Date:   Fri, 9 Sep 2022 08:42:25 +0000
From:   "Tian, Kevin" <kevin.tian@...el.com>
To:     Ethan Zhao <haifeng.zhao@...ux.intel.com>,
        Zhenyu Wang <zhenyuw@...ux.intel.com>,
        "Wang, Zhi A" <zhi.a.wang@...el.com>,
        Jani Nikula <jani.nikula@...ux.intel.com>,
        Joonas Lahtinen <joonas.lahtinen@...ux.intel.com>,
        "Vivi, Rodrigo" <rodrigo.vivi@...el.com>,
        Tvrtko Ursulin <tvrtko.ursulin@...ux.intel.com>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Eric Farman <farman@...ux.ibm.com>,
        Matthew Rosato <mjrosato@...ux.ibm.com>,
        Halil Pasic <pasic@...ux.ibm.com>,
        Vineeth Vijayan <vneethv@...ux.ibm.com>,
        "Peter Oberparleiter" <oberpar@...ux.ibm.com>,
        Heiko Carstens <hca@...ux.ibm.com>,
        Vasily Gorbik <gor@...ux.ibm.com>,
        Alexander Gordeev <agordeev@...ux.ibm.com>,
        Christian Borntraeger <borntraeger@...ux.ibm.com>,
        Sven Schnelle <svens@...ux.ibm.com>,
        Tony Krowiak <akrowiak@...ux.ibm.com>,
        Jason Herne <jjherne@...ux.ibm.com>,
        Harald Freudenberger <freude@...ux.ibm.com>,
        Diana Craciun <diana.craciun@....nxp.com>,
        "Alex Williamson" <alex.williamson@...hat.com>,
        Cornelia Huck <cohuck@...hat.com>,
        Longfang Liu <liulongfang@...wei.com>,
        Shameer Kolothum <shameerali.kolothum.thodi@...wei.com>,
        Jason Gunthorpe <jgg@...pe.ca>,
        Yishai Hadas <yishaih@...dia.com>,
        Eric Auger <eric.auger@...hat.com>,
        "Kirti Wankhede" <kwankhede@...dia.com>,
        Leon Romanovsky <leon@...nel.org>,
        Abhishek Sahu <abhsahu@...dia.com>,
        Christoph Hellwig <hch@...radead.org>,
        "intel-gvt-dev@...ts.freedesktop.org" 
        <intel-gvt-dev@...ts.freedesktop.org>,
        "intel-gfx@...ts.freedesktop.org" <intel-gfx@...ts.freedesktop.org>,
        "dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-s390@...r.kernel.org" <linux-s390@...r.kernel.org>,
        "kvm@...r.kernel.org" <kvm@...r.kernel.org>
CC:     "Liu, Yi L" <yi.l.liu@...el.com>
Subject: RE: [PATCH v3 01/15] vfio: Add helpers for unifying vfio_device life
 cycle

> From: Ethan Zhao <haifeng.zhao@...ux.intel.com>
> Sent: Friday, September 9, 2022 4:24 PM
> 
> Hi, Kevin,
> 
> 在 2022/9/9 18:22, Kevin Tian 写道:
> > The idea is to let vfio core manage the vfio_device life cycle instead
> > of duplicating the logic cross drivers. This is also a preparatory
> > step for adding struct device into vfio_device.
> >
> > New pair of helpers together with a kref in vfio_device:
> >
> >   - vfio_alloc_device()
> >   - vfio_put_device()
> 
> To be honest, this pair of functions make me confusing to understand their
> 
> behaviour from wording point of view:
> 
> - vfio_alloc_device(),  Okay, it allocates the vfio device, no reference
>   count thing. but,

I think it's quite common to have an alloc() helper initialize refcount, e.g.
vfio_group_alloc() both initialize its user refcount and also call
device_initialize()  to gets kref initialized. Similar example in
ib_alloc_device(), etc.

> - vfio_put_device()
>   seems it will decrease reference count and then if it is zero, free it.
>   so they are not of one *pair* about wording.
> 
> How about
> 
> - vfio_alloc_device() / - vfio_free_device()
> or
> - vfio_get_device() / - vfio_put_device(), perhaps not match their behviour
> in following code.
> 
> 
> 
> Thanks,
> Ethan
> 
> 
> >
> > Drivers can register @init/@...ease callbacks to manage any private
> > state wrapping the vfio_device.
> >
> > However vfio-ccw doesn't fit this model due to a life cycle mess
> > that its private structure mixes both parent and mdev info hence must
> > be allocated/freed outside of the life cycle of vfio device.
> >
> > Per prior discussions this won't be fixed in short term by IBM folks.
> >
> > Instead of waiting for those modifications introduce another helper
> > vfio_init_device() so ccw can call it to initialize a pre-allocated
> > vfio_device.
> >
> > Further implication of the ccw trick is that vfio_device cannot be
> > freed uniformly in vfio core. Instead, require *EVERY* driver to
> > implement @release and free vfio_device inside. Then ccw can choose
> > to delay the free at its own discretion.
> >
> > Another trick down the road is that kvzalloc() is used to accommodate
> > the need of gvt which uses vzalloc() while all others use kzalloc().
> > So drivers should call a helper vfio_free_device() to free the
> > vfio_device instead of assuming that kfree() or vfree() is appliable.
> >
> > Later once the ccw mess is fixed we can remove those tricks and
> > fully handle structure alloc/free in vfio core.
> >
> > Existing vfio_{un}init_group_dev() will be deprecated after all
> > existing usages are converted to the new model.
> >
> > Suggested-by: Jason Gunthorpe <jgg@...dia.com>
> > Co-developed-by: Yi Liu <yi.l.liu@...el.com>
> > Signed-off-by: Yi Liu <yi.l.liu@...el.com>
> > Signed-off-by: Kevin Tian <kevin.tian@...el.com>
> > Reviewed-by: Tony Krowiak <akrowiak@...ux.ibm.com>
> > Reviewed-by: Jason Gunthorpe <jgg@...dia.com>
> > Reviewed-by: Eric Auger <eric.auger@...hat.com>
> > ---
> >   drivers/vfio/vfio_main.c | 92
> ++++++++++++++++++++++++++++++++++++++++
> >   include/linux/vfio.h     | 25 ++++++++++-
> >   2 files changed, 116 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c
> > index 27d9186f35d5..adc1b697bb78 100644
> > --- a/drivers/vfio/vfio_main.c
> > +++ b/drivers/vfio/vfio_main.c
> > @@ -498,6 +498,98 @@ void vfio_uninit_group_dev(struct vfio_device
> *device)
> >   }
> >   EXPORT_SYMBOL_GPL(vfio_uninit_group_dev);
> >
> > +/* Release helper called by vfio_put_device() */
> > +void vfio_device_release(struct kref *kref)
> > +{
> > +	struct vfio_device *device =
> > +			container_of(kref, struct vfio_device, kref);
> > +
> > +	vfio_uninit_group_dev(device);
> > +
> > +	/*
> > +	 * kvfree() cannot be done here due to a life cycle mess in
> > +	 * vfio-ccw. Before the ccw part is fixed all drivers are
> > +	 * required to support @release and call vfio_free_device()
> > +	 * from there.
> > +	 */
> > +	device->ops->release(device);
> > +}
> > +EXPORT_SYMBOL_GPL(vfio_device_release);
> > +
> > +/*
> > + * Alloc and initialize vfio_device so it can be registered to vfio
> > + * core.
> > + *
> > + * Drivers should use the wrapper vfio_alloc_device() for allocation.
> > + * @size is the size of the structure to be allocated, including any
> > + * private data used by the driver.
> > + *
> > + * Driver may provide an @init callback to cover device private data.
> > + *
> > + * Use vfio_put_device() to release the structure after success return.
> > + */
> > +struct vfio_device *_vfio_alloc_device(size_t size, struct device *dev,
> > +				       const struct vfio_device_ops *ops)
> > +{
> > +	struct vfio_device *device;
> > +	int ret;
> > +
> > +	if (WARN_ON(size < sizeof(struct vfio_device)))
> > +		return ERR_PTR(-EINVAL);
> > +
> > +	device = kvzalloc(size, GFP_KERNEL);
> > +	if (!device)
> > +		return ERR_PTR(-ENOMEM);
> > +
> > +	ret = vfio_init_device(device, dev, ops);
> > +	if (ret)
> > +		goto out_free;
> > +	return device;
> > +
> > +out_free:
> > +	kvfree(device);
> > +	return ERR_PTR(ret);
> > +}
> > +EXPORT_SYMBOL_GPL(_vfio_alloc_device);
> > +
> > +/*
> > + * Initialize a vfio_device so it can be registered to vfio core.
> > + *
> > + * Only vfio-ccw driver should call this interface.
> > + */
> > +int vfio_init_device(struct vfio_device *device, struct device *dev,
> > +		     const struct vfio_device_ops *ops)
> > +{
> > +	int ret;
> > +
> > +	vfio_init_group_dev(device, dev, ops);
> > +
> > +	if (ops->init) {
> > +		ret = ops->init(device);
> > +		if (ret)
> > +			goto out_uninit;
> > +	}
> > +
> > +	kref_init(&device->kref);
> > +	return 0;
> > +
> > +out_uninit:
> > +	vfio_uninit_group_dev(device);
> > +	return ret;
> > +}
> > +EXPORT_SYMBOL_GPL(vfio_init_device);
> > +
> > +/*
> > + * The helper called by driver @release callback to free the device
> > + * structure. Drivers which don't have private data to clean can
> > + * simply use this helper as its @release.
> > + */
> > +void vfio_free_device(struct vfio_device *device)
> > +{
> > +	kvfree(device);
> > +}
> > +EXPORT_SYMBOL_GPL(vfio_free_device);
> > +
> >   static struct vfio_group *vfio_noiommu_group_alloc(struct device *dev,
> >   		enum vfio_group_type type)
> >   {
> > diff --git a/include/linux/vfio.h b/include/linux/vfio.h
> > index 0e2826559091..f67cac700e6f 100644
> > --- a/include/linux/vfio.h
> > +++ b/include/linux/vfio.h
> > @@ -47,7 +47,8 @@ struct vfio_device {
> >   	struct kvm *kvm;
> >
> >   	/* Members below here are private, not for driver use */
> > -	refcount_t refcount;
> > +	struct kref kref;	/* object life cycle */
> > +	refcount_t refcount;	/* user count on registered device*/
> >   	unsigned int open_count;
> >   	struct completion comp;
> >   	struct list_head group_next;
> > @@ -57,6 +58,8 @@ struct vfio_device {
> >   /**
> >    * struct vfio_device_ops - VFIO bus driver device callbacks
> >    *
> > + * @init: initialize private fields in device structure
> > + * @release: Reclaim private fields in device structure
> >    * @open_device: Called when the first file descriptor is opened for this
> device
> >    * @close_device: Opposite of open_device
> >    * @read: Perform read(2) on device file descriptor
> > @@ -74,6 +77,8 @@ struct vfio_device {
> >    */
> >   struct vfio_device_ops {
> >   	char	*name;
> > +	int	(*init)(struct vfio_device *vdev);
> > +	void	(*release)(struct vfio_device *vdev);
> >   	int	(*open_device)(struct vfio_device *vdev);
> >   	void	(*close_device)(struct vfio_device *vdev);
> >   	ssize_t	(*read)(struct vfio_device *vdev, char __user *buf,
> > @@ -161,6 +166,24 @@ static inline int vfio_check_feature(u32 flags,
> size_t argsz, u32 supported_ops,
> >   	return 1;
> >   }
> >
> > +struct vfio_device *_vfio_alloc_device(size_t size, struct device *dev,
> > +				       const struct vfio_device_ops *ops);
> > +#define vfio_alloc_device(dev_struct, member, dev, ops)
> 		\
> > +	container_of(_vfio_alloc_device(sizeof(struct dev_struct) +
> 	\
> > +					BUILD_BUG_ON_ZERO(offsetof(
> 		\
> > +						struct dev_struct, member)),
> 	\
> > +					dev, ops),
> 	\
> > +		     struct dev_struct, member)
> > +
> > +int vfio_init_device(struct vfio_device *device, struct device *dev,
> > +		     const struct vfio_device_ops *ops);
> > +void vfio_free_device(struct vfio_device *device);
> > +void vfio_device_release(struct kref *kref);
> > +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);
> 
> --
> "firm, enduring, strong, and long-lived"

Powered by blists - more mailing lists