notifications are
> still supported.
> 
> Note: the ivshmem implementation doesn't allow switching between MSI and
> IO interrupts, this patch doesn't either.
> 
> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>

Paolo could you also take a look at this one?
Seems to work, but I am not familiar with the kvm msi irqfd primitives.

One comment below.

> ---
>  hw/misc/ivshmem.c | 175 ++++++++++++++++++++++++++++++++++++++++++++++++++++--
>  1 file changed, 169 insertions(+), 6 deletions(-)
> 
> diff --git a/hw/misc/ivshmem.c b/hw/misc/ivshmem.c
> index 73644cc..39c0791 100644
> --- a/hw/misc/ivshmem.c
> +++ b/hw/misc/ivshmem.c
> @@ -19,6 +19,7 @@
>  #include "hw/hw.h"
>  #include "hw/i386/pc.h"
>  #include "hw/pci/pci.h"
> +#include "hw/pci/msi.h"
>  #include "hw/pci/msix.h"
>  #include "sysemu/kvm.h"
>  #include "migration/migration.h"
> @@ -68,6 +69,7 @@ typedef struct Peer {
>  
>  typedef struct MSIVector {
>      PCIDevice *pdev;
> +    int virq;
>  } MSIVector;
>  
>  typedef struct IVShmemState {
> @@ -293,13 +295,73 @@ static void fake_irqfd(void *opaque, const uint8_t *buf, int size) {
>      msix_notify(pdev, vector);
>  }
>  
> +static int ivshmem_vector_unmask(PCIDevice *dev, unsigned vector,
> +                                 MSIMessage msg)
> +{
> +    IVShmemState *s = IVSHMEM(dev);
> +    EventNotifier *n = &s->peers[s->vm_id].eventfds[vector];
> +    MSIVector *v = &s->msi_vectors[vector];
> +    int ret;
> +
> +    IVSHMEM_DPRINTF("vector unmask %p %d\n", dev, vector);
> +
> +    ret = kvm_irqchip_update_msi_route(kvm_state, v->virq, msg);
> +    if (ret < 0) {
> +        return ret;
> +    }
> +
> +    return kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, n, NULL, v->virq);
> +}
> +
> +static void ivshmem_vector_mask(PCIDevice *dev, unsigned vector)
> +{
> +    IVShmemState *s = IVSHMEM(dev);
> +    EventNotifier *n = &s->peers[s->vm_id].eventfds[vector];
> +    int ret;
> +
> +    IVSHMEM_DPRINTF("vector mask %p %d\n", dev, vector);
> +
> +    ret = kvm_irqchip_remove_irqfd_notifier_gsi(kvm_state, n,
> +                                                s->msi_vectors[vector].virq);
> +    if (ret != 0) {
> +        error_report("remove_irqfd_notifier_gsi failed");
> +    }
> +}
> +
> +static void ivshmem_vector_poll(PCIDevice *dev,
> +                                unsigned int vector_start,
> +                                unsigned int vector_end)
> +{
> +    IVShmemState *s = IVSHMEM(dev);
> +    unsigned int vector;
> +
> +    IVSHMEM_DPRINTF("vector poll %p %d-%d\n", dev, vector_start, vector_end);
> +
> +    vector_end = MIN(vector_end, s->vectors);
> +
> +    for (vector = vector_start; vector < vector_end; vector++) {
> +        EventNotifier *notifier = &s->peers[s->vm_id].eventfds[vector];
> +
> +        if (!msix_is_masked(dev, vector)) {
> +            continue;
> +        }
> +
> +        if (event_notifier_test_and_clear(notifier)) {
> +            msix_set_pending(dev, vector);
> +        }
> +    }
> +}
> +
>  static CharDriverState* create_eventfd_chr_device(void * opaque, EventNotifier *n,
>                                                    int vector)
>  {
>      /* create a event character device based on the passed eventfd */
>      IVShmemState *s = opaque;
> -    CharDriverState * chr;
> +    PCIDevice *pdev = PCI_DEVICE(s);
>      int eventfd = event_notifier_get_fd(n);
> +    CharDriverState *chr;
> +
> +    s->msi_vectors[vector].pdev = pdev;
>  
>      chr = qemu_chr_open_eventfd(eventfd);
>  
> @@ -484,6 +546,53 @@ static bool fifo_update_and_get(IVShmemState *s, const uint8_t *buf, int size,
>      return true;
>  }
>  
> +static int ivshmem_add_kvm_msi_virq(IVShmemState *s, int vector)
> +{
> +    PCIDevice *pdev = PCI_DEVICE(s);
> +    MSIMessage msg = msix_get_message(pdev, vector);
> +    int ret;
> +
> +    IVSHMEM_DPRINTF("ivshmem_add_kvm_msi_virq vector:%d\n", vector);
> +
> +    if (s->msi_vectors[vector].pdev != NULL) {
> +        return 0;
> +    }
> +
> +    ret = kvm_irqchip_add_msi_route(kvm_state, msg); /*  */
> +    if (ret < 0) {
> +        error_report("ivshmem: kvm_irqchip_add_msi_route failed");
> +        return -1;
> +    }
> +
> +    s->msi_vectors[vector].virq = ret;
> +    s->msi_vectors[vector].pdev = pdev;
> +
> +    return 0;
> +}
> +
> +static void setup_interrupt(IVShmemState *s, int vector)
> +{
> +    EventNotifier *n = &s->peers[s->vm_id].eventfds[vector];
> +    bool with_irqfd = kvm_msi_via_irqfd_enabled() &&
> +        ivshmem_has_feature(s, IVSHMEM_MSI);
> +    PCIDevice *pdev = PCI_DEVICE(s);
> +
> +    IVSHMEM_DPRINTF("setting up interrupt for vector: %d\n", vector);
> +
> +    if (!with_irqfd) {
> +        s->eventfd_chr[vector] = create_eventfd_chr_device(s, n, vector);
> +    } else if (msix_enabled(pdev)) {
> +        if (ivshmem_add_kvm_msi_virq(s, vector) < 0) {
> +            return;
> +        }
> +
> +        if (!msix_is_masked(pdev, vector)) {
> +            kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, n, NULL,
> +                                               s->msi_vectors[vector].virq);
> +        }
> +    }

Is an } else { here missing?

Should we add IVSHMEM_DPRINTFs to help the developer understand which of these paths have been taken?

> +}
> +
>  static void ivshmem_read(void *opaque, const uint8_t *buf, int size)
>  {
>      IVShmemState *s = opaque;
> @@ -587,11 +696,10 @@ static void ivshmem_read(void *opaque, const uint8_t *buf, int size)
>      IVSHMEM_DPRINTF("eventfds[%ld][%d] = %d\n", incoming_posn,
>                      new_eventfd, incoming_fd);
>      event_notifier_init_fd(&peer->eventfds[new_eventfd], incoming_fd);
> +    fcntl_setfl(incoming_fd, O_NONBLOCK); /* msix/irqfd poll non block */
>  
>      if (incoming_posn == s->vm_id) {
> -        s->eventfd_chr[new_eventfd] = create_eventfd_chr_device(s,
> -                   &s->peers[s->vm_id].eventfds[new_eventfd],
> -                   new_eventfd);
> +        setup_interrupt(s, new_eventfd);
>      }
>  
>      if (ivshmem_has_feature(s, IVSHMEM_IOEVENTFD)) {
> @@ -666,10 +774,65 @@ static int ivshmem_setup_msi(IVShmemState * s)
>      return 0;
>  }
>  
> -static void ivshmem_write_config(PCIDevice *pci_dev, uint32_t address,
> +static void ivshmem_enable_irqfd(IVShmemState *s)
> +{
> +    PCIDevice *pdev = PCI_DEVICE(s);
> +    int i;
> +
> +    for (i = 0; i < s->peers[s->vm_id].nb_eventfds; i++) {
> +        ivshmem_add_kvm_msi_virq(s, i);
> +    }
> +
> +    if (msix_set_vector_notifiers(pdev,
> +                                  ivshmem_vector_unmask,
> +                                  ivshmem_vector_mask,
> +                                  ivshmem_vector_poll)) {
> +        error_report("ivshmem: msix_set_vector_notifiers failed");
> +    }
> +}
> +
> +static void ivshmem_remove_kvm_msi_virq(IVShmemState *s, int vector)
> +{
> +    IVSHMEM_DPRINTF("ivshmem_remove_kvm_msi_virq vector:%d\n", vector);
> +
> +    if (s->msi_vectors[vector].pdev == NULL) {
> +        return;
> +    }
> +
> +    /* it was cleaned when masked in the frontend. */
> +    kvm_irqchip_release_virq(kvm_state, s->msi_vectors[vector].virq);
> +
> +    s->msi_vectors[vector].pdev = NULL;
> +}
> +
> +static void ivshmem_disable_irqfd(IVShmemState *s)
> +{
> +    PCIDevice *pdev = PCI_DEVICE(s);
> +    int i;
> +
> +    for (i = 0; i < s->peers[s->vm_id].nb_eventfds; i++) {
> +        ivshmem_remove_kvm_msi_virq(s, i);
> +    }
> +
> +    msix_unset_vector_notifiers(pdev);
> +}
> +
> +static void ivshmem_write_config(PCIDevice *pdev, uint32_t address,
>                                   uint32_t val, int len)
>  {
> -    pci_default_write_config(pci_dev, address, val, len);
> +    IVShmemState *s = IVSHMEM(pdev);
> +    int is_enabled, was_enabled = msix_enabled(pdev);
> +
> +    pci_default_write_config(pdev, address, val, len);
> +    is_enabled = msix_enabled(pdev);
> +
> +    if (kvm_msi_via_irqfd_enabled() && s->vm_id != -1) {
> +        if (!was_enabled && is_enabled) {
> +            ivshmem_enable_irqfd(s);
> +        } else if (was_enabled && !is_enabled) {
> +            ivshmem_disable_irqfd(s);
> +        }
> +    }
>  }
>  
>  static void pci_ivshmem_realize(PCIDevice *dev, Error **errp)
>

On Wed, Sep 30, 2015 at 1:47 PM, Claudio Fontana
<claudio.fontana@huawei.com> wrote:
> Is an } else { here missing?

No, but I added a comment: "it will be delayed until msix is enabled,
in write_config "

>
> Should we add IVSHMEM_DPRINTFs to help the developer understand which of these paths have been taken?

ok