Re: [PATCH v5 5/5] vhost-user-blk: default num_queues to -smp N

2020-07-10 Thread Raphael Norwitz
On Fri, Jul 10, 2020 at 5:53 AM Stefan Hajnoczi  wrote:
>
> On Thu, Jul 09, 2020 at 11:02:24AM -0700, Raphael Norwitz wrote:
> > On Mon, Jul 6, 2020 at 7:00 AM Stefan Hajnoczi  wrote:
> > > diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c
> > > index a00b854736..39aec42dae 100644
> > > --- a/hw/block/vhost-user-blk.c
> > > +++ b/hw/block/vhost-user-blk.c
> > > @@ -420,6 +420,9 @@ static void vhost_user_blk_device_realize(DeviceState 
> > > *dev, Error **errp)
> > >  return;
> > >  }
> > >
> > > +if (s->num_queues == VHOST_USER_BLK_AUTO_NUM_QUEUES) {
> > > +s->num_queues = 1;
> > > +}
> >
> > What is this check for? Is it just a backstop to ensure that
> > num_queues is set to 1 if vhost-user-blk-pci doesn't update it?
>
> For the non-PCI VIRTIO transports that do not handle num_queues ==
> VHOST_USER_BLK_AUTO_NUM_QUEUES themselves.
>

Got it. Looks good then.

Reviewed-by: Raphael Norwitz 

> Stefan



Re: [PATCH v5 5/5] vhost-user-blk: default num_queues to -smp N

2020-07-10 Thread Stefan Hajnoczi
On Thu, Jul 09, 2020 at 11:02:24AM -0700, Raphael Norwitz wrote:
> On Mon, Jul 6, 2020 at 7:00 AM Stefan Hajnoczi  wrote:
> > diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c
> > index a00b854736..39aec42dae 100644
> > --- a/hw/block/vhost-user-blk.c
> > +++ b/hw/block/vhost-user-blk.c
> > @@ -420,6 +420,9 @@ static void vhost_user_blk_device_realize(DeviceState 
> > *dev, Error **errp)
> >  return;
> >  }
> >
> > +if (s->num_queues == VHOST_USER_BLK_AUTO_NUM_QUEUES) {
> > +s->num_queues = 1;
> > +}
> 
> What is this check for? Is it just a backstop to ensure that
> num_queues is set to 1 if vhost-user-blk-pci doesn't update it?

For the non-PCI VIRTIO transports that do not handle num_queues ==
VHOST_USER_BLK_AUTO_NUM_QUEUES themselves.

Stefan


signature.asc
Description: PGP signature


Re: [PATCH v5 5/5] vhost-user-blk: default num_queues to -smp N

2020-07-09 Thread Raphael Norwitz
On Mon, Jul 6, 2020 at 7:00 AM Stefan Hajnoczi  wrote:
>
> Automatically size the number of request virtqueues to match the number
> of vCPUs.  This ensures that completion interrupts are handled on the
> same vCPU that submitted the request.  No IPI is necessary to complete
> an I/O request and performance is improved.  The maximum number of MSI-X
> vectors and virtqueues limit are respected.
>
> Signed-off-by: Stefan Hajnoczi 
> Reviewed-by: Cornelia Huck 
> ---
>  include/hw/virtio/vhost-user-blk.h | 2 ++
>  hw/block/vhost-user-blk.c  | 6 +-
>  hw/core/machine.c  | 1 +
>  hw/virtio/vhost-user-blk-pci.c | 4 
>  4 files changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/include/hw/virtio/vhost-user-blk.h 
> b/include/hw/virtio/vhost-user-blk.h
> index 34ad6f0c0e..292d17147c 100644
> --- a/include/hw/virtio/vhost-user-blk.h
> +++ b/include/hw/virtio/vhost-user-blk.h
> @@ -25,6 +25,8 @@
>  #define VHOST_USER_BLK(obj) \
>  OBJECT_CHECK(VHostUserBlk, (obj), TYPE_VHOST_USER_BLK)
>
> +#define VHOST_USER_BLK_AUTO_NUM_QUEUES UINT16_MAX
> +
>  typedef struct VHostUserBlk {
>  VirtIODevice parent_obj;
>  CharBackend chardev;
> diff --git a/hw/block/vhost-user-blk.c b/hw/block/vhost-user-blk.c
> index a00b854736..39aec42dae 100644
> --- a/hw/block/vhost-user-blk.c
> +++ b/hw/block/vhost-user-blk.c
> @@ -420,6 +420,9 @@ static void vhost_user_blk_device_realize(DeviceState 
> *dev, Error **errp)
>  return;
>  }
>
> +if (s->num_queues == VHOST_USER_BLK_AUTO_NUM_QUEUES) {
> +s->num_queues = 1;
> +}

What is this check for? Is it just a backstop to ensure that
num_queues is set to 1 if vhost-user-blk-pci doesn't update it?

>  if (!s->num_queues || s->num_queues > VIRTIO_QUEUE_MAX) {
>  error_setg(errp, "vhost-user-blk: invalid number of IO queues");
>  return;
> @@ -531,7 +534,8 @@ static const VMStateDescription vmstate_vhost_user_blk = {
>
>  static Property vhost_user_blk_properties[] = {
>  DEFINE_PROP_CHR("chardev", VHostUserBlk, chardev),
> -DEFINE_PROP_UINT16("num-queues", VHostUserBlk, num_queues, 1),
> +DEFINE_PROP_UINT16("num-queues", VHostUserBlk, num_queues,
> +   VHOST_USER_BLK_AUTO_NUM_QUEUES),
>  DEFINE_PROP_UINT32("queue-size", VHostUserBlk, queue_size, 128),
>  DEFINE_PROP_BIT("config-wce", VHostUserBlk, config_wce, 0, true),
>  DEFINE_PROP_END_OF_LIST(),
> diff --git a/hw/core/machine.c b/hw/core/machine.c
> index 845f6476cb..31bfaacdb5 100644
> --- a/hw/core/machine.c
> +++ b/hw/core/machine.c
> @@ -30,6 +30,7 @@
>
>  GlobalProperty hw_compat_5_0[] = {
>  { "vhost-scsi", "num_queues", "1"},
> +{ "vhost-user-blk", "num-queues", "1"},
>  { "vhost-user-scsi", "num_queues", "1"},
>  { "virtio-balloon-device", "page-poison", "false" },
>  { "virtio-blk-device", "num-queues", "1"},
> diff --git a/hw/virtio/vhost-user-blk-pci.c b/hw/virtio/vhost-user-blk-pci.c
> index 4f5d5cbf44..a62a71e067 100644
> --- a/hw/virtio/vhost-user-blk-pci.c
> +++ b/hw/virtio/vhost-user-blk-pci.c
> @@ -54,6 +54,10 @@ static void vhost_user_blk_pci_realize(VirtIOPCIProxy 
> *vpci_dev, Error **errp)
>  VHostUserBlkPCI *dev = VHOST_USER_BLK_PCI(vpci_dev);
>  DeviceState *vdev = DEVICE(>vdev);
>
> +if (dev->vdev.num_queues == VHOST_USER_BLK_AUTO_NUM_QUEUES) {
> +dev->vdev.num_queues = virtio_pci_optimal_num_queues(0);
> +}
> +
>  if (vpci_dev->nvectors == DEV_NVECTORS_UNSPECIFIED) {
>  vpci_dev->nvectors = dev->vdev.num_queues + 1;
>  }
> --
> 2.26.2
>