| Message ID | 1605223150-10888-2-git-send-email-michael.christie@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/1] qemu vhost scsi: add VHOST_SET_VRING_ENABLE support | expand |
On Thu, Nov 12, 2020 at 05:19:00PM -0600, Mike Christie wrote: > +static int vhost_kernel_set_vring_enable(struct vhost_dev *dev, int enable) > +{ > + struct vhost_vring_state s; > + int i, ret; > + > + s.num = 1; > + for (i = 0; i < dev->nvqs; ++i) { > + s.index = i; > + > + ret = vhost_kernel_call(dev, VHOST_SET_VRING_ENABLE, &s); > + /* Ignore kernels that do not support the cmd */ > + if (ret == -EPERM) > + return 0; > + if (ret) > + goto disable_vrings; > + } The 'enable' argument is ignored and this function acts on all virtqueues, while the ioctl acts on a single virtqueue only. This function's behavior is actually "vhost_kernel_enable_vrings()" (plural), not "vhost_kernel_set_vring_enable()" (singular). Please rename this function and drop the enable argument.
On Thu, Nov 12, 2020 at 05:19:00PM -0600, Mike Christie wrote: > diff --git a/linux-headers/linux/vhost.h b/linux-headers/linux/vhost.h > index 7523218..98dd919 100644 > --- a/linux-headers/linux/vhost.h > +++ b/linux-headers/linux/vhost.h > @@ -70,6 +70,7 @@ > #define VHOST_VRING_BIG_ENDIAN 1 > #define VHOST_SET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x13, struct vhost_vring_state) > #define VHOST_GET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x14, struct vhost_vring_state) > +#define VHOST_SET_VRING_ENABLE _IOW(VHOST_VIRTIO, 0x15, struct vhost_vring_state) OK so first we need the kernel patches, then update the header, then we can apply the qemu patch. > /* The following ioctls use eventfd file descriptors to signal and poll > * for events. */ > -- > 1.8.3.1
> On Dec 2, 2020, at 3:59 AM, Michael S. Tsirkin <mst@redhat.com> wrote: > > On Thu, Nov 12, 2020 at 05:19:00PM -0600, Mike Christie wrote: >> diff --git a/linux-headers/linux/vhost.h b/linux-headers/linux/vhost.h >> index 7523218..98dd919 100644 >> --- a/linux-headers/linux/vhost.h >> +++ b/linux-headers/linux/vhost.h >> @@ -70,6 +70,7 @@ >> #define VHOST_VRING_BIG_ENDIAN 1 >> #define VHOST_SET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x13, struct vhost_vring_state) >> #define VHOST_GET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x14, struct vhost_vring_state) >> +#define VHOST_SET_VRING_ENABLE _IOW(VHOST_VIRTIO, 0x15, struct vhost_vring_state) > > OK so first we need the kernel patches, then update the header, then > we can apply the qemu patch. Hey Michael, Don’t waste any more of your time on this patch and the kernel related ones. I’m working on the userspace initiated threading approach discussed the other week.
diff --git a/hw/scsi/vhost-scsi.c b/hw/scsi/vhost-scsi.c index 4d70fa0..bbb2ba3 100644 --- a/hw/scsi/vhost-scsi.c +++ b/hw/scsi/vhost-scsi.c @@ -93,6 +93,12 @@ static int vhost_scsi_start(VHostSCSI *s) return ret; } + ret = vsc->dev.vhost_ops->vhost_set_vring_enable(&vsc->dev, 1); + if (ret) { + error_report("Error enabling vhost-scsi vqs %d", ret); + vhost_scsi_common_stop(vsc); + } + ret = vhost_scsi_set_endpoint(s); if (ret < 0) { error_report("Error setting vhost-scsi endpoint"); diff --git a/hw/virtio/vhost-backend.c b/hw/virtio/vhost-backend.c index 88c8ecc..e190c8e 100644 --- a/hw/virtio/vhost-backend.c +++ b/hw/virtio/vhost-backend.c @@ -102,6 +102,35 @@ static int vhost_kernel_set_mem_table(struct vhost_dev *dev, return vhost_kernel_call(dev, VHOST_SET_MEM_TABLE, mem); } +static int vhost_kernel_set_vring_enable(struct vhost_dev *dev, int enable) +{ + struct vhost_vring_state s; + int i, ret; + + s.num = 1; + for (i = 0; i < dev->nvqs; ++i) { + s.index = i; + + ret = vhost_kernel_call(dev, VHOST_SET_VRING_ENABLE, &s); + /* Ignore kernels that do not support the cmd */ + if (ret == -EPERM) + return 0; + if (ret) + goto disable_vrings; + } + + return 0; + +disable_vrings: + s.num = 0; + for (i--; i < dev->nvqs; ++i) { + s.index = i; + vhost_kernel_call(dev, VHOST_SET_VRING_ENABLE, &s); + } + + return ret; +} + static int vhost_kernel_set_vring_addr(struct vhost_dev *dev, struct vhost_vring_addr *addr) { @@ -302,6 +331,7 @@ static const VhostOps kernel_ops = { .vhost_scsi_get_abi_version = vhost_kernel_scsi_get_abi_version, .vhost_set_log_base = vhost_kernel_set_log_base, .vhost_set_mem_table = vhost_kernel_set_mem_table, + .vhost_set_vring_enable = vhost_kernel_set_vring_enable, .vhost_set_vring_addr = vhost_kernel_set_vring_addr, .vhost_set_vring_endian = vhost_kernel_set_vring_endian, .vhost_set_vring_num = vhost_kernel_set_vring_num, diff --git a/linux-headers/linux/vhost.h b/linux-headers/linux/vhost.h index 7523218..98dd919 100644 --- a/linux-headers/linux/vhost.h +++ b/linux-headers/linux/vhost.h @@ -70,6 +70,7 @@ #define VHOST_VRING_BIG_ENDIAN 1 #define VHOST_SET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x13, struct vhost_vring_state) #define VHOST_GET_VRING_ENDIAN _IOW(VHOST_VIRTIO, 0x14, struct vhost_vring_state) +#define VHOST_SET_VRING_ENABLE _IOW(VHOST_VIRTIO, 0x15, struct vhost_vring_state) /* The following ioctls use eventfd file descriptors to signal and poll * for events. */
This patch made over the master branch allows the vhost-scsi driver to call into the kernel and tell it to enable/disable a virtqueue. The kernel patches included with this set, will create a worker per IO vq when multiple IO queues have been setup. Signed-off-by: Mike Christie <michael.christie@oracle.com> --- hw/scsi/vhost-scsi.c | 6 ++++++ hw/virtio/vhost-backend.c | 30 ++++++++++++++++++++++++++++++ linux-headers/linux/vhost.h | 1 + 3 files changed, 37 insertions(+)