Patchwork virtio-pci: Disable virtio-ioeventfd when !CONFIG_IOTHREAD

login
register
mail settings
Submitter Stefan Hajnoczi
Date Jan. 25, 2011, 1:58 p.m.
Message ID <1295963888-17275-1-git-send-email-stefanha@linux.vnet.ibm.com>
Download mbox | patch
Permalink /patch/80361/
State New
Headers show

Comments

Stefan Hajnoczi - Jan. 25, 2011, 1:58 p.m.
It is not possible to use virtio-ioeventfd when building without an I/O
thread.  We rely on a signal to kick us out of vcpu execution.  Timers
and AIO use SIGALRM and SIGUSR2 respectively.  Unfortunately eventfd
does not support O_ASYNC (SIGIO) so eventfd cannot be used in a signal
driven manner.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
---
 hw/virtio-pci.c |   21 ++++++++++++++++++++-
 1 files changed, 20 insertions(+), 1 deletions(-)
Michael S. Tsirkin - Jan. 25, 2011, 2:13 p.m.
On Tue, Jan 25, 2011 at 01:58:08PM +0000, Stefan Hajnoczi wrote:
> It is not possible to use virtio-ioeventfd when building without an I/O
> thread.  We rely on a signal to kick us out of vcpu execution.  Timers
> and AIO use SIGALRM and SIGUSR2 respectively.  Unfortunately eventfd
> does not support O_ASYNC (SIGIO) so eventfd cannot be used in a signal
> driven manner.
> 
> Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>

I'd rather have the CONFIG_ ... mess contained in kvm-all.c
which tests CONFIG_IOTHREAD anyway, than spread out
to devices. Can we make kvm_has_many_ioeventfds check this?

> ---
>  hw/virtio-pci.c |   21 ++++++++++++++++++++-
>  1 files changed, 20 insertions(+), 1 deletions(-)
> 
> diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
> index d07ff97..e921eda 100644
> --- a/hw/virtio-pci.c
> +++ b/hw/virtio-pci.c
> @@ -640,6 +640,25 @@ static int virtio_pci_set_host_notifier(void *opaque, int n, bool assign)
>      return virtio_pci_set_host_notifier_internal(proxy, n, assign);
>  }
>  
> +static bool virtio_pci_can_use_ioeventfd(void)
> +{
> +    if (!kvm_has_many_ioeventfds()) {
> +        return false;
> +    }
> +
> +    /* Use ioeventfd for virtqueue kick only if we have an I/O thread to
> +     * perform out-of-line processing.  Otherwise we might as well do
> +     * synchronous virtqueue kicks and in fact we have to since eventfd does
> +     * not support SIGIO.  Without the I/O thread a signal would be required to
> +     * kick the vcpu out of guest code.
> +     */
> +#ifdef CONFIG_IOTHREAD
> +    return true;
> +#else
> +    return false;
> +#endif
> +}
> +
>  static void virtio_pci_vmstate_change(void *opaque, bool running)
>  {
>      VirtIOPCIProxy *proxy = opaque;
> @@ -705,7 +724,7 @@ static void virtio_init_pci(VirtIOPCIProxy *proxy, VirtIODevice *vdev,
>      pci_register_bar(&proxy->pci_dev, 0, size, PCI_BASE_ADDRESS_SPACE_IO,
>                             virtio_map);
>  
> -    if (!kvm_has_many_ioeventfds()) {
> +    if (!virtio_pci_can_use_ioeventfd()) {
>          proxy->flags &= ~VIRTIO_PCI_FLAG_USE_IOEVENTFD;
>      }
>  
> -- 
> 1.7.2.3
Stefan Hajnoczi - Jan. 25, 2011, 2:20 p.m.
On Tue, Jan 25, 2011 at 2:13 PM, Michael S. Tsirkin <mst@redhat.com> wrote:
> On Tue, Jan 25, 2011 at 01:58:08PM +0000, Stefan Hajnoczi wrote:
>> It is not possible to use virtio-ioeventfd when building without an I/O
>> thread.  We rely on a signal to kick us out of vcpu execution.  Timers
>> and AIO use SIGALRM and SIGUSR2 respectively.  Unfortunately eventfd
>> does not support O_ASYNC (SIGIO) so eventfd cannot be used in a signal
>> driven manner.
>>
>> Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
>
> I'd rather have the CONFIG_ ... mess contained in kvm-all.c
> which tests CONFIG_IOTHREAD anyway, than spread out
> to devices. Can we make kvm_has_many_ioeventfds check this?

Good idea.

Stefan

Patch

diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index d07ff97..e921eda 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -640,6 +640,25 @@  static int virtio_pci_set_host_notifier(void *opaque, int n, bool assign)
     return virtio_pci_set_host_notifier_internal(proxy, n, assign);
 }
 
+static bool virtio_pci_can_use_ioeventfd(void)
+{
+    if (!kvm_has_many_ioeventfds()) {
+        return false;
+    }
+
+    /* Use ioeventfd for virtqueue kick only if we have an I/O thread to
+     * perform out-of-line processing.  Otherwise we might as well do
+     * synchronous virtqueue kicks and in fact we have to since eventfd does
+     * not support SIGIO.  Without the I/O thread a signal would be required to
+     * kick the vcpu out of guest code.
+     */
+#ifdef CONFIG_IOTHREAD
+    return true;
+#else
+    return false;
+#endif
+}
+
 static void virtio_pci_vmstate_change(void *opaque, bool running)
 {
     VirtIOPCIProxy *proxy = opaque;
@@ -705,7 +724,7 @@  static void virtio_init_pci(VirtIOPCIProxy *proxy, VirtIODevice *vdev,
     pci_register_bar(&proxy->pci_dev, 0, size, PCI_BASE_ADDRESS_SPACE_IO,
                            virtio_map);
 
-    if (!kvm_has_many_ioeventfds()) {
+    if (!virtio_pci_can_use_ioeventfd()) {
         proxy->flags &= ~VIRTIO_PCI_FLAG_USE_IOEVENTFD;
     }