[Intel-gfx] [PATCH 13/15] vfio/ccw: Use the new device life cycle helpers
Tian, Kevin
kevin.tian at intel.com
Sat Aug 27 10:39:04 UTC 2022
This missed a Suggested-by from Jason. Will add in next version.
> From: Tian, Kevin <kevin.tian at intel.com>
> Sent: Sunday, August 28, 2022 1:11 AM
>
> ccw is the only exception which cannot use vfio_alloc_device() because
> its private device structure is designed to serve both mdev and parent.
> Life cycle of the parent is managed by css_driver so vfio_ccw_private
> must be allocated/freed in css_driver probe/remove path instead of
> conforming to vfio core life cycle for mdev.
>
> Given that use a wait/completion scheme so the mdev remove path waits
> after vfio_put_device() until receiving a completion notification from
> @release. The completion indicates that all active references on
> vfio_device have been released.
>
> After that point although free of vfio_ccw_private is delayed to
> css_driver it's at least guaranteed to have no parallel reference on
> released vfio device part from other code paths.
>
> memset() in @probe is removed. vfio_device is either alreary cleared
> when probed for the first time or cleared in @release from last probe.
>
> The right fix is to introduce separate structures for mdev and parent,
> but this won't happen in short term per prior discussions.
>
> Remove vfio_init/uninit_group_dev() as no user now.
>
> Signed-off-by: Kevin Tian <kevin.tian at intel.com>
> ---
> drivers/s390/cio/vfio_ccw_ops.c | 52 +++++++++++++++++++++++++----
> drivers/s390/cio/vfio_ccw_private.h | 3 ++
> drivers/vfio/vfio_main.c | 27 +++------------
> include/linux/vfio.h | 3 --
> 4 files changed, 53 insertions(+), 32 deletions(-)
>
> diff --git a/drivers/s390/cio/vfio_ccw_ops.c
> b/drivers/s390/cio/vfio_ccw_ops.c
> index 4a806a2273b5..9f8486c0d3d3 100644
> --- a/drivers/s390/cio/vfio_ccw_ops.c
> +++ b/drivers/s390/cio/vfio_ccw_ops.c
> @@ -87,6 +87,15 @@ static struct attribute_group *mdev_type_groups[] = {
> NULL,
> };
>
> +static int vfio_ccw_mdev_init_dev(struct vfio_device *vdev)
> +{
> + struct vfio_ccw_private *private =
> + container_of(vdev, struct vfio_ccw_private, vdev);
> +
> + init_completion(&private->release_comp);
> + return 0;
> +}
> +
> static int vfio_ccw_mdev_probe(struct mdev_device *mdev)
> {
> struct vfio_ccw_private *private = dev_get_drvdata(mdev-
> >dev.parent);
> @@ -98,9 +107,9 @@ static int vfio_ccw_mdev_probe(struct mdev_device
> *mdev)
> if (atomic_dec_if_positive(&private->avail) < 0)
> return -EPERM;
>
> - memset(&private->vdev, 0, sizeof(private->vdev));
> - vfio_init_group_dev(&private->vdev, &mdev->dev,
> - &vfio_ccw_dev_ops);
> + ret = vfio_init_device(&private->vdev, &mdev->dev,
> &vfio_ccw_dev_ops);
> + if (ret)
> + return ret;
>
> VFIO_CCW_MSG_EVENT(2, "sch %x.%x.%04x: create\n",
> private->sch->schid.cssid,
> @@ -109,16 +118,33 @@ static int vfio_ccw_mdev_probe(struct
> mdev_device *mdev)
>
> ret = vfio_register_emulated_iommu_dev(&private->vdev);
> if (ret)
> - goto err_atomic;
> + goto err_put_vdev;
> dev_set_drvdata(&mdev->dev, private);
> return 0;
>
> -err_atomic:
> - vfio_uninit_group_dev(&private->vdev);
> +err_put_vdev:
> + vfio_put_device(&private->vdev);
> atomic_inc(&private->avail);
> return ret;
> }
>
> +static void vfio_ccw_mdev_release_dev(struct vfio_device *vdev)
> +{
> + struct vfio_ccw_private *private =
> + container_of(vdev, struct vfio_ccw_private, vdev);
> +
> + /*
> + * We cannot free vfio_ccw_private here because it includes
> + * parent info which must be free'ed by css driver.
> + *
> + * Use a workaround by memset'ing the core device part and
> + * then notifying the remove path that all active references
> + * to this device have been released.
> + */
> + memset(vdev, 0, sizeof(*vdev));
> + complete(&private->release_comp);
> +}
> +
> static void vfio_ccw_mdev_remove(struct mdev_device *mdev)
> {
> struct vfio_ccw_private *private = dev_get_drvdata(mdev-
> >dev.parent);
> @@ -130,7 +156,17 @@ static void vfio_ccw_mdev_remove(struct
> mdev_device *mdev)
>
> vfio_unregister_group_dev(&private->vdev);
>
> - vfio_uninit_group_dev(&private->vdev);
> + vfio_put_device(&private->vdev);
> + /*
> + * Wait for all active references on mdev are released so it
> + * is safe to defer kfree() to a later point.
> + *
> + * TODO: the clean fix is to split parent/mdev info from ccw
> + * private structure so each can be managed in its own life
> + * cycle.
> + */
> + wait_for_completion(&private->release_comp);
> +
> atomic_inc(&private->avail);
> }
>
> @@ -592,6 +628,8 @@ static void vfio_ccw_mdev_request(struct vfio_device
> *vdev, unsigned int count)
> }
>
> static const struct vfio_device_ops vfio_ccw_dev_ops = {
> + .init = vfio_ccw_mdev_init_dev,
> + .release = vfio_ccw_mdev_release_dev,
> .open_device = vfio_ccw_mdev_open_device,
> .close_device = vfio_ccw_mdev_close_device,
> .read = vfio_ccw_mdev_read,
> diff --git a/drivers/s390/cio/vfio_ccw_private.h
> b/drivers/s390/cio/vfio_ccw_private.h
> index cd24b7fada91..63d9202b29c7 100644
> --- a/drivers/s390/cio/vfio_ccw_private.h
> +++ b/drivers/s390/cio/vfio_ccw_private.h
> @@ -88,6 +88,7 @@ struct vfio_ccw_crw {
> * @req_trigger: eventfd ctx for signaling userspace to return device
> * @io_work: work for deferral process of I/O handling
> * @crw_work: work for deferral process of CRW handling
> + * @release_comp: synchronization helper for vfio device release
> */
> struct vfio_ccw_private {
> struct vfio_device vdev;
> @@ -113,6 +114,8 @@ struct vfio_ccw_private {
> struct eventfd_ctx *req_trigger;
> struct work_struct io_work;
> struct work_struct crw_work;
> +
> + struct completion release_comp;
> } __aligned(8);
>
> int vfio_ccw_sch_quiesce(struct subchannel *sch);
> diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c
> index 9485da17f2e6..15a612153c13 100644
> --- a/drivers/vfio/vfio_main.c
> +++ b/drivers/vfio/vfio_main.c
> @@ -482,21 +482,6 @@ static struct vfio_device
> *vfio_group_get_device(struct vfio_group *group,
> /*
> * VFIO driver API
> */
> -void vfio_init_group_dev(struct vfio_device *device, struct device *dev,
> - const struct vfio_device_ops *ops)
> -{
> - init_completion(&device->comp);
> - device->dev = dev;
> - device->ops = ops;
> -}
> -EXPORT_SYMBOL_GPL(vfio_init_group_dev);
> -
> -void vfio_uninit_group_dev(struct vfio_device *device)
> -{
> - vfio_release_device_set(device);
> -}
> -EXPORT_SYMBOL_GPL(vfio_uninit_group_dev);
> -
> /*
> * Alloc and initialize vfio_device so it can be registered to vfio
> * core.
> @@ -543,20 +528,18 @@ int vfio_init_device(struct vfio_device *device,
> struct device *dev,
> {
> int ret;
>
> - vfio_init_group_dev(device, dev, ops);
> + init_completion(&device->comp);
> + device->dev = dev;
> + device->ops = ops;
>
> if (ops->init) {
> ret = ops->init(device);
> if (ret)
> - goto out_uninit;
> + return ret;
> }
>
> kref_init(&device->kref);
> return 0;
> -
> -out_uninit:
> - vfio_uninit_group_dev(device);
> - return ret;
> }
> EXPORT_SYMBOL_GPL(vfio_init_device);
>
> @@ -577,7 +560,7 @@ void vfio_device_release(struct kref *kref)
> struct vfio_device *device =
> container_of(kref, struct vfio_device, kref);
>
> - vfio_uninit_group_dev(device);
> + vfio_release_device_set(device);
>
> /*
> * kvfree() cannot be done here due to a life cycle mess in
> diff --git a/include/linux/vfio.h b/include/linux/vfio.h
> index b0928a81b45a..a36a65b966b5 100644
> --- a/include/linux/vfio.h
> +++ b/include/linux/vfio.h
> @@ -160,9 +160,6 @@ static inline void vfio_put_device(struct vfio_device
> *device)
> kref_put(&device->kref, vfio_device_release);
> }
>
> -void vfio_init_group_dev(struct vfio_device *device, struct device *dev,
> - const struct vfio_device_ops *ops);
> -void vfio_uninit_group_dev(struct vfio_device *device);
> int vfio_register_group_dev(struct vfio_device *device);
> int vfio_register_emulated_iommu_dev(struct vfio_device *device);
> void vfio_unregister_group_dev(struct vfio_device *device);
> --
> 2.21.3
More information about the Intel-gfx
mailing list